Meta-informative Centering in Utterances

Studies in Language Companion Series (SLCS)

This series has been established as a companion series to the periodical
Studies in Language.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/slcs

Editors
Werner Abraham Elly van Gelderen
University of Vienna / Arizona State University
University of Munich

Editorial Board
Bernard Comrie Christian Lehmann
Max Planck Institute, Leipzig University of Erfurt
and University of California, Santa Barbara
Marianne Mithun
William Croft University of California, Santa Barbara
University of New Mexico
Heiko Narrog
Östen Dahl Tohuku University
University of Stockholm
Johanna L. Wood
Gerrit J. Dimmendaal University of Aarhus
University of Cologne
Debra Ziegeler
Ekkehard König University of Paris III
Free University of Berlin

Volume 143
Meta-informative Centering in Utterances
Between Semantics and Pragmatics
Edited by André Włodarczyk and Hélène Włodarczyk
Meta-informative Centering
in Utterances
Between Semantics and Pragmatics

Edited by

André Włodarczyk
Université Charles de Gaulle & Université Paris-Sorbonne

Hélène Włodarczyk
Université Paris-Sorbonne

John Benjamins Publishing Company

Amsterdam / Philadelphia
 TM
The paper used in this publication meets the minimum requirements of
8

the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.

CIP data is available from the Library of Congress.

Studies in Language Companion Series, issn 0165-7763 ; v. 143

isbn 978 90 272 0610 7 (Hb ; alk. paper)
isbn 978 90 272 7114 3 (Eb)

© 2013 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
 Table of contents

Preface vii
Introduction ix
André Włodarczyk & Hélène Włodarczyk

part 1.  Associative semantics and meta-informative centering

Roles and anchors of semantic situations 3
André Włodarczyk
Frames of semantic situations 21
André Włodarczyk
Grounding of the meta-informative status of utterances 41
André Włodarczyk
Attention-centered information in language 59
Hélène Włodarczyk

part 2.  Neuropsychological evidence for the MIC theory

Semantic and episodic memory by reference to the ontological grounding
of the old and new meta-informative status 103
Franz J. Stachowiak
Tracing the role of memory and attention for the meta-informative
validation of utterances 121
Franz J. Stachowiak

part 3.  Meta-informative centering in languages

It-clefts in the Meta-Informative structure of the utterance in modern
and ­present-day English 145
Ana Elina Martínez-Insua & Javier Pérez-Guerra
Discourse coherence and referent identification of subject ellipsis
in Japanese 167
Shigeko Nariyama
 Meta-informative Centering in Utterances

Structure of centre of attention in a multi-party conversation in Japanese:

Based on the data of a review meeting concerning a Science Café held
in ­Hiroshima 183
Miki Saijo
Verbal aspect in Slavic languages between semantics and pragmatics 193
Hélène Włodarczyk
The position in the utterance and the melodic realisation of object
and reflexive pronouns in classical modern literary Russian 231
Olivier Azam
Accented and unaccented pronouns in Ancient Greek: A pragmatic choice
by the speaker 259
Jean-Christophe Pitavy
Personal subject pronouns and the meta-informative centering of utterances
in classical Latin 285
Perrine Vedrenne-Cloquet
Glossary of defined terminology 297
Index 303
 Preface

The papers in this volume are a selection of lectures delivered at three consecutive
MIC Sorbonne workshops in Paris: Discourse coherence – text and theory (2008),
Context-bound Communication (2010) and New standards for language studies
(2012). Another selection relating specifically to Slavic languages has been pub-
lished in French in the Revue des Études Slaves (2009, vol. 80/1–2). The editors and
co-authors benefitted a great deal from discussions with the programme commit-
tee members and participants before, during and after the workshops.
Sachiko Ide (former President of the International Pragmatics Association,
Women’s University, Tokyo), Istvan Kecskes (President of the American Pragmatics
Association, New York State University, Albany), Masao Aizawa (Vice-President
of The Japanese Language National Institute, Tachikawa) and Yasunari Harada
(Director at the Institute for DECODE = Digital Enhancement of Cognitive
Development, Waseda University, Tokyo) have played an important role in the
­dissemination and development of various aspects of the MIC theory.
Grants for a research exchange program with the Jagiellonian University
(Cracow) as well as invited lectures delivered by the editors at Waseda University
(Tokyo), The Institute of Slavic Studies of the Polish Academy of Sciences, Warsaw
University (Warsaw), The Japanese Language National Institute (Tachikawa), and
New York State University at Albany allowed us to discuss many MIC theoretical
issues.
We have been very privileged to have received help and advice from dis-
tinguished scholars. We wish to thank especially professors Werner Abraham
(Vienna University), Elisabeth Leiss (Ludwig-Maximilians-Universität München),
Radosław Katarzyniak (Wrocław University of Technology, Poland) and Robin K.
Burk (USMA, West Point, Dept. of Electrical Engineering & Computer S­ cience).
Last but not least, the editors also held discussions on various occasions with
their colleagues from CELTA Sorbonne research group: Jan Pekelder, Paul-Louis
Thomas, Stéphane Viellard and the late Jean Breuillard and Michel Viel, all of
them being professors at Paris-Sorbonne University, as well as Claude Delmas
­(professor at Sorbonne Nouvelle University) and Claude Guimier (professor at
Caen University). They have all contributed a great deal in one form or another to
the realisation of our research plans.
A special and personal thank you goes also to our friend, Dr Rosemary
­Masters, for her translations of a few chapters, careful rereading of, and insightful
comments on all the chapters of this book.
 Meta-informative Centering in Utterances

Acknowledements

The publisher and editors wish to thank the Slavic Studies Institute (PAN – P
­ olish
Academy of Science) for permission to print the paper by André Włodarczyk
(“Roles and Anchors of Semantic Situations”, Études Cognitives – Studia kogni-
tywne 8, SOW, Warszawa 2008, p. 53–70) as well as the English translation of
the paper by the same author (“Les Cadres des situations sémantiques”, Études
­Cognitives – Studia Kognitywne 5, Warszawa 2003, p. 35–51).
 Introduction

André Włodarczyk & Hélène Włodarczyk

Université Charles de Gaulle / Université Paris-Sorbonne,
Centre de Linguistique Théorique et Appliquée (CELTA Sorbonne)

1.  Information and meta-information

At the beginning of the 21st century, the study of communication in human lan-
guages still remains under the influence of the theory of information structure
(Lambrecht 1994) which elaborates on the concepts of theme, rheme, and commu-
nicative perspective. The meta-informative centering (MIC) theory is an alternative
framework based on the concepts of centre of attention (CA), meta-information
and the meta-informative status of information. Importantly, the MIC theory is
built on associative semantics (AS) in which the concept of information is defined
as a compound relational structure, which to some extent makes it compatible
with the definition of information as often used in the field of computer (infor-
mation) science. In order to explain how it is possible for the hearer to interpret
a linguistic message (i.e. to build a mental representation of the situation spoken
about) a mapping must be established between the linguistic form and its content,
using some formalised representation of meaning. From the semantic point of
view, the logical motivation for the formal representation of situations consists in
reducing multi-argument relations solely to unary and binary ones, and combin-
ing them in compound configurations.

2.  Attention centering

The origins of the MIC theory go back to Włodarczyk 1999 when we used the
concept of “centre d’intérêt” (Fr. centre of interest) for the first time. This French
term has since been replaced by that of “centre of attention” under the influence of
the American centering theory (Grosz & Sidner 1986), and it underwent a few revi-
sions and reassessments in the succeeding papers published in English. According
to the MIC theory, because of the linear order of speech, no judgment may be
uttered without selecting at least one centre of attention; thus centering is con-
sidered as a structuring operation not only at the discourse/text level but also at
 André Włodarczyk & Hélène Włodarczyk

the level of utterance (cf. Chomsky’s binding theory) because of its necessarily
linear (sequential) nature. In the American centering theory, centres of attention
are defined at the discourse/text level: one constituent of an utterance is treated
as a forward or backward looking centre in order to maintain the cohesive flow of
information from one utterance to its successor. Forward and backward looking
centres make it possible to give an account of the relations which bind utterances
together into coherent discourses/texts. Although the centering theory precedes in
time the meta-informative centering theory, these theories were created separately
from each other. However, the results of the American theory can be straightfor-
wardly integrated into the MIC theory. Since the anaphoric and cataphoric moti-
vations of the meta-informative old and new status of the utterance during the
communication act match perfectly well the concepts of backward- and forward-
looking centres, the MIC theory is not an alternative to the American centering
theory. Rather, the latter is a complementary approach with respect to the MIC
which is more general.
Although the main characteristics of utterances consist in the fact that the
priority in their generation/understanding is given to attention-driven processes,
utterances are built choosing the most suitable ready-made verb valence schemata
which are used by the speaker in order to communicate the old or new status of
expressed content. Therefore, linguistic expressions which correspond to gram-
matical notions such as subject, object, topic and focus are defined as attention-
driven phrases (ADP); i.e. phrases governed by the centres of attention (1) in base
utterances where merely either an old or new status of information can be envis-
aged for both the subject and predicate in each utterance (but where no contrast
between the old and new status of information is expressed), and (2) in extended
utterances where the topic and focus are defined as dually opposed concepts;
(a) the topic being defined as governed by an old meta-informative status contrast-
ing with the new one of the comment and (b) the focus as governed by a new meta-­
informative status contrasting with the old one of the background.
Although the importance of attention for cognitive functions of the brain has
recently also been emphasised by neuropsychologists (part 2 in this volume by
Franz J. Stachowiak), this faculty of the human brain has not yet been explicitly
used to account for linguistic objects in mainstream cognitive linguistics. In the
proposed theory, one kind of attention is a component of pragmatics and since the
semantic content of linguistic messages is not directly accessible, the interpreta-
tion of the truth value of utterances needs to be postponed and priority is given to
the resolution of the meta-informative status of communicated chunks of infor-
mation. In other words, in linguistic communication, truth valuation of the con-
tent of an utterance takes place only behind – so to speak – communication, i.e.
after the meta-informative old or new status has been assigned using such means
 Introduction 

of expression as intonation, word order, declension, determination, modality and

aspect, etc. It seems important for us to mention the endeavours of the psycholo-
gist Russell Tomlin (1997) who tried to check whether attention might be of use in
research on languages, but the erroneous interpretation of the results his Japanese
collaborators obtained led him to abandon the initial idea, stating that there was
no direct link between attention and expression.

3.  Between semantics and pragmatics

Languages provide speakers with ready-made schemata of expressions having

verbs for pivots (known as verb valences), learned by native speakers from the
very beginning of their linguistic activity. These schemata enable the speakers
to communicate whatever they “mean” about situations of the “world out there”
(as these situations are conceptualized in their mind). However, once the schema
of an expression has been selected, the speakers’ freedom as regards concentrat-
ing their attention on a desired chunk of information as represented in their
mind is considerably limited. Nevertheless, they may transform the schema they
have “at hand” in such a way that eventually it fits their attentional purposes.
For example, they may change the word order or the voice of the verb in case
they need the selected chunk of information be expressed by an attention-driven
phrase (ADP).
Currently, combining notions such as “argument structure” and “verb valence”
linguists attempt to capture the mappings between noun phrases in the syntactic
plane and the roles enacted by the participants in the semantic plane of discourse.
Here, this correspondence is recognised as a more complex problem the solution
of which cuts across another dimension of language, namely, the pragmatic one
where attention-driven saliences (centres of mental representations) are estab-
lished independently of semantic roles and situations. It is for this reason that
subjects and objects are defined as ADPs, being therefore considered as more than
merely ‘formal’ syntactic positions. They are the result of a compromise between
the pragmatic motivations of the speaker and the available linguistic resources
expressing the effects of globally and locally centred attention in base utterances.
Indeed, since the syntactic structure of utterances depends on the pragmatic
(meta-informative) constituency, there is a need to insist on the relationship with
some important pragmatic issues on the one hand, as well as with morphology
and syntax on the other. For this reason, we can observe that the meta-informative
structure of utterances and discourse is very closely related to the linguistic form.
Most pragmatic approaches are rather ‘distant’ from the linguistic form, they deal
with the inferences drawn by the hearer about the speaker’s intentions in order
 André Włodarczyk & Hélène Włodarczyk

to interpret utterances. Obviously, such approaches, along with praxemics (where

language is used in a perlocutory act), also belong to the study of language.
Although these aspects of pragmatics are not characteristic of the MIC theory,
the centrality of pragmatics in this approach leads to an important consequence
concerning the compositionality principle which has been largely admitted in
contemporary linguistic theories since the work of Montague R. Compositionality
is often spelt out as follows: “The meaning of a complex expression is determined
by the meaning of its structure and the meanings of its constituents.” (May 2013,
http://plato.stanford.edu) It is however obvious that in an important number of
uses of language such as that of stereotypes, this principle may be questioned. In
the MIC-theoretical framework this problem can be treated in an innovative way
by establishing a direct relationship between syntax and pragmatics (less directly
with semantics as is often intuitively taken for granted). We consider that the
constituency structure of utterances is a direct mirror of the relationship between
the subject (the globally centred attention-driven phrase) and the predicate (any
information communicated about the subject). We wish to emphasise that in our
view, both subject and predicate are concepts belonging essentially to pragmatics.
However, what makes communication extremely fast and efficient is that there
exists a default relationship between the subject (as attention-driven phrase) and
one of the semantic roles which varies depending on language family. In nomina-
tive languages, the subject of an utterance with a verb in the active voice (which
is the “unmarked” voice in this case) is by default related to active semantic roles.
Thus, when a speaker selects a verb in order to refer to a situation (a state of affairs
in the world out there) s/he can retrieve from his/her memory a syntactic struc-
ture containing the information about the semantic role enacted by its subject;
e.g. in English, the verb eat in the active voice requires as its subject the choice
of a noun referring to an animate being which plays an active role (eater). This
default relationship is different in ergative languages where speakers are provided
with default information concerning the semantic passive role expressed by the
noun phrase in the ergative case. Nevertheless subject remains a pragmatic con-
cept, which makes it possible for languages to use alternative verb voice forms
or verbal lexemes in which the subject does NOT play the default semantic role
relevant to a given type of languages (for example, the active role in nomina-
tive languages). Among verb categories, aspect and tense are essential for meta-­
informative strategies since they enable speakers to choose a point of view or a
point of reference concerning the situation they talk/predicate about. In other
words, these categories indicate which chunk of information (stage or moment,
in the case of aspect and past, present or future, in the case of tense) is being/has
been chosen to fit the centre of attention.
 Introduction 

In conclusion, restrictions on the compositionality principle ultimately boil

down to pragmatic restrictions, or putting it more explicitly, semantic interpre-
tations are dependent on the context of the utterance, be it linguistic or extra-
linguistic. Thus semantic theories cannot deal with the problem without including
pragmatic considerations. The MIC theory – which incorporates pragmatic moti-
vations into the constituency structure of utterances – is an attempt to provide a
systematic account of the limitations of compositionality that can be observed in
human languages. For the same reason, we claim that the pragmatic validation of
utterances as conveying information with either new or old communicative status
precedes in natural languages the truth logical valuation of sentences (true or false),
the latter being the main preoccupation in formal languages.
According to the pragmatic relevance theory (Sperber & Wilson 1986), lin-
guistic communication depends more on inferences about speakers’ intentions
and representations than on the decoding of linguistic expressions. We claim
however that the pragmatic dimension of utterances cannot be modelled without
the syntactic and semantic dimensions. Nevertheless, although semantic content
is the main motivation of linguistic communication, utterances do not express all
the information humans would probably be willing to transmit, all the more so
since speakers cannot communicate content other than from their point of view
and to the best of their knowledge. The attention-driven meta-informative, hence
subjective, dimension of natural languages is represented by the rich choice of
paraphrases of various kinds; e.g. passive or active voices, personal or impersonal
(anonymous) constructions, topicalisation and focalisation, etc. Therefore, human
languages should be treated using more sophisticated theoretical frameworks than
the ones which were elaborated, in an effort to build a bijective (non ambigu-
ous) mapping between form and meaning, for artificial and formal languages, pre-
serving expression from subjectivity and ambiguity. The not fully compositional
conception of natural language semantics lying at the heart of the AS and MIC
theories brings them, in a sense, closer to the relevance theory than might at first
glance seem to be the case.
The syntax and morphology of natural languages make it possible to encode
semantic content which is inextricably interwoven with the pragmatic attitudes of
the speaker. However, in order to analyse expressions made up of heterogeneous
elements (belonging to various dimensions of language), the proposed theoreti-
cal framework is an attempt to explain how morphological forms and syntactic
structures of a particular language are mapped into semantic and pragmatic repre-
sentations. Although some very important concepts (such as subject and predicate)
sound astonishingly familiar and hackneyed, they are integrated in an innovative
coherent theory and their scope is fixed by strict terminological conventions.
 André Włodarczyk & Hélène Włodarczyk

4.  Overview of the contributions to this volume

The first part of this volume contains the presentation of essential issues of the
associative semantics (AS) and meta-informative centering (MIC) theories, the
second part is an attempt at a neuropsychological foundation of this theoretical
framework, and the third part deals with its application to several linguistic phe-
nomena in different languages. Due to its origins dating back to investigations
on typologically different languages and because it is attuned to contemporary
research in the field of computer science, the ASMIC theoretical framework exhib-
its both theoretical and experimental characteristics.

4.1  Associative semantics and meta-informative centering

The AS and MIC theories are characterised by an interdisciplinary approach;
therefore linguistic problems are tackled in connection with neuropsychological
approaches and computer science (artificial intelligence: multi-agent communica-
tion) approaches. As regards language theory, our theoretical approach calls into
question the autonomy of syntax not only in relation to semantics but also to prag-
matics. The first chapter concerns the problem of semantic roles which constitute
the core of the semantic layer of natural languages. The second chapter is devoted
to a formalised semantic description of types of situations in diverse languages,
applying a subset of a universal ontology: this issue constitutes the basis for aspect
studies (Chapter 10). The grounding of the meta-informative old and new status
in discourse and in ontology is tackled in the third chapter, whereas the fourth
chapter concerns attention centered information in different types of languages.

4.2  Neuropsychological evidence for the MIC theory

The second part is an attempt by a neurolinguist Franz Stachowiak, to bring evi-
dence from neuropsychological research on memory and attention to bear on the
meta-informative centering in language. The chapter entitled Semantic and epi-
sodic memory with respect to the ontological grounding of the old and new meta-
informative status aims at integrating a MIC component into a neurolinguistic
model of language production and understanding. The second chapter of this part
is devoted to attention centering in communicative processes and to its relation to
word order and syntax. Stachowiak claims that “verbal processing of information is
strongly influenced by or even rests on the capacity and mode of operation of working
memory and other types of memory and is intricately related to attentional processes,
which play a role in directing the interest of a communication partner in spoken or
written language.”
 Introduction 

4.3  Meta-informative centering in languages

In the third part of the volume are presented the fruits of the application of the
MIC theory to various languages regarding a number of problems which until now
have found no satisfactory solution in approaches where syntax is not connected
with meta-information.

4.3.1  Discourse coherence in English and Japanese

The MIC theory provides a set of definitions of meta-informative operations inde-
pendently of their expression in different languages, be it by lexical, grammatical,
syntactic and/or prosodic markers. Thus, careful distinctions are made between
meta-informative operations and their expression(s). Indeed, probably in any nat-
ural human language, topicalisation is a sort of meta-informative operation, it is
characterised by the old status of the topic contrasting with the new status of the
comment part of the utterance, whatever the linguistic means of expression. For
example, left dislocation is one of them but, as a syntactic device, it may also be
used to express a focus. Unfortunately, as a result of the difficulty of separating the
‘form’ from the ‘content’ in linguistic analyses, syntactic constructions have often
been used to refer to the meta-informative operations they represent; because
of the lack of a formalised axiomatic theory it is impossible to identify common
structures among different languages.
In the chapter by Ana E. Martínez-Insua & Xavier Pérez Guerra, the English it-
cleft is “presented as a device of focalisation used for establishing a ­meta-informative
contrast with the second part of the utterance, making it possible, in a Strict Word
Order (SWO) language as English, to put this focalised constituent at the front of
the utterance (thus contradicting the neutral order: given-before-new)”. This corpus
based study brings evidence for the quite recent consolidation (in the history of
English) of the mechanism of clefting as a meta-informative strategy.
In her paper about subject ellipsis in Japanese Shigeko Nariyama discusses two
ellipsis-related issues using the MIC theory: (1) ellipsis as a discourse coherence
marker, which answers the question: why use ellipsis? (i.e. the pragmatic issue);
and (2) referent identification of ellipsis, which answers the question: how do we
know whom/what a zero encoding of ellipsis refers to? (i.e. a semantic issue). It
shows that centre of attention plays an important role in providing an adequate
and consistent explanation of the mechanism of ellipsis. It also attests to the prom-
inence of implicit lexical knowledge in Japanese that contributes to the referent
identification of ellipsis.
The chapter by Miki Saijo is an attempt to use the concept of centre of atten-
tion for the analysis of a multi-party conversation in Japanese. It is shown that
the CA structure in the multi-party conversation in this case is identical to the
 André Włodarczyk & Hélène Włodarczyk

extended utterance structure of the MIC theory, as a result of analysing discourse

examples collected from the actual meeting conducted in Japanese in a framework
of topicalisation as well as central and peripheral kinds of attention.

4.3.2  Verbal aspect in Slavic languages between semantics and pragmatics

The main function of the linguistic category of aspect is perfectly reflected by the
traditional term “aspect” or “view” which means that the speaker chooses a view of
the situation s/he is speaking about. This view of a situation, or “viewpoint”, is first
of all reflected by an analysis of the internal parts of the situation: moments and
stages. This necessary choice can be compared to that of attention centering with
the purpose of building an utterance (cf. the definitions of subject and object in
Chapter 4 in this volume). As such, aspect is an essential component of the meta-
informative structure of utterances. The internal view of the situation is further
completed by external view parameters concerning its repetition, the modification
of its flow or intensity, and the composition of several situations into one complex
situation.
The ASMIC theory is very helpful in dealing with the blurred boundaries
between semantics and pragmatics as far as the category of aspect use is con-
cerned, making it possible to propose a tentative but comprehensive way out of
endless debates in the field of Slavic aspectology: the problem of aspect pairs, the
difference between aspect and Aktionsart, the amazing differences in the use of
imperfective verbs in Slavic languages and the use of the imperfect tense in French
or progressive forms in English, etc.

4.3.3  T onic and atonic personal pronouns in modern Russian,

classical Greek and Latin
The last three chapters bring an innovative explanation to personal pronoun usage
in modern literary Russian, classical Greek and Latin. Personal pronouns are
markers of discourse coherence both on the semantic and pragmatic levels. On
the one hand, they are used for the identification of entities spoken about ­(taking
part in world situations) in relation to the participants of the discourse situation.
On the other hand, most Indo-European languages have at their disposal two
series of personal pronouns: one of them is unstressed, used in base utterances,
and the other one is stressed, used in extended utterances. Historically, this dis-
tinction corresponds to that of atonic and tonic pronouns in classical grammar,
but with time the use of tonic pronouns as stressed and atonic as unstressed was
blurred either by the disappearance of atonic forms (in modern Russian) or the
incorrect use (with respect to the previous norm) of atonic pronouns as stressed
(e.g. in contemporary Polish, cf. Włodarczyk H. 2012). Therefore, the theoretical
basis provided by the MIC theory makes it possible to distinguish methodically
 Introduction 

between morphology (tonic vs. atonic), prosody (stressed vs. unstressed) and the
pragmatic status (old vs. new) of different types of pronoun uses in discourse. As
pointed out by Jean-Christophe Pitavy concerning classical Greek: “the use of the
accented personal pronoun in conjunction with the verbal form allows the speaker to
construct an extended utterance.” Similarly, Perrine Vedrenne-Cloquet shows that
in Latin: “the pronouns ego and tu featured in discourse have a highly important role
in the construction of meta-informative coherence, since they can be used as topicali-
sations, but also as focalisations. Moreover, they can be used as additional markers:
their second position gives to the first position element a pragmatic emphasis which
it would not otherwise have.”
The three chapters about pronouns emphasise that reference to the old and
new meta-informative status makes it possible to distinguish between topicalised
and focalised uses of tonic pronouns. Moreover, the description of word order in
utterances containing topicalised and focalised pronouns provides evidence for
the fact that topic is the global CA of extended utterances. Olivier Azam con-
cluded from the observation of his Russian corpus that “topicalisation does indeed
appear to be incompatible with postpositioning”. Let us add (this is mentioned in
passing in Chapter 4) that the ellipsis of pronouns and zero-form pronouns (often
referred to indistinctly as pro-drop) should be thoroughly distinguished: the latter
can be attested to only in languages in which the person is relevantly marked by
verb endings (most Indo-European languages) whereas ellipsis is possible both
in Indo-European languages and in languages with no person category in verb
flexion (e.g. Japanese).
The volume is closed by a glossary of AS and MIC terms.

References

Grosz, Barbara J. & Sidner, Candace L. 1986. Attention, intentions, and the structure of dis-
course. Computational Linguistics 12(3): 175–204.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus, and the Mental
Representations of the Discourse Referents. Cambridge: CUP.
Sperber, Dan & Wilson, Deirdre. 1986. Relevance: Communication and Cognition. Oxford:
Blackwell.
Tomlin, Russell S. 1997. Mapping conceptual representations into linguistic representations:
The role of attention in grammar. In Language and Conceptualization, Jan Nuyts & Eric
­Pederson (eds), 162–189. Cambridge: CUP.
Włodarczyk, André. 1999. La validation informative des énoncés et la quantification linguis-
tique. Études cognitives/Studia kognitywne 3: 121–133. Warszawa: SOW, PAN.
Włodarczyk, Hélène. 2012. L’ emploi des pronoms personnels en polonais (par contraste avec le
russe et le français). Études à la mémoire de J. Breuillard sous la dir. de S. Viellard, Revue des
Études Slaves, t.LXXXIII, fasc.2–3, 617–648. Paris: Institut d’Etudes Slaves.
part 1

Associative semantics
and meta-informative centering
 Roles and anchors of semantic situations*

André Włodarczyk
Université Charles de Gaulle (Lille 3) / Université Paris-Sorbonne,
Centre de linguistique théorique et appliquée (CELTA)

Although neither theoretical nor computational linguists did provide

sufficiently careful insight into the problem of semantic roles, recently
some progress is being achieved in robotics (study of the simulation of
human interaction), and mostly in multi-agent systems. Taking advantage
of this motivation and applying it to the study of languages, I distinguish
between various abstract ontological levels. Instead of using such concepts as
agentive, objective, experiencer, etc., on the highest (generic) ontological level,
I postulate generalised agents which are defined by the following ontological
features, among others: (1) features of control (autonomy): goal and feedback,
(2) features of emotion (character): desire and intention, (3) epistemic features
(reason): belief and cognition, (4) communication features (language faculty):
verbal and visual. In accordance with such ontological concepts, natural and
artificial entities are obviously suited to fulfil the semantic roles of agents and
figures respectively in the widest sense of these terms. I further propose to
distinguish between three classes of generic ontological roles, namely active,
median or passive. Here are examples of generic roles: (1) active role (Initiator,
Causer, Enabler, Benefactor, Executor, Stimulant, Source, Instigator etc.),
(2) passive role (Terminator, Affect, Enabled, Beneficient, Executed, Experiencer,
Goal, etc.) and (3) median role (Mediator, Instrument, Benefit, Motor, Means
etc.). Figures can play quasi-active (Q-active) roles.

1.  Human communication as a composite partial function

From a philosophical point of view (phenomenology), it is interesting to observe

that while signs and situations belong to the universe of noumena (they are imma-
nent), representations belong to the realm of phenomena (they are transcendental).

*  Research on this topic has been partly sponsored by a grant of the Hakuhodo Foundation
during the periods from October, 1st 2006 to January, 31st 2007 and from July, 15th 2007 to
September, 14th 2007 when the author was invited researcher at the National Institute of
Japanese Language (Tokyo).
 André Włodarczyk

For this reason, infons can be seen as elements of semantic knowledge (declarative
know-what) as derived (in disorder) from linguistic interactions (communications)
of man. However, once they are arranged in a sequential order and completed by
sets of complementary infons they are enriched by ontological knowledge (proce-
dural know-how). It should be emphasized that this idea is similar to our earlier
proposal1 to organize declarative semantic attributes and their values into Feature
Structures according to their ontological grounding. It is therefore hypothesized
that ontology lies at the basis of any natural language semantics. What is seman-
tic proper for a given natural language is the selection and declarative organisa-
tion of ontological situations, their participants and anchors (locations in time and
space).
Language is a partial function from infons to expressions or vice versa (i.e.:
only a part of what the speaker means can be put into words, on the one hand,
and the hearer must complete every concept he picks up during the commu-
nication process, on the other hand). Thus, the theory of semantic situations
that is sketched hereafter seeks to clarify the fundamental composability (­ partial
compositionality) of invertible functions which can be played by signs and their
combinations in order to enable their users to convey meanings about the enti-
ties of the World. Using the notion of composability we shall attempt to explain
the well-known and now largely accepted fact that linguistic expressions con-
vey only partially the information that the speaker has in mind. In this con-
nection, we shall use the couple of notions encoding/decoding in connection
with the communication of meaning both as it is conveyed in a linguistic mes-
sage (expression) uttered by a speaker and as it means something to the hearer
(a ­signified content).
As an immediate consequence of the above claim, it was necessary to revise
the current linguistic theory of predication. Indeed, it is no longer possible to
maintain that predication takes place when the content of a sentence is “com-
plete”. Instead, our theory postulates the neat separation of the truth conditions
(which are characteristic of information or situation with its roles or, more tra-
ditionally, argument structure relations) from the communicative new/given
statuses of expressions. Thus, it appears necessary to make the following distinc-
tion between two language levels: informative (relational in logical sense) level
and meta-informative (­constitutive in syntactic sense) level. Indeed, in the MIC
(­Meta-Informative Centering) theory, Predication is defined as uttering about

.  Cf. the framework of the research project CASK (Computer-aided Acquisition of
­Semantic Knowledge), Centre for Theoretical and Applied Linguistics (CELTA), Université
Paris ­Sorbonne.
 Roles and anchors of semantic situations 

a State of Affairs (SOA) with either new or given statuses. It becomes there-
fore clear that what is ­communication proper concerns the meta-informative
­status of truth-conditional (relational) information (Włodarczyk A. & H 2006a,
2006b, 2008).

2.  Language and ontology

Since F. de Saussure, natural languages are supposed to be “systems in which every-

thing is interconnected” (systèmes où tout se tient). In modern philosophy ontolo-
gies are no longer concerned with elementary entities which exist in isolation.
Consequently, entities are no more single beings/objects but consist of interrelated
elements and are considered today to be structures. In the case of living beings
viewed as structured objects, they are said to be able to occur in various environ-
ments to which they adapt and with which they interact. Such living beings are
known as agents and their interaction with the universe as well as between them-
selves is called their behaviour (cf. Ingarden R. – 1981). Entities viewed as struc-
tures of elements which develop their behaviour in environments are also agents.
They can be described in terms of the theory of general systems.
Finally, general systems can be formally defined using the following tuple of
concepts:
S = (Wld, Rel, Env, Bhv)

where:

Wld – Set of objects

Rel – Relations between the objects
Env – Environments
Bhv – Behaviours

The behaviour of more than one entities may include interaction with common
goals. Such interaction is known as cooperation and necessarily yields com-
munication. Humans are endowed with a particularly efficient communication
device: language. In the years 1980, a Polish mathematician J. Pogonowski (1979)
proposed a hierarchical reconstruction of structural linguistic theory in which
language is seen as a general system whose sorts are all its levels, objects – all its
(concrete and abstract) units and signature – all its syntagmatic, paradigmatic and
inter-level relations. Because natural languages are general systems themselves they
also exhibit properties of those systems. Consequently, they must be studied in
connection with the ontologies of their bearers (users). As we shall see later, in the
present theory the universe (Wld) contains objects (i.e.: animate and inanimate
 André Włodarczyk

entities) which are related (Rel) through their behaviour (Bhv) to situation roles
within situation frames or environments (Env).
Currently Artificial Intelligence (AI) specialists agree that intelligent pro-
cessing of information requires utilisation of ontological knowledge; i.e.: such
knowledge which may be organized in structured and intelligent databases.
Indeed, ontologies are necessary for designing any information system which is
supposed to interact with humans. According to another tenet of the AI special-
ists, regardless of the overall architectures of information systems, such systems
often should contain more than one ontology. In order to cope with such com-
plex cases, at least two kinds of ontologies are postulated: (1) upper ontologies
on the highest levels of abstraction, and (2) domain ontologies on the lowest
(concrete) levels.

2.1  Semantic theory of natural language from the ontological perspective

Due to the basic partiality of natural languages, we claim that semantic descrip-
tions are motivated by ontologies. If semantics has to deal with meanings expressed
in a given natural language, then there cannot be such thing as “universal seman-
tics” precisely because natural languages differ from each other. Nevertheless,
it seems possible to postulate the feasibility of a useful universal formal generic
(or upper) ontology and of as many as necessary domain ontologies. As a matter
of fact, natural languages are themselves parts of ontologies and represent their
various realisations. Ontologies which are associated with natural languages are
comparable to the generic and domain ontologies of other information systems.
I claim that the cognitive description of the semantic structure of a given natural
language must integrate both generic and domain-specific (in this case language-
specific) ontological knowledge.
In the frame 1 we sketch out a few upper-ontological distinctions in a clas-
sical classification-like manner. However, ontological classifications are seldom
tree-like classifications. For example, there are serious difficulties connected
with the distinction Animate Agent/Inanimate Figure and Natural/Artificial
entities (see Figure 1).

Animate Inanimate

Natural Artificial

Figure 1.  An example of heterarchical (inheritance) relations in ontologies

 Roles and anchors of semantic situations 

UPPER (or GENERIC) ONTOLOGY

1. Entity
1.1. Agent - Animate entity
1.1.1. Human
1.1.2. Non Human
1.2. Figure – Inanimate entity
1.2.1. Material
1.2.2. Immaterial
2. Situation (fragment of the World)
2.1. Situation Frames
2.1.1. State (space without time)
2.1.2. Action (space with time flow)
2.1.2.1. Event (space with time flow but without progression)
2.1.2.2. Process (space with time flow and progression and/
or granularity)
2.2. Situation Roles
2.2.1. Active role (Entities: Initiator, Benefactor, Speaker ...)
2.2.2. Median role (Entities: Mediator, Means, Tool, OSA (the One
Spoken About except the Speaker and the Hearer) ...)
2.2.3. Passive role (Entities: Terminator, Beneficient, Hearer ...)
2.3. Situation Anchors
2.3.1. Start etc. (for spatial anchors), Beginning etc. (for temporal
anchors)
2.3.2. Path etc. (for spatial anchors), Course etc. (for temporal
anchors)
2.3.3. Arrival etc. (for spatial anchors), End etc. (for temporal
anchors)

Frame 1.  Upper (or generic) ontology (for language semantics)

Both natural and artificial entities may be either animate or inanimate. Thus,
the set A of animate entities intersects with the set X of natural entities and the
set Y of artificial entities respectively. Inanimate entities thus correspond to the
difference between natural and artificial entities X–A = {x | x ∈X and not x ∈A} as
regards inanimate natural entities and to the difference between artificial and ani-
mate entities Y–A = {x | x ∈Y and not x ∈A} as concerns inanimate artificial enti-
ties. Inanimate entities are those which belong either to X or Y but not to A. Such
a definition breaks away from the traditional logical classification but may be dealt
with in multiple inheritance systems or in any system able to process m ­ embership
approximation.
 André Włodarczyk

2.2  Entities: Agents and figures

I propose to borrow the generalised notion of agent from Artificial Intelligence and
will call Agent any animate participant of a situation, whereas I will call Figure any
inanimate participant. Both agents and figures may be explicit or implicit. Thus all
situation animate protagonists are agents but linguists generally call them “agents”,
“patients” (or “contra-agents”), “experiencers”, “beneficients”, etc. Agents interact
in particular situations and it is precisely this interaction (interdependency) that
determines their abstract quality of agents in the generic sense.
Animate entities (Agents) are either human (+Hum) or non human (-Hum)
and Inanimate entities (Figures) are material (+Mat) or immaterial (-Mat). It
­happens, however, that situation participants are transformed from Figures to
Agents (agentivation, traditionally known as “personification”) or from Agents to
Figures (figuration, traditionally known as “depersonification”).

Table 1.  The most characteristic features of semantic agents

Characteristic Control Emotive Epistemic Communication
features (autonomy) (character) (reason) (language)
Cognition
Intention
Feedback

Purpose

Verbal
Desire

Visual
Belief

AGENTS

Human + + + + + + + +
Non human + + ± – – – – –

Table 1 enumerates some characteristic features of agents on the highest level

of abstraction. Agents of a situation are defined by several semantic features:
(1)  control features (autonomy): goal and feedback; (2) emotive features (char-
acter): desire and intention, (3) epistemic features (reason): belief and cognition),
(4) communication features (language faculty): verbal and visual.

3.  Relations, roles and anchors

It was probably the theory of Case Grammar2 (Fillmore Ch. 1968, 1971) which
was most influential and very frequently referred to in linguistics. In ­particular,

.  “Case grammar has not become a mainstream part of generative grammar, perhaps because
the phenomena it deals with do not yield to the kind of neat formal analysis that the paradigm
demands. There has been a tendency to push the issues off by arguing that they are ‘semantic’
 Roles and anchors of semantic situations 

it introduced concepts such as agentive, objective, experiencer, etc., in order to

explain meanings of the morphological paradigm of cases (nominative, accusa-
tive, dative etc.). Indeed, Case Grammar can be considered as the first attempt at
elucidating universal semantic roles (known also as thematic roles or argument
structures). Quite recently, Case Grammar reappeared with substantial modifica-
tions in the Berkeley Net Frame (BNF) project, thus becoming the foundational
approach in research in the field of semantic networks. However, the most inter-
esting issues in this domain may be found in Sowa J. (1999) who proposed his
theory of semantic roles following some original ideas of Somers H. (1987) and
Dick J. P. (1991).
As we have seen, from the ontological point of view, everything (Things, Enti-
ties), elements of the world, abstractions (including the constituents of linguistic
expressions to which attributes can be assigned) can be seen as structures. Thus,
attributes assigned to objects appear to be relations over those objects. However,
in the case when objects (things) happen to be relations between other objects,
it is necessary to distinguish between two kinds of attributes: attributes which
designate component parts of relations known as arguments in logic (figures in
mathematics) and attributes which represent the literals in logic (grounds in math-
ematics) of the above mentioned relations.
In order to explain the linear ordering of basic utterances we propose to
model the semantic situation as having three constitutive components:3 frames,
roles and (spatio-temporal) anchors. This model describes information con-
tained in linguistic messages usually referred to as utterances. It suffices that
static semantic situations be represented as spatio-temporally located frames
(“spaces”, fragments of universe) but the representation of dynamic situations
must include also the way situations internally develop in space and time. More-
over, roles determine the “places” (in the frames) that can be occupied by the
entities (called participants when fulfilling some situation roles). We formalize
this as a projection from mental roles of semantic situations into a set of entities
which thereby become their (also mentally represented) participants. However,
it is necessary to distinguish between explicit participants – that have to be rec-
ognized by the hearer at the stage of signification, i.e. signified participants – and
implicit participants – that, in order to be understood, have to be added by the
hearer at the stage of interiorisation or even deeper at the stage of categorisation.

rather than to deal with them in the syntax. However, case grammar has been widely discussed
and has influenced the design of many computer systems for natural language.” (Winograd T.
1983, p. 311)
.  Note that in the proposal recently put forward by Pustejovsky J. [1995] the term “struc-
ture” is used instead of “component”.
 André Włodarczyk

This h­ ypothesis is essential for the present theory of semantic situations which, as
it takes into account the partiality of meaning of linguistic expressions, entails the
need to use, for their formal representation, stratified structures (namely hyper-
graph structures) instead of simple tree structures. Thus, the semantic partiality
hypothesis may be seen as an alternate theory to that of surface/deep structures
because it makes it possible to elucidate incomplete contents, on the one hand,
and to take into account their basically heterogeneous character (conveyed by
linear language expressions), on the other hand.

3.1  Role component

The basic structure of a dictionary being roughly speaking classification of usages
(i.e.: typified uses in their contexts), the usages being named semions, semantic
situations are valid with respect to mental situations. The semantic part of semi-
ons can be represented symbolically using sets of infons.4 Infons are triples 〈i, a, t〉
where i = infon’s name (attribute or relation), a = list of arguments and t = truth
value5 (true and false). In this approach, we add a third truth value, namely the
instantly indeterminate value (or anonymous value) in order to distinguish between
the speaker’s and the hearer’s communication acts. When modelling the speaker’s
activity, the truth value will be most frequently “true” unless he states it other-
wise. In the understanding process of the hearer, that value will be “indeterminate”
unless expressed overtly as “true” or “false” by the speaker. Elementary infons are
unary infons. Sets of elementary infons with the same relation names are called
compound infons. Thus, the names of compound infons are generalisations of the
names of component infons, i.e.: each relation name of a compound infon being
considered as the general name of a given semantic situation. In linguistic expres-
sions, they correspond roughly to verbs.
Here is an example of a compound infon which represents the utterance:
“Brutus killed Caesar.”
S1 |= < kill, (killer: Brutus, killee: Caesar), true >

The above compound infon can be developed as two elementary infons provided
that both of them hold in the same mental situation S1.

.  Cf. Devlin K. (1991).

.  Note that the presence of truth value makes the only difference between infons and predi-
cates in classical logic.
 Roles and anchors of semantic situations 

S1 |= < kill, (killer: Brutus), true >

S1 |= < kill, (killee: Caesar), true >

Now, if we want to represent more information which is associated with the main
information contained in the above utterance, we may wish to add some other
infons which hold in S1.

S1 |= < kill, killer: Brutus, killee: Caesar, true >

S1 |= < at, loc.time: Antiquity, loc.place: Rome, true >

The mental entity ‘Caesar’ could be represented by infons such as:

Caesar |= < name, (family: Caesar, given: Julius), true >

Caesar |= < position, (rank: consul, city: Rome), true >
Caesar |= < kill, (killer: Brutus, killee: Caesar), true >
Caesar |= <     at, (loc.time: Antiquity, loc.place: Rome),
true >
Caesar |= …

We call Roles the binary relations which correspond to elementary infons. Roles
are component relations within the frames of more general situations. Roles either
equal or are included in situations (Roles ⊆ Situations), i.e.: roles are either rela-
tions or proper subsets of relations. Every semantic situation is therefore com-
posed of infons which are partially determined by attributes from both (1) upper
ontologies and (2) domain ontologies. Situation participants of upper ontologies
are seen logically as semantic arguments and were represented in the theory of
situation semantics (Barwise, J. and Perry J. – 1983) as functions from roles to
entities named “anchors”. Note that the term anchor will be used here in quite a
different way.
In this semantic theory roles are defined as pairs of participation functions as
follows:

–– a type participation function from role types to the high level ontological
entities (TP: Role type –> Entity type)
–– an instance participation function from the role instances to the individual
entities (IP: Role instance –> Entity instance)

The instant roles inherit the properties from their types (abstractions: hypernyms,
super-classes). But, due to the partiality principle mentioned above, when generat-
ing or compiling a particular (concrete) role, language users either may be uncon-
scious of many inheritable abstractions or abstractions of entities may be taken
into consideration whenever their instances are indeterminate.
 André Włodarczyk

The Role Component of semantic situations may contain from one to three
elements from the set of three kinds of roles whose generic names are: active,
­passive and median. This theoretical shift6 has important consequences regarding
the very nature of roles which are defined here qualitatively rather than substan-
tially or relationally (as relations between terms). We must not forget that logical
arguments are first of all terms. However, such terms are not order-free. If we want
to make the sequential order of terms free, it is necessary to split each of them
into two parts: the role and its filler (often referred to as participant). We will
repeat this shift as regards our second component of semantic situations, which is
Anchor Component.
There are two kinds of participant types: Agents (animate entities) and F
­ igures
(inanimate entities).

Table 2.  Roles of semantic actions

Roles active median passive
Participants

Agent (animate) Initiator Mediator Terminator

Figure (inanimate) Source Means Goal

Roles enumerated in the Table 2 represent the most usual ones. Here are some
agents or figures: (1) in active roles (Initiator, Causer, Enabler, Benefactor, Execu-
tor, Stimulant, Source, Instigator etc.), (2) in passive roles (Terminator, Causee,
Enabled, Beneficient, Executed, Experiencer, Goal, etc.) and (3) in median roles,
(Mediator, Instrument, Benefice, Motor, Means, Matter etc.). While active and
passive roles are considered here as primitive roles, median roles are definable by
introducing two embedded associated situations (a-situations) in which (a) the
participants of median roles fulfil passive roles in the first of the associated situa-
tions and (b) they fulfil active roles in the second associated situation. I distinguish
therefore role associations (see the above discussion about infons) as explicitly
expressed by linguistic utterances from situation associations which are implicitly
“responsible” for the creation of median roles (but note, however, that associations
of semantic situations are also often marked in languages; i.e.: have explicit mor-
phological markers, for example prefixing or suffixing verbs or using compound
verbs). Let us also add that there may be more than one median roles in the same

.  Note however the rather exceptional idea of Tesnière L. (1959) who theorized about roles
using the abstract name “actant” (acting entity) and enumerated three of them: first actant,
second actant and third actant.
 Roles and anchors of semantic situations 

utterance, nevertheless this is rather rare. In general, linguistic information con-

tains no more than maximum 3–4 roles and 3–4 anchors. This point is crucial for
the present theory of associative semantics.
In general, active roles are filled by agents and passive roles by figures, but
this is not a rule. Initiator and Terminator are entities fulfilling Active and Passive
Roles respectively. Agents typically fit the dynamic situations (Actions) whereas
figures fit static situations (States). When it is not so, shallow level i.e.: partially
specified semantic level is probably needed. When the Subject of a default (active
or passive) diathesis sentence does not correspond to the default (active or pas-
sive in that order) semantic role of a given situation, the shallow level7 of meaning
must be introduced. In the cases of agentivation or figuration (see above), the shal-
low role names will be prefixed by the capital letter Q-… (as in quasi-). For exam-
ple, “Q-initiator” “Q-source” will be said to designate the figure in an active role
(which normally fit to agents) and “Q-source” will be said to designate the agent in
an active role (which normally fit to figures). Research concerning the similarity
of Proper Roles with Quasi-roles would probably require detailed exploration of a
number of analogies in the ontological domain.
Hereafter is an example of transformation of the information contained in
“The key opened the door” using participation role types only.
Shallow level semantics:

< open, (quasi-active-role: ‘key’, terminator: ‘door’), true >

Standard level semantics:

< open, (initiator: x, means: ‘key’, terminator: ‘door’), true >

Note that in standard level semantics, it is necessary to recognise that there is an

indeterminate initiator x and that the former quasi-initiator has been transformed
into a median role (because it is passive with respect to the initiator and again
quasi-active with respect to the terminator).

3.2  Anchor component

Indeed, it is possible to model the Anchor Component of semantic situations (see
Table 3) in a similar way as their Role Component. However, we must be aware
of the fact that while the participation functions are projections into the entities,
anchoring functions are projections into other situations.

.  Note, however, that due to the general principle of partiality of linguistic expressions both
typified participants and their instances may happen to be indeterminate.
 André Włodarczyk

Table 3.  Anchors of semantic situations

Anchors initial intermediary terminal
Localisation

Space Start Path (Itinerary) Arrival

Time Beginning Course End

It is namely because both Components of semantic situations (both Roles and

Anchors) can be defined as pairs consisting of (a) the Role and (b) the Participant,
on the one hand, and those of (a) the Anchor and (b) the Location, on the other
hand, that in the past some linguists have had the intuition of the localistic view8
of the semantic structure of linguistic expressions.

4.  Associative semantics: Examples of analysis

In this theory, hypergraphs are thought of as the most adequate representation

device for describing associative meaning of information (as expressed by linguis-
tic utterances). Because set-theoretical formalisms generally used in theories with
hypergraph representations are rather cumbersome, I provisionally introduce the
following (not fully formal) representation in which I will abandon to use quite
explicitly the above described Type/Instance participation functions. The body of
the associative situation (A-Situation) semantics being recursive, in order to show
the flavour of this analysis I defined the following simple semantic representation
language inspired by the CDL:9

SITUATION := [Sit: HEAD + BODY + FOOT]

HEAD := {relation + variable(s)}
BODY := {participant(s)} + {role(s)} and (optionally) SITUATION
FOOT := {(neutral) role < {relation}}
FOOT := {active role < {relation} > passive role}

While HEAD contains shallow level valence FOOT corresponds to the standard
level valence. BODY may contain other situations which are thus associated with
the main (linguistically marked) situation. The above formalism enables the analy-
sis of the semantic situation (meaning) of the utterance #1 as follows:

.  Note that in this approach the analogy goes from Roles to Anchors rather than backwards.
.  CDL (Concept Description Language) is designed by Yokoi T. et al. (2005–2007).
 Roles and anchors of semantic situations 

#1. Brutus killed Caesar.

[SIT0: ‘kill’
    HEAD kill x y
    BODY
   {kill relation property=asymetric effect=death:
     {x Brutus participant=agent: }
     {y Caesar participant=agent: }
     {Role1 x=killer generic=active type=initiator }
     {Role2 y=killee generic=passive type=terminator }
     {Role1 < kill > Role2}}
    FOOT
   {Brutus < kill > Caesar}]
#2. Brutus killed Caesar with a knife.
[SIT0: ‘kill’
    HEAD kill x y z
    BODY
   {kill relation property=asymetric effect=death:
     {x Brutus participant=agent: }
     {y Caesar participant=agent: }
     {z knife participant=figure: }
   {Role1 x=killer generic=active type=initiator }
   {Role2 y=killee generic=passive type=terminator }
   {Role3 y=weapon generic=median type=origin }
[SIT1:
    HEAD use x y
    {use prop=asym purpose=tool:
       {Role1 x=user generic=active type=initiator: }
       {Role2 y=weapon generic=passive type=terminator:}
       {Role1 < use > Role2 }}
    FOOT
       {Brutus < SIT1:use > knife}]
[SIT2:
    HEAD causeDie x y
   {causeDie prop=asym effect=death:
       {Role1 generic=q-active type=origin object=weapon: }
       {Role2 generic=passive type=terminator }
       {knife < causeDie > Caesar }}
    FOOT
       {knife < SIT2:causeDie > Caesar: }]
FOOT
{Brutus < SIT1:use > knife*knife < SIT2:causeDie > Caesar}]
 André Włodarczyk

Note that in the analysis the median role (fulfilled by ‘knife’) has been split into
two primitive roles (passive and q-active) each belonging to another associated
situation (use x y and causeDie x y).
Let us analyse now the difference between #3a and #3b beneath.

#3a. Peter gave a book to Mary.

#3b. Peter gave Mary a book.

Using the distinction between the informative and meta-informative levels

(Włodarczyk A. & H. 2006a, 2006b), it is possible to interpret the above utter-
ances as follows. Due to the English syntax word order, “Peter” in both utterances
can be easily analyzed as their global Centre of Attention (Subject). It is clear that
“a book” in #3a and “Mary” in #3b correspond to the local Centre of Attention
(Object). Let us mention in passing that the so-called “Oblique Object” should
not be considered as a centre of attention. Obviously, this is the only difference
between #3a and #3b. From the point of view of the information conveyed by the
utterances, we cannot but recognize one single meaning which can be represented
by the following recursive blocks:

[SIT0: #3a and #3b

    HEAD give0 x y z
    BODY
   {Peter participant=agent: }
   {Mary participant=agent: }
   {book participant=figure: }
[SIT1: ‘give1’
    HEAD give1 x y
    BODY
   {give1 x y
     {Role1 x=giver generic=active type=initiator }
     {Role2 y=object generic=passive type=terminator }
     {Role1 < give1 > Role2}
   FOOT
   {Peter < give1 > book}]
[SIT2: ‘reward’
    HEAD reward’x y:
    BODY
    {Role1 x=agent generic=active type=benefactor}
    {Role2 y=agent generic=passive type=beneficient }
    {Role1 < reward > Role2}
FOOT
      {Peter < reward > Mary}]
 Roles and anchors of semantic situations 

[SIT3: ‘make_happy’
    HEAD make_happy x y
    BODY
   {Role1 x=Q-agent generic=active type=benefactor}
   {Role2 y=agent generic=passive type=beneficient }
   {Role1 < make_happy > Role2}
FOOT
    {book < make_happy > Mary}]
FOOT
{Peter < SIT2:give1 > book*book < SIT2:reward > y = Mary}]
or
{Peter < SIT3:give1 > book*book < SIT3:make_happy > y = Mary}]

The noun “book” has been interpreted as fulfilling the median roles defined by two
pairs of associative situations “give1” and “reward” (meaning 1) as well as “give1”
and “make_happy”.
Here are a few other examples of semantic situations containing median roles
(see Table 4). It is important to note that the difference between “to give” and “to
receive” is meta-informative only, but the difference between “to sell” and “to buy”
concerns in addition the informative contents. This becomes obvious if we con-
sider that there might be another median role (fig-value object=money) attached
to both the pair “to sell” and “to buy”, but not to “to give” and “to receive”. It is
also for the same reason that the standard level meaning of both “to sell” and “to
buy” includes associations of their reciprocal situations (i.e.: “to sell” requires the
embedded a-situation [SIT1: ‘buy”], on the one hand, and “to buy” requires the
embedded a-situation [SIT1: “sell”], on the other hand).

Table 4.  Examples of semantic roles

Active role Median role Passive role

x gives y to z x = agt-initiator y = fig-mediator z = agt-terminator

z receives y from x (giver) (given object) (receiver)
x sells y to z x = agt-initiator y = fig-mediator z = agt-terminator
(seller) (sold object) (buyer)
x buy y from z x = agt-initiator y = fig-mediator z = agt-terminator
(buyer) (bought object) (seller)

As we have said, we have called shallow level the semantic interpretation

of Valence in which, as a rule, one participant cannot fulfil but one single role
(note that Fillmore’s semantics is shallow in this sense). If we admit that it
is acceptable that one participant fulfil more than one role in the described
situation while our meta-linguistic analysis remains still declarative, we will
 André Włodarczyk

consider such interpretation as standard level meaning, leaving room for the
procedural extensions giving rise to deep semantics.
We now present the analysis of an utterance containing the median role which
results from the association of more deeply embedded a-situations.

#4. “Peter bought a car for his daughter.”

[SIT0: ‘buy’
 HEAD buy x y z
 BODY
  {Peter participant=agent: }
  {car participant=figure: }
  {daughter participant=agent: }

[SIT1: ‘acquire’
 HEAD acquire x y
 BODY
 {acquire x y
    {Role1 x=buyer generic=active type=initiator }
    {Role2 y=bought generic=passive type=terminator }
    {Role1 < acquire > Role2}}
 FOOT
 {Peter < acquire > car}]

[SIT2: ‘offer’
 HEAD offer x y z:
 BODY
    {Role1 x=donor generic=active type=benefactor}
    {Role2 y=receiver generic=passive type=beneficient }
    {Role1 < offer > Role2}

[SIT2.1: ‘give’
 HEAD give x y
 BODY
    {Role1 x=giver generic=active type=owner }
    {Role2 y=gift generic=Q-active type=ownedObject }
    {Role1 < give > Role2}
 FOOT
    {Peter < give > car }]
[SIT2.2: ‘reward’
 HEAD {reward x y }
 BODY
    {Role1 x=rewarder generic=Q-active type=initiator }
    {Role2 y=rewarded generic=passsive type=terminator }
 Roles and anchors of semantic situations 

 FOOT {car < reward > daughter }

  {Role1 < reward > Role2}
  {car < reward > daughter} ]
FOOT
    {Peter < offer:give > car }
    {car < offer:reward > daughter} ]
FOOT
  {Peter < {buy:acquire} > car }
  {Peter < {buy:offer:give } > car}
  {car < {buy:offer:reward } > daughter}]

STANDARD LEVEL VALENCE SEMANTICS:

In the above representation, the agent ‘Peter’ fulfils two active roles as ‘buyer’
{Peter < buy:offer:give > car} and as ‘donor’
{car < buy:offer:reward > daughter}

while the figure ‘car’ fulfils the median role because in the [SIT1 ‘buy:acquire’]
it fulfils a passive role as ‘bought’ in {Peter < buy:acquire > car} and it fulfils a
Q-active role in {car < {buy:offer:reward } > daughter}.

5.  Further research

At the basis of ontological theories there are classifications. Theories and Logic
provide formal languages which can determine whether the given ontological
classifications are satisfiable with respect to the knowledge as conveyed by well-
formed expressions of a formal language. It is clear that such kind of classification
with a language is in fact a meta-classification (cf. Kent R. 1998).
However, when the formal language expressions are seen as type descriptors
and the uses of natural (human) language expressions are seen as instances, the
corresponding classifications are partial with respect to the ontological meta-­
classifications from which the formal descriptions are derived. In such cases, the
ontological type description formal language is a semantic meta-language describ-
ing the given natural language.

References

Barwise, Jon & Perry, John. 1983. Situations and Attitudes. Cambridge MA: The MIT Press,
Bradford Books.
Devlin, Keith. 1991. Logic and Information. Cambridge: CUP.
Dick, Judith P. 1991. A Conceptual Case-Relation Representation of Text for Information
Retrieval. Ph.D. dissertation, University of Toronto.
 André Włodarczyk

Fillmore, Charles J. 1968. The case for Case. In Universals in Linguistic Theory, Emmon Bach &
Robert T. Harms (eds), 1–90. New York NY: Holt, Rinehart & Winston.
Fillmore, Charles J. 1971. Types of lexical information. In Semantics, Danny Steinberg & Leon
A. Jokobovitz (eds), 370–392, Cambridge: CUP.
Ingarden, Roman. 1981. Spór o istnienie świata, Tom III : O strukturze przyczynowej realnego
świata (The Controversy over the Existence of the World, Vol. III: On the Causal Structure
of the Real World), translated from German into Polish by Danuta Gierulanka. Warszawa:
PWN.
Kent, Robert. 1998. The model theory of ontologic. 〈http://www.ontologos.org/Papers/Papers.
html〉
Pogonowski, Jerzy. 1979. Formal methods in linguistics. Buffalo Papers in Linguistics 1(3): 31–83.
Pustejovsky, James. 1995. Generative Lexicon. Cambridge MA: The MIT Press, Bradford Books.
Somers, Harold. 1987. Valency and Case in Computational Linguistics [Edinburgh Information
Technology Series 3]. Edinburgh: EUP.
Sowa, John F. 1999. Knowledge Representation: Logical, Philosophical, and Computational Foun-
dations. Pacific Grove CA: Brooks Cole.
Tesnière, Lucien. 1959. Éléments de syntaxe structurale, Paris: Klincksieck.
Winograd, Terry. 1983. Language as a Cognitive Process, Vol. 1: Syntax. Reading MA:
Addison-Wesley.
Włodarczyk, André & Włodarczyk, Hélène. 2006a. Subject in the meta-informative centering
theory. Études cognitives / Studia kognitywne 7: 39–64. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2006b. Focus in the meta-informative centering
theory. In La Focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André & Włodarczyk, Hélène. 2008. The Pragmatic Validation of Ulterances. In
Études cognitives/Studia kognitywne 8: 117–128. Warszawa: SOW, PAN.
Yokoi, Toshio et al. 2005–2007. Concept Description Language – CDL. Core Specifications, Ver-
sion 1.0. Tokyo: Institute of Semantic Computing (ISeC).
 Frames of semantic situations

André Włodarczyk
Université Charles de Gaulle (Lille 3) / Université Paris-Sorbonne,
Centre de linguistique théorique et appliquée (CELTA)

In different languages, the means of expressing a situation are shared

variously between grammar and lexicon and must be represented by semantic
configurations specific to a particular language (or family of languages).
Situations are semantic ‘nests’ for several verbs which allow for the expression
of a variety of aspects in the course of communication. For this reason, from a
theoretical point of view, semantic situations can be seen as independent of the
language in which they are expressed. Research into aspect in diverse languages
showed that an aspect could not be explained without first describing the
semantic situation. Thus numerous attempts have been made to classify semantic
situations,1 especially after the classification proposed by Z. Vendler (1957 and
1967) for explaining aspect in English. As verbs in natural languages normally
express complex situations, semantic situations are made up of other situations,
with the result that it is possible to speak of situational aggregates. In other
words, situations are mixed up with one another. We maintain, however, that it
is possible to separate out a number of situation types and their constituent parts
by considering the possibility of classifying them from two points of view. We
will hypothesise a distinction between frame situations and role situations. In this
article we will consider frame situations alone. Role situations are dealt with in the
preceding chapter. Some of the criteria used in our approach have been variously
taken into account by different theoreticians, but most have confused the frame/
role distinction and have thus arrived at less homogeneous classifications than
our own. Looking somewhat further ahead, we think that situations can equally
be classified according to their role components, which could be by showing
their interactions (situations that are intransitive, transitive, convertible etc.) or
their relationship with, amongst other things, centres of attention that are global
(subject) or local (object), or again, the nature, countable or otherwise, of the

.  François J. and Denhiere G. (1997, p. 125) list – from 1967 to 1994 – 34 different pro-
posals for the conceptual classification of verbal predicates; amongst which are the works of
W. Chafe, S. Dick, Ch. Fillmore, J. François, R. Jackendoff, R. Martin, Z. Vendler, A.R. Verkuyl
etc. In the domain of Slavonic linguistics alone one may cite amongst many others, Bulygina
T.V. (1982), Laskowski R. (1998), Piernikarski C. (1969), etc.
 André Włodarczyk

participants, but we will not be dealing with these here. The aim of this study
is to define situations from the point of view of their “internal construction”
(i.e. without taking into account modality, tense, aspect etc.) It is indeed desirable
(1) to put forward a coherent system of classifying situations based on a small
group of well defined primitives and (2) to show the way in which different verbal
expressions, used in context, may inherit characteristics from partially organised
(hierarchised) situations, taking into account the primitives that have been
identified.

1.  The structure of semantic situations

This research on semantic situations follows on from our work based on the
hypothesis of parallel processes which are supposed to characterise linguistic and
cognitive operations. Furthermore, as a working hypothesis, we postulate that
semantic situations (i.e. situations expressed by linguistic means) have a discrete
character (as indeed does everything that we perceive) in spite of the fact that
epistemic agents give the impression that they capture the continuum. We also
propose (as is frequently put forward) that it must be possible to classify all situa-
tions by using as small a group of primitives as possible, and that these, as such, do
not directly represent any situation.
Definition 1: A “situation” is any state or course of events contained within limits
called “frames”.

1.1  Space
Definition 2: There are two types of situation: static (the state of things) and
dynamic (the course of events or actions).
All situations have this in common, that they are located in space (their prop-
erties are plotted in three-dimensional space), and that they are distinctive in the
way in which they are concerned with time. State is distinguished from the three
other kinds of situation known generically as action situations (c.f. hereafter) by
the criterion of dynamism, that is, by the fact that the passing of time brings about
change therein.
Definition 3: The constituent elements of a static situation are its “place” (in the
universe of entities) and its “periphery” (places of untimed transition)2

.  Transition is here used as it is understood in the network of bipartite graphs, known as
“Petri nets”, see later, in the section on representation and formalisation.
 Frames of semantic situations 

1.2  Time
All situations (static and dynamic) are surrounded in time by neighbouring situa-
tions: preceding or subsequent. As we will see later, neighbouring situations act as
transitions for the situations in question.
Proposition: Dynamic semantic situations are structured.
Definition 4: The constituent elements of a dynamic situation are its “moments”
(timed transitions) and its “stages” (the interval from one transition to the next).
From a conceptual point of view, all situations should only be characteristic.
A situation can be said to be characteristic when all the moments and all the stages
of its “life cycle” are taken into account, although semanticians often emphasise that
languages rarely express characteristic situations without selecting within them any
moment or stage (for example, in French, the verb “bouger” would represent the
characteristic situation expressing the abstract concept of “movement”), while the
language vocabularies are made up of more words which contain aspectual semes3
(for example, the French verb “arriver” would express the final phrase of the charac-
teristic situation “bouger”, or more precisely, of one of its hyponyms “se déplacer”).
Thus certain utterances express what we will call whole situations, whereas others
present situations that are analysed into moments and/or stages.
Not only are situations analysed – as we have said – into “internal” moments
and stages, but they are also related to neighbouring situations that are often taken
into account in linguistic expression. Indeed, what happens in the world is never
perceived “in a vacuum” but in relation to what comes before and follows on after-
wards. Many linguists consider that situations can be seen as subject to “life cycles”
with the result that within them we can see the following three – inner – stages:
“the starting stage” (begin), “the middle stage” (run) and the “final stage” (end). It
is to these three fundamental stages in particular that we add two segments, one
on either side, representing the preceding neighbouring situation, (before) and the
following neighbouring situation, (after). These two preceding and following situ-
ations must be considered as outer stages of the situation in question; they are
therefore its immediate preparation and consequence.4

.  In fact the categories of tense and aspect in a given language involve semantic ­situations;
they allow the user, by diverse inflectional, derivational or syntactical means, to choose
­different ways of looking at semantic situations.
.  According to the triple of sequential processes (Hoare C.A.R., 1969), operations are
­preceded by conditions (known as “preconditions”) and followed by consequences (known
as “postconditions”).
 André Włodarczyk

Proposition: Dynamic semantic situations are complex. This complexity results

from the fact that situations do not exist without participants or without the roles
they play.
Given that we are for the time being only dealing with situation frames, these
may seem to be simple (i.e. representing one single process). In this case, by rep-
resenting the situation frame as a straight line (following the linear time model ),
we obtain the distribution of the elements that make up situations as in Figure 1.
Each of the constituent elements is here represented by points (showing moments)
and by intervals limited by markers (showing stages). The same linear time model
is normally sufficient when representing parallel processes, thus it can also be used
for complex situations, since the straight line in Figure 1 represents the course of
events in a complete situation and is not a measure of time.

MOMENTS
initial start enter exit finish terminal

< before > < begin > < run > < end > < after >
STAGES

Figure 1.  Sequential cycle of a simple dynamic situation

Our definitions of semantic situations are cognitive in nature (and not onto-
logical or phenomenological). We distinguish between the four different situa-
tions: (1) states, (2) events, (3) ordinary processes and (4) refined processes. It should
be remembered that despite certain similarities, these terms are either used with
different meanings in other theories (states, events, processes) or are our own terms
(ordinary processes/refined processes).
In dynamic situations, entities undergo changes as a result of the fourth
dimension (time). All dynamic situations develop or progress (a) by moments:
start, enter, exit and finish and (b) by stages (ordinary and refined processes have
three distinct stages: begin, run and end.) The middle stage (run) does not appear
in events where the beginning is immediately followed by the end, because events
are perceived without internal duration.

1.3  Progression
Ordinary and refined processes are characterised not only by time but also by
progression,5 which is defined as the movement from one stage to another.
Although time is here a necessary condition, progression must be understood as

.  cf. Mazurkiewicz A. (2000).

 Frames of semantic situations 

a succession of stages. Mazurkiewicz A. (2000) sees progression as “continuous

change” which is added to the interpretation of states and events.6 The problems
of progression (as well as those of regression) are widely discussed in the theory of
situation calculus, by both logisticians and roboticists7

1.4  Granularity
The fourth concept (but the sixth dimension) which we feel we must bring into the
situations model concerns the cognitive granularity8 of stages. Here we must ascer-
tain (clarify) the way in which progressions evolve. Granularity can be defined as a
conjunction of selectors of semantic features (such as intensity, speed, size, weight,
strength etc.). For example, for situations where the granules involve the speed of
development, the semantic features can be either acceleration (ti = vj & ti+1 = vj+1,
i.e. in ti time, speed is vj and in time ti+1 speed is increased and is equal to vj+1) or
deceleration (ti = vj & ti+1 = vj–1, i.e. in ti time, speed is vj and in time ti+1 speed is
decreased and is equal to vj–1). Most situations with granulated stages are dual
pairs of selectors of semantic features concerning (a) the orientation of physical
movement (left/right,9 up/down,10 forwards/backwards), and (b) the orientation
of psychological movement (movement of the spirit), polarity (yes/no,11 more/
less) and intensity (strong/weak)12 etc. Although the concept of granularity is close
to that of discreteness, we will take care not to confuse them.

2.  The partial ordering of semantic situations

The regularity of the defining differences between state and actions (event, ordi-
nary and refined process) can be represented in a table showing their position in
a hierarchy (Table 1): the more complex a situation, the more defining features it

.  cf. Mazurkiewicz A. (1986).

.  cf. Reiter R. (2001).
.  The author of the metaphor “a granule of knowledge” was Zadeh L. (1996. 1997). This
concept, first reinterpreted within the framework of the theory of rough sets by Pawlak Z.
(1981, 1982, 1991), was then generalised by Polkowski L. and Skowron A. (1999).
.  e.g. In French “se balancer” is used for a pendulum, but “en avant/en arrière” for a child’s
swing.
.  e.g. In Polish “machać ręką” (to wave one’s hand) but in Japan the gesture designated by
the expression “te wo furu” is oriented left/right.
.  e.g. In French “hésiter” between a positive or a negative decision or between a range of
decisions.
.  e.g. In French “clignoter” or in Polish, “migotać”.
 André Włodarczyk

has. The hierarchical relationships thus elucidated can be compared to the order
of classic logic where propositional logic is zero order (order 0), predicate logic is
first order (order 1) etc. In the same way, states (static situations) can be seen as
being order 0, events, and ordinary and refined processes are successive orders.
Thus different situations are included within one another and organise themselves
in the following way:
Situation = State ⊆ Event ⊆ Ordinary Process ⊆ Refined Process

This formula reads as: “Situations are ordered: state ⊆ event ⊆ ordinary process ⊆
refined process, from the smallest to the largest”. Interlocked situations make up a
hierarchical order where each kind of situation inherits the properties of the one
that precedes it. The inclusions that we propose in order to define the hierarchical
order of types of situations are only concerned with their properties (dimensions).

Table 1.  The hierarchy of semantic situations (tabular lay-out)

Characteristic SITUATIONS
properties Static situations Dynamic situations (ACTIONS)
(dimensions)
STATE EVENT ORDINARY PROCESS REFINED PROCESS

Space (3D) + + + +
Time – + + +
Progression – – + +
Granularity – – – +

This table brings to mind certain existing classifications of verbs into semantic
groups. As both Vendler Z. (1967) and his successors realised, semantic situations
are not notions expressed by verbs alone. The problem is a little more complex
since we are here dealing with semantic types which sometimes have no direct
equivalent in the form of a linguistic expression, but which can be found amongst
the semes contained in (a) verbs themselves, (b) verbal periphrases, (c) presuppo-
sitions, (d) communicative acts, (e) knowledge etc. Verbs are not therefore neces-
sarily being dealt with for their properties as lexemes, but as verbs employed in
utterances for the purpose of designating situations. Depending on the context13
in which it is used, the same verb can express different kinds of semantic situations.

.  By context, we mean not just the aspect-temporal inflections of a verb, but also the
participants and the anchors with which it is used.
 Frames of semantic situations 

2.1  The hierarchy of types of situation

One can equally well represent situation types by binary oppositions (Figure 2).

SITUATION FRAME

– +
without time with time

– +
without progression with progression

– +
without granularity with granularity

STATE EVENT Ordinary Granular

PROCESS PROCESS

Figure 2.  Hierarchy of semantic situation frames (flow chart)

It goes without saying that the inclusion relationships which we have shown
between the different kinds of semantic situations, allow situations to be formed
in two ways: successively or in parallel. When one situation follows another, their
make-up follows the law of transitivity in inclusion relationships. Winston M. E.,
Chaffin R. and Hermann D.J. (1987) identify the inclusion relationship of classes
which precedes the inclusion relationship of composition in the hierarchy, and this
in turn precedes the spatial (topographical) inclusion relationship. These authors
observed that mixed inclusion relationships are valid if, and only if, the conclusion
expresses the weakest relationship in the sense of the hierarchical order existing
between these inclusion relationships.
This formal property of inclusion relationships has an interesting application
in the description of semantic situations expressed in languages. In the case that
we are dealing with here, concrete verbs, even when they express situations of an
identical type (e.g. processes), must obey the transitivity of inclusion if their lexi-
cal meanings refer to inclusion of different sorts. The weakest inclusion is: state ⊆
refined process. That is why states often follow processes or, to put it another way,
processes are often in a composition relationship with states (c.f. result aspect) but
in the case of the inclusion state ⊆ event, it is obvious that it is events which should
be composed with states.
 André Włodarczyk

2.2  The situation frames model

Our model of semantic situation frames is, by its very nature, cognitive. For
example, the granules of progression allow us to look at the cognitive refinement
with which certain situations (refined processes) can develop (not only nuances of
expression). Thus we define a situation as a quadruple containing (1) the name of
the situation, (2) its type, (3) its frame and (4) its roles.
Situation = { Name, Type, Frame, Roles }

Names are the names of the verbs (or adjectives) and verbal (or adjectival) phrases.
There are four types: states and events, ordinary and refined processes. So as to
comply with the theorising tradition in linguistics, however, we also distinguish
between static situations (states) and dynamic situations (actions: events and
processes).
Situation frames have dimensions which determine their type, to which the
parameters “moment” and “stage” are added.
Situation Frame = { Dimension, Moment, Stage }

Figure 3 shows the situation frame seen as a “whole situation” or analysed by

“moments” and “stages”.

Components of a situation frame cycle

WHOLE MOMENT STAGE

initial | start | enter | exit | finish | terminal before | begin | run | end | after

Figure 3.  Conceptual view of components of situation frame

States only have the “space” dimension, they are states of things but they can
be transformed into events or even processes. Events are characterised by the fact
that they have no “middle” element, and as a result, no “run” stage. Processes show
all moments and all stages, but Ordinary processes are distinguished from Refined
processes by the absence of granularity in their progression. The order of the differ-
ent types of situation frames can be seen in Table 2.
It is important to note that in spite of the presence of the parameters nec-
essary for the representation of aspect (understood as including “modes of
action”), the structure of the situation does not in itself show any characteristics
of aspect. It corresponds to the semantic nest which is the idealised meaning of
 Frames of semantic situations 

Table 2.  Distinctive features of semantic situation frames

DIMENSION MOMENT STAGE

SITUATION TYPES

GRANULARITY
PROGRESSION

TERMINAL
BEFORE
INITIAL

FINISH
ENTER

AFTER
BEGIN
START
SPACE
TIME

EXIT

RUN
END
STATE + – – – – – – – – – – – – – –
EVENT + + – – – + – – + – + – – – +
ORDINARY PROCESS + + + – + + + + + + + + + + +
REFINED PROCESS + + + + + + + + + + + + + + +

a family of verbs or adjectives (or indeed verbal or adjectival expressions). At

the level of usage, expressions containing the same verbs or adjectives can cor-
respond to different semantic nests in turn, thanks to those mechanisms (the
addition of adverbs, p­ refixes etc.) which allow a choice between the different
parameters.

3.  Examples of types of semantic situations

The dimensions which we have noted (space, time, progression and granularity)
have allowed us to identify four kinds of situations and we will subsequently give
some language examples of these. In particular, we show that certain aspect pro-
cedures (prefixes, conjugated forms, syntactical constructions) allow us to change
the association of a verb from one semantic type to another, for example a verb can
be changed in this way from a verb of process into a verb of event. In this article,
we will simply give some examples without going into the details of the prob-
lem of links between aspect parameters and semantic types. We must nonetheless
emphasise the fact that the situation structures that we have defined, in particular
moments and stages, whilst remaining theoretical, must, independently of aspec-
tualisation, be considered inherent to the situations.

3.1  States
STATE is the simplest kind of situation, and is defined as a situation in space which
lasts for a greater or lesser period of time but which is not affected by the passing
 André Włodarczyk

of time since it remains identical throughout. We therefore consider that the time
parameter is not applicable to states. In a state, it is not possible to distinguish
between successive stages. Although states can have a duration, we do not make
any distinction between the state-moment and the state-duration.
    Mary is feeling bad.
    Marie se sent mal.
    Maria czuję sie niedobrze.
    At that time all inhabitants of the district were in their beds.
    A cette heure-là tous les habitants du quartier étaient couchés.
    O tej porze wszyscy mieszkańcy tej dzielnicy leżeli w łożku.

It is possible to consider the start of a state (viewed as another situation preceding

it) or the end of a state (considered as another situation following it). Languages
have various means of expressing these different ways of looking at states.

3.2  Events
An EVENT, which is a momentary dynamic situation (that is, without progres-
sion), has a “start” and a “finish” that are so close that it is not possible to distin-
guish a run stage between them. Indeed, even if, in reality, nothing can happen
without occupying an interval of time, however minimal that may be, in linguistic
representation man conceives of events as instantaneous and does not take into
account the real time interval which they occupy. On the other hand, events can
be used to express the start or finish moment of a state, an ordinary process or a
refined process.
Depending upon the situation which precedes or follows an event, it is pos-
sible to identify three kinds of event. A type E1 event occurs and vanishes with no
relationship to a preceding or subsequent event, and can be thought of as a very
short moment of time within a state, no sooner starting than finishing.
   Suddenly he shouted.
   Nagle wrzasnął.
    Soudain il poussa un hurlement.

A type E2 event precedes a state or an ordinary process the duration of which is

not stated, and it can therefore be thought of as a starting point.
    Yesterday Peter fell ill.
   Piotr wczoraj zachorował.
    Pierre est tombé malade hier.

A type E3 event ends a state or an ordinary process: it can be seen as a terminal

point.
 Frames of semantic situations 

    Peter stopped being irritating.

    Piotr przestał być dokuczliwym.
    Pierre a cessé d’être énervant.

3.3  Ordinary processes

PROCESSES are defined by progression and can therefore be analysed in three
stages (start, middle and end). An ORDINARY PROCESS is defined as a situa-
tion located in space and it develops with the passage of time (progression). Three
stages can be seen in an ordinary process: initial, intermediate and final.
A dark shape was approaching in their direction.
Jakaś ciemna sylwetka coraz wyraźniej zbliżała się w ich stronę.
Une silhouette sombre s’approchait d’eux de plus en plus nettement.

3.4  Refined processses

Processes which contain a supplementary parameter, granularity, which repeats
itself during progression, are defined as refined processes.
Machali do przejeżdżających rękami, krzyczeli coś.
Ils agitaient les mains en direction des voyageurs, criaient quelque chose.
They were waving at passing travellers and were shouting something.

The semantic features of the granules in certain refined processes can be quantita-
tive in nature. Because of the variation in the size of these features, it is possible
to subdivide these processes into three phases14 which make up their “life cycle”
(Figure 4).

MOMENTS
enter exit
< run > <e
initial start > nd finish terminal
gin >
< be
< before > STAGES < after >

Figure 4.  Diagram of the life cycle of one specific case of a refined process

More generally, the “begin” stage is thus the phase in which the specific size
is increasing (represented by a rising segment), the “run” – the phase where the

.  A diagram of this kind is used in numerous works on aspect, e.g. Sémon 1986; Barentsen
1995.
 André Włodarczyk

specific size seems to reach equilibrium (represented by a horizontal segment)

and the end stage – the phase where the specific size decreases to the end point
of the process (represented by a falling segment). For example, a situation such as
“the door is opening” could be analysed in this way in three phases: (a) the initial
speed of the progression granule being equal to zero, in the course of the “begin”
stage this speed increases, (b) in the “run” stage it is stabilised, then (c) it decreases
during the “end” stage.

4.  Representations and formalisation

The primary source of inspiration for defining the internal stages of a s­ emantic
situation was for us algorithmic in nature. In the course of working out the
order of situations, however, we thought it useful to consult theories of modal
logic: (1) of action (von Wright G.H. 1967), (2) of the structure of events
(­Winskel G. 1982 & 1983; Kowalski R.A. & Sergot M.J. 1986) and (3) of situa-
tions (­McCarthy J. & Hayes P.J. 1969). It was above all the relationship of these
theories to computer science which seemed to us to be able to contribute to
the development of theories of meaning in linguistics. This continuing dialogue
between logic and computer science is, moreover, equally profitable to logic
itself. As far as the contribution of logic and computer science to linguistics is
concerned, it is barely necessary to point out that generative linguistics owes its
appearance to the development of computer science techniques (in particular,
to the intensive research carried out on formal languages, including “Chomsky’s
hierarchy of languages”), and that since then, a great deal of research on natural
language processing has been inspired either directly by logic or by theories in
the area of computer science on the treatment of symbols.
In order to construct our concept of situation structure, however, we relied
above all on sequence structure as defined by Hoare C.A.R. (1969). It was no doubt
this same concept which inspired Moens M. (1987), and then Moens and ­Steedman
(1988), to propose the “tripartite structure of events” called “nucleus”. This nucleus
was then adopted by theoreticians of discourse representation: Kamp and Reyle
(1993), and other specialists of modes of action (Aktionsarten): B ­ lackburn et al.
(1993) and Gagnon and Lapalme (1995).
As a result of the intensive use in computer science of both (1) Petri nets and
(2) the Event Structures15 of Winskel G. (1982 and 1983), and because these two

.  The semantics of event structures was formulated by Winskel G. for the CCS – Calculus
of Communicating Systems of Milner R.A. (1980).
 Frames of semantic situations 

representations are useful for describing parallel structures,16 we can compare the
results of our research on situation with the research in this area prompted by
Petri nets and event structures. Petri nets were first applied to linguistic analysis
by Mazurkiewicz A. (1986, 2000). As we will see, however, our use of the Petri
model differs from that of Mazurkiewicz A. (1986) in that we do not deal with
inter-situational relationships (i.e. involving multiple situations), and from that of
Mazurkiewicz A. (2000) in that we identify two levels of analysis: micro-structural
(lower) and macro-structural (higher). In order to present macro-structural con-
cepts we will also use the formalism of “event structures”. Nevertheless, the latter
will be augmented in the following ways: (a) by adding the time component17 and
(b) by completing the structure of “events” (“semantic situations” in our terminol-
ogy) with the “terminal moment” and thus, with the “after” stage.

4.1  Semantic situations and Petri nets

There are at least three different ways of introducing the Petri net system (set-
theoretical, graphic and algebraic). We are using the graphic method because of its
visual quality. For our current needs we will limit ourselves to a shortened version
of Petri nets, known as “ordinary nets”, which are made up of the pair (place, tran-
sition) and an set of links running from places to transitions and from transitions
to places. Places are represented by circles and transitions by squares. A study of
the different interpretations of Petri’s axioms (Banaszak Z. et al. 1993; Suraj Z &
Szpyrka M. 1999; Grafcet 1979) shows that places are often interpreted not only
as “states” but also as “positions”, “conditions”, “local states” or “stages” and transi-
tions as “events”, “operations” or “actions”.
By interpreting places as “states” and transitions as “events”, Mazurkiewicz A.
(1986) showed a direct correspondence between situations and Petri primi-
tives, based on the idea that all semantic situations could be reduced to just two
­categories (states and events). If we consider the situation model used from 1986
to  2000, ­following Mazurkiewicz A, by Koseka-Toszewa V. (1986) and others
(e.g. ­Laskowski R. 1986 or Bojar B. 1986), we first of all see that this model, based
upon the classic Petri nets computation (i.e. without any enrichment of a temporal
or stochastic kind for example), does not distinguish between (1) the micro-struc-
tural level (lower subconceptual level) and (2) the macro-structural level (higher
conceptual level). In a Petri net, a place is always followed or preceded by a tran-
sition and a transition by a place, in other words two places or two transitions

.  We consider that interpretations are parallel processes (Włodarczyk A. 2003).

.  Winkowski J. (1992).
 André Włodarczyk

never follow one another. This formal constraint imposes the order of succession
(precedence) at the lower level which, taken at the higher level, led Mazurkiewicz
A. (1986, p. 10) to consider that there might be a natural order characterising the
relationship between places (interpreted as “states”) and transitions (interpreted
as “events”).
In our opinion, it is above all the assimilation of the primary concepts of Petri
nets with semantic situations (expressed by languages) which leads Mazurkiewicz
to propose a relationship of succession (precedence) at the higher level (though
removed at the lower level) and to conclude that this generates order (this rela-
tionship would be above all transitive and asymmetric) at the higher level also. To
realise that this is not in fact the case, one only has to remember that the relation-
ships (of succession or even of concurrence) between semantic situations (in daily
life) in fact depend upon expression or are simply unforeseeable.18
A “transition” in Petri nets is not in itself a dynamic situation. In order to
make transitions dynamic, computer scientists proposed extended Petri nets
(Murata 1989, 1991). In addition, and by this time through the study of aspect in
languages, researchers at Berkeley19 proposed introducing more enrichments into
the initial formalism: “our extensions to the basic Petri Net formalism include typed
arcs, hierarchical control, durative transitions, parametrization, typed (individual)
tokens and stochasticity”.
Independently of this, Mazurkiewicz A. (2000) also proposed the extension
of Petri nets by introducing the concept of “progression” (represented by a tri-
angle) which allowed him to time transitions. Hence the need to identify the fol-
lowing four kinds of progression: (1) open progression [state1 → progression →
state2], (2) closed progression [event1 → progression → event2], (3) progression
closed from behind [event → progression → state], (4) progression closed in front
[state → progression → event].20 Mazurkiewicz’s proposal has, however, profound
consequences for the very structure of Petri nets, in particular the need to intro-
duce a third graph, that of triangles. Thus extended, Petri nets would no longer
be systems of transitions and places (bipartite graphs),21 but of systems of transi-
tions, places and progressions (tripartite graphs). Unless the progression defined
by Mazurkiewicz A. (2000) is a “macro-stage” (cf. Grafcet) and consequently has

.  This observation has already been made by Bojar B. (1986, p. 85)
.  Chang N. (1997), Narayanan S. (1997), Chang N., Gildea D. and Narayanan S. (1998)
.  N.B. [state → progression → event] ≡ [event ← progression ← state].
.  Following the explanation given by Sowa J. (2000), Petri Nets are a fusion of Flow Chart
and State Machine structures, each of which can be represented by a graph. In particular, it is
the way in which this fusion was carried out which makes us think of bipartite graphs.
 Frames of semantic situations 

a semantic interpretation based on the idea of the continuity of operations which

define change.
Since we are dealing with languages (including natural languages) the seman-
tic content of which has to do with the “expression” of situations and not their
“execution”, each kind of situation must be modelled (at the higher level) by using
formulae which make it discrete (lower level). The distinction between the macro-
scopic (higher) and microscopic (lower) levels is capital when one has to compile
languages. Thus, if we wanted to represent our situations model, the model pro-
posed by Mazurkiewicz in 1986 (which is based upon a minimum net made up of
places and transitions) would not be sufficient, because our model is in fact mac-
roscopic. We would therefore need to define a certain number of macro-objects
which would be the elementary units of our net. This would entail distinguishing
between untimed transitions (t1, t2,…tn) and timed ones (m1, m2,…mn).22 Timed
transitions, called moments, depend on the timing system, a fact which makes
them dynamic.

STATE t1 t2

EVENT m1 m2

ORDINARY PROCESS m1 m2

REFINED PROCESS m1 m2

Fig. 5.  Four kinds of situation represented by an extended Petri formalism

In Figure 5, state is represented as a “place” between two transitions (untimed)

and event as a “place” between two timed transitions. The two kinds of process
(ordinary and refined) are characterised by progression represented by a triangle,
which here is intended to be a synthesis of the four kinds of progression defined by

.  “… any timed Petri net can be translated by a t-timed net. Formally, a duration of sensiti-
zation di = [di min; di max] is associated with each ti transition in the net”.“In the case of t-timed
nets, the interval associated with the transitions characterises their duration. It is the time
during which the place tokens on entry are no longer present (they are reserved) but during
which the tokens produced are not yet visible in the place of exit” (Pradin-Chézalviel B. &
Valette R. – 2000).
 André Włodarczyk

Mazurkiewicz A. (2000). In the refined process, granularity is represented by the

lines subdividing the progression triangle.
Even the representations that we have just proposed, however, are too sche-
matic to take account of the situations model that we described informally in
­Sections 1–3 above. Figure 6 represents in an intuitive way a complex situation23
in which the arrows correspond to paths (in time), some of which are parallel.
These parallels reflect the idea that situations contain roles which, once played, can
develop in times that are relative to one another, and all of this is contained within
the internal time frame of each situation.

S0: John bought a car (from Peter) for Mary.

S1a: John gives money. S1a

S1b: Peter receives money. S1b
S2a: Peter gives his car. S2a
S2b: John receives the car. S2b
S3a: John offers a car. S3a
S3b: Mary accepts the car. S3b
S4a: The car is property of Mary. S4a
S4b: Mary is happy. S4b

Figure 6.  The distribution of elementary situations making up a complex situation

The model in Figure 6 compared to that in Figure 1 simply adds parallels, with-
out changing the time model itself. The internal time of a situation can be simply
modelled by linear time. But this is not the case with time in inter-situational rela-
tionships. Human experience shows that to take account of these relationships we
would have to use the branching model24 (but not, for example, the circular model).

4.2  Semantic situations and “event and configuration structures”

Semantic situations can also be represented by the concepts of the sequential logic
of Hoare C.A.R. (1969) and of extended event structures.25 According to event

.  In the context of organic programming (GAEA system), complex situations are also
defined as being made up of primitive situations (Nakashima H. – 1998). cf. also Nakashima H.,
Ohsawa I. and Kinoshita Y. (1991) and Harada Y. and Nakashima H. (1995).
.  This important observation comes from Kayser D. (personal communication).
.  We call “extended event structures” the event structures of Winksel G. (1982 & 1983)
integrated into the event and configuration structures of Winkowski J. (1992).
 Frames of semantic situations 

structures, any action26 which takes place in time (timed action) is a quadruple
made up of:

–– the action itself

–– enabling time: t0
–– start time: t1
–– completion time: t2

The three defined times are ordered: t0 > t1 > t2. Taking inspiration from Hoare’s
logic of sequential processes, according to which all sequences are made up of a
prefix {ф} (expressing preconditions), an operation α and a suffix {ψ} (express-
ing postconditions or effect): {ф} α {ψ}, we add to the event structures a 4th time
(effect or resulting time: t3). Only those operations α whose preconditions {ф} are
true can give results prescribed by postconditions {ψ}.
In relation to “event structures”, in our model of semantic situations what we
call state is not an event structure because the predicate of the state expresses the
property of the objects or entities without the state being characterised by any
internal structure. What we have described as an event can be defined in the for-
malism of event structures as a situation in which the enabling time t0 coincides
with the start time t1, and the completion time t2 coincides with the effect time t3
(in short, t0 = t1 and t2 = t3). Thus an event can be defined as a particular case of
event structures which we would call critical situations. Ordinary processes, on the
other hand, can be described by event structures, on condition, however, that a
middle phase is added between the start and completion times.
Translated from French by Rosemary Masters

References

Banaszak, Zbigniew, Kuś, Janusz & Adamski, Marian. 1993. Sieci Petriego – Modelowanie, Ste-
rowanie i Synteza Systemów Dyskretnych, Zielona Góra: Wydawnictwo Wyższej Szkoły
Inżynierskiej.
Barentsen, Adrian. 1995. Trexstupenčataja model’ invarianta soveršennogo vida v russkom
jazyke. In Semantika i struktura slavjanskogo vida, Stanisław Karolak (ed.). Kraków:
Wydawnictwo Naukowe WSP.
Blackburn, Patrick; Gardent, Claire & de Rijke, Maarten. 1993. “Back and forth through time
and events.” Proceedings of the Ninth Amsterdam Colloquium, December 14–17, 1993.
Amsterdam: ILLC, University of Amsterdam, 161–75.

.  In this paragraph we use the terms actions and events in their accepted sense in the
“theory of event structures” and not as we define them in our typology of semantic situations.
 André Włodarczyk

Bojar, Bożenna. 1986. Werbalne wykładniki leksykalne relacji temporalnych w językach pols-
kim i bułgarskim. In Studia Gramatyczne bulgarsko-polskie, Tom I: Temporalność, 83–91.
Wrocław: Ossolineum.
Bulygina, Tatiana V. 1982.  K postrojenju tipologii predikatov v russkom jazyke. In Semantičeskie
tipy predikatov, O.N. Seliverstova (ed.). Moskva: Nauka.
Chang, Nancy. 1997. A Cognitive Approach to Aspectual Composition. Ms, UC Berkeley Com-
puter Science Division and International Computer Science Institute. Report ICSI TR
97–034. 〈http://www.icsi.berkeley.edu/NTL/papers/cog.pdf〉
Chang, Nancy, Gildea, Daniel & Narayanan, Srini. 1998. A dynamic model of aspectual com-
position.  Proceedings of the Twentieth Annual Meeting of the Cognitive Science Society
­COGSCI-98. Madison WI.
Dowty, David. 1979. Word Meaning and Montague Grammar: The Semantics of Verbs and Times
in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel.
François, Jacques & Denhière, Guy (eds). 1997. Sémantique linguistique et psychologie cognitive,
Grenoble: Presses de l’Université de Grenoble.
Gagnon, M. & G. Lapalme. 1995. “From Conceptual Time to Linguistic Time.” Computational
Linguistics 22: 91–127.
Grafcet, Collectif ADEPA. 1979. Le Grafcet, diagramme fonctionnel des automatismes séquentiels,
rapport de la commision ADEPA sur la normalisation du GRAFCET, Montrouge.
Harada, Yasunari & Nakashima, Hideyuki. 1995. Situated disambiguation with properly speci-
fied representation. In Semantic Amibuity and Underspecification, Kees van Deemter &
Stanley Peters (eds). Stanford CA: CSLI.
Hoare, Charles A.R. 1969. An axiomatic basis for computer programming. Comm. ACM 12:
576–583.
Kamp, Hans & Reyle, Uwe. 1993. Formal Logic and Discourse Representation Theory, Introduc-
tion to Modeltheoretic Semantics of Natural Language. Berlin: Springer Verlag. Series: Stud-
ies in Linguistics and Philosophy, Vol. 42.
Koseska-Toszewa, Violetta. 1986. O temporalności inaczej. In Studia Gramatyczne bułgarsko-
polskie, Tom I: Temporalność, 39–64.
Kowalski, Roger A. & Sergot, Marek J.A. 1986. Logic-based calculus of events. New Generation
Computing, Vol 4.
Laskowski, Roman. 1986. System temporalno-aspektowo-modalny języka polskiego a sieci
Petriego. In Studia Gramatyczne bułgarsko-polskie, Tom I: Temporalność, 23–37.
Laskowski, Roman. 1998. Uwagi o znaczeniu czasowników. In Gramatyka współczesnego języka
polskiego, Morfologia, wyd. 2 zmienione. T. 1, 152–157. Warszawa: P.W.N.
Maggiolo-Schettini, Andrea & Winkowski, Józef 1992. Towards an Algebra for Timed Behaviours
[Report 677. Institute of Computer Science of the Polish Academy of Sciences. January
1990]. Also Theoretical Computer Science 103: 335–36.
Mazurkiewicz, Antoni. 1986.  Zdarzenia i stany: Elementy temporalności (Events and states: Ele-
ments of temporality). In Studia Gramatyczne bułgarsko-polskie, Tom I: Temporalność, 7–21.
Mazurkiewicz, Antoni. 2000. Oral communication on progression at the meeting of Polish-
French Research Group “Etudes cognitives”. Paris: LACITO.
Mccarthy, John & Hayes, Patrick J. 1969. Some philosophical problems from the standpoint of
artificial intelligence. In Machine Intelligence 4, Bernard N. Meltzer & Donald Michie (eds).
Edinburgh: EUP.
Milner, Robin A. 1980. A Calculus of Communicating Systems. Berlin: Springer.
Moens, Marc. 1987. Tense, Aspect and Temporal Reference, Ph.D. dissertation, University of
Edinburgh.
 Frames of semantic situations 

Moens, Marc & Steedman Mark 1988. Temporal ontology and temporal reference. Journal of
Computational Linguistics 14(2): 15–28.
Murata, Tadao. 1989. Petri nets: Properties, analysis, and applications. In Proceedings IEEE-89,
Vol. 77, 541–576.
Murata, Tadao. 1991. Planning with Petri nets. In Information and Control.
Nakashima, Hideyuki. 1998. GAEA Version 2.3 Manual, Revision 011, Cooperative Architec-
ture Project Team, Electrotechnical Laboratory (ETL), Japan.
Nakashima, Hideyuki, Ohsawa I. & Kinoshita Y. 1991. Inference with mental situations,
TR-91-7: A technical report of Electronical Laboratory, Tsukuba.  〈http://ci.etl.go.jp/
­
kyocho/Papers〉
Narayanan, Srini. 1997. Talking the talk is like walking the walk: A computational model of ver-
bal aspect. In Proceedings of the Nineteenth Annual Meeting of the Cognitive Science Society
COGSCI-97. Stanford CA: Stanford University Press.
Pawlak, Zdzisław. 1981. Information systems – Theoretical foundations. Informations Systems
6: 205–218.
Pawlak, Zdzisław. 1982. Rough sets. International Journal of Information and Computer Science
11: 344–356.
Pawlak, Zdzisław. 1991. Rough Sets: Theoretical Aspects of Reasoning about Data. Dordrecht:
Kluwer.
Petri, Carl A. 1966. Kommunikation mit Automaten, Schriften des IIM Nr 2, Institut für
Instrumentelle Mathematik. Bonn 1962, English translation: Technical Report RADC-
TR-65-377, Grifftiths Air Force Base, New York, Vol. 1, Suppl. 1.
Piernikarski, Cezar. 1969. Typy opozycji aspektowych czasownika polskiego na tle słowiańskim,
Ossolineum, Wrocław.
Polkowski, Lech  & Skowron, Andrzej. 1999. Towards adaptive calculus of granules. In Comput-
ing with Words in Information /Intelligent Systems [Studies in Fuzziness and Soft Comput-
ing 30], Lofti A. Zadeh & J. Kacprzyk (eds), 201–228. Berlin: Physica-Verlag.
Pradin-Chézalviel, Brigitte & Valette, Robert. 2000. Accessibilité de marquage et logique linéaire
dans un réseau de Petri t-temporel, IUT A. Toulouse: Université Paul Sabatier.
Reiter, Raymond. 2001. Knowledge in Action – Logical Foundations for Specifying and Imple-
menting Dynamical Systems. Cambridge MA: The MIT Press.
Sémon, Jean-Paul. 1986. Postojat’ ou la perfectivité de congruence, définition et valeurs textu-
elles. Revue des Etudes Slaves, T. 58/4, Institut d'Etudes Slaves, Paris .
Sowa, John F. 2000. Knowledge Representation: Logical, Philosophical, and Computational Foun-
dations. Pacific Grove CA: Brooks Cole.
Suraj, Zbigniew & Szpyrka, Marcin. 1999. Sieci Petriego i PN-Tools (Narzędzia do konstrukcji i
analizy sieci Petriego), Rzeszów: Wydawnictwo Wyższej Szkoły Pedagogicznej.
Vendler, Zeno. 1967. Verbs and times. In Linguistics in Philosophy, Zeno Vendler, 97–121. Ithaca
NY: Cornell Univer­sity Press. (Revised version of Vendler Zeno. 1957. Verbs and times. The
Philosophical Review 66: 143–160).
Von Wright, Georg H. 1967. The logic of action: A sketch. In The Logic of Decision and Action,
Nicholas Rescher (ed.). Pittsburgh PA: University of Pittsburgh Press.
Winkowski, Józef 1992. Modelling timed behaviours with the aid of event and configuration
structures. ICS PAS Reports, N0 715, 34.
Winskel, Glynn. 1982. Event Structure Semantics for CCS and Related Languages. Springer LNCS
140: 561–576.
Winskel, Glynn. 1983. Event structure semantics for CCS and related languages. Aarhus Univer-
sity, DAIMI PB-159, April.
 André Włodarczyk

Winston, Morton E., Chaffin, Roger & Hermann, Douglas J. 1987. A taxonomy of part-whole
relation. Cognitive Science 11(4): 417–444. Amsterdam: Elsevier.
Włodarczyk, André. 2003. Linguistique dynamique: évolution du discours dans le temps. In
Etudes linguistiques romano-slaves offertes à Stanislaw KAROLAK, 497–510. Kraków:
Edukacja.
Zadeh, Lofti A. 1996. Fuzzy logic = computing with words. IEEE Transactions On Fuzzy Systems
4: 103–111.
Zadeh, Lofti A. 1997. Toward a theory of fuzzy information granulation and its certainty in
human reasoning and fuzzy logic. Fuzzy Sets and Systems 90: 111–127.
Grounding of the meta-informative
status of utterances

André Włodarczyk
Université Ch. de Gaulle (Lille 3) / Université Paris-Sorbonne, Centre de
linguistique théorique et appliquée (CELTA)

In order to represent any kind of linguistic relationship (syntactic, semantic

or pragmatic), linguists are keen to borrow the formal definition of predicate
from First Order Logic (FOL). We claim however that the etymological sense of
predicate (in its original conception made by Aristotle) reflects the pragmatic
nature of elementary linguistic declarative sentences. The ASMIC (Associative
Semantics and Meta-Informative Centering) theory makes it possible to restore
the original meaning of “speaking about” to the term predicate, introducing
the notion of information (more specifically, meta-information) into that of
predication. Indeed, it is more judicious to use the term information to refer to
the semantic content of linguistic utterances rather than to their pragmatic status
which should be defined as meta-information. Thus, the semantic validation
of linguistic utterances as true or false is indirect (it relies on their semantic
interpretation) and goes far beyond the meta-informative (bearing the old or new
status of information) formatting of the content they convey. Paradoxically, the
old or new status of information is primary whilst the true or false validation is
secondary, and not the other way round as might at first sight be thought. In the
MIC framework, several different kinds of grounding of the meta-informative
old or new status can be distinguished: (1) inter-utterance (speech act bound
concepts of anaphora and cataphora), (2) discourse (known vs. unknown) related
to the process of knowledge acquisition and (3) ontological representing the
speaker’s knowledge as stored in his/her long-term memory. The memorised
situations are either abstract (types of situations having such properties as
generic, general, habitual or potential) or concrete (instances of situations defined
as counterparts of types having such properties as specific, particular, occasional
or actual respectively). The first two opposed pairs of concepts (a) generic vs.
specific and (b) general vs. particular, concern (whole) situations, whereas the
other two oppositions, (c) potential vs. actual and (d) habitual vs. occasional
are participant-oriented. Besides its importance for the theory of discourse, the
grounding theory of the meta-informative status of utterances sheds new light on
the rather puzzling pragmatic usages of some important grammatical categories
of human languages such as noun determination (eg. articles and quantifiers) and
verb modification (tense and aspect).
 André Włodarczyk

1.  Introduction

Linguistic terminology and research methods are elaborated following the para-
digm shifts of social sciences, starting with the approaches known as normativist,
evolutionist, structuralist and mathematical but ending up with the computational
one. However, the recent computational shift did not bring the desirable system-
atic insight into the really hard problems that the research on natural languages
requires from linguists. For this reason we claim that by using computational
­metaphors and tools as well as by taking into account discoveries in the field of
neurology, reconstructions of many linguistic notions can gain much in coherence
and adequacy in the future. Consequently, the proposed approach postulates the
need for an interdisciplinary collaboration between linguists, logicians, psycholo-
gists, neurologists and computer scientists.
Human communication uses a basically aerial channel so that the repre-
sentations resulting from both sequential and parallel mental operations must
be converted to/from a series of signs, i.e.: they must be encoded/decoded in
sequences of sounds (these being modelled/recognized due to the phonemic
abstractions). Obviously, as linguistic units cannot be uttered other than by
­following a sequential order, the result of both sequential and parallel operations
on expressions is known as a syntactic construction. Undoubtedly, subject and
object actually exhibit the syntactic organization of linguistic expressions, but
this is far from being their primary role, hence syntactic functions of both sub-
ject and object are secondary in our theory. Let us recall that what – f­ollowing
the Prague School t­ radition – linguists usually refer to as “information” (cf. the
notion of “Information Structure”) we name meta-information, reserving the
term information to refer to the semantic content of linguistic utterances. In our
approach ­meta-information means literally “information about another informa-
tion” because it is mainly responsible for the constitutive structure of predication
(aboutness) and ultimately determines the sequential ordering of the information
flow in utterances.
When modelling semantic situations, it is convenient to apply elements of
dependency1 of some kind in order to render the non-linear nature of meaning.
In addition to this, it seems that a separate meta-informative layer/dimension is
necessary in order to achieve the ordering of non-linear, mental representations

.  Dependency is a (directed tree-like) relation first applied to syntax by S.D. Katsnelson
(1948) and by L. Tesnière (1953), both using the same term valence in order to represent the
relations between the units of an utterance.
 Grounding of the meta-informative status of utterances 

in the form of sequentially ordered phrases in utterances or texts (sequences of

utterances). To communicate information about a semantic (non linear) situation
which the speaker has in mind, s/he needs to select at least one item (expressible
by a phrase) which corresponds to his/her centre of attention (CA) and predicate
about it, i.e. to build an attention-driven phrase (ADP) and attach another phrase
which bears information about that information to which the ADP points. Thus
the speaker, following his attentional attitude, has to select from amongst frames,
participants and anchors of the semantic situation (cf. Associative Semantics – AS,
Chapters 1 and 2 in this volume). Obviously, the primary element which happens
to be selected according to the attentional attitude of the speaker corresponds to
the subject of the utterance. The utterance may also possibly contain one more
secondary CA representing what is expressed on the syntactic level by the object.
In the syntactic structure of an utterance as represented by a phrase structure tree,
the primary centre of attention is global whereas the secondary one is local. In
fact, at least as far as the SVO languages are concerned, the hierarchical distinc-
tion between global and local CAs comes directly from the phrase structure of
sentences as represented by a tree structure in which the upper (left) node NP cor-
responds to the subject, and the lower node NP corresponds to the object. Thus, in
the MIC theory, the syntactic structure of utterances as imposed by (verb) valence
schemata reflects essentially pragmatic content, assuming that semantic informa-
tion (except names and traces of relations) relies basically on the lexical meaning
of constituents.
Let us emphasise the fact that, in our view, although both constituency and
dependency relationships are completely different ways of representing syntactic
relations, they can be used together for representing the structure of utterances.
Suffice it to mention that while constituency is often mainly used in descriptions
of languages with a “restrictive word order” (RWO), dependency characterises
descriptions of “free word order” (FWO) languages.2 Nonetheless, both graphic
representations of sentences (often considered as alternative) as used in syntactic
theories match in fact two different layers/dimensions of utterances in the MIC
theory; constituency concerns the meta-informative layer/dimension whereas
dependency relationships might be used as a symbolic representation of the infor-
mative semantic layer/dimension (the one with “predicate-argument structures”
in Elementary Logic).

.  Among them Slavic languages for which Mel’čuk I. (1988, 2003) proposed a dependency
syntax. The standard syntactic structure theory proposed by Chomsky N. (1957) was based
on constituency.
 André Włodarczyk

2.  Information and meta-information

As we put it above, predication in its specifically linguistic sense, i.e. as it has been
defined within the MIC theoretical framework, consists of uttering an information
about another information which takes the form of the subject of an utterance.
The first theoreticians of the subject-predicate relationship in natural languages
were obviously Plato and Aristotle, and many grammarians and linguists followed
them, giving rise to the foundation of what is known today as “traditional syn-
tax” preceding the theoretical effort made within the Generative Grammar frame-
work. The difficult (almost impossible) mutual understanding between linguists
and logicians concerning “predication” comes from the fact that logicians defined
predication as a formal relationship between terms, withdrawing from the original
concept of predication viewed as aboutness, i.e. in the way that fits the purposes of
artificially built formal, as opposed to natural, languages. In logic, predication may
be used in two ways. Firstly, predication is understood as the assertion P(a) where
P is a predicate and a is an individual term to which it applies. Thus the assertion
“Peter is a student” is interpreted using the set-theoretical framework (where an
element a belongs to a set) as “the individual called Peter belongs to the set of
students”. Secondly, in predicate calculus, predication is considered as the logical
formula of a relation: P(a, b) where the terms a and b fill in the argument places of
the predicate P; this is the way logicians interpret “Peter bought a book”: buy(Peter,
book). None of these formulae, neither P(a) nor P(a, b), takes account of the ‘linear
order’ (more properly, sequential order of constituents) of linguistic predicative
expressions (utterances) because they reflect the semantic relational information
itself, leaving aside the quite easily observable fact that what is explicitly stated in
the linguistic form is not semantic information but another information about it.
Thus, as such, logical formulae can prove to be more or less convenient for the
formal notation of semantic content but not for grasping the meta-informative
structure of natural utterances.
Since the last quarter of the 20th century, the genuinely linguistic view on predi-
cation has been making a come-back in a more and more formalised way, especially
in the fields of logical pragmatics and computer science, because it became clear
that a purely syntactic framework, and even the so-called “­semantico-syntactic”
frameworks, were insufficient to account for natural language real utterances. The
syntactic and semantico-syntactic models designed on the basis of generative gram-
mar led to very complicated theories, often disconnected from linguistic intuition
and evidence. The most relevant objections to purely syntactic approaches came
not only from the semanticists but also from the logicians [Thomason R. H. – 1977]
who emphasised that only a formalised pragmatic theory (on the basis of Grice’s
pragmatics) might contribute to simplifying syntax and semantics.
 Grounding of the meta-informative status of utterances 

In the MIC theory, information is produced when relationships are estab-

lished between entities, while predication takes place when the speaker refers to
the status of what s/he utters about the selected chunk of information, i.e. when the
speaker produces linguistic expressions in which some distinguished segments are
highlighted as centres of attention. Therefore, although ­meta-information belongs
to the pragmatic dimension of language, it is not a mere supplement to the seman-
tic and syntactic structures of utterances (as the pragmatic level is often treated in
linguistic theories). We find it necessary to emphasise that, in our view, no utter-
ance can be formed in natural language without choosing a CA and assigning to it
a meta-informative Old or New status (see also Chapter 4 in this volume).
Let us also emphasise that the specificity of the MIC theory consists of tak-
ing into consideration not only syntagmatic relations – as is the case in linguistic
theories which concern exclusively syntax – but also paradigmatic relations. In this
respect, the MIC theory takes advantage of the Saussurian tradition which pointed
out that linguistic (sequentially structured) messages contain both syntagmatic and
paradigmatic relations. Taking into account the two axes (dimensions) of linguis-
tic oppositions which reflect the selection and combination of linguistic units, we
distinguish two domains of semantics and two domains of pragmatics (Table 1).

Table 1.  Properties of selection and combination in the process of meaning creation
Bi-axial nature of discourse Semantics Pragmatics

Information Meta-information
Selection Property Comparison Centering
(paradigmatic oppositions) a ∈ A (attribute space) (distinguish e)
Combination Situation Configuration Predication
(syntagmatic relations) r(x) (relation on x) (say p about e)

Semantics operates both on the paradigmatic axis in the form of property

comparison (entities can be opposed to other entities through the comparison of
their sets of features/attributes) and on the syntagmatic axis in the form of situ-
ations viewed as relationship configurations with individual participants (agents
and figures) and spatio-temporal anchors. It is worth mentioning at this point the
results obtained in the psychology of aphasia where for various reasons two gen-
eral classes of this mental disease have been discovered, giving rise to the notional
pair we borrowed here: “selection/combination” (Jakobson & Halle 1956) and
“property comparison/scene creation”3 (Wisniewski 1997). As for p ­ ragmatics,

.  In our terminology “scene creation” corresponds to “situation configuration”.

 André Włodarczyk

we claim that it has a similar bi-axial (two-fold) nature: namely, it consists of

(a) ­centering, i.e. selecting and highlighting something amongst other things and
(b) predicating, i.e. uttering about the chosen something selected as the subject
of the utterance.
The following schema (Figure 1) shows how ontology (mental universe),
semantics and pragmatics (discourse) are linked together according to the
ASMIC theory. Before we adopt the concept of world “ontology”, we need to
make the following distinction concerning (a) signs and real or fictitious worlds
on the one hand, and (b) mental representations of information and knowledge
on the other.

Participant Role Global/Local

or Location or Anchor Centre of attention

Ontology Semantics Pragmatics

e r c
Entity/Location inf Active/Passive m-inf Subject/Oject

inf(r) ≅ e m-inf(c) ≅ r

Figure 1.  Information (inf) and meta-information (m-inf) can be seen as a composi-
tion of functions (m-inf ⋅ inf) from discourse to ontology or reference (with semantics as a
­common term)

Figure 1 represents what we consider to be a composition of the following

functions: in order to transmit or express information natural languages use (a)
a meta-informative function m-inf: D → S (mapping the centres of attention c
onto the roles r of semantic situations) and (b) an informative function inf: S → U
from the semantic situations S to the (mental) universe U (mapping the roles r
of semantic situations onto the entities e taking part in these situations in some
possible world of the universe). Since we understand semantics as a function map-
ping from linguistic expressions onto representations of the possible world situa-
tions, we consider that it is necessary to relate linguistic meaning to ontology. In
this respect our approach differs drastically from that of structural linguists who
consider that linguistic science is not concerned with ontology or so-called “ency-
clopaedic knowledge”. Moreover, introducing the distinction between informa-
tion and meta-information makes it possible to explain the connection between
semantics and pragmatics within the limits of simple linguistic units (utterances)
and not exclusively within paragraphs and/or texts.
 Grounding of the meta-informative status of utterances 

3.  Oldness and newness revisited: The meta-informative

status of utterances

The introduction of meta-information into the theory of grammar leads to the

important distinction between the semantic and the pragmatic validation of utter-
ances. The pragmatic (in particular communicative) Old (viz. given) or New status of
utterances is assigned independently from their possible truth-propositional value.
From the logical point of view, the truth-propositional values are assigned only in
semantic (information) interpretation of utterances. There are no morphemes or
syntactic structures used to mark the opposition between true or false, and even neg-
ative statements can be true or false as well. On the other hand, we claim that every
linguistic declarative utterance must be assigned a m ­ eta-informative (either old or
new) status. It may happen that the meta-informative status is assigned to an utter-
ance either simultaneously or consecutively with respect to its ­truth-propositional
value in the process of communication exchange. In other words, the assignment
of a meta-informative status to an utterance may precede the semantic stage. More-
over, both the assignment of true/false values and the validation of old/new status
are necessary elements of linguistic predication. Consequently, the fragments of an
expression (phrases) and text parts (utterances, paragraphs, chapters) which cor-
respond to the centres of attention ((a) subjects and objects, (b) topics and focuses,
(c) general and particular themes) need to be taken into consideration when estab-
lishing argumentation strategies of various kinds. In saying this, we do not accept
the view that topicalisation and focalisation are specific to colloquial style.
As a matter of fact, in the pragmatic validation process, the meta-informative
status of utterances makes it possible to communicate any content regardless of
its uncertainty or whether it is true or false (Włodarczyk & Włodarczyk 2008).
We claim that the main difference between natural languages and logic languages
consists in making it possible to talk about everything, be it true or false, certain
or uncertain (rather true or rather false), suspended (simply not yet definitely clas-
sified as such in the speaker’s mind) or whatever else. In other words, speakers do
not need to have at their disposal true and only true knowledge about the world in
order to be able to communicate.
Many linguistic approaches emphasise the importance of the alternation of
old and new information status as essential for the study of discourse coherence,
e.g.: “Connected speech unfolds as an unbroken sequence of ‘messages’, in which the
speaker is alternating between elements of given and elements of new; these map
into the structures of the other grammatical units, most powerfully into those of the
clause.” (Halliday & Greaves 2008: 42). We totally agree with the statement that the
old/new alternation in the text structure is relevant for clause structure and believe
 André Włodarczyk

that this problem reveals the real nature of linguistic predication. We propose
therefore that more attention be paid, in linguistic studies, to old/new alternation
within the syntactic structure of utterances.
Nevertheless, since the notions of old and new have given rise to a number of
interpretations in the literature,4 it is crucial to emphasize the original contribu-
tion of the concept of meta-informative old/new status to linguistic studies of this
sort of problem. The old/new opposition is often seen as representing alternative
terms either for anaphoric/cataphoric or known/unknown. We consider that the
meta-informative status of an utterance, on condition that it is in accordance with
the verb schema (valency), depends on the pragmatic choice of the speaker and,
as it is shown below (Table 2), that it is motivated by communicative and cogni-
tive acts as well as by ontology. The reason for this is that oldness and newness are
neither values nor features of information, rather they concern the way information
is treated in communication. It is precisely this treatment that we call “assigning
a meta-­informative old/new status to information (or its chunks)”. Consequently,
we propose replacing the notion of ‘old/new information values’ with that of ‘meta-
informative old/new status’ of any information being a part of an utterance or a text.
Let us look at a few important consequences of the proposed theoretical
shift. In order to elucidate the discourse grounding of newness, let us recall, as an
example, that the life and work of a famous scientist, say Albert Einstein, belong
in our time to the shared knowledge of most people with high school education.
However, in a text Einstein’s name may be freely introduced by the speaker with
either an old or new meta-informative status. In the following utterance: It is Albert
Einstein who discovered the law of Relativity, the subject Albert Einstein is focalised
as a new chunk of information contrasting with what is treated as the old informa-
tion: “there is a person who discovered the law of Relativity (supposedly although
you do know that this law has been discovered, what you do not know is the name of
that physicist) and I am telling you: that person’s name is Albert Einstein”. The aim
of the following example is to show that text or discourse itself is not a sufficient
condition of the meta-informative old or new status. A noun phrase which has
already appeared in a text/discourse may well be treated as having the new status
in a subsequent part of the same text/discourse just because the entity it refers to is
supposed to enter into a new situation (becomes a participant of a new role). For
instance, within the general theme of a text on Chopin’s life, the name “Chopin” (or

.  For a review of the most frequently used meanings of the terms given and new, see [Prince,
E. – 1981]. She established that given has at least three different meanings: (1) predictability,
the speaker assumes that the hearer is able to predict the occurrence of an item within a sen-
tence (Halliday 1967), (2) saliency, assumed by the speaker to be present in the consciousness
of the hearer (Chafe1976) and (3) shared knowledge, known and accepted as true (Clark &
Haviland 1977).
 Grounding of the meta-informative status of utterances 

“Frederic”) may nevertheless occur with the meta-informative new status despite
the fact that this name has already been quoted many times and that Chopin is a
very famous composer, as in the following example: “It was Chopin whom Georges
Sand invited to the party.” This is an important point making it possible to explain
also why pronouns, although most often used as anaphoric or cataphoric devices
in a text, may be used to refer alternatively to old or new chunks of information:
the chapters concerning personal pronouns in this volume provide evidence for
the need to distinguish between the old/new status as ‘recoverable’ from the text
and as that which is grounded in the World and refinable by shared knowledge.
Obviously, although first and second person pronouns are unambiguously iden-
tified as the speaker and hearer, in every speech situation, they may be treated
as referring to either old or new meta-informative status of utterances. Similarly,
although the third person pronouns normally refer to the entities mentioned in
previous parts of the text, they may nonetheless appear in the text with either new
or old status. Many examples of such usage of personal pronouns can be found in
Latin, Greek or Russian utterances which are analysed in the present volume (see
chapters by Azam O., Pitavy J.-Ch. and Vedrenne P.). This suggests that in utter-
ances occurring in different contexts, personal pronouns may encode either old or
new meta-informative status: sometimes (a) as the subject or object of utterances
with totally new or old status and sometimes (b) as a focus (new information) con-
trasting with its background (the status of which is old) or a topic (old information)
contrasting with its comment (the status of which is new).

4.  Grounding of the meta-informative old/new status

The meta-informative status of a linguistic message (or a part of a message) is the

way speakers introduce information into their discourse, treating it as conveying
either new or old information. Besides the ontological entailments (S-situations or
P-situations see Tables 2 and 3), the old/new meta-informative status originates in
the discourse strategies (following from the pragmatic attitude) as established5 by
the speaker. Moreover, on the semantic (information) layer, the situation ­spoken
about in an utterance bears a cognitive status connected with the knowledge of the
speaker and what s/he considers as belonging to the knowledge of the hearer (see
the discussion of this problem in (Wooldridge 2000: 113). The grounding of the
meta-informative status can be thought of as an entailment relationship between

.  Note that such discourse strategies are partly fixed by syntactic rules of languages but
here we focus only on the range of possible choices the speakers have at their disposal when
introducing their own points of view.
 André Włodarczyk

Table 2.  Grounding of the meta-informative old or new status

Memory Type of Grounding of Grounding of

information old status new status
and knowledge

Immediate Communicative anaphoric cataphoric

(Adjacent
utterances)

Intermediate Cognitive known unknown

(Recently
acquired
information)

Permanent Ontological generic specific

general particular
(Mental Σ Π
potential actual
reference) habitual occasional

information (which resides in either intermediate or long term memory) and

meta-information (what is just said about information).
We distinguish between three kinds of grounding of meta-informative old and
new status as follows:

a. Immediate Communication grounding is adjacent utterances bound: the

situation spoken about is either connected, by anaphora, to the one which has
been mentioned before (adjacent prior) or, by cataphora, to the one which
was not yet mentioned (adjacent posterior). This kind of grounding is in fact
a particular case of the discourse communicative ones.
b. Cognitive grounding is speech/text bound: it is related to the process of knowl-
edge acquisition: known (already registered) or unknown (yet unregistered)
during the given instance (period) of communication. These operations con-
cern contents registered in intermediate term memory. When – ­during com-
munication – the speaker acquires knowledge about P-situations (instances)
and – whenever possible – s/he transforms this knowledge into that of types
(classifications, perhaps heterarchies), s/he can speak about such more
abstract situations as known, producing utterances with an old meta-infor-
mative status. Mutatis mutandis, when the speaker acquires knowledge about
S-situations (types) and s/he uses this knowledge in order to produce situa-
tion instances, s/he can speak about such individual situations as unknown,
producing utterances with a new meta-informative status.
 Grounding of the meta-informative status of utterances 

c. The ontological6 grounding depends on the speaker’s knowledge about situa-

tions stored in his/her long term memory. From the ontological point of view,
(a) S-situations (types) can be determined as being either generic, general,
habitual or potential (b) and P-situations (instances) can be determined as
being either specific, particular, occasional or actual. Obviously, because of the
contrastive (contrary) status introduced either by the topic or the focus in
an utterance, the ontological grounding of this utterance is determined by
the predication (the ‘subject’ and the ‘predicate’ global aboutness status). The
confusion frequently made in linguistic studies between old/new, anaphoric/
cataphoric and known/unknown distinctions can be explained by the above
mentioned grounding of the meta-informative status of attention-driven
phrases (ADP) or utterances.

Importantly, the idea of distinguishing between S-situations (types) and

P -situations (instances) can be seen as a generalisation of the distinction between
“generic actions” and “individual actions” (Von Wright G. H. 1963), the former
being performed repeatedly on different occasions, the latter concerning actions
that are performed on a concrete occasion and at a certain moment. In brain neu-
rology, the above distinction of S/P -situations is comparable to that of “semantic
memory” vs. “episodic memory” (cf. Stachowiak, two chapters of the 2nd part
of this volume), while in Description Logic (Baader et al. 2003), the general idea
which lies behind the distinction “TBox” (terminological box for describing axi-
oms) vs. “ABox” (assertive box for describing instances) seems to be close to our
ontological distinction. Finally, our distinction concerns the supposedly shared
knowledge of human communicating agents. Last but not least, the problem of the
old/new status of utterances boils down to pointing out the typicality/individual-
ity of ontological situations or the distinction between the Universal Types and
Existential Types in abstract data type theory.
Further analysis of the ontological knowledge allows us to distinguish between
two kinds of situation modalities: (A) generic > general vs. specific > particular and
(B) potential7 > habitual vs. actual > occasional (Table 3).

.  We understand ontology as the representation of the world in the speaker’s mind. The
ontology which is activated in the speaker’s mind and, at the same time, matches the speech/
text situation and its environment, is said to be grounded.
.  The term potential is understood here as the capacity of an entity to play a role in a given
situation (not as an equivalent of possible which in modal theory is opposed to necessary).
 André Włodarczyk

Table 3.  Ontological grounding of aboutness

Kinds of situations

Σ-situations P-situations
(situation types) (situation instances)

A generic > general specific > particular

B potential > habitual actual > occasional

Pairs in A differ from pairs in B by the fact that while the former concern
situations, the latter describe the entities (participants which enact roles). Saying
so, we combine the two views on ontology (namely, the situation-oriented view
of ­Wittgenstein and the entity-oriented view of Aristotle, see Wolniewicz 1969).
Indeed, ontological grounding is conditioned by the categorisation of the world.
It is well-known that Aristotle distinguished between substance and accidence,
giving rise to the definition of entity (object ontology). In modern ontology,
­Wittgenstein L. has shifted the Aristotelian view of ontology from entities to facts.
“Aristotle’s ontology is an ontology of substances, Wittgenstein’s ontology is an
ontology of facts. But (…) both ontologies lay down conditions which in view of our
vocabulary appear to be identical.” And farther: “…let us note in passing that their
symmetric character seems to be itself an additional manifestation of the parallelism
discussed.”.[Wolniewicz B. – 1982]

Nonetheless, for Ingarden (1974), entities are no more isolated beings/objects but
consist of interrelated elements and should be viewed as structures. In the case of
living beings viewed as structured objects, they are said to be able to occur in vari-
ous environments or situations to which they adapt and with which they interact.
Such living beings are known as agents and their interaction with the universe as
well as between themselves is called their behaviour (Ingarden 1974). Structures
built of elements which develop their behaviour in environments can be described
in terms of the theory of general systems. Consequently, in Associative Semantics
theory the universe contains objects (i.e.: animate and inanimate entities) which
are related to situation roles through their behaviours within situation frames or
environments. In addition to the above, entities as well as facts may each be seen
either as types (classes) or as instances (tokens).
On the other hand, due to the fact that the First Order Logic8 (FOL) has
been derived from reflections on natural languages, linguists often study jointly

.  The notions of types and instances (occurrences) of situations are sometimes compared
to the universally (∏) and existentially (∑) quantified arguments of logical predicates, though
in First Order Logic (FOL) the symbols are inverted (∏ denotes the universal quantifier
 Grounding of the meta-informative status of utterances 

two logical concept quantifiers (some/all) and descriptors (a/the) as being closely
related linguistic devices. But natural languages have the possibility of expressing
ontological features and relations using the markers of noun determination (quan-
tifiers, articles) as well as those of verb aspects and modalities.

5.  Information centering structure

In figure 2 below, the foundational concepts and their interdependencies as defined

in the MIC theoretical approach are visualised. Only the meta-informative notions
are represented (merely the abstract pair S/P represents the ontological knowl-
edge). It is easy to see, for example, that the Subject is global (i.e. refers to the global
centre of attention) and that its meta-informative status can be either old or new.
Observe also that neither the status validation of the Topic nor that of the Focus are
represented by the ontological knowledge concerning the S/P-situation.

New Old

Comment Background

Local Global
Π Σ

Focus Predicate Topic

Object Subject

Figure 2.  The formal concept analysis context9 of information centering in utterances

while ∑ denotes the existential one). We follow here the set-theoretical tradition in which the
symbols have been inverted.
.  The Formal Concept Analysis (FCA) Context (Wille R. 1982) makes it possible to visu-
alise, in lattice form, heterarchies of concepts representing information systems (binary tables
or bipartite graphs with objects, attributes and assignments).
 André Włodarczyk

We claim that from a general perspective, to be felicitous, utterances need

to contain prosodic and/or syntactic markers which introduce the information
they convey as bearing either old or new meta-informative status. Thus, the meta-­
informative (pragmatic) validation of utterances is an “added value” with respect
to its informative (semantic) validity as true or false. Nevertheless, the meta-level
old or new status of whole utterances (viz. predications) can be traced back to the
type/instance distinction which is characteristic of entities and situations. In this
sense, the ontological grounding of an old or new status can also license the truth or
falsity of the communicated information: in base utterances, the hearer can infer
either (a) from the old meta-informative status of predication that the underlying
situation is typical (i.e. is a S-situation) or (b) from the new meta-informa-
tive status of predication that the underlying situation is individual (i.e. is a
P-situation).

6.  Aboutness and agentivity

It is a well-known fact that the semantic agent often corresponds “quite n­ aturally” –
one would be tempted to say – to the subject of a sentence. This statement gives
rise to a quite understandable confusion which consists of taking every subject for
an agent because of the impression that the scope of agentivity can be extended as
many times as needed due to the rhetorical operation called metaphor. Neverthe-
less, although many linguists claim that this is not the case, bearing in mind that
agent and subject belong to two different kinds of notions, semantic and syntac-
tic, respectively, some linguists distinguish between “syntactic” (“formal”) subjects
and “notional” (“logical”) ones. We argue however that the subject as represent-
ing a global centre of attention and the object defined as a local centre of atten-
tion, in fact indicate that their functions are essentially pragmatic, namely they
point to some chunks of information thus playing a meta-informational role in
communication. In our view, semantic denotations (information) and pragmatic
selections (meta-information) are parallel (more specifically probably concurrent)
operations.
Both the centered information (which is spoken about) and the about-­
information are supposed to be always endowed with communicative old/new
status. The aboutness relating to (a) S-subjects and P-subjects, in base utterances,
and to (b) Topics and Focuses, in extended utterances, is expressed in the second
part of the utterance; i.e. as S-predicates and P-predicates, on the one hand, and as
comments and backgrounds, on the other. The aboutness of (general and particular)
themes are “summaries” of the units of texts/discourses which are bigger than an
utterance. Their relationship with the units they represent is syntagmatic as well as
paradigmatic.
 Grounding of the meta-informative status of utterances 

7.  Conclusion

To sum up, meta-informative old and new status are motivated by (a) the flow
of information in an inter-utterance communication (anaphora/cataphora), (b)
­concerns resulting from the discourse (known/unknown) and (c) beliefs pertaining
to ontological knowledge. Anaphora and cataphora are not simply syntactic devices.
They are operations driven by the centered information which is expressed either
within the limits of an utterance or in two adjacent utterances. Consequently, as such,
these statuses should not be confused with their grounding itself. In an older version
of the MIC theory, grounding types were not clearly distinguished. The anaphora/
cataphora pair was inserted into a dual list of totally ordered concepts without dis-
criminating between inter-utterance communicative, discourse (recently acquired
information) and ontological grounding. We did this in order to account for lin-
guistic morphemes10 expressing alternatively all these concepts. In fact, we were
not wrong to consider the above mentioned list of concepts as representing roughly
two “graduated scales of opposite values of old and new information” (Włodarczyk
& Włodarczyk 1998). Indeed, all these notional pairs can be perceived form a dual
perspective ranging over some strength criterion and its opposite.
We did not develop the operational strategy (internal mental representations
in the human brain) which underlie the information centering of utterances. We
think however that the three following operations seem to be quite useful for far-
ther inter-disciplinary approaches.

1. Attention – one of the psychological preconditions of predication (declara-

tive speech act); its role is to determine which information chunk(s) will be
selected by the speaker as his/her main (possibly also one or more subsidiary)
subject, topic or focus of communication.
2. Aboutness – meta-informative (often complex and stratified) set of con-
tents (logical formulae) which can be assigned to the centered chunk(s) of
information.
3. Mapping of the contents of information together with the assigned communi-
cative status onto the patterns of utterance schemata (valence templates).

Nevertheless, the notion of attention-centered concepts (subject, object, topic,

focus, theme etc.) cannot be used simply as a substitute for belief-like modalities.
In ­language, what matters in fact is that some utterance segments (phrases) either
originate in centres of attention (in synthesis mood) or are supposed to create
them (in analysis mood).

.  Japanese noun particles (Włodarczyk A. 1980, 1996, 1998) and Polish and Russian verb
aspect (Włodarczyk H. 1994, 1997).
 André Włodarczyk

Acknowledgment

The author owes many thanks to Radosław Katarzyniak (professor at Wrocław

­University of Technology), Jan Pekelder (professor at University Paris-Sorbonne  –
CELTA) and Franz Stachowiak (professor at Charles Sturt University Albury,
­Australia) for their insightful remarks and, especially, to Rosemary Masters (Ph.D.)
for checking the English in this paper.

References

Baader, Franz, Calvanese, Diego, McGuinness, Deborah L., Nardi, Daniele & Patel-Schneider,
Peter, F. (eds). 2003. The Description Logic Handbook: Theory, Implementation, and Applica-
tions. Cambridge: CUP.
Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of
view. In Subject and Topic, Charles Li (ed.). New York NY: Academic Press.
Chomsky, Noam. 1957[2002]. Syntactic Structures. Berlin: Mouton de Gruyter.
Clark, Herbert H. & Haviland, Susan E. 1977. Comprehension and the given-new contract. In
Discourse Production and Comprehension, Roy O. Freedle (eds). Hillsdale NJ: Lawrence
Erlbaum Associates.
Halliday, Michael A.K. 1967. Notes on Transitivity and Theme in English (Part 2). Journal of
Linguistics 3, p. 206.
Halliday, Michael A.K. & Greaves, William S. 2008. Intonation in the Grammar of English.
­London: Equinox.
Ingarden, Roman. 1974. Der Streit um die Existenz der Welt, Vol. 3: Über die kausale Struktur der
realen Welt. Tübingen: Niemeyer.
Jakobson, Roman & Halle, Morris. 1956. Two aspects of language and two types of aphasic
­disturbances. In Fundamentals of Language, Roman Jakobson. The Hague: Mouton.
Katsnelson S. D. 1948. O grammatičeskoj kategorii. Vestnik LGU 1948 N°2. (Second edition
1986, in Obscee i tipologičeskoe jazykoznanie. Leningrad).
Mel’čuk, Igor. 1988, Dependency Syntax: Theory and Practice. Albany NY: The SUNY Press.
Mel’čuk, Igor. 2003. Levels of dependency in linguistic description: Concepts and problems.
Dependency and Valency. An International Handbook of Contemporary Research, vol.
1, Vilmos Agel, Ludwig M. Eichinger, Hans Werner Eroms, Peter Hellwig, Hans Jurgen
­Herringer & Henning Lobin (eds), 188–229. Berlin: Walter de Gruyter.
Prince, Ellen F. 1981. Toward a taxonomy of given-new information. In Radical Pragmatics
­223–255. New York NY: Academic Press.
Tesnière, Lucien. 1953. Esquisse d'une syntaxe structurale. Paris: Klincksieck.
Thomason, R.H. 1977. Where pragmatics fits in. In Proceedings of the Texas Conference on
­Performatives, Presuppositions and Implicatures, Andy Rogers, Robert Eugene Wall & John
Peter Murphy (eds), 161–166. Washington DC: Center for Applied Linguistics.
Von Wright, Georg H. 1963. The Logic of Preference. Edinburgh: EUP.
Wille, Rudolf. 1982. Restructuring lattice theory: An approach based on hierarchies of con-
cepts. In Ordered Sets, Ivan Rival (ed.), 445–470. Dordrecht: Reidel (Reprint in: ICFCA
‘09: Proceedings of the 7th International Conference on Formal Concept Analysis, Berlin,
Heidelberg, 2009, 314).
 Grounding of the meta-informative status of utterances 

Wisniewski Edward J. 1997. Understanding novel noun phrases. In Proceedings of IWHIT 1997,
The University of Aizu, 97–101.
Włodarczyk, André. 1980. 主題から主語へ、そして主語から主題へ　–　ハとガ (Shudai kara
shugo e, soshite shugo kara shudai e – wa to ga) (From Topic to Subject and from Subject
to Topic – wa and ga). In 言語月刊 (Language monthly), Vol. 9, No 8/80, 大修館 Tokyo:
Taishûkan.
Włodarczyk, André. 1996. La thématisation en japonais et l’article français. Studia Kognity-
wne: Sémantique et Linguistique Contrastive, Warszawa: Editions Slawistyczny Ośrodek
Wydawniczy.
Włodarczyk, André.  2003. Linguistique dynamique: Évolution du discours dans le temps. In
Etudes linguistiques romano-slaves offertes à Stanisław Karolak, 497–510. Kraków: Oficyna
Wydawnicza Edukacja.
Włodarczyk, André. 2008. Roles and anchors of semantic situations. Etudes cognitives/Studia
kognitywne VIII, 53–70. Warszawa: SOW.
Włodarczyk, André & Włodarczyk, Hélène. 1998. Graded Informative content of linguistic
messages. In Proceedings of the 16th International Congress of Linguists. Oxford: Pergamon.
(CD-ROM: Paper N0 0431).
Włodarczyk, André & Włodarczyk, Hélène.  2006a. Focus in the meta-informative centering
theory. In La Focalisation dans les langues. André Włodarczyk & Hélène Włodarczyk (eds).
Paris: Coll. Sémantiques, L’Harmattan.
Włodarczyk, André & Włodarczyk, Hélène. 2006b. Subject in the meta-informative centering
theory. Études cognitives/Studia kognitywne 7. Warszawa: SOW.
Włodarczyk, André & Włodarczyk, Hélène. 2008. The pragmatic validation of utterances.
Études Cognitives/Studia Kognitywne 8. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2012. Subjecthood and topicality are both prag-
matic issues. An International Workshop on Linguistics of BA and the 11th Korea-Japan
Workshop on Linguistics and Language Processing. December 10–11, 2011. Tokyo: Waseda
University.
Włodarczyk, Hélène. 1994. L’Aspect verbal slave et les domaines du donné et du nouveau. In
Etudes cognitives, Vol.1: Sémantique des catégories d’aspect et de temps, 113–130. Warszawa:
Polska Akademia Nauk, Instytut Slawistyki, Slawistyczny Ośrodek Wydawniczy.
Włodarczyk, Hélène. 1997. L’Aspect verbal dans le contexte en polonais et en russe. Paris: Institut
d’Etudes Slaves.
Wolniewicz Bogusław. 1969. A parallelism between Wittgensteinian and Aristotelian ontolo-
gies. In Boston Studies in the Philosophy of Science, Vol. IV, Robert S. Cohen & Marx W.
Wartofsky (eds), 208–210. Dordrecht: Reidel.
Wolniewicz, Bogusław. 1982, A formal ontology of situations. Studia Logica 41: 381–413.
Wooldridge, Michael J. 2000. Reasoning About Rational Agents. Cambridge MA: The MIT Press.
 Attention-centered information in language

Hélène Włodarczyk
Université Paris-Sorbonne, Centre de linguistique théorique
et appliquée (CELTA)

In this chapter we outline the concept of attention centering in the MIC

theory, and explain how it is linked with the old or new meta-informative
status. We distinguish two meta-informative levels giving rise to two different
meta-informative types of utterances: base and extended utterances (in the
latter, the speaker establishes a contrast between old and new meta-informative
status). We define subject and object, topic and focus as global and local
attention-driven phrases (ADP) respectively, expressing attention-centered
information in both kinds of utterances: base and extended ones. We raise
the problem of the truly predicative distinction between subject and object,
as well as arguing that subjects, be they used in an active or passive voice
utterance, should be defined as one and the same concept, namely, as the
attention-driven phrase pointing at the attention-centered information. We
examine the possibility of leaving the subject implicit and distinguish this
problem from the possibility of using anonymous (impersonal) subjects (AnS).
We tackle the difference between subjects in base utterances and topicalised
objects in extended utterances. We claim that it is necessary today to revise, in
the light of meta-informative centering, the typological distinction between
subject-prominent and topic-prominent languages. The last paragraph briefly
raises the matter of word-order in connection with attention-centering. In the
so-called rigid word-order languages, word-order concerns mainly the first
meta-informative level (that of the subject and object of base utterances) whereas
in the so-called free word-order languages, word-order is used essentially (but not
exclusively) as a marker of the second meta-informative level (that of extended
utterances with topic and/or focus).

1.  Introduction

In current linguistic research, argument structure and verb valency theories are
attempts to capture the mappings between noun phases in the syntactic plane
and the roles enacted by the participants in the semantic plane of discourse. In
the MIC theory this correspondence is recognised as a more complex phenom-
enon, the solution of which cuts across another dimension of language, namely,
 Hélène Włodarczyk

the ­pragmatic one where attention-centered mental representations concern the

frames, roles and anchors of semantic situations. Thus, subjects and objects are
defined not only as positions in the syntactic structure of the utterance but as
expressions motivated by the pragmatic choice of the speaker, respectively the
global and local centres of attention of utterances.
The relation between topic/comment structure and subject/predicate struc-
ture is one of the main theoretical problems tackled by linguists who carry out
advanced research on “information structure” (Lambrecht 1994). In the MIC
theory, we propose to treat this problem on two levels: (a) the components of the
subject-predicate structure belong to the first meta-informative level and (b) the
topic/comment and the focus/background structures are components of the sec-
ond meta-informative level. Since the MIC theory deals with the core of the basic
syntactic structure (NP+VP) of utterances as determined by meta-information in
the “speak-about” sense, it puts much more emphasis than do syntax-based lin-
guistic theories on the pragmatic dimension of language. Indeed, the concept of
the subject of an utterance as an entity spoken about had its origin in the classical
view of sentence structure in Greek philosophy and, more recently, in its revival in
the 19th century (Paul H. 1880).
However, in the MIC theory, the subject, defined as the global centre of atten-
tion of basic utterances, and the topic, as the global centre of attention of extended
utterances, are quite original and innovative because both these concepts have
been based on cognitive capacities of the human mind, i.e. on the relation between
language and the role of attention underlying such cognitive human activities as
perception and reasoning.

2.  Centres of attention and attention-driven phrases

The function of attention in human cognition has been a major topic in experimen-
tal psychology since the very beginning of this science, and it is still a fundamental
topic of neuro-cognitive research (Bunting et al. 2008; Cowan 2008; Shelton et al.
2008; Stachowiak in this volume). Let us quote a statement from a psychological
study about attention: “… the experience of attending to an object becomes an
experience of being aware of that object when it is conjoined with attending to a
representation of the self. “(LaBerge 2003: 291). In linguistic studies this statement
from neuro-cognition can be related to the meta-informative level which neces-
sarily accompanies any utterance. This level is represented in the MIC theory by
the two properly meta-linguistic (“meta-informative”) “predicates” (in the sense
of relations): speak with (x, y) and speak about (x, z) which are necessarily added
to the predicate representing the situation spoken about, this situation a­ ppearing
 Attention-centered information in language 

as the z variable in the predicate speak about (x, z). The “representation of the
self ” mentioned by psychologists has its counterpart in the representation of the
speaker (and his point of view) in linguistic utterances.
In linguistics, we need a general concept to capture what is common to the
concepts of subject, object, topic and focus. In the MIC theory, these concepts are
put together under the heading attention-driven phrases (ADP) and are considered
as segments of linguistic utterances (on the level of syntax) pointing at “centres of
attention” (CA) defined as psychological phenomena (on the level of cognition).
Paradoxically, as we shall see below, without this common concept of ADP, the
specificity of each of them as subject, object, topic and focus is difficult to grasp.
As applied to language studies, the concept of attention centering was first
introduced within the framework of computational linguistics (known as center-
ing theory, Grosz et al. – 1986, 1995) with the aim of automatically discovering
coherent information from text documents. For this purpose, the terms forward-
and backward-looking centres have been coined, taking it for granted that informa-
tion is kept coherent in text, and more specifically in adjacent pairs of utterances.
In this theory, the concept of centre of attention concerns only the text/discourse
level and, even more importantly, attention focusing is not considered as one of
the most important operations underlying the structure of a single utterance. In
other words, the computational Centering Theory did not encompass definitions
of the subject and object as attention-driven phrases governed by the centres of
attention on the first meta-informative level (where no contrast is proposed by
the speaker between the old and new information),1 neither were the topic and
focus defined as dually opposed concepts, respectively the topic (as governed by
an old meta-informative status contrasting with the ‘new” one of the comment)
and the focus (as governed by a new meta-informative status contrasting with the
old one of the background). It should be stressed that the meta-informative cen-
tering (MIC) theory is not an alternative to the computational one. Rather, it is
a complementary approach; in the MIC approach, the anaphoric and cataphoric
motivations of the meta-informative old and new status of the utterance during the
communication act (cf. Chapter 3 in this volume), match perfectly the concepts of
backward and forward looking centres in the American centering theory, which can
therefore be integrated into the MIC framework.
Although the American theory precedes the European one in time, it is
important to state that these theories have separate origins. The origins of the MIC
theory go back to 1999 when we first used the concept of centre of interest (“­centre

.  The definition of old and new meta-informative status is introduced in Chapter 3 in this
volume.
 Hélène Włodarczyk

d’intérêt”) in two different papers (Włodarczyk A. 1999 and Włodarczyk H. 1999).

Since that pioneering time, the term “centre of interest” has been replaced by
that of “centre of attention” under the influence of the Grosz-Sidner centering-­
theoretical framework. Let us now define two foundational concepts belonging to
the MIC theoretical approach:
Centre of attention (CA): a mentally distinguished element of a semantic situation
or a whole semantic situation is said to be centered, i.e. treated by the speaker as
his/her centre of attention.
Attention-driven phrase (ADP): a linguistic expression (phrase, part of a text/dis-
course) which points to the information chunk (conceptual representation) cor-
responding to the centre of attention.
Research in neuro-cognitive sciences provides psychological and neurological
grounds for the importance of attention in human cognition. A group of research-
ers under the direction of Russell Tomlin (1995, 1997) investigated attention in the
relationship between visual perception and linguistic expression. Tomlin’s follow-
ers, Myachykov et al. (2009) remind us of what linguists should know about atten-
tion as defined in psychology: “Everyone knows what attention is. It is the taking
possession of the mind in clear and vivid form of one out of what seem several
simultaneous objects or trains of thought”. (James 1890: p. 402). They quote also a
modern author: “Attention defines the mental ability to select stimuli, responses,
memories, or thoughts that are behaviourally relevant, among the many others
that are behaviourally irrelevant” (Corbetta 1998: p. 831).
Psycholinguistic tests have been devised by Russell Tomlin (1997) to study the
relationship between perceptual attention and the subject of an utterance: “it is
likely that the most accessible (e.g. visually most salient) referent will be articulated
before other referents taking part in the event and that it will be assigned as the
most prominent grammatical constituent, for example, the Subject.” (Myachikov
et al. 2009). This research programme was named: “Attention and Grammar in lan-
guage production”. Native English speakers were shown a short animated film in
which one fish swallows the other. In the first part of the experiment the active fish
is cued, in the second one, the passive fish is cued. When asked to describe what
they see, speakers spontaneously choose the cued fish as the subject of the utter-
ance, be it active or passive. In the latter case, the verb is in the passive form. This
experiment provides evidence for the definition of subject as centre of attention,
and the author’s conclusion is that “an attentionally detected referent is coded as
the syntactic subject in English.” In this view, subject is thus considered “the most
prominent grammatical constituent” of the sentence (Myachikov et al. 2009). The
same “Fish film” was then shown to speakers of different languages, among them
Japanese and Russian. Results confirmed the attention motivation for ­choosing a
 Attention-centered information in language 

participant of a situation as the subject of the utterance, describing the situation

thus: “attentionally detected referents map onto syntactic subject” (Hayashi  A.,
Tomlin R. S. & Yokota T.). However, as was pointed out in (Myachykov et al. 2009),
some problems arise in the interpretation of the linguistic choice of speakers of so
called free word order languages (FWO) such as Russian. From the MIC theory
perspective we put forward the hypothesis that difficulties in interpreting tests
by Russian native speakers are caused by the lack of objective criteria in syntac-
tic theories making it possible to distinguish in FWO languages simple subjects
from focalised or topicalised subjects: in fact, in FWO languages, the difference
between simple subject and focalised or topicalised subject is mainly expressed
by suprasegmental markers (intonation, pauses, word order) and, in most syn-
tactic descriptions, these markers are not taken into account. Criticisms of the
fish film experiment are based on linguistic approaches which posit an indepen-
dent grammatical level of role assignment and sentence structuring independent
of its meta-informative structuring. In our view, syntax is merely a means to the
end consisting of communicating information about a situation represented in the
speaker’s mind.
In the MIC theory we consider that any linguistic utterance is concerned
with “attention”, not only those utterances that are the direct wording of visual
perception but also those referring to any mental representation. The importance
of attention is in fact put forward in a great deal of linguistic research in differ-
ent fields (Oakley 2009). It is worth comparing the concept of centre of attention
with the concept of salience (Chafe 1994) used in pragmatic theories of discourse.
According to Chafe (1994: 100), contextual salience concerns the distinguished
or highlighted position that a referent occupies in the speaker’s mind. In cogni-
tive linguistics (Langacker 1987, 1991; Talmy 2000), the so-called “attentional”
view asserts that the speaker chooses to express only that part of a situation or
that point of view which holds his attention. In the AS-MIC theory, similarly, we
claim that the important role of attention is supported by the specificity of the
semantic content of linguistic utterances, consisting in their partiality by contrast
with expressions of artificial formal languages which are explicit. Linguistic mes-
sages are always partial with respect to reference: when communicating some-
thing about a situation in the world, in order to be brief and relevant (Grice 1975),
speakers have to choose a point of view and communicate explicitly only this point
of view. The whole situation must be completed by the hearer on the basis of his/
her knowledge both of the speech-act situation and of the world in general. Infor-
mation which must be inferred or completed by cognitive processes is implicit
(context-dependent). It is well known that the context-dependency of speech is
fundamental for explaining the partiality of linguistic expressions. As linguists,
we study in the AS-MIC framework all grammatical and lexical devices (prosody,
 Hélène Włodarczyk

phonology, morphology, syntax) that take part in the centering of attention in

order to explain how information content is enriched by meta-informative point-
ers when formulating utterances. Not only linguistic markers must be taken into
account as a meta-informative device but also gestural behaviour (eye movements,
amongst others); in fact, utterances that are partial as regards their semantic con-
tent are in natural, spontaneous, communicative circumstances complemented by
all sorts of gestures (with cultural specificities). In remote written communication,
the complementation of partial information relies much on context and world
knowledge, the latter being also partly different from one culture to another.
We must point out that in natural language, partiality of information can also
be considered as the reason why canonical syntactic schemes themselves can easily
undergo ellipsis. We define ellipsis as the omission of old chunks of information
from a syntactic scheme. Only old chunks of information (which are recoverable
from the context or the shared knowledge) may be omitted.2 In Example (1) we
put in square brackets the part of the utterance which normally undergoes ellipsis
because of its old meta-informative status
(1) Where is Mary? – [Mary is] in the kitchen.

Moreover, in the MIC theory, we consider yet another cause of partiality, namely
in natural utterances, there is an implication between local and global ADPs; a
local ADP is dependent on a global one which may remain implicit, i.e. when the
object is present, the subject may remain implicit and when the focus is present,
the topic may be omitted. This implication has very important consequences both
on the level of base and extended utterances (see below).

3.  Attention-Driven Phrases (ADP) in base and extended utterances

Base and extended utterances are defined as communicative discourse units. Each
utterance can contain more than one centered segment: we distinguish between
global and local CAs and between base and extended utterances, the latter includ-
ing second level ADPs (topic and focus) added or superimposed on first level
ADPs (subject and objects). Utterances in which the subject and the predicate
have the same meta-informative status (old or new) are said to be base utterances:
we call this the first meta-informative level.

.  Two reasons for subject omission in Japanese are taken into account in the chapter here-
after by Nariyama Shigeko: (1) the morpho-syntactic ground: the Japanese verb has no per-
sonal category (no personal endings) (2) the meta-informative ground: the subject is omitted
when it is old.
 Attention-centered information in language 

Table 1.  Three layers of linguistic information (from Włodarczyk A. & H. 2006a)
informative layer typical semantic units: situation schemata
meta-informative layer linguistic units uttered in a context:
1st m-inf. level: base utterances
2nd m-inf. level: extended utterances
abstract cognitive layer arrangements of sets of utterances:
texts, monologues, dialogs, discourses etc.

In the MIC theoretical framework, the theme is not merely a synonym of topic
(Table 2). As a matter of fact, we define topic and theme by reference to the repre-
sentation layer of linguistic information to which they respectively belong. We use
the word “theme” as a term referring to texts or discourses as organised, linguisti-
cally coherent sets of utterances. Thus, making reference to the meta-informative
status of base and extended utterances respectively, it is possible to capture and
better explain the difference and, at the same time, the similarity which character-
ise the subject and the topic.

Table 2.  Pivots of discourse (from Włodarczyk A. & H. 2008a)

TYPE OF EXPRESSION CENTRES OF ATTENTION

Global Local

1.1. Basic utterance subject object

1.2. Extended utterance topic focus
2. text/dialog general theme particular theme

The second meta-informative level is an extension of the predication: it consists

of adding contrasting attention-centered phrases to a base utterance. Extended
utterances consist of two contrasting parts of opposed meta-informative status;
their centre of attention is an emphasized noun phrase contrasting with the rest
of the utterance: an old topic contrasting with a new comment or a new focus with
an old background. Some extended utterances entail both topic and focus. From
a meta-informative perspective, the speaker can highlight a centre of attention
either through focalisation or, on the contrary, through topicalisation. On the
syntactic level, an extended utterance may take the form of either a single inde-
pendent clause or a complex sentence with main and subordinate clauses, one of
which may be treated as topic or focus.3 In many languages, centres of attention

.  Topicalisation and focalisation of subordinate clauses in Polish have been investigated in
the MIC framework by Miladi L. 2006, 2009.
 Hélène Włodarczyk

of the second meta-informative level are often realized as left or right dislocated
constituents, i.e. they occur out of the clause boundaries on the left (in utterance
(2), three topicalised noun-phrases are separated from the clause by the dash ) or
on the right (in (3), the topic is placed at the end, after the comma).
Transport of the mails, transport of the human voice, transport of flickering
(2) 
pictures – in this century as in others our highest accomplishments still have
the single aim of bringing men together. (Antoine de Saint-Exupéry, Wind,
Sand and Stars)
(3) They bought a new car, Susan and Tom.

The problem of clause boundaries is closely related to what we call “extension of

the utterance”. This makes it possible to account for some types of focalisation, for
instance in German where a strictly syntactic view of the utterance makes it dif-
ficult to tackle the problem of noun phrases which are out of the syntactic clause
boundaries.4 It is important to emphasize that dislocation is only a syntactic device
used to mark utterance extension. In linguistic analyses, in order to keep separate
from one another morphological and syntactic markers on the one hand and their
functions on the other, one should be aware that many other markers can be used
to mark topicalisation or focalisation, the first and most universal one being into-
nation (Halliday & Greaves 2008; Morel & Danon-Boileau 1998).

3.1  Attention-Driven phrases of base utterances: Subject and object

The subject is defined as that nominal phrase of a base utterance which is centered
as the global centre of attention and functions as an attention-driven phrase. Let
us recall that due to the linear nature of linguistic messages conveying informa-
tion about non-linear mental situations, the speaker must necessarily identify
a participant of a situation, no matter what its role, word it as a noun phrase
and make it the subject of a sentence, what we call the global ADP pointing at
the main CA. Subject, in this sense, is understood independently of its semantic
content but in agreement with the classic linguistic tradition as “what is spoken
about”. Choosing one of the participants of the situation entails distinguishing it,
making it hierarchically more important than the others. This hierarchical supe-
riority in the structure of the utterance is generally represented by the dominant
place of the subject NP in the uppermost left node of a constituency tree, in which
object NPs are in lower nodes as daughters of VPs. On this basis, we proposed
a definition of subject, which is universal for nominative and ergative languages
(Włodarczyk & Włodarczyk 2006b): subject is the global attention-driven phrase

.  See the book chapter “Extensions de l’énoncé” in [Cotte et al. – 2004].
 Attention-centered information in language 

(of a base utterance) pointing at the main centre of attention (in the mental rep-
resentation of a situation).
Object is also defined as an attention-driven phrase referring to a centre of
attention but, as it is a local one in the constituent structure of the utterance, it is
secondary with respect to the subject. Thus, the object is dependent on the subject
and its occurrence in an utterance entails a subject. Talmy Givón, among contem-
porary linguists, also takes into account the importance of attention and, in this
regard, his approach is close to our point of view: “the subject and DO (direct
object) may be viewed as the grammaticalized primary and secondary topic of
the discourse at the time when the clause in which they appear is being processed.
(Givón 2001: 198)” In fact, what Givón calls the pragmatic function “topic” we call
more generally the “centre of attention”. However this author relates the necessity
of selecting a “primary and a secondary topic” exclusively to the cognitive limita-
tions imposed by attention in the human brain, but not explicitly to the sequential
(“linear”) nature of units in linguistic utterances.
In an utterance the subject of which has been dropped5 – because it has an “old”
meta-informative status and can be easily recovered from the context – the pres-
ence of the object allows us to infer the existence of the absent subject. However, the
possibility of allowing the subject be completely absent (without any prosodic, mor-
phemic or syntactic marker) or to represent it by a pronoun, is language-specific.
(4) Mary entered the store and bought a book.

In English utterance (4), no subject pronoun is necessary in front of the verb

bought. But in (5), when uttered after a pause as the continuation of (4), the 3rd
person pronoun (she referring back to Mary) becomes necessary in the English
version, whereas in the Polish translation (6) it is not (since the verb ending -ła
contains also the feature “female”).
(5) She wanted to read it during the journey.
(6) Pol.
Chiała ją
Ø pro 3pers V 3rd p sg fem past pro 3pers sg acc
she wanted it
czytać podczas podróży.


V inf prep. n gen.
read during journey

.  N.B.: We treat in two different ways the possibility of dropping an “old” status subject in
the third person, noun or pronoun, and the use of the zero form of a subject pronoun in the
first or second person in languages where verbs exhibit explicit personal endings making it
possible to identify the subject as the speaker or hearer (Włodarczyk, H. 2012).
 Hélène Włodarczyk

Polish utterance (6) may be uttered for instance as the continuation of (7) hereaf-
ter, which is a translation of English (4):
(7) Pol.
Mary weszła do sklepu
N nom fem sg Vpast 3pers sg fem prep n gen mas sg
Mary entered into the store

i
kupiła książkę.

Conj Vpast 3pers sg fem n acc fem sg
and bought a book

Both subject and object, as attention-driven phrases, are independent of the

semantic role of the participant they point to; subject does not necessarily point
to the active participant of the situation (“agent”), neither does the object refer
obligatorily to the passive participant (“patient”). However the matching between
centres of attention and semantic roles is immediate in linguistic communication
because utterance schemata (based on verb valence) provide the speakers and the
hearers with the sufficient amount of knowledge for assigning semantic roles to
both subject and object.6 Let us notice however that verb valence is a complex
mixture of linguistic functions that are syntactic (the place in the utterance and/
or case-form of the NPs which depend upon the verb valence), pragmatic (centres
of attention) as well as semantic (frames, roles and anchors). A native speaker or
hearer acquires at the same time the knowledge concerning both the meaning of
verbs and their valence, e.g. a person who knows the English verb to read also
knows that it requires an animate entity (an agent) able to read (as the participant
playing the active role) and an inanimate entity (a figure) consisting in a written
word or text as the participant in the passive role. As a consequence, although the
subject does not obligatorily point to the active role nor the object to the passive
one, we claim that there is a default relationship in nominative languages between
the subject and the active role and in ergative languages between the subject (in
the absolutive case) and the passive role (Włodarczyk & Włodarczyk 2006b).
The properties shared by subject and direct object are important; they were
mentioned by linguists (Jespersen 1924) long before we proposed the concepts

.  It is however sometimes possible to postpone the semantic interpretation of the subject.
In such cases, the speakers just communicate information about an entity which appears to
them as their global centre of attention. The subject appears first as a pseudo-active partici-
pant until a deeper semantic interpretation is possible, e.g.: the utterance “A car (pseudo-active
role) hit a pedestrian.” can be interpreted as “A driver (active role) hit a pedestrian with his car.”
(Włodarczyk & Włodarczyk 2012).
 Attention-centered information in language 

of centre of attention and attention-driven phrase. Several morpho-syntactic

properties of subject and direct object in various languages bring evidence
for the difference between these NPs and other NPs. In many languages (e.g.:
­English, French, German, Polish etc.) the object of the active voice may be
transformed into the subject of the passive voice. In Polish, both subject and
object require the genitive case when they are within the scope of negation or
quantification, see Examples (8) (9), (10) and (11) below.

(8) Pol.
Jest ojciec w domu.
V 3pers sg pres n mas sg nom prep n mas sg loc
is father at home
“Father is at home.”

(9) Pol.
Nie ma ojca w domu.
neg V 3pers sg pres n mas sg gen prep+n mas sg loc
not is father at home
“Father is not at home.”

(10) Pol.
Student kupił książkę.
n mas sg nom V 3pers sg past n fem sg acc
student bought book
“The student bought a/the book.”

(11) Pol.
Student nie kupił książki.
n mas sg nom neg V 3pers sg past n fem sg gen
student bought book
“The student did not buy a/the book.”

As a matter of fact, the same participant of the semantic situation can be alterna-
tively introduced into the discourse by the speaker either as the subject or as the
object of the utterance, depending on his/her viewpoint or, more precisely, on the
attention he/she pays to it. Traditional occidental grammarians were well aware
of the interchangeability of subject and object in the utterance. The passive voice
is one of the devices which natural languages use to realize such interchangeabil-
ity without changing the information content of the utterance,7 e.g. “The Greeks
defeated the Persians.” (active construction) vs “The Persians were defeated by the

.  Nor its truth validity (see Włodarczyk & Włodarczyk 2008b).
 Hélène Włodarczyk

Greeks.” (passive construction). An important consequence is that no linguistic

utterance conveys information objectively or in a neutral way; even when he/she
speaks of a third person, there is always some chunk of content which shows that
the speaker is present in his/her own utterance since the subject and the object
phrases are results of his/her (weaker or stronger) centres of attention organised
hierarchichally from his/her point of view.

3.2  Attention-Driven Phrases of extended utterances: Topic and focus

As was shown above, in order to analyse an ADP as the topic or the focus of an
utterance, it is necessary that the meta-informative status of the rest of the utter-
ance has an opposite (contrasting) status: the comment is always “new” and the
background is always “old”. The speaker introduces as the topic the constituent
s/he wishes the hearer (a) to pay attention to and (b) to consider as having the
“old” status. Thus, in the MIC theory, topic is defined as a prominent attention-
centered phrase with an “old” meta-informative status. It is only the comment part
of the utterance which introduces information with a “new” meta-informative
status. The comment itself may, in some cases, be further divided into two parts:
focus and background (but see below). What is introduced as a topic is what the
speaker intends to be “taken for granted”, presupposed to be known; therefore it
is an important argumentation device used in discourse strategy to influence the
addressee. Only the comment is meant to introduce new information. Thus, the
topic is – or is part of – the common ground supposed to be shared by speaker and
hearer; it is the part of the utterance with “old” meta-informative status to which it
is possible to add new information.
The term focus requires some further comment. In some linguistic approaches,
every part of an utterance which contributes “new” information is named a focus:
in the MIC theory, however, focus is a term which is defined not only (a) as new
information (a focus points to new information) but also (b) as an attention-
driven phrase, i.e. a part of the utterance which is distinguished or highlighted
among other parts of the same utterance and (c) as contrasting with another part
of the utterance having an opposite meta-informative status. Thus, not every part
of an utterance which contains new information should be considered as a focus.
A new chunk of information must furthermore be emphasized by some pointer to
become a focus. It should be borne in mind that in the current linguistic literature
of the field known as “information structure”, the word focus is extremely poly-
semous. It is sometimes used as a part of the common vocabulary of the English
language, either as a noun in the broad sense of “centre of interest” or as a verb
in the sense of “pay particular attention”. Sometimes, the word “focus” is used as
 Attention-centered information in language 

a part of specialized linguistic terminology and defined as “the part of a sentence

given prominence, usually for emphasis or contrast”.8 However we consider that
prominence is the effect (not the cause) of the contrary (inverse) meta-informative
status of the extension ADPs in an extended utterance with respect to the rest of
the utterance.
As was already mentioned, the implication relation between global and local
CAs may lead to ellipsis: since a local CA implies a global CA, the latter may
remain implicit, i.e. when the object is present, the subject may remain implicit
and, similarly in extended utterances, when the focus is present, the topic may be
omitted.

3.3  Interpreting a noun phrase as subject or object

Both the subject and the object are defined only on the meta-informative level
as global and local ADPs respectively. Their semantic roles are not expressed
directly but need to be “reconstructed” (inferred) taking into account several
pieces of knowledge, among which the knowledge of the verb valence and of the
fact that in nominative (as opposed to ergative) languages, in active voice, the
subject of a verb points by default to the active participant of the situation and
the object to the passive participant. When this default value of the semantic
roles which are pointed at by subject and object fails to yield a satisfactory inter-
pretation, hearers have to turn to their general knowledge of the world and/or to
the particular knowledge concerning the situation spoken about. In this regard,
understanding an utterance depends on inferences the hearer is able to draw in
order to go beyond the expressed participants by the ADPs and to reconstruct the
roles they play in the situation. Although these inferences are guided to a certain
degree by explicit linguistic indices present in the utterance, they always depend
ultimately on the knowledge of the represented world. In some cases, even in
inflectional languages when there occurs some homonymy in declension forms,
interpretation can rely merely on common knowledge. It is the case, for instance
in German (and Slavonic languages), of feminine and neuter nouns which have
homonymic forms for the nominative and the accusative cases. When such a
noun appears in an utterance, hearers are forced to rely on their common sense
knowledge in order to interpret it as subject or object. An example from the Ger-
man language is given in Mohr S. (2005). We quote this utterance (12) hereafter
and propose its interpretation within the MIC framework.

.  This definition is quoted from Webster’s online dictionary.

 Hélène Włodarczyk

(12) Ger.
Die Regierung betrachtet
n fem sg nom/acc V 3pers sg pres act
the government considers
das Volk als korrupt.
art n neu sg nom/acc conj + adj
the people as corrupt
“The people consider the government corrupt.”

This sequence of morphemes has several possible interpretations depending on

the different intonation and prosody it exhibits, which itself is related to speech
situation and text coherence. When intonation makes it possible to consider (12)
as a base utterance, the first level meta-informative gloss is
subject + predicate (verb + object + predicative adjective)

and the semantic interpretation:

“The government considers the people to be corrupt.”

But, if we interpret this utterance as an extended one, we have two possible meta-
informative glosses:

a. topicalized object + comment (verb + subject + predicative adjective)

which can be interpreted as (literally): “As for the government, people consider it
as corrupt.” If we put a pitch accent on the first NP we can interpret the utterance
as ­containing a focus:

b. Focalised object + background (verb +subject + ­predicative adjective)

Consequently, utterance (12) would be then translated into English by: “It is the
government that the people consider as corrupt.”
The following utterance (13) exhibits a different word order.
(13) Ger.
Das Volk betrachtet die Regierung als korrupt.
n neu sg nom/acc V 3pers sg pres active n fem sg nom/acc conj + adj
the people considers the government as corrupt
“The people consider the government corrupt.”

Furthermore, with a stress on the first NP in utterance (13), we get the following
interpretation:

“It is the people who consider the government corrupt.”

 Attention-centered information in language 

In utterances (12) and (13) both noun phrases have homonymic forms of nomi-
native and accusative cases because they are feminine and neuter nouns respec-
tively: Die Regierung (the government) and das Volk (the people). Nevertheless,
whatever be the word-order (either 12 or 13), any German native speaker prefers,
even out of context and in a written form with no prosodic marker, the read-
ing in which the people fill the active role and the government the passive role
because, as the author of the dissertation puts it, “… it is part of our knowledge
of the world that ­usually ­people consider the government corrupt and that it is not
the other way round.” (Mohr 2005). It is necessary to add that, in oral speech,
prosodic markers enable us to interpret both 12 as 13 either as base utterances
(with a subject+predicate structure) or as extended ones (with a topic+comment
or focus+background structure).
As a matter of fact, the MIC theory makes it possible to predict that, in written
German, each of these two utterances has theoretically possible various interpre-
tations among which the adressee can only choose when s/he takes into account
the common sense knowledge which makes it possible to reject the option of con-
sidering the two NPs alternatively as filling the active or passive role. The differ-
ent readings of each sentence depend on the prosody which makes it possible to
distinguish between base and extended utterances (with topic and/or focus). In
the above German examples, knowledge of the world makes it possible to interpret
utterance (12) as an extended utterance even in its written form, without prosodic
markers. Without prosodic markers (in the written form of the utterance consid-
ered out of context), it is impossible to decide whether it is a topic-comment utter-
ance in which the object NP “die Regierung” is topicalised or a focus-background
utterance in which the object NP is focalised (and therefore put in the first posi-
tion) or whether it is a focus-background utterance in case there is a stress on the
first NP. In German when the topicalised or focalised object appears in the first
position, as a consequence, the subject (here: das Volk) appears in the postverbal
position because German syntax puts the verb in the second place. Depending on
intonation, utterance (12) is interpreted as a base (subject-predicate) utterance or
as an extended (focus (here: focalised subject) – background) utterance. Thus the
different possible interpretations (and prosodic realisations) of a linear sequence
of morphemes depend ultimately on the text structure which in turn depends on
centering as cohesion device.
Moreover, let us point at the fact that in case marking languages (for Slavic
languages, see Włodarczyk H. 2009) any change of word order is sufficient to lead
to the extension of the utterance, i.e. the topicalisation or focalisation of the “dis-
placed” constituent. In German not only the first position in the utterance but also
the “after last” position (Nachfeld) can be used for topicalization or focalization
(Vinckel 2004).
 Hélène Włodarczyk

3.4  S ubject of passive utterances compared with topicalised Object

of active utterances
In nominative (as opposed to ergative) languages, both the subject of an utterance
in passive voice and the topicalised object of an utterance in active voice (with
an explicit subject or with an anonymous subject)9 point to the passive role of the
semantic situation expressed by the verb. The difference lies in that the topicalised
object is the global ADP of an extended utterance, whereas the subject is the global
ADP of a basic utterance. The choice of one or the other utterance depends on the
text coherence.
In French, problems of the possible equivalence of passive constructions and
of utterances with a “left dislocated” object are mentioned by Gaatone (1998: 239)
who quoted the following utterance (14) in which the topicalised object (“cet
effort”) is “left disocated”.

Fr. L’effort pour coordonner les données tactiles et visuelles, reste l’un
(14) 
des plus prodigieux de ma vie […]. Cet effort, chaque enfant en bas âge
l’accomplit.
“The effort to coordinate tactile and visual data remains one of the most tre-
mendous efforts in my life […] Such an effort, each infant has to ­accomplish
it.”

In (14) the Object NP “cet effort” is “left dislocated” (in our terms “topicalised”)
but Gaatone stressed that it could just as well be the subject of an utterance in
­passive voice, as in (15) below.

(15) Fr. Cet effort est accompli par chaque enfant en bas âge.
“This effort is accomplished by each infant.”

In our view, both utterances (14) and (15) fit in the text coherence but the choice
between them makes it possible to put more or less emphasis on the noun phrase
“cet effort”. The topicalised NP is more highlighted than the subject of the utter-
ance in passive voice and, last but not least, it has an old meta-informative status
which can be motivated by anaphora (cf. Chapter 3 in this volume). This is prob-
ably the reason why the topicalised object was used by the author of the original
version of this utterance and why, in fact, utterances (14) and (15) are not simply
equivalent.

.  See below § 4.5.

 Attention-centered information in language 

In Polish, as in other languages often characterised as “free  word order”

languages,10 because of the noun declension forms, the topicalisation of the object
is very frequent, although the passive voice may also be used to treat a NP refer-
ring to a passive role as subject. Compare utterances (16) and (17).
(16) Pol.
Książka została podarowana Jankowi
n fem sg nom V 3pers fem sg past pas n mas sg dat
book was offered to Janek
na urodziny (przez ojca).
prep n pl acc prep n mas sg acc
for birthday by father
“The book was offered to Janek for his birthday (by his father).”
(17) Pol.
Książkę to11 ojciec podarował
n fem sg acc MI part n mas sg nom V 3pers mas sg past act
book father offered
Jankowi na urodziny.
n mas sg dat prep + n -hum pl acc
to Janek for birthday
Lit. “The book, it is father who offered it to Janek for his birthday.”

Hereafter, the impersonal utterance (18) – with an anonymous subject – is close to

the passive utterance (16) in that both make it possible not to mention at all the
NP referring to the active role (ojciec: father), however utterance (18) puts more
emphasis on the first NP (książkę: the book) which therefore should be interpreted
as a topic.

.  Polish scholars insist that in Polish (and other Slavonic languages) the word order is “free”
but not “arbitrary”: “Układ wyrazów w zdaniu w języku polskim jest swobodny, j­akkolwiek
niezupełnie dowolny.” (“The word order in a Polish sentence is free, however it is not com-
pletely arbitrary”.) (Szober 1967). We claim, however, that each different order has a different
meta-informative interpretation.
.  In Polish, the use of the particle to after a topicalised NP and/or before a focalised NP
is not obligatory. More research should be conducted in order to demonstrate on large data
whether the use of to is really characteristic of oral communication and less frequently used
in writing (literary) style. Actually utterance (17) has two different readings depending on the
intonation: either only the first NP is topicalised and the rest of the utterance is the comment,
or the first topicalised NP is followed by a focalised NP as we interpreted it in the gloss we
proposed.
 Hélène Włodarczyk

(18) Pol.
Książkę podarowano
n fem sg acc V 3pers neu sg past imprs
book [it] was offered
Jankowi na urodziny.
n mas sg dat prep + n -hum pl acc
to Janek for birthday
Lit. “The book [someone] offered to Janek for his birthday.”

The topicalisation of an object (by “left dislocation”) is frequent in written Pol-

ish because it is stylistically neutral and relies only on a change in word-order. In
French (or English), it requires most frequently the use of an anaphoric pronoun
in the comment part of the utterance.12 It is usually assumed that in these lan-
guages, the topicalisation of objects is more frequent in spoken than in written
language however our observations of French literary, artistic, philosophial or sci-
entific styles suggest that topicalisation is an important discourse strategy device
which is widely used in different styles.13
Moreover, in many languages, it is also possible to topicalise even the subject
of a passive utterance. See for example the following Polish utterance (19).

(19) Pol.
Książka ta została podarowana
n fem sg nom det fem sg nom V Passive 3pers fem sg past
book this was offered
Jankowi na urodziny (przez ojca).
n mas sg dat prep + n -hum pl acc  prep + n mas sg acc
to Janek for birthday by father
“As for the book, it was offered to Janek for his birthday (by his father).”

In French, the topicalisation of the subject of a passive utterance is also possible in

the same way as the topicalisation of the subject or object NP of an active utterance.

(20) Fr. Le livre, il a été offert à Janek pour son anniversaire.

Lit. “The book, it was offered to Janek for his birthday.”

Such a topicalisation as in utterances (19) and (20) is often used to express a

contrast between the topicalised NP and another NP previously mentioned in
the text, this could be the case of “voiture” (car) in (22) and “samochód” (car)

.  Concerning French/Polish contrasts of this kind see (Włodarczyk, H. 2004).

.  This problem should be investigated using data from huge corpora.
 Attention-centered information in language 

in utterances (21) if put in contrast with “livre” in French utterance (20) and
“książka” Polish (19).
(21) Pol.
Samochód Janek
n mas sg acc n mas sg nom
car Janek
dostał na Boże Narodzenie.
V 3pers sg mas past act prep+n neu sg acc
got for Christmas
Lit. “The car, Janek got it for Christmas.”
(22) Fr. La voiture, Janek l’a eue pour Noël.
Lit. “The car, Janek got it for Christmas.”

Translators must be aware of the different respective uses of subject and topic as
text coherence devices, in order to render not only the semantic content of the
original text but also its pragmatic force.

4.  Subject in different types of languages

Linguists encounter many difficulties when searching for a universal definition of

the subject which would be valid for different types of languages. We propose to
define the subject along the three following dimensions:

1. As a syntactic constituent, it is realised as a noun phrase filling a place in the

verb valence with specific morphological markers, be it a case marker or the
position before or after the verb etc.
2. From a pragmatic perspective, the subject is defined first of all as an
attention-­driven phrase (ADP), i.e. the second argument of the speak_about
meta-informative relationship. However, we need to keep in mind that besides
this MIC theoretical innovation the subject corresponds also to another meta-­
informative relationship: speak_with. As a matter of fact, in many languages,
the subject of the utterance may be correlated with one of the participants
of the speak-with situation: either the speaker or the hearer.
3. The verb semantics provides a rather covert knowledge making it possible to
assign a semantic role to the participant of the described situation. The subject
itself by default points to the active role (in nominative languages).

In the next section, we will briefly consider the problems of explicit and non-
explicit subjects, of impersonal utterances and the relationship between the sub-
ject and the topic.
 Hélène Włodarczyk

4.1  Subject and category of person: Explicit subject

Subjects are always expressed explicitly in languages with verbs that have personal
inflection; i.e. in languages in which verbs mark the matching of the subject14
(the global ADP) of the utterance with the participants of the speech act (speaker,
hearer or participant of the semantic situation spoken about). In utterances (23),
(24) ad (25) hereafter, the subject is neither the speaker nor the addressee but a
person spoken about, referred to by a noun phrase (the student), therefore the verb
is in the third person form.
(23) The student is listening.
(24) Fr. L’ étudiant écoute.
(25) Pol. Student słucha.

In structural linguistics, (e.g. Martinet 1962, 1967), subject is defined as the obliga-
tory NP that cannot be deleted from an utterance without resulting in agram-
maticality. In our view, this criterion is difficult to apply because of the frequent
ellipsis that may affect any constituent of an utterance which bears the old meta-
informative status in a coherent text. For this reason, we prefer to put forward the
morphological factor. Subject is explicit when it is marked by a morpheme in the
verb ending, regardless of which semantic role it refers to, that is to say that subject
may be explicit not only in Indo-European nominative languages but also in erga-
tive languages (provided that we deal with a language of this group exhibiting verb
morphemes pointing at the participants of the speech-act).
However, in languages like contemporary English or French, in which the per-
sonal inflection of the verb is not clearly distinctive (neither in spoken nor written
language), the subject must be made explicit by the use of a personal pronoun. As
a matter of fact, in contemporary English, only the 3rd person singular form of
verbs in the present tense is distinct from all other persons in both oral and writ-
ten language.
(26) I | you | we | you | they think [as opposed to] he thinks.

In contemporary French, personal morphemes in the conjugation (of regular

verbs) in the present tense are distinctive only in the written form of the language:
(27) Fr. Je pense, tu penses, il pense, nous pensons, vous pensez, ils pensent.

.  In some languages verbs also have personal affixes for expressing not only subject but
also objects with the speech act participants, e.g. in Basque “there is agreement between the
case-markers of the NPs and the personal affixes in these three cases [the absolutive, the erga-
tive and the dative case]”, Brettschneider G. “Typological Characteristics of Basque” in Plank
1979: 372.
 Attention-centered information in language 

In oral contemporary French, in which 1st person plural form “nous pensons” is
frequently replaced by “on pense”, the only personal distinction concerns the 2nd
person plural form:
(28) Fr. Je/tu/il/on/ils pense(s) {pãs} [as opposed to] vous pensez {pãsε}.

In languages in which verbs have distinct morphemes for each person, the pres-
ence of the subject in the utterance may be marked only by the personal form of
the verb. This is the case in many Indo-European languages like Latin, Spanish and
Italian. In Polish15 where the personal morphemes of the verb are distinctive in all
moods and tenses, the first and second person pronoun in the nominative form
do not appear in the utterance unless the speaker intends to put an emphasis on
them, i.e. until s/he treats them not as a simple subject but either as a topicalised
or focalised subject (cf. (29) to (32)).
Pol. (from a novel by Maria Nurowska, Panny i wdowy,
(29) 
Warszawa 1992, p.117)
Można sie było tego spodziewać.
Ø AS -hum V past 3pers neu sg pro neu sg gen V inf.
[it] possible was this to foresee
Ty zawsze na siebie
pro 2pers sg nom Top adv prep+refl pro Ø pro 2pers sg nom
you always on yourself [you]
bierzesz winę.
V 2pers sg pres n fem sg acc
take [the] fault
“It was easy to foresee. It’s always you who take the fault on yourself.”
(30) Pol. (from a novel by M.Musierowicz, Pulpecja, Signum, Kraków 1993, p.133)
Coś mi tu nie
Indef pro neu sg nom pro 1pers dat adv neg
something for me here does not
pasuje w tym pokoju.
V pres 3pers sg g prep det n mas sg loc
fit in this room
“— There is something I don’t like in this room. Look at the curtains, how
they hang.”

.  In this respect, Polish differs from Russian where personal morphemes have disappeared
in past tense, thus speakers have to use personal pronouns for marking subjects. Compare the
Polish verb paradigm: widziałem, widziałeś, widział with the Russian one: я видел, ты видел,
он видел (I saw, you saw, he saw).
 Hélène Włodarczyk

To nie ja je tak
MI Part neg pro 1pers 3pers pro pl acc adv
it is not me them this way
powiesiłam tylko Julia.
V past 1pers sg part n sg nom
hung only Julia
“— It isn’t I/me16 who hung them but Julia.”

The same usage occurs in Italian where the personal pronoun is used only when it
is necessary to treat the subject as topic or focus.
(31) It. Pago …
V pres 1pers sg
“I pay…”
(32) It. Io pago …
pro 1pers sg V pres 1pers sg
“It’s me who pays …”

As a matter of fact, many languages exhibit (like Italian) two different series of
stressed and unstressed forms of personal pronouns in all cases of the declension
(see Chapters 12, 13 & 14 in this volume). Even in English or French where declen-
sion has almost entirely disappeared in nouns, there still exist different forms for
subject or object pronouns: I/me, he/him etc. Moreover, stressed forms sometimes
differ from the unstressed one; e.g.: in French moi/je, toi/tu etc. Stressed forms
are used in order to topicalise or focalise pronouns.17 In languages like Italian or
­Polish, the unstressed nominative 1st and 2nd person pronouns are simply omit-
ted (this kind of ellipsis can be considered as a “zero form” of pronouns). The pos-
sibility of omitting the personal pronoun which points to the subject function in
languages with personal morphemes in verbs is considered a universal device (e.g.
Encyklopedia 1999: 414). This omission also takes place in typologically different
languages and concerns not only subject pronouns but also object pronouns in
languages with two-fold (subject and object) personal morphemes in verbs (for
example in Basque). It is necessary to underline that the absence of the subject
in languages with verb personal endings should not be interpreted as a case of
implicit subject but rather as the use of the unstressed (zero) form of the 1st and

.  It seems that, from the MIC perspective, the conditions of use of either I or me as the
form of the topicalised 1st person pronoun in English should be reconsidered.
.  Out of context and in a written text (without intonation markers) it is not easy to
­interpret a stressed pronoun as a topic or a focus.
 Attention-centered information in language 

2nd person pronoun in the nominative case. We treat in a radically different way
the absence of subject in languages with no personal inflection (see below: implicit
subject).
From a diachronic perspective, it is worth adding that, concerning personal
inflection, modern personal morphemes of Indo-European verbs are former per-
sonal pronouns that were agglutinated to the verb radical.18

4.2  Implicit subject

In Japanese, there is no agreement between verb and subject because the Japanese
verb has no personal inflection and consequently, subject may remain implicit,
that is to say, when it is not expressed it must be inferred from the context or from
the situation in which the speech act takes place (Kuroda 1965; Ikegami 2007;
Nariyama in this volume). When only the object is present in the utterance, it
implies (as local ADP) that the subject (global ADP) must be inferred from the
context or situation as in utterances (33) and (34) hereafter:

(33) Jap.　魚を食べます。(Sakana o tabemasu.)

Sakana o tabemasu.
Ø subject N object part V pres polite
--- fish eat
“[I/you/he/she/we/you/they] eat fish.”

(34) Jap. 日本では、魚を食べます。 (Nihon de wa Sakana o tabemasu.)

Nihon de wa sakana o tabemasu.
N de locative part wa MI part Ø subject N + object part V pres polite
in Japan fish eat
“In Japan, [people/we] eat fish.”

In the situation expressed by the verb to eat (utterances 33, 34), the object noun
fish, referring to a passive participant, implies that animate being(s)19 are involved
in the situation as active agent(s). But it is also possible to express the subject
explicitly as in utterance (35).

.  In a more recent time, in the history of Slavic languages, Polish reinforced the plural first
person morpheme of the verb by adding -y to the -m inherited from Proto-slavic, e.g. Russian
идём (let’s go | we go), pol. idziemy (we go). Thus, the morpheme -my in modern standard
Polish is completely homonymous with the pronoun my (we).
.  In Japanese the distinction between singular and plural (excepts for the plural form -tachi
(達) which is used almost exclusively for humans) is not marked by grammatical morphemes.
 Hélène Włodarczyk

(35) Jap. 日本では、皆が寿司をよく食べます。

(Nihon de wa minna ga sakana o yoku tabemasu.)
“In Japan, everybody eats often fish.”
Comment
Topic Subject Predicate
Object Verb
Nihon de wa minna ga sakana o yoku tabemasu
N de locative part wa MI part pro subject N o object part adv V pres polite
in Japan everyone fish often eat

In Japanese, although there can be no agreement between the verb and the
subject, the honorific morpheme of the verb is, however, chosen in agreement
with the subject,20 be it only implicit or explicitly expressed in the utterance. In
­Example (36), it is possible to infer from the honorific form that the one who has
arrived is a person to whom the speaker owes respect.
(36) Jap. 鈴木先生がいらっしゃいました。
(Suzuki sensei ga irasshaimashita.)
Subject Predicate
Suzuki sensei ga irasshaimashita.
proper Noun professor nom. part V past polite deferent
professor Suzuki has arrived
“Professor Suzuki arrived.”

Depending on intonation and prosody, “Professor Suzuki” can be interpreted as a

subject similarly as above in the base utterance (36) or as a focus contrasting with
the background in the extended utterance (37):
(37) Jap. 鈴木先生が, いらっしゃいました。
(Suzuki sensei ga, irasshaimashita.)
Focus Background
Suzuki-sensei ga , irasshaimashita.
Suzuki professor ga focus part (pause) V past Polite Deferent
[It is] professor Suzuki [who] has arrived
“It is professor Suzuki who has arrived.”

Hereafter is a short dialog in Japanese with its English translation. None of the
following Japanese utterances has a subject neither is there any personal form of
the verb which would indicate the subject but in translation subject pronouns in
English or personal verb forms in Polish must be added.

.  Verb appreciative honorifics agree with the subject [Włodarczyk A. 1996].
 Attention-centered information in language 

Jap. A — 今日は！ B — やあ！　 A — どこに行くの？ B — 橋の下に魚を釣

(38) 
りにくんだ！
(A — Konnichi wa ! B — Yaa! A — Doko ni iku no? B — Hashi no shita ni
sakano o tsuri ni iku n’da.)
“A — Good morning/afternoon ! B — Hello ! A — Where are [you] going?
B —[I am] just going over there for fishing under the bridge.”
Pol. “A — Dzień dobry! B — Witaj ! A — Dokąd idziesz [2pers sg]? B —
Właśnie idę [1pers sg] pod most na ryby.”

4.3  Implicit subject and topic

In Japanese, a language with possibly implicit subject, extended utterances with
a topic are frequently used. In the framework of attention-centering this can be
explained by the fact that both subject and topic are global ADPs. When the global
ADP of the base utterance (the subject) remains implicit, the utterance is often
extended and entails its own global ADP (the topic). Let us analyse a few examples
of Japanese utterances (39–43) with a topic corresponding to the subject or to the
object. In English translations, the explicit subject or object needs to be reintro-
duced at least as an anaphoric pronoun (it appears in the following literal transla-
tions given in square brackets).

(39) Jap. その戦争は三年間続いた。(Sono sensoo wa san-nen-kan tsuzuita.)

Sono sensoo wa san-nen-kan tsuzuita.
Det N wa MI part Num N V past
This war three years lasted
Lit. “As concerns this war, [it] lasted three years.”
“This war lasted three years.”

(40) Jap. 日本はこの十年の間に大きく変わった。

(Nihon wa kono nijuu-nen no aida ni ookiku kawatta.)
Nihon wa kono nijuu-nen no aida ni ookiku kawatta.
N wa MI part det Num no part ni part adv V past
as concerns Japan during the last twenty years a lot [it] changed
Lit. “As concerns Japan, [it] has changed a lot during the last twenty years.”
“Japan has changed a lot during the last twenty years.”

(41) Jap. 若い者は、寒いと、酒ばかり飲んでいるよ。

(Wakai mono wa samui to sake bakari nonde iru yo.)
Wakai mono wa samui to sake bakari nonde iru yo.
N + wa MI part adj to N adv V pres yo modal part
as concerns Japan when cold alcohol only drink
Lit. “As concerns young people, when it is cold, [they] drink only alcohol.”
 Hélène Włodarczyk

(42) Jap. 田中さんからは、もう返事をもらいました。

(Tanaka-san kara wa moo henji wo moraimashita.)
Tanaka-san kara wa moo henji wo moraimashita.
N polite suffix from adv N wo V past polite
wa MI part object part deferent
as concerns from Mr Tanaka also already answer [I] received
“As concerns Mr Tanaka, [I] have already received his answer.”

(43) Jap. この本は、昨日買ったばかりです。

(Kono hon wa, kinoo katta bakari desu.)
Kono hon wa kinoo katta bakari desu.
det N wa MI part adv V past adv deferent
as concerns this book yesterday [I] bought only
“As concerns this book, [I] bought it only yesterday.”

As the translations of the above examples suggest, Japanese utterances where the
implicit subject has the same referent as the explicit topic, can often be trans-
lated into English by a basic utterance with a subject, (utterances 39, 40, 41),
instead of the literal translation with a topic. In other cases, the topic remains
present in the English translation alongside the co-referring subject or object
(utterances 42 and 43). But the choice of an utterance with or without a topic
in the translation from Japanese into a language with an obligatorily explicit
subject ultimately depends on the text coherence (on the place of a given utter-
ance within the whole text), and this cannot be shown on the basis of isolated
utterances taken out of context. Further on, translation with or without a topic
depends also on stylistic conventions which vary from language to language (see
§ 4.4. below).
On the other hand, it is necessary to stress the fact that in Japanese, notwith-
standing a widespread opinion among linguists, the particle wa is not exclusively
a marker of topic. Consequently, in Japanese discourse, topic-comment utterances
are not as frequent as occurrences of wa. As a matter of fact, the particle wa is also
used to mark the subject bearing the old meta-informative status matching a predi-
cate of the same meta-informative status (cf. Włodarczyk A. 2005). In base utter-
ances expressing general laws or observations, the wa particle can also be used as a
marker of generic, general, potential or habitual subject (Włodarczyk A. 1980, 1996,
1998). In utterance (44) below, pronounced without a pause after the wa particle,
both sora (sky) and aoi (is blue) bear the old meta-informative status motivated by
the situation type, i. e. the whole situation spoken about is considered as general.
Utterance (44) thus means “it is the characteristic of the sky to be blue” and not
“[today] the sky is blue”.
 Attention-centered information in language 

(44) Jap.　そらは青い。(Sora wa aoi.)

Sora wa aoi desu.
N wa MI part predicative Adj deferent
the sky is blue
“The sky is blue.”

The wa particle occurs in entirely old base utterances in which the old meta-­
informative status of subject and of the verb is motivated by the generic o
­ ntological
status.

(45) Jap. 　地球は太陽の周囲を回転する。

(Chikyuu wa taiyou no shuui wo kaiten-suru.)
Chikyuu wa taiyoo no shuui wo kaiten suru.
N subject wa particle N no particle N wo particle V pres
The Earth the Sun of round turns
“The Earth turns around the Sun.”

The difference between the use of wa as a a topic marker – as in utterances (39) to

(43) above – and its use to mark a general, generic, habitual or potential subject –
as in utterances (44) and (45) – lies in the intonation and prosody: the topicalised
phrase is pronounced with a specific prosody and separated from the comment by
a short pause.
Similarly, it is worth emphasizing that the ga particle (which is currently opposed
in linguistic studies to wa as a marker of non topicalised subject) is not an exclusive
marker of subject; very often this particle is used to mark the Focus of an utterance
as in Example (37) above (Włodarczyk A. 2005). As a matter of fact, the ga particle
can be used in two different types of utterances: to mark either a new subject noun
phrase in a base (entirely new) utterance or a focalised NP in an extended utterance.
Therefore the Japanese wa and ga particles should not be defined merely as
markers of topic and subject but more generally as meta-informative particles since
they can be used both on the first and the second meta-informative level in order
to mark either a simple new or old status subject in a base utterance or a topicalised
(old) or focalised (new) noun phrase (topic or focus) in an extended utterance. Thus,
we refute the most famous theory of the wa particle in the history of linguistics,
the Japanese wa particle is NOT simply a topic particle but more generally a meta-
informative particle (Włodarczyk A. 1998) bearing the old meta-informative status.

4.4  Topic prominency and subject prominency

In the light of the definitions of subject, object, topic and focus as ADPs in the MIC
framework, it is necessary to revisit the typological classification of world languages
 Hélène Włodarczyk

into subject- and topic-prominent types (Li & Thompson 1976). In European gram-
mars, from traditional up to more recent ones, topicalisation and focalisation were
either not paid any attention at all (when marked only by prosody or word-order),
or considered as typical of colloquial speech when marked by dislocation and cleav-
ing (Blanche-Benveniste et al. 1990; Miller & Fernandez-Vest 2006). Thanks to the
possibility of exploring large corpuses it will be possible to reappraise ready-made
ideas in linguistics according to which in languages like French and English cleav-
ing and dislocation (markers of extended utterances in RWO languages) belong
only to colloquial speech. This view takes its origin in the syntax of European lan-
guages, itself rooted in classical (Greek and Latin) grammatical tradition which is
focused on base utterances, namely on the subject-predicate (or NP+VP) structure,
leaving aside extended utterances. The classical European grammatical tradition
was continued in syntactic studies during the 20th century (up to Noam Chomsky’s
generative syntax), therefore most syntactic theories still lack concepts and terms to
deal with extended utterances. In the MIC theory, subject and topic structures are
both meta-informative: the subject-predicate structure belongs to the first meta-
informative level whereas the topic-comment belongs to the second one. The prob-
lem will have to be thoroughly explored on large corpora and it will be necessary
to observe whether one or the other of two types of utterances is really (i.e.: with
statistical proof) more frequently used in some languages than in others. To be of
some evidential value such exploration must be conducted both on original utter-
ances in their discourse context and with huge parallel corpora. In order to achieve
this task, the MIC theory offers reliable formalised criteria for identifying topics
and distinguishing simple subjects from topicalised ones.

4.5  Anonymous Subject

It is well known that languages with explicit subjects also exhibit subjectless
utterances under the name of impersonal ones. As we were working on a for-
mal model of impersonal sentences in Polish and Russian (Włodarczyk 1994)
we were led to put forward the hypothesis that such utterances in fact contain
a “hidden” subject, making it possible to give an account of different types of
impersonal sentences that are distinguished by native speakers. We proposed
calling this “hidden” subject anonymous.21 In languages in which verb inflection

.  For more arguments in favour of the anonymous subject hypothesis, especially in Russian
and Polish, please see Włodarczyk, H. 1994, 1996. That hypothesis was elaborated and tested
on a computer using Prolog. This program contains rules generating and analysing different
kinds of ‘impersonal’ utterances in Polish and Russian, these rules being homologous to the
ones which generate utterances with a subject.
 Attention-centered information in language 

displays distinctive p
­ ersonal morphemes (e.g. Slavonic languages, but also some
Romance languages), those morphemes are the only mark of the presence of this
anonymous subject in the utterance, as is the case in Polish (46) and Latin (47)
utterances.
(46) Pol. Pada.
(47) Lat. Pluit.
“It is raining.”

In French or English an anonymous subject is marked by an impersonal pronoun.

The pronoun referring to an anonymous subject is sometimes called “dummy
subject” in syntactic theories: in languages with personal inflection,22 the term
“dummy subject” captures the fact that in an utterance the place of the subject is
empty (Ø).
(48) Fr. Il pleut.
(49) It is raining.

Although German verbs have personal morphemes in the present tense, a neutral
pronoun must be used to mark the anonymous subject.
(50) Ger. Es regnet.

Evidence for the hypothesis that in languages with explicit subjects, impersonal
utterances do entail an anonymous subject is corroborated by the existence of
two distinct semantic types of anonymous subjects: one refers semantically to
the set of animate non-human entities (–hum), the other one to human beings
(+hum).23 This distinction can be observed in many European languages, for
instance: English, French, German, Russian, Polish. As a matter of fact, besides
the -hum anonymous subject (like the one in utterances 46–50) above which is
used in order to refer to -hum entities as, for instance, atmospheric phenom-
ena, there exists another type of anonymous subject referring to human beings.
Furthermore, within the category of human anonymous subjects, languages

.  Mel’čuk (1988) put forward the concept of “zero subject”.

.  As regards Polish, cf. Grzegorczykowa et al. (1988: 200): ”Podczas gdy konstrukcje nieo-
sobowe na się i -no/-to: chodzi się, wykonano, milczano, bito, implikują istnienie osobowego
wykonawcy czynności lub nosiciela stanu czy procesu, chociaż go bliżej nie określają, konstrukcje
nieosobowe typu zasypało (drogę) wykluczają osobowego sprawcę czynności (agensa).“In
English: “Whereas such [Polish] impersonal utterances with forms się and -no/-to (e.g. chodzi
się, wykonano, milczano, bito) entail the existence of a personal agent of the action (or that of a
person involved in a state or process) without closer determining this person, other impersonal
utterances of the type zasypało (drogę) exclude a personal agent of the action (agens)”.
 Hélène Włodarczyk

are endowed with different devices enabling speakers to distinguish between

an anonymous human subject considered as the generic human being (utter-
ances 51–54) and a somewhat specified human being belonging to a socially
­identified group (the police, a company, the government, etc. as in (56) below).
In the f­ormer case the speaker may identify herself/himself 24 as belonging to
the generic human person and s/he frequently uses a pragmatic device (i.e. the
second person verb form and pronoun) in order to involve not only her/himself
but also the addressee, cf. (52) and (54).
(51) One never knows what may happen.
(52) You never know what may happen.
(53) Pol.
Nigdy się nie wie, co może się zdażyć.
adv Ø 3pers Refl pro neg Rel pro V V inf
neu sg V 3pers sg neu 3pers neu 3pers
+hum +hum +spk pres sg nom sg G
Never [one] does not know what may happen

(54) Pol.
Nigdy nie wiesz, co może się zdażyć.
adv Ø 2pers G neg V 3pers sg G rel pro 3pers V 3pers V inf
sg +hum +hum +spk pres neu sg nom sg G
Never [you] do not know what may happen

On the other hand, the speaker may insist on the fact that she herself (he himself)
does not belong to the mentioned group (cf. Polish utterances 55–56 below).
(55) Pol.
Zatrzymano go na drodze.
Ø3pers neu V 3pers sg neu pro 3pers masc prep + N fem
sg +hum +hum -spk sg acc sg loc
[someone] stopped him on [the] road
“He was stopped (by the police) on the road.”

.  It is perhaps worth mentioning that in English, as in most European languages, the
­masculine personal pronoun he was until recently (Halliday & Hasan 1976), the unmarked
gender form for the generic human being, and it was therefore not necessary to use the
­unpronounceable form s/he.
 Attention-centered information in language 

(56) Pol.
Zatrzymali go na drodze.
Ø3pers pl V 3pers pl pro 3pers prep + N
+hum +hum -spk masc sg acc fem sg loc
[They] stopped him on [the] road
Lit. “He was stopped (by the police) on the road.”

Utterances with +hum anonymous subject in Slavic languages, like Polish utter-
ance (57) can often be rendered into English or French by a passive utterance
in which the participant playing the active role remains implicit as soon as
the object of the passive (by whom) is not expressed: “He was stopped (by the
police) on the road.” Moreover, as has already been pointed out for Polish, lan-
guages may use several different types of anonymous subject referring either
to a generic human being or to a more limited group of persons: the choice
between them is determined by semantic nuances as well as pragmatic and sty-
listic conditions.25
Only when taking into account the semantic distinction between the anony-
mous +hum and –hum subject, is it possible to understand the communicative
purposes for which impersonal utterances are used. Such utterances allow us to
identify the global CA very generally, merely as a participant (of the situation)
belonging to a class of entities (+hum or -hum) without selecting a particular ele-
ment of one of these classes. As is the case with an explicit subject, the semantic role
of the participant referred to by the anonymous subject may be various and it has
to be reconstructed by the hearer when interpreting the utterance (Włodarczyk H.
1996). Just as the semantic role pointed to by an explicit subject, the one to which
the anonymous subject points, is not expressed in the utterance; it is a part of
covert information.
In languages with explicit subjects, the anonymous subject corresponds to a
linguistic sign with both (a) the form (the “morphonological zero form” inclusive,
e.g. in Polish, Russian, Italian; or the invariant form of an indefinite pronoun as
man, es in German, on, il in French) and (b) the content (it refers to an indefinite
either human or non-human entity). But anonymous subjects differ from indefi-
nite pronouns (such as someone or something) in that they may occur only in the
position of the subject of an utterance (or exist only in the nominative case-form
in languages with declension). The presence of an ­anonymous subject in an utter-

.  About Polish, Russian and French, see Włodarczyk, H. 1996. We distinguish several types of
+hum and –hum AnS in Russian and Polish, for a detailed discussion see Włodarczyk, H. 1994.
 Hélène Włodarczyk

ance (even when it has a zero morphonological form) is manifested by agreement

constraints on the verb. For instance, when automatically generating imper-
sonal utterances in Russian or Polish we must take into account these ­agreement
­constraints in order to produce acceptable utterances (Włodarczyk H. 1994).
The concept of the anonymous subject makes it possible to explain French
utterances in which the active participant is demoted to the position of the object
(local CA) when the position of the subject is occupied by an anonymous subject.
Such utterances enable the speaker to introduce an active participant as a local CA
(i.e. as an object !) even when the verb is otherwise intransitive, as in utterances
(57) and (58) hereafter.

(57) Il vient des étrangers.

“There come strangers.”

(58) Il arrive de drôles de choses.

“There happen strange things.”

The choice of such constructions can be explained in the MIC framework by

­discourse coherence factors:

“On the other hand, TCs [there Clauses] are close to inverted constructions in
that the information presented in the postposed segments (e.g. the Object or local
CA) is consistently less familiar within the discourse than that presented in the
preposed segments (e.g. the Subject or global CA). The use of TCs is, in this sense,
the result of the speaker’s choice to postpone unfamiliar information and place
it in the local CA position of the utterance for the sake of maintaining discourse
coherence.”(Martinez-Insua 2011: 103–104)

Thus impersonal constructions can be used in the so called strict word-order lan-
guages (like French or English) to postpone until the end of an utterance with an
entirely new status the introduction of a NP with a new meta-informative status.
The French constructions with the pronoun il or the English there constructions
used to express an anonymous subject make it possible to treat the participant
introduced after the verb in both languages as a local CA (i.e. as a forward-looking
centre) although the use of a direct construction with this participant worded as
an explicit subject would imply that the participant is treated in the utterance as
a backward-looking centre, relying on information previously introduced into the
discourse.
To return to the problem of the anonymous subject, let us stress that there
is in fact no significant difference between those anonymous subjects which
have a “morpho-phonological zero form” and those which have the form of an
 Attention-centered information in language 

indefinite pronoun occuring in the nominative case only. On the contrary, the
semantic distinction between +hum and –hum anonymous subjects is attested
cross-linguistically regardless of their form (either a morpheme or a zero form).
Necessarily, however, impersonal utterances with an anonymous subject occur
only in languages with an explicit subject. In such languages however, utterances
with an anonymous subject must not be confused with utterances in which, due
to the text coherence, the subject undergoes ellipsis, e.g. utterance (60) used in
answer to (59).

(59) – What is Bill doing?

(60) – Watching TV.

It is worth adding that in languages with an implicit subject (like Japanese) the
concept of the anonymous subject is simply absurd since there is no need (gram-
matical obligation) to replace the subject by some abstraction.

4.6  I nterpreting topicalised objects in active impersonal utterances

(with an anonymous subject)
The isomorphism between subject and topic has often led linguists to interpret
a topicalised object as a subject, especially in Indo-European languages when a
sentence has an anonymous subject (Haspelmath 2001; Fischer 2004). Due to the
frequent topicalisation of the object of impersonal utterances, topicalised objects
are often interpreted as subjects in an oblique case or as objects with “subject-
like” properties. Observe that reasons given in favour of such an interpretation are
mostly of a syntactic nature (Fisher 2004). In the MIC framework, we call Subjects
only those NPs which are in the unmarked case form (nominative26 or absolutive).
However, in theoretical frameworks where the presence of an anonymous subject
is not taken into account, it has been argued that when there is no explicit Sub-
ject in the nominative form another NP in an oblique case may be considered as
subject called “oblique subject”, “dative subject”, “quasi-subject”, “subject-like NP”
or “quirky subject” in the typology of Germanic languages (Haspelmath 2001;
Givón 2001; Fisher 2004). We consider that this interpretation of NPs in oblique

.  The only exception we admit for Russian and Polish concerns genitive NPs (with a parti-
tive meaning), in fact this genitive is an ‘adnumeral’ form of the nominative or accusative cases
with numerals or with negation. For details see Włodarczyk, H. – 1992, 1993, and 1996. The
concept of ‘adnumeral’ or “second” genitive was introduced into French linguistic studies on
slavic languages by Menantaud 2008.
 Hélène Włodarczyk

cases not as objects but as subjects is motivated by the fact that such NPs are very
often topicalised (“fronting” in Fisher’s terms) and that topic is taken for a subject
because it shares with the subject the property of being a global CA.
Let us compare the following Polish utterances hereafter: (60) with a subject,
(62) with an anonymous subject and (63) with both an anonymous subject and a
topicalised object.

(61) Pol.
Śnieg zasypał drogi.
N sg masc nom V 3pers sg past N pl fem pl acc
snow covered roads
“Snow covered the roads.”

(62) Pol.
Śniegiem zasypało drogi.
N sg masc instr Ø 3pers sg neu -hum V 3pers sg past N pl fem. pl acc.
[by] snow [it] covered roads
“The roads have been covered by snow.”

In (62), scholars admitting the existence of “oblique subjects” consider that the
noun snow (śniegiem) may be considered as an “instrumental subject” due to its
semantic quasi-active role and probably because of its being topicalised. In the MIC
framework however, we claim that the noun śniegiem is not a subject but a topi-
calised (indirect) object, and that the subject has the zero form of a -hum anony-
mous subject (expressed by “it” in English).27 Notice that though, in utterance (63),
it is the direct object NP drogi (roads) corresponding to a semantic passive role
which is topicalised, this NP cannot be interpreted as an “oblique subject” either.

(63) Pol.
Drogi zasypało śniegiem.
N fem pl acc Ø 3pers neu sg -hum V 3pers sg past N sg mas instr
roads [it] covered [by] snow
“The roads have been covered by snow.”

Thus, in MIC, we claim that in impersonal sentences of Indo-European languages

where the position of the subject is occupied by an anonymous subject, a noun
phrase in the case form of an object (in the dative or accusative or instrumental

.  Polański (2003) defends the same point of view, considering that NPs in oblique cases
appearing in Polish impersonal utterances should not be treated as subjects.
 Attention-centered information in language 

case) cannot be interpreted as the so-called “oblique subject”. For this reason the
concept “oblique subject” looses its theoretical validity.

5.  Word-order as a meta-informative marker

In current linguistic theories, word order (WO) is considered to be a marker of

syntactic relations. For instance, we can read in the handbook of syntax that “both
morphology and word order are used to code the grammatical roles of ­subject and
object” [Givón 2001: 233]. In addition, today’s word order typology (Dryer 2007)
is based on the assumption that word-order concerns the ­syntactic ­constituents of
expressions and is therefore governed by syntactic rules.
In the MIC theory, we consider that the syntactic structure of a base u ­ tterance
reflects directly the pragmatic definition of subject and object as attention-driven
phrases referring to centres of attention, and the pragmatic definition of predicate
as “what is said about the subject”. As was shown at the beginning of this chapter,
we claim that the core of syntax does not reflect directly information (the semantic
dimension) but meta-information (the pragmatic dimension). Syntax is a means of
expression (along with morphology, phonology and prosody) used for ­linearising
the information content as communicated by an utterance; this linearisation
would be impossible without the selection of a global centre of ­attention in the
semantic situation spoken about. Therefore, it is meta-information (not informa-
tion) that is directly reflected by the elementary syntactic structure of an utterance.
The fact that information is independent of meta-information allows variation in
the combinability of centres of attention with semantic roles. ­Meta-informative
competence consists of the ability to present the same information in different
meta-informative word orders, i.e. to establish the mappings between centres of
attention and semantic roles.
Most linguists can agree that each language has a canonical word order.
Indeed, it is possible to define the canonical word order of a given language by the
linear arrangement of the main constituents (subject, object, verb) of a declarative
(as opposed to interrogative, injunctive etc.) utterance in which both subject and
object are expressed by noun phrases (specific WO rules apply when subject or
object are expressed by pronouns or subordinate clauses). Moreover, the utter-
ance in canonical word-order is pragmatically neutral, i.e. none of its constituents
receives special prominence: this corresponds to what we call “base utterance”.
What is more, pragmatic neutrality is expressed in each language by a declarative
intonation characteristic solely of base utterances.
In modern typology, languages are divided into two types: rigid word order
(RWO) languages (e.g.: English, French, Japanese etc.) and free word order (FWO)
 Hélène Włodarczyk

languages (e.g. Germanic, Slavic languages, Turkish etc.) The latter are considered
to have different degrees of word-order freedom. It should be added that, even in
FWO languages there exists a canonical WO of the base utterance. Any change
with respect to this canonical WO produces a pragmatic effect (among which are
­topicalisation and focalisation),28 consequently the “freedom” of FWO languages
is not arbitrary.
In the MIC framework, both so called RWO and FWO languages can be
dealt with in a consistent way. According to this theory, in both types of lan-
guages ­word-order is a meta-informative marker used to highlight attention-
driven phrases and to distinguish between old and new information. In order to
determine the canonical WO in a language we must only take base utterances
into consideration. In French and English, WO is used mainly to distinguish
between the subject and the object of a base utterance. In order to extend an
utterance, French or English speakers much refer to explicit segmental markers
(adverbs or particles) which lead to important changes in the syntactic structure
of the utterance: cleft sentences, dislocation, etc. In FWO (for example in Slavic)
languages morphological case markers of subject and object make it possible to
use the word order on both (1) the first meta-informative level – in order to dis-
tinguish entirely new from entirely old base-utterances and (2) the second meta-
informative level where the mere dislocation of the subject or object can induce
its topicalisation or focalisation (Włodarczyk H. 2009). However, even in this
type of languages, the change in word-order often depends on explicit segmental
markers (particles, such as “to” (lit. “this”) in Polish and Russian).
Thus, in the MIC approach, we replace the typological distinction between
RWO and FWO languages by the distinction between languages in which the
word order concerns mainly the first meta-informative level, i. e. the base utter-
ance (like in French and English), and those in which the word order is used
mainly as a marker of the second meta-informative level (like in Slavic languages).
But this classification must be understood as a weighed heterarchy because there
is probably no language in which word order is used exclusively for marking the
meta-informative status in base utterances. In addition, let us emphasize the fact
that besides word order, intonation is the most important marker of the second
meta-informative level (it is a favourite expressive device for marking the topic or
focus). Since, however, it may be sometimes difficult to perceive intonation (even
in oral communication) and since intonation is not sufficiently marked in most
writing systems, most languages use additional, explicit meta-informative m­ arkers

.  For the MIC approach to these problems in Polish and Russian see Włodarczyk, H. 2009.
 Attention-centered information in language 

of the results of topicalisation and focalisation: these are mostly particles and spe-
cial syntactic constructions.

6.  Conclusion

Linguistic communication is only a part of diverse activities performed by humans

in their environment so that, in this regard, language can be considered auxiliary to
action. Nevertheless, in language use, meaning is the most important motivation.
But utterances do not express meaning directly in so far as speakers cannot commu-
nicate information objectively but only with respect to their own point of view and
to their own knowledge. Because of their meta-informative, hence subjective dimen-
sion (exhibited by numerous passive or active, personal or impersonal, topicalised
or focalised paraphrases etc.), natural languages should be considered diametrically
different from artificial languages which are designed for establishing a bijective
mapping between form and meaning in order to free expression from subjectivity.
However, this does not mean that logical language should not be used to represent
the information dimension of natural language; as a matter of fact, first order logic is
an excellent device for representing many aspects of semantic situations.
Yet, in natural languages, morphology and syntax convey semantic contents
which are inextricably interwoven with pragmatic ones. This makes human commu-
nication fast and efficient. But in order to analyse expressions made up of elements
which belong to various dimensions, the ASMIC theory attempts to represent them
separately, explaining how morphologic and syntactic forms of a particular language
are matched to semantic and pragmatic representations. Although many concepts
(such as subject and predicate) used in the ASMIC framework sound astonishingly
traditional and hackneyed, they are integrated in a new coherent formalised theory
in which their scope is fixed by strict terminological conventions. For this purpose,
this theory does not exclude any formal representation device used in other lin-
guistic, logical or computational approaches to language. Tree structures (both con-
stituency and dependency ones), graphs and lattices are regarded not as alternative
but as complementary representations. The attempt at formalising thoroughly the
ASMIC theory will further depend on the use computer scientists make of it in their
attempts at building artificial models of linguistic communication.

Acknowledgement

I am grateful to André Włodarczyk who provided me with Japanese language

examples and their interpretation in the MIC-theoretical framework.
 Hélène Włodarczyk

Abreviations

/ or lat. Latin
1pers first person loc locative
2pers second person mas masculine
3pers third person mi meta-informative
acc accusative n noun
act active voice neg negation
adj adjectif neu neuter
adp attention-driven phrase nom nominative
art article np noun phrase
ca centre of attention num numeral
conj conjunction part particle
dat dative pas passive voice
det deictic pf perfective
en. English pl plural
fem feminine pol. Polish
fr. French pp prepositional phrase
g gender as a variable (fem, mas or neu) prep preposition
ge. German pres present
gen genitive pret preterite
±hum human/non-human pro pronoun
imprs impersonal refl reflexive
inf infinitive rel relative (pronoun)
instr instrumental sg singular
ip imperfective ±spk speaker involved or not
it. Italian v verb
jap. Japanese

References

Blanche-Benveniste, Claire, Bilger, Mireille, Rouget, Christine & van den Eynde, Karel. 1990. Le
français parlé: Études grammaticales. Paris: Éditions du CNRS.
Brettschneider, Günter. 1979. Typological characteristics of Basque. In Ergativity: Towards a
Theory of Grammatical Relations, Frans Plank (ed.), 371–384. London: Academic Press.
Bunting, Michael F., Cowan, Nelson & Colflesh, Gregory J. H.. 2008. The deployment of atten-
tion in short-term memory tasks: Tradeoffs between immediate and delayed deployment.
Memory & Cognition 36(4): 799–812. 
Chafe, Wallace L. 1994. Discourse, Consciousness and Time. Chicago IL: The University of
­Chicago Press.
Corbetta, M. 1998. Frontoparietal cortical networks for directing attention and the eye to visual
locations: identical, independent, or overlapping neural systems. Proc. Natl. Acad. Sci. USA
95, 831–838.
Cowan, Nelson. 2008. What are the differences between long-term, short-term, and working mem-
ory? In Progress in Brain Research, Vol. 169: Essence of Memory, Wayne S. Sossin, Jean-Claude
Lacaille, Vincent F. Castellucci & Sylvie Belleville (eds), 323–338. Amsterdam: Elsevier.
 Attention-centered information in language 

Dryer, Matthew S. 2007. Word order. In Clause Structure, Language Typology and Syntactic
Description, Vol. 1, 2nd edn, Timothy Shopen (ed.). Cambridge: CUP.
Encyklopedia językoznawstwa ogólnego, 2nd edn. 1999. Kazimierz Polański (ed.), Ossolineum,
Wrocław – Warszawa – Kraków.
Fischer, Susann. 2004. The diachronic relationship between quirky subjects and stylistic front-
ing. In Non-nominative Subjects [Typological Studies in Language 60], Peri Bhaskararao &
Karumuri V. Subbarao (eds), 193–212. Amsterdam: John Benjamins.
Gaatone, David. 1998. Le Passif en Français. Paris: Editions Duculot, Champs Linguistiques.
Givón, T. 1988. The pragmatics of word order: Predictability, importance and attention. In Stud-
ies in Syntactic Typology [Typological Studies in Language 17], Michael Hammond, Edith
A. Moravcsik & Jessica Wirth (eds). Amsterdam: John Benjamins.
Givón, T. 1988. 1994. The pragmatics of voice: Functional and typological aspects. Voice and
Inversion [Typological Studies in Language 28], T. Givón (ed.), 3–44. Amsterdam: John
Benjamins.
Givón, T. 2001[1984]). Syntax: An Introduction, Vol.1. Amsterdam: John Benjamins.
Grice, H. Paul 1975. Logic and conversation. In Syntax and Semantics, Vol 3, Peter Cole & Jerry
L. Morgan (eds). New York NY: Academic Press.
Grosz, Barbara J. & Sidner, Candace L. 1986. Attention, intentions, and the structure of d
­ iscourse.
Computational Linguistics 12(3): 175–204.
Grosz Barbara J., Joshi, Arawind K & Weinstein, Scott. 1995. Centering: A framework for
­modeling the local coherence of discourse. Computational Linguistics 21(2): 203–226.
Grzegorczykowa, Renata; Laskowski, Roman & Wróbel Henryk. 1998. Gramatyka współczesnego
języka polskiego, Morfologia, wyd. 2 zmienione, T. 1–2. Warszawa: P.W.N.
Halliday, Michael A.K. & Hasan, Ruqaiya. 1976. Cohesion in English. London: Longman.
Halliday, Michael A.K & Greaves, William S. 2008. Intonation in the Grammar of English.
­London: Equinox.
Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages.
In Non-Canonical Marking of Subjects and Objects [Typological Studies in Language
46], Alexandra Y. Aikhenvald, Robert M.W. Dixon & Masayoki Onishi (eds), 53–83.
­Amsterdam: John Benjamins.
Hayashi, A., Tomlin, Russell S. & Yokota, T. Attention detection and Japanese (Japanese Fish
Film Paper) 〈http://logos.uoregon.edu/tomlin/in_progress.html〉
Ikegami, Yoshihiko. 2007. Subjectivity, ego-orientation and subject-object merger. In Japa-
nese Linguistics European Chapter, Ikegami Yoshihiko, Viktoria Eschbach-Szabo & André
Włodarczyk (eds). Tokyo: Kurosio.
James, William. 1890. The Principles of Psychology, 2 vols. Dover Publications 1950.
Jespersen, Otto. 1924, The Philosophy of Grammar. London: George Allen & Unwin.
Kuroda, Shigeyuki. 1965. Subject, in Japanese Generative Grammar – Syntax and Semantics,
Vol. 5, Masayoshi Shibatani (ed.), 1–16. New York NY: Academic Press.
LaBerge, David. 2003. Attention, awareness and the triangular Circuit. In Essential Sources in the
Scientific Study of Consciousness, Bernard J. Baars, William P. Banks & James B. Newman
(eds). Cambridge MA: The MIT Press, Bradford Book.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus, and the Mental
Representations of the Discourse Referents. Cambridge: CUP.
Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, Vol. I: Theoretical Prerequisites.
Stanford CA: Stanford University Press.
Langacker, Ronald W. 1991. Foundations of Cognitive Grammar, Vol. II: Descriptive Application.
Stanford CA: Stanford University Press.
 Hélène Włodarczyk

Li, Charles N. & Thompson, Sandra A. 1976. Subject and topic: A new typology of language. In
Subject and Topic, Charles N. Li (ed.), 457–61. New York NY: Academic Press.
Martinet, André. 1962, A Functional View of Language. Oxford: Clarendon. (In French: Langue
et fonction, Denoël, Paris 1969).
Martinet, André. 1967, Éléments de linguistique générale. Paris: Armand Colin.
Martínez-Insua, Ana E. 2011. When discourse matches syntax: On meta-informative centering
theory and discourse coherence in the recent history of English. International Journal of
English Studies (IJES) 11(2): 91–111.
Menantaud, Henri. 2008. Génitif premier et génitif second: Faut-il diviser le génitif polonais
en “sous-cas”? L’Enseignement du polonais en France: 2èmes Assises, 49–54. Paris: Institut
d’Études Slaves.
Miladi, Lidia. 2006. Topicalisation et focalisation des complétives et/ou des infinitives dans les
constructions verbales du polonaise. Actes du colloque international ‘Constructions verbales
et production de sens’ [Collection «Recherches en linguistique étrangère], 224–241. Besan-
çon: Presses universitaires de Franche-Comté.
Miladi, Lidia. 2009, Les fonctions discursives de la particule énonciative to du polonais dans
les constructions à segment détaché à gauche. Revue des Etudes Slaves, numéro s­pécial:
La  cohérence du discours dans les langues slaves: Linguistique théorique et textuelle,
LXXX(1–2): 87–103.
Miller, Jim & Fernandez-Vest, Jocelyne. 2006. Spoken and written language. In Eurotype: Typol-
ogy of Languages in Europe Pragmatic Organisation of Discourse in Languages in Europe,
Giuliano Bernini & Maria Schwartz (eds), 9–64. Berlin: Mouton de Gruyter.
Mohr, Sabine. 2005. Clausal Architecture and Subject Positions: Impersonal Constructions in
the Germanic Languages [Linguistik Aktuell/Linguistics Today 88]. Amsterdam: John
Benjamins.
Morel, Marie-Annick & Laurent, Danon-Boileau. 1998. Grammaire de l’intonation, L’exemple du
français. Paris: Faits de Langues, Ophrys.
Myachykov, Andriy, Garrod, Simon & Scheepers, Christoph. 2009. Attention and syntax in sen-
tence production: A critical review. Discours, Revue de linguistique, psycholinguistique et
informatique 4: Linearization and Segmentation in Discourse. 〈http://discours.revues.org/
index7594.html〉
Oakley, Todd, 2009. From Attention to Meaning. Explorations in Semiotics, Linguistics, and
­Rhetoric. Frankfurt: Peter Lang.
Polański, Kazimierz (ed.). 1999 (wydanie 2). Encyklopedia językoznawstwa ogólnego. Wrocław.
Warszawa. Kraków. Zakład Narodowy im. Ossolińskich.
Polański, Kazimierz. 2003. Na marginesie dyskusji o podmiocie. Etudes linguistiques romano-
slaves offertes à Stanisław Karolak, 409–416. Kraków: Oficyna Wydawnicza Edukacja.
Paul, Hermann. 1880[1995]. Prinzipien der Sprachgeschichte, 10th edn. Tübingen: Niemeyer.
Shelton, Jill T., Elliott, Emily M. & Cowan, Nelson. 2008. Attention and working memory: Tools
for understanding consciousness. Psyche, 14(1).
Talmy, L. 2000. Toward a Cognitive Semantics. Cambridge, MA: The MIT Press.
Tomlin, Russell S. 1995. Focal attention, voice, and word order: An experimental, cross-­
linguistic study. In Word Order in Discourse [Typological Studies in Language 30], Pamela
A. ­Downing & Michael Noonan (eds), 517–554. Amsterdam: John Benjamins.
Tomlin, Russell S. 1997. Mapping conceptual representations into linguistic representations:
The role of attention in grammar. In Language and Conceptualization, Jan Nuyts & Eric
­Pederson (eds), 162–189. Cambridge: CUP.
 Attention-centered information in language 

Vinckel, Hélène. 2004, La ‘projection à droite’ en allemand contemporain (Ausklammerung).

In Énoncer – l’ordre informatif dans les langues, Pierre Cotte, Martine Dalmas & Hélène
Włodarczyk (eds). Paris: Coll. Sémantiques, L’Harmattan.
Włodarczyk, André. 1980. 主題から主語へ、そして主語から主題へ　—　ハとガ (Shudai kara
shugo e, soshite shugo kara shudai e – wa to ga) (From Topic to Subject and from Subject
to Topic – wa and ga). In 言語月刊 (Language monthly), Vol. 9, No 8/80, 大修館 Tokyo:
Taishûkan.
Włodarczyk, André. 1996. La thématisation en japonais et l’article français. In Sémantique et
Linguistique Contrastive. Warszawa: Slawistyczny Ośrodek Wydawniczy (SOW) Polska
Akademia Nauk (PAN).
Włodarczyk, André. 1998. The proper treatment of the Japanese “wa” and “ga” particles. In
Proceedings of the International Workshop on Human Interface Technology 1998. (IWHIT
‘98)  – Aizu-Wakamatsu, Japan.
Włodarczyk, André. 1999. La validation informative des énoncés et la quantification linguis-
tique. Études cognitives/Studia kognitywne 3: 121–133. Warszawa: SOW, PAN.
Włodarczyk, André. 2003a. Les Homotopies du topique et du focus. In Ordre et distinction dans
la langue et le discours, Actes du colloque international de Metz (18,19 et 20 mars 1999), Ber-
nard Combettes, Catherine Schnedecker & Anne Theissen (eds), 513–526. Paris: Honoré
Champion éditeur.
Włodarczyk, André. 2003b. Linguistique dynamique: Évolution du discours dans le temps. In
Études linguistiques romano-slaves offertes à Stanisław Karolak, 497–510. Kraków: Oficyna
Wydawnicza Edukacja.
Włodarczyk, André. 2004. Centres d’intérêt et ordres communicatifs. In Énoncer, l’ordre infor-
matif dans les langues, Pierre Cotte, Martine Dalmas & Hélène Włodarczyk (eds), 13–32.
Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André. 2005. From Japanese to general linguistics – Starting with the ‘wa’ and ‘ga’
particles. In Paris Lectures on Japanese Linguistics, André Włodarczyk (ed.), 27–62. Tokyo:
Kurosio Shuppan.
Włodarczyk, André. 2008. Roles and anchors of semantic situations. Études cognitives/Studia
kognitywne 8: 53–70. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 1998. Graded informative content of linguistic
­sessages. In Proceedings of the 16th International Congress of Linguists. Oxford: Pergamon.
(CD-ROM: Paper N0 0431).
Włodarczyk, André & Włodarczyk, Hélène. 2006a. Focus in the meta-informative centering
theory. In La Focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André & Włodarczyk, Hélène. 2006b. Subject in the meta-informative centering
theory. Études cognitives/Studia kognitywne 7: 39–64. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2008a. Roles, anchors and other things we talk
about: Associative semantics and meta-informative centering theory. Intercultural Prag-
matics 5(3): 345–365.
Włodarczyk, André & Włodarczyk, Hélène. 2008b. The pragmatic validation of utterances.
Études Cognitives/Studia Kognitywne 8: 117–128. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2012. Information centering: Subjecthood and
topicality. In An International Workshop on Linguistics of BA and The 11th Korea-Japan
Workshop on Linguistics and Language Processing, 10–11 December 2011. Tokyo: Waseda
University.
 Hélène Włodarczyk

Włodarczyk, Hélène.1994. Les phrases à sujet anonyme et verbe réfléchi en polonais et en russe,
Organisation sémantique de la phrase dans les langues slaves, Revue des Études Slaves T.
LXVI (3): 669–680, Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 1996. Les traits sémantiques du sujet “anonyme” en polonais, russe et fran-
çais, Semantyka a konfrontacja językowa: 179–198. Warszawa: SOW, PAN.
Włodarczyk, Hélène. 1998. Wykładniki wartości informacyjnej wypowiedzenia w j. polskim i
francuskim (aspekt, okreslonosc, modalnosc), Congrès des Slavistes Cracovie 1998, Revue
des Etudes Slaves T. LXX(1): 53–66. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 1999. Les marqueurs de la validation des énoncés en français et polonaise.
Études cognitives/Studia kognitywne 3: 135–162. Warszawa: SOW, PAN. (Erratum: dans cet
article, il faut lire S pour P et P pour S).
Włodarczyk, Hélène. 2003. L’Interprétation dynamique des centres d’intérêt dans les dialogues
français et polonais. In Études linguistiques romano-slaves offertes à Stanisław Karolak,
­511–528. Kraków: Oficyna Wydawnicza Edukacja.
Włodarczyk, Hélène. 2004. Les centres d’intérêt de l’énoncé en polonais et en français. In Énon-
cer, l’ordre informatif dans les langues, Pierre Cotte, Martine Dalmas & Hélène Włodarczyk
(eds), 33–48. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, Hélène. 2006. Struktura meta-informacyjna wypowiedzenia w języku polskim, III
Kongres Polonistyki Zagranicznej, Poznan 2006; Postcriptum 2006/2 Nr 52: 181–185. Kato-
wice: Szkoła Języka i Kultury Polskiej Uniwersytetu Śląskiego.
Włodarczyk, Hélène. 2008. Du bon (et du mauvais) usage des formes accentuées et atones
des pronoms personnels en polonaise. L’Enseignement du polonais en France, IIes Assises:
39–48. Société française d’Etudes Polonaises. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 2009. Les centres d’intérêt de l’énoncé et la cohérence textuelle en polonais
et russe. Revue des études slaves LXXX(1–2): 13–32. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 2012. L’emploi des pronoms personnels en polonais – par contraste avec le
russe et le français. In Études à la mémoire de J. Breuillard sous la dir. de S. Viellard, Revue
des Études Slaves, t. LXXXIII(2–3): 617–648. Paris: Institut d’Études Slaves.
part 2

Neuropsychological evidence
for the MIC theory
Semantic and episodic memory by reference
to the ontological grounding of the old
and new meta-informative status

Franz J. Stachowiak
Charles Sturt University Albury, NSW, Australia

The present paper argues for a model of language production and comprehension
in which a MIC-component interacts with working memory, episodic memory,
and semantic memory as resources for the meta-informative processing of
new and old information. This component is also interconnected with the
grammatical encoding system which produces “surface” structures. Focused
attention plays a decisive functional role in this system as it directs the processes
which transform intentions into messages. Although MIC theory is essentially
a linguistic theory it offers itself to direct mapping with notions from cognitive
and clinical neuropsychology. Levelt (1989) dedicates pioneering work to the
attention resources, the focus of the message and the macro-planning of the
information in the early stages of speech production, however, his model remains
vague with respect to the formal status of the cognitive systems involved. MIC
theory complemented by findings from cognitive neuropsychology can fill these
gaps. Despite ample evidence for the relative independence of the different
memory systems (e.g. double dissociations between episodic and semantic
memory; particular vulnerability of episodic memory in medial temporal lobe
impairment and diseases such as Alzheimer’s), recent research accentuates the
interaction between these systems in building up generic “semantic” knowledge
from autobiographical experience and vice versa facilitating the retrieval of
information from episodic memory by contributing generic information from
semantic memory. Input of information into the MIC-component from episodic
memory is more likely to be attributed “new status” than inputs from semantic
memory, although attention processes can redirect the meta-informative status
of that information at any time depending on outcomes from ToM (Theory of
Mind), which is attributed an active role in this model. ToM is the capacity of
humans to understand mental states of others including the ability to judge the
contents of their respective memory systems. In verbal interaction estimates of
the communication partners’ memory systems’ contents form an input to the
MIC component of the speaker who selects CAs. This ability is seen in early
childhood and understood as a key factor in Tomasello’s (2000) social-pragmatic
theory of word learning and language acquisition. The model proposed here will
 Franz J. Stachowiak

be further scrutinized by examining cognitive communication disorders caused

by lesions of the right hemisphere, which will be explained as disruptions of the
interaction of the modules described in the MIC based model.

1.  Introduction

In contrast to other linguistic theories the AS-MIC theory considers pragmat-

ics as “not a mere supplement of the semantic and syntactic structures of utter-
ances” (Włodarczyk, A., Chapter 3 in this volume). The pragmatic competence
of a speaker/listener according to this view consists of his/her ability to produce
and understand verbal expressions whose predication structure exhibits meta-
information serving to draw attention to and to detect the “subject” of the mes-
sage. The spotlighted centres of attention are related to the world (“ontology”,
“encyclopaedic knowledge” etc.) via the semantic system (lexical, sentential, text
semantics). One of the main features of communication according to this model is
that the information to be conveyed is treated as new or old, the newness or old-
ness being its meta-informative status depending on the discourse strategies (i.e.
the pragmatic attitude) of the speaker with respect to the addressee of the utter-
ance. “To sum up, meta-informative old and new status are motivated by (a)  the
flow of information in an inter-utterance communication (anaphora/cataphora),
(b) ­concerns resulting from the discourse (known/unknown) and (c) beliefs pertain-
ing to ontological knowledge” (ib. conclusion). Levelt (1989) dedicated pioneering
work to the attention resources, the focus of the message and the macro-planning
of the information in the early stages of speech production, however, his model
remains vague with respect to the formal status of the cognitive systems involved.
MIC theory complemented by findings from cognitive neuropsychology can fill
these gaps. Insofar as notions such as information, knowledge, attention etc. are
related to the way our mind works, this theory offers itself to a direct mapping with
notions from cognitive neuropsychology and psycholinguistics and opens up an
interdisciplinary approach to understanding the interrelationship of the modules
or components of this system and the flow of information between them when
meta-informative steps are undertaken to shape an utterance.
Basic processes of attention and working memory interact to determine the
typical positions of Centres of Attention (CAs) in the meta-informative forma-
tion of utterances. I have shown in the chapter devoted to the role of memory
and attention for the meta-informative validation of utterances (next chapter
in this volume) that well known phenomena such as the primacy and recency
effects known from studies of human memory influence the decision where to
locate a CA.
 Semantic and episodic memory w.r.t. old and new meta-informative status 

2.  New and old information in utterances

The present paper examines the role of semantic and episodic memories as
ontological source or grounding of old vs. new information. It starts from the
assumption that these memory systems (in connection with the working memory
system) are continuously being addressed (storage and retrieval processes) dur-
ing the production and perception of utterances. It needs to be noted, however,
that the newness or oldness of information is dependent on the speaker’s wish (or
decision or compulsion) which meta-informative status to attribute to it. Gen-
eral public knowledge of a name like “Albert Einstein”, which would “normally”
be considered as old information as it stands for a famous figure in science, can
nevertheless acquire a meta-informative status as “new”. To use an example from
A. Włodarczyk (Chapter 3 in this volume): “In the following utterance: It is Albert
Einstein who discovered the law of Relativity, the subject Albert Einstein is focalised
as a new chunk of information contrasting with what is treated as the old infor-
mation: there is a person who discovered the law of Relativity (supposedly although
you do know that this law has been discovered, what you do not know is the name
of that physicist) and I am telling you: that person’s name is Albert Einstein.” This
oscillation between new and old information and especially the drawing of atten-
tion to something as new on the background of known information is reminiscent
of the classical distinction of Gestalt-Psychology between figure and ground. The
focalization on Albert Einstein in the sentence above makes only sense in the con-
text of someone for whom this information is new. The speaker assumes that the
addressee has attributed this fact to another person, for instance to “Otto Hahn”
or “Madame Curie” or does not know at all, who was the discoverer of Relativ-
ity. In a TV quiz with multiple choice questions, in which the three names are
given as possible solutions to the question “Who discovered Relativity?” this sen-
tence would be understood as the solution to a trivial question, either confirming
a correct guess or correcting a wrong answer. In this case “Albert Einstein” would
be given the status of new information although the name had been mentioned
before in that context, namely as one of the possible solutions.
Interestingly, the discovery of Relativity was an event in Albert Einstein’s life,
which must have left strong traces in his and his colleagues’ episodic memories,
but the important fact that it happened became general knowledge, i.e. semantic
knowledge of billions of traditionally educated people. In contrast, most private
events of Einstein’s life are not known to the public. The fact that he got divorced
from his wife Mileva in order to marry his cousin Elsa, cannot be considered as
general knowledge. However, if this information is predicated to his name, new
epistemic knowledge is produced. The MIC theory distinguishes clearly between
two kinds of situations: Σ-situations: generic > general > potential > habitual on
 Franz J. Stachowiak

the one hand and Π-situations: specific > particular > actual > occasional on the
other. The events of the divorce and the remarriage belong to the Π-type of situa-
tion. In neuropsychological terms one could speak of information retrieved from
episodic memory, whereas the names and well known facts such as the discovery
of Relativity are said to be stored in semantic memory. However, there is a certain
problem in linking the linguistic approach to the neuropsychological one in our
example. The fact that Einstein married twice may have been part of his own and
his wives’, friends’ and family’s episodic memories but not of other persons’ except
biographers’, historians’ or simply curious people who find this information in the
internet. When distinguishing between old and new information we make refer-
ence not only to our own memory, but also to what we assume to be stored in the
memories of our communication partners.
This in mind, let us first examine what the nature and structure of episodic
vs. semantic memories is and how they interact in forming our autobiographi-
cal memories, world knowledge or encyclopaedic knowledge (to use some of the
usual terms). Later, on that basis, we will consider how they interact with what
I would like to call the MIC-component of the pragmatic competence of a speaker/
listener.

3.  Episodic and semantic memory

The term “episodic memory” was introduced by Tulving (1972), who assumed
it to be a late evolutionary development out of semantic memory and unique to
humans, the essence of which “lies in the conjunction of three concepts – self,
autonoetic awareness, and subjectively sensed time” (Tulving 2002: 5). Episodic
memory pertains to the learning, storing and retrieval of autobiographical facts,
i.e. unique personal experiences, to events in our lives, occurrences, careers,
encounters, talks with others, accidents, etc. The question “Where were you and
what did you do when September 11, 2001 happened?” is a question about an
episode in one’s personal life, which might be easy to retrieve as it must have
been connected with world breaking news at the time the incidence happened.
­Ethnographic analysis (Reeves, Kuper & Hodges 2008) offers ideas on the orga-
nizational categories which determine the structuring (and restructuring) of
­episodic memory:

–– Space – Physical layout of the place(s)

–– Actor – Range of people involved
–– Activity – A set of related activities that occur
–– Object – The physical things that are present
 Semantic and episodic memory w.r.t. old and new meta-informative status 

–– Act – Single actions people undertake

–– Event – Activities that people carry out
–– Time – The sequencing of events that occur
–– Goal – Things that people are trying to accomplish
–– Feeling – Emotions felt and expressed

Following Tulving (2002) these aspects would have to be related to the autobio-
graphic self, but as will be shown later, should also be related to the “self ” of others
from the viewpoint of an observer.
Using MIC terminology, “September 11, 2001” would be a Π-situation.
In contrast: semantic memory, although it is – like episodic memory – a form
of long term declarative (factual) memory, stores knowledge that can be learnt
independently of reference to time and place. According to the classical view it
is not relevant where and when it was learnt. Evolution has given animals trait
memory, which might be an early form of semantic memory in Tulving’s sense, but
humans, of course, build up knowledge from learning. The fact of the discovery of
Relativity can be learnt at school without any personal involvement. “Encyclopae-
dic knowledge of information such as the features of objects (e.g. apples are usu-
ally red), categories (e.g. oranges and bananas are both types of fruit), historical
events, mathematical tables, cognitive maps, and similar types of information are
considered to be stored in semantic memory systems of the brain” (Dickerson &
Eichbaum 2010: 87).
The term semantic memory is not to be interchanged with what “semantics”
stands for in linguistics. The subjective lexicon can certainly be considered as part
of semantic memory. However, the semantic system relevant for language pro-
duction and comprehension is a functionally autonomous system. The meaning
of words, the selectional restrictions for the combining of word meanings, the
phrasal semantics etc. are relevant for the functioning of the language system and
can be disturbed in the case of aphasia without a general loss of semantic memory
or encyclopaedic knowledge. Thus a person’s knowledge about trees in general is
wider than the meaning of the word “tree”, and semantic disturbances in aphasia
compromise primarily the linguistically relevant semantic network (or feature sys-
tem or decision table system (Miller and Johnson-Laird 1976) depending on the
model you want to use) related to the “lemma” (Levelt 1989) of the lexical entry
“tree”, but not the encyclopaedic or generic knowledge about trees in its entirety.
Interestingly, closed class words (function words such as articles, prepositions etc.)
can be severely affected in Broca’s aphasia, i.e. retrieval of them becomes very dif-
ficult in oral speech and reading, whereas open class words do not pose major
difficulties for these patients. Frontal lesions as in the case of Broca’s aphasia hit
function words as they play a major role in the syntactic processing of ­utterances.
 Franz J. Stachowiak

­emporo-parietal lesions leading to Wernicke’s aphasia can compromise the

T
semantic differentiation of words severely. Modern imaging technology shows
that the linguistically relevant semantic system is subserved by an extended neural
network covering the whole language area and its wider connections. However,
the views on the exact nature of the relationship between semantic memory in
the wider neuropsychological sense and the narrower linguistic description of the
semantic system remain controversial and a matter of further research.

4.  Findings from cognitive neuroscience

Clinical findings indicating a dissociation between retrieval from semantic and

episodic memory have been cited as evidence for the psychological reality and
relative functional independence of the two systems. First observations were made
on the famous patient H. M., whose medial temporal lobes were removed as a cure
against intractable epilepsia (Scoville & Milner 1957). This involved the bilateral
resection of his rostral medial temporal lobes (MTLs), including the hippocampal
formation, the amygdala, and overlying cortex. As a result of the surgery H. M.
showed a profound anterograde and retrograde amnesia with preservation of
remote memory, whereas other aspects of his intellect and personality remained
intact. His lexical knowledge, as part of his semantic memory, was not impaired.
But also his test results on general world knowledge (Kensinger, Ullman & Corkin
2001) were apparently in the normal range. The latter showed no effect across time
so that it was concluded that his semantic memory was not negatively affected by
the surgical lesion. His main problem was to acquire new, long-term declarative
memories including episodic and semantic memory, which initiated the discus-
sion about the role of the hippocampus in acquiring and integrating new informa-
tion into long term memory. Using modern imaging technology the localization
of episodic and semantic memory functions in the brain has become more precise.
Grewe, Rossum and Donaldson (2007) demonstrated electrophsiologically that
semantic memory and episodic memory interact in normal functioning, not only
by modulating recollection, but also by enhancing familiarity. Prince, Tsukiura
and Cabeza (2007) found that:
“within the left temporal lobe, the hippocampus was associated with successful EE
(Episodic Memory Encoding), whereas a posterior lateral region was associated
with SR (Semantic Memory Retrieval). Within the left inferior prefrontal cortex,
a posterior region was involved in SR, a mid region was involved in both SR and
EE, and an anterior region was involved in EE, but only when SR was also high.
Thus, the neural correlates of EE and SR are dissociable but interact in specific
brain regions.”(p. 144)
 Semantic and episodic memory w.r.t. old and new meta-informative status 

Accordingly recent models assume a rather complex feedback system between

episodic and semantic memory. They accentuate the interaction between these
systems in building up generic “semantic” knowledge from autobiographical
experience and vice versa facilitating the retrieval of information from episodic
memory by contributing generic information from semantic memory. Episodic
information is vulnerable to forgetting. Most of us who are members of research
associations and attend annual scientific conferences which take place in different
places every year cannot remember – after a few years – the exact sequence of con-
ferences, where they took place, in which hotels we stayed and which contents were
dealt with in which year. But we store the script of the conference, how it is usually
organized and we develop a kind of generic conceptual system of what this confer-
ence is normally about, in fact we could summarize this in a few sentences and tell
others why we go there and what kind of information we expect to get there. This
generic kind of information helps us to prepare for the next actual meeting, to
judge its proposed contents and meet certain people. It also helps us to remember
the last occurrences of this conference by providing us with a structured concep-
tual framework into which we fit the information about the latest episodes. Andrew
M. Nuxoll and John E. Laird (2007) have pointed to the role of episodic memory as
a critical component for cognitive architectures that support general intelligence.
They propose a model in which episodic memory overlaps with case-based rea-
soning and supports noticing novel situations, detecting repetition, virtual sensing
(“to remember seeing a coffee shop just around the corner from here”), action and
environment modelling, predicting successes and failures, managing long term
goals, retroactive learning and reanalysis of knowledge, explaining behaviour and
boosting learning. In their model a close interaction between working memory
and episodic memory is assumed, the term semantic memory is, however, not
used. Klein et al. (2009) come to similar conclusions. For them episodic memories
enable us to re-evaluate our past in light of new information and thus contribute to
human social interaction. In their model they relate episodic to semantic memory:
“Episodes that are inconsistent with a trait summary can provide boundary con-
ditions on its scope – they provide information about the situations in which the
generalization does not apply” (p. 306). In general, semantic memory would speed
up decisions, whereas episodic memory would provide accuracy. An interesting
new model, which was presented by B ­ attaglia and Pennartz (2011), follows these
lines. They consider the construction of semantic memory as grammar based rep-
resentations learned from relational episodic information:
“Experimental evidence points to a gradual transformation of memories, following
encoding, from an episodic to a semantic character. This may require an exchange
of information between different memory modules during inactive periods. We
propose a theory for such interactions and for the formation of semantic memory,
 Franz J. Stachowiak

in which episodic memory is encoded as relational data. Semantic memory

is modelled as a modified stochastic grammar, which learns to parse episodic
configurations expressed as an association matrix. The grammar produces tree-
like representations of episodes, describing the relationships between its main
constituents at multiple levels of categorization, based on its current knowledge
of world regularities. These regularities are learned by the grammar from
episodic memory information, through an expectation-maximization procedure,
analogous to the inside–outside algorithm for stochastic context-free grammars.”
(Battaglia & Pennartz 2011: 1)

In this sense it seems reasonable to assume a continuum between episodic and

semantic memory, starting with fresh memory of the events of the hour and the
day and stretching to general knowledge such as how a day is usually spent. I had
the occasion to meet a patient with global amnesia, who had lost practically all
of his episodic memories up to the point of the onset of his illness. He showed
also signs of semantic memory loss. The 42 year old teacher of mathematics and
physics had suffered a heart arrest followed by a thalamic lesion. He had to believe
other people that he was married, had children, had built a house, had worked at
a particular school etc. as all of that information was lost from his memory. He
had no aphasia and showed no semantic disorder, but when asked what a chicken
is, he was very uncertain whether it is a kind of bird or another animal. The sen-
tence “Chickens lay eggs” was fully familiar to him. In a pizzeria he noted in an
astonished way: “Ah, that’s a Pizza”. When eating it he said that he knows the taste.
He was able to demonstrate – after some preparation – that he could still teach
mathematics. His behaviour showed the typical demeanour of a teacher, and he
understood the contents well enough to make them plausible to the students. He
understood the rules of soccer and watched matches on TV, but did not know any
player and could not remember, whether Germany had won the World champion-
ship. In comparison to the deep loss of his episodic memory (except for a short
period in his childhood), his semantic memory and world knowledge was intact
to such an extent that he could lead a rather independent life. His anterograde
amnesia improved, and he was able to learn new things such as the fact that he
was married. He admitted that he had sensed some intimacy with the lady who
purported to be his wife and that he had once called her spontaneously “love”, but
that he had to believe her everything that she told him about their family. This
case matches very well with the theory of Battaglia and Pennartz (2011). It was
particularly interesting with respect to the way how the patient communicated
with others. Much of what was said to him was completely new for him, even if
he did not admit it. Thus he often appeared to have come from another planet.
He masked this by giving very general answers to questions thus relying on the
semantic knowledge that he had. But in some cases he seemed to be agnostic,
 Semantic and episodic memory w.r.t. old and new meta-informative status 

i.e. not knowing or recognizing things. The “normal” pragmatics of selecting and

recognizing centres of attention in the sense of the MIC theory was widely sus-
pended in this case.

5.  Grounding, mirror neurons and theory of mind

What is attributed the meta-informative status of new or old information in ver-

bal communication, cannot be related directly to the different types of memory.
As A. Włodarczyk has shown (Chapter 3 in this volume) decisions on this status
are speech act bound (inter-utterance communicative grounding), related to the
recency of knowledge acquisition (discourse grounding) or related to the type of
information stored in episodic and semantic memory (ontological grounding).
Expanding on the last point it should be kept in mind that according to Tulving’s
definition the contents of episodic memory are restricted to what we experience
in our life time. Generally speaking one cannot conclude from this that semantic
memory contains old information and episodic memory stores new information.
The age of acquisition of some autobiographical information can be relatively old
as some people remember childhood episodes from as early as 2 years of age. In
contrast, some new semantic knowledge can be acquired till the end of life. How-
ever, in terms of the difference between figure and ground, it seems that semantic
memory lends itself rather to form the ground, whereas information from epi-
sodic memory typically forms the figure as in the example “It was Chopin whom
Georges Sand invited to the party” (Włodarczyk A., Chapter 3 in this volume). The
invitation to the party is an event stored in the couple’s and their friends episodic
memories and has as such been narrated to the world after them.
Basic to human interaction is that we continuously try to read other peoples’
mind. We are guided by presumptions about what others know, think, feel and
intend and on what kind of experiences their thinking, knowledge, emotions and
intentions are based upon. This includes naturally that we try to scan their epi-
sodic and semantic memories in order to find out what is new to them and what
is known by them. Two related theories try to explain this behaviour: The mir-
ror neuron theory (Gallese 2001) provides the basis for the human capacity of
empathy and the theory of mind (Dennet 1977) describes the ability of humans
to acknowledge that others have minds by analogy to their own. It implies that we
interpret others’ activities as goal directed and based on particular beliefs, desires,
knowledge etc. This intentionality or “aboutness” is a continuous driving force for
directing attention in human communication. Tomasello (1999) has shown exper-
imentally that the ability to share attention with other human beings emerges in
non-linguistic form near the end of the first year of life:
 Franz J. Stachowiak

“Thus, many different studies have found that children begin to develop joint
attentional skills at around 9 to 12 months of age, including such things as
following the gaze direction and gestures of adults, imitating adult actions on
objects, and directing adult attention to outside objects using various kinds of
gestural signals (see Tomasello 1995a, for a review). Most children also show their
first signs of comprehending language at this same age, with the first linguistic
productions coming soon after”. (Tomasello 1999: 406)

At the same time these theories offer the solution to the problem that episodic
memory is considered as related to the self and to autobiographical events. When
distinguishing between old and new information we try to match our own episodic
and semantic memories with those of others, and narratives about the actions of
others are based on the original contents of their episodic memories which were
either transmitted to others or extrapolated by communication partners.

6.  C
 ognitive communication disorders as disturbances
of the MIC-component

There is clinical evidence from diverse syndromes that the system of meta-­
informative shaping of an utterance can be disturbed so that communication with
these patients becomes rather cumbersome. Cognitive communication disorders
following from right hemisphere lesions, caused for instance by Traumatic Brain
Injury (TBI), stroke or tumour result from the underlying cognitive deficits caused
by the neurological impairment, i.e. from disorders of attention, memory, organi-
zation, information processing, problem solving, and executive functions. These
difficulties in communicative competence are distinct from aphasia. Phonological,
grammatical or lexical-semantic problems are generally not symptoms of cogni-
tive communication disorders. Impaired attention, difficulties processing com-
plex information, difficulties learning new information, problems distinguishing
between relevant and irrelevant information, inefficient retrieval of stored infor-
mation, problem solving difficulties and inappropriate social behaviour due to lack
of empathy and failing to read others’ minds as well as inflexibility and impulsivity
have an impact on the way these patients communicate. Abusamra, Coré, Joanette
and Ferreres (2009) found that the expression of patients with right hemisphere
damage is often less informative than that of control subjects and that this infor-
mation is organized more simply. Characteristic is “lack of coherence and strong
tendency to tangential discourse” (p. 74) and “what distinguishes patients with
RHD from aphasic patients is their tendency to make inappropriate comments or
to stray from the topic of the story.” (p. 74).
 Semantic and episodic memory w.r.t. old and new meta-informative status 

The following conversational example, in which a male patient fails to inter-

pret a metaphor is given by the authors (pp. 77–78):

E: What does this phrase mean: My friend’s mother-in-law is a witch?

P: Let’s change also one word: My son-in-law’s mother-in-law is a witch?
E: And so what does it mean?
P: I know she is a person who hasn’t had a pleasant life, throughout her
­marriage. That ….that she’s about to be separated from her husband;
I’m ­referring to the mother-in-law of my son-in-law (ha, ha, ha)
E: OK it’s not important – it’s the same.
P: Certainly! The mother-in-law of my son-in-law. The mother-in-law of my
­son-in-law is a witch!
E: What does being a witch mean?
P: Because the woman is separated, because all her life she has criticized her
husband for the way he is: only seen in his defects, who has kept his daughter
all her life under a glass bell and she’s now a poor lady because she can’t find
the fiancé her mother would like.
E: So what does witch mean, then?
P: What does it specifically mean? It means being tied-down to religious sects, to
religions, to umbanda …who knows, there are so many.
E: So therefore, “the mother-in-law of my son-in-law is a witch”? Does it
mean the mother-in-law of my friend practices black magic? And the
mother-in-law of my friend has many brooms and she is also a bad
­person and rude?
P: It’s absolutely clear. My friend’s mother-in-law has many brooms…no!
My friend’s mother-in-law practices black magic.

The patient obviously refers the question to his own family and seems to approach
a correct description of why he considers his wife a witch, underlining this with
information from his episodic memory, but when asked again he drifts away from
the topic and refers to generic information pertaining to religion etc, which is
semantically related but rather to the literal meaning of “witch” than to the meta-
phorical meaning which the examiner has in mind. In the end the patient takes
only up what has been said in the context and confirms the question (contextual
grounding).
Similarly schizophasia is characterized by a problem to uphold attention to
relevant information, patients often produce chains of associations to what was
said in the context and drift away from the topic. “Among the more noticeable
abnormalities in the speech samples she (Chaika) collected were ‘production of
new sentences according to phonological and semantic features of previously
uttered discourse rather than according to topic’ and ‘failure to monitor own
 Franz J. Stachowiak

speech.” (Mitchell & Crow 2005: 968). The following example from Ian Thompson
(2001: see also ­discourse in neuropsychiatry) shows this typical kind of sheering
off from the centre of attention:
“Parents are people that raise you. Anything that raises you can be a parent. Parents
can be anything, material, vegetable or mineral, that has taught you something.
Parents would be the world of things that are alive, that are there. Rocks, a person
can look at a rock and learn something from it, so that would be a parent.”

These clinical examples show that the meta-informative centering of information

in running speech can be disrupted by certain lesions to the brain. The right hemi-
sphere obviously plays a major role in this, it’s cooperation with left hemisphere
language functions seems to be indispensable. The following flow chart (Figure 1)
tries to capture the interrelationship between the MIC component and the major
cognitive modules processing an utterance such as “It is Einstein who discovered
the law of Relativity”.

7.  Empirical evidence for the model

As outlined above and in my next chapter (in this volume) clinical evidence shows
the relative functional autonomy of the modules depicted: aphasia can destroy the
grammatical component and the linguistic semantic component, cognitive com-
munication disorders can compromise the MIC-component as a result of atten-
tion and memory problems. Different brain injuries can impair working memory,
episodic memory and semantic memory in rather isolated ways leading to various
dissociations of disturbances. In the intact brain all these systems interact, as mod-
ern imaging technology proves. A research gap needs to be closed by investigating
the relationship between episodic memory and verbal behaviour. How are narra-
tives structured which have to rely on variable contents of episodic and semantic
memory?
An interesting study – however without the view of a combined linguistic and
neuropsychological model as presented here – has been undertaken in cognitive
neuropsychology. In an experiment in which subjects had to recall three different
types of story: a previously unknown story, a well-known fairy tale, and a modi-
fied well-known fairy tale, Dalla Barba, Attali and La Corte (2010) were able to
show that the interference of strongly represented, overlearned information in epi-
sodic memory recall is implicated in the production of confabulations of old age
people. This effect turned out to be particularly prominent when the to-be remem-
bered episodic information showed strong semantic similarities with ­preexisting,
 Semantic and episodic memory w.r.t. old and new meta-informative status 

Referential World – Events – Ontology

SENSORY INPUTS INPUTS FROM MEMORY

PERTINENCE Central EPISODIC MEMORY-Figure

executive
• Space–Physical layout
Visuo-spatial Episodic Phonological of the place(s)
STORAGE sketch-pad buffer loop
• Actor–Range of
people involved
Visual Episodic Language
semantics LTM • Activity–A set of
related activities that occur
Fluid systems Crystalized systems
EINSTEIN • Object–The physical
things that are present

• Act–Single actions
people undertake

ATTENTION • Event–Activities that

people carry out

• Time–The sequencing
Verbal context| of events that occur
• Goal–Things that
people are trying to
MIC-COMPONENT accomplish

DETERMINATION OF CA’S • Feeling–Emotions felt

NEW: EINSTEIN OLD: Discovery of relativity

SEMANTIC MEMORY -
Ground

GRAMMATICAL: It cleft construction Generic knowledge: Law of

relativity
MORPHOLOGICAL
Scripts

PHONOLOGICAL FORMALIZER Encyclopedic knowledge

“It is ...
LINGUISTIC SEMANTICS
Lexicon etc

VERBAL OUTPUT (ORAL AND WRITTEN)

“It is Einstein who discovered the law of relativity”

Figure 1.  Embedding of MIC-Component in production of utterance

 Franz J. Stachowiak

­ verlearned information. One of the three narratives was a modified version of the
o
well known (overlearned) fairy tale “Little red riding hood”. In this version Little
Red Riding Hood is not eaten by the wolf. The new version was as follows:

“Once upon a time/there was a little girl/called/Little Red Riding Hood/because of

her little/red hood./On her way/to visit her grandmother,/Little Red Riding Hood/
found/a nice/wolf cub/lost/in the forest./Little Red Riding Hood/decided/to take the
wolf cub/ to her grandmother’s/to feed it./Very scared,/the grandmother/implored/
Little Red Riding Hood/to put the animal back/in its den./Little Red Riding Hood/
was very sad/when she left/her grandmother’s house/with the wolf cub,/but then,
fortunately,/in a clearing/of the forest/she found/the mother wolf.” (ib. p. 660)

As expected old age subjects who had to recall this story produced more con-
fabulations in this modified story than in a completely new everyday story. The
confabulations consisted in saying that “the wolf eats Little Red Riding Hood,”
as it is known from the original version of the story, rather than recalling the
modified version in which Little Red Riding Hood is not eaten by the wolf.”
(ib. P. 658). In the above model the confabulatory recall could be explained as
due to vitiated attention processes related to the distinction between new and
old information. The grounding of the activity of the MIC-component would
be compromised by the uninhibited retrieval of old information, which should
have a structural effect on the distribution of information in the stories as they
were retold by the subject.
Another way of gathering evidence for the relationship between types of
memory and the meta-informative status of the information retrieved from mem-
ory in a certain narrative task has been attempted in a small pilot study conducted
for this presentation: Eight 3rd year students were asked to give a short report
about their memories of the first day at Charles Sturt University.
Task: Please write a report of your first visit to Charles Sturt University.
Describe this event with the details you remember. Describe in a few sentences how
memory influences your verbal report, for instance if you can’t remember details.
The responses were evaluated using the organizing categories of episodic mem-
ory described above (Reeves, Kuper & Hodges 2008). The distribution of infor-
mation related to these categories was classified as new or old (­meta-informative
status) and further sub-classified according to sentence initial or final position.
For the sake of consistency and simplicity the same categories were assumed for
semantic-generic information and applied according to the same scheme. The
centres of attention (or the relevant parts of information) in the different sen-
tential positions were identified and cross-classified according to the categories
mentioned (see Figure 3). The following example (Figure 2) shows the attempts
to classify the different phrases. The scheme used (Figure 2) obviously needs
 Semantic and episodic memory w.r.t. old and new meta-informative status 

more elaboration and more categories, in particular categories related to the

­features of objects etc.
A full study on these lines would definitely need strict criteria as to the appli-
cation of the categories used, so that the analysis would guarantee some degree of
intersubjective reliability. My first attempt yielded, however, some results.

Figure 2.  Example of student report with classification remarks between lines

Episodic information Semantic-generic information

New status Old status New status Old status
New New Old Old New New Old Old
initital final initial final initial final initial final

Space–Physical layout of the place(s) 2 13 11 7 1 1

Actor–Range of people involved 6 4 26 5 3
Activity–A set of related activities that occur
Object–The physical things that are present 4 7 1 9
Act–Singleactionspeople undertake 22
Event–Activities that people carry out 4 10 3 2 1
Time–The sequencing of events that occur 1 1 3
Goal–Things that people are trying to
3 1
accomplish
Feeling–Emotions felt and expressed 4 16 1
SUM 21 76 29 19 23 1 1

Figure 3.  Frequency of informational categories (episodic vs. semantic-generic)

in relationship to their meta-informative status and their position in 8 student reports
about their first day at University
 Franz J. Stachowiak

This simple count demonstrates that in the specific task, which asked for
information from episodic memory, most of the relevant new information appears
in sentence final position (the expected place for new information). This applies
to all categories used. But it is obvious that the categories “Act – Single Actions
people undertake”, “Events – activities that people carry out” and “Feeling – emo-
tions felt and expressed” relate to information that is expressed by the verb phrase.
The category “Activitiy – a set of related activities that occur” did not apply as no
activities were reported that were not directly related to the main event, the first
day at the University. Interestingly information from all other categories could
appear in this position, but also in sentence initial position, where it seems to get
the status of focused attention. The actor “usually the first person “I”, which was
considered as old information, appeared most frequently in initial subject posi-
tion (26 instances), as can – of course – be expected. Generic information gets the
status of new information in several instances and relates new information to old
one as described above.
These data are, of course, too sparse to arrive at sound conclusions, but they
seem to indicate that the mapping between information stored in episodic and
semantic memory on the one hand and centres of attention on the other as a result
of the meta-informative processing of utterances – is guided by structural prin-
ciples. There is no 1:1 mapping of memory entries into fixed positions in the meta-
informative structuring of utterances, in contrast, there is a lot of flexibility. In
FWO languages such as Jiwarli1 (Austin 2001) any kind of information from the
above categories can be moved into initial position as centre of attention. However
some types of information seem to lend themselves to predilected positions on the
way from intentions into utterances.

References

Abusamra, Valeria, Côté, Claude H., Joanette, Yves & Ferreres, Aldo. 2009. Communication
impairments in patients with right hemisphere damage. Life Span and Disability XII(1):
67–82.
Austin, Peter K. 2001. Word order in a free word order language: The case of Jiwarli. In Forty
Years On: Ken Hale and Australian Languages, Jane Simpson, David Nash, Mary Laughren,
Peter Austin & Barry Alpher (eds), 205–323. Canberra: Pacific Linguistics.
Battaglia, Francesco P. & Pennartz, Cyriel M.A. 2011. The construction of semantic memory:
Grammar-based representations learned from relational episodic information. Frontiers in
Computational Neuroscience 5: 1–22.

.  Jiwarli is an Aboriginal language traditionally spoken in the north-west of Western

­Australia, inland from the town of Carnarvon.
 Semantic and episodic memory w.r.t. old and new meta-informative status 

Dalla Barba, Gianfranco, Attali, Eve & La Corte, Valentina. 2010. Confabulation in healthy
aging is related to interference of overlearned, semantically similar information on epi-
sodic memory recall. Journal of Clinical and Experimental Neuropsychology 32(6): 655–660.
Dennet, Daniel. 1977. The Intential Stance, 6th edn. Cambridge MA: The MIT press.
Dickerson, Bradford C. & Eichenbaum, Howard. 2010. The episodic memory system: Neurocir-
cuitry and disorders. Neuropsychopharmacology Reviews 35: 86–104.
Gallese, Vittorio. 2001. The “Shared Manifold” hypothesis: From mirror neurons to empathy.
Journal of Consciousness Studies 8(5–7): 33–50.
Greve, Andrea, Van Rossum, Marc C.W. & Donaldson, David I. 2007. Investigating the func-
tional interaction between semantic and episodic memory: Convergent behavioral and
electrophysiological evidence for the role of familiarity. NeuroImage 34: 801–814.
Kensinger, Elizabeth A., Ullman, Michael T. & Corkin, Suzanne. 2001. Bilateral medial temporal
lobe damage does not affect lexical or grammatical processing: Evidence from amnesic
patient H.M. Hippocampus 11: 337–346.
Klein, Stanley B., Cosmides, Leda, Gangi, Cynthia E., Jackson, Betsy, Tooby, John & Costabile,
Kristi A. 2009. Evolution and episodic memory: An analysis and demonstration of a social
function of episodic recollection. Social Cognition 27: 283–319.
Levelt, Willem J.M. 1989. Speaking. From Intention to Articulation. Cambridge MA: The MIT
Press.
Miller, George A. & Johnson-Laird, Philip N. 1976. Language and Perception. Harvard Univer-
sity Press. Cambridge, Mass.
Mitchell, Rachel L.C. & Crow, Tim J. 2005. Right hemisphere language functions and schizo-
phrenia: The forgotten hemisphere? Brain 128: 963–978.
Nuxoll, Andrew M. & Laird, John E. 2007. Extending Cognitive Architecture with Episodic
Memory. In the Proceedings of the 21st National Conference on Artificial Intelligence (AAAI).
Prince, Steven E., Tsukiura, Takashi & Cabeza, Roberto E. 2007. Distinguishing the neural
­correlates of episodic memory encoding and semantic memory retrieval. Psychological
­Science 18(2): 144–151.
Reeves, Scott, Kuper, Ayelet & Hodges, Brian D. 2008. Qualitative Research Methodologies:
­Ethnography. BMJ 337:a1020.
Scoville William B. & Milner, Brenda. 1957. Loss of recent memory after bilateral hippocampal
lesions. Journal of Neurology, Neurosurgery & Psychiatry 20: 11–21.
Thompson, Ian. 2001. Discourse analysis in neuropsychiatry. Communication and Mental
­Illness, Jenny France & Sarah Kramer (eds), 393–406. London: Jessica Kingsley Publishers.
Tomasello, M. 1999. The Cultural Origins of Human Cognition, Harvard University Press. Cam-
bridge, Mass.
Tulving, Endel. 1972. Episodic and semantic memory. In Organization of Memory, Endel
­Tulving & Wayne Donaldson (eds), 381–403. New York NY: Academic Press.
Tulving, Endel. 2002. Episodic memory: From mind to brain. Annual Review of Psychology
53: 1–25.
 Tracing the role of memory and attention for
the meta-informative validation of utterances

Franz J. Stachowiak
Charles Sturt University Albury, NSW, Australia

This chapter is an attempt to establish a connection between some aspects of

the MIC theory and neuropsychological findings about memory and attention.
It is argued that verbal processing of information is strongly influenced by or
even rests on the capacity and mode of operation of working memory and other
types of memory and is intricately related to attentional processes, which play a
role in directing the interest of a communication partner in spoken or written
language. In fact the author assumes that the basic slot structure for marking
the information as new or old in verbal utterances derives from operational
principles, limits and strategies (to overcome these limits) of working memory
as a gate to long term memory.

1.  Introduction

In the following I will try to establish a connection between some aspects of the
MIC theory and neuropsychological findings about memory and attention. I will
argue that verbal processing of information is strongly influenced by or even rests
on the capacity and mode of operation of working memory and other types of
memory and is intricately related to attentional processes, which play a role in
directing the interest of a communication partner in spoken or written language.
In fact I assume that the basic slot structure for marking the information as new or
old in verbal utterances derives from operational principles, limits and strategies
(to overcome these limits) of working memory as a gate to long term memory. In
particular I will support the following claim by Hélène and André Włodarczyk
(MIC 2010 conference paper):
We claim that it is impossible to produce a (pragmatically) felicitous utterance
without introducing the information it conveys as either old or new, this
distinction being expressed as much by the prosody and the syntactic structure
of the utterance as by the choice of its grammatical and lexical units. As a matter
of fact, the meta-informative (pragmatic) validation of an utterance as old or new
is different from its informative (semantic) validation as true or false.
 Franz J. Stachowiak

In a seminal article, in which they examine the role of memory in complex cogni-
tive tasks, Ericsson and Kintsch (1995) have presented a model of text compre-
hension, in which they assume an interaction of two kinds of processes: the short
term activation of information while text segments are being processed and the
storage and retrieval of information that forms episodic text structures in long
term memory. New information from the text is continually integrated into these
episodic text structures, but relevant parts of them need to remain accessible dur-
ing reading or auditive comprehension. Ericsson and Kintsch claim that the acces-
sible portions of this structure in long term memory serve as extended working
memory, called “long term – working memory”. In order to overcome the limits
of working memory capacity, when massive rapid information flow takes place,
we need to hold only a few relative parts of the information in working memory,
which serve as cues to retrieve the entire information and expand the episodic text
structures. Sometimes texts are not expressed in such a way that understanding
is smooth and easy. Therefore “items from the episodic text memory under con-
struction must be reinstated in the focus of attention to ensure the coherence of the
memory representation”. (Ericsson & Kintsch 1995: 30). Various devices are used to
achieve this such as “anaphoric and cataphoric elements in the episodic model of the
current text, generic lexical knowledge, as well as contextual features”:
In fact, language possesses a variety of graded syntactic devices to indicate to the
processor just where in the text information that is to be reactivated is located.
These are not only anaphoric cues but also cataphoric cues (e.g. referents marked
with an indefinite “this” will recur as a central concept in the subsequent text).
Syntax also instructs the reader when not to look for prior links but to start
a new thematic unit by means of a variety of switching devices. For example,
a plain “and” signals continuity (only 16% of the occurrences of “and” were
associated with switches in topic in Givón’s corpus), whereas an “And” following
a period signals the beginning of a new thematic unit (100% switches). Thus,
the cues present in a reader’s focus of attention not only make possible retrieval
from LT-WM but also indicate to the reader when to attempt such retrieval and
when not to. (Ericsson & Kintsch 1995: 30)

Although due reference is made to these syntactic devices Ericsson and Kintsch
pursue a psychological line of argument bypassing the detailed linguistic dis-
cussion with respect to universally relevant notions such as topic – comment
structure and the different pragmatic strategies of marking information as new
or old.
In the following I will argue that a detailed linguistic analysis of the infor-
mational structure of texts, as proposed in the MIC theory, is a necessary pre-
condition for the further elaboration also of the approach taken by Ericsson and
Kintsch. In fact, it can be shown that a clear distinction between semantic and
 Memory and attention in MIC 

pragmatic aspects of the content side of the information structure and their delin-
eation from “syntagmatic and paradigmatic aspects of form” can lead back to basic
neuropsychological foundations that have not been identified in Ericsson’s and
Kintsch’s paper.

2.  Some examples from German as a FWO language

German is considered a Free Word Order Language (FWL). The first Example (1)
shows that German seems to allow the rather free movement of the constituents
of a sentence into or towards positions in which they strike the attention of the
recipient. My notion of CA thus implies not only the functional aspect of bringing
something to someone’s attention, by marking it as “interesting, new, important,
surprising” (for instance by primary or secondary stress in almost any position in
the sentence) but also the local aspect of finding the optimal place in a linear order,
which by its mere position supports the recipient of the utterance in processing the
information. The classical topic-comment distinction is covered by my consider-
ations. This distinction is treated in the MIC theory as the distinction between
chunks of old and new information to which the relevant feature of prominence
is added. A chunk of information which is both old and prominent is called a
topic, whereas a chunk of information which is both new and prominent is called
a focus. Non-prominent old and new chunks of information are called respectively
the comment and the background.
Let us consider the German sentence (1)
Mein Mann kann morgens gut Kaffee kochen.
My husband can in the morning well coffee cook.
msgposs nom 3sgprs adv adv n acc inf

This sentence exhibits the default word order of a sentence with the finite verb
“kann” in second position (Gerdes & Kahane 2006). At least 20 different accept-
able sentences can be formed from the same set of words and constituents. These
different possibilities can be explained in the MIC framework as different prag-
matic (meta-informative) structuring of the same propositional information,
expressed as a base utterance or extended utterance (Włodarczyk & Włodarczyk
2008). A base utterance is entirely new or old; it contains only a subject (and even-
tually an object) as centre(s) of attention. An extended utterance is characterised
by the opposition (1) between old information (background) and new information
(focus) or (2) between new information (comment) and old information (topic).
Some extended utterances contain both a topic and a focus; in such cases, the
topic comes at the beginning and the focus may follow it immediately or appear
 Franz J. Stachowiak

as a stressed constituent in any place in the utterance. Hereafter, I propose the

interpretation of a few possible changes in word order of sentence (1). However,
one must be aware that, in order to interpret properly the pragmatic structure of
an utterance, a broader context is needed.
Hereafter are only a few examples:
(1.1) Kann mein Mann morgens gut Kaffee kochen? (question)
Can my husband in the morning well coffee cook?
We have here the default finite verb first position of the German interroga-
tive sentence.
(1.2) Morgens kann mein Mann gut Kaffee kochen.
(correct answer, if morgens had been stressed in the question)
This statement might be (depending on the context) either an entirely new
information or a statement with a focus on Morgens.
(1.3) Morgens gut Kaffee kochen kann mein Mann.
We can interpret this as a focus on Mein Mann.
(1.4) Morgens Kaffee kochen kann mein Mann gut.
We can interpret this as a topic on „Morgens Kaffee kochen“ and a focus
on gut. The part kann mein Mann is new but not stressed information:
it is the comment.
(1.5) Gut Kaffee kochen kann mein Mann mórgens. (but probably not in
the ­evening)
We can interpret this as a topic on Gut kaffee kochen and a focus on
­Morgens.
(1.6) Gut kochen kann mein Mann Kaffee morgens.
We can understand this as a topic on gut kochen and a focus on morgens.
Gut kann mein Mann morgens Kaffee kochen. (Confirmation: he can do
(1.7) 
it well)
This may be a topicalisation of gut.
(1.8) Kann morgens gut Kaffee kochen, mein Mann. (positive remark about
­husband)
Mein Mann is the postposed topic.
etc.

In addition to moving the constituents almost every word can be accentuated,

which leads to different interpretations. I have indicated some of them after the
sentences. If I stress, for instance, mórgens in (1.5) it means that the husband can
do it well in the morning but probably not at another time.
One can see that the subject is usually moved towards the middle or the end of
the sentences in all non-default versions. In (1.3) and (1.8) the subject is moved to
 Memory and attention in MIC 

the very far end of a sentence and gets particular attention there either as a focus
or a postposed topic.
However, very few decisions, in fact one or two fixed positions and some
inflections can considerably reduce the number of possible word orders in
­German, even down to only a single one. If some morpho-syntactic processing is
under way and a few word positions have been determined, the pragmatic free-
dom of arranging the linear order of the words diminishes or disappears. Thus if
a sentence has been started with “Gut Kaffee…” as in (1.5), only one particular
linear word order can follow. A similar phenomenon seems to occur in Russian, as
was pointed out by Myachykov and Tomlin (2008) in a study on the relationship
between perceptual priming and structural choice in Russian sentence produc-
tion. Despite the morphological richness of Russian and its rather free word order,
which allows speakers to preferably assign the sentential starting point, but not the
subject, to the perceptually primed referent (in a visual task) alternating between
the agent-initial and the patient-initial structural alternatives, Russian speakers
were more reluctant than speakers of English (despite the narrower inventory of
structural options in English) to alternate structure as a function of perceptual
prime. “This tendency may result from the necessity to maintain early commit-
ments to the case-marked noun forms, which effectively binds structural selection
to a much smaller number of available alternatives than the normative grammar
of Russian suggests” (Myachykov & Tomlin 2008: 1). This confirms the position
adopted in the MIC Theory that in language production the pragmatic decisions
come before the grammatical ones, but it is also evident that the grammatical ones
restrict the pragmatic ones. In language comprehension automatic grammatical
parsing comes first and is resolved in very short time – between 300–500 ms after
hearing the sentence (Friederici & Jacobson 1999; Friederici, Hahne; Saddy 2002).
Thus unexpected deviations from canonical word order or violations of phrase
structure are detected rather early and checked with respect to semantic and prag-
matic functions some 300 msec. later. I assume that on a scale between conscious-
ness and automaticity, pragmatic decisions are rather more conscious, whereas the
grammatical ones are clearly more automatic.
It must be emphasized that in the MIC theory, the basic syntactic structure
(represented as NP-VP in a constituency tree) corresponds to the base (entirely
new or old) utterance schema: a subject and what is predicated about it, i.e. a verb
with eventually an object. This syntactic structure reflects directly the pragmatic
structuring of a base utterance which consists in the choice of a global centre of
attention, i.e. a subject, and eventually a local one, the object. This makes it pos-
sible for speakers to produce and hearers to interpret base utterances very quickly,
even before they analyse the situation on the deep semantic level, i.e. even before
they understand clearly, which participant plays an active role and which a passive
 Franz J. Stachowiak

one. On the other hand, it seems probable that extended utterances (consisting
of two different old and new information chunks) require more conscious and
deliberate pragmatic decisions which result in non-canonical word order. Further
psycholinguistic testing is necessary to corroborate this hypothesis.
Obviously the initial and final fields in the topological model of the German
sentence (cf. Zifonun et al. 1997) are important slots for the strategic position-
ing of the centers of attention in distributing the information that one wants to
convey in the most effective and successful way. Apparently new and given infor-
mation can appear in both locations. It depends on how one presents them meta-
informatively and how one wants to locate them with regard to the preceding and
subsequent information.
The languages of the world apply a whole range of different grammatical
devices to mark the centers of attention or the highlights of information (­Gundel
et al. 2004). Constructions such as topic and focus fronting, cleft sentences, dis-
location to the periphery, topic markers (as in Japanese and Korean), sentence-
linking features such as ellipsis, anaphora, concord, word repetition, pro-forms,
adverbial contrasts and, last but not least, prosodic features are applied, but con-
sidering the diversity of all these devices it becomes apparent that there is no one
to one mapping between these techniques and pragmatic functions or subfunc-
tions of manipulating the distribution of information in the chain of speech.
Of course, it need not be elaborated here that – if we take the sentence or an
utterance as a cut out section of a longer text or stretch of speech – the first element
of this section, prototypically refers backwards or hooks up to what has been men-
tioned (“anaphoric grounding”), what is already known (“discourse grounding”)
or what is referred to (“ontological referential grounding”) in a more general sense
(see Table 1 below). The final element of the cut out section prototypically refers
to what is to follow in the text or communication (“cataphoric grounding”), to
information not yet known (“discourse grounding”) or it refers to specific, particu-
lar, actual or occasional aspects of ontological knowledge as stored in long term
memory. On the phonetic level this backward and forward processing is paralleled
by the way in which the realization of speech sounds is influenced by the preced-
ing and following sounds (Coarticulation).
Similarly – although on a higher cognitive level – the flow of information needs
to be handled in such a way that our “cognitive apparatus” can deal with it effec-
tively. André and Hélène Włodarczyk have described this as a meta-­informative
process, depicted in their model of the grounding of the old or new status of infor-
mation (see Table 1) presented in the meta-informative mode. This corresponds
quite well with approaches in cognitive neuropsychology, and I would like to com-
plement their scheme by some further distinctions. Their notions of intermediate
and permanent memory point clearly to the fact that the marking of information
 Memory and attention in MIC 

Table 1.  Grounding of the meta-informative old or new status, according

to the MIC Theory (cf. Włodarczyk A. Chapter 3 in this volume)

Memory Type of Grounding of Grounding of

information old status new status
and knowledge

Immediate Communicative anaphoric cataphoric

(Adjacent
utterances)

Intermediate Cognitive known unknown

(Recently
acquired
information)

Permanent Ontological generic specific

general particular
(Mental Σ Π
potential actual
reference) habitual occasional

as new or old involves online processes as well as the retrieval of stored memory
contents from the different facets of memory. I think that it is important to relate
this psycholinguistic aspect to the structural aspects which play a role in the lin-
earization of speech. Before going into the details of neuropsychological research
on memory and attention I would like to briefly discuss some basic aspects of the
processing of old and new information and the role of human memory in this.

3.  New and old information from a neurobiological perspective

Like other animals we are continuously confronted with an enormous flow of infor-
mation via the different senses such as hearing, sight, touch, smell, taste etc. With-
out the possibility of encoding and storing information in memory and retrieving
it in order to match old and new information and thus to learn from experience,
coping with the environment would be much more difficult. Of course, relevant
information for the survival of a species is coded in genetic memory, but in many
species, especially in primates and in particular in humans, different brain based
memory systems are operating in order to collect, evaluate and mark new relevant
information for social and cognitive behaviors and decisions.
Ernst Pöppel (1998), who has done pioneering work on the temporal struc-
ture of perception and cognitive processing found that the flow of information
 Franz J. Stachowiak

input can be handled only by allowing it within certain time windows, which form
oscillating frames for parcelling the information into processable bits: “Humans
integrate information in a time window of approx. 3 seconds and generate a state
of awareness that can be referrred to as “subjective present”. This applies also to the
processing of language:
”Erkennbar wird unser “Gegenwartsfenster” beim Sprechen. Unsere Sprache ist
rhythmisch gegliedert, wobei Satzaussagen – im Übrigen unabhängig von der
Sprache und auch vom Alter – nur wenige Sekunden dauern, jeweils getrennt durch
Planungspausen für die folgende Aussage“. (Pöppel 1998)
“The window of our subjective present becomes recognizable in speaking. Our
language is structured rhythmically, and predications (propositions) last –
independently of language and the speaker’s age – only a few seconds, each of
them separated by planning pauses for the next statement”.

4.  H
 uman memory as a complex differentiated system and its relation
to text comprehension

Human memory is not a unitary system. Depending on the duration of the reten-
tion of information and the storage capacity, various kinds of memory must be
differentiated (see Figure 1). Sensory or iconic memory retains information only for
the blink of an eye, a very short moment. Short term memory allows for memory
in the range of up to 15 seconds and passes information on to long term mem-
ory where it can possibly be stored for a whole life. Modern neuropsychological
research deals with the so called working memory (Cowan 1999, 2005), a variant
model of short term memory. It is understood as a system that temporarily (in
the range of 1–3 seconds ) retains and manipulates information during cogni-
tive tasks such as learning, understanding, reasoning or calculation operations.
Working memory stores auditive information, for instance when comprehending
a verbal utterance, but also visual information, for instance when making deci-
sions in a complicated traffic situation. Working memory is very closely related
to attentional processes (Cowan et al. 1999). It is directly relevant to the notion of
center of attention in the MIC theory.
In long term memory information is often retained throughout a whole life.
It classifies and integrates information in a system of concepts. This pertains to
information, which is relevant for our personal biography (episodic memory)
such as occurrences, experiences, careers, encounters, accidents etc., but also to
information which is indispensable for our general knowledge, our education and
intelligent behavior (semantic memory). Semantic memory comprises knowledge,
as for instance the fact that I know that Paris is the capital of France. Episodic
 Memory and attention in MIC 

MEMORY

Long-term memory Short-term memory

Sensory memory
Short-term/workingmemory

Declarative memory Non declarative memory

(explicit memory) (implicit memory)

Events Facts Procedural Perceptual Classical Non associative

(episodic memory) (semantic memory) memory representation conditioning learning
system

Specific personal World knowledge, Conditioned

Skills Perceptual Habituation
experiences from object knowledge, response
(motor and priming sensitization
a particular language knowledge, between two
cognitive)
time and place conceptual priming stimuli

Medial temporal lobe Basal ganglia Perceptual Skeletal Reflex

Middle diencephalon and cerebellum and association muscle pathways
Neocortax, especially neocortex
the prefrontal cortex

Figure 1. Long-term and short-term memory Acc. to 〈http://huehueteotl.wordpress.com/2009/08/〉

 Franz J. Stachowiak

memory refers for instance to my memories of my very first stay in Paris when
I was 19 years old. A distinction has also to be made between explicit and implicit
memory. The former, also called declarative knowledge, signifies entries in long
term memory, which can be recalled purposely and consciously, as for instance a
mathematical formula such as a² + b² = c². In contrast, the results of procedural
learning, as for instance one’s competence to ride a bicycle or peel a potato, are
considered as implicit knowledge.
Semantic memory implies explicit knowledge. It is closely related to language
and word meanings. Our vocabulary is part of semantic memory. That a poodle
is a dog and that a dog is a mammal is part of my world knowledge, which I can
retrieve deliberately and “declare” as a matter of fact, but it is also part of my concep-
tual and notational knowledge. In the same way general knowledge, as for instance
expressed in “take-home-messages such as “The year has 12 months”, is considered
as part of semantic memory. Semantic memory exceeds episodic memory, as it
is beyond the scope of autobiographical memory contents. In our brain seman-
tic memory is organized as a conceptual network. However, semantic memory
interacts with episodic memory or is even extracted as a kind of yield or upshot
from episodes in life. It is quite obvious that verbal utterances frequently refer to
contents of the different kinds of memories in order to give them – depending
on the context – the status of new or old information. The opposition between Σ
(types) and Π (occurences) in the ontological grounding of the meta-informative
old or new status (Włodarczyk A. Chapter 3 in the present volume), could there-
fore be related directly to these forms of long term memory. If the distribution of
new and old information in an utterance serves the purpose of matching old and
new memory entries and of learning, it can be assumed that relevant information
is passed on into episodic memory and from there to semantic memory, while at
the same time information from semantic memory is fed into working memory in
order to decode a message semantically.
Of central importance to the MIC theory is the notion of “working memory”.
The best known model was developed by Baddeley and Hitch (1974), which has
attracted an immense amount of neuropsychological research.
Working memory refers to our capacity to simultaneously process and store
information. According to Baddeley’s model (Figure 2), working memory con-
sists of several subcomponents which play an active role in the process of holding
new information while processing or changing it in the visual and the verbal sys-
tem (for instance in text production and and comprehension). The central execu-
tive serves the attentional control of information that is actively dealt with in the
visual or verbal subsystems and it controls the transfer of information from and
to long term memory. In other words, this component is a good candidate for
a mechanism that is engaged in pragmatic decisions on the meta-informative
 Memory and attention in MIC 

Central
executive

Visuo-spatial Episodic Phonological

sketch-pad buffer loop

Visual Episodic Language

semantics LTM

Fluid systems Crystalized systems

Figure 2.  Working memory acc. to Baddeley’s three-component model

of working memory (2003)

status of information as organized in utterances. The phonological loop allows

a speaker/hearer to rehearse phonological information like a telephone number
that was just heard, and the visual sketchpad has a parallel function for incoming
visual information as for instance in a picture story.
The episodic buffer was introduced by Baddeley in 2000. It is meant as a tem-
porary multimodal store, which integrates information from long term memory
with information from the visual sketchpad and the phonological loop, in order
to create short term representations of coherent events as for instance in narra-
tives. In so far as the metainformative organization of an utterance requires active
maintenance of information in the face of ongoing morphosyntactic and seman-
tic decisions and processes, this model offers some psycholinguistic explanations
for phenomena that are difficult to understand if analyzed solely in classical lin-
guistic terms. The situational model, introduced by Ericsson and Kintsch (1995),
which integrates textual information with background knowledge (as for example
in politeness systems) would alo be subserved by the episodic buffer. The box
model (Table 1 above), referring to the types of grounding of metainformative
information in memory, offers a good basis for further elaboration with respect
to the interplay, i.e. feedback and feed forward of information, between what is in
the text and what is already in the speaker’s or hearer’s head or what comes from
the environment and alters their state of mind during language processing. Thus
it must be assumed that a very complex system of dynamic processes involving
memory, attention and verbal processing (“the language system” in the narrower
sense) interact so that we need some arrows between the boxes to indicate the
 Franz J. Stachowiak

direction of information flow. This interconnected functional complexity is also

the reason why fMRI1 studies of language processing exhibit the activation of mul-
tiple connected areas in the brain and why it is so difficult to delineate the activa-
tion of individual levels of processing (syntactic, semantic, pragmatic) in ERP2
studies investigating the time windows (P150, 350 or P600), in which each level or
phase of processing takes place.

5.  W
 orking memory capacity and processes from the basis
of the meta-informative processing of linguistic utterances

How much information can be held in working memory? In a very careful lit-
erature review, in which he brought together a wide variety of data on capac-
ity limits, Nelson Cowan (2001) came to the conclusion that a single, central
capacity limit between three and four unrelated chunks, which can be attended
at one time, characterizes human short term memory. Chunks are understood
as a collection of concepts that have strong associations to one another and
much weaker associations to other chunks currently in use. With respect to
language processing this implies that on average not more than four centers of
attention, about which the speaker intends to predicate, can be processed at the
same time.
The objective of many studies on the limits of working memory was to obtain
data, which allow working memory span to be used as a measure to predict the
performance of tested individuals in a whole range of cognitive tasks such as
reading comprehension, text recall, reasoning and problem solving. The following
­figure (Figure 3) shows the decline in performance depending on the ­number of
intervening items. After 12 intervening items the retention rate declines towards
zero.
Hence, one can conclude that the interference from subsequent items causes
the deletion of information in short term memory. It is also evident that this study
falls short of accounting for memory processing of complex information in cog-
nitively higher demanding tasks. Scanning a brief list of digits contrasts sharply
with a task in which the information presented has some relationship to other

.  Functional magnetic resonance imaging

.  ERPs (event related potentials) are “electrophysiological” responses of the brain to
experimental stimuli such as syntactically or semantically deviant sentences to which
the subject responds in a typical time frame Thus ERPs provide a reliable measure of the
­sequence and timing of the brain (here: the language processing areas) when processing
linguistic information.
 Memory and attention in MIC 

1
0,9
0,8
0,7
Proportion correct

0,6
1 digit/s
0,5
4 digit/s
0,4
0,3
0,2
0,1
0
1 2 3 4 5 6 7 8 9 10 11 12
Number of intervening items

Figure 3.  Proportion of correct recall of items as a function of the number of intervening
items and presentation rate. Presentation rate was either one digit per second or four digits
per second (Acc. to Waugh & Norman 1965)

information which might be available to a person within a communicative setting

or a situational context or which he might have stored in long term memory as
a result of learning and experience. With regard to text comprehension it can be
assumed that with the mass of information exchanged in verbal communication
an exhaustive serial search of all items of a given linguistic surface structure would
prevent the decoding of the message rather than sharpen it. Memory decay would
be greatest in exhaustive serial search processes in sequences of many items such
as long texts.
Considering these results together, it becomes understandable why centers
of attention within a text are used as boot straps against the decay of information
on the one hand and against inefficient unterminated scanning processes on the
other.
At the same time grammar appears to be a formal ordering system to generate
chunks and relate them to each other in order to facilitate information processing.
The dynamic memory processes discussed so far can be depicted in the fol-
lowing model (Figure 4). The relationship between working memory and long
term memory as far as it is relevant in text processing can be understood as a
transfer process in which incoming information is processed in working memory
and partly transferred to long term memory (Ericsson and Kintsch use the term
“Long-term working memory”; A. and H. Włodarczyk prefer the term “Intermedi-
ate memory”) in order to form episodic text structures, and as a retrieval process
from long term memory with regard to the meaning of lexical items and episodic
and semantic memory in general.
 Franz J. Stachowiak

Rehearsal
Long-term
Information, Short term & memory
Input, working Transfer
Text memory Episodic
text
structures

Forgetting,
Decay

Figure 4.  Dynamic memory processes during the processing of information

This model does not explicitly specify a component that attributes relevance
or importance to the information that is being processed. The Central Executive
Component in Baddeley’s working memory model would, however, be a candidate
for performing the task of meta-informative centering in the sense of the MIC
Theory (see further elaboration below).
One of the best known psychological findings to which the linguistic distinc-
tion between old and new information in discourse can be related is the serial
position effect. This refers to the fact that in free list recall – i.e. when subjects
are asked to recall the items of a list of digits, words or other symbols in free
order – the accuracy of the recall depends on the position of the items in the list.
In immediate recall the last items of a list are usually better remembered than
the first ones (Recency Effect). The first items of the list are better recalled than
the middle ones (Primacy Effect). An explanation of the recency effect is that in
short-term memory the processing of new items overwrites old ones. So at the
time of recall the last ones have not been overwritten and are still held active in
working memory. The primacy effect is explained with the hypothesis that initial
items are more readily encoded into long-term memory because they could be
rehearsed before new items came to be processed. Thus in recall the first items
are retrieved from long term memory. The first item in a list can also be processed
by itself, whereas the following ones have to be incrementally dealt with together
with the subsequent items. The explanations are supported by the finding that
the recency effect is eliminated when an interfering task is given before recall. If
for instance an arithmetical task is to be solved before the list items are recalled,
the recency effect disappears, as the new task requires attention and processing
capacity. Fast presentation of the items reduces the primacy effect. Later recall
increases it.
The following figure (Figure 5) shows the serial position effect.
 Memory and attention in MIC 

100

Words recalled (%)

50

0
Primacy Intermediate Recency
Position in sequence

Figure 5.  Serial position effect. The percentage of items recalled as a function of their
position in the list (Based on Murdock 1962)

If the explanations given above are correct, the distinction between primacy
and recency effect emphasizes the concept of transfer between working memory
and long term memory. Further findings add more details. Forward serial recall
generates a stronger primacy effect, whereas backward recall is characterized by
an increased recency effect. Although comprehending a verbal text is a much more
complex task, backward recall in list learning experiments resembles to some
degree the recall of segments of text that have already passed through working
memory but need to be checked again in order to fully understand the nature
of information to be conveyed. This could be the process of scanning backwards
from the comment towards the topic in a sentence. It is also conceivable that rel-
evant information is shifted to the end of a sentence in order to profit from the
recency effect especially if too much information had to be processed in the pre-
ceding paragraphs. Oberauer (2003: 471) holds that “one mechanism for generating
both primacy and recency effects is retroactive interference. Retroactive interference
during encoding (input interference) naturally produces a recency effect … Retroac-
tive interference during a series of outputs (output interference), on the other hand,
provides an advantage for the items retrieved first, because the early items interfere
with the later items in the output sequence”.
Although these studies do not investigate meta-informative strategies in sen-
tence and text processing directly – a task that calls for more linguistically based
research – it appears that some basic neuropsychological preconditions for dealing
with new and old information in texts can be traced in these results on working
memory capacity. The limitation of working memory capacity is a factor in a cog-
nitively demanding task such as text production and comprehension. It imposes
the necessity to select centers of attention as bootstrapping devices to cope with
 Franz J. Stachowiak

the workload of simultaneously storing and manipulating bits of information and

arranging the linear surface structure in such a way that communication partners
are supported in decoding the gist of the intended message. In this sense the skil-
ful filling of the slots of a sentence’s topological field structure, as for instance in
sentence 2, is part of necessary and universal properties of language induced ulti-
mately by human biology and cognitive capabilities.

6.  Focussed attention on centers of informational saliency in texts

In the preceding chapters it has already become apparent that attentional pro-
cesses are an integral part of working memory functions. Tasks that engage work-
ing memory at short notice, require attentional behaviour in the form of alertness,
focussed attention and also divided attention. Tasks that require divided attention
implicate at least two sources of attraction or stimuli that need to be regarded
simultaneously or sequentially in a fast alternation of the focus of attention.
The traditional models of attention (Broadbent 1958) assumed that only one
piece of information at a time would pass a perceptual filter and then be processed
semantically. The most critical part of this model was that it did not allow for
parallel processing of stimuli. Problems in the older models were overcome by
the introduction of a pertinence factor by Norman (1968). Pertinence refers to
the level of significance or salience, which the incoming information has for the
receiver. Certain kinds of inputs, such as the sound of our name, can be personally
important to us and are attributed the status of high level pertinence, others will
have changing levels of pertinence depending on the requirements and estimation
of ongoing information flow in a given context. Figure 6 shows the components of
Norman’s model and their interaction.

Sensory inputs

Processing
Storage

Pertinence

Selection

Attention

Figure 6.  Selection and attention Acc. to Norman’s model (1968)

 Memory and attention in MIC 

This means that pertinence can be given to a percept in such a way that it
catches the attention of the receiver. Relevant chunks of information in human
language need to be marked in such a way as to catch the attention of the commu-
nication partner. Word order change and intonation may be primary instruments,
but equally and additionally effective are “adverbial” lexical focus markers such as
“eben” (“just”), “gerade” (“exactly”), “insbesondere” (“especially”) which can even
be used cumulatively in order to draw the attention to that part of the information
that is considered as particularly relevant by its producer, as the following text
passage from a talk by my colleague Henning Lobin at the conference “Digital
­Science” (20.09.2010 in Cologne) shows:

2. “.. Und auch hier wird gerade im BMBF ja eine Art Integration all dieser
Förderungen diskutiert, eh.. eine Zusammenführung dieser eh..Fördervorhaben, die
eh.., wenn man sie alle zusammenrechnet, ein unglaublich hohes Fördervolumen
auch im Moment aufweisen, um eben gerade insbesondere im Bereich der
Gesellchaftswissenschaften, Geistes- und Kulturwissenschaften so etwas wie ein
E-Science Initiative ins Leben zu rufen”

literal translation
“… and also here is in the BMBF (Federal Minsitry of Education and Research)
surely a kind of integration (of) all those eh – fundings being discussed, eh… a
joining (of) those support projects, which eh…, if one them all together counts,
an unbelievably high funding volume also at the moment exhibit, in order to just
exactly especially in the area of the social sciences, arts and cultural sciences (so)
something like an E-Science initiative into life to call ”.

The important point that he is making in this part of his talk – after speaking
about other research areas – is that the cultural sciences will get a lot of funding
within an initiative to support digital science. The relevant new information is
located in the right periphery, which starts with the final conjunction “um” and is
­followed by the adverbial focus markers “eben gerade insbesondere” in cumula-
tive order calling for intensified attention to the chunks of information given in
the final position, which guarantees the recency effect in the working memory of
the listeners.
In order to further substantiate the usefulness of Norman’s pertinence model
of attention for the concept of “Center of attention” in the MIC Theory, it is appro-
priate to recall John Morton’s influencing Logogen model (Morton 1969), in which
he introduced the notion of “threshold”. A logogen is an abstract unit that contains
information about a lexical item such as its sound, form and meaning and con-
nected images. In a language task, for instance in comprehending a word or using it
in a naming task the logogen will be activated by verbal or other stimuli and context
information, if the incoming information is consistent with the information stored
with the logogen and if a certain threshold level is overcome. Transferred to the
 Franz J. Stachowiak

notion of “Center of Attention” one can similarly assume that certain information
is necessary and certain “verbal tricks” must be applied in order to reach the level
at which pertinence is given to a partular constituent of a sentence. Focus markers
in this sense serve to reach the threshold at which focussed attention is activated.
Moving the subject of a sentence, which is also its topic and the agent of a
verbalized action into the very last position of a sentence captures the attention of
a hearer or reader, because this “trick” is – at first – against his or her expectations.
Thus it can be expected that in a FWO language the positions of new and old infor-
mation are exchanged for the purpose of attention capture. Neurologically this
leads to the activation of certain neural circuits in the brain. Experiments using
imaging technology seem to indicate that prefrontal, frontal, parietal and tempo-
ral areas of the brain are activated when attention is captured (Watkins et al. 2007).

7.  Th
 e MIC component in models of language production
and comprehension

If one focusses on language production and comprehension as a process of the

exchange of information in which new ideas are entered into our memory sytems
with the purpose of integrating new facets of information into the knowledge that
we already possess and hook concepts together in a new way, then it becomes
apparent that the speaker and the listener have different tasks in realizing this pro-
cess and achievings its goals. The speaker must present the information in such a
way that the listener is able to recognize what is new. Clark and Clark (1977) have
described this as the “Given – New – Contract”:

“The speaker agrees (a) to use given information to refer to information she thinks
the listener can uniquely identify from what he already knows and (b) to use new
information to refer to information she believes to be true but is not already known
to the listener”. (1977: 92)

In order to fulfill this contract the speaker’s task is to apply all the necessary means,
which her language offers in order to make it easy for the listener to identify what
is new and what is old information. The MIC theory investigates the devices that
the languages of the world offer to their speakers in order to achieve this goal.
Linguistic research shows that these devices depend on different structural pos-
sibilities which are determined by properties such as free versus restricted word
order and other typological aspects. In a model of language production a compo-
nent needs to be identified which represents the functions that a speaker fulfills in
determining the centers of attention (CA) in the linear order of a given sentence
in a given context.
 Memory and attention in MIC 

The listener’s task or goal is to determine the meaning of words, phrases

and clauses and to decode their syntactic relationships, in order to understand
the propositional content of a sentence. But this is only a basic part of the total
comprehension process, which requires relating sentences to each other and inte-
grating their contents in order to understand the gist of what has been said or
written in a larger context. According to Haviland and Clark (1974) listeners apply
a “given-new-strategy”, which includes the following steps:

–– Identify the given and new information in the linear structure of an utterance
–– Search your memory for information that matches the “given” information
(antecedent)
–– Integrate the new information into memory by substituting or equalizing the
new information with the antecedent knowledge
–– If there is no antecedent in memory, form bridging assumptions or implica-
tures in order to get the message
–– Rely on your world knowledge, beliefs and imagination and external situ-
ational information to decode the message

It follows also that a model of language comprehension must include a functional

component, which deals with the centers of attention at a certain point in the
sequence of processes that result in comprehension.
A. and H. Włodarczyk have proposed a functional explanation of the relation-
ship between centers of attention, their semantic counterparts and their referents
in a possible world (Figure 7). In order to express information natural languages
use a meta-informative function m-inf: D -> S (mapping centres of attention c
onto roles r of semantic situations) and informative function inf: S –> U from the
semantic situation S to the (mental) universe U (mapping, the roles r of semantic
situation onto the entities e taking part in relations in some possible world of the
universe), (acc. to A. Włodarczyk, Chapter 3 in this volume).

Participant Role Global/Local

or Location or Anchor Centre of attention

Ontology Semantics Pragmatics

e r c
Entity/Location inf Active/Passive m-inf Subject/Oject

inf(r) ≅ e m-inf(c) ≅ r

Figure 7.  Information (inf) and meta-information (m-inf ) as a composition of functions

 Franz J. Stachowiak

Transferred into a psycholinguistic model of language production and com-

prehension this functional system would be understood as a component that maps
information from different levels of processing – top down from cognitive and
pragmatic processes to the linearization of actual sentences. Whether the mapping
process, as indicated by the arrow in their model, can be reversed for the process
of language comprehension, awaits further discussion.
If the MIC component, included in a model of language production, decides
to which items attentional saliency is given, this will take place before positioning
the words in the linear order or as a part of this process. In the well known speech
production model by Garret (1980) this would still be part of the functional rep-
resentational level, possibly the last processing step before the activation of the
positional representational level.
A MIC component in a model of language comprehension would have the
function of enhancing or suppressing the attention given to sections in a text that
are marked or not marked as centers of attention. It would either be part of or be
closely related to the central executive of Baddeleys working memory model and to
Norman’s attention model. The exact nature of their interrelationsship is a matter
of further research. Neurolinguistic research (Stachowiak et. al. 1977; S­ tachowiak
1985) seems to indicate that the MIC component can be preserved under the con-
dition of aphasia, whereas grammatical processing is impaired. Psychiatric distur-
bances on the other hand resulting in psychotic speech or in schizophasia seem to
show the reverse picture, namely preserved language “tools” but severely impaired
processes of focussed attention on centers of informational saliency in texts.

References

Baddeley, Alan D. & Hitch, Graham J. 1974. Working memory. In The Psychology of Learning
and Motivation, Vol. 8, Gordon H. Bower (ed.), 47–89. London: Academic Press.
Baddeley, Alan D. 2000. The episodic buffer: A new component of working memory? Trends in
Cognitive Sciences 4(1): 417–423.
Baddeley, Alan D. 2003. Working memory and language: An overview. Journal of Communica-
tion Disorders 36(3): 189–203.
Broadbent, Donald. 1958. Perception and Communication. London: Pergamon Press.
Clark, Herbert H. & Clark, Eve V. 1977. Psychology and Language: An Introduction to Psycholin-
guistics. New York NY: Harcourt Brace Jovanovich.
Cowan, Nelson. 1999. An embedded-processes model of working memory. In Models of Work-
ing Memory, Akira Miyake & Priti Shahi (eds), 62–101. Cambridge: CUP.
Cowan, Nelson. 2001. The magical number 4 in short-term memory: A reconsideration of
­mental storage capacity. Behavioral and Brain Sciences 24: 87–185.
Cowan, Nelson. 2005. Working Memory Capacity. New York NY: Taylor and Francis.
Cowan, Nelson, Nugent, Lara D., Elliot, Emily, Ponomarev, Igor, & Saults, John Scott. 1999. The
role of attention in the developement of short-term memory: Age differences in the verbal
span of apprehension. Child Developement 70: 1082–97.
 Memory and attention in MIC 

Ericsson, Anders K. & Kintsch, Walter. 1995. Long-term working memory. Psychological Review
102: 211–245.
Friederici, Angela D. & Jacobsen, Thomas. 1999. Processing grammatical gender during lan-
guage comprehension. Journal of Psycholinguistic Research 28(5): 467–484.
Friederici, Angela D., Hahne, Anja, & Saddy, Douglas. 2002. Distinct neurophysiological
­patterns reflecting aspects of syntactic complexity and syntactic repair. Journal of Psycho-
linguistic Research 31: 45–63.
Garrett, Merrill F. 1980. Levels of processing in sentence production. Language Production:
Vol. 1, Speech and Talk, Brian Butterworth (ed.), 177–220. San Diego CA: Academic Press.
Gerdes, Kim & Kahane, Sylvain. 2006. A polynomial parsing algorithm for the topological model:
Synchronizing constituent and dependency grammars, illustrated by German word order
phenomena. In Proceedings of the 21st International Conference on Computational Linguis-
tics and the 44th annual meeting of the Association for Computational Linguistics, 1097–1104.
Gundel, Jeanette K. & Fretheim, Thorstein. 2004. Topic and focus. In Handbook of Pragmatic
Theory, Larry Horn & Gregory Ward (eds), 174–196. Malden MA: Blackwell.
Haviland, Susan E., & Clark, Herbert H. 1974. What’s new? Acquiring information as a process
in comprehension. Journal of Verbal Learning and Verbal Behavior 13: 512–521.
Morton, John. 1969. Interaction of information in word recognition. Psychological Review 76:
165–178.
Murdock, Bennet B. Jr. 1962. The serial position effect of free recall. Journal of Experimental
Psychology 64: 482–488.
Myachykov, Andriy & Tomlin, Russell. 2008. Perceptual priming and structural choice in
­Russian sentence production. Journal of Cognitive Science 6(1): 31–48.
Norman, Donald E. 1968. Toward a theory of memory and attention. Psychological Review 75:
522–536.
Oberauer, Klaus. 2003. Understanding serial position curves in short-term recognition and
recall. Journal of Memory and Language 49: 469–483.
Pöppel, Ernst. 1998. Wie kommt die Zeit in den Kopf. Zeitschrift für KulturAustausch 3: 29–31.
Stachowiak, Franz J. 1985. Metaphor production and comprehension in aphasia. In The Ubiquity
of Metaphor [Current Issues in Linguistic Theory 29], Wolf Paprotté & René Dirven (eds),
559–599. Amsterdam: John Benjamains.
Stachowiak, Franz J., Huber, Walter, Poeck, Klaus & Kerschensteiner, Max. 1977. Text compre-
hension in aphasia. Brain and Language 4: 177–195.
Telljohann, Heike, Hinrichs, Erhard W., Kübler, Sandra, Zinsmeister, Heike & Beck, Kathrin.
2006. Stylebook for the Tübingen Treebank of Written German (TüBa-D/Z). Tübingen: Semi-
nar für Sprachwissenschaft, Universität Tübingen.
Watkins, Susanne, Dalton, Polly, Lavie, Nilli & Rees, Geraint. 2007. Brain mechanisms mediat-
ing auditory attentional capture in humans. Cerebral Cortex 17(7): 1694–1700.
Waugh, Nancy C. & Norman, Donald A. 1965. Primary memory. Psychological Review 72(2):
89–104.
Włodarczyk, André & Włodarczyk, Hélène. 2008. Roles, anchors and other things we talk
about: Associative semantics and meta-informative centering theory. Intercultural Prag-
matics 5(3): 345–365. Berlin: Mouton de Gruyter.
Zifonun, Gisela, Hoffmann, Ludger, Strecker, Bruno & Ballweg, Joachim. 1997. Grammatik der
deutschen Sprache. Berlin: De Gruyter.
part 3

Meta-informative centering in languages

 It-clefts in the meta-informative
structure of the utterance in Modern
and Present-day English*

Ana Elina Martínez-Insua & Javier Pérez-Guerra

Universidade de Vigo

This chapter focuses on the it-cleft construction and aims at, first, profiling it
from the perspective adopted in Meta-Informative Centering Theory (hereafter,
MIC) and, second, describing the major tendencies yielded by a diachronic
corpus-based analysis. It-clefts deserve a specific place in an account of English
constructions because of at least two specific characteristics: the expletive nature
of its introducer and the difficulty of accounting for the postverbal subordinate
clause by resorting to grammatical concepts such as relativisation, detached
postmodification, or even some kind of right dislocation.
  The English it-cleft will be presented here as a device of focalisation used for
establishing a meta-informative contrast with the second part of the utterance,
making it possible, in a Strict Word Order (SWO) language as English, to put this
focalised constituent at the front of the utterance (thus contradicting the neutral
order: given-before-new). The study presented here illustrates the quite recent
consolidation of the mechanism of clefting as a thematising meta-informative
strategy in the English language. This consolidation of the it-cleft construction
is corroborated both quantitatively and qualitatively. By means of this system,
the speaker focuses on the referent of the Local CA constituent that occupies the
X-position and places heavy (normally available) information in final location
within the background rightmost clause.

1.  Introduction

In the literature the label ‘cleft’, as suggested in Jespersen (1909–49, 1937), is com-
monly used with structures such as (1) to (5), which, irrespective of their syntactic

*  The research reported in this article was funded by the Spanish Ministry of Science and
Innovation and the European Regional Development Fund (grant no. FFI2009-11274/FILO),
and Autonomous Government of Galicia (Directorate General for Scientific and ­Technological
Promotion, Grant No. CN2011/011).
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

organisation, are so called “because of the apparent dismemberment of a single

sentence entailed in their derivation” (Delahunty 1982: 5):
(1) What this paper describes is the cleft construction.
(2) The cleft construction is what this paper describes.
(3) It is the cleft construction that this paper describes.
(4) That is the cleft construction (that) this paper describes.
(5) They are real researchers that tackled this issue.

The construction1 in (1) is called basic pseudo-cleft or basic wh-cleft, (2) illustrates
the reversed or inverted pseudo- or wh-cleft, the structure in (3) is known as cleft
or it-cleft, the one in (4) is known as th-cleft, and the example in (5) illustrates
a pronominal cleft. The treatment of the constructions in (1) to (5) as clefts is
grounded not only on the formal similarities between the constructions but also
on semantic, pragmatic and communicative considerations. To illustrate this, it
is commonly assumed that it- and pseudo-clefts share the same truth-conditions
or, as Prince (1978: 884) puts it, the “same objective information”, which implies
that they are semantically identical. In this vein, Akmajian (1970: 149) maintains
that it-clefts and pseudo-clefts “are synonymous, share the same presuppositions,
answer the same questions, and in general they can be used interchangeably”.
Declerck (1988: 209), however, claims that “there are numerous pragmatic factors
that may induce the speaker to prefer one type of cleft to another in a particular
context” [our italics] (see Traugott 2008: Section 3 for semantic and informative
differences between it- and pseudo-clefts).
It-clefts deserve a specific place in an account of English constructions because
of at least two specific characteristics: the expletive nature of its introducer and the
difficulty of accounting for the postverbal subordinate clause by resorting to gram-
matical concepts such as relativisation, detached postmodification, or even some
kind of right dislocation. First, on strictly syntactic grounds, basic pseudo-clefts
can be analysed as (unmarked) sentences in which a relative clause,2 be it headless
or headed, functions within the subject of the sentence. Second, with the same

.  For the consideration that it-clefts constitute a grammatical construction type, see, among
others, Davidse (2000), Lambrecht (2001) and Traugott (2008).
.  Although the majority of the scholars agree with the consideration that the wh-clause in
a pseudo-cleft is a relative clause (see references in favour of this view in Dikken et al. 2000),
there are some who think differently. To cite an example, Quirk et al. (1985: 1388) ground their
disagreement on the grammaticality of why pseudo-clefts, which have no relative counterpart,
as in ?Why we left was because we were tired. As we see it, the unacceptability of such examples
is due to the fact that headless why ‘relative clauses’ do not exist any longer, which corroborates
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

argument in mind, a reversed pseudo-cleft can be claimed to contain a relative

clause fulfilling the syntactic role of subject predicative. Third, in th- and pro-
nominal clefts the introductory (non-expletive) referential items together with the
coreferring postverbal subject predicatives can constitute a complete clause, and
the postverbal subordinate clauses can be said to be linked either to the subjects or
to the subject predicative and thus constitute an example of postmodification (or
even some sort of right dislocation).
This chapter focuses on the it-cleft construction and aims at, first, profiling
it from the perspective adopted in Meta-Informative Centering theory (hereaf-
ter, MIC) and, second, describing the major tendencies yielded by a diachronic
­corpus-based analysis. The English it-cleft will be presented here as a device of
focalisation used for establishing a meta-informative contrast with the second
part of the utterance, making it possible, in a Strict Word Order (SWO) language
as English, to put this focalised constituent at the front of the utterance (thus
­contradicting the neutral order: given-before-new). The study is organised as fol-
lows: in Section 2 we outline the barebones of MIC. Section 3 is devoted to the
MIC-conformant meta-informative analysis of it-clefts (see Section 2 below for
the characterisation of meta-information as the sequential ordering of informa-
tion). In Section 4 we focus on the semantic and informative characteristics of
it-clefts. In Section 5 we account for the data on which our corpus analysis is based
and analyse the degree of variation undergone by the construction in Modern and
Present-day English. Finally, Section 6 is devoted to the summary of the findings
and the concluding remarks.

2.  The theoretical framework: The MIC theory

Following the tradition of the Prague theory of Functional Sentence Perspective

(Daneš 1974 or Firbas 1992, among others), the MIC theory contains also ele-
ments (anaphor vs. cataphor) which correspond to the backward and forward
looking centres in the American Centering Theory (Grosz & Sidner 1986). Even
if both theories were created separately, as Włodarczyk & Włodarczyk (personal
communication) state, MIC easily integrates within the European framework a
number of aspects and results of the American Centering Theory, there being

the analysis of the wh-components of pseudo-clefts as relative clauses. In fact, why pseudo-
clefts were possible in older English, as shown in (i), retrieved from the Helsinki Corpus:

(i) And why he is moyst, is, that it should be the more indifferenter and abler to euery
thing that shoulde be reserued or gotten into him: (Vicary: 33).
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

several points of coincidence between both theories. The American framework

may be understood as “an account of one aspect of discourse processing, local
discourse structure, that makes specific claims about both processing complex-
ity and discourse anaphora” (Walker, Joshi & Prince 1998: preface). Centering is
“a model of the conversants’ center of attention in discourse that is concerned with
the relationship of attentional state, inferential complexity, and the form of refer-
ring expressions” (Walker, Joshi & Prince 1998: 1). Such an approach to language
aims to model discourse processing factors that might explain the differences in
the degree(s) of coherence perceived by addressees when they receive and inter-
pret messages.
The origins of the MIC theory go back to 1999, when Włodarczyk (see
Włodarczyk & Włodarczyk 1999) used the concept ‘centre d’intérêt’ for the first
time. Since then, this French term has been replaced by that of ‘centre of atten-
tion’ under the influence of the Grosz-Sidner centering-theoretical framework,
on the one hand, and it underwent a few important revisions and reassessments
in the subsequent papers published in English, on the other hand. In Grosz,
Joshi & Weinstein (1995), centres of attention are defined at the text level: one
constituent of an utterance is treated as a ‘forward’ or ‘backward looking centre’
in order to maintain the cohesive flow of information from on utterance to its
successor. Forward and backward looking centres make it possible to give an
account of the relations which bind utterances together into a coherent text.
MIC assumes that, because of the linear order of speech sounds in human lan-
guages, no judgment may be uttered without selecting at least one centre of
attention, and centering is regarded as a structuring operation at the levels of
both text and utterance.
In the MIC model, the main focus is on ‘meta-information’, which is the
label for the sequential ordering of ‘information’, that is, for the sequential order-
ing of the content of linguistic utterances. In this sense, the meta-informative
level “is necessary in order to achieve the ordering of non-linear mental repre-
sentations as texts (sequences of linguistic utterances)” (Włodarczyk, this vol-
ume). In order to communicate such mental non-linear situations, users of the
language select the element(s) they will be treating as the centre(s) of attention
in their messages and predicate something about it/them. Once the centres of
attention (henceforth, CAs) have been selected from the referents or ‘anchors’
of the situation, the user establishes a hierarchy and turns the primary (most
important) one into the subject of the utterance, while the secondary centre(s)
is/are turned into the object(s). According to MIC, in English the primary CA
is global and preverbal (represented in upper left nodes of phrase-markers),
whereas the secondary ones are local (located in lower right nodes within the
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

tree structure). The specific labels for one and the others are ‘Global CA’ and
‘Local CAs’, respectively.
Centering and predication are crucial notions within the MIC Theory. While
centering involves selecting and highlighting one entity among others, predication
implies saying something about the entity selected by the speaker as the global
CA of the message. Thus, predication takes place when speakers produce linguis-
tic expressions in which “some distinguished segments are highlighted as centres
of attention” (Włodarczyk, this volume). From such a perspective, “no utterance
can be formed in a natural language without choosing a CA and assigning to it
a meta-informative Old or New status” (Włodarczyk, this volume). This implies
that the meta-informative status of the CAs, as treated in this framework, con-
cerns their oldness or newness. In other words, the meta-informative status of
information concerns the way in which it is treated in the communicative event,
the way speakers introduce it into their discourse. Ultimately, the old/new meta-
informative status originates in the discourse strategies (partly fixed by syntactic
rules) established and employed by the speaker (Włodarczyk, this volume).
MIC has already been presented elsewhere as an appropriate framework
for the analysis and study of English thematically marked sentences, especially
those where an Anonymous Subject (henceforth, AnS) occupies the Global
CA position that another semantically full constituent would occupy in the
unmarked version of the sentence (see, Martínez-Insua 2011, Martínez-Insua &
Pérez-Guerra 2012). This chapter aims at highlighting the appropriateness of
this theoretical framework for the description of clefting as a focusing meta-
informative strategy.

3.  Description of it-clefts

This section tries to characterise it-clefts against the background of the MIC
approach to language, bearing in mind MIC’s distinction between Global and Local
CAs. In Section 3.1 we describe the meta-informative organisation of the it-cleft
sentence and pay special attention to the distribution of the CAs. In ­Section 3.2
we describe the structural and grammatical features of the construction.

3.1  A MIC-compliant representation of it-clefts

From a representative point of view, it-clefts such as (3) above, repeated here for
convenience, might be described in two ways:
(3) It is the cleft construction that this paper describes.
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

(i) Sentences where the expletive and semantically empty AnS it3 fills the Global
CA slot, while a semantically full constituent functions as Local CA and is
followed by a clause that constitutes its background. The meta-informative
layout of the it-cleft would then be described as in (6):
(6) [ItAnS]Global CA [be] [Xi]Local CA (FOCUS) [introducer +
[clause …n.p.i…]]BACKGROUND

where the rightmost clause is the background contrasting with the focus expressed
in the cleft part of the utterance. This clause contains a null pointer [n.p.i] left by
the focalisation of one of its constituents [Xi] which acts as the Local CA.

(ii) Sentences where the expletive and semantically empty AnS it fills the Global
CA slot, while the semantically full Local CA is a clausal constituent whose
unmarked meta- informative layout has been altered. As a result of such word-
order alteration in the (main) clause, one of the components of the (subordi-
nate) clause has been focalised and brought to initial position, leaving a null
pointer behind. The meta-informative layout of the it-cleft is then described
as in (6′):
(6′) [ItAnS]Global CA [be] [[Xi] [introducer + [clause …n.p.i…]]]Local CA

where the Local CA is made up of a (rightmost) clause from which one of the
constituents has been topicalised and placed before the introducer [Xi], leaving a
null pointer behind [n.p.i].
The first one of these two characterisations, where the rightmost clause is
taken as the background of the utterance, seems to be the most appropriate one, as
we see it. The role of the rightmost clause as a second-level CA is in line with the
role of constituents such as a book, in Mary gave John a book, to John in Mary gave
a book to John, or with his key in John opened the door with his key. Knowing that
MIC does not allow for the presence of “neither double Local nor two-­members
single Local CAs” (Włodarczyk & Włodarczyk, personal communication), the
rightmost clause has to be interpreted as a second-level CA (Włodarczyk &
Włodarczyk 2006: 8). In other words, according to the MIC theory, which bases
its premises on actual surface structure, we argue that a link construction such

.  From a contemporary synchronic perspective, the expletive nature of the so-called ‘intro-
ducer’ (it) is justified by the fact that agreement does not hold between the two main constitu-
ents of the copular construction (it as the grammatical subject and the focus or ‘X’), as in It is
my parents that I never trust (see, in this respect, Pérez-Guerra 1999: Chapter 4, Section 3.2).
By using historical data, a number of scholars maintain that it keeps referential (cataphoric)
status (see Pérez-Guerra 1999: 170–171).
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

as the it-cleft construction (see Section 4.1 in this respect) consists of the linking
of two second-level (focus and background) CAs by means of a syntactic design
governed by an almost semantically bleached linking verbal operator, namely the
expletive it.4

3.2  A structural and grammatical description of it-clefts

For a construction to be considered as it-cleft, the following conditions should
be met:5

(i) The Local CA ([X] in the representation in (6) above and the null pointer
are coreferring. For this condition to hold, X must materialise, in Givón’s
(1984: 731) words, an entity; otherwise referentiality would be out of the ques-
tion.6 The characterisation of X as an entity-denoting constituent excludes
sentences like (7), adapted from Declerck (1988), from the class of clefts:
(7) It is always expensive what Cambridge University Press sells.

Example (7) illustrates the extraposition of the Subject what Cambridge University
Press sells and the filling of the empty Global CA slot with the AnS it.

(ii) The ‘introducer’ in the rightmost clause can be one of the following: Ø
(­Visser’s 1970: Chapter I apo koinou constructions), that, who or which. In the
literature, other expressions – for instance, what, when (see Declerck 1997)
or where – are frequently included in the set of possible introducers of the
rightmost clauses of it-clefts. From a syntactic perspective, such (apparent)
clefts with wh-forms are not syntactically different from extraposed headless
relative clauses or pseudo-clefts, as shown in (8) and (9) vs (10) (similarly,
Delahunty 1982: 268ff & Ball 1994a: 181):

.  This proposal is, in fact, in keeping with syntactic solutions which rely on the fact that the
syntactic relation holding between the focus and the clause is subordination (see Section 3.2
in this respect).
.  Notice the markedly syntactic nature of the criteria and terminology employed in the
following list.
.  Example (i) will only be acceptable if a genius is understood as an entity-denoting
‘nominal,’ that is, in an identifying (probably contrastive) way:

(i) It’s a genius that he is.

If a genius is interpreted predicatively, the sentence becomes unacceptable. Notice that only
under very special circumstances can predicative nominals undergo clefting (see (iii) in the
main text).
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

(8) It is phrase-markers what I drew on the blackboard.

What I drew on the blackboard is phrase-markers.
(9) It was before the 1957–58 tour of South Africa, when Bagenal said
half-jokingly before some of the team: ‘I wish I was going with you
so-and-so’s.’ (Findlay: 210)
(10) It is phrase-markers that I drew on the blackboard.
*That I drew on the blackboard is phrase-markers.

(iii) The null pointer within the rightmost clause may fulfil a large array of func-
tions: Global CA (in (11)), Local CA (in (3) above and in (12)), (non-sentence)
adverbial (in (13) and (14), adverbial complement or obligatory adverbial
(in (15)), prepositional complement of a verb (as in (16)), of an adjective
((17)) or of a noun ((18)),7 complement of a preposition (in (19)), predica-
tive complement of the subject or of the object in very special environments
(examples (20) and (21),8 taken from Declerck 1988), predicative9 (in (22)), or
particle of a phrasal verb (in (23)).
(11) It is a gap that occurs in initial position.
(12) It is {#Ø/10 to} me that he dedicated the book.
(13) It was with much attention that I checked the last proofs of the article.
(14) It was only reluctantly that he agreed to help me with them at all.
(15) It is to Boston that she went.
(16) It is to my article that she was referring.
(17) It was about that Minister that the President was angry.

.  See Dik (1997: 309–310) for the analysis of clefts with PP-foci as non-prototypical clefts.
.  In examples (20) and (21), pretty and a teacher are clearly contrastive and thus the meaning
of the construction is identifying, as required by condition (i) in the main text, ­otherwise the
clefting of a predicative complement would be disallowed.
.  Inflected verbs cannot be cleft (see (i) below) and (uninflected) VPs are accepted as Local
CAs in clefts only in informal Irish English (as in (22) in the main text; see Givón 1984: 731
for a detailed account):

(i) *It’s wore that John e a white suit at the lecture.

.  Huddleston (1984: 460) regards non-prepositional indirect objects in X-position of clefts
like *It was Liz that I bought the flowers as ungrammatical. Delahunty (1982: 87) only accepts
them if they include stranded prepositions, as in It was Liz that I bought the flowers for. This
example should be included in the category of complement of a preposition (see (iii) in the
main text).
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

(18) It was of Syntactic Structures that he was the writer.

(19) That was the doctor I was speaking to.
(20) It’s pretty that my mother-in-law is, more than anything else.
(21) It’s a teacher that he is, not a butcher!
(22) ?It was teach English that he did at that time.
(23) It was (obviously) off that I turned the computer when I suspected it was
infected.

Disjuncts and conjuncts cannot occur in the Local CA position of clefts, as shown
in examples (24) and (25), respectively:
(24) *It was frankly that he didn’t expect to discover the philosopher’s stone
in his study.
(25) *It is nonetheless that conjuncts can be cleft in languages other than
­English.

(iv) The null pointer/initiator belongs to one of the following categories:

–– NP, as in (3) above,

–– PP, as in (13), (15), (16), (17) or (18),
–– Adverb Phrase, as in (14),
–– Particle of a phrasal verb, as in (23) above.
–– The Local CA in a cleft does not normally belong to the category clause,11 VP
(see footnote  10) or (non-contrastive) AP (in (26) below; see, however, (7)
and (20) for acceptable APs in Local CA position):
(26) *It’s very unhappy that Ph.D. students are.

.  The acceptability of the clefts in (i), (ii) and (27) in the main text, including, respectively,
an -ing, a that- and an infinitive clause in X-position, indicates that the restriction on clausal
clefting is maybe due not to the theoretical organisation of the construction but to perfor-
mance preferences:

(i) It was changing the word-processor that I most hated.

(ii) ?It was that I preferred a four-section article that I told him. [special intonation is
required]

Delahunty (1982, 1984) justifies the acceptability of that-clauses by contending that they are
NPs. In fact, only that-clauses which are equivalent to entity-NPs, namely factive that-clauses,
are eligible for clefting. *It was that he was going to finish the book in two years that he thought
is ungrammatical because it includes a non-factive cleft clause in X-position.
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

(v) The rightmost clause can be either finite (supposedly a that-clause), as in (3)
above, or nonfinite (-ing or infinitive clause), like the one in (27):
(27) [Within the United States,] it is Robinson to appear like a Jones. (Gibb:115)

(vi) The construction is introduced by an AnS, materialised by dummy or exple-

tive it. In our opinion, precisely because the anonymous nature of it is taken
as one of the identifying features of the it-cleft construction,12 examples like
(4) or (5) cannot be included in the label ‘cleft’. By contrast, they will be given a
quasi-right-dislocation analysis, in which the rightmost constituent functions
as a displaced postmodifier of the pronominal Global CA. This view receives
support from the fact that, whereas the example in (4) is able to undergo
reversion, which is a defining feature of identifying utterances, clefts cannot
reverse their Global and Local CAs around be ((3′)):
(4′) The cleft construction (that) this paper describes is that.
(3′) *The cleft construction that this paper describes is it.

4.  Informative aspects of it-cleft sentences

Despite their markedness, it-clefts, like other thematic systems (e.g. most there-
sentences), normally have the truth-conditions of their unmarked counterparts,
even though minor informative differences between both versions can be detect-
ed.13 In this section, devoted to the study of certain semantic and pragmatic
aspects of the constructions under analysis, we pay attention, first, to the iden-
tifying nature of it-clefts (Section 4.1) and, second, to the consequences which
the meta-informative layout of an it-cleft has for the informative design of the
sentence (Section 4.2).

.  Visser (1970: 50) claims that examples of the type they are my brothers that… were clefts
from the sixteenth to the nineteenth centuries.
.  Atlas and Levison (1981: 2), when they deal with the cleft in (i) and the noncleft in
(ii), point out that “[(i)] exhibits presuppositional behaviour that [(ii)] does not, namely, the
preservation of the inference to [(iii)] under denial and questioning of [(i)]” (see Halvorsen
1978: 18–19 for a similar account):

(i) It was John that Mary kissed.

(ii) Mary kissed John.
(iii) ‘Mary kissed someone’
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

4.1  Clefts as identifying structures

Informatively speaking, and being a subtype of be-structures, clefts must be clas-
sified in one of the two groups of link constructions, namely, identifying (or
specifying)14 and attributive (or predicational) (see Enkvist 1979 or Declerck
1988 for the semantic and formal differences between both classes; see Halliday &
­Matthiessen 2004 or Thompson 2004 for the functional characterisation of both
these processes). In a word, these types of link constructions can be described
simply by saying that identifying be-sentences specify a value for a variable15 (as
suggested by Declerck, the semantic scheme of identifying sentences is ‘x=y’, inter-
pretable as ‘assign the value y to x’, the value occupying the Local CA position
and the variable occurring in the Global CA position), while attributive structures
predicate a property of their Local CAs in a non-identifying way. Most clefts seem
to illustrate the former type, whereas pseudo-clefts may have both readings, that
is, identifying and attributive. This leads us to disallow examples like (7) above,
repeated here for convenience, as members of the class of clefts:
(7) It is always expensive what Cambridge University Press sells.

It is obvious that (7) has not got an identifying meaning, since expensive is not a
value but a predication.
Let us now deal with the clefts in (28) to (31), which are taken as attributive
in the literature:
(28) It is a poor heart that never rejoices.
(29) It’s a wise child that knows its own father.
(30) It would be a brave man who marries her, won’t it?
[taken from Declerck (1988)]
(31) Was it an interesting meeting you went to last night?

Regarding proverbial examples like (28) and (29), and less proverbial ones such as
(30), whose reading is clearly attributive, we agree with Declerck that these exam-
ples are not clefts but sentences involving some sort of relative clauses. Out of the
several reasons which Declerck adduces in favour of his proposal, we will, just for
the record, mention two. On the one hand, these examples cannot be uncleft ((28′)

.  Declerck (1988) distinguishes between ‘specificationally-identifying’ and ‘descriptionally-­

identifying’ be-structures. Since we are not aware of any instance of descriptionally-­identifying
cleft, we will not make use of Declerck’s distinction.
.  In Declerck’s (1988: 228) words, clefts “suggest that an answer is being given to a question
that is implicitly present in the mind of the hearer”.
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

to (30′)); on the other, the selection of relative pronouns is closer to clear relative
clauses than to the usual pattern of clefts ((28″) and (29″) vs. (32)):

(28′) A poor heart never rejoices. [the meaning is different from that of (28)]
(29′) A wise child knows its own father. [the meaning is different from that of (29)]
(30′) A brave man would marry her. [the meaning is absolutely different from
that of (30)]
(28″) It is a poor heart that/which/who never rejoices.
(30″) It would be a brave man that/who marries her, won’t it?
(32) It will be John that/#who will marry her, won’t he?

As far as (31) above is concerned, Declerck concludes that, owing to its hybrid nature
between the identifying and the attributive subclasses, this example illustrates a
new type of it-be construction. Most of the features cited by Declerck evincing the
attributive reading of the construction are semantic, whereas those signalling the
identifying interpretation are grammatical or syntactic. From both a grammatical
and a semantic point of view, the (supposedly) new class represented by (31) eas-
ily finds its way in the general class of (identifying) clefts. As we see it, Declerck’s
judgements are influenced by the obvious attributive nature of the adjective inter-
esting premodifying X in (31), which he extends up to the whole construction.
Declerck mentions the fact that the Local CAs (X-segments) in examples like (31)
can be quantified either by inserting no or by grading the adjective, and adds that
this is a consequence of the attributive character of the sentence. Adjectives, both
in attributive and in identifying contexts, can be expanded by constituents which
are modifiers typical of attributive items and not of the contexts in which these
adjectives are involved. In other words, the possibility of either inserting no before
the adjective or of grading interesting in (31) simply demonstrates the attributive
meaning of interesting, not of the whole example. Moreover, the supposedly ‘odd’
example (33) given by Declerck is perfectly acceptable in English, at least by the
native speakers consulted:
(33) It is John that/who is going to lead and an interesting subject that is going to
be discussed.

If our informants are correct in their predictions, the coordination of the clear cleft
It is John who is going to lead and (It is) an interesting subject that is going to be dis-
cussed in (33) corroborates the non-sylleptical semantic equivalence of the struc-
ture, or, in other words, the inclusion of (31) in the class of (identifying) clefts.
Declerck also compares (31) with other structures comprising verbs other
than be which are clearly attributive:
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

(34) [A couple of weeks ago it seemed clear which candidate deserved voting for,
but now many people think that] it has become a difficult choice they have
to make.
(35) It used to be a really good painter who painted their portraits.
We agree with the attributive nature of these examples, which is demonstrated
by their impossibility to be formulated as non-clefts, as shown in (34′) and
(35′):
(34′) They have to make a difficult choice. [not equivalent to (34)]
(35′) A really good painter {used to paint / painted} their portraits.
[not equivalent to (35)]
By contrast, ‘identifying’ (31) has a corresponding non-cleft:
(31′) Did you go to an interesting meeting last night?

Finally, Declerck claims that neither exhaustiveness nor contrast is involved

in (31). In this connection, we concur that the justification of the acceptability
of indefinite NPs in the Local CA of (identifying) clefts is the contrastive content
of their modifiers. Put another way, an interesting meeting, even though formally
indefinite, is specific at least from the speaker’s viewpoint. The contrast expressed
by the example is confirmed by its intonation pattern, recognised as such by
Declerck:
(31″) Was it an INTERESTING meeting that you went to last night? [–No, it was
a BORING meeting…]

Once the contrastiveness reading of (31) has been accepted, the meaning of
exhaustiveness can be justified. Finally, as we see it, tense restrictions affect not
only attributive sentences, as Declerck tries to show in (36), but also identifying
ones like (37):
(36) It would be/*is a more interesting subject that we would be discussing if
John had his way.
(37) It would be/??is Linguistics that we would be discussing if John had his way.
Summing up, we have tried to demonstrate that Declerck’s mixed type of identify-
ing/attributive clefts belongs to the general one of (identifying) clefts. Apart from
semantics, already discussed, all the grammatical factors (existence of non-cleft
counterparts, intonation and even pronominal connection) in Declerck’s study
also point towards the aforementioned conclusion:
(38) Was it a MEETING that you went to last night?
(39) It was an excellent description of himselfi/*himi that Johni wrote.
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

We have agreed with Declerck about the exclusion of certain it-be structures from
the class of clefts. To attempt the analysis of these examples is beyond the aims of
this chapter. It is their exclusion from the class of clefts that must remain uncon-
troversial for the purposes of the corpus study.

4.2  Meta-informative structure

From a communicative/pragmatic point of view, clefting rearranges the topic-
comment structure of the sentence. On the one hand, the information in the
rightmost that-clauses is presupposed (Prince’s 1978 or Declerck’s 1988 ‘informative-­
presuppositional’ clefts) and, on the other, the Local CA is focal (Declerck’s ‘stressed-
focus’). This dual communicative nature of clefts is summarised in the following
quotation by Enkvist (1979: 151): “clefted elements (…) express new information
and evoke presuppositional sets” [our italics].
It is normally assumed that clefts are structures that determine pragmatic
functions in a meta-informative way. More specifically, in adopting the meta-
informative cleft pattern, one is formally recognising that the that-clause follow-
ing the Local CA is pragmatically presupposed or given (Engelkamp & Zimmer
1983: 40; Brömser 1984: 330) from the speaker’s viewpoint, even though it is actu-
ally new for the hearer.16 However, as Lambrecht (1994: 70–71) points out, in a
sentence like (40) [based on his (2.21)], uttered out of the blue by a lecturer, the
information conveyed by that said that… does not have to be presupposed by the
audience, which contradicts both Hetzron’s (1975: 361) assertion that clefts cannot
be “unprecedented”, and Rochemont’s (1986) and Declerck’s (1988: 231) predic-
tions that clefts cannot be used out of the blue because they require the presup-
posed part to be in the hearer’s consciousness:
(40) It was George Orwell that said that the best books are those which tell you
what you already know.

We contend that even though the content of the that-clause in (40) is clearly not
shared by the audience, it is – and here lies the justification of the cleft meta-­
informative organisation – presented as such. Harold (1995: 158) claims that the

.  See Croft (1991: 119), who claims that “the information in the backgrounded clause
refers back to an already open cognitive file, namely, a file established under ‘generally known
facts’. Thus, even if the backgrounded clause is new information to the hearer, a cognitive
file has been opened for it”. Similarly, this new/given duality is supported by Engelkamp and
Zimmer (1983: 64) when they say that “cleft sentences are used in particular communicative
situations, i.e. when the speaker fills an ‘active’ [author: givenness dimension] gap [author:
newness dimension] in the hearer’s knowledge” [our italics].
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

information in the that-clause of a cleft “is generally not in the listener’s mind. It
may be known or inferable or it may be totally new”. In this latter case, as pointed
out before (see also Declerck 1988), by means of the clefting device, the new infor-
mation is presented as if it were old. This leads us to conclude that the aim of
English clefts is to trigger the interpretation of the that-clause after the Local CA
as given, be it a carrier of old information or not.
Clefts are also claimed to bring forward to the attention of the hearer/reader a
certain (post-be) theme of the discourse as the focus of attention (the Local CA),
which makes clefts be considered meta-informative devices for focus-marking
(Rochemont’s 1986 ‘constructional focus construction’). This ‘presentative’ func-
tion is outlined in, for instance, Hetzron (1975: 362–364), Van Oosten (1986: 59)
or Declerck (see the section devoted to so-called ‘stressed-focus clefts’ in his 1988
study). Consequently, the meta-informative layout of the cleft makes the hearer
interpret George Orwell as focal in, for Example (40) above. Once it has been char-
acterised as such, the referent of the Local CA may become the ensuing discourse
theme.
The principle of end-focus (as part of the general one of given-before-new),
when applied to the it-cleft pattern, predicts that, on the one hand, the rightmost
clause of the it-cleft must contain information either unavailable or less predict-
able than the information conveyed by other constituents preceding the final
clause, and, on the other, which is in fact a consequence of the previous one, the
referentiality potential of the segment in postverbal Local CA should be, commu-
nicatively speaking, more referring than that of the final clause.

5.  The data

This section accounts for the data employed in our analysis of it-cleft sentences in
the history of English. In Section 5.1 we account for the frequency of the construc-
tion from Late Middle English to the present. Section 5.2 deals with the informa-
tive status of the major constituents in the it-clefts in the database.

5.1  Frequency of it-clefts

Table 1 gives information about the database on which this investigation is based.
The Late Middle (LME) and Early Modern English (EModE) data have been
retrieved from The Helsinki Corpus of English Texts. The Late Modern English
(LModE) examples have been taken from A Representative Corpus of Histori-
cal English Registers (ARCHER; see Biber et al. 1994) and the Corpus of English
Dialogues (CED). Finally, the contemporary data have been selected from The
Lancaster-Oslo/Bergen Corpus of British English (LOB, texts dated in 1961).
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

Table 1.  The corpus (raw data and normalised frequencies [n.f.] per 100,000 words
and 1,000 clauses)
Period words clauses it-clefts n.f./100,000 words n.f./1,000 clauses

LME: 1420–1500  71,097  4,751  3  4.21 0.63

EModEI: 1500–1570  61,219  3,891  4  6.53 1.02
EModEII: 1570–1640  75,762  5,729  5  6.59 0.87
EModEIII: 1640–1710  62,940  4,360 11 17.47 2.52
LModE: 1710–1900  67,962  6,247 13 19.12 2.08
PDE: 1961  98,007  6,974 36 36.73 5.16
436,987 31,952 72 16.47 2.25

It-clefting has not been a productive thematic mechanism in the history of the
English language. As Ball (1994b: 610) observes, a few “clefts and cleft-like con-
structions are attested in (…) OE and Early Middle English”. Whereas in LME the
number of it-clefts is not significant at all in the corpus employed (0.63), at the end
of the Modern period, their frequency is slightly above two out of one thousand
clauses. In PDE the proportion of clefts reaches and exceeds the figure of 5 out of
one thousand clauses.

5.2  Information in it-clefts

In Section 4.1 we concluded that end-focus, when applied to the it-cleft pattern,
predicts that the rightmost clause of the it-cleft conveys information which is not
predictable and that the informative status of the Local CA should be more refer-
ring than that of the final clause. Tables 2 and 3 contain the data for the referential-
ity potential of, respectively, the rightmost clauses and the Local CAs of it-clefts.
The basic taxonomy of informative content is as follows:

–– referring (ref) constituents, either linguistically/textually or deictically, when

the content of the linguistic expression has already been mentioned or alluded
to in the discourse, belongs to universal knowledge, or is a current situational
element. In keeping with Ariel (1996: 23ff), an expression will be regarded as
linguistically/textually referring if it is not new in a span of seven sentences
prior to its occurrence.
–– low-referring (low-ref) constituents, when only non-head components of the
expression i.e. its complements or modifiers – are informationally available,
when the entity denoted by the expression is derivable from the linguistic con-
text, or when its referent has been alluded to in the previous discourse in a
span of more than seven clauses.
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

–– non-referring (non-ref) constituents, when the referent can neither be

recalled from the discourse domain linguistically – in the previous seven
clauses –, ­situationally or permanently, nor derived from a previous referring
expression.

Table 2.  Referentiality of the sentence-final clause

period Referentiality

non-ref low-ref ref

LME 1  2
EModE 3  8  9
LModE 1  2 10
PDE 4 17 15

Table 3.  Referentiality of the local CA

period Referentiality

non-ref low-ref ref

LME  1 1  1
EModE  9 5  6
LModE  7 4  2
PDE 15 8 13

According to the information displayed in Table 2, the informative content of

the sentence-final clauses of the it-clefts in the database is referring in the major-
ity of the cases. What is more, if the results of low-referring information, that is,
of constituents whose head is not referring and whose modifiers/complements
are referring, are computed together with the ones of referring information, the
proportion of either completely or partially referring segments in final position
is even higher. These data lead to the rejection of end-focus since such a prag-
matic principle does not stand up well under the onslaught of the high propor-
tion of available information conveyed by the rightmost clauses of the clefts under
examination. In Atlas and Levinson’s (1981: 16) words, the it-cleft “contravenes the
­convention that old information precede new information”.
As regards the referentiality potential of the other major constituent of a cleft,
namely, the Local CA, the scores in Table 3 evince a situation which is quite dif-
ferent from the one depicted in Table 2. Whereas most final clauses contain avail-
able referents, the Local CAs of the it-clefts analysed are normally non-referring,
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

especially from EModE onwards. These results make it possible to corroborate the
informative characterisation of it-clefts as focusing meta-informative strategies.

6.  Summary and concluding remarks

In Section 1 we defined the concept ‘cleft sentence’, which was restricted to the meta-
informative pattern known as ‘it-cleft’ since pseudo- or wh-clefts were regarded
as identifying or attributive copulative constructions comprising sentence-­initial
wh-clauses realising the function Global CA, and pronominal and th-clefts were
treated as examples of non-restrictive relativisation or even (pseudo-) right dislo-
cation. In Sections 1 and 2 we described the basic structure of an it-cleft as follows:
Global CA in the form of the AnS it followed by a link verb, then the Local CA
and finally either a nonfinite or a finite non-meta-informatively-­centered clause,
the latter being preceded by one of the following introducers: Ø, that, which or
who. Several conditions and remarks seem in order here concerning this structure,
which were discussed in detail in the course of this paper:

(i) existence of a null pointer within the rightmost clause: One of the constituents
of the rightmost clause is promoted to Local CA position.
(ii) categories in Local CA position: Most categories can occur in Local CA posi-
tion, non-contrastive adjective phrases (APs) and verbal constituents being
excluded.
(iii) syntactic functions related to the Local CA segments: The Local CA segments
can be associated with null pointers fulfilling most syntactic functions within
the rightmost clause, with the exception of verb-based functional roles, dis-
juncts and conjuncts. The usual functions realised by the null pointers of
the sentence-final clauses of the corpus were subject and adjunct, which was
accounted for on the basis that the constituents fulfilling these functions can-
not easily be focalised by means of other strategies.
(iv) information: The information conveyed by the Local CA segment is unavail-
able (either new or, as is claimed in the literature, ‘presented as new’ by the
speaker) to the hearer in the majority of the cases. Such an assertion, which
was corroborated by the data, characterised the cleft construction as a focusing
meta-informative device, not conditioned by the given-before-new principle.

The mechanism of clefting has then become a focalising meta-informative strat-

egy which has consolidated itself in the English language quite recently. By means
of this system, the speaker focuses on the referent of the Local CA constituent
that occupies the X-position and places heavy (normally available) information
in final location within the background rightmost clause. The consolidation of the
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

it-cleft construction has been corroborated both quantitatively and qualitatively.

On the one hand, even though the strategy of clefting is not a frequent construc-
tional option in the periods investigated, an important increase in the number of
it-clefts could be noticed from EModE onwards. On the other hand, the qualitative
spread of the meta-informative pattern was corroborated by means of the grow-
ing number of both syntactic functions and categories associated with the Local
CA segment. In fact, with a few exceptions, almost every category can occupy the
position reserved for the Local CA segments, and the Local CAs constituents can
be related to null pointers realising almost every syntactic role in the rightmost
clauses (subject, object, predicative, prepositional complement, adverbial comple-
ment, adjunct).

Sources

ARCHER = A Representative Corpus of Historical English Registers 1990–1993/2000. Compiled

under the supervision of Douglas Biber and Edward Finegan at Northern Arizona Univer-
sity, University of Southern California, University of Helsinki and Uppsala University.
CED = A Corpus of English Dialogues 1560–1760. Compiled under the supervision of Merja
Kytö (Uppsala University) and Jonathan Culpeper (Lancaster University). (The) Helsinki
Corpus of English Texts – Kytö, M. comp. 1996. Manual to the diachronic part of the H­ elsinki
Corpus of English Texts. Coding conventions and lists of source texts. Helsinki: University of
Helsinki (Department of English).
(The) Lancaster-Oslo/Bergen Corpus of British English – Johansson, S. 1978. Manual of infor-
mation to accompany the Lancaster-Oslo/Bergen Corpus of British English, for use with digi-
tal computers. Oslo: University of Oslo (Department of English).

References

Akmajian, Adrian. 1970. On deriving cleft sentences from pseudo-cleft sentences. Linguistic
Inquiry 1: 149–68.
Ariel, Mira. 1996. Referring expressions and the ±coreference distinction. In Reference and Ref-
erent Accessibility, Thorstein Fretheim & Jeanette K. Gundel (eds), 13–35. Amsterdam: John
Benjamins.
Atlas, Jay David & Levison, Stephen C. 1981. It-clefts, informativeness, and logical form. In
Radical Pragmatics. Peter Cole (ed.), 1–61. New York NY: Academic Press.
Ball, Catherine N. 1994a. Relative pronouns in it-clefts: The last seven centuries. Language Vari-
ation and Change 6: 179–200.
Ball, Catherine N. 1994b. The origins of the informative-presupposition it-cleft. Journal of Prag-
matics 22: 603–628.
Biber, Douglas, Finegan, Edward & Atkinson, Dwight. 1994. ARCHER and its challenges: Com-
piling and exploring ‘A Representative Corpus of Historical English Registers’. In Creating
and Using English Language Corpora, Udo Fries, Gunnel Tottie & Peter Schneider (eds),
1–14. Amsterdam: Rodopi.
 Ana Elina Martínez-Insua & Javier Pérez-Guerra

Brömser, Bernd. 1984. Towards a functional description of cleft constructions. Lingua 62:
325–348.
Croft, William. 1991. Syntactic Categories and Grammatical Relations. The Cognitive Organiza-
tion of Information. Chicago IL: University of Chicago Press.
Daneš, Frantisek. 1974. Functional sentence perspective and the organization of the text.
In Papers on Functional Sentence Perspective, Frantisek Daneš (ed.), 106–28. Prague:
Academia.
Davidse, Kristin. 2000. A constructional approach to clefts. Linguistics 38(6): 1101–1131.
Declerck, Renaat. 1988. Studies on Copular Sentences, Clefts and Pseudo-clefts. Leuven: Foris.
Declerck, Renaat. 1997. When-clauses and Temporal Structure. London. Routledge.
Delahunty, Gerald Patrick. 1982. Syntax and Semantics of English Cleft Sentences. Bloomington
IN: Indiana University Linguistics Club.
Delahunty, Gerald Patrick. 1984. The analysis of English cleft sentences. Linguistic Analysis 13:
63–113.
Dik, Simon C. 1997. The Theory of Functional Grammar, Part 2: Complex and Derived Construc-
tions, edited by Kees Hengeveld. Berlin: Mouton de Gruyter.
den Dikken, Marcel, Meinunger, André & Wilder, Chris. 2000. Pseudoclefts and ellipsis. Studia
Linguistica 54(1): 41–89.
Engelkamp, Johannes & Zimmer, Hubert D. 1983. Dynamic Aspects of Language Processing.
Focus and Presupposition. Berlin: Springer.
Enkvist, Nils Erik. 1979. Marked focus: Functions and constraints. In Studies in English Linguis-
tics for Randolph Quirk, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik (eds), 134–152.
London: Longman.
Firbas, Jan. 1992. Functional Sentence Perspective in Written and Spoken Communication.
­Cambridge: CUP.
Givón, Talmi. 1984. Syntax. A Functional-typological Introduction. Amsterdam. John Benjamins.
Grosz, Barbara J. & Sidner, Candace L. 1986. Attention, intentions, and the structure of
­discourse. Computational Linguistics 12(3): 175–204.
Grosz, Barbara J., Joshi, Aravind K. & Weinstein, Scott. 1995. Centering: A framework for
­modeling the local coherence of discourse. Computational Linguistics 21(2): 203–226.
Halliday, Michael A.K. & Matthiessen Christian M.I.M. 2004. An Introduction to Functional
Grammar. London: Arnold.
Halvorsen, Per-Kristian. 1978. The Syntax and Semantics of Cleft Constructions [Texas Linguistic
Forum 11]. Austin TX: University of Texas.
Harold, Bruce B. 1995. Subject-Verb word order and the function of early position. In Word
Order in Discourse [Typological Studies in Language 30], Pamela Downing & Michael
Noonan (eds), 137–161. Amsterdam: John Benjamins.
Hetzron, Robert. 1975. The presentative movement or why the ideal word order is VSOP. In
Word Order and Word Order Change, Charles N. Li (ed.), 347–388. Austin: University of
Texas Press.
Huddleston, Rodney. 1984. Introduction to the Grammar of English. Cambridge: CUP.
Jespersen, Otto. 1909–1949. A Modern English Grammar on Historical Principles. Vols. I–VII.
Heidelberg: Ejnar Munksgaard.
Jespersen, Otto. 1937. Analytic Syntax. London: Allen and Unwin.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Cambridge: CUP.
Lambrecht, Knud. 2001. A framework for the analysis of cleft constructions. Linguistics 39(3):
463–516.
 It-clefts in the meta-informative structure of the utterance in Modern and Present-day English 

Martínez-Insua, Ana Elina. 2011. When discourse matches syntax: On meta-informative

­centering theory and discourse coherence in the recent history of English. International
Journal of English Studies 11(2): 97–117.
Martínez-Insua, Ana Elina & Pérez-Guerra, Javier. 2012. Passivisation and extraposition as
meta-informative strategies: On active-passive variation in the recent history of English. In
Encoding the Past, Decoding the Future. Corpora in the 21st Century, Isabel Moskowich &
Begoña Crespo (eds.), 121–146. Newcastle upon Tyne: Cambridge Scholars.
Pérez-Guerra, Javier. 1999. Historical English Syntax. A Statistical Corpus-based Study on the
Organisation of Early Modern English Sentences. Muenchen: Lincom Europa.
Prince, Ellen F. 1978. A comparison of wh-clefts and it-clefts in discourse. Language 54: 883–906.
Quirk, Randolph, Greenbau m, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A Comprehen-
sive Grammar of the English Language. London: Longman.
Rochemont, Michael S. 1986. Focus in Generative Grammar [Studies in Generative Linguistic
Analylsis 4]. Amsterdam: John Benjamins.
Thompson, Geoff. 2004. Introducing Functional Grammar. London: Arnold.
Traugott, Elizabeth Closs. 2008. ‘All that he endeavoured to prove was’: On the emergence of
grammatical constructions in dialogual and dialogic contexts. In Language in Flux: Dia-
logue Coordination, Language Variation, Change and Evolution, Robin Cooper & Ruth
Kempson (eds), 143–177. London: Kings College Publications.
Van Oosten, Jeanne. 1986. The Nature of Subjects, Topics and Agents. A Cognitive Explanation.
Bloomington IN: Indiana University Linguistics Club.
Visser, Fredericus Theodorus. 1970. An Historical Syntax of the English Language, Part I: Syntac-
tical Units with one Verb. Leiden: E.J. Brill.
Walker, Marilyn A., Joshi, Aravind K. & Prince, Ellen P. 1998. Centering in naturally occurring
discourse: An overview. In Centering Theory in Discourse, Marilyn A. Walker, Aravind K.
Joshi & Ellen F. Prince (eds), 1–28. Oxford: Clarendon Press.
Włodarczyk, Hélène. 1999. Les marqueurs de la validation des énoncés en français et polonais.
Etudes Cognitives / Studia Kognitywne III: 135–162.
Włodarczyk André & Włodarczyk, Hélène. 2006. Focus in the meta-informative centering
theory, La focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
 Discourse coherence and referent
identification of subject ellipsis in Japanese

Shigeko Nariyama
The University of Melbourne, Australia

This paper discusses two ellipsis-related issues using the MIC theory as the
backbone of explanation: (1) ellipsis as a discourse coherence marker, which
answers the question, why use ellipsis (i.e. the pragmatic issue); and (2) referent
identification of ellipsis, which answers the question, how do we know
who/what a zero encoding of ellipsis refers to? (i.e. the semantic issue.) The
former is an issue also related to the speaker’s cognition whether or not to utilize
ellipsis for enhancing a discourse coherence based on the speaker’s assumption
of the addressee’s knowledge and ability to make sense of the zero encoding. On
the other hand, the latter deals with a matter stemming from the other side of
the same issue; namely, the addressee’s cognition to interpret and retrieve the
referential identity of ellipsis. The paper examines the mechanisms of referent
identification of ellipsis found in Japanese, the language that is said to utilize
ellipsis with extremely high frequency but without conventional cross-referencing
systems. It shows that Centre of Attention plays a central role in providing an
adequate and consistent explanation of the mechanism of ellipsis. It also presents
the importance of implicit lexical knowledge of Japanese that contributes to
referent identification of ellipsis.

1.  Introduction

The Meta-informative Centering (MIC) theory developed by Włodarczyk and

Włodarczyk (2006a, 2006b, 2008a, 2008b) places an unprecedentedly strong
emphasis on the importance of pragmatics (meta-information) more than other
linguistic theories. This is so as to provide better and more consistent explanations
for understanding texts and coherence of discourse that are bound to the linear
nature of morphonological form of natural language. As such, Centre of attention
portrayed in the MIC theory is a notion that is able to adequately capture one of
the central issues in linguistics – the mechanism of ellipsis.
Ellipsis is defined broadly as implicitly encoded entities that nonetheless syn-
tactically and semantically constitute an integral part of understanding a sentence
and discourse. Zero encoding of ellipsis means that the working and the referential
identification of ellipsis tend to be intangible and evasive. This is a long standing
 Shigeko Nariyama

issue for languages, such as Japanese and other East Asian languages, as the gram-
mar of these languages freely allows zero encoding of entities without recourse
to the conventional cross-referencing systems. For instance, Romance languages
commonly rely their referent identification on subject-verb agreement features
in terms of person, number and gender. In contrast, the following Japanese sen-
tence taken from an article contains at least four ellipses without any conventional
agreement features displayed elsewhere in the overt part of the sentence. Nonethe-
less, the meaning of the sentence is unambiguous to the speakers of Japanese who
will all interpret the sentence as the given English translation.
(1) 妻に話すと同意してくれた。
Tuma ni hanasu to, douisite kureta.
wife to talk when, agree gave
‘When (I) talked to (my) wife, (she) agreed with (me)’.

This paper discusses two ellipsis-related issues using the MIC theory as its
backbone of explanation: (1) Ellipsis as a discourse coherence marker, which
answers the question, why use ellipsis (i.e. the pragmatic issue); and (2) Referent
identification of ellipsis, which answers the question, how do we know who/what
a zero encoding of ellipsis refers to? (i.e. the semantic issue.) The former is an issue
also related to the speaker’s cognition whether or not to utilize ellipsis for enhanc-
ing a discourse coherence based on the speaker’s assumption of the addressee’s
knowledge and ability to make sense of the zero encoding. On the other hand,
the latter deals with an issue from the other side of the same issue; namely, the
addressee’s cognition to interpret and retrieve the referential identity of ellipsis.
The remaining of this paper is organized as follows to address these two issues.
Section 2 observes the first issue – the role of ellipsis in discourse and its extent of
contribution towards creating a discourse coherence. Section 3 examines the second
issue – the mechanisms of referent identification of ellipsis found in J­apanese, the
language that is said to utilize ellipsis at the extremely high frequency nonetheless
without conventional cross-referencing systems. It shows that Centre of ­Attention
plays the central role in providing an adequate and consistent explanation on the
mechanism of ellipsis. It also presents the importance of implicit lexical knowledge
of Japanese that contributes to referent identification of ellipsis.

2.  Ellipsis as a discourse coherence marker

One of the major functions of ellipsis, or anaphors more generally, that is to say,
the use of reduced referential form instead of repeating the exact same words
 Subject ellipsis in Japanese 

over and over again, is pertained to creating discourse coherence (e.g. Halliday &
­Hassan 1976; Fox 1996). This claim is plausible, in that the use of ellipsis creates
cognitive states of coherence on the grounds that the addressee has to do context
search for the interpretation of the missing referent based on reduced information
provided, and this search links the current sentence to other sentences and meta-
information, and thus creates coherence. This is vindicated by the converse. Fox
(1996) reports that full expressions (not using reduced referential form, including
ellipsis) for inferable referents create paragraph/story boundary and discontinue
coherence.
Speakers in general are not conscious of making their sentences linguistically
or cognitively coherent. Rather they simply do not want to repeat the same words
or what is understood and obvious to the addressees, or such a sentence will sound
too tedious and uneconomical. This works fine in communication because of the
following. As will be elaborated in Section 3 in the discussion of the role of topic in
discourse, the global topic and old information, i.e. the entities with an old meta-
informative status and Centre of Attention, are most commonly encoded as ellip-
sis, and this reduced encoding leads to achieving the optimal coherence. This is an
utterly logical consequence, in that discourse is about talking about ‘something’,
i.e. ‘topic’ and linking sentences with the topic as the pivot, which leads to topic
continuity and hence coherence (Hinds 1983). As a consequence, the speaker is
able to encode the topic as ellipsis and the addressee can identify the ellipsis as
referring to the topic (see Section 3).
Note the differences between coherence and cohesion. Halliday and Hasan
(1976) in their book on Cohesion in English differentiate coherence from cohe-
sion, both of which function to enhance readability of sentences, i.e. ‘easy to
understand/read well’. According to their classification, coherence is created by
meta-linguistic means, such as commonsense knowledge, cognitive state, and
logical argumentation (cause and effect, question and answer, and so forth),
while cohesion is enhanced using grammatical means, such as anaphora and
ellipsis, substitution, and information structures including new/old information
and theme and rheme. That is to say, Halliday and Hasan regard ellipsis as part
of cohesion.
However, as Section 3 will show in explaining the mechanism of referent iden-
tification found in Japanese ellipsis, understanding referential identity of ellipsis
requires not only grammatical means hence establishing cohesion, but also meta-
linguistic means thereby requiring no antecedent either endophorically (within
texts) or exophorically (outside texts), hence establishing coherence. In the regard,
the MIC theory correctly and comprehensively encompasses coherence and cohe-
sion as meta-informative.
 Shigeko Nariyama

3.  Referent identification of ellipsis in Japanese

Japanese is taken as a pilot language in this study for its abundant use of ellipsis
nonetheless without familiar agreement features, such as subject-verb agreement
and clitic pronouns. According to a 1955 research report by the National Language
Institute for Japanese Language, the subject was omitted 74% of the time in con-
versation, 37% in written texts, and 20% in novels in Japanese.1 A recent report on
the topic in 2008 by Ide (forthcoming) showed that the rate of omission was 68.7%
for the subject and 40.1% for the object. In comparison, the rates of its English
counterpart were 15.4% and 8.0% respectively.2
How are speakers of Japanese able to understand the referential identity of
so many ellipses contained in a single sentence in the discourse without cross-
referencing systems, as in (1)? Nariyama (2003) elucidated the mechanisms of
Japanese ellipsis with its approach ranging from lexical semantics to pragmatics,
which are very different from agreement systems found in European languages.
She classified the mechanisms as the interplay of the following three tiers of
device of restrictions: (1) Argument-inferring predicates (Word-level devices),
(2)  Argument structure (Sentence-level devices that specify how to structure
a sentence), and (3) Discourse topic oriented structure (­Inter-sentential-level
devices that specify how to sequence sentences coherently following one
another). These devices impose restrictions on the selection of words and sen-
tence structures and how to sequence sentences. These restrictions, thus in turn,
signal the referential identity of ellipsis. More recently, Nariyama (2009) added
various types of implicit lexical knowledge to Predicate devices that contribute
to identifying the referent of ellipsis. These devices utilize both the informative
and meta-informative means.
Each device is briefly summarised below. For simplicity and logical argu-
mentation of this paper, these devices are explained in the order of Argument
structure (Sentence-level) (in Section 3.1), Discourse topic oriented structure
(Discourse level) (in Section 3.2), and Argument-inferring predicates (Word-
level) (in S­ ection 3.3).

.  See Nariyama (2003: 23–26) for the rates of the subject omission in other genres and the
rates of the object ellipsis.
.  These results are based on experiments using a corpus called Mister O corpus. Subjects
participating in the experiments are asked to look at a series of pictures containing a story line
with no speech. In task 1, they were asked to construct a story. In task 2, they formed pairs to
make conversations using questions prepared by the researchers. The two tasks were done in
Japanese and also in English.
 Subject ellipsis in Japanese 

3.1  Argument structure

Nariyama (2003) vindicated that Japanese sentences must form a certain argu-
ment structure by following what is named ‘the principle of direct alignment’.
This principle specifies that Japanese sentences follow restrictions on the type of
noun that can be the subject and on structuring the sentence so as to express the
selected noun as the subject. That is to say, the subject must be higher than the
non-subject in terms of either the person/animacy hierarchy (Silverstein 1976)
and/or discourse salience (topicality and referentiality including old informa-
tion and definiteness), as illustrated in Table 1. The role of the person/animacy
hierarchy and discourse salience corresponds to Centre of Attention in the MIC
theory, in that an entity with the old meta-informative status is chosen to be the
subject.3

Table 1.  The Principle of Direct Alignment

Japanese sentences must form SB > nonSB

in terms of the person/animacy and/or discourse salience (topicality)

Subject ← → Non-Subject

Person/Animacy [1st Person > 2nd P > 3rd P > animate > inanimate]

Discourse Salience   [topic > non-topic]

These restrictions imposed by the principle of direct alignment are exempli-

fied in the following two sets of Examples: (2) for demonstrating the importance
of the person/animacy for argument structure and (3) of Discourse Salience. (2a)
is a natural Japanese sentence, while (2b) is awkward. When (2b) ‘He found me’ is
rephrased as (2c) ‘I got found by him’ using an intransitive verb, (2c) sounds natu-
ral. This difference can be explained that while (2b) violates the principle, (2c) now
follows the principle that the subject is higher than non-subjects in the sentence in
terms of the person/animacy. When the propositional content (information) is in
inverse order, the inverse verbs -kureru or -iku can be suffixed to the main verb to
indicate that the order of the sentence is in inverse order, as in (2d) (see Nariyama
2003: §4.6 for the treatment of other types of constructions).

.  I thank André Włodarczyk for personally revealing his unpublished work on this issue
in the MIC theory. It shed light on how closely my own research is related to the MIC theory
and how it can be represented using the MIC theory that captures a wider scope of language
mechanisms.
 Shigeko Nariyama

(2a) ✓ [1>3] Watashi-wa kare-o mituketa.  私は彼をみつけた。

I-Top him-obj found (transitive)
‘I found him.’
(2b) ? [3>1] Kare-wa watashi-o mituketa.  彼は私をみつけた。
He-Top I-obj found (transitive)
‘He found me.’
(2c) ✓ [1>3] Watashi-wa kare-ni mitukatta.  私は彼にみつかった。
I-Top him-by found (intransitive)
‘I got found by him.’
(2d) ✓ [3>1 with the inverse verb]
Kare-wa watashi-o mitukete-kureta.  彼は私をみつけてくれた。
He-Top I-obj found (transitive)-Inverse
‘He found me.’
Similarly for the restrictions of discourse salience on argument structure,
(3a) below sounds somewhat strange causing a lack of coherence, while (3b)
is natural. This difference between the two is attributed to the definiteness and
topicality of the two referents: ‘lost child’ and ‘husband’. Unlike many European
languages, the definiteness is grammatically not required and generally implicit
in J­ apanese, except when nouns are accompanied by demonstratives. Note, how-
ever, the definiteness can be inferred from the type of postpositional particles to
mark the subject. Generally, ga is used to mark an indefinite subject noun, while
wa, ­so-called ‘the topic marker’, is used for topicalised/identifiable less indefinite
noun (see Włodarczyk 1980, 1998, 2005, 2007; Włodarczyk & Włodarczyk 2006a).
This difference is statistically reported in Watanabe (1989) that 95% of wa-marked
subjects represented old information and 99.5% definite nouns, while those for
­ga-marked subjects did 45% and 61.6% respectively.
Moreover, possessives are not generally marked unless otherwise the meaning
is believed to become unambiguous or when a focus is placed on the noun. The
possessor of kinship terms and body parts, however, is usually understood as the
speaker or the subject in Japanese (Kuno 1973, inter alia). Hence, ‘lost child’ is less
definite than ‘husband’.
Thus, when the subject with a less definite noun ‘lost child’ is made into the
subject against a more definite noun ‘my husband’ as in (3a), the sentence is wired.
On the other hand, (3b) is natural because the sentence follows the principle that the
subject is higher than the non-subject in terms of discourse salience. However, when
‘lost child’ is marked by the topic marker wa as shown in (3c), it becomes a topic
and increases its discourse salience, and hence it can be the subject of the sentence.
(3a) ?Maigo-ga otto-ni mitukatta.  迷子が夫にみつかった。　
 Lost child -sbj husband-by found (Intransitive)
‘A lost child got found by (my) husband.’
 Subject ellipsis in Japanese 

(3b) ✓ Otto-ga maigo-ni mituketa.  夫が迷子をみつけた。

 husband–sbj Lost child -by found (Transitive)
  ‘(My) husband found a lost child.’
(3c) ✓ Maigo-wa otto-ni mitukatta.  迷子は夫にみつかった。
  Lost child -Top husband-by found (Intransitive)
  ‘The lost child got found by (my) husband.’

These restrictions on argument structure, therefore, in turn signal the referential

identity, when arguments are omitted. Because the subject must be higher than
non-subject in terms of the person/animacy and/or discourse salience, the overt
argument acts as a reference to search for the referent of the ellipsis.
Moreover, there is also a restriction on the order of omission of arguments
when a sentence contains multiple arguments. A set of Examples (4) show that
when the subject is omitted while the object is overt, as in (4a), the sentence
sounds natural. However, the reverse as in (4b) sounds incomplete and something
missing. Arguments both omitted as in (4c) are acceptable.
(4a) ✓ Kare-o mituketa.  彼をみつけた。
 He-obj found (Transitive)
 ‘ϕ found him.’
(4b) ? Watashi-wa mituketa.  私は みつけた。
   I-Top   found (Transitive)
   ‘I found ϕ.’
(4c) ✓ Mituketa. (Transitive)　みつけた。
 ‘ϕ found ϕ.’

The differences among the three sentence types are captured in the principle of
ellipsis – The higher the referent is in terms of the person/animacy and/or dis-
course salience, the more prone it is to ellipsis. This is summarised in Table 2
(Nariyama 2003: 242).

Table 2.  The Principle of Ellipsis

The higher the referent is in terms of the person/animacy and/or discourse salience, the
more prone it is to ellipsis; namely:
Higher referent    ϕ

This principle also helps the addressees to identify what ellipses refer to. That is
to say, when two arguments are omitted as in (4c), the interpretation always f­ ollows
the principle of ellipsis – the argument high on the person/animacy and/or dis-
course salience in Japanese (e.g. ‘I found him’), and the meaning cannot be inter-
preted in the inverse order. One exception to this is when a referent is focused on
providing new information after a Wh-question, as in “Who did it?” “I did ø”.
 Shigeko Nariyama

Thus, the two principles, the principle of argument structure and the principle
of ellipsis, jointly claim – (1) an argument high on the animacy hierarchy and/or
discourse salience is made into the subject, and (2) this subject is most prone to
ellipsis. In other words, a referent with old meta-informative status is made into
the subject and this subject is most prone to ellipsis.
This explains why the subject ellipsis occurs more frequently than the non-
subject ellipsis in actual texts. Table 3 shows statistically from analysing 3718
naturally occurring sentences from various genres. 93.5% of those sentences with
ellipsis have subject ellipses and overt object, and the rest 6.5% have non-subject
ellipses. Among 6.5%, 4.6% of those sentences have both arguments omitted, and
only 1.9% of those sentences violated the principle. This violation is the result of
placing a focus on the overt subject, such as the answer to a wh-question.

Table 3.  The proportion of ellipsis (Nariyama 2003: 246)

Subject ellipsis [ø nonSB V] 93.5%
Non-subject ellipsis   6.5%
  (with SB ellipsis [ø ø V]   4.6%)
  (without SB ellipsis [SB ø V] due to focus, e.g. after WhQ)   1.9%)

In short, the higher an argument is in terms of the person/animacy hierarchy

or topicality, the more likely it is chosen to be the subject, and the more prone it
is to ellipsis, unless the focus is placed on it. That is to say, the referent with old
meta-informative status receiving Centre of Attention is prone to ellipsis. Thus,
these restrictions on argument structure in turn signal the identity of the subject,
and hence the addressees are able to infer the identity of ellipsis.

3.2  Topic-oriented discourse structure

First, observe the following minimal pair of discourse with subject ellipses with
the only difference being the postpositional particle marking the first mentioned
subject shown by the underlines; namely, (5a) with wa and (5b) with ga.

(5a) 「太郎iは　日本に帰ってから　(ϕi) 働いた。　(ϕi) 今忙しくして

いる。」
Taroi-wa nihon-ni kaettte kara, øi hataraita.
Taro-Top Japan-to return after, worked.
ϕi ima isogashikushiteiru.
now busy.
‘Taroi worked after (hei) returned to Japan. (Hei) is now busy.’
 Subject ellipsis in Japanese 

(5b) 「太郎i が　日本に帰ってから (ϕj) 働いた。 (ϕj) 今忙しくし

ている。」
Taroi-ga nihon-ni kaettte kara, øj hataraita.
Taro-sub Japan-to return after, worked.
ϕj ima isogashikushiteiru.
now busy.
‘(I/Someonei) worked after Taro returned to Japan. (I/Hei) is now busy.’

(5a) and (5b) show that ellipted subjects refer to the topicalised subject. The ellipses
in (5a) refer to the topic. On the other hand, (5b) does not have the mention of the
topic, and hence the ellipses do not refer to the non-topicalised subject, instead
refer to the global/discourse topic that must have been mentioned prior to (5b)
or the first person as the default topic. In other words, sentences in Japanese are
sequenced with the topic marked by wa as the pivot and the subsequent mentions
of the same topic are ellipted. This omitted referents referring to the topic makes
the sentences coherent and allows the hearer to infer the referent correctly and
easily. This is illustrated in Table 4.

Table 4. 
Basic rule of Discourse Structure: ‘X-wa … . øx … .’

Once again, this difference in interpretation is attributed to the difference

between the two particles. Ga that mainly marks the subject of the sentence is
used when the current subject is new and different from the previous as well as the
following, e.g. in the case of a discourse with three sentences, [SB1. SB2-ga. SB3].
On the other hand, wa that mainly marks the topic of the sentence is used when
the current subject is different from the previous but the same as the following
which is omitted, i.e. [SB1. SB2-wa. (SB2)] (see Watanabe 1989; Włodarczyk 2005;
Włodarczyk & Włodarczyk 2006a; Nariyama 2009). The subject is omitted when the
current subject is the same as the previous and following subjects. In other words,
the global topic and old information, i.e. a referent with old m
­ eta-informative sta-
tus, is encoded as ellipsis, and this creates the optimal coherence.
Similarly, the ga-marked subject in the subordinate clause indicates a different
subject from the main subject, so that the omitted subject in the second sentence
must be different from the overt subject and probably refers to the topic men-
tioned in the previous sentences (see ii below).
The following is a summary of this section showing that the interpretation of
ellipses derives from the location of ellipsis (in the main clause or subordinate)
and the type of subject marking (wa or ga). (3) is the case of complex s­ entences
 Shigeko Nariyama

involving conjuctives indicating that the adjoining subjects are the same or
­different (see Section 3.3.5, and also Nariyama 2003: §5.5 for the detailed treat-
ment of other means in complex sentences).
(i) ‘[ øx …,] X-wa ….’ (cataphora: 13%)
or ‘X-wa [ øx …,] ….’ (anaphora: 87%)
(ii) ‘X-wa …. [ Y-ga …,] øx ….’
or ‘X-wa …. [ øx …,] Y-ga ….’
or ‘X-wa …. [ øx …,] øx ….’
(iii) ‘{[ øy … SS conjuctive,] Y-ga … }.’

The following discourse, a contrived structure of hypothetical sentences, demon-

strates an example of how the above summary of rules determines the interpreta-
tion of ellipsis:
X-wa …. ø1x …, {[Y-ga … SS,] ø2y … .}
ø3x …. Z-ga …. ø4x …. A-wa …, ø5A ….
(new topic)

3.3  Argument-inferring predicates

Each lexeme conveys considerable amount of implicit semantic and pragmatic
information, much more than any existing dictionaries have descriptions for. In
particular, some predicates in Japanese impose restrictions on type of arguments,
mainly subjects, that they can take. These restrictions can be seen as a type of
cross-referencing in a broad sense, although they differ significantly from familiar
cross-referencing systems. Such implicit knowledge from lexical semantics helps
identify the referent of ellipsis in Japanese. In other words, implicit lexical knowl-
edge is one meta-informative means of identifying referent.
Below describes selected types of implicit lexical knowledge that have a direct
relevance to referent identification of ellipsis.

3.3.1  Restrictions on person

Intentions and emotions can only truly be known to the person who holds them,
and other people can only observe and guess them. The Japanese language thereby
makes clear linguistic distinctions, indicating whether the intentions and emo-
tions are the speaker’s own (i.e. the first person) as in (6a), or others (i.e. non-first
person) as in (6b). The predicates, thus, display subject-verb agreement in terms of
first person or non-first person, providing an identity of the subject.
(6a) ø Uresii.  うれしい　
(sub) happy
‘(I/*someone) am happy.’
 Subject ellipsis in Japanese 

(6b) ø Uresi-soo-da.  うれしそうだ

(sub) happy-look-be
‘(Someone) looks happy (to my observation)’

3.3.2  Restrictions on gender

Some predicates (verbs and adjectives) semantically restrict the gender of the sub-
ject, providing a type of subject-verb agreement in terms of gender. Many of these
reflect the culture of the language. Hence, this type of gender agreement may be
language dependent, and the interpretation of some predicates may be language
specific. For Example (7), the subject ellipsis has two candidates for the ellipses:
John and Mary. But it refers to SHE (Mary), because in Japanese ‘stubbornness’ is
associated strongly with female character (i.e. being stubborn is not lady-like in
the Japanese society), hence giving that interpretation. This implicit knowledge
from lexical semantics is often enough to identify the referent of ellipsis in the
given context.

(7) ジョンとメリーがけんかしている。（？）気が強いな。
John to Mary ga kenka shiteiru. (?) kiga tsuyoina.
and sub argue doing temper strong
‘John and Mary are having an argument. (He/She) is stubborn.’

Some other examples of female subject-referring predicates are: おてんば otenba

(tomboy), 男勝り otokomasari (female acting masculine), 妊娠 ninshin (pregnant),
while for male subject-referring predicates are: 男らしい otokorashi (manlike),
りりしい ririshii (manly), いさぎよい isagiyoi (graceful/manly) (Nariyama 2009: 72).

3.3.3  Fine-grained selectional restrictions: Representative arguments

Some predicate imposes fine-grained selectional restrictions on the type of
arguments that a predicate typically takes. These restricted arguments are
referred to as ‘representative arguments’ (Nariyama et al. 2005a). For ‘arrest’ for
instance, the typical subject is ‘police officer’ and the typical object is ‘suspect of
crime’. Similarly, ‘diagnose’ typically forms ‘medical doctor diagnose patience’s
illness’. As these representative arguments are semantically so related with some
predicates, predicates alone are often enough to identify the referents when
they are omitted. Nariyama et al. (2005a) extracted from dictionary definition
sentences such 10,076 predicates (verbs, verbal nouns, adjectives) that have
typical arguments in Japanese (Nariyama 2009: 68–70). Nariyama et al. (2005b)
further reported the results from an experiment that 78.8% of these representa-
tive arguments are ­readily ­transferable to English, and hence indicated that the
relationship of those representative arguments and predicates may be language
independent.
 Shigeko Nariyama

3.3.4  Restrictions on seniority

Some predicates reflect the seniority of the subject. This information helps identify
the referent of ellipsis.
(I) Use of honorifics
For example, ‘to eat’ taberu can be described as meshiagaru if a senior
­person is eating, and itadaku if a junior person is eating.
(8a) ø Mesiagaru. [subject honorific]
‘(Honouree) eats.’
(8b) ø Itadaku. [non-subject honorific]
‘(Honourer) eats (something offered by or at the presence of
­honouree.)’
(II) High(senior)/low (junior) subject-inducing verbs
Some verbs semantically restrict the interpretation that the subject is
­socially higher [H] or lower [L] than the non-subject. Take an example from
English for a clarity of illustration, verbs like ‘command’ typically have a
subject that is senior and an object that is junior, while verbs like ‘mutiny’
implies that the subject is junior and the object is senior.
(9) [H] “Chairman instruct/command officer”
(10) [L] “Soldier mutiny commander”
This seniority information narrows down the candidates that are amenable
to the interpretation, and hence helps to identify the referent of ellipsis. 189
High-subjects (H) inducing verbs and 160 Low-subjects (L) inducing verbs
have been identified for Japanese verbs (Nariyama 2009: 71).
(III) Use of causative and permission constructions
Constructions such as causative and permission are used when the ­subject
is more senior than non-subject in Japanese (Nariyama 2009: 71). This
signals that the the actor must be more senior to cause a junior to do
­something or grant a junior a permission to do something. Hence, when
­arguments are omitted, the construction types can indicate the identity
of the arguments in terms of seniority, which helps to narrow down the
­correct referents from the context.

3.3.5  Restrictions on same/different subjects in complex sentences

Some lexemes impose restrictions whether or not two subjects in adjoining clauses
are the same or different. This is referred to as switch-reference systems. Switch-
reference systems are generally understood as a syntactic mechanism that indi-
cates whether the subject of a dependent clause is the same as or different from the
subject of the main clause in complex sentences or clause chaining (Wilkins 1988;
Stirling 1993). They are recognised as one of the reference tracking devices that
identify or signal the identity of referent.
 Subject ellipsis in Japanese 

One common means of switch-reference systems is manifested in the use of

different conjunctive particles, and some conjunctive particles in Japanese have
been found to possess this function (Nariyama 2002). For example, the conjunc-
tive particle nagara in (11) signals that the subjects in the adjoining clauses are the
same [SS], while to in (12) implies that the subjects in the adjoining clauses are
different [DS]. Other conjunctive particles that signal SS are -nagara, -te, -si, -tutu,
-ø, tameni, and DS -to, -tara, -ga, -node, yooni.

[SS] (11) ø Terebi-o mi nagara, haha-ga naitei-ta.

(SB) TV-OB watch while[SS], my mother-SB crying-past
‘My motheri was crying, while (shei) watched TV.’

[DS] (12) ø Miru to, haha-ga naitei-ta.

(SB) look when[DS], my mother-SB crying-Past
‘My motheri was crying, when (someonej) looked at (heri).’

Furthermore, some verbs are found to induce the same/different subjects in com-
plex clauses. That is to say, some verbs semantically restrict the subject in the
embedded clause to be the same as [SS] as in (13), or different from [DS] the
matrix subject as in (14).

(13) [SS] “X refrain from (X) doing A”

“X make an excuse for (X) doing A.”
(14) [DS] “X help Y cook.”
“X instruct Y to do A.”

This switch-reference function reflected in some verbs help identify the referents.
57 same-subject verbs and 273 different-subject verbs have been identified in
­Japanese (Nariyama 2009: 72).

4.  As a conclusion: Implicit lexical knowledge for cross-linguistic studies

This paper has observed various aspects of ellipsis in light of discourse coher-
ence and referent identification, and has shown that the notion of Centre of Atten-
tion in the MIC theory captures the core mechanisms of ellipsis. It is claimed
that anaphora including ellipsis is universal and that no human language has
been found to repeat the same referent (Van Valin 1987: 513). Statistically, Gil-
ligan (1987) reported the finding from examining 100 languages that 93 languages
of those permit ellipsis in finite clauses. As such, studies of ellipsis in relation to
coherence constitute an important cross-linguistic study and a useful test base for
cross-lingual comparative studies.
 Shigeko Nariyama

Various language dependent as well as language independent aspects of

ellipsis were presented in this paper. For example, some elements of implicit lexi-
cal knowledge, such as implicit restrictions on gender and seniority, outlined in
­Section 3.3, are language and culture specific.

References

Fox, Barbara. 1996. Studies in Anaphora [Typological Studies in Language 33]. Amsterdam:
John Benjamins.
Gilligan, Gary. 1987. A Cross-linguistic Approach to the Pro-drop Parameter. Ph.D. disserta-
tion, University of Southern California.
Halliday, Michael A.K. & Hasan, Ruqaiya. 1976. Cohesion in English. London: Longman.
Hinds, John. 1983. Topic continuity in Japanese. In Topic Continuity in Discourse: A Quanti-
tative Cross-language Study [Typological Studies in Language 3], T. Givón (ed.), 43–93.
­Amsterdam: John Benjamins.
Ide, Sachiko. Forthcoming. Research project entitled Empirical and theoretical studies on culture,
interaction, and language in Asia.
Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge MA: The MIT Press.
Nariyama, Shigeko. 2002. The WA/GA distinction and switch reference for ellipted subject iden-
tification in Japanese complex sentences. Studies in Language 26(2): 369–431.
Nariyama, Shigeko. 2003. Ellipsis and Reference-tracking in Japanese [Studies in Language Com-
panion Series 66]. Amsterdam: John Benjamins.
Nariyama, Shigeko et al. 2005a. Extracting representative arguments from dictionaries for
resolving zero pronouns. In Proceedings of Machine Translation Summit X, 3–10. Phuket
Thailand.
Nariyama, Shigeko et al. 2005b. Building a cross-lingual referential knowledge database using
dictionaries. In Proceedings of Recent Advances in Natural Language Processing, 354–360.
Sofia Bulgaria.
Nariyama, Shigeko. 2009. How can we know ‘who did what to whom’ in Japanese? The ­Grammar
of omission: Less is more, 「誰が? 誰に? 何を?：日本語の省略がわかる本」Meiji Shoin
明治書院 Tokyo.
Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Grammatical categories in
Australian languages. Robert MW Dixon (ed), 112–171. Canberra: Australian Institute of
Aboriginal studies.
Stirling, Lesley. 1993. Switch-reference and Discourse Representation, Cambridge: CUP.
Van Valin, Robert. 1987. Aspects of the interaction of syntax and pragmatics: Discourse corefer-
ence mechanisms and the typology of grammatical systems. In The Pragmatic Perspective
[Pragmatics & Beyond Companion Series 5], Jef Verschueren & Marcella Bertuccelli Papi
(eds), 513–531. Amsterdam: John Benjamins.
Watanabe, Yasuko. 1989. The Function of “WA” and “GA” in Japanese Discourse. Ph.D. disserta-
tion, University of Oregon at Eugene.
Wilkins, David. 1988. Switch-reference in Mparntwe Arrernte: Form, function, and problems of
identity. In Complex Sentence Constructions in Australian Languages [Typological Studies
in Language 15], Peter Austin (ed), 141–175. Amsterdam: John Benjamins.
 Subject ellipsis in Japanese 

Włodarczyk, André. 1980. 主題から主語へ、そして主語から主題へ　—　ハとガ (Shudai kara

shugo e, soshite shugo kara shudai e – wa to ga) (From Topic to Subject and from Subject
to Topic – wa and ga). In 言語月刊 (Language monthly) 9(8/80) 大修館 Tokyo: Taishûkan.
Włodarczyk, André. 1998. The proper treatment of the Japanese “wa” and “ga” particles. In Pro-
ceedings of the International Workshop on Human Interface Technology 1998. (IWHIT ‘98),
63–70. – Aizu-Wakamatsu, Japan.
Włodarczyk, André. 2005. From Japanese to general linguistics – starting with the ‘wa’ and ‘ga’
particles. In Paris Lectures on Japanese Linguistics, André Włodarczyk (ed.), 27–62. Tokyo:
Kurosio Shuppan.
Włodarczyk, André. 2007. 「日本語の二重主語文 – 発話におけるメタ情報の中心化理論から」
(Nihonogo no nijuu shugo bun – Hatsuwani okeru meta jouhou no chuusinnka riron kara)
(Double subject sentences in Japanese – discourse from the Meta-informative Centering
Theory)、国立国語研究所 Tokyo: The National Japanese Language Institute.
Włodarczyk, André & Włodarczyk, Hélène. 2006a. Subject in the meta-informative centering
theory. Études cognitives / Studia kognitywne 7: 39–64. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2006b. Focus in the meta-informative centering
theory. In La focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André & Włodarczyk, Hélène. 2008a. The pragmatic validation of utterances.
Études Cognitives / Studia Kognitywne 8: 117–128. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2008b. Roles, anchors and other things we talk
about: Associative semantics and meta-informative centering theory. Intercultural Prag-
matics 5(3): 345–365. Berlin: Mouton de Gruyter.
 Structure of centre of attention
in a multi-party conversation in Japanese
Based on the data of a review meeting
concerning a Science Café held in Hiroshima

Miki Saijo
Tokyo Institute of Technology

Applying Włodarczyk A. & Włodarczyk H.’s Meta-Informative Centering

Theory, the paper describes how a multi-party conversation in Japanese is
formed among participants as they keep adding new pieces of information.
It is shown here that the CA structure in the multi-party conversation in this
case is identical to the expanded speech structure of the MIC Theory, as a result
of analyzing discourse examples collected from the actual meeting conducted
in Japanese in a framework of topics and comments as well as central and
peripheral attentions.

Keywords:  Meta-Informative Centering Theory; Multi-Party Conversation;

Science Café held in Hiroshima; Centre and Peripheral Attention; Shared Centre
of Attention

1.  Introduction

The purpose of this paper is to show how various topics and focuses of plural par-
ticipants brought into a Japanese conversation can be controlled to produce a con-
sistent flow of conversation, explaining it with reference to the MIC Theory. The
data to explain this issue is a staff review meeting of science café held in Hiroshima
(hereafter “Hiroshima data”). There were six individuals,1 and each participant
stated his or her opinion prompted by the chair of the meeting, forming a multi-
party conversation.

.  Four persons appear in the example discourse used for the analysis.
 Miki Saijo

2.  Theoretical framework

A multi-party conversation is a conversation held among three or more par-

ticipants. In a multi-party conversation, it often happens that topics brought
up by the participants and focuses that are derived therefrom (topics as old
information) are generated in parallel simultaneously as a series of sequences.
In order for the conversation to be sustained without causing a breakdown
despite such a condition, those simultaneously arising focuses and topics must
be managed in a way that is understandable by all the participants. Two levels
are required concerning the generation of “aboutness” in order to describe this
mechanism. One of them is a level of information generation (1st order) and the
other is a meta-informative level (2nd order) concerning how the p ­ articipants
mutually handle the generated information in their discussions. In this paper,
Włodarczyk A. & Włodarczyk H.’s Meta-Informative Centering Theory, here-
inafter “MIC theory” (Włodarczyk A. & H. 2006a, 2006b, Włodarczyk A. 2007)
is applied to a multi-party conversation conducted in Japanese, to show how
the participants generate the shared centre of attention (CA) in the particular
multi-party conversation. MIC theory is the theory to describe the information
structure (predicate structure) of an utterance, but in this study, we would like
to try to apply it to discourse level by using an idea of ‘extended utterance by
plural participants’.
The previous studies on multi-party conversations conducted in Japanese,
treating the conversations occuring in real social situations as data, have hith-
erto been primarily sociological and psychological analyses as exemplified in an
analysis of miscommunication in work places (Nambu, Harada, Suto, Shigemori,
Uchida 2006), an analysis of how group consensus formation that is required for
executing a task is achieved through cooperation among participants (Morimoto,
Saijo, Nohara, Takagi, Otsuka, Suzuki, Okumura 2006). On the other hand, vari-
ous studies on multi-party interactions have been carried out primarily by artifi-
cial intelligence researchers focusing on improvements to man-machine interface
and human communication support (see Bono & Takanashi 2007 for review).
Those research activities are trying to describe interactions among three or more
participants in terms of paralinguistics such as prosodies and prominence as
well as non-verbal communication such as gestures and eye contact. At present,
the means of description and units of analyses themselves are still the targets of
reseach in those artificial intelligence-related studies, and the question as to how
the analytical concepts accumulated in the discourse research can contribute to
the study of artificial intelligence has yet to be seen through future developments
(Bono & Takanashi 2007).
 Centre of attention in multi-party conversation in Japanese 

As can be seen from the above, the analysis of multi-party conversation in

Japanese is still in the trial and error stage in various academic fields, and this
paper is one of them. This paper attempts an analysis of how a shared centre of
attention is generated in an actual conversation by treating the exchange of short
speech as data.

3.  Background to the Hiroshima data

This data is an excerpt from a review meeting of “Science Café Hiroshima”. Science
Café is a talk event between scientists and non-scientists so that they can get to
know each other. The city of Hiroshima is one of the only two cities in the world
that experienced nuclear bombing in August 1945. The damage from nuclear
bombs was so extensive and tragic that citizens started up activities to fill the gap
between average citizens and scientists in order to prevent science from again
becoming something that citizens could not control. Science Cafe Hiroshima is
one of such activities that were managed by citizen volunteers to make it pos-
sible for scientists and non-scientists to communicate. The theme of the particular
­Science Café held in June 2007 was the role that hydrogen plays as a future energy
source. A scientist presented a speech about challenges for using hydrogen energy
as an energy source for electric power generation and motor vehicles, which was
then followed by a workshop-style discussion on how to use hydrogen energy.
After the presentation and workshop, the participants had a chance to observe
RX-8, one of the only six cars in the world equipped with a hybrid engine which
operates on hydrogen and gasoline. We will refer to this car, RX-8 (see Figure 1),
in the data, which will be shown later.

Figure 1.  Observation of a RX-8

 Miki Saijo

4.  Data

Out-line of the ‘Hiroshima data’ is a multi-party conversation in a meeting with

6 participants, held to review the ‘Science Café Hiroshima’ which took place just
before this meeting. The following is an excerpt of a sequence of utterances related
to RX-8. The reason for the choice of this excerpt is that this sequence is relatively
short and easy to present. Figure 2 shows the transcript of the excerpt, Figure 3 is
the English translation, and Figure 4 is the excerpt in literal Japanese. Translation
into English substantially changes the structure of the utterances.
Figure 4 is the original transcript except that certain things are added in the
areas that are considered missing from a grammatical standpoint. The words in
brackets are the parts that the author added to fill in the missing areas. The follow-
ing is the explanation of the points added in each utterance.
In 01 utterance, the subject “minasan (you).” is added. The reason for adding it
is that we noticed a sentence final particle “yone (wasn’t it?). “Yone (wasn’t it?)” has
a function as a marker of an effort to share certain information or emotion. The
antecedent of “yone (wasn’t it?)” is a speaker’s presumption (that everyone wanted

01 A : Yappari RX-8 (âru ekkusu eito) tte ichidowa untensite mitaidesuyo, ne.
02 B : Soreto Are nenryo, suisono jite, jidôsha janakute tada RX-8ga kitadakede mou.
03 C : Sô desuka.
04 D : Sugoi Gan gan totteta monne.
05 A : Sugoi nanka so, nanka, otakunakanjide, (Warai)Tôkyô môtâshô mitaina kanjidesuyone.
06 B : Konna kamera motte.
07 D : Yokuiru otakuno shônen.
08 C : Â sôdesune.

Figure 2.  Excerpt 1 in transcript

01 A : It was a dream to drive RX-8, wasn’t it?

02 B : And, uh, the fuel, hydrogen, … Everyone was so excited about just seeing the RX-8, itself.
03 C : Is that so?
04 D : T’was awesome! You were taking a lot of pictures, weren’t you?
05 A : Was great, kind o’ like, geeky. [giggle] It was like being in The Tokyo Motor Show.
06 B : Carrying around such a camera.
07 D : Kind of like a typical boy, a geek?
08 C : Uh, more or less.

Figure 3.  Excerpt 1 in English translation

 Centre of attention in multi-party conversation in Japanese 

01 A : [Minasan] yappari RX-8 tte icidowa untensite mitai desuyone.

02 B : Soreto Are nenryo, suisono jite, jidôsha janakute tada RX-8ga kitadakede mo mô
[minna kôfunshite imashita].
03 C : Sô desuka.
04 D : Sugoi, [anatawa shashin wo] gan gan totteta monne.
05 A : Sugoi nanka so, nanka, [anataga shahin wo totteiru sugataga] otakuna kannjide, (Warai)
Tôkyô môtâshô mitaina kanji desuyone.
06 B : Konna kamera motte.
07 D : [Anatawa] yokuiru otakuno shônen [noyou].
08 C : Â sôdesune.

Figure 4.  Excerpt 1 in literal Japanese

to drive RX-8), and the presence of “yone (wasn’t it?)” allows us to assume that
the person A wished to share his presumption with others, so that the subject of
the descriptive part “untensite mitai (wanted to drive)” has to be “minasan (you)”
rather than “I.” We also note that the utterance has no object. Instead, we find a
marker of reference “tte’ for introducing the topic. This part (RX-8tte) showed
a strong prominence in the actual speech, showing that it is indeed the topic.
In 02 utterance, “minna kôfunshite imashita (everyone was excited).“is added.
“Mô” which is an adverb meaning “already (or readily)” is placed here to modify
­“kôfunshite imashita (was excited).” Although this utterance has neither subject
nor verb, the exclamation mark is enough to convey B’s excitement to C. The cor-
rectness of the assumption is clear from the fact that the person C acknowledged
the receipt of the information by responding, “Sô desuka (Is that so?)” in 03. How-
ever, nobody knows if the information is true or false. In 04 utterance, “Monne”
consists of a sentence final participle “mon” that expresses information sharing
to which is added another final particle “ne,” and it represents D’s thinking that
B’s picture taking of RX-8 is clear evidence of his excitement. In 05 utterance, a
noun phrase “anataga shahin wo totteiru sugataga (the view of your taking pic-
tures)” is added. Although it is the topic of this utterance, it is omitted. “Otakuna
kannjide (the geeky feeling)” and “Tôkyô môtâshô (the Tokyo Motor Show)” are
both expressed with strong prominence, but they are not presented as the topic
but rather as the speaker’s comments. This ends the topic introduced in 01, RX-8.
The topic of the following utterances is geeky feeling or geek. The features of the
excerpt are as follows:
Subject is implicit.
Object, verb and adjective may remain implicit.
Noun phrase which shows topic remains implicit.
By prominence on Noun Phrase, participants can infer the implicit topic.
Sentence final particle shows speaker’s attitude to conversation.
 Miki Saijo

In short, in this excerpt at the level of simple utterance, canonical grammatical

elements remain implicit. How, then, can people continue a conversation with so
much insufficiency of grammatical information? In order to maintain a conversa-
tion with at least a certain level of coherence, the participants must be cognizant of
the topic of the conversation, but how can it be achievable in the given example? In
order to think about the mechanism, a concept, “Meta-Information,” is necessary.

5.  C
 entre of attention and peripheral attention in multi-party
conversation

In order to continue a conversation, a shared identification of the “aboutness,” i.e.

what is being talked about, is required in the chain of utterances between the pres-
ent speaker and the previous speaker. Moreover, in order to convey a message to the
hearer or addressee, the speaker needs to mark what he wants to focus on.
According to the MIC theory, the speaker’s CA (centre of attention) is the sub-
ject in a simple utterance, the topic if the utterance is an extended, and the theme if
the utterances is a discourse level (Włodarczyk A. & H. 2006b). What the speaker
wants to focus on is the centre of attention, the individual CA, and it is necessary
to have a shared CA on a level higher than the level of the individual CA in a con-
versation. In the multi-party conversation, the shared CA is divided between the
general theme and the topic. How did the participants of the Hiroshima data mark
the CA and the peripheral attention (PA)?
The four speakers’ central and peripheral centres of attention are shown sepa-
rately where they appear in their utterances in Figure 5. Although the general theme
of the conversation is not explicitly shown, it is understood that it is the “impression
of the science café as a whole” judging from the chain of utterances continuing from
the previous section of conversation. In 01, RX-8 is introduced as a topic with a ref-
erential expression “RX8 tte” and remains as a shared topic in 01–03. The point here
is that it is difficult to determine whether the 01 utterance of “RX8 tte” is a topic or
a subject. It is a new piece of information considering the fact that this noun, RX8,
appears for the first time in this review meeting, and as the subject of this utterance.
On the other hand, however, the participants have seen this hydrogen-driven car
called RX8 before this meeting started, so that they have recognized it although it
appears as a word for the first time in the conversation. In other words, RX8 is an
old piece of information. This is obvious from the fact that an adverb “yappari”
(as you and I know) is placed in front of the verb phrase “notte mitai” (want to
ride in it) as a modifier.2 Thus, in this paper we ­consider “RX8 tte” to be both the

.  Professor Sachiko Ide kindly suggested this point to the author in MIC Conference 2008.
 Centre of attention in multi-party conversation in Japanese 

01 A : [Minasan] yappari RX-8 tte (CA:topic)) ichidowa unten

sitemitai (PA:comment) desuyone (PA:comment).
02 B : soretoare nenryô suisono jite jidôsha janakute tada RX8
(CA:focus) ga kitadakedemo mô (PA:background)[minna
kôfunshite imashita].
03 C : Sô desuka.
04 D : Sugoi, [anatawa shashinwo] (CA:focus) gan gan totteta
(PA:background) monne (PA:background).
05 A : Sugoi nanka so, nanka, [anataga shahino totteiru sugataga] (CA:focus) otakuna
kannjide (PA:comment), Tôkyô môtâshô (PA:comment) mitaina
kanjidesuyone (PA:backgraund).
06 B : Konna kamera motte (PA:comment).
07 D : [Anatawa] yokuiru otakuno shônen [noyou] (PA:comment).
08 C : Â sôdesune.

Figure 5.  Excerpt 1 in literal Japanese with CA and PA indication

CA and the topic in this utterance. Simultaneously, 01 expresses the wish to share
the comment “unten shitemitai” with the participants using the sentence final par-
ticle “yone,” marking it as a peripheral attention. In this paper, we consider that the
predicates, comments as new information and backgrounds as old information, are
peripheral attentions (PA). In 02, the person B establishes RX-8 as the CA by the
repetition of the word, but converts the PA of 01 to a focus conveying the situation
of “minna kôfunshite ita” using the adverb “mou”. In response to the implicit PA in
02 that everyone was excited or moved, a topic, the pictures of RX-8, is presented
in 04 as an expansion of the focus of RX-8, and the fact that the person B is taking
a lot of pictures is presented as PA. In 05, the phrase, the pictures of RX-8, is fur-
ther extended and a topic of “the person B taking pictures of RX-8” is introduced.
But it is not shown explicitly. The CA can be restored from the PAs expressed as
comments such as “geeky feeling” and “like the Tokyo Motor Show.” Figure 6 is a
graphical representation of the topics, focuses, comments and backgrounds in the
chain of utterance from 01 through 08 and how they shift. In the Hiroshima data,
one topic is shown as the individual CA and in the same utterance PAs are also
presented. And the next speaker selects one or more preceding attentions, and adds
comments on them as new information. In this way, in the Hiroshima data, plural
speakers extended a shared topic in a discourse.
Looking at the issues in this way, there is a noticeable similarity (see
Figure 7) between the CA structure of the extended utterance indicated by
Włodarczyk 2007 and the CA structure of the sequence of utterances found in
the ­Hiroshima data.
 Miki Saijo

01A
RX-8 Want to drive Wasn’t it?
CA: topic PA: comment PA: comment

02B
RX-8 Already (everyone was exited)
CA: focus PA: background

03C
Is that so?
PA: background

04D
Photo of RX-8 Taking photo Weren’t you?
CA: topic PA: comment PA: background

05E
View of Your taking Geeky Like being in Tokyo Isn’t it?
Photo CA: Topic PA: comment motor show? PA: comment PA: background

06B
Having a camera
PA: comment

07D
View of Your Boy, a geek
taking photo CA: focus PA: comment explanatory notes
gradation: new
08C white: old
That’s right explicit: solid line
PA: background implicit: dotted line

Figure 6.  Graphical representation of topics, focuses, comments, backgrounds in the excerpt

Extended utterance Sequence of utterances

topic comment topic comment

general shared
CA focus background CA focus background
local individual
CA PA
Włodarczyk (2007)

Figure 7.  Structural isomorphism between extended utterance and discourse

6.  Conclusion

Since this paper remains as an analysis of one example, it cannot be generalized as

an analysis of multi-party conversations in general, but it is safe to say that as far
as this Hiroshima data is concerned, the participants shared the vehicle named
 Centre of attention in multi-party conversation in Japanese 

RX-8, which they can recall, as the CA, explicitly or implicitly, and expanded
the discourse by presenting their comments about it as the PA. Moreover, the
­particular topic-comment and the central and peripheral structure of attention
are identical to the structure of the expanded utterances of the MIC Theory in
the utterance level. It is suggested here that the participants can maintain con-
versation despite the fact that utterances in a multi-party conversation in Japa-
nese often fail to indicate major elements grammatically because the participants
keep repeating the previous speaker’s utterance, or expanding it by comments,
or inviting other parties’ empathy using final particles, so that they adjust their
own CA with other parties’ CA. The technique used in this case, i.e. reviewing the
structure of conversation by supplementing the grammatically missing elements
using markers of other parts of the utterances in order to correlate the newness
vs. oldness of information as well as central vs. peripheral attentions, carries the
risk of arbitrary manipulation, that it cannot be restored if there are too many
missing parts. Also, we note that the excerpt is indexical in that the shared CA
is the vehicle that all the participants saw immediately prior to the conversation
so that there was no negotiation among the participants on that point. Therefore,
it is our intention to analyze in our future research, possibly co-working with
researchers in the field of information processing, whether we can see the struc-
ture of the information management by participants even in a case where there
is no clear-cut reference in the participants’ awareness by assuming a shared CA
in discourse. In addition, we wish to bring the research on self utterance in long,
coherent discourse and expression with references on other parties’ utterances
(Saijo 1999) into this framework to tackle the issue of consistency of discourse in
a multi-party conversation.

References

Bono, Mayumi & Takanashi, Katsuya. 2007. What is necessary in analysis of multi-party interac-
tion? Trends and current state of interaction studies. Transaction of the Japanese Society for
Artificial Intelligence, 703–710. (In Japanese)
Morikoto, Ikuyo, Saijo, Miki, Nohara, Kayoko, Takagi, Kotaro, Otsuka, Hiroko & Okumura,
Manabu. 2006. How do ordinary Japanese reach consensus in group decision making? Iden-
tifying and analyzing “Naïve Negotiation.” Group Decision and Negotiation 15: 157–169.
Nambu, Misako, Harada, Etsuko, Suto, Satoru, Shigemori, Masayoshi & Uchida, Kaori. 2006.
Risk sharing communication in medical settings: Analyses of nurses’ conversation. Cogni-
tive Sudies 13(1): 62–79. (In Japanese).
Saijo, Miki. 1999. The Role of Meta-linguistics in Discourse. Tokyo: Kazamashobo. (In Japanese)
Włodarczyk, André & Włodarczyk, Hélène. 2006a. Focus in the meta-informative center-
ing ­theory. In La focalisation dans les langues, 1–27. Paris: Collection Semantiques,
L’Harmattan.
 Miki Saijo

Włodarczyk, André & Włodarczyk, Hélène. 2006b. Subject in the meta-informative centering
theory. In Etudes cognitive. Studia kognitywne VII, 7–32. Warszawa: SOW, PAN.
Włodarczyk, A. 2007. 「日本語の二重主語文－発話におけるメタ情報の中心化理論か
ら」, Nihongo no ’nijû-shugo-bun’ – hatsuwa ni okeru meta-jôhô no chushinka-riron kara
(The ‘Double Subject’ Theory – from the viewpoint of the meta-informative centering in
utterances). http://perso.numericable.com/andre.wlodarczyk/
Verbal aspect in Slavic languages
between semantics and pragmatics

Hélène Włodarczyk
Université Paris-Sorbonne, Centre de linguistique théorique
et appliquée (CELTA)

The main function of the linguistic category of aspect is perfectly reflected by

the traditional term “aspect” or “view” which means that the speaker chooses
a view of the situation s/he is speaking about. This view of a situation, or
“point of view”, is first of all reflected by an internal analysis of the situation
into parts: moments and stages. This necessary choice can be compared to
that of a centre of attention in order to build an utterance (cf. the definition of
subject and object in Chapter 4 in this volume). As such, aspect is an essential
tool of the meta-informative structure of the utterance. The internal view of
the situation is further completed by external view parameters concerning its
repetition, the modification of its flow or intensity, the composition of several
situations into one complex situation. This approach aims at integrating into
a cohesive whole the great variety of uses described in the huge literature on
verbal aspect in Slavic languages. The ASMIC theory is of great help in dealing
with the blurred borderline between semantics and pragmatics in aspect usage,
making it possible to propose some tentative way out of endless debates on
Slavic aspectology: the problem of aspect pairs, the difference between aspect
and Aktionsart, the amazing differences in the use of imperfective (IPF) verbs
in Slavic languages and the use of the imperfect tense in French or progressive
forms in English, etc. By reference to the three sorts of parameters we have
defined (concerning situation types, situation internal and external view) we
can distinguish precisely the different possible semantic types of perfective
(PF) partners that can be derived from a simple IPF verb in Slavic languages
depending on the type of semantic situation to which the simple verb refers
(in a given context). The reference to the different values of the aspect
parameters also makes it possible to distinguish among derived PF verbs those
which can be considered as pertaining to grammatical aspect, as opposed to
the lexical classes of derived verbs formed with prefixes having not only an
aspectual perfectivising meaning but adding also various (spatial or abstract)
meanings to the root verb.
 Hélène Włodarczyk

1.  Introduction

As concerns the universal concept of “aspect” (independently of its specific realisa-

tions in sundry languages), we define it in close relation to the principle of semantic
partiality and meta-informative structure of human language (Chapters 2, 3 and 4
in this volume): in order to communicate, a speaker is obliged to choose a view of
the situation represented in his/her mind. Verbal aspect is a grammatical category
making it possible to choose a point of view in relation to a mental situation. Thus,
the main function of the linguistic category of aspect is perfectly reflected by the
traditional term “aspect” or “view”1 which means that the speaker chooses a view
of the situation he is speaking about. This view or “point of view” is first of all
reflected by the selection of a component of a situation: a moment or a stage. This
internal view of the situation is further completed by external view parameters
concerning its repetition, the modification of its flow or intensity, the composition
of several situations into one complex situation. In Slavic languages, a rich collec-
tion of different aspect morphemes makes it possible (and compels speakers each
time they use a verb) to select a part of the situation and a modification parameter.
This necessary choice can be compared to that of a centre of attention in order to
build an utterance (cf. the definition of subject and object in Chapter 4 in this vol-
ume). As such, aspect is an essential tool of the meta-informative structure of the
utterance. We propose the AS and MIC theory (Chapters 2 and 3 by Włodarczyk,
A. in this volume) as common theoretical foundations to explain the semantic and
pragmatic usage types of verbal aspect in Polish and Russian. This approach does
not aim at describing newly discovered types of aspect use, but to integrate into a
cohesive whole the great variety of uses described in the huge literature on verbal
aspect in Slavic languages.
Our approach is based on the description of aspect uses in Polish and ­Russian
as representative of Eastern and Western Slavic subfamilies (Włodarczyk H. 1997).
Homophonic prefixes inherited from Proto-Slavic are mostly used with differ-
ent nuances even in closely related languages (e.g. the study of the usage types of
the prefix po- in Czech and Polish by Piernikarski 1975). However, despite many
important differences in detail, all Slavic languages share common general mor-
phological, semantic and pragmatic features of aspect, which make it possible to
oppose globally verbal aspect in Slavic to verbal aspect in other Indo-European
languages. As concerns the latter, we will turn for comparison to some aspect
usage types in French or English, nonetheless keeping in mind the important dif-
ferences between the aspect categories of these two languages.

.  “vid” in Russian is the exact translation of “view”.

 Verbal aspect in Slavic languages between semantics and pragmatics 

The ASMIC theory is of great help in dealing with the blurred borderline
between semantics and pragmatics in aspect use, putting forward a tentative way
out of some endless debates on Slavic aspectology: the problem of aspect pairs, the
need to revisit the distinction between aspect and Aktionsart, the amazing differ-
ences in the usage types of imperfective (IPF) verbs in Slavic compared to that of
the imperfect tense in French or progressive forms in English, etc.

2.  O
 ntology and Semantics: Explaining specific meanings
with universal concepts

Does it make sense to ask again and again what is universal and what is specific in
the verbal aspect of a given language? We consider that aspect belongs to universal
linguistic categories that can be defined by bundles of univocally defined attribute-
feature structures. Such attributes and their different values belong to the ontology
of human languages. The semantic content of a linguistic expression is a function
mapping this expression onto ontological concepts. In order to be able to explain
the semantic content of various aspectual expressions in different languages, we
need a general set of ontological abstract attribute-feature structures from which
we can choose specific subsets of semantic feature structures. We consider lin-
guistics as a theoretic-experimental science which consists of explaining specific
meanings with universal concepts, the relevance of which can be corroborated by
their use in the description of large collections of data (Włodarczyk H 2009). The
bundles of features serve as descriptions of individual uses of aspect forms in con-
text: all uses which could be described by the same feature bundles are considered
as occurrences of the same type: thus they are reduced to one type of use, we call
a usage type of aspect.
It is worth emphasizing that aspect categories in different languages differ
both in form and content so that important discrepancies must be taken into
account. As concerns form: on the one hand, the forms expressing aspect mean-
ings greatly differ from one language family to another; on the other hand, the
boundaries between lexicon and grammar and the degree of grammaticalisation
of aspect meanings are different in various languages. Let us take as an example the
contrast between French or English on the one hand and Slavic languages on the
other. InFrench and English, aspect morphemes belong to verbal flexion and are
amalgamated with tense forms; in Slavic languages, aspect morphemes are deri-
vational and therefore they are present in different tenses and moods of the con-
jugation even in non finite forms. Moreover, several aspectual meanings currently
conveyed by prefixes in Slavic languages are expressed by semi-auxiliary aspectual
verbs in French and English, e.g. the beginning of a situation can be expressed
 Hélène Włodarczyk

by a prefix in Polish and by auxiliary verbs in French and English, cf. utterances
(1, 2, 3) hereafter.
(1) Pol. Nagle zaśpiewali pieśń ludową.
(2) Fr. Soudain, ils se mirent à chanter une chanson populaire.
(3) En. Suddenly they began to sing a folk song.

In order to be able to explain the semantic content of various aspectual expres-

sions in different languages, we need a general set of abstract ontological attribute-­
feature structures (cf. Section 6 of this chapter) from which we can choose a
specific subset of semantic feature structures.

3.  The borderline between semantics and pragmatics

There exists a long-lasting debate concerning the borderline between the seman-
tics and pragmatics of aspect. As a matter of fact, it is always difficult to distinguish
the two domains from one another because the paradox of verbal aspect is that its
meaning seems to be completely pragmatic from the very beginning, since aspect
is literally the “viewpoint” of the speaker on the situation spoken about.
On the other hand, in structural linguistics, most contextually-bound uses
of semantic categories were regarded as belonging to the pragmatic level, taking
it for granted that systemic meanings, identified as the system of semantic oppo-
sitions characteristic of a given language or family of languages (be they gram-
matical or lexical), belong to “langue”, whereas on the contrary all uses in context
were considered as pertaining to “parole”, thus as belonging to the pragmatic level
defined as the “usage” of the linguistic system by a speaker in a speech act. Many
structuralist linguists insisted on the fact that a general definition of the mean-
ing of the category of aspect (on the level of “langue” or language system) is not
sufficient to give account of its numerous uses in context (on the level of “parole”
or language usage), therefore they opposed the general and particular ­meanings
(Jakobson 1932, 1936), the primary and secondary meanings (Kuryłowicz 1972,
1977), or the semantic potential and particular meanings of a grammatical ­category
(Bondarko 1971a, 1971b). But it is important to underline that not all particular
or ­secondary contextually-bound meanings are strictly pragmatic, many of these
uses are semantic.
Indeed, the development of pragmatic studies in logic and linguistics in the
20th century led to a quite new understanding of langue as the linguistic system
and parole as its usage. In the ASMIC framework integrated into the ­methodology
 Verbal aspect in Slavic languages between semantics and pragmatics 

of interactive linguistic research,2 we consider that the access to language sys-

tems obligatorily goes through the experimental exploration and retrieval of large
amounts of language uses in order to reduce them to types. From the ASMIC
perspective, the properly pragmatic uses of a category are those directly related
to the speech act, its agents (the speaker and the addressee), and the old or new
status assigned by the speaker to the utterance or to its constituents. Let us recall
at this point, that in the framework of the MIC theory, the term information
concerns the semantic content of utterances, while meta-information (MI) con-
cerns the presentation of information as it is linearised in utterances and texts,
which is made possible by the choice (by the speaker) of a centre of attention
among the participants of the situation spoken about, and the assignment (by the
speaker again) of the old or new status to the utterance and its centres of atten-
tion (Włodarczyk & Włodarczyk 2008). As concerns aspect, we will consider as
strictly pragmatic (in a narrow sense) the usage types of the category in old and
new utterances.

4.  Aspect and Aktionsart

It is necessary to keep in mind and to insist on the specificity of verbal aspect in

Slavic languages in contrast with the other languages of the Indo-European family:
on the one hand, aspect in Slavic languages is a grammatical category since each
verb form used in an utterance is obligatorily either perfective (PF) or imperfec-
tive (IPF), but on the other hand, this category is characterised by derivational –
not inflectional – markers, the latter specificity having important consequences
as regards the semantics of the category of aspect in Slavic languages. In other
words, although the category of Slavic aspect is binary (opposing only two classes
of verbs: PF and IPF) there is a great diversity of perfective meanings depend-
ing on the prefix which is used with a verb root. Probably because this diversity
of meanings is related to derivational markers, i.e. originally lexical markers,
aspectual meanings expressed by various prefixes were considered as pertaining
to “lexical aspect” (and called Aktionsart). Only the most frequent realisations3

.  Information on the SEMANA software designed for interactive linguistic research
(making use of artificial intelligence tools to build databases) is available on CELTA’s website:
http://celta.paris-sorbonne.fr/anasem/indexIL.html
.  As we will explain below, generally only the PF verbs derived from IPF dynamic telic verbs
with the sense of “reaching the finish moment” are considered grammatical aspect partners.
 Hélène Włodarczyk

of this abstract binary opposition between PF and IPF were granted the status of
“grammatical aspect”.
In the history of Russian aspect theories, V. V. Vinogradov (1947–1972:
379–390) pointed to A. A. Potebnia (1835–1891) as one of the first ­aspectologists
who separated aspect as a binary opposition from Aktionsart which he called
“duration degrees” (stepeni dlitel’nosti). Potebnia based his conception on
­
­diachronic observations; the typically Slavic aspect opposition developed gradu-
ally on the semantic ground of Aktionsart which existed in other Indo-European
languages, in which however the same binary aspect opposition does not prevail.4
In Slavic languages, the ­Perfective/Imperfective opposition is the consequence of
the clustering into only two classes of all Aktionsart verbs based on their abstract
common properties regarding situations (Netteberg 1953, Włodarczyk H. 1997).
The need to treat aspect and Aktionsart as manifestations of one category was
put forward by many linguists in the last quarter of the 20th century5 (among oth-
ers, by Karolak 1997). The binary category of aspect in Slavic languages emerged
out of the great semantic diversity of the derivational lexical markers that are
used as aspect morphemes: prefixes and suffixes. The use of prefixes and suffixes
to express aspectual meanings is a potentiality within many Indo-European lan-
guages but aspectual categorisation developed diversely in different subfamilies. In
Latin (as pointed out by Veyrenc J. [1968] in his Russian grammar), there existed
a frequentative suffix (jacere → jactare) which however did not evolve (transform)
into an imperfectivising morpheme. Likewise, prefixes could be used to express
a more precise, more definite in time (hence non-durative and non-repeated)
meaning than that of the simple verb (jacere → ejacere ‘throw away’). In modern
­Germanic languages this sort of evolution concerns the use of prepositions to cre-
ate verbal affixes, be they prefixed or postfixed or used as so-called verbal particles
to create compound or phrasal verbs. In English, the meanings of verbs with par-
ticles (called “phrasal verbs”) pertain to aspectual semantics (Brinton 1988) and
it would be of great utility for teaching Slavic languages to English speakers to

.  However it cannot be excluded that the phenomen of reducing multiple Aktionsarts to a
binary category is cyclic. Diachronic studies of verbal categories in Germanic languages (Leiss
1992, 2002) suggest that the history of aspectuality in Germanic languages has already passed
through an early phase (in Gothic) of binary aspect opposition expressed by pairs of perfec-
tive and imperfective verbs. Moerover, I notice that the use of the ga- prefix (ge- in modern
German) with the original meaning of “together” to form perfective aspect partners in Gothic
(Leiss 2002) can be compared with that in Slavic languages of the po- perfectivising prefix
with the original meaning of distribution and totalisation.
.  As concerns the history of the concept of Aktionsart in Slavic linguistics, see Włodarczyk,
H. 1997.
 Verbal aspect in Slavic languages between semantics and pragmatics 

undertake a systematic comparison with the so-called Aktionsart of Slavic lan-

guages (Kardela 1997). Several contrastive German-Polish studies (Kątny 1994;
Czochralski 1975; Czarnecki 1998) showed that the meanings of German mobile
verbal prefixes are comparable to the various Aktionsart meanings expressed by
Polish prefixes. Czochralski (1975) suggested that the abstract meaning of the
binary PF/IPF aspect category by contrast to the particular meanings of differ-
ent prefixed or suffixed verbs could be compared to a phoneme and its phonetic
realizations (allophones): different PF verbs derived from one simple IPF verb can
be regarded as “allograms” of their IPF partner. The specificity of Slavic languages
among other modern Indo-European languages consists in the fact that the variety
of Aktionsart meanings is further organised into a binary grammaticalized opposi-
tion of only two abstract meanings (the IPF and PF).

5.  Aspect in Slavic languages as a “hypercategory”

We have proposed (Włodarczyk & Włodarczyk 2001, 2003, 2006) considering

aspect in Slavic languages as a hypercategory in order to include in this category
both what was considered in previous studies as “grammatical” aspect6 on the
one hand and “lexical” aspect or Aktionsart on the other (Agrell 1908). This view
takes account of the diachronic emergence of the abstract binary category of ver-
bal aspect in Slavic languages out of a multiplicity of Aktionsart classes of verbal
lexemes. In other Indo-European languages there exists also a great diversity of
verbal morphemes (used as prefixes, suffixes or postfixes) which express similar
meanings (e.g. “begin”, “interrupt”, “continue”, etc., cf Brinton 1988; Ter Meulen
1995; Kątny 1994) but which do not give rise – at least in the synchrony of modern
European languages – to a binary opposition of two abstract meanings: the Perfec-
tive (PF) and Imperfective (IPF) aspect.
The concept of hypercategory makes it possible to explain that in Slavic lan-
guages, the perfective meaning never occurs without added semantic features
(e.g. of beginning, result, interruption, intensity, etc.) and therefore, as concerns
a majority of verbal roots, perfectivity may be expressed by several different per-
fective verbs derived from the same root. On the basis of an investigation of the
Polish frequency dictionary (SFPW 1980), we were able to show (Włodarczyk &
­Włodarczyk 2001) that what was generally considered as “grammatical aspect”
was the most frequently used type of perfective derived verbs, i.e. PF verbs with

.  Also called “viewpoint aspect”, cf. Abraham 2013.

 Hélène Włodarczyk

resultative aspectual meaning derived from IPF telic action verbs (in the MIC
framework these verbs are called “ordinary process” verbs, see Section 6 hereaf-
ter). Indeed, the “resultative” meaning seems to be considered the prototypical
meaning of the PF aspect because it is very frequent. Nevertheless one should
not forget that not all perfective verbs express the result of a telic active situation;
using the semantic features designed in a previous study,7 we describe “resultative”
PF verbs as verbs denoting the finish moment of an ordinary process and/or, in an
appropriate context, the after stage which starts after the finish moment. However,
the frequency of ordinary process verbs (generally called telic action verbs) in text
corpuses is so important that in Slavic aspectology it kept out of sight the existence
of less frequent state or refined process verbs whose perfective derived partners
express most frequently either the beginning (the start moment or begin stage)
or the interruption of a state or process in one of their stages. This is the reason
why in Slavic grammars static and refined process verbs are mostly considered as
imperfectiva tantum; as a matter of fact since these types of verbs have no differ-
entiated stages nor moments, they lack a PF partner pointing at a finish moment
(cf. the list of imperfectiva tantum verbs in the Polish grammar of the Academy
of science, Laskowski 1998). Moreover, the other way round, all perfective verbs
with any other meaning than that of result are considered as perfectiva tantum, as
a consequence such verbs are cast aside from the category of “grammatical aspect”
into that of “lexical aspect” (Aktionsart). As we shall see in Section 7 below, the
description of Slavic verbs based both on the concept of hypercategory and on the
universal semantic features proposed in the ASMIC framework (Chapter 2 in this
volume and Section 6 hereafter) makes it possible to reduce the huge number of
perfectiva and imperfectiva tantum to a few peripheral verbs. In our view Imperfec-
tiva tantum stricte sensu are IPF verbs with the meaning of abstract static relations,
e.g. Pol. znaczyć IPF, Rus. znachit’ IPF, “to mean”; Pol. należeć, Rus. prinadlezhat’
IPF, “to belong”). In our approach, on the other hand, perfectiva tantum are mostly
those event verbs which lack suffixal IPF derived partners. However, there exist
event PF verbs from which it is possible to derive an IPF partner denoting the
habitual repetition of the event. The list of Polish and Russian perfectiva tantum
verbs in the ASMIC restrictive acceptance is even shorter than that of imperfectiva
tantum in the same restrictive acceptance. In this perspective, it is possible to give
an exhaustive enumeration of both imperfectiva and perfectiva tantum verbs for
each Slavic language. Nevertheless, due to historically motivated morphological

.  All terms in italics are the labels of the semantic features designed for the aspect category
in Włodarczyk & Włodarczyk 2003. We give a brief summary of theses features in §6 imme-
diately hereafter.
 Verbal aspect in Slavic languages between semantics and pragmatics 

restrictions, the list is not exactly the same in all Slavic languages8 although the
(im)possibility of deriving a certain type of aspectual partner relies basically on
the same semantic conditions in all Slavic languages.

6.  Universal semantic features for the description of aspect usage types

To describe the different usage types of aspect forms we use three kinds of seman-
tic features: Situation Type, Situation Internal View and Situation External View.9
First, we interpret each contextual use of a verb form by reference to a formalised
classification of situations among which we distinguish states and actions, which
are further divided into events and processes that are of two sorts: ordinary or
refined, i.e. consisting in repeated quanta of the same process (Table 1 of Chapter 2
in this volume, repeated hereafter).

Table 1.  Hierarchy of semantic situations

Characteristic SITUATIONS
properties Static situations Dynamic situations (ACTIONS)
(dimensions)
STATE EVENT ORDINARY PROCESS REFINED PROCESS

Space (3D) + + + +
Time – + + +
Progression – – + +
Granularity – – – +

Most semantic classifications of situation types proposed for Slavic languages

(Avilova 1976; Laskowski 1998) and other languages (the most famous for English
by Vendler 1967) have much in common with that proposed by Włodarczyk A.
(2003) (Chapter 2 in this volume); they differ mostly in that they take into account
not only the frame of the situation itself but also the participants of the situation

.  Verbs recently borrowed from non Slavic foreign languages are very often either imper-
fectiva tantum or biaspectual verbs (i.e. the same form may be used alternatively in different
contexts as perfective or imperfective).
.  In former versions of this theory of aspect (Włodarczyk & Włodarczyk 2003, 2006) we
used the term “situation control” in place of situation external view. We decided to change this
term because, in linguistics, “control” is mainly used to refer to the ability of an active agent to
“control” a situation consciously and voluntarily.
 Hélène Włodarczyk

and the role they play in the situation. The classification we use keeps apart frames
and participants playing roles in situations; however the relevance of participants
(and even anchors) for aspectual interpretation is taken into account in the analy-
sis of the context in which a verb occurs. The parameters used in Table 1 organise
situations into an ordered set.
In addition to situation types, we describe aspect uses with two sorts of seman-
tic features concerning the internal and external view of the situation. The internal
view of the situation is the point of view consisting of the selection of a moment
or stage (Figure 1).

MOMENTS
initial start enter exit finish terminal

< before > < begin > < run > < end > < after >
STAGES

Figure 1.  Sequential cycle of simple dynamic situations (ordinary processes)

The external view of the situation is made up of three types of features con-
cerning (1) the quantification of the situation (is it repeated or not), (2) the modi-
fication of the flow and the modification of intensity and (3) the composition of
several situations into one situation.10
For the description of Polish and Russian verbs we distinguish five different
flow modifications as follows:

1. resumption of a situation already begun and interrupted: resume

2. continuation (beyond the expected term) of a situation already in progress:
keep
3. interruption of a situation in progress in one of its stages: interrupt
4. unfolding of a situation from the first to the last moment, including all stages:
trans
5. unfolding of a situation intermittently: off-and-on

In many languages, the intensity modification (by reference to normal intensity) is

expressed by adverbs but in Slavic languages prefixes can express this meaning. In
Polish and Russian we found only three types of intensity modification: uniformly
strong or weak (stable) or increasing (variable) intensity.
Two or several situations can be composed into one. The component situa-
tions may be parallel or sequential.

.  A description of these features can be found in Włodarczyk & Włodarczyk 2003, 2006.
 Verbal aspect in Slavic languages between semantics and pragmatics 

The external view features are shown in Figure 2.

EXTERNAL VIEW

REPETITION MODIFICATION COMPOSITION

Flow Intensity Sequential Parallel

Figure 2.  Parameters of situation external view

What we call external view corresponds only partly to the verb nuances which
are generally named Aktionsart in Slavic aspectology. Our concept of external view
is slightly narrower in that we consider that any verb nuance called an Aktionsart
but concerning the internal flow of the situation should be classified as belonging
to the internal view, i.e. to aspect in the narrow sense. For instance, we consider
that the Aktionsart verbs known as “ingressive” verbs, which draw attention to the
beginning of the situation (either the start or enter moments), belong to the inter-
nal view. On the other hand, we claim that any semantic nuance which modifies
the flow of a situation, its normal sequential cycle, belongs to the “external” view.
Among such nuances we put first of all any of the five modifications of the normal
flow of a situation listed above: interrupt, resume, off-and-on, keep and trans. The
modifications of intensity (strong or weak, or increase) are generally described as
Aktionsart (“augmentative” or “attenuative”) but it must be emphasized that they
always combine with the internal view, i.e. with the selection of a moment or stage.
Moreover, in Slavic languages, an intensity modification is also very often associ-
ated with a flow modification. In the third type of internal view parameters we put
the composition into one complex situation of several similar situations. On the
one hand, we use the composition parameter to give an account of verbs which
are considered in Slavic aspectology as belonging to the “distributive Aktionsart”:
the same situation is repeated several times by the same active participant but on
different passive participants or by different active participants in different places
or times, cf. utterances
(4) Pol. Wzsyscy pasażerowie już powychodzili PF.
(5) Rus. Vse pasazhyry uzhe povychodili PF.
“All the passengers have already come out one after the other.”

On the other hand, composed situations concern verbs in which the prefix points
at one situation and the root at another one, e.g. in the following verbs: Pol.
wyciągnąć PF and Rus. vytjanut’ PF, “to pull out”, the prefix points at the path of
 Hélène Włodarczyk

motion (out of something) and the root at the manner of motion (pulling). All
verbs expressing a composed situation are perfective because they combine the
composition parameter with the selection of the finish moment of the last situa-
tion which is composed into the complex one.
We must emphasize that our approach differs from the Slavic theory of
Aktionsart in that we treat the external view parameters as ontologic concepts not
as classes of derived verbs. As a matter of fact, in order to describe the semantic
content of verbs which are generally considered in Slavic aspectology as Aktion-
sarts we use both internal and external view parameters. As an example let us look
at the description of so-called “augmentative” Aktionsart verbs.

Pol. Nie wyspałem się. Wczoraj póżno w nocy u sąsiqdów rozpłakało się
(6) 
dziecko i już nie mogłem zasnąć.
Rus. Ja ne vyspal’sja. Nochju u sosedov rebënok razplakalsja i ja uzhe potom
(7) 
nie mog usnut’.
“I could not get enough sleep. Last night my neighbours’ child began crying
loudly and I could not fall asleep.”

We describe the complex meaning of the verb “rozpłakać się PF, (generally called
“augmentative” Aktionsart) using the following parameters:
Type of situation: refined process
Internal view: enter moment
External view: not repeated, increasing intensity, flow modification: keep.

The chosen parameters give a much more precise account of the meaning of the
verb than the traditional label of “augmentative” Aktionsart. As a matter of fact
such a verb denotes not only the increasing intensity of the situation but also the
fact that the situation has reached the enter moment (last moment of the begin
stage and first moment of the run stage) and that the run stage is continuing
beyond its normal duration.
All the nuances described by the three sorts of parameters we mentioned (sit-
uation types, internal and external view) are mostly expressed in Slavic languages
by amalgamating morphemes (prefixes and suffixes) in which various types of
these meanings are combined, as in the Examples (6) and (7) on which we have
just commented.
We use these three types of features to give an account of the numerous con-
textually-bound semantic usage types of the aspect category: aspect features make
it possible to typify individual uses (occurrences) into usage types. Each aspect use
can therefore be described by a semantic feature bundle consisting of two parts:
situation internal view and situation external view. All uses described by the same
feature bundle belong to the same usage type. The situation type is considered as
 Verbal aspect in Slavic languages between semantics and pragmatics 

a condition for the usage of aspect. Thus we proposed the following formalised
definition of aspect usage:
Aspect Usage = {Situation Internal View, Situation External View} condition:
Situation Type

It is important to emphasize that the category of aspect plays an important part in

the meta-informative construction of utterances as it makes it possible to choose a
view of the situation spoken about. In particular, the analysis of a dynamic situation
into moments and stages makes it possible to select11 one of these moments or stages
as the “centre of attention” to which the speaker wishes to draw the hearer’s atten-
tion. Thus, subject and object (cf. Chapter 4 in this volume) are grammatical devices
pointing to centres of attention in an utterance, while aspect is a device pointing at a
centre of attention (one moment or stage) within the verbal situation itself.
As an example, the “resultative” meaning of the PF aspect can be considered
as a semantic contextual usage: it is possible to describe it as the emphasis put, in
an appropriate context, on the after stage of an ordinary process having reached its
finish moment.
(8) Pol. Wysprzedałem wszystkie bilety.
“I have sold out all the tickets.” implying: “I now have no tickets left.”

In Example (8), the after stage is the state when there are no tickets left, the process
leading to this stage is that of selling tickets. Likewise, other contextual meanings
related to the semantic features of the aspect category (the internal and the exter-
nal view features) belong to the semantics of the category.
In order to test the above mentioned features, we are conducting an experi-
ment which consists of describing aspect uses in Polish with the SEMANA soft-
ware12 comprising a database builder and computational data analysis tools (formal
concept analysis, rough set analysis, statistical data analysis). The computer-aided
method of acquisition of semantic knowledge (Włodarczyk A. 2007, 2009) con-
sists of collecting numerous contextual uses of a linguistic category and describ-
ing each of them by a feature bundle. The interactive method makes it possible
to add or remove from the database builder a feature or one of its values, when
it appears necessary for the relevant description of the corpus of examples. The
linguist c­ ontinues to describe more and more samples until no new ­configuration

.  It is necessary to nuance the choice of the speaker: when aspect belongs to grammaticized
resources of a given language, the “choice” of a moment or stage in the situation analysis is
very often restricted by stereotyped communication habits which impose on a speaker’s mind.
.  Information about interactive linguistics and the SEMANA software can be found at
〈http://celta.paris-sorbonne.fr/anasem/indexIL.html〉
 Hélène Włodarczyk

of features is added; when this is achieved, the “general” meaning of the category
will be defined as the common core of features used to describe the different uses.
Usage types are obtained when reducing all identical uses to only one by deleting
duplicates.
This method is currently being applied to the description of context uses of PF
and IPF verbs in Polish in order to discover experimentally the configuration of
features which characterize, respectively, the PF and the IPF aspect (Włodarczyk H.
2009). Each configuration characteristic of one of the two aspects consists of a core
and a periphery. The core can be interpreted as corresponding to the most typi-
cal and frequent usage types, the periphery to less frequent and less typical usage
types. With respect to internal view parameters (the analysis of the situation into
moments and stages), IPF verbs mostly denote stages, whereas moments charac-
terise more frequently the usage of PF verbs. However it is important to note that
some PF verbs (mostly those derived with a prefix which are ususally classified as
“Aktionsart verbs”) point also at stages but these stages are additionally character-
ised by one parameter of the external view (repetition, modification of flow and
intensity or composition of situations). For instance, the combination of the inner
stage of an ordinary process (the run stage) with the interrupt parameter leads to
verbs which do not point at the finish moment but nevertheless are PF verbs, called
“limitative” Aktionart verbs, e.g. the verbs Pol. poczytać, Rus. pochitat’ (“to read for
a while”) are both perfective but they are used to denote ordinary processes that
are interrupted in their run stage, hence before they reach their finish moment.
The combination of the run stage with the interrupt parameter is comparable to
the combination of the same interruption parameter with the parameter of unana-
lysed situations (considered as wholes) expressed by IPF verbs denoting states and
refined processes. Since IPF verbs denoting states and refined processes simply do not
have any finish moment, it is impossible to derive from them a PF verb pointing at
the finish moment; however it is possible to derive from these IPF verbs aspectual
PF partners with the meaning of interruption of the whole situation (see §7.1 and
7.2 hereafter). Thus, when collecting samples of aspect uses (Włodarczyk H. 2009)
we were able to record that not all verbs denoting a stage are imperfective verbs,
some are perfective but, in that case, they combine the stage meaning with another
parameter, e.g. interrupt, resume, intensity, composition of situations.

7.  Aspect pairs and aspect clusters

It is worth underlining that the concept of hypercategory can serve as formal foun-
dation for the notion of “verb arsenal” (“arsenal verbal”) proposed by [Sémon J.-P.
1986] or that of “verb clusters” proposed by [Janda 2007] to replace the traditional
 Verbal aspect in Slavic languages between semantics and pragmatics 

concept of aspect pair in which an imperfective verb is associated with one derived
perfective partner only. As a matter of fact, an aspect pair is only a particular real-
ization of the more general case of aspectual derivation which consists in an aspect
cluster: one simple IPF verb has generally more than one PF partner (Włodarczyk
& Włodarczyk 2001; Włodarczyk H. 2009). As a consequence of the definition of
PF Aspect as a hypercategory, we consider that all those prefixed PF verbs that can
be derived from one simple IPF verb – adding any aspectual nuance13 (beginning,
limitation, intensity, etc.) to the root verb – should all be considered as aspectual
partners of the root verb. We call such a group of verbs an aspect cluster, e. g. from
the simple Russian verb pisat’ IPF (to write) it is possible to create several derived
perfective verbs: napisat’ PF (to write a text up to its end), this verb belonging
to the semantic class of so called “resultative” perfectives ; popisat’ PF (to write
during a certain time), this verb belonging to so called “limitative” perfectives;
razpisat‘sja PF (to write too much), this verb being an example of so-called “inten-
sive” perfectives.14 As we mentioned above, in the traditional treatment of aspect
as opposed to Aktionsart, only resultative perfectives are considered as “real” or
“proper” aspectual partners of the simple IPF verb from which they are derived
with a prefix. On the contrary, we consider that it is possible to derive from an
IPF simple verb more than one prefixed PF verb and that all derived PF verbs can
be ­considered as grammatical PF partners on condition that the prefix does NOT
change the lexical meaning of the IPF root verb but adds to it only an aspectual
nuance. We consider as aspectual those nuances which correspond to one of the
values of the aspect parameters presented in Section 6 in this chapter. The refer-
ence to the different values of the aspect parameters we have defined makes it
possible to distinguish among derived PF verbs those which can be considered
as pertaining to grammatical aspect, as opposed to the lexical classes of derived
verbs formed with prefixes having not only an aspectual perfectivising mean-
ing but adding also various (spatial or abstract) meanings to the root verb. As
an example, we consider a derived PF verb the prefix of which adds the sense of
the start moment to the meaning of its IPF simple verb as a “grammatical aspect”
partner; this is the case of verbs Pol. zapłakać PF and Rus. zaplakat’ PF, “begin to
cry”, formed by adding the prefix za- to IPF simple verbs Pol. płakać IPF and Rus.
plakat’ IPF, “to cry”. However the same prefix za- can be used to create a derived

.  In this approach, we define as “aspectual nuance” one of the aspect parameters defined in
the MIC theory (Section 6 in this chapter).
.  We use the expression “so-called” because in our approach based on the description of
aspect semantics using feature structures, we replace each Aktionsart name or aspectual label
with a bundle of features.
 Hélène Włodarczyk

PF verb with a new lexical meaning from another simple IPF verb; eg. in Polish
zamówić PF, “to order”, from verb mówić IPF “to speak”; in Russian zakazat’ PF
“to order (something)” from verb kazat’ IPF “to give orders (to some one)”. Thus
aspect clusters are a subclass of all derived PF verbs created by the adjunction of
prefixes to simple verbs.
Aspectual clusters can be made up of only one PF partner or up to six or seven
PF partners with various prefixes. E.g. the simple Polish verb spać IPF (to sleep) has
several perfective partners which are used in different contexts depending on the
semantic feature added by the prefix: pospać PF (to sleep during a certain limited
period of time), wyspać się PF (to sleep as much as needed to feel well), zaspać PF (to
miss something because of sleeping too long, oversleep), odespać PF (to recover sleep-
ing, sleep off), dospać PF (to keep sleeping until a certain moment), przespać PF cały
film (to spend a whole period of time sleeping, sleep over). It is sometimes possible to
further derive an IPF partner by suffixation from a prefixed PF verb belonging to
an aspect cluster. E.g. the verb wysypiać się IPF may be derived from wyspać się PF
“to get enough sleep” and used as its IPF partner in contexts of habitual repetition
as in the following utterance.

(9) Pol.
On się wysypia tylko podczas weekendu.
he Pr 3pers N sg sleeps enough V IPF pres 3pers sg only during week-end
“He gets enough sleep only during the week-end.”

As a matter of fact, we admit (Włodarczyk H. 2003, 2008, 2009) – as many scholars

did before us (e.g. Veyrenc 1980) – that one should continue to treat as aspectual
pairs those consisting of one PF and one IPF verb which is derived from the PF
verb by a suffix. Derivating an IPF verb from a PF verb using a suffix produces
only one apectual partner. E.g. the suffix -a- is the imperfectivizing morpheme in
the following pairs: Pol. rzucić PF & rzucać IPF, Rus. brosit’ PF & brosat’ IPF, “to
throw”. The most productive imperfectivizing suffix is the morpheme -ywa- in
Pol., -yva- in Rus., e.g. in the following pairs: Pol. przepisać PF & przepisywać IPF,
Rus. perepisat’ PF & perepisyvat’ IPF, “to copy”.
The concepts of hypercategory and aspect cluster (in place of the too restrictive
notion of aspect pair) make it possible to reconsider those Slavic verbs which were
considered to have no aspectual partner of the opposite value (named perfectiva
and imperfectiva tantum).
The semantic features defined in the AS theory (Włodarczyk A. 2003;
Włodarczyk & Włodarczyk 2003, and Chapter 2 in this volume) lead to sev-
eral innovative explications of the contextual semantic usage types of aspect in
Slavic languages. By reference to the three sorts of aspect parameters ­concerning
 Verbal aspect in Slavic languages between semantics and pragmatics 

s­ituation types, situation internal view and situation external view we can dis-
tinguish precisely the different possible semantic types of PF partners that can
be derived from a simple IPF verb in Slavic languages (see Piernikarski 1969),
depending on the type of semantic situation to which the simple verb refers (in a
given context). The four kinds of semantic situations we defined, i.e. state, event,
ordinary process and refined process, allow different semantic types of derived PF
in Slavic languages. It is generally assumed in the Polish and Russian aspectologi-
cal ­literature15 that “real” or “exact” perfective partners (having exactly the same
lexical ­meaning as their imperfective root) are prefixed PF verbs (often called
“resultative ­perfectives”) pointing at the finish moment of ordinary process verbs.
However, as shown above in §5, in the derived PF clusters of ordinary process IPF
verbs, the derived PF verb pointing at the finish moment, called “result” of the pro-
cess is the most frequent (Włodarczyk & Włodarczyk 2001), but generally NOT
the only possible one. As mentioned above in §5, such resultative verbs cannot
be derived either from stative verbs or from refined process verbs which, for this
reason, are generally considered as imperfectiva tantum. If we rely both on the
formalised list of aspectual semantic features defined in the AS theory and on the
two step concept of aspect as a hypercategory, it is possible to put an end to end-
less debates16 concerning the semantic identity of a derived aspectual partner with
the base verb. As a matter of fact no derived verb is completely identical with the
base verb and each type of semantic situation that can be expressed by a base verb
requires ­different aspect partners.

7.1  The perfective partners of stative IPF verbs

First, let us look at the possibility of deriving PF partners from IPF static verbs
which are generally considered as imperfectiva tantum. It is obvious that it is
impossible to derive resultative PF partners from stative verbs since the situation
they refer to does not undergo any change as time passes, which excludes the pos-
sibility for such a situation to evolve towards a finish moment. Nevertheless Slavic
languages can form and use derived PF verbs with another meaning than that of
result. As a matter of fact, it is often possible to derive from an IPF state verb a
PF prefixed verb with the added meaning of limitation of the state duration; both
Polish and Russian use the prefix po- with this meaning, e.g. Pol. poleżeć PF, Rus.
polezhat’ PF, “to lie”; Pol. posiedzieć PF, Rus. posidet’ PF, “to sit”; Pol postać PF, Rus.

.  Avilova 1976; Laskowski 1998.

.  For details, see Włodarczyk, H. 1997, p. 40–50 “La Paire aspectuelle et le sens lexical”,
where the most important works on the topic are briefly outlined.
 Hélène Włodarczyk

postat’ PF, “to stand”. Verbs derived with the prefix po- and conveying the mean-
ing that a situation is limited in duration were considered as “non ­grammatical,
­limitative” Aktionsart derived from “imperfectiva tantum stative verbs”. These
“limitative Aktionsart” verbs, be they derived from stative, ordinary or refined
process IPF verbs, were for the first time treated as “congruency perfectives”17 by
(Sémon 1986) and named “aspectual Aktionsart” by (Mehlig 1994).

7.2  The perfective partners of refined process IPF verbs

A refined process consists of a series of sequential identical grains of the same
situation repeated one after the other. Such situations have no “natural” finish
moment: they are not telic situations and therefore, it is impossible to derive a per-
fective “resultative” partner verb denoting the finish moment and the after stage
of the process. The possible perfective partners of imperfective verbs denoting a
refined process are verbs referring either to (a) the interruption of the refined pro-
cess after a certain duration as in Example (10) below, or (b) to only one of the
grains of the refined process,18 see Example (11).

(10) Pol. pokołysać (się) PF, Rus. pokachat’(sja) PF, “to swing for a certain time”;
Pol. pomachać rękoma PF, Rus. pomachat’ rukami PF, “to wave arms for a
certain time”.

(11) Pol. machnąć PF, Rus. maxnut’ PF, “to wave once”.

It goes without saying that other, although less frequent, PF verbs are sometimes
derived from refined process verbs, e.g. one can derive from the Pol. IPF verb
machać, Rus. machat’ (“to wave“) the PF derived verb with intensity meaning Pol.
namachać się PF, Rus. namachat‘sja PF (“to wave a lot and get tired of it“).
Discussion of the lexical or grammatical status of derived PF verbs of types
(a) and (b) has occupied the community of Slavic aspect specialists for at least
half a century. As far as I know, the proposal to treat the so-called limitative verbs
(with prefix po-) as PF partners appeared first in (Sémon 1986). The treatment of
“semelfactive” verbs (with suffix -nu- in Russian) as perfective partners is argued
for in (Xrakovskij 1997).

.  Such derived prefixed verbs are used as PF partners of the simple IPF verb when it is
impossible to derive a resultative partner.
.  Such verbs are generally called “semelfactive Aktionsart” verbs.
 Verbal aspect in Slavic languages between semantics and pragmatics 

 e imperfective partners of event verbs: The problem of perfectiva

7.3  Th
tantum
The suffixal IPF verb derived from a simple PF verb expressing an event can be
used only in one type of usage characteristic of IPF verbs: the habitual repetition
type. The usage of an IPF verb as referring to the run stage of an ongoing ordinary
process is impossible, because events are characterised by the lack of time progres-
sion between the start and the finish moment. E.g. the imperfective verbs Pol.
wybuchać IPF or Rus. vzryvat’sja IPF derived from event PF verbs Pol. wybuchnąć
PF and Rus. vzorvat’sja PF, “to explode” may be used only with the meaning of
habitual repetition.
The different types of aspect partners we just mentioned briefly for state,
refined process or event verbs show that, because of the derivational nature of
aspect morphemes (prefixes and suffixes), the binary PF/IPF aspect opposition
in Slavic languages strongly depends on the semantic situation expressed by each
verbal lexeme and consequently can be expressed not exclusively by verb pairs but
also by verb clusters.

8.  The pragmatic usage types of aspect

As has already been pointed out, we admit that the properly pragmatic usage
types of the aspect category are related to the interaction between speaker and
hearer, especially to the meta-informative (MI) old or new status, which a speaker
assigns to an utterance. In Polish and Russian, the relation between verbal aspect
and meta-informative status can be observed in the contextual uses of IPF and
PF verbs in old and new utterances. Namely, it has been observed (Forsyth 1970;
Bondarko 1971b) that PF verbs are more frequently used in a new utterance and
IPF verbs in an old one. More precisely, we observed (Włodarczyk H. 1997) that
IPF verbs appear in both old and new contexts whereas the PF verbs are used
almost exclusively in new context, the only exception being the expressive usage
type of a PF verb in an old MI context (see §8.5 hereafter). However, the configu-
rations of semantic features which are characteristic of IPF and PF aspect respec-
tively remain valid whatever the meta-informative status of the utterance in which
a verb form occurs. The interaction between the semantic and pragmatic usage
types of aspect is extremely complex partly because the meta-informative status
of a verb phrase depends on the type of utterance in which it occurs, either a base
utterance (entirely new or old) or an extended utterance (with a contrast between
new and old constituents).
 Hélène Włodarczyk

 spect usage types and the grounding of the old/new

8.1  A
meta-informative status
We treat the pragmatic usage types of aspect by reference to the m ­ eta-informative
status of the utterance and its different types of grounding as presented in ­Chapter 3
in this volume. This framework makes it possible to account for seemingly incon-
sistent (at least for non-native speakers) uses of imperfective and perfective verbs
in Polish19 and Russian. As outlined in Chapter 3 in this volume, the old and new
status are not only grounded in discourse (anaphoric/cataphoric) and in the shared
knowledge of the speech act agents (known/unknown), but are also motivated by
reference to the represented world (language ontology) opposing generic to spe-
cific, general to particular, potential to actual, habitual to occasional. (cf. Table 2
from Chapter 3, repeated hereafter for convenience).

Table 2.  Grounding of old and new meta-informative status

Memory Type of Grounding of Grounding of

information old status new status
and knowledge

Immediate Communicative anaphoric cataphoric

(Adjacent
utterances)

Intermediate Cognitive known unknown

(Recently
acquired
information)

Permanent Ontological generic specific

general particular
(Mental Σ Π
potential actual
reference) habitual occasional

As shown in Table 2, one of the possible types of grounding of the old or new
meta-informative status is ontological knowledge. When applied to the usage
of aspect in old and new utterances, the ontological reference accounts for the
usage of verbal aspect which opposes situations seen as types to situations seen
as occurrences or tokens (cf. Table 3 from Chapter 3 in this volume repeated

.  The implication of old/new status for the usage of IPF past forms in Polish in contrast
with French imparfait forms is discussed in [Walkiewicz & Włodarczyk 2012].
 Verbal aspect in Slavic languages between semantics and pragmatics 

here for convenience). The opposition of types and occurrences concerns both
situations and their participants, above all those playing an active role, expressed
in most utterances by the subject of the verb. The number form (singular or
plural) of the subject noun denoting the active participant and its determination
(in/definiteness) in the utterance are indices leading to the interpretation of the
situation as type or token.

Table 3.  Ontology-based informative motivation of aboutness

Situation types (∑) Situation occurrences (∏)

generic > general specific > particular

potential > habitual actual > occasional

In order to apply the concepts schematically represented in Tables 2 and 3

to the description of the pragmatic usage of verbal aspect, let us recall impor-
tant observations which were already introduced in Chapter 3. First, the relation
between genericity and generality is a degree relation; in a generic situation all ele-
ments of a class (of situations and participants) without exception are concerned,
whereas only most elements of a class are representative of a general situation. It
should be stressed that the potential meaning should not be identified with the
modal meaning of possibility (possible vs. necessary): we understand it as the abil-
ity of a participant to play a role in a situation. The relation between potentiality
and habituality is causal: a participant who is used to playing a role in a situation
is assumed to be, at the same time, able to play this role. In some utterances, the
speakers point at the indefinite number of iterated situations they could or can
observe. In other utterances, they speak about the conclusion they could draw
from this observation about the ability of the agent(s) taking part in the iterated
situations. From a habitual situation, we may infer a potentiality and when this
potentiality concerns the whole class of entities referred to by the subject of the
verb without exceptions, it may lead to a generic or general interpretation as in
Examples (12) to (15) hereafter.

8.2  U
 sage types of IPF and PF verbs in ontology-motivated
old and new utterances
We use the concepts presented in Table 3 to explain the usage of IPF and PF verbs
in ontology-motivated old and new utterances (cf. Włodarczyk H. 1997). As con-
cerns the ontological grounding of the old/new meta-informative status, perfective
verbs appear almost exclusively in Π contexts whereas imperfective verbs are used
mostly (but not exclusively) in Σ contexts. There is an obvious relation between
the semantic and pragmatic usage types of PF and IPF verbs, namely in Slavic
languages, PF verbs – most of which are derived by prefixation from a simple
 Hélène Włodarczyk

IPF verb – denote a more specific point of view on a situation spoken about than
their IPF partner. As a matter of fact, a perfectivizing prefix (a) adds information
about one chosen moment of the situation (the start, enter, exit or end moment)
whereas the IPF verb may be used to refer to the whole situation without analys-
ing it into parts, and (b) very often the prefix also adds information about what
we call ­“external” aspect, especially the modification of the flow (interrupt, resume,
trans)20 or of the intensity of the situation, eventually a prefix makes it possible
to compose several distinct situations into one composed situation (we call this
the composition parameter, Section 6 in this chapter). By contrast, the IPF verbs
contain less specification about the same situations than their derived PF part-
ners, thus they can be used easily in generic and general utterances (often called
“­gnomic statements” by linguists.)
Beginning with the first ontological motivation, IPF verbs can appear in utter-
ances referring to generic (12 and 13) and general (14, 15) situations.
(12) Pol. Sól rozpuszcza się (IPF) w wodzie.
(13) Rus. Sol’ raspuskaetsja (IPF) v vode.
“Salt dissolves in water.21

(14) Pol. Francuzi jedzą (IPF) ser.

(15) Rus. Francuzy jedjat (IPF) syr.
“French people eat cheese.”

In utterances (12) to (15) IPF verbs used in generic or general statements denote sit-
uations seen as wholes: either non modified (by any of the external aspect param-
eters) states or refined processes, or ordinary processes seen as wholes (including
all their moments and stages).
On the other hand, we find PF verbs used in utterances (16, 17, 18, 19, 20) to
point at specific and particular situations.
(16) Pol. Polski poeta Adam Mickiewicz zmarł w Istambule w 1855 r.
“The Polish poet Adam Mickiewicz died in Istanbul in 1855.”
Rus. 28/02/2013 – Proslavlennyj latyshskij poet Imats Ziedonic skonchalsja v
(17) 
sredu v vozraste 79 let.
“2013-02-28 – The famous Latvian poet Imats Ziedonic passed away on
wednesday at the age of 79.”

.  Two of the flow modification parameters: the off-and-on parameter of intermittence and
the keep parameter do not concern PF verbs but IPF verbs.
.  When two Polish and Russian utterances can be approximatively translated by one and
the same English utterance we give it after the second example.
 Verbal aspect in Slavic languages between semantics and pragmatics 

(18) Pol. Wczoraj zginął (PF) w Afganistanie kolejny polski żołnierz.

“Yesterday, one more Polish soldier was killed in Afghanistan.”
(19) Pol. Nasza japońska przyjaciółka zjadła (PF) wczoraj pierwszy raz kawałek
francuskiego sera.
(20) Rus. Nasha japon’skaja prijatel’nica s’’ela (PF) vchera v pervyj raz kusochek
francuskogo syra.
“Yesterday, our Japanese friend ate a piece of French cheese for the first
time.”

This specific or particular pragmatic usage combines with the different semantic fea-
tures that can be expressed by PF verbs: moments, modification of flow or intensity,
composition. In utterances (16) to (20) all PF verbs can be characterised semanti-
cally by the finish moment and the non repetition (uniqueness) of the situation.
On the other hand, IPF verbs are used in utterances referring to habitual sit-
uations, i.e. situations which are supposed to be repeated regularly as habits an
indefinite number of times.
(21) Pol. Ekonomiści czytają (IPF) prasę codziennie.
(22) Rus. Ekonomisty chitajut (IPF) pressu kazhdyj den’.

“Economists read the press every day.”

In habitual contexs, IPF verbs keep their semantic content: they point at the whole
situation (utterances 21 and 22) or at its run stage (in the sense of ongoing process).
In opposition, PF verbs occur in utterances denoting occasional situations, i.e.
situations which occur only on particular occasions.22 In utterances (23, 24), the
temporal adverb “yesterday” pointing only at one particular day, the singular form
of the noun “economist” and the particular noun “article” opposed to the general
noun “press” from utterances (21, 22), trigger an occasional reading of the situation.
(23) Pol. Wczoraj ekonomista przeczytał (PF) ciekawy artykuł.
(24) Rus. Vchera ekonomist prochital (PF) interesnuju stat’ju.
“Yesterday the economist read an interesting article.”

The occasional pragmatic usage combines with the semantic parameter charac-
teristic of the PF verb in the given context, in utterances (23) and (24) the finish
moment of the process of reading.
It has also been noticed that IPF verbs may be used to denote potential situa-
tions, i.e. processes which some agents are able to perform:

.  Occasional situations can be repeated a definite number of times (for instance: yesterday
he met his cousin and kissed her three times.).
 Hélène Włodarczyk

(25) Pol. Jego siostra dobrze śpiewa (IPF).

(26) Rus. Jego sestra xorosho poët (IPF).
“His sister sings well (is a good singer).”

In such utterances with potential meaning, IPF verbs refer to the whole situation
spoken about.
The opposite usage of PF verbs appears in contexts where the situation they
refer to is depicted as actual. But since in Polish and Russian, PF verbs cannot be
used to refer to present but only to past or future situations, their actual usage is
restricted to past or future tenses.23 For instance, when asked a few days after a
conversation in which the decision was taken that one of the interlocutors should
write a letter to their common director, utterances (27) and (28) in which the PF
verb “to write” is used in the past are questions aiming at making sure that the
intended action to write a letter was actually achieved.
(27) Pol. Napisał (PF) pan list do dyrektora?
(28) Rus. Vy napisali (PF) pis’mo direktoru?
“Did you write the letter to the director?”

In Examples (27) and (28) the PF verbs point at the finish moment and can be
given a resultative interpretation.
Obviously, the pragmatic usage types of aspect are extremely context-­
dependent but the influence of context is even more decisive for IPF than PF verbs.
Since the IPF is the unmarked member of the aspect opposition (Jakobson 1932),
its ontology-based usage types (the different nuances of the ∑ motivations) appear
alternatively depending on the context which makes it possible to enhance one of
them. Thus, the generic interpretation of a verb form depends on the generic inter-
pretation of the co-occurring subject and object noun phrases; the habitual inter-
pretation is triggered mostly by temporal adjuncts denoting regular repetition; the
potential interpretation can be caused by the semantic content of the verb itself
(an action which an agent may be able to perform). On the other hand, since the
PF is the marked member of the aspect opposition, its ontology-based uses are not
so much context-dependent; as a consequence, the four ∏ meanings of situation
tokens co-appear with a PF verb, e.g. the specific, particular, actual and occasional
meanings in (16, 17, 18, 19, 20), although one of them may be highlighted by the
context, for instance by an adjunct, e.g. the occasional meaning of the PF verb in
utterances (19, 20, 23, 24). Thus, it is important to emphasize that such examples
as (18, 19, 20) pertain simultaneously to particular, actual and occasional mean-
ings, one of them being possibly highlighted by a constituent of the utterance.

.  As we will show in 8.4 below, in the present tense, only IPF verbs can be used with actual
meaning.
 Verbal aspect in Slavic languages between semantics and pragmatics 

Although the four above mentioned ∑/∏ oppositions serve as motivation for
the old or new MI status, they concern the ontology of the situations referred to,
i.e. ultimately the semantics of the situations. This seems to be one more reason
why it is difficult to draw a clear line between semantics and pragmatics since the
meta-informative level itself is partly motivated by the informative level (as mental
reference to ontological knowledge stored in long term memory).

8.3  Discourse-based pragmatic usage types of PF and IPF verbs

The discourse-based pragmatic usage of aspect is motivated by anaphora/­cataphora
in a text or by the knowledge shared in a speech situation. The communicative and
cognitive motivations reflect directly the point of view of the speaker and the way
the speaker intends to draw the hearer’s attention to some information by treating
it as new or old. As a matter of fact newness or oldness of information motivated
only by discourse is always more “subjective” than that based on ontology because
it depends on a relatively free choice of the speaker, namely on the discourse strat-
egy s/he chooses towards the addressee in a given speech situation.
We understand the terms anaphoric and cataphoric in a strict sense: we con-
sider that a verb form is used anaphorically when it occurs after another verb form
of the same verb used previously in the same discourse; whereas cataphoric are
those verb forms introduced for the first time into the text. The anaphoric motiva-
tion can cause the use of IPF verb forms even when the situation is a token not a
type. As we shall see further, anaphoricity and knowledge override the ontological
motivation opposing particular situations to types.

(29) Pol. A – Opowiedz mu jakąś bajkę, on nie może usnąć. B – Dobrze, którą
mam opowiadać (IPF)?
“A – Tell him a tale, he cannot get to sleep. B – Well, which tale should
I tell (IPF)?”

(30) Rus. A – Ob”jasni (PF) emu, pozhalujsta, on chto-to ne ponimaet v

­uprazhnenii. B – Nu, chto ob”jasnjat’ (IPF)?
“A – Please, explain to him, there is something he does not understand
in the exercise.
B – Well, what should I explain (IPF)?”

Examples of the cataphoric use of PF verbs can be found at the beginning of com-
municative exchanges, discourses or texts, as in utterances (31) and (32) hereafter
which are respectively the first sentence of a short story and a novel.

(31) Pol. Przyjechał (PF) kiedyś do mnie mój daleki kuzyn, misjonarz.
(S. Mrożek, Z gawęd wuja, p. 123)
“Once a cousin of mine came to see me, he was a missionary.”
 Hélène Włodarczyk

(32) Pol. Na początku zimy przyjechała (PF) z Dorpatu z Estonii babka

­Dilbinowa i pokój, który zajęła, bardzo nęcił Tomasza. (Cz. Miłosz,
­Dolina Issy, p. 61)
“At the beginning of winter grandmother Dilbinow arrived from Dorpat in
Estonia and the room she occupied was very tempting to Tomasz.”

The cataphoric usage of PF verbs is described in aspect studies as the possibility

of using PF verbs to introduce into a discourse new situations which are char-
acterised in stylistic approaches as “moving the story forward”. Forsyth (1970)
refers to this usage of the PF aspect as the “kinetic presentation” of situations. The
anaphoric/cataphoric usage of verbs is comparable to the distinction between old
and new centres of attention: when choosing to insist on the newness of a centre
of attention (a focus contrasting with an old background) the speaker gives it a
certain prominence. In the same way by choosing a PF verb the speaker gives
some prominence to a situation introduced as new; on the contrary, the use of
an IPF verb may result in keeping the situation in the background24 with an old
status.
The cognitive motivation (known/unknown) is defined by reference to the
knowledge shared by the speaker and hearer in a verbal interaction. This knowl-
edge may be acquired quite recently in the preceding conversation or text – and
thus be shared only by the participants of the speech situation – or belong to
the common knowledge in a given society, culture or quite broadly in all man-
kind, at least at a given epoch. While the communicative motivation (anaphoric/­
cataphoric) relies essentially on text constituents, the cognitive motivation is more
difficult to bring to light since it requires investigation into the knowledge of the
speech participants, be it immediately acquired or be it stored in their long term
memory, the latter depending on a vast cultural background. It is so closely depen-
dent on the cultural background that the pragmatic usage of IPF verbs motivated
by the knowledge of a situation differs even in such genetically close languages as
Polish and Russian.25 Moerover, it is important to emphasize that the discourse-
based cataphoric and unknown uses of PF verbs combine with their uses based on
ontology without cancelling them. The discourse-based motivations of the new
MI status of PF verb forms do not override the ontology-based motivations. For
this reason, PF verbs used cataphorically always concern situations seen as tokens
(with specific, particular, actual and/or occasional meanings).

.  This can be compared to [Weinrich 1973, 1989] who distinguishes in French tenses used
in the foreground or background.
.  Examples of cultural differences of the known/unknown motivation in aspect usage types
are given in Włodarczyk, H. 1997 p. 187.
 Verbal aspect in Slavic languages between semantics and pragmatics 

By contrast, the use of an IPF verb in an utterance with old MI status can be
motivated by the fact that the speaker considers that the situation s/he speaks
about belongs to the knowledge of the hearer. For instance, it is possible to use
an IPF verb as referring to a known situation to remind the hearer of a fact
which the speaker considers to be a part of the hearer’s knowledge. Let us take
as an example the following conversations (33 in Polish and 34 in Russian) in
which the IPF verb “to say” (Pol. mówiłem, Rus. ja govoril) is used to suggest that
person A should keep in mind a piece of knowledge already communicated by
person B.
(33) Pol. A – Dlaczego nie zaprosiłeś Piotra? B – Mówiłem (IPF) ci, że wyjechał
na dwa miesiące.
(34) Rus. A – Pochemu ty Petra ne priglasil? B – Ja zhe govoril (IPF) tebe, chto on
ujechal na dwa mesjaca.
“A – Why didn’t you invite Peter? B – I already told (simple past) you that
he left for two months.”

Another condition of the cognitive motivation of the IPF aspect may be the refer-
ence to common knowledge: it is possible to use an IPF verb to refer to a situation
which is assumed to be known in a given environment. For instance, a waiter in
a big restaurant (with numerous staff) may ask (using an IPF verb) guests already
seated at a table whether they have already ordered something. In this case, all the
usual episodes of a restaurant visit, including the “necessary” phase of placing an
order is perceived as known by its participants.
(35) Pol. Czy panowie już zamawiali (IPF)?
(36) Rus. Vy uzhe zakazyvali (IPF)?
“Have you already ordered (present perfect)?”

Another example of the known motivation of the use of an IPF verb form is given
in utterances (37) and (38) in which the IPF verb “to open” is used because the
act of opening the window is assumed to be the cause of the sensation of cold in a
room where the window is already closed.
(37) Pol. Ale tu zimno, otwierałeś (IPF) okno?
(38) Rus. Kak zdes’ xholodno! Ty otkryval (IPF) okno?
“It’s so cold here, did you open (simple past) the window?”

In such known usage of the IPF, the situation expressed by the verb consists in old
information (has the old MI status), because either what is new in the utterance is
only the positive vs negative sign accompanying the verb (did the situation actu-
ally take place or not?) or the new information concerns only one of the partici-
pants or anchors but not the situation itself.
 Hélène Włodarczyk

In the anaphoric and known usage types of IPF aspect illustrated by utter-
ances (29, 30 and 35 to 38) we observe at the same time a shift in the point of view
in relation to the situation, a change consisting of the selection of another part
of the situation to that selected by the PF partner. A PF verb points at the finish
moment of a process and frequently conveys the meaning of a resulting state (we
call it the after stage) beginning after a process has reached its finish moment. On
the contrary the IPF aspect used with an anaphoric or known meaning induces at
the same time a shift of the attention from the finish moment of a process to its
run stage, leaving unexpressed or implicit the finish moment – which leaves open
diverse interpretations: either the finish moment has not been actually reached
(because the process failed or was interrupted) or it has been cancelled by the
moment we call terminal, consisting in putting an end to the resulting state initi-
ated by the finish moment of a process.
The anaphoric and known discourse motivations of the old status account for
usage types of the IPF aspect which are generally regarded as “untypical” because
they seem to contradict what is considered its most typical semantic usage, namely
the usage of the IPF pointing at ongoing processes.26 The reference to the meta-
informative status makes it possible to give a more coherent account of those prag-
matic usage types of IPF verbs which perplex non native speakers27 for whom it
is hardly possible to puzzle out a logical link between such usage types and the
traditional concept of the imperfective aspect as durative. As a matter of fact, the
usage of IPF aspect in old utterances motivated by discourse or knowledge seems
to override its ontology-based usage types denoting generic, general, potential or
habitual situations. Thus, in an old utterance with anaphoric or known motiva-
tion, IPF verbs denote specific, particular or actual situations. Thus, IPF verbs
can express both typical (∑) and individual (∏) situations. In their anaphoric and
known usage types IPF verbs point at particular tokens of situations (instead of
types), namely they express particular, actual and occasional situations just as PF
verbs but without the specific semantic features added by PF morphemes, mostly
prefixes adding a precision concerning the analysis or mode of the situation
referred to by the verbal root (e.g. the start or finish moment, the interruption or
resuming, etc.). In former studies this use of the IPF verb was explained by the fact

.  At the same time these pragmatic usage types have no regular correspondents in the
usage types of the French IMP or English progressive tenses. However a similar anaphoric or
known usage of the English progressive form has been described by Adamczewski H. 1978. It
seems that what is called “imparfait pittoresque” in French grammars is probably an anaphoric
or known usage (Włodarczyk H. 1995).
.  As regards the non-correspondence between such uses of the IPF aspect in Slavic and
the use of imparfait in French see (Walkiewicz & Włodarczyk 2012).
 Verbal aspect in Slavic languages between semantics and pragmatics 

that the IPF verb is the most abstract name of a situation without any of the par-
ticular aspect features which can be added by a perfectivizing morpheme.28 The
discourse-grounded anaphoric and known usage types of imperfective verbs refer
to the same situations as those denoted by a perfective verb expressing a particular,
actual or occasional situation, but they appear in case this situation had already
been mentioned in the previous discourse or belongs to the knowledge universe of
the discourse participants.

8.4  Usage of Slavic imperfective verbs in both old and new contexts
As regards the different usage types of the Slavic imperfective aspect, we must
bear in mind that they should not be confused with the usage types of the French
imparfait or with that of the English progressive forms. The non-correspondance
originates mostly in the fact that some semantic meanings of aspectual forms in
different languages are comparable whereas some pragmatic usage types remain
different. Without deeper insight into the problem of the borderline between
what is semantic and what is pragmatic it is practically impossible to compare the
usage types of aspectual forms in different languages. One of the main theoretical
problems in the history of Slavic aspect studies has its origin in the fact that the
imperfective aspect has been globally identified with the Imperfectum of classical
languages (e.g. Lomonossov’s Russian grammar).29 This concerns first of all what
is considered the primary (or most important) usage of the imperfective aspect in
Slavic. A  huge aspectological tradition considers as the primary meaning of the
imperfective aspect its usage denoting an actual ordinary process in its run stage
or an actual state lasting through a certain period; this usage is called the “dura-
tive” or “ongoing process” meaning of the IPF and it is currently opposed to the
“momentary” meaning of the PF (Karolak 1997). In the history of aspect studies
this opposition is represented by the spatial metaphor of a line (representing the
IPF) opposed to a point (representing the PF). However investigations into Polish
and Russian aspect usage in context (Włodarczyk H. 1997) led us to the conclusion
that the usage of the Imperfective aspect in Slavic languages to denote an actual
ongoing process is only one of its possible different meanings in context. It should
not be considered as the primary meaning of the imperfective if we admit that the
imperfective is the unmarked member of the privative aspect opposition. What can
be stated about the unmarked member is that it exhibits many different contextual
usage types, some of which are more frequent than the others. As the unmarked

.  Forsyth J., 1970 calls this use of the IPF aspect simple denotation, Bondarko 1971b general
factual (obobshchenno-fakticheskoe znachenie).
.  On the history of Slavic aspectology cf. Włodarczyk 1997; Archaimbault 1999.
 Hélène Włodarczyk

member, the imperfective aspect has no primary usage, more precisely, no default
usage appearing in a minimal context. To explain it, let us draw attention to the fact
that he following utterances cannot be interpreted without further context:
(39) Pol. Ojciec czyta gazetę.
(40) Rus. Otec chitaet gazetu.
N mas sg nom V IPF pres 3pers sg N fem sg acc
“Father reads (or is reading) the newspaper.”

In such a short utterance, there is no reason to choose the actual durative meaning
rather than the habitual or potential meaning of the IPF verb. As a consequence,
out of context, it is impossible to choose the simple or progressive present in the
English translation.
The actual durative reading becomes explicit only in the following utterances
(41) and (42):
(41) Pol. Nie przeszkadzaj! Ojciec czyta gazetę.
(42) Rus. Ne meshaj! Otec chitaet gazetu.
neg. V Imperative N mas sg nom V IPF pres N fem sg acc
2pers 3pers sg
“Do not disturb! Father is reading the newspaper.”

On the other hand, the habitual meaning can be made explicit when the IPF verb
is used with an adjunct phrase specifying the habitual repetition:
(43) Pol. Ojciec czyta gazetę codziennie.
(44) Rus. Otec chitaet gazetu kazhdyj den’.
N mas sg nom V IPF pres 3pers sg N fem sg acc Prep P or Adverb
“Father  reads the newspaper every day.”

In an appropriate context, a potential meaning may be also expressed by an IPF

verb.
(45) Pol. Jego siostra pięknie śpiewa.
(46) Rus. Sestra jego prekrasno poët.
N fem sg nom possessive adverb V IPF pres 3pers sg
“His sister beautifully sings.”
“His sister sings beautifully.”

However, an actual situation, on the other hand, may also be expressed by an IPF
verb form:

(47) Pol. Posłuchaj!  Jego siostra śpiewa pieśń.

(48) Rus. Poslushaj! Sestra jego poët pesnju.
V PF Imperative N fem sg nom V IPF pres N fem sg acc
2sg ­possessive 3pers sg
“Listen! his sister is singing a song.”
 Verbal aspect in Slavic languages between semantics and pragmatics 

Hence, the same IPF Slavic verb form may be used to point alternatively at an
actual occurence of a situation (47, 48) – then it has to be translated into English
by a progressive form – or at a potential sitation (45, 46) – which translates into
English by a non progressive verb form.
The actual meaning is a pragmatic contextual usage of the imperfective aspect
which can be used regularly to refer to individual situations (situation tokens)
seen as unanalysed whole or as actually ongoing processes or lasting states (in any
tense: past, present or future). But, on the contrary, as was shown in Section §8.2.,
the ontological motivation makes it equally possible (in other contexts) to use
imperfective verb forms to refer to typical (generic, general, habitual or potential)
situations.
When comparing IPF verb forms in Slavic with imparfait tense forms in
French we propose to keep in mind the definition by Jerzy Kuryłowicz (1972),
according to whom the imparfait tense (in Latin and Romance languages) is a
marked form aiming at transferring into the past the point of view of the Pres-
ent tense in its actual (opposed to atemporal) reading, i.e. a situation going on at
speech time. To explain the actual meaning of progressive tense forms in English,
we must use another framework, the theory proposed by Reichenbach (1947) who
introduced a reference point (R) distinct both from the speech moment (S) and
the moment of the event (E) talked about. A reference point (R) anterior to the
speech act moment (S) and simultaneous with the moment of event (E) is neces-
sary to account for the French imparfait or English progressive simple past. This
reference point is nothing but another situation’s moment. On the contrary, in
Slavic languages, imperfective past verb forms are defined only by reference to the
speech act moment but not to another event: they are opposed to PF past forms
in that PF verb forms do not point either at unanalysed situations or at stages
but refer either to selected moments or modified stages. The interpretation of IPF
verbs as denoting actually ongoing situations depends on information added in
the context, i.e. information about another situation which is simultaneous and
serves as a reference point.

8.5  The expressive usage of the PF Aspect in old contexts

In their most frequent pragmatic usage, PF verb forms refer to tokens (individual
occurrences of situations) while IPF verb forms refer to types of situations and
therfeore can be used in gnomic utterances. However, there exists a less frequent
usage of the PF aspect in utterances with the old meta-informative status; this
usage pertains to metaphor and adds expressivity to the utterance. Thus PF verb
forms – which normally refer to specific, particular, occasional (i.e. non habitually
repeated) and actual situations – can be used with an expressive value to denote
generic, general, potential or habitual situations. This usage of PF verbs with
 Hélène Włodarczyk

generic, general, habitual or potential meaning is described as stylistically expres-

sive; it can be explained by the possibility of using a unique situation as the proto-
typical example of an indefinite number of similar situations (cf. ­Włodarczyk H.
1997). In other words, the PF verb expressing an individual occurrence of a situa-
tion (an item or a token) is used metaphorically as a typical example, a prototype
(Rosh 1973) of a whole set of similar situations. This usage is called “exemplary
meaning” in Russian aspectology (in Russian nagljadno-primernoe znachenie, cf.
Bondarko 1971b: 22).
Thus, the generic and general usage of the PF aspect occurs very often in prov-
erbs or idioms which are supposed to be true in any circumstance and serve as a
model of behaviour. Such idioms contain indefinite nouns with a generic inter-
pretation (any element of the class referred to), as is the case for “woman” in (49)
and for “mother” in (50). Such utterances also point at the unlimited number of
repetitions of a situation. It is worth emphasizing that in this type of general state-
ments, PF verbs used in the present tense are not interpreted as pointing at future
situations but at atemporal situations which are always true at any time and there-
fore can take place in past, present or future.

(49) Pol. Gdzie diabeł nie może, tam babę pośle (PF). (J. Krzyżanowski, Mądrej
głowie dość dwie słowie, t. I, p. 48)
Lit. “Where the devil is helpless he sends a woman.”

(50) Rus. Mat’ i vysoko podymet (PF), da ne bol’no opustit (PF) ruku. (Dal’, Po-
slovicy russkogo naroda)
Lit. “Although a mother raises her hand it does not hurt when she lets it
down.”

The general usage of the PF can be triggered by the impersonal use of the second
person singular denoting an indefinite anonymous human subject with the mean-
ing “anyone ” (cf. Chapter 4 in this volume).
(51) Pol. Coś na Bóg spuścił (PF 2pers sg), to już stracono (PF).
Lit. “What you left to God is already lost.”
(52) Rus. Bez truda ne vytashchish’ (PF 2pers sg) rybku iz pruda. 〈http://www.
mista.ru/pogovorki.htm〉
Lit. “Without effort you cannot pull a fish out of a pond.”

In the same way, a PF verb used with a potential meaning always contains an
expressive nuance emphasizing that the action is not easy to perform but that the
subject has such qualities that he is able to perform it at any moment.
(53) Pol. On jest bardzo oczytany: on odpowie (PF) na każde pytanie.
“He is well-read: he will (can?) answer any question.”
 Verbal aspect in Slavic languages between semantics and pragmatics 

(54) Rus. On vam ljubuju spravku dast (PF), vsë ob”jasnit (PF) – ochen’
­erudirovannyj chelovek.
“He can give you any information, explain anything; he is a very learned
person.”

A PF verb used in a habitual context also adds some expressivity to the utter-
ance: the situation expressed by a PF verb is not so regularly repeated as when it is
expressed by an IPF, it can thus be called a sporadic repetition. This usage of a PF
verb is often used with the expressions Rus. byvaet, byvalo and Pol. bywa, bywało
(it occurs/occurred that).
(55) Pol. Chociaż jestem bardzo zapracowany, bywa, że wyjdę (PF) na krótki
spacer z przyjaciółmi.
Rus. Xotja u menja mnogo raboty, no i s druzjami inogda byvaet vyjdu (PF)
(56) 
poguljat’
“Although I am very busy, I will sometimes go out for a short walk with
friends.”

In all these types of uses the present tense forms of PF verbs do not have their
usual future meaning but that of atemporal present characteristic of general
statements.

9.  Conclusion

The pragmatic usage types of the aspect category based on ontology (∑ and ∏
uses) are close to their semantic usage and comparable in different languages
(and therefore easier to grasp for non-native speakers) whereas the usage types
based on discourse (anaphoric/cataphoric) and knowledge (known/unknown) are
more language specific. As was shown in this chapter, the different pragmatic and
semantic usage types of PF and IPF verbs can combine diversely in context. As
regards PF verbs their different usage types do not override. On the other hand
some usage types of IPF verbs can override each other and this is the cause of
non-native speakers’ difficulties with some usage types of the IPF verbs classifed
as “­untypical”  – although their real frequency remains to be studied on large
corpuses.
Beyond the description of verbal aspect itself, the ASMIC theory can serve as
tertium comparationis to bring together verbal and nominal categories which are
involved in the marking of the old and new meta-informative status, thus allowing
cross-category comparisons in the pragmatic dimension of language. Former stud-
ies comparing the nominal and verbal categories were based on semantic investi-
gations, bringing together nouns and verbs in the perspective of heterogenous vs
 Hélène Włodarczyk

homogenous types of situations (Mehlig 1994, 1996); the semantic (often gram-
maticized) behaviour can be compared to that of count and mass nouns. The rela-
tion between the aspect category and the meta-informative old or new status can
serve as a basis for investigating what is common between article and aspect (Leiss
2000)30 more specifically in the pragmatic dimension (Włodarczyk H. 1997), and
thus opens up a wide range of possibilities not only in theoretical linguistics but
also in applied linguistics where reference to the motivations of the old/new meta-
informative status may serve to improve the translation and teaching methods of
Slavic languages in non-Slavic environments (Włodarczyk H. 1998). This is an
important application for the contrast between Slavic and non-Slavic languages.
As a matter of fact, when contrasting aspect usage in two modern languages,
one should take into account both semantic and pragmatic properties of aspect
forms. As an example, French imparfait and Slavic past imperfective are both fit
alternatively for old and new contexts. However, although the Slavic past imper-
fective occurs in similar old and new contexts as the French imparfait, it is used
in a wider range of new contexts (cf. Walkiewicz & Włodarczyk 2012). In a larger
typological perspective, taking into account both the semantic and pragmatic fea-
tures of aspect may help in understanding and comparing some aspect uses in
different languages31 which seem puzzling when considered only semantically.32
Let us emphasise that in all types of languages, aspect is an essential tool of the
meta-informative structure since it compels speakers to choose a point of view
(a centre of attention) in relation to a situation they talk about.

References

Abraham, Werner. 2013. The developmental logic of the analytic past in German and Polish
An issue of universalism or areal contact? In: Lohndal, Terje (ed.), In Search of Univer-
sal Grammar: From Old Norse to Zoque, 175–194. Linguistik Aktuell/Linguistics Today.
Amsterdam: John Benjamins.

.  The comparison proposed by Leiss 2000 makes use of mereology.

.  Let us mention the two different usage types of -te iru verb forms in Japanese which seem
contradictory at first glance since such forms may be used alternatively with a so-called “pro-
gressive” meaning and with a so-called “resultative” meaning. It is quite probable that what is
common to these two usage types is the possibility of them occuring in new utterances. The
new status is motivated by the actual character of the situation spoken about, be it an actual
ongoing process or an actual state resulting from a telic process which has reached its finish
moment.
.  The usage of the English progressive form with old meta-informative status motivated by
the discourse (by anaphora or knowledge) has been described in Adamczewski 1978.
 Verbal aspect in Slavic languages between semantics and pragmatics 

Adamczewski, Henri. 1978. Be+ing dans la grammaire de l’anglais contemporain. Paris: Librairie
Honoré Champion.
Agrell, Sigurd. 1908. Aspektänderung und Aktionsartbildung beim polnischen Zeitworte. Lund:
Håkan Ohlssons Buchdruckerei.
Archaimbault, Sylvie. 1999. Préhistoire de l’aspect verbal. L’émergence de la notion dans les
­grammaires russes. Paris: CNRS Editions, Coll: Sciences du Langage.
Avilova, Natalia S. 1976. Vid glagola i semantika glagol’nogo slova. Moskva: Nauka.
Bondarko, Aleksandr V. 1971a. Grammaticheskaja kategorija i kontekst. Leningrad: Nauka.
Bondarko, Aleksandr V. 1971b. Vid i vremja russkogo glagola. Moskva: Prosveshchenie.
Brinton, Laurel J. 1988. The Development of English Aspectual Systems. Aspectualizers and Post-
verbal Particles. Cambridge: CUP.
Czarnecki, Tomasz. 1998. Aspektualität im Polnischen und Deutschen. Bedeutungen und Formen
in einer konfrontativen Übersicht. Gdańsk: Wydawnictwo Uniwersytetu Gdańskiego.
Czochralski, Jan A. 1975. Verbalaspekt und Tempussystem im Deutschen und Polnischen, Eine
konfrontative Darstellung. Warszawa: PWN.
Forsyth John. 1970. A Grammar of Aspect, Usage and Meaning in the Russian verb. Cambridge:
CUP.
Janda, Laura A. 2007. Aspectual clusters of Russian verbs. Studies in Language 31(3): 607–648.
Jakobson, Roman. 1932. Zur Struktur des russischen Verbum. Charisteria Guilelmo Mathe­sio…
oblata: 74–84. Prague: Cercle linguistique de Prague.
Jakobson, Roman. 1936. Beitrag zur allgemeinen Kasuslehre (Gesamtbedeutung der russischen
Kasus). TCLP VI: 240–288. Prague.
Kardela, Henryk. 1997. Telicity as a perfectivising category: Notes on aspectual distinctions in
English and Polish, Language History and Linguistic Modelling: A Festschrift for Jacek Fisiak
on his 60th birthday, Vol.2: Linguistic Modelling, Raymond Hickey & Stanislaw Puppel
(eds). Berlin: Mouton de Gruyter.
Karolak, Stanisław. 1997. Arguments contre la distinction: aspect/modalité d’action (­ Aktionsart).
Etudes cognitives/Studia kognitywne Vol. 2: 175–192. Warszawa: Slawistyczny Ośrodek
Wydawniczy (SOW) Polska Akademia Nauk (PAN).
Kątny, Andrzej. 1994. Zu ausgewählten Aktionsarten im Polnischen und deren Entsprechungen im
Deutschen, Rzeszów: Wyd. Wyższej Szkoły Pedagogicznej.
Kuryłowicz, Jerzy. 1972. Miejsce aspektu w systemie koniugacyjny. Symbolae Polonicae in
Honorem Stanislai Jodłowski, 93–99. Wrocław, Warszawa, Kraków, Gdańsk: WPAN
Ossolineum.
Kuryłowicz, Jerzy. 1977. Problèmes de linguistique indo-européenne, Warszawa: WPAN
Ossolineum.
Laskowski, Roman. 1998. “Uwagi o znaczeniu czasowników”. In Gramatyka współczesnego
języka polskiego, Morfologia, wyd. 2 zmienione, T. 1: 152–171. Warszawa: PWN.
Leiss, Elisabeth. 1992. Die Verbalkategorien des Deutschen. Ein Beitrag zur sprachlichen Katego-
risierung. Berlin: Walter de Gruyter.
Leiss, Elisabeth. 2000. Artikel und Aspekt: die grammatischen Muster von Definitheit. Berlin:
Walter De Gruyter.
Leiss, Elisabeth. 2002. Der Verlust des aspektuellen Verbpaare und seine Folgen im Sprach-
wandel des Deutschen: ein Überblick. In Grammatische Kategorien aus sprachhistorischer
und typologischer Perspektive. Akten des 29 Linguistenseminars, Kyoto 2001, Mitsonobu
Yoshida (eds), 9–25. München: Iudicium.
Mehlig, Hans R. 1994. “Gomogennost’ i geterogennost’ v prostranstve i vremeni”. Revue des
Etudes Slaves 66: 595–606.
 Hélène Włodarczyk

Mehlig, Hans R. 1996. Some analogies between the morphology of nouns and aspect in Russian.
Folia Linguistica XXX: 87–109.
Netteberg, Kristine. 1953. Études sur le verbe polonais. Copenhagen: Rosenkilde og Bagger.
Piernikarski, Cezar. 1969. Typy opozycji aspektowych czasownika polskiego na tle słowiańskim.
Wrocław: Ossolineum.
Piernikarski, Cezar. 1975. Czasowniki z prefiksem po- w języku polskim i czeskim. Warszawa:
PWN.
Potebnja, Aleksandr A. 1941. Iz zapisok po russkoj grammatike (1899), t. 4. Moskva - Leningrad:
Izd. A.N.S.S.S.R.
Reichenbach, Hans. 1947. Elements of Symbolic Logic. New York NY: Macmillan & Co.
Rosch, Eleanor H. 1973. Natural categories. Cognitive Psychology 4(3): 328–350.
Sémon Jean-Paul. 1986. Postojat’ ou la perfectivité de congruence, définition et valeurs textu-
elles. Revue des Etudes Slaves 58(4): 609–636.
SFPW: Słownik frekwencyjny polszczyzny współczesnej, 1980. red. Kurcz, Ida, Lewicki, Andrzej,
Sambor, Jadwiga, Szafran, Krzysztof, Worończak, Jerzy. (eds). Kraków: Instytut Języka
­Polskiego PAN.
Ter Meulen, Alice G.B. 1995. Representing Time in Natural Language: The Dynamic Interpreta-
tion of Tense and Aspect. Cambridge MA: The MIT Press, A Bradford Book.
Vendler, Zeno. 1967. Verbs and times. On Linguistics and Philosophy, Ithaca, Zeno Vendler,
97–121. New York NY: Cornell Univer­sity Press.  (Revised version of Vendler, Zeno. 1957.
Verbs and times. The Philosophical Review 66: 143–160).
Veyrenc, Charles Jacques. 1968. Grammaire du russe. Paris: Presses universitaires de France
(coll. Que sais-je?, n0 1278).
Veyrenc, Charles Jacques. 1980. Études sur le verbe russe. Paris: Institut d’études slaves.
Vinogradov, Viktor Vladimirovitch. l947, 2nd ed. l972. Russkij Jazyk. Grammaticheskoje uchenije
o slove. Moskva - Leningrad: Izd. A.N.S.S.S.R.
Walkiewicz, Aleksandra & Włodarczyk, Hélène. 2012. Revisiting the correspondence between
the Polish Imperfective Aspect and the French Imparfait in the Distributed Grammar (DG)
framework. In Studies in Polish Linguistics 7, 5–36. Kraków: Institute of Polish Language,
Polish Academy of Science, Lexis.
Weinrich, Harald. 1973. Le Temps. Paris: Le Seuil.
Weinrich, Harald. 1989. Grammaire textuelle du français. Paris: Didier/Hatier.
Włodarczyk, André. 2003. Les cadres des situations sémantiques.  Études cognitives/Studia
­kognitywne V, 35–52. Warszawa: SOW.
Włodarczyk, André. 2007. CASK – Computer-aided acquisition of semantic knowledge project
(in Japanese). Japanese Linguistics 21. Tokyo: The National Institute for Japanese Language.
English version downloadable at 〈http://celta.paris-sorbonne.fr/anasem/papers/〉
Włodarczyk, André. 2009. Interactive discovery of ontological knowledge for modelling
­language resources. In Representing Semantics in Digital Lexicography. Proceedings of
MONDILEX, the 4th Open Workshop, Violetta Koseska-Toszewa, Ludmilla Dimitrova &
Roman Roszko (eds). Warszawa: SOW. 〈http://celta.paris-sorbonne.fr/anasem/papers/〉
Włodarczyk, André & Włodarczyk, Hélène. 2001. La préfixation verbale en polonais I. Le statut
grammatical des préfixes, II: L’Aspect perfectif comme hyper-catégorie. Études cognitives/
Studia kognitywne 4: 93–120. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2003.  Les paramètres aspectuels des situations
sémantiques. Études cognitives/Studia kognitywne 5: 11–34. Warszawa: SOW, PAN.
 Verbal aspect in Slavic languages between semantics and pragmatics 

Włodarczyk, André & Włodarczyk, Hélène. 2006. Semantic structures of aspect (A cognitive
approach). In Od fonemu do tekstu, prace dedykowane Profesorowi Romanowi ­Laskowskiemu,
389–408. Kraków: Instytut Języka Polskiego Polskiej Akademii Nauk. Lexis.
Włodarczyk, Hélène. 1994. L’Aspect verbal slave et les domaines du donné et du nouveau.
Études cognitives, Vol.1: Sémantique des catégories d’aspect et de temps, 113–130. Warszawa:
SOW, PAN.
Włodarczyk, Hélène. 1995. L’apport de la linguistique théorique à la traduction du polonais en
français:  Le problème du temps et de l’aspect au passé. Les contacts linguistiques franco-
polonais, 13–27. Lille: Presses Universitaires de Lille. Travaux et recherches, Collection
UL3.
Włodarczyk, Hélène. 1996. Les traits sémantiques du sujet ‘anonyme’, en polonais, russe et fran-
çais. Semantyka a konfrontacja językowa: 179–198. Warszawa: SOW, PAN.
Włodarczyk, Hélène. 1997. L’Aspect verbal dans le contexte en polonais et en russe. Paris: Institut
d’Études Slaves.
Włodarczyk, Hélène. 1998. Wykładniki wartości informacyjnej wypowiedzenia w j. polskim i
francuskim (aspekt, okresloność, modalność), Congrès des Slavistes Cracovie 1998. Revue
des Études Slaves T. 70(1): 53–66. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 2003. L’Aspect perfectif comme hypercatégorie (approche cognitive), com-
munication au XIIIe congrès des slavistes à Ljubljana en août 2003. Revue des Études Slaves
74(2–3):327–338. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 2008. La place du temps dans la théorie cognitive de l’aspect (perfectifs
d’achèvement et d’interruption en polonais et en russe). In Le Temps construit, Mélanges
offerts à Jean-Paul Sémon, Jean Breuillard (ed.): 109–131. Paris: Institut d’Études Slaves.
Włodarczyk, Hélène. 2009. From ontological attributes to semantic feature structures – Experi-
mental research on aspect in Polish. Cognitive studies – Études Cognitives 9: 15–32.
Warszawa: SOW, PAN.
Xrakovskij, Viktor S. 1997. Mul’tiplikativy i semel’faktivy (problema vidovoj pary). Semantika i
struktura slavjanskogo vida II, Stanisław Karolak (ed.), 227–240. Kraków: Wyd. Naukowe
WSP.
 The position in the utterance and the melodic
realisation of object and reflexive pronouns
in classical modern literary Russian

Olivier Azam
École Normale Supérieure, Paris / Université Paris-Sorbonne, Centre de
linguistique théorique et appliquée (CELTA)

The numerous studies on word order and intonation in modern Russian have
primarily focused on the “weightiest” elements in the utterance, appearing
to almost systematically ignore the pronouns. Yet these are among the most
common words in a language. The absence of rules governing their use, in a
language in which word order, though unfixed, is in no way arbitrary, poses a
serious problem for the non-native speaker.
  By focusing on the position of object and reflexive pronouns, the current
study aims to respond to this absence of critical work by presenting, in the
light of the metainformative theory of centres of attention, the criteria which
determine the enunciator’s choice of a particular word order and prosodic
realisation. We generally distinguish two groups of pronouns complements
(differentiated only by the prosodic criteria in modern Russian, though they
are also morphologically distinct in Old Russian.) In place of this traditional
binary opposition (tonic pronouns versus atonic pronouns) which has limited
practical use, the current study proposes a more precise opposition of three types
of expressed pronouns – stressed, tonic and atonic – to which must be added
elliptical pronouns, particularly common in Russian. This new distinction allows
a more nuanced analysis of utterances, which in turn renders it possible to prove
the degree to which pragmatic strategy, word order, intonational curve and the
tonic of the latter are closely linked in modern lliterary Russian.

1.  The problem

Although the issue of word order in Russian has been widely explored in linguis-
tics for many decades, and even though some rules, however clumsily formulated,
which govern this order have become a kind of mantra which is reproduced even
in handbooks for foreigners, it is evident that the problem of placing pronouns
in the utterance is almost never tackled in Russian linguistics. When it comes to
 Olivier Azam

e­ lement order in the sentence, it is almost always a question of nouns and nominal
groups, so that it is very difficult to determine whether specific factors such as
rhythm or the balance of the sentence may play a role in the choice of pronoun
position in the utterance.
For foreigners it is striking that Russian uses far fewer object pronouns than
French, German or English. In this way, such commonplace sentences as “give it to
me” or “I told you it” can almost never be literally translated into Russian. Rather
than using two object pronouns (a possible, though rare, construction), Russian
only expresses one of these complements, judging that the context will suffice to
establish the second, or else it will replace the object pronoun with a demonstra-
tive. It may also express one of the complements by means of a substantive. The
primary difficulty is therefore to determine whether or not it is necessary to use
an object pronoun. This is a specific question which will be the subject of a spe-
cial study. Within the limits of this article, which forces me to limit my analysis
to those governing a single object or reflexive pronoun,1 my aim is to resolve the
double bind presented by the following stage, once it has been established that the
pronoun should be used: where should it be placed within the utterance and how
should one realise it on the melodic plane, or in other words, how should it be
pronounced?

2.  Preliminary observations

2.1  The choice of the Gospels and the oral use of a written corpus
In these circumstances the choice of the Gospels as a corpus might initially seem
surprising. It undoubtedly presents some minor drawbacks.
The first, of which one must simply take note, is that the most widely accepted
modern Russian translation of the New Testament is the so-called Synodal Trans-
lation, which dates from the beginning of the nineteenth century. Now it seems
apparent that one of the aspects of the language which has significantly altered over
the past one hundred and fifty years is the position of object pronouns, de facto if
not de jure (I will return to this distinction), and even if this evolution appears less
radical than that which has affected the use of possessives (see Ch. Bonnot 2008).
In spite of all this, the Synodal Translation is indisputably w ­ ritten in m
­ odern

.  I will also leave aside particular structures whose word order is more or less fixed in the
corpus, such as participial phrases or non-nominative clauses (zero subject or impersonal
clauses such as instrumental clauses) in which the choice of the position of the pronoun is
subject to specific restraints.
 Object and reflexive pronouns in modern Russian 

l­iterary Russian, even if it must be specified that it is classical not contemporary

literary Russian.2
The second drawback of the Gospels is that they are, in the highest sense of the
word, Scriptures. One can thus legitimately question the extent to which analysis
of a written corpus is relevant to the study of a phenomenon eminently concerned
to the study of intonation. To answer this objection one might simply evoke the
nature of the Gospels themselves, which are not just narrative texts but also and
above all a compilation of Christ’s words. There is a large proportion of situational
dialogue and direct speech and certain passages (in particular Christ’s sermon in
the Gospel of St. John) abound with plays on pronouns.
The most important point is not the nature of the text, however, but the way in
which it has been analysed. I have in fact undertaken, for the purposes of the cur-
rent article, an exclusively oral study of the four Gospels, and I will insist strongly
upon this point. Ideally, moreover, the collection of examples would have con-
sisted exclusively not of written quotations but of the acoustic excerpts to which
they correspond.3 The collection and classification of the examples cited here was
preceded in fact by the audition of three complete recordings of the Gospels given
by three different professional Russophone readers. It is therefore a question of
prepared readings in which one can assume that the intonative choices are justified
by an analysis of the sense of the piece. Attentive audition of these three record-
ings is accompanied by the notation of all the pronouns in these four texts, and
by a technical analysis, when in doubt, of the corresponding sound file. Naturally
the variations and different choices made from reader to reader are taken into
account. I have established however that these were only minor variations which

.  It is precisely this disjunction between the language of the Russian Synodal Bible and
the most contemporary usage which I wished to emphasise by deliberately conserving, to the
point of transliteration, the original spelling which is used as a visual signal. From a desire
to simplify I have however chosen not to transliterate the hard sign at the ends of words. It
will suffice to remember that all words which end, in Soviet and post-Soviet spelling, with a
consonant (other than yod) which is not followed by a soft sign, until the 1917–1918 reform,
finished with a hard sign.
.  I have used a play on typographical conventions to illustrate certain prosodic functions:
the enclitic pronoun is in exposant; the simply tonic pronoun is rendered in italics, the stressed
pronoun in bold. Elements other than the pronoun which are emphasised may be in bold
too. Finally, where it has seemed necessary to give the wider context, the part of the quota-
tion which constitutes the actual example is underlined. In order to render this article more
concise, the examples have not been translated. Since each verse is accompanied by its Bible
reference (the number of the verse in bold at the beginning of the quotation, the chapter and
name of the evangelist in brackets at the end of the example) the reader can easily consult his/
her Bible translation of choice.
 Olivier Azam

in no way undermine the conclusions elaborated here and only demonstrate that
certain pronominal positions are ambiguous and open to interpretation.
In spite of the limited drawbacks consonant with the choice of the Gospels,
this oral use of a prepared reading of texts which, though undoubtedly written,
nevertheless contain a large proportion of direct speech, seemed a more valid
option than to rely on ad hoc examples or on the point-blank interrogation of
Russophones on this or that utterance when the intonation does not seem evident.
Moreover the Gospels boast other advantages: three out of the four are synoptic.
The observation of pronouns can therefore be conducted in near-identical, yet
slightly differing contexts, which sometimes change nothing and sometimes add
or lack an infinitesimal detail which itself has an effect on the use of the pronouns.
These very similar contexts allow a considerable degree of refinement within the
analysis.

2.2  Elliptical, atonic, tonic and stressed object pronouns in modern Russian
By object pronoun (without further precision) I mean all object pronouns which
have a strong rection, in other words all object pronouns whose presence, at least
elliptical (often used in Russian), is necessary in order for the utterance to be com-
plete.4 This study will show that with regard to these object pronouns, whether or
not they are introduced by a preposition has no effect, in Russian, on their position
within the utterance.
The only object pronouns studied in this article are the reflexive and personal
pronouns.5 It is generally accepted that in modern Russian pronouns can be atonic
or stressed, though it must rightly be clarified that the pronoun may also not be
expressed at all. In reality, this study shows that the object pronoun in modern
Russian may occur in not three but four distinct forms: (1) it may not occur at all,
this would be termed an elliptical pronoun; (2) it may be present in the utterance as
an atonic or more precisely enclitic pronoun (since the enclitic pronoun retains its
lexical stress); (3) it may occur in a simply tonic form and finally (4) it may occur
in its stressed tonic form.
By enclitic pronoun I mean precisely that which Jean Breuillard refers to as
“the melodic enclitic position of the pronoun,” which is defined by the fact that the
unit in an enclitic position “cannot be spoken in a higher tone than the unit which
precedes it” (J. Breuillard 2008: 55).

.  Thus in an utterance such as On podošël k nemu (he came towards him), we would be
dealing with an object pronoun (k nemu) just as in On uznal eë (he recognised her).
.  For the sake of convenience the term “pronoun” without further precision will henceforth
refer to object and personal pronouns. In all other cases the type of pronoun will be specified.
 Object and reflexive pronouns in modern Russian 

By a simple tonic pronoun I am referring, in modern Russian, to a pronoun

which is not in a “melodic enclitic position”, and yet which does not carry the cen-
tre of the intonation curve of the segment to which it belongs (nor, a fortiori, does
it carry a contrastive stress).
By stressed pronoun, on the other hand, I mean a pronoun which carries a
contrastive stress. In this definition a “stressed” pronoun is therefore always “tonic”
but a “tonic” pronoun is not necessarily “stressed”.

2.3  Base and extended utterances

Handbooks which deal with word order in modern Russian such as those of
­Xavronina and Krylova (Krylova – Xavronina 1986) have popularised the long-
cherished notion of the “actual division”, and the distinction, for the most part fairly
nebulous, between utterances which can be divided into “theme” and “rheme” –
infelicitous terms – and “indivisible” utterances which are purely “­rhematic”. In the
first category of utterances, the handbooks for foreigners or prescriptive Russian
grammar books generally affirm that the “theme” should precede the “rheme” in a
word order deemed “neutral” but that this order can be “inverted” in what is deemed
a more “expressive” word order. In this instance the “rheme,” which carries a more
marked final nuclear stress than in its habitual place, is preposed to the “theme”.
The relevance of this terminology which dates back to the Prague School has
for a long time been rightfully called into question, notably by Jean-Paul Sémon
(Sémon 1995–1996). Undoubtedly there is a great risk of confusing the two levels
of analysis: that which relies on the morphosyntax and that which comes under
the domain of pragmatics. It was to emphasise the distinction between these two
that Jean-Paul Sémon introduced the concept used here of the nexus as a substitute
for the “predicate” in order to refer to the organising centre of the clause at a mor-
phosyntactic level:6 the “predicate,” a term borrowed from logic, refers in informal
language to “that which is said of a subject” and concerns the pragmatic level.7
More recently, Hélène and André Włodarczyk (2004, 2006, 2008) have, in an
elaboration of their meta-informative centering theory (M.I.C.), underlined the
necessary distinction to be drawn between information (the semantic content of
the utterance itself) and meta-information, which is information on the information
itself contained within the utterance. What J.-P. Sémon terms the field of ­incidence
(Sémon 1996: 205), which amalgamates both the context and the entirety of the

.  “Nexus” therefore has a different meaning here to that understood by Otto Jespersen, who
first used the term in linguistics (see A Modern English Grammar, 1909).
.  This is also the meaning of the term “predicate” in the M.I.C. theory.
 Olivier Azam

e­ xtralinguistic knowledge the speaker may have, combines at once both actual
information (that which has actually been spoken or written) and meta-informa-
tion. Yet it is at the same time the field of incidence as the speaker perceives it and
the knowledge s/he ascribes to the listener, (that is to say, still with reference to J.-P.
Sémon, to the subjective representation s/he constructs of the objective recipient
of the utterance) which motivate the pragmatic choices of the speaker, choices
which, in the Slavic languages, in turn govern word order and therefore have a
direct influence on syntax.
Finally the M.I.C. theory also has the merit of recalling the degree to which
the nebulous notion of “theme” is dangerous: too often it can be confused with
that of the subject.
As all of these criticisms seem entirely justified, M.I.C. terminology will be
adopted here. Thus, rather than “indivisible” utterances and “entirely rhematic”
sentences, I will discuss base utterances (M.I.C.), which is to say utterances in
which the nexal relations (established by the nexus with the different elements
which it controls and between these different elements themselves) are new or old.
The other utterances, those it was maintained could be divided into “theme” and
“rheme” are extended utterances in the M.I.C. theory: from a meta-informative
perspective, the speaker foregrounds a centre of attention through focalization or
on the contrary through topicalization, and sometimes through both.8 Focaliza-
tion consists in treating a centre of attention as new while a topicalized centre
is treated as given. However, in written texts, the distinction between focaliza-
tion and topicalization is not always easily grasped: for lack of a dedicated term
which would subsume both notions within the M.I.C. theory, I will simply refer to
“emphasis” when this distinction is dispensable. In the majority of cases however,
intonation allows one to distinguish between the two. Reusing the two adjectives
proposed by J.-P. Sémon (Sémon 1997) I will refer to tensive intoneme for the
topic and detensive intoneme for the focus.9

.  In this case, the focalization of one element is accompanied by the topicalization of
another.
.  In Russian the most common tensive intoneme (which the 1982 Grammar Book of the
Academy terms IK3) is not only ascending: the sharp rise produced on the syllable which
carries the centre of the intonation contour is followed, in the very next syllable, by a fall in the
voice which returns to a lower level than before the rise. The other intonation contours which
are rarer variants of IK3 (IK4 and in a more pompous style, IK6) are in fact mostly (IK4) or
entirely (IK6) ascending. In the vocabulary of the Grammar Book of the Academy, tensive
intonemes express nezaveršënnost’ (incompletion) and detensive intonemes zaveršënnost’
(completion). IK1 (neutral) or IK2 (a contrastive intoneme, an expressive variant of the pre-
vious term) in an enunciative sentence.
 Object and reflexive pronouns in modern Russian 

It is important to note that it is not only the constituents of the utterance, but
also the relationships maintained between these constituents through the nexus,
which can be new or old (already known). This precision is particularly important
when dealing with pronouns, since by definition a pronoun replaces a noun which
the recipient of the message should be able to restore. Except in the case of an
unusual effect, the referent of the pronoun is therefore always known. If the new
character, which means an utterance is extended, was only related to the identity
of its constituents, then the part of the utterance represented by the pronoun could
never, as it were, constitute the new element.

3.  The canonical place and displacement of the object pronoun

The definition of the canonical place of the object pronoun will serve as the prem-
ise for this study, which will present all of the potential variants in word order in
relation to this place.

3.1  The canonical place of the object pronoun

From a statistical perspective, it so happens that the canonical place is by far the
most frequent position of the object pronoun. It is important to clarify straight
away however that the canonical place understood here is not defined in statistical
terms. To draw out the legal metaphor suggested by the term “canonic,” one might
say that this is the place the pronoun occupies de jure and not de facto. Undoubt-
edly in this corpus there is a foreseeable coincidence between the rule and the fact.
From my perspective however it was in no way automatic.
That which I term the canonical place of the pronoun can only be uncovered
in a negative fashion: it is a “default” setting, unmarked, the place the pronoun
occupies “in all other cases,” that is to say in all those that will not be studied
here. From a methodological point of view, the definition of the canonical place
of the object pronoun in the Synodal Translation therefore constitutes the conclu-
sion of this analysis, even though it is used here as the starting point of the study
(the hypothesis to be proved). Finally I will add that the term canonical place
is conventional and reductive because it is not merely a question of place in the
local sense. In fact canonical place is defined by a dual criterion: a local criterion –
that of the relative position of the pronoun in relation to the verb which governs
it – and a prosodic criterion – the way in which the pronoun is expressed on the
melodic plane.
The systematic analysis of all the utterances of the four Gospels containing
object or reflexive pronouns has made it possible to determine with absolute
 Olivier Azam

­certainty that the pronoun is in its canonical place when it is in a melodic enclitic
position directly after the verbal nexus which governs it.
(0) 15. gorjačka ostavila ee; i ona vstala i služila im. (Mt 8)

Nevertheless it may sometimes occur that the pronoun in a melodic enclitic posi-
tion is itself preceded by a subject pronoun which is also enclitic (modern Russian
like Slavonic Russian allows two successive enclitics).
(0bis) 6. Nа vsjakij prаzdnik otpuskаl on im odnogo uznikа. (Mc 15)

The phenomenon is exceptional: in this case the presence of the subject pronoun
would be disregarded and it would be considered that the object pronoun still
occupied its canonical place.
Having posed the question of the canonical place of the pronoun it remains
to stipulate the conditions of its use. As I have stated, it is a default position that
the speaker will use whenever s/he does not feel the need to displace the pronoun
or to modify its melodic realisation, that is to say, in concrete terms, in two types
of utterances:

1. base utterances and

2. extended utterances in which the pronoun does not constitute the new element.

Everywhere else, I will argue that the pronoun is displaced.10

3.2  Th
 e displacement of the pronoun in an utterance containing a verb
governing a single expressed complement
In an utterance which only contains a single object complement of strongly
expressed rection, the pronoun, when it is not in its “canonical place,” may occupy
two locations:11 it can either be preposed, or postposed and stressed. The latter will
only be studied in the fifth segment, and I will begin here with the case of prepo-
sitioning (points 3 and 4).
When preposed to its governing verb, the pronoun complement in classical
Russian is always part of an extended utterance, never a base one.

.  By the displacement of the object pronoun I therefore mean not only a change of place
but also any deviation in relation to the two criteria which define its canonical place: the dis-
placement and/or change in its melodic materialisation.
.  When the context is sufficiently clear I will use the term place (“last place,” “change of
place”) when dealing simply with the “geographical” location of the pronoun in the linearity
of the utterance; but in ambiguous cases and if I wish to specify that I am not dealing with its
melodic realisation, I will rather use the words location or position.
 Object and reflexive pronouns in modern Russian 

However among utterances with a preposed pronoun, it is again necessary

to distinguish two distinct types of preposing which correspond to two kinds of
extended utterances which will be studied successively:

1. Either the pronoun is preposed in an extended utterance whose meta-­

informative charge is focused on an element other than the pronoun, in which
case the personal or reflexive pronoun will be simply tonic;
2. or else it is preposed in an extended utterance where the meta-informative
charge is focused on the pronoun itself, in which case the pronoun will not
only be tonic but stressed.

4.  The preposed tonic pronoun complement

The pronoun complement can remain in its canonical position when it appears in
an extended utterance in which it does not constitute the new element. However
it so happens that this is situated after the verb, and therefore towards the end of
the utterance.
Even if it is perfectly possible to place the new emphasised element of an
extended utterance after the postposed enclitic personal object pronoun (which
occurs when the pronoun is left in its canonical place) the speaker can chose to ret-
rograde the pronoun before the verb – the term is deliberately pejorative – in order
to completely free up the end of the sentence which is, as we know, the place of hon-
our, since in a neutral word order12 in Russian the old always precedes the new.
Thus displaced, the pronoun is no longer in the position of melodic enclitic. It
becomes tonic according to the definition given earlier without, however, becom-
ing stressed. It does not carry the centre of a neutral intonation contour at the
end of the sentence (of IK1 type), nor, a fortiori, is it the carrier of the centre of a
contrastive intoneme of type IK2 since the aim of its preposition here is precisely
to “render it forgettable” – to speak coarsely – in favour of a more important
element.
Examples 1–6 illustrate this kind of prepositioning of the tonic pronoun, in
which one can distinguish different cues for the emphasis of an element other than
the pronoun which justifies its displacement.

.  In speaking of word order here I am only referring to the higher level of m
­ eta-informative
analysis (the paleophoric – old or already known – versus the neophoric group) and not to the
order of terms within each group (here the pronoun would be part of the paleophoric group).
 Olivier Azam

4.1  The tonic preposed pronoun complement of a dependent infinitive

This cue can be the presence of a modal verb or another verb controlling an infini-
tive of which the pronoun itself is the complement. In this kind of construction
which contains an infinitive which is both governing and controlled, the pronoun
most often appears between the verb controlling the infinitive and the infinitive
to which the pronoun yields last place. In this type of utterance the focalisation
generally affects the nature of the action or the state expressed by the infinitive.
(1) 2. Gospodi! esli xočeš′, možeš′ menja očistit′. (Mt 8)

Clearly if the focalisation affects the pronoun itself and not the infinitive then the
pronoun can be displaced; it will be stressed and pushed in front of the modal or
other verb controlling the infinitive (see eg. 22) or conversely stressed and post-
posed to the infinitive (see eg. 49).

4.2  P
 reposed tonic pronoun in an interrogative utterance of empty
emphatic term
An even more telling indication of the emphasis of an element other than the
pronoun is the use of an empty emphatic term (an interrogative pronoun, adjective
or adverb formed from the root 〈*k〉) in an interrogative utterance, often accom-
panied by the preposition of the tonic pronoun.
(2) 46. Bože Moj, Bože Moj! dlja čego Ty Menja ostаvil? (Mt 27)

In this case of course one cannot assume a concern on the part of the speaker to
free up the last place for something else – since this location will remain unoccu-
pied – but perhaps a desire to disambiguate the utterance, above all if it is not spo-
ken but intended to be read: certainly the use of an empty emphatic does not, for
all that, exclude the possibility that another element of the utterance could carry
the highest meta-informative charge. Now if the pronoun were left in its habitual
location, one might believe (if one had not heard the original utterance spoken)
that it was the neophoric element and that it should be stressed.
In this corpus I have pointed out the preposed tonic pronouns used with
empty emphatic terms which mean that the focalization affects the action (eg. 2),
but also the object complement (Jn 8, 53), the means (Mc 9, 50) or the manner
(Mt 22, 17).

4.3  P
 reposed tonic pronoun in an utterance with a topicalized or focalized
subject
Remaining with the interrogative utterance, but this time with a full emphatic
term, the emphasis may affect the subject followed by the interrogative particle li:
 Object and reflexive pronouns in modern Russian 

(3) 34. vo grěxаx ty ves′ rodilsja, i ty li nаs učiš′? (Jn 9)

The emphasis which encourages the prepositioning of the object pronoun can also
focus on a stressed personal subject pronoun, which is particularly evident in a
negation when it directly affects the nominative pronoun, particularly in the case
of two independent elements coordinated by a which contain the same nexus: the
identity of the subject, denied in the first clause, is corrected by the second. In both
clauses, the subject pronoun carries the nuclear stress and the object pronoun is
preposed.
(4) 16. Ne vy Menja izbrаli, а Ja vаs izbrаl (Jn 16)

The emphasis of the subject can also take the form of the topicalization of the sub-
ject by means of the correlation kto…, (tot). Here again the pronoun is preposed
and tonic in both clauses:
(5) 11. On otvěčаl im: Kto menja iscělil, Tot mně skаzаl: voz′mi postel′ tvoju
i xodi. (Jn 5)

The topicalization of the subject in this kind of construction is accompanied by a

focalization of the stressed nexus, for which the last place is freed up by preposing
the pronoun.
(6) 26. Kto Mně služit, Mně dа poslěduet. (Jn 12)

The use of i as a discourse or pragmatic marker affecting the subject (in the
­following example i oni) is another means of underlining its emphasis, once again
through topicalization.
(7) 12. kogdа dělаeš′ oběd ili užin, ne zovi druzej tvoix, ni brаt′ev tvoix, ni
­rod­stven­ni­kov tvoix, ni sosědej bogаtyx, čtoby i oni tebja kogdа ne pozvаli,
i ne polučil ty vozdаjanija. (Lc 14)

When the focalized subject is a nominal group and it appears in its normal place –
in last place in a neutral word order – the pronoun is naturally displaced and
preposed.
(8) 54. esli Ja Sаm Sebja slаvlju, to slаvа Moja ničto. Menja proslаvljaet Otec
Moj, o Kotorom vy govorite, čto On Bog vаš. (Jn 8)

The emphasis on the subject (běsnovatye hereafter) can be underlined by one or

more appositions:
(9) 28. I kogdа On pribyl nа drugoj bereg v strаnu Gergesinskuju, Ego vstrětili
dvа běsnovаtye, vyšedšie iz grobov, ves′mа svirěpye, tаk-čto nikto ne směl
proxodit′ těm putem. (Mt 8)
 Olivier Azam

In an expressive word order, the focalized subject nominal group can be brought
to the beginning of the utterance and affected by a contrastive detensive intoneme:
(10) 25. Otče prаvednyj! i mìr Tebja ne poznаl (Jn 17)

In spite of their ostensibly near-identical lexical content it is important not to con-

fuse the previous example with an utterance such as
(11) 10. V mìrě byl, i mìr črez Nego nаčаl byt′, i mìr Ego ne poznаl. (Jn 1)

It is the wider context and above all the oral use of the corpus (the audition of the
recordings) which allows one to dissipate the ambiguity. In Jn 17, 25, the subject
is focalized and this insistence is strengthened by i which is not the coordinator
but has an adverbial value and means “also, and even”. In verse Jn 1, 10, on the
other hand, i is a coordinator with a consecutive value. In the second quotation
the subject is not focalized but topicalized, the focus affecting the nexus or, more
precisely, the nexus and the sign (“and though the world was made through him,
the world did not recognise him”). Whether there is focalization of the sign and
the nexus or focalization of the subject as in the first example, however, the result
is the same for the pronoun, which is preposed and tonic.

4.4  Preposed tonic pronoun in an utterance with a focalized nexus

Owing to the fact that in a base utterance with a transitive nexus and a nominative
complement the word order is of SVO type, it is sometimes difficult to distinguish
a base utterance from an extended one. When the complement is a pronoun how-
ever the distinction is easier to determine, to the extent that the speaker generally
decides to preposition if the nexus is focalized. Audition of the example given
below confirms that it is indeed this circumstance which brings about the preposi-
tion of the pronoun since the nexus is affected by a detensive contrastive intoneme.
(12) 26. Kto Mně služit, Mně dа poslěduet (Jn 12)

4.5  P
 reposed tonic pronoun in an utterance with a focalized object
complement
It is hardly surprising that an object complement would be focalized: even the
recourse to a predicative construction betrays the importance the speaker places
on the ascribed quality (expressed here by the adjective odin “alone/only”):13
(13) 32. vy rаzsěetes′ kаždyj v svoju storonu, i Menja ostаvite odnogo; (Jn 16)

.  In this example the emphasis could have affected the pronoun (through topicalization of
the object and the non-focalization of its complement): the pronoun would then have been
 Object and reflexive pronouns in modern Russian 

Far more exceptional is the use of a cardinal number object complement, a con-
struction impossible in English:
(14) 70. Iisus otvěčаl im: ne dvěnаdcаt′ li vаs izbrаl Ja? no odin iz vаs diаvol.
(Jn 6)

4.6  O
 ther focalizations which entail a recourse to the preposed tonic
pronoun
The focalization can affect the temporal (eg. 15) or locative phrase (eg. 16).
(15) 18. Esli mìr vаs nenаvidit, znаjte, čto Menja prežde vаs voznenаviděl.
(Jn 15)
(16) 7. i pojdite skorěe, skаžite učenikаm Ego, čto On voskres iz mertvyx i
predvаrjaet vаs v Gаlileě; tаm Ego uvidite. Vot, ja skаzаl vаm. (Mt 28)

5.  Stressed preposed pronoun

We have just seen one of the cases in which the object or reflexive pronoun can be
preposed: the speaker, at liberty to leave the pronoun in its canonical place, decides
to move it back before the nexus: the pronoun is then tonic but not stressed. This
occurs in an extended utterance whose informative charge is focused on an ele-
ment other than this pronoun.
The second case of prepositioning of the pronoun is also found in extended
utterances, but this time in those whose meta-informative charge is focused on the
pronoun itself. The speaker can then make the choice to prepose the pronoun to an
end diametrically opposed to that outlined in point 3: here the aim of the preposi-
tioning is not to “retrograde” the pronoun but on the contrary to emphasise it. In
this case, the speaker will not merely prepose the pronoun, s/he will also stress it,
that is to say will focus the intonation contour on it. The latter is usually contras-
tive, and it can be verified (infra point 6) that when the stressed pronoun precedes
the governing verb, the intoneme is usually tensive.
Examples 17 to 32 demonstrate the use of the preposed and stressed object
pronoun as well as the diversity of cues which can underline its emphasis. It is
unsurprising that many among them have already been mentioned before in rela-
tion to the emphasis, by focalization or topicalization, of an element other than the
pronoun. Naturally in the absence of formal cues, there remain many cases where

stressed and the sentence would have meant “As for myself, you shall leave me alone”. This was
not the choice of the readers of these texts however.
 Olivier Azam

only the field of incidence allows us to determine that the meta-informative charge
is focused on the object pronoun.
Since this study is concerned with recordings rather than on the written
­Gospel text, the classification of examples reflects finally the choices which the
readers were forced to make when faced with utterances which could be inter-
preted in two different ways. Here, however, ambiguity and hesitation can only
affect the choice between an extended utterance with a stressed pronoun and an
extended utterance whose informative charge is focused on an element other than
the pronoun: prepositioning to the nexus means confusion with a base utterance
is not possible.

5.1  Th
 e “placing in parallel” and prepositioning of the stressed complement
pronoun
All kinds of “placing in parallel” – ranging from basic constriction to opposition
and passing through several degrees of contrastive comparison and ­confrontation –
encourage the emphasis or “selection of Centres of Attention” (M.I.C.) within the
utterance, and, among the different possible selections, the emphasis of the pro-
noun through focalization or topicalization.
This recourse to the stressed preposed pronoun is particularly frequent when
this type of “placing in parallel” occurs between two clauses (most often coordi-
nated or juxtaposed, more rarely dependant) within the compound sentence.

5.2  U
 se in a compound sentence which contains independent elements
coordinated by no (topicalization or focalization of the object)
Thus the preposed stressed focalized pronoun is readily used in an independent
element coordinated by the adversative conjunction no with another independent
element of a different sign which corrects the information contained in the first
clause and more precisely the identity of the referent of the complement.
(17) 44. Iisus že vozglаsil i skаzаl: věrujuščij v Menja ne v Menja věruet, no v
Poslаvšаgo Menja; (Jn 12)
(18) 37. kto primet odno iz tаkix dětej vo imja Moe, tot prinimаet Menja; а kto
Menja primet, tot ne Menja prinimаet, no Poslаvšаgo Menja. (Mc 9)

Elsewhere, the focalization of the object pronoun affected by the negation can
even be reinforced by the use of the particle že (Mc 10, 40; Jn 17, 20).
It is also possible for the preposed stressed pronoun to appear in the clause
introduced by this aforementioned no. In the example given here the opposition
introduced by the adversative conjunction does not focus on the identity of the
object referent but on the entirety of the preceding clauses and the object pronoun
 Object and reflexive pronouns in modern Russian 

is not focalized but very clearly topicalized: the English translation of Lc  24,  24
(spoken with a distinct pause between ego and ne) would be “him they did not see.”
(19) 24. i pošli někotorye iz nаšix ko grobu i nаšli ták, kák i ženščiny govorili;
no Ego ne viděli. (Lc 24)

5.3  U
 se in an extended utterance containing independent elements
coordinated by a (focalization or topicalization of the object)
The use of the adversative conjunction a assumes both topicalization and focal-
ization, since there is a reprise of an element from the preceding clause and the
correction of another element, the corrected element being focalized in the clause
introduced by a.
In this corpus, the preposed stressed pronoun only constitutes the focus of
a clause introduced by a in one example (whose topic is precisely a first person
preposed and stressed personal subject pronoun which is followed by a pause):
(20) 11. Ja uže ne v mìrě, no oni v mìrě, а Ja k Tebě idu. (Jn 17)

Everywhere else, the object pronoun itself constitutes the topic of the proposition
introduced by a. In the examples given the focalized element of this second clause
can be:

–– an adverb (here vsegda)

(21) 11. ibo niščix vsegdа iměete s soboju, а Menja ne vsegdа iměete; (Mt 26)

–– or of course the sign of a clause

(22) 31. Podobno i pervosvjaščenniki s knižnikаmi, nаsměxаjas′, govorili drug
drugu: drugix spаsаl, а Sebja ne možet spаsti. (Mc 15)14
(23) 43. Ja prišel vo imja Otcа Moego, i ne prinimаete Menja; а esli inoj pridet vo
imja svoe, ego primete. (Jn 5)

5.4  U
 se in a compound sentence containing independent elements
coordinated by i (topicalization of the object)
In the following example where the conjunction i is used in a very similar man-
ner to a, the focalization of izbavim in the second clause is accompanied by the
topicalization of vas.

.  Compare on the one hand with Example 1, in which the pronoun is tonic and situated
before the infinitive (but after the modal) and, on the other hand, with Example 49, where the
pronoun is accented and situated after the infinitive.
 Olivier Azam

(24) 14. i esli slux ob ėtom dojdet do prаvitelja, my ubědim ego i vаs ot
­ne­prijatnosti izbаvim. (Mt 28)

In spite of the slight nuance introduced by the change of conjunction (the idea
of opposition is toned down) for this analysis the example should be considered
together with those of paragraph 4.3.

5.5  U
 se in a compound sentence containing juxtaposed independent
elements which have the same verb (focalization of the object reinforced
by i used as a discourse marker)
In spite of appearances the following example has only the prepositioning and stress
of the object pronoun in common with the previous one. Indeed here, even though
it is found at the beginning of the clause, i is used as a particle and not as a conjunc-
tion, the clauses are juxtaposed rather than coordinated. In this kind of coordinated
compound sentence where the two independent elements contain the same verb or
two synonymous verbs, the pronoun is no longer the topic but the focus:
(25) 1. Dа ne smuščаetsja serdce vаše; věrujte v Bogа, i v Menja věrujte. (Jn 14)

We will observe (infra point 5) that in this example it can also be postposed and
stressed.

5.6  Use in a compound sentence whose clauses are linked by subordination

Parataxis and coordination are not the only means of emphasizing the object pro-
noun. “Placing in parallel” can also be expressed through subordination and dif-
ferent kinds of relations of subordination can allow a topicalization of the subject,
a topicalization which, as I have shown, very often itself goes hand in hand with
the focalization of the object pronoun. The topicalization of the subject which pro-
vokes the prepositioning of the stressed form of the complement pronoun is thus
regularly obtained through the use of the correlation kto…(tot)15 (26) or through
its equivalent (27), a nominalised active present participle:
(26) 5. i kto primet odno tаkoe ditja vo imja Moe, tot Menja prinimаet; (Mt 18)
(27) 20. Istinno, istinno govorju vаm: prinimаjuščij togo, kogo Ja pošlju, Menja
prinimаet; а prini­mаjuščij Menja prinimаet Po­slаvšаgo Menja. (Jn 13)

However subordination can also allow the focalization of the object pronoun
without necessarily producing a topicalization of the subject. This is notably the
case in the setting of a conditional clause:

.  See supra 3.3.

 Object and reflexive pronouns in modern Russian 

(28) 8. Ja skаzаl vаm, čto ėto Ja; itаk, esli Menja iščete, ostаv′te ix, pust′ idut;
(Jn 18)

5.7  F
 actors which underline the focalization of the preposed pronoun
within the propositional frame
The justification for a recourse to the preposition of the stressed pronoun can be
found within the proposition itself, and several cues allow one to determine that
the pronoun is focalized here. These cues can be:

a. The presence of the adjective odin (“alone”) which serves to underline the
singularisation of the object;
(29) 10. Gospodu Bogu tvoemu poklonjajsja, i Emu odnomu služi. (Lc 4)

b. The presence, just before the pronoun, of an i used as an adverb and meaning
“also”:
(30) 45. Učitel′! govorja ėto, Ty i nаs obižаeš′. (Lc 11)

c. The presence following on from it of the interrogative particle li which indi-

cates that the interrogation affects the preceding word:
(31) 10. Pilаt govorit Emu: mně li ne otvěčаeš′? ne znаeš′ li, čto ja iměju vlаst′
rаspjat′ Tebja, i vlаst′ iměju otpustit′ Tebja? (Jn 19)

5.8  N
 o formal cue underlining the emphasis of the preposed pronoun
complement; this is deduced from the field of incidence
It can also occur however that no formal cue underlines the emphasis of the pre-
posed pronoun in the written text. The only formal indication of its emphasis is its
oral accentuation: the detensive contrastive intonation contour is focused on it. In
a written text intended to be read aloud such as this corpus, the readers’ choice to
stress the pronoun (illustrated by the example below) suggests a decision based on
a prior meta-informative analysis of the field of incidence.
(32) 5. Sej est′ Syn Moj Vozljublennyj, v Kotorom Moe blаgovolenie; Ego
slušаjte. (Mt 17)

6.  The postposed stressed complement pronoun

We have just seen (supra point 4) that in an extended utterance whose principle
meta-informative charge is focused on the pronoun (whether by topicalization or
focalization), the speaker can choose to accentuate it by preposing it to the nexus.
 Olivier Azam

However this preposition – unlike the accentuation16 which accompanies it – is

not systematic. The speaker can also choose to leave the pronoun in the position
it occupies when in its canonical place, that is to say after the nexus. This time
however the postposed pronoun will no longer be found in the position of melodic
enclitic: on the contrary it will always carry a contrastive stress.
When written, in order for the speaker to decide against prepositioning the
complement pronoun which carries the emphasis, the context must be suffi-
ciently clear to dissipate all ambiguity – unless, of course, this ambiguity is delib-
erately sought. In the absence of a sufficiently unequivocal context and field of
incidence the recipient of the message, misled by the position occupied by the
pronoun, could risk interpreting an extended utterance with the emphasis on
the pronoun as a base utterance, since the reader, unlike the listener, is denied
the crucial information provided by the intonation which alone allows one to
distinguish the stressed pronoun from one occupying its canonical place. In real-
ity, certain contexts are ambiguous and two interpretations of the postposed pro-
noun are possible. This restriction stated, however, it seems initially that for an
identical meta-informative analysis (extended utterance with emphasis on the
pronoun) the prepositioning and postpositioning of the stressed pronoun are two
interchangeable variants.
Verse Jn 18, 40 seems to support this, with an elliptical nexus but a stressed
pronoun:

(33) 39. Est′ že u vаs obyčаj, čtoby ja odnogo otpuskаl vаm nа Pаsxu; xotite li,
otpušču vаm Cаrja Iudejskаgo? 40. Togdа opjat′ zаkričаli vsě, govorja:
ne Ego, no Vаrаvvu. Vаrаvvа že byl rаzbojnik. (Jn 18)

If the verb otpustit’ were reestablished in verse 40, where would the complement be
positioned in relation to it? As long as the pronoun remained stressed, the meta-
informative analysis would remain unchanged whether it came before or after.
More troubling still, chiasmi, often considered as simple stylistic forms, also
appear to confirm that only the accentuation of the pronoun is relevant. There
are numerous chiasmi in the Gospels and certain extracts which contain them,
already used to illustrate the prepositioning of the stressed pronoun, could equally
be used here to illustrate how it remains postposed. One can add the following
examples to the previously cited chiasmi (eg. 17–18 and 27–28):

(34) 37. Vse, čtó dаet Mně Otec, ko Mně pridet; i prixodjaščаgo ko Mně ne
izgonju von; (Jn 6)

.  That is to say the fact that the pronoun is stressed in the sense I have defined it.
 Object and reflexive pronouns in modern Russian 

(35) 20. Pomnite slovo, kotoroe Ja skаzаl vаm: rаb ne bol′še gospodinа svoego.
Esli Menja gnаli, budut gnаt′ i vas; esli Moe slovo sobljudаli, budut
sobljudаt′ i vаše. (Jn 15)
(36) 48. i skаzаl im: kto primet sie ditja vo imja Moe, tot Menja prinimаet; а kto
primet Menja, tot prinimаet Poslаvšаgo Menja; ibo kto iz vаs men′še vsěx, tot
budet velik. (Lc 9)
(37) 16. Slušаjuščij vas Menja slušаet, i otvergаjuščijsja vas Menja otvergаetsja; а
otver­gаjuščijsja Menja otvergаetsja Poslаv­šаgo Menja. (Lc 10)

These chiasmi illustrate the two possible combinations: the preposed then post-
posed stressed pronoun (which is symbolized by A+P) and the postposed then
preposed stressed pronoun (P+A). As I will show however chiasmi are rarely lim-
ited to their most basic expression. In general they appear in complex combina-
tions such as P+A | P+P (supra eg. 27 Jn 13, 20); A+P+P (Lc 9, 48) or even in a
series of parallel and sequential chiasmi (P+A | P+A followed by a P+P combina-
tion in Lc 10, 16) or symmetrical and embedded chiasmi of P+A+P type in Jn 12,
44 (supra eg. 17) or P+A | A+P – a chiasmus within a chiasmus – in Mc  9, 37
(supra eg. 18). An important detail in this corpus is that the only isolated ­chiasmi
are of A+P type and they are comparatively plentiful (Jn 18, 8; Jn 15, 20 and Jn 37,
6): this frequency and the absence of isolated chiasmi of P+A type can be easily
explained by the ambiguity of the postposed pronoun in the written text as men-
tioned. When such a pronoun appears in a clause which follows a clause where it
was preposed and stressed, the reader, undoubtedly deprived of acoustic support
but alerted by the first prepositioning of a stressed pronoun, could easily see the
effect of the chiasmus and deduce from this that the postposed pronoun of the sec-
ond clause should also be stressed. On the other hand, if the context is not abso-
lutely unequivocal, it is harder to immediately discover a chiasmus of P+A type
on the first reading: glancing over the first clause, the reader may not notice that
the postposed pronoun is not in its canonical place but stressed, and only if the
chiasmus is involved in a wider play of oppositions (Jn 13, 20; Lc 10, 16; Jn 12, 44;
Mc 9, 37) will the reader discover, sometimes a posteriori, that the first postposed
pronoun should also be stressed.
In spite of these common sense restrictions on the use of isolated chiasmi
(P+A), the very existence of the chiasmi and their diversity appears to confirm
what the utterances with elliptical nexus and stressed pronoun suggested (eg. 33):
the relative place of the latter in relation to the nexus is not important as long as
the pronoun is stressed. It would therefore appear that we are dealing here with two
almost identical variants, what traditional Russian grammar complacently terms
ravnopravnye varianty. Linguistics will not be satisfied by the idea that there could
be no difference in a language between two distinct forms of expression however,
nor will it settle for admitting that the difference could be merely “stylistic”.
 Olivier Azam

It is therefore necessary to go further, but the analysis of examples of utter-

ances with a postposed stressed pronoun initially appear disappointing: it allows
one, effectively, to recapitulate point by point the classification of factors which
underline the emphasis of the stressed pronoun when the latter was preposed.
It appears then that the factors cited in point 4 are those which accompany the
accentuation of the pronoun and that they are in no way linked to its preposition-
ing, which seems to support the theory that here we are indeed dealing with two
interchangeable variants of word order and that the only relevant factor remains
the accentuation of the pronoun. Yet another reason to resume the linear clas-
sification proposed in point 4: this is the only possible way, should the occasion
arise, to display the differences between the two variants which might otherwise
go unnoticed.

6.1  “ Placing in parallel” and postpositioning of the stressed pronoun

complement: Use in a compound sentence containing independent
elements coordinated by no (focalization of the object)
In a clause introduced by no (not underlined), the stressed pronoun is also
postposed:
(38) 28. dščeri Ierusаlimskija! ne plаč′te obo Mně, no plаč′te o sebě i o dětjax
vаšix; (Lc 23)

Here, however, a second cue is added to the first: the coordination of the pro-
nominal complement with a nominal group of the same function i o dětjax vašix
(see. infra 5.5.d). Again in this example one finds a chiasmus. There are several
other examples of stressed postposed pronouns with no in the chiasmi already
quoted.

6.2  U
 se in a compound sentence containing independent elements
coordinated by a (topicalization of the object)?
Though the examples are relatively numerous when the stressed pronoun is pre-
posed, there is only one example where the pronoun is apparently postposed.

(39) 28. Tаm budet plаč i skrežet zubov, kogdа uvidite Avrааmа, Isааkа i Iаkovа i
vsěx prorokov v Cаrstvii Božiem, а sebja izgonjaemymi von. (Lc 13)

In fact the postpositioning is illusory: if one considers the entirety of the meta-
sentence (or coordinated compound sentence) in linear form, uvidite does indeed
precede sebja, but they do not appear in the same clause; uvidite is not reused in
the second clause and sebja is the complement of an elliptical nexus. As above
in the example of Jn 18, 39–40 (eg. 33) it is absolutely impossible to say whether
 Object and reflexive pronouns in modern Russian 

sebja should precede or follow the nexus in the case of the restoration of uvidite.
As we will see later (see infra point 6) sebja will in fact probably be preposed, from
whence the question mark which accompanies the title of this subheading.

6.3  U
 se in a compound sentence containing juxtaposed independent
elements containing common elements (focalization of the object)
The common element here is the noun mir (“peace”):17
(40) 27. Mir ostаvljaju vаm, mir Moj dаju vаm; ne tаk, kаk mìr dаet, Ja dаju vаm.
(Jn 14)

6.4  U
 se in a compound sentence whose clauses are linked
by subordination
Here again the topicalization of the subject often accompanies the focalization
of the object expressed by the stressed postposed pronoun. This topicalization is
obtained by the use

–– of the correlation kto…(tot)…

(41) 40. Kto prinimаet vаs, prinimаet Menja; а kto prinimаet Menja, prinimаet
Poslаvšаgo Menja; (Mt 10)

–– or of the active nominalised present participle:

(42) 45. i vidjaščii Menja vidit Poslаvšаgo Menja. (Jn 12)

As seen with the preposed stressed pronoun, subordination can also allow its
focalization when it is postposed without necessarily producing a topicalization of
the subject. This is particularly evident with a conditional clause:

(43) 15. Esli ljubite Menja, sobljudite Moi zаpovědi. (Jn 14)

Of course here as when the pronoun is preposed (see eg. 4.6), the use of a con-
ditional clause in itself does not provoke a recourse to the stressed form of the
pronoun, but it demonstrates that the presence of the former generally indicates
focalization. Here (eg. 43) the condition which must be fulfilled is that the action
expressed by the subordinate verb affects a specific object (the conditional clause
then underlining the importance of the selection of the object by focalization in
order to accomplish the action described in the main clause).

.  Without any ambiguity of meaning thanks to the classical spelling.

 Olivier Azam

6.5  F
 actors underlining the focalization of the postposed pronoun
within the propositional frame
The pronoun is still focalized here. The cues which accompany its focalization
can be:

a. the presence of the adjective odin (“alone”):

(44) 2. I po prošestvii dnej šesti, vzjal Iisus Petrа, Iаkovа i Ioаnnа, i vozvel nа
goru vysokuju osobo ix odnix, i preobrаzilsja pred nimi. (Mc 9)
b. the presence, just before the pronoun, of an i used as an enunciative particle
meaning “also”:
(45) 7. Bylo u nix i nemnogo rybok; blаgosloviv, On velěl rаzdаt′ i ix. (Mc 8)
When i affects the complement pronoun, it often combines with other focalization
cues such as a conditional clause (Jn 5, 46; 15, 20), a comparative relative clause
(Lc 6, 38) or even, as shown below, the coordination of the pronoun with an accusa-
tive nominal group (as in Jn 15, 24);

c. double negative ni

The place of the negation just before the complement pronoun is an obvious sign
of its focalization. With ni – a double negation – however, the cue itself is also
double because ni implies in addition the coordination of the stressed pronoun
with a nominal group of the same function (see 5.5.d).
(46) 19. vy ne znаete ni Menja, ni Otcа Moego; (Jn 8)
d. When the stressed pronoun is postposed, one often finds a cue as to its focal-
ization which was not found when it was preposed: the presence of a second
complement in the same case.

This second complement can be coordinated with the stressed pronoun. It may be
a question of a second stressed pronoun:
(47) 9. i zvаvšij tebja i ego, podošed, ne skаzаl by tebě: ustupi emu město; i togdа
so stydom dolžen budeš′ zаnjat′ poslědnee město. (Lc 14)
Or else a coordinated nominal group:
(48) 33. tаm rаspjali Ego i zloděev, odnogo po prаvuju, а drugаgo po lěvuju
storonu. (Lc 23)
Finally, the second accusative noun phrase could simply be apposed to the stressed
complement pronoun:
(49) 40. A teper′ iščete ubit′ Menja, Čelověka, skazavšago vam istinu, kotoruju
slyšal ot Boga. Avraam ėtogo ne dělal. (Jn 8)
 Object and reflexive pronouns in modern Russian 

It is evident in this last example that the pronoun is the complement of an infini-
tive and its place should be compared with the one it occupies in Example 1, on the
one hand, and 22, on the other.

6.6  N
 o formal cue underlining the emphasis of the postposed pronoun:
This is deduced from the field of incidence
As when the stressed complement pronoun is preposed, it can also occur that no
textual cue underlines its emphasis when it is postposed, with a greater concurrent
risk however. This risk is that the field of incidence may not be unequivocal enough
and will lead to an erroneous meta-informative analysis which would result in a
pronoun which should be stressed being read as a postverbal enclitic pronoun. In
fact it is precisely when focalization is not accompanied by easily identified textual
cues such as those I have described that readers hesitate and their interpretations
most often diverge.
As regards the classification of quotations under this heading I have therefore
followed a single reader (the same for all the examples), but this choice and the
divergences noted do not undermine the findings. On the contrary, they serve to
prove (1) that written utterances with postposed complement pronouns are struc-
turally ambiguous if no textual cue confirms that the pronoun focuses on itself
the highest meta-informative charge; (2) that it is indeed the intonation, (and not
the place of the pronoun) which plays the most important role, since without it,
doubt over the emphasis of the pronoun remains and (3) that the postpositioning
of a stressed pronoun is possible in spite of everything, even in the absence of cues
and despite the risk of confusion when written between the stressed postposed
pronoun and the pronoun in its canonical place.
In the following example the focalization of the object goes hand in hand with
the topicalization of the subject, affected by a tensive intoneme and separated from
the nexus by a pause:

(50) 7. itаk, esli Ty pokloniš′sja mně, to vse budet Tvoe. (Lc 4)

In Mc 2, 14, on the other hand, the reader’s decision to focalize the complement
was less foreseeable:

(51) 14. Proxodja, uviděl On Levija Alfeevа, sidjaščаgo u sborа pošlin, i govorit
emu: slěduj zа Mnoju. I on, vstаv, poslědovаl zа Nim. (Mc 2)

The reader could make this choice because he was reading a written text… As
soon as the text is spoken, however, no further doubt is possible: thanks to the
intonation the structurally ambiguous utterance becomes an extended utterance
with focalization of the pronoun.
 Olivier Azam

The final example is particularly interesting:

(52) 28. Vy slyšаli, čto Ja skаzаl vаm: idu ot vаs i pridu k vаm. (Jn 14)

One might indeed ask why, in the first clause (idu ot vas) the reader decided to
accentuate the pronoun, when he has the nuclear stress fall on the verb in the
second where the pronominal complement (k vam) is no more than an enclitic
occupying its canonical place. The reason for this choice is based on the change of
verb. What distinguishes the two clauses – and thus decides where the emphasis
is placed – is the direction of the motion. Now with a displaced verb without a
preverb such as idu the mono-oriented displacement can occur in several direc-
tions; this would then be expressed by the complement ot vas which would be
focalized, from whence the accentuation of the pronoun. On the other hand, in the
second clause, the direction of the mono-oriented displacement is already given
by the verb itself or, more precisely, by its preverb. The complement k vam there-
fore teaches us nothing more than 〈pri〉 already taught us, and there is therefore no
reason to focalize it.

7.  T
 opicalization or focalization of the preposed and postposed stressed
pronoun?18

As feared, the classification of cues which accompany the prepositioning and post-
positioning of the object pronoun are surprisingly similar… And yet on closer
examination it is possible to distinguish a fundamental difference: while the
examples of the prepositioning of the stressed object pronoun gathered together
examples of topicalized and focalized pronouns, it is striking to find that there are
no examples of a postposed topicalized pronoun. In this respect, a comparison of
points 4.2 and 5.2 is revealing: when used in a compound sentence which contains
independant clauses coordinated by a, all of the stressed pronouns are topical-
ized and are relatively numerous. Yet I have been able to establish that with a no
stressed postposed pronoun is found. It is scarcely likely that the total absence of
postposed stressed topicalized pronouns in the entirety of the Gospels is down
to chance; undoubtedly we hold the key here to the principle difference between

.  I would particularly like to thank Ch. Bonnot and H. Włodarczyk for having set me on
this path by suggesting that I explore the distinction between emphasis through topicalization
and focalization.
 Object and reflexive pronouns in modern Russian 

preposed and postposed stressed pronun: topicalization does indeed appear to be

incompatible with postpositioning.
It seems possible to advance as a rule that a stressed topicalized pronoun can
only be preposed. Does this mean that we are confronted here with a perfectly
balanced opposition, with prepositioning reserved for topicalization and post-
positioning for focalization? Certainly not, as the list of examples quoted in the
fourth section of this study proves. If a stressed topicalized pronoun is necessarily
preposed, the opposite is not true: the focalized pronoun can be either preposed
or postposed.
It seems part of the problem has been resolved, yet one question remains.
What difference is there between prepositioning and postpositioning of the focal-
ized stressed object pronoun? The meta-informative analysis of the sentence is in
fact the same here …
Without questioning all of the reservations expressed at the beginning of the
article over what constitutes the commonly taught pedagogic “mantra” of word
order in modern Russian, I believe it is nevertheless necessary to return here –
taking all necessary precautions – to the distinction made between a “neutral”
word order and one termed “expressive”.19 Indeed, the difference between the
focalized postposed and preposed pronoun, between On ljubit menja and On
menja ljubit (“I am the one he loves”) or, to return to one of the chiasmi seen
previously in which all of the pronouns are emphasized, between esli menja iščete,
ostav′te ix (“If it is I whom you seek, then leave them”) and the three other poten-
tial variants (esli iščete menja, ostav′te ix / esli iščete menja, ix ostav′te / esli menja
iščete, ix ostav′te) appears to be of the same type as that which distinguishes Knigu
kupil otec (with a contrastive accent) from Otec kupil knigu:20 both utterances
respond to an identical meta-informative analysis (“It is (my) father who bought
the book”), but the second constitutes a variant which, for convenience and lack
of anything better at the moment, will be qualified following tradition as more
“expressive”.

.  The concept of “expressivity” in a variant of word order is of course only admissible if
this term is understood in a relative sense: “expressive” here means “more expressive” than
another, since all word orders necessarily “express” something.
.  I will willingly take as a comparative element an extended utterance whose focus car-
rying the contrastive accent otec is a noun filling a different syntactic function – here, that
of subject – in order to underline that neither the syntactic function of the pronoun, nor its
pronominal nature have any bearing here on my argument.
 Olivier Azam

The comparison of utterances with focalized preposed and postposed pro-

nouns leads me to conclude that in the event of the focalization of the pronoun, a
neutral word order, unmarked, will maintain the stressed object pronoun in the posi-
tion it would occupy if it were in its canonical place. On the other hand for greater
expressivity the speaker can choose to prepose the stressed focalized pronoun.
Certainly, the preposition of the focalized pronoun is surprising, not because of
the place of a stressed pronoun within the utterance – since the stressed topicalized
pronoun is always preposed – but because of its melodic realisation.
In spite of appearances, formal differences persist between the preposed topi-
calized stressed pronoun and the preposed focalized pronoun in a more expressive
variant of word order. Let us return to a variant of the chiasmi (esli menja iščete,
ix ostav′te) in which the second pronoun ix could be interpreted as a topicalized
pronoun. For this, ix must be followed by a pause and the intonation must rise
sharply on the syllable of the pronoun which carries the lexical stress (the voice
can remain high or even fall sharply as soon as early as the following syllable if the
pronoun has a post-tonic syllable as in the instrumental plural imi):21 esli menja
iščete, ix↗| ostav′te↘. (If it is I whom you seek, then leave them.”). There would then
be two intonation segments and two different contours in the second clause. On
the other hand, if ix is focalized there is no pause, the entirety of the second clause
forms a single unit and the falling intoneme is focused on the pronoun: esli menja
iščete, ix↘ ostav′te.
When the stressed topicalized pronoun is preposed to the verb, it is in its
place, it carries a contrastive accent of the tensive type and is usually followed
by a pause (in a rapid delivery the pause can be imperceptible but the intonation
remains); conversely, when the stressed preposed pronoun is a focus, there is no
pause and the pronoun carries a detensive contrastive accent.22 Yet however “free”
word order is in Russian, a detensive intoneme or an end-phrase intoneme at the

.  See supra 1.4 and Note 9.

.  To be entirely exhaustive it is necessary to specify that very exceptionally a stressed post-
posed focalized pronoun can bear a tensive intoneme. This only happens in a very specific
situation: when the pronoun is the focus of the clause in which it occurs but this clause is itself
the topic of the entirety of the compound sentence. The tensive intoneme which indicates the
topicalization of the first clause focuses on the stressed syllable of the word which carries the
highest meta-informative charge – here the focalized pronoun – whose own intoneme, which
should have been detensive, it “crushes”: the melodic expression of the meta-informative or-
ganisation of the compound sentences takes precedence over the melodic expression of the
clause. This is what occurs in two of the variants of the chiasmi which I have studied (esli
iščete menja↗ |, ostav′te ix ↘ and esli iščete menja↗ |, ix↘ostav′te).
 Object and reflexive pronouns in modern Russian 

beginning of the utterance – and a fortiori a contrastive accent at the end of the
sentence – is less usual than a detensive intoneme at the beginning of the utterance,
the surprise it generates is therefore of a nature to more easily capture the attention
of the listener and fix it on the pronoun which carries the centre of this detensive
contour: this is what leads me to conclude that the prepositioning of the pronoun
constitutes the “most expressive” variant of word order when the stressed pronoun
is focalized.

8.  C
 onclusion: The position and melodic realization of pronouns in base
and extended utterances

The ultimate goal of the analysis of this corpus was to define the canonical or
default place of the pronoun: that which it occupies when it is placed just after the
nexus in the position of a melodic enclitic. I have in the first place asked the reader
to accept the veracity of this conclusion by presenting it as a principle which was
taken as the starting point and basis of this study.
Undoubtedly, by definition, a principle cannot be proven. It is perfectly pos-
sible to prove however that its negation is false: for this the reader must simply
examine each of the other possible realizations of the pronoun successively (pre-
posed tonic, preposed or postposed stressed) and to reexamine the entirety of this
study in order to determine that none of them could be seen as the default position.
Having established the definition of canonical place, from then on it was pos-
sible to present an algorithm which follows the rules which allows one to deter-
mine both the positioning and the melodic realization of the personal and reflexive
pronouns in an utterance which only contains a single expressed complement.23

1. In an non-extended utterance, the pronoun occupies its canonical place.

2. In an extended utterance
a. whose emphasis is not focused on the pronoun, the speaker can
–– either leave the latter in its canonical place
–– or choose to retrograde the pronoun by placing it before the nexus in
order to free up the end of the utterance. The pronoun is then preposed to
the nexus, tonic but not stressed.

.  Although it seems reasonable to believe that the rules given below can be applied to the
entirety of Russian classical literature, their validity will nevertheless only be vouched for in
relation to the corpus studied here: the Synodal Translation of the Gospels.
 Olivier Azam

b. where the emphasis is focused on the pronoun, the latter is not just tonic
(that is, non-enclitic) but stressed (bearing the centre of a contrastive
intoneme).
–– If the emphasised pronoun constitutes the topic of the utterance, it will be
preposed, bearing a tensive intoneme and usually followed by a more or
less perceptible pause.
–– If the pronoun constitutes the focus of the utterance, it will
–– bear a detensive intoneme and usually be postposed to the nexus (in
an unmarked word order)
–– but still bearing the centre of a detensive intoneme, it can also be pre-
posed to the nexus (in a more expressive word order), and in this case
it cannot be followed by any pause.
Translation from French by Dúnlaith Bird

References

Bonnot, Christine. 2008. Un cas ‘d’inversion’ de l’ordre canonique en russe moderne. La post-
position du pronom possessif épithète. In Questions de linguistique slave. Études offertes
à Marguerite Guiraud-Weber, 39–54. Aix-en-Provence: Publications de l’Université de
Provence.
Breuillard, Jean. 2008. Un cas d’enclise du sujet en russe. À propos des phrases du type Perevo-
dila Irina bystro. In Questions de linguistique slave. Études offertes à Marguerite Guiraud-
Weber. 55–65. Aix-en-Provence: Publications de l’Université de Provence.
Krylova, Ol′ga & Xavronina, Serafima. 1986. Porjadok slov v russkom jazyke. Moskva: Russkij
jazyk.
Lehfeldt, Werner. 2003. Akzent und Betonung im Russischen, 2nd edn. München: Otto Sagner.
Russkaja grammatika.Tom I. 1982. 99–122 (§155–171). Moskva: Nauka.
Sémon, Jean-Paul. 1995–1996. Interrogation et informativité. Paris: Cours polycopié de prépara-
tion à l’option linguistique de l’agrégation de russe à l’université Paris-Sorbonne.
Sémon, Jean-Paul. 1996. Imparfait français et prétérit imperfectif russe. In Études russes:
mélanges en l’honneur de Louis Allain, 205–216. Lille: Presses universitaires du Septentrion.
Sémon, Jean-Paul. 1997. Lexique de la chronopoïèse. Paris: service de reprographie de l’université
Paris-Sorbonne.
Włodarczyk, André. 2004. Centres d’intérêt et ordres communicatifs. In Enoncer, l’ordre infor-
matif dans les langues, Pierre Cott, Martine Dalmas & Hélène Włodarczyk (eds). Paris:
L’Harmattan.
Włodarczyk, André & Włodarczyk, Hélène.  2006. Focus in the meta-informative centering
theory. In La focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds).
Paris: L’Harmattan, collection Sémantiques.
Włodarczyk, Hélène. 2008. Du bon (et du mauvais) usage des formes accentuées et atones des
pronoms personnels en polonaise. In L’Enseignement du polonais en France, IIes Assises.,
39–48. Paris: Société française d’Etudes Polonaises, Institut d’Etudes Slaves.
Accented and unaccented pronouns
in Ancient Greek
A pragmatic choice by the speaker

Jean-Christophe Pitavy
Université Jean Monnet, Saint-Etienne

This chapter is an attempt to apply elements arising from Meta-Informative

Centering Theory (Włodarczyk 2004 and Włodarczyk & Włodarczyk 2006) –
to the situation in ancient Greek.
  The accented 1st or 2nd person pronoun may represent
  –  either a centre of attention refering to a topicalised colocutor
  –  or a centre of attention refering to a focalised colocutor
The choice between the accented form and the unaccented form is not
insignificant, and is motivated by the meta-informative structure of the utterance.
When the speaker has a choice between two forms (in cases where the corpus
offers us differentiated and alternating forms), the use of the accented form
corresponds in some cases to an extended utterance structure, and therefore to
an expressive utterance from the point of view of the opposition between one
part referring to “old” information and one part referring to “new” information.
In other words, this sets up a formal contrast – morphophonological and
syntactical in this particular case. Deciding whether one is dealing with a
topic or a focus brings into play numerous criteria which go beyond simple
pronoun usage. Choosing between these two possibilities refers us back to
meta-informative organisation: it is the speaker who decides on the pragmatic
role which a constituent part of an utterance must play in relation to the rest,
and in relation to the act of speaking itself, by taking into account the structure
and the elements available in the realm of discourse, and it is this which ensures
the coherence of what is said. There is therefore a minimum amount of freedom.
A centre of attention can only be topicalised or focalised relative to its counterpart
in the utterance (commentary and background respectively), but it must also
take into account all the relationships between the representations, information
and meta-information established and updated in the successive stages of the
discourse. We are therefore dealing with the case of a typical figure, well attested
in languages in which not a simple utterance, but an extended utterance is
chosen. Here the pronoun is the ingredient which allows the utterance to be
extended. A large amount of data, of which a small sample is presented here,
 Jean-Christophe Pitavy

seems to confirm that the use of the accented personal pronoun in conjunction
with the verbal form allows the speaker to construct an extended utterance.
The verification of this mechanism requires a meticulous examination of the
meta-informative structure of a very large number of examples of usage. There are
obviously exceptions, notably – and I have cited some of these – as regards certain
formulaic structures or those which are inserted at the level of meta-informative
commentary. P1 personal pronouns which engage the speaker as the necessary,
privileged reference point, are commonly introduced in such cases. Moreover, it
is not enough to examine the meta-informative structure of each usage in order
to decide whether we are dealing with topicalisation or focalisation. We must
establish more clearly the relationship between the choice apparently made by
the speaker and the processes supporting this structuring: certain parameters –
purely syntactical (the order of words) and pragmatic (particles) – have been
called upon in relation to pronominal marking. Beyond, however, presenting the
elements of the complexity of the situation in ancient Greek (elements arising
from both the uses of the language and the historical and material conditions
affecting access to the corpus), the aim has been to show the advantages of the
Meta-Informative Centering Theory in evaluating the use of certain strong
pronominal forms. Where it is recognised that these pronouns form part of an
extended utterance, it has proved possible and easier to differentiate between
focalisation and topicalisation.

0.  Introduction

In this chapter I would like to contribute to the theme of coherence by trying to

apply elements of the Meta-Informative Centring Theory1 to the use of accented
and unaccented pronouns in Ancient Greek. I will therefore refer both to certain
aspects of a system whose complexity is increasingly evident and to the varied
reality of usage to be found in literary corpuses,2 in an approach which seeks to

.  For the framework and corresponding terminology (centre of attention, focus, topic,
­extended utterance…), I rely, unless otherwise stated, on the Meta-Informative Centering
theory as presented in Włodarczyk (2004) and Włodarczyk and Włodarczyk (2006).
.  In this study I will not enlarge upon the particular conditions of usage and description
which apply to a language that is no longer spoken, and for which we cannot use the assess-
ments and other tests of acceptability available for modern spoken languages. These difficul-
ties can be partially compensated for by taking into account the variety of uses shown and by
the special place accorded to hypothesis.
To this we must add the representative nature, with regard to usage, of the literary
corpuses and the problem of the very value of these texts as testimony, since they have been
transmitted to us over a long period of time and have been subject to numerous accidents
 Accented and unaccented pronouns in Ancient Greek 

investigate the organisation of the different means offered by the language to mark
the category of person, inviting the elaboration of a theoretical hypothesis.
I will draw upon certain points – of two kinds, in language and in discourse –
as regards the use of accented pronouns in classical Greek that can be ascertained
from the corpuses. The linguistic field at the level studied here is characterised by
complex organisation and uneven illustration.
The corpus used consists of a collection of classical Greek prose: Plato’s dia-
logues and a choice of orators (mainly Lysias and Demosthenes). As the study
deals with P1 and P2 pronouns,3 it is obviously Plato’s texts which will provide
most of the examples.

1.  The theoretical dimension of the problem

1.1  The link between focalisation, topicalisation and coherence

From the point of view of the analyst, the value of focalising or topicalising opera-
tions is clearly linked to the meta-informative relationships between the constituents
and their referents in the successive stages of the discourse. The old or new status of
a centre of attention depends directly on the global coherence of the discourse for
the speaker and the addressee (Charaudeau et al. 2002: 99–100, Halliday & Hasan
1976: 4). I will therefore include coherence as a co-textual relationship, that is, a rela-
tionship between the intrapredicative elements and the extrapredicative information
of the universe of discourse. Furthermore, in the case of personal pronouns, it is also
important to suggest an “extra-textual” relationship with the referents of the centres
of attention involved, that is to say, the participants in the discourse.
For the pronouns dealt with here, information (semantic elements) is neces-
sarily available, since personal marks refer to each one of the participants actu-
ally involved in the egocentric speech event (see Lyons 1968: 275–276; Langacker
1987: 130). Modifications to the (meta)informative structure thus involve organis-
ing the greater or lesser role (the change from the status of old to new, for example)
that these elements are to play.4

along the way. I will not touch upon these general points, other than the last one, which has
consequences for the study of certain pronouns.
.  Cf. Abbreviations used at the end of this chapter.
.  One may add that these referents illustrate the “meta-informative” meaning of a feature as
new, for example, since the speaker, an indispensable reference point for the discourse, cannot
be unknown.
 Jean-Christophe Pitavy

1.2  Hypothesis
The initial idea is to use the propositions of the theory of centres of attention to
account for certain uses of accented pronouns in Ancient Greek. Rather than pro-
posing that the accented (tonic) form represents a focalisation (or an emphatic
structure, a different though similar operation), or inversely, that we are deal-
ing with topicalisation, in fact rather than considering that the meta-informative
structure is unrelated to the pronominal morphology, I prefer to acknowledge, by
following the distinctions of the theory mentioned earlier, that the accented pro-
noun may represent either a centre of attention referring to a topicalised speech
participant; or a centre of attention referring to a focalised speech particpant.
As a basic principle it is proposed that since the choice of the accented rather
than the unaccented form is not without importance, this choice is not made with-
out consequence or purpose, but is motivated by the meta-informative structure
of the utterance. As in many languages, it is not always easy to determine the fac-
tors behind the use of accented forms rather than their unaccented equivalents.
I suggest that, when the speaker has a choice between two forms (in cases where
the corpus allows us to see differentiated and alternating forms), the use of the
accented form corresponds in certain cases to an extended utterance structure,
and thus an utterance that is marked from the point of view of the opposition
between one part referring to “old” information and one part referring to “new”
information. In other words, this sets up a formal contrast – one that is, as it hap-
pens, morpho-phonological and syntactical. Working out whether one is dealing
with a topic or a focus brings into play numerous criteria which go beyond the
simple use of pronouns.

1.3  Formal criteria v. textual criteria

Some of these criteria are formal, and have to do with the way in which the
(extended) utterance is structured. We can thus mention
intonation;
word order;
the use of “particle” markers;
morphological marks, that is, the choice between a short or long “verb form” (in a
wider sense)

The different processes pose numerous and varied problems which clearly cannot
here be given the development that they deserve, as this would risk ­modifying
considerably the purpose of this study which concentrates on pronominal mor-
phology. Furthermore, we cannot take all of these processes into account in the
same way. Intonation marks in Ancient Greek, for example, are not at all well
understood. The corpuses at our disposal include signs of accentuation, added
 Accented and unaccented pronouns in Ancient Greek 

to the texts several centuries after their composition.5 These signs are thus only
the tip of an iceberg, about which we know almost nothing and for which we are
reduced to hypotheses about the system and its likely consequences in discourse.
At the same time, it goes without saying that the different processes are interde-
pendent and that for a variety of reasons none of them will be completely ignored.
Intonation, for example, is clearly linked to the other processes mentioned: the
word order (the place of unaccented units is determined by those that have an
accent), particles and the use of certain morphemes. As we will see, accents are a
deciding factor at certain levels of morphosyntactical organisation (certain mor-
phemes are made up on two series of forms, accented and unaccented).
Other criteria, which I would like to illustrate here, invite us to take account
of the relationships between the centres of attention organised in the utterance in
question and the information available co-textually. In my view this kind of exter-
nal relationship for the extended structure concerns the coherence of the discourse.

2.  The pronoun system in Ancient Greek

2.1  The pro-drop system: The case of the subject constituent

Since Ancient Greek is a pro-drop language, the person referred to by the subject
is frequently marked by the verb form alone. Pronoun forms in the nominative
therefore theoretically stand in contrast to the zero form:
(1) Ø esthíeis v. esthíeis sú
[Ø you-eat] [you-eat P2N]
“(you) eat”
But the situation varies according to the person. Thus, the contrast between the full
subject pronoun and the zero form is quite clear apart from the “third ­person” pro-
nouns (P3 and P6). Furthermore, except the subject, as concerns the morphological
material (language as a system of signs), one can distinguish at least three cases:

P1 and P2 pronouns, which have two series of forms which I will provisionally
call “weak” and “strong” at this initial stage;
P3 and P6, which have no specific personal pronoun;6

.  It is generally considered that the accentuation and punctuation of classical texts appeared
at the same time, both normally attributed to the Alexandrian grammarian Aristophanes of
Byzantium, in the 3rd century BCE. See Ildefonse (1997: 21).
.  This is true for the language system. The situation in discourse is in reality more complex.
For different grammatical functions classical Greek, like other Indo-European languages such
as Latin, relies on various forms of demonstratives (deictic and anaphoric forms from the
 Jean-Christophe Pitavy

P4 and P5, for which the contrast between accented and unaccented forms is a
theoretical one; it is assumed to exist at the language level by grammarians, but is
badly represented in the corpuses (where accented forms mainly dominate). This
contrast, virtually neutralised in discourse in classical prose, is inaccessible and
thus cannot be used in analysis.

2.2  P1 and P2

If therefore we set aside what cannot be used in practice, the opposition between
two series of forms only exists for P1 and P2. These two persons present similari-
ties and differences.

2.2.1  Similarities
P1 and P2 have two characteristics in common:

–– in the nominative, there is only one possible form (egô, sú);

–– for the other cases (accusative, genitive, dative), there are two series of forms.

The first characteristic is in fact common to all persons.

2.2.2  Differences
For P2 (Table 1), the opposition is only marked from the point of view of the tone
(a syllable of identical structure is either accented or unaccented):

Table 1.  The two series of P2 pronominal forms

Unaccented Accented

accusative se sé
genitive sou soû
dative soi soí

This morphological organisation poses serious problems for the practical

exploitation of literary corpuses dating from Ancient Greece. The texts have been
transmitted in a form somewhat different from their initial state, in particular as
far as the notation of accented forms is concerned. Thus for P2, only the accent

nearest to the furthest: hóde, hoûtos, ekeînos, see Perdicoyanni-Paléologou 2002). In several

­utterances quoted in this chapter – (3), (4) and (11) for example – a personal pronoun (P1, P2)
and a demonstrative replacing P3 are used in a similar way.
 Accented and unaccented pronouns in Ancient Greek 

differentiates the two series. Originally accents were not noted in the written form:
texts have thus often been given accents several centuries after their composition,
according to criteria imposed by successive editors. This historical peculiarity
means that we cannot trust the available forms of this pronoun as much as we can
trust the others. The contrast between the two series must be retained in discourse
by means of other, more or less complex criteria, relying upon the presence of a
certain kind of constituent, for example, or the position of the pronoun.
For a hypothetical understanding of the use of one kind of form, we can now
formulate the principle of the symmetry of value and function between the equiva-
lent pronouns of different persons. We can work out the distribution of unaccented
pronominal forms based on the more trustworthy distribution of the P1 pronoun.
In fact, for P1 (Table 2), the contrast from the point of view of tone is coupled
with that of the series of forms that I will here call “short” and “long” (this termi-
nology is not commonly used). This difference, which is actually morphological
(formal structure, number of syllables) is usually better preserved in the texts that
have come down to us:

Table 2.  The two series of P1 pronominal forms

Unaccented forms (“short”) Accented forms (“long”)

accusative me emé
genitive mou emoû

The opposition between the two series is normally more reliable in the
corpuses.
For both personal pronouns, the unaccented form is obviously clitic (­Creissels
1995: 24–28; Givón 1984: 83; Riegel et al. 1994: 200–204; Schiering 2006: 17–20;
Zwicky 1996: xii, xvii). There is therefore for the same information (the person)
“strong” and “weak” marking, in spite of the different means used:
P1: “long” accented form v. short clitic form;
P2: accented form v. clitic form.

2.2.3  Nominative v. other cases

We can liken the use of the subject pronoun with a verb form to the use, as a
complement, of an accented form in place of an unaccented form (Table 3):

Table 3.  Subject pronoun and accented complement pronoun (P1)

accented form egṑ légō emoì légeis
v. □ v.
unaccented form Ø légō moi légeis
 Jean-Christophe Pitavy

In both cases (Table 4), there is an accented (“strong”) pronoun form in one of
the contrasted terms instead of minimal marking (verbal inflection for the subject,
clitic form for the object):

Table 4.  Hypothesis of association between pronominal morphology and accentuation

for marking the person
“strong” marking “weak” marking

nominative accented subject pronoun + verb ending verb ending alone

other cases accented complement pronoun unaccented clitic complement

An important difference between the nominative and other cases can be seen
in the fact that there is a redundant structure for expressing the subject: the subject
constituent is marked twice, that is by the verb form and a separate pronoun.7
We may have reservations about likening the marking potential in one case to
that of another, but

–– in the language system, these structures are complementary: unlike other

cases, there is no enclitic subject pronoun in the nominative, and inversely,
there is no verbal mark for functions other than the subject;
–– more interestingly: forms also seem to correspond in discourse, so that we
can give the example of structures or expressions, with a specific, discursive
purpose – such as comment clauses –, characterized by the almost systematic
use of a subject pronoun (therefore in the nominative, Table 5):

Table 5.  The subject pronoun in comment clauses

P1: hōs ego oîmai P2: hōs sù oîei
as P1N I-think (P1) as P2N you-think (P2)
“in my opinion” “in your opinion”

These are parallel structures for both persons.

We therefore find the same distribution, that is, the use of the accented rather
than the clitic form for structures in which the subject constituent in the initial
expression becomes the complement of an impersonal expression:

.  A classic and well established analysis of pro-drop languages, suggest that the use of the
subject pronoun in this case is deliberate because there is “emphasis”.
 Accented and unaccented pronouns in Ancient Greek 

hōs emoì dokeî

(2) P1:
as P1D it-seems
“in my opinion”

2.3  The value of contrasting markings (strong v. weak form)

Our initial hypothesis, well established, is that grammar (morphology) gives the
speaker two possible ways of marking a personal constituent. That does not mean,
of course, that the speaker has absolute freedom in every case, but if we retain the
distinction between the purely semantic level (the expression of information) and
the pragmatic (meta-informative) level, we can on the face of it express it as two
points:

–– From an informative (semantic) point of view, the two solutions amount to

the same thing. Whether the personal constituent is accented (and redundant
in the nominative) or unaccented (and marked by the subject verb ending),
makes no difference: there is no more information given in the strong form
than in the weak form (c.f. for Example (1));
–– On the other hand, the choice between these two possibilities relates to the
meta-informative organisation: it is the speaker who decides on the pragmatic
role that a constituent should play (see Nølke 2001: 88–90) in relation to the
rest of the utterance, and in relation to the act of speaking itself, by taking into
account the structure and the elements available in the universe of discourse, a
decision which ensures its coherence. There is thus a very small amount of free-
dom. A centre of attention can be topicalised or focalised only in relation to its
corollary in the utterance (respectively comment and background), but also in
relation to all the inter-relationships of representations, information and meta-­
information established and updated in the successive stages of the discourse.

We therefore find ourselves dealing with a typical case, well attested in languages
in which one can choose not just a base utterance, but an extended one. Here, the
pronoun is the ingredient which allows the utterance to be extended.
Italian, a pro-drop language, is known for its pragmatic exploitation of the
use of the subject pronoun, combined with other parameters such as word order.
It thus sets up, for example, a contrast between three types of utterance in the
­following way:
(2) (a) l’ho fatto, “I did it”, simple utterance;
io, l’ho fatto, extended utterance in which the subject pronoun is moved
(b) 
to the left and plays the role of topic, roughly equivalent to the English
“As for me, I did it” (French “moi, je l’ai fait”);
 Jean-Christophe Pitavy

l’ho fatto io, extended utterance in which the same pronoun is moved,
(c) 
this time to the right, and plays the role of focus “I’m the one who did
it” , “I did it” (“c’est moi qui l’ai fait”).

We can of course compare the different solutions adopted by English and French,
which are non pro-drop language and combine a separated pronoun with syntac-
tical structures that also involve the order of the constituents (cf. Dik 1989: 131;
Lambrecht 1994: 115–116).
It goes without saying that these different processes are not enough in them-
selves: it is not enough to use structure (b) or (c) for valid topicalisation or focali-
sation. It is the coherence with the elements of the universe of discourse which
makes one or another type of meta-informative structure possible.

3.  Topicalisation v. focalisation

If we therefore say that the use of the strong form of a personal pronoun allows
for either the topicalisation or the focalisation of the constituent representing the
speaker or the addressee, the first criterion that determines the nature of the meta-
informative structuring is a recognition of the relationships of coherence with the
co-text or the immediate context.

3.1  Topic pronouns

There are examples in which it is possible to analyse the personal constituent as the
topic, that is, the set of information on the basis of which the comment is built. In
the following example, the speaker defines what each of the two listeners must do:
(3) SŌ. … ei mḕ boúletai Prōtagóras apokrínesthai, hoûtos mèn erōtátō,
ego dè apokrinoûmai…
I ptc I-will-reply
… epeidàn dè ego apokrínōmai hopós’ àn hoûtos boúlētai erōtân, pálin
hoûtos emoì lógon
huposkhétō homoíōs. Plato, Protagoras, 338 d 1–5.
“So.: […] If Protagoras does not want to reply, let him ask the questions;
I will reply […]. When I have replied to all the questions that he wants to
ask me, he should justify himself in the same way.”

Socrates proposes a change of role both for Protagoras (he is to ask the questions
rather than answer them) and for himself. The new information, for each one
of the stages, is not the identification of who must reply (answer: me), but what
the speaker (I) must do. The recognition of the semantic-pragmatic relationship
of this utterance to the co-text thus allows us to interpret the value of the two
 Accented and unaccented pronouns in Ancient Greek 

personal markers as representing the two parts of the topic.8 This is also the case
in the following example:

(4) Hēmeîs gàr epethumḗsamen… Theodótou, Plataïkoû meirakíou,

kaì ego mèn eû poiôn autòn ēxíoun eînai moi phílon,
conj. P1N ptc well treating P3acc I-judged to be P1D friend acc
hoûtos dè hubrízōn kaì paranomôn ṓieto anagkásein autòn poieîn hó ti
boúloito.(Lysias, Against Simon, 5, 2)
“We fell in love with Theodotus … a young boy from Plataiai. I wanted to
win his friendship by treating him well, whereas he (Simon) thought he
could force him to do what he wanted by treating him badly and being
violent towards him.”

The information that forms the basis of the topic is given in the preceding sen-
tence, and thus constitutes a shared topic (P4: “we [fell in love]”) which is split
into two parts in the following sentence: the pronoun “we” is taken up by the two
pronouns (egṑ “I”, hoûtos “the latter”). Each member of this pair (he, I) has his own
comment.
The following example is a little more complex:

SŌ. […] eboulómēn gàr án moi toùs lógous ménein kaì akinḗtōs hidrûsthai
(5) 
mâllon ḕ pròs têi Daidálou sophíai tà Tantálou khrḗmata genésthai. Kaì
toútōn mèn ádēn;
epeide dé moi dokeîs sù truphân,
conj. ptc P1d you-seem P2N to weaken
autós soi sumprothumḗsomai deîxai hópōs án me didáxē(i)s perì toû hosíou.
(Plato, Euthyphro, 11e1)
“So.: […] I would prefer my arguments to remain in place, that they should
be firmly fixed, rather than receive in addition to the wisdom of Dedalus,
the wealth of Tantalus. But enough of that. Since it seems to me that you
are weakening, I will join my efforts to yours to show you how you could
teach me about holiness.”

The subject pronoun is here used to return to the addressee (Euthyphro), after an
expression putting an end to further development. One can thus assume that he
is putting together a new topic: the sentence reintroduces Euthyphro (you), who
becomes (again) an object of worry for the speaker (Socrates).

.  Note the parallel structure used for each of the participants (he, I):

hoûtos mèn erōtátō egṑ dè apokrinoûmai…

[dem N ptc reply (imp P3)] [P1N ptc reply fut P1)]
 Jean-Christophe Pitavy

These few examples may be enough to illustrate the fact that it is not the use
of the subject pronoun itself, together with the verb form, which determines the
value of topic or focus. Such uses of pronouns seem to show that the pronoun
marks the topic according to the meta-informative structure – in this case the
available co-textual information about the referent corresponding to the pronoun.

3.2  Focalised pronouns

In other cases, the accented pronoun can be analysed as constituting new informa-
tion in relation to background that is available in the preceding co-text. This is the
case in the following example:
Eîen dḗ, ê d’hòs ho Sokrátēs, tà mèn Harmonías hēmîn tês Thēbaïkês ­híleá
(6) 
pōs, hōs éoike, metríōs gégonen: tí dè dḕ tà Kádmou, éphē, ô Kébēs, pôs
­hilasómetha kaì tíni lógōi? –
Sú moi dokeîs, éphē ho
P2N P1D you-seem said art
Kébēs, exeuresein.(Plato, Phaedo, 95 a 7)
NPr inf. fut.
“Come, said Socrates, now that Harmony, the goddess of Thebes, has seen
fit to be favourable to us, in moderation, how will we win the good grace
of Cadmus and using what language – it seems to me, said Cebes, that it is
you, Socrates, who will discover this.”

The addressee (sú “you”) appears as the element modifying the descriptions in
the preceding sentence (spoken by Socrates). Cebes corrects their roles, making
Socrates the one who must reply (the verb is P2 singular). He does not tell us that
there will be a reply to the question, but he indicates who will be responsible for
discovering it.
This kind of focalisation is common in Plato’s dialogues, notably when dealing
with the delicate task of replying or defining who should take on a specific role. In
the following example, on the other hand, there is no informative change as far as
either the process or the role of one of the participants is concerned:

(7) SŌ. […] ei gàr ek mèn tôn állōn tà zônta gígnoito, tà dè zônta thnḗiskoi, tís
mēkhanḕ mḕ oukhì pánta katanalōthênai eis tò tethnánai? – Oudè mía moi
dokeî, éphē ho Kébēs, ô Sṓkrates,
allá moi dokeîs pantápasin alēthê légein.(Plato, Phaedo, 72 d 5)
conj P1D you-seem absolutely true to say
“So. […] if in fact living creatures come from other living creatures, and
living creatures die, is there any escape from the final result that all things
would end in death? – None, in my opinion, Socrates, said Cebes, on the
contrary, you seem to say what is perfectly true.”
 Accented and unaccented pronouns in Ancient Greek 

The verb form is used with no personal subject pronoun (Ø  dokeîs légein “you
seem to say”). We have here a confirmatory utterance: in the final phrase, Cebes
brings nothing “new”, either about what is happening or about the role that
Socrates (“you”) is playing.
The case of an utterance involving focalisation because the variable of a con-
stituent is changed, can be illustrated by the following example:
(8) SŌ. Tí oûn tis ereî pálin ex arkhês epistḗmēn? ou gár pou aperoûmén gé pō? –
THEAI. Hḗkista, eánper me sú
conj neg P2N
ge apagoreúēis. (Plato, Theaetetus, 200 d 6)
ptc you-give-up
“So.: Therefore, going back to the beginning, how can we define science?
For we are not going to give up, are we? – theaetetus: Certainly not, at
least as long as you yourself don’t give up.”
The change (new information) involves the agent of the process “giving up” (apag-
oreúein: the two speakers use different forms of the same verb). The change is from
P4 to P2. Theaetetus introduces a stark contrast between the continuation of the
conversation between the two of them and the decisive, marked role played by
Socrates alone.
This change of person in what is an unchanging process can be seen in the
following injunctions:
(9) SŌ. Ê kaì hēúreté ti toioûton?
THEAI. Émoige dokoûmen; skópei dè
P1D+ptc we seem see ptc
kaì sú.(Plato, Theaetetus, 147e4)
also P2N
“So.: And have you found something [comparable to the roots]? – THE.:
I believe in fact that we have found something; but see if you also agree.”
The unchanging element here is the judgement asked for (in Socrates’ question:
“have you found…?”) and expressed (in Theaetetus’ reply). But this reply includes
a variable which corresponds to the person responsible for the stated point of view
(the subject of the judgement: “I believe…”). Theaetetus’ reply thus creates a con-
trast between two opinions: that of the speaker (émoige “[it seems] to me”), and
the one that is demanded of Socrates (skópei sú “you, see”).
(10) SŌ. Tò erōtṓmenon ou pánu moi phaínetai apokrínesthai. –
GOR. Allà sú, ei boúlei, eroû autón. (Plato, Gorgias, 448d6)
conj P2N if you-want ask P3A
“So.: I have the impression that [Polus] is not replying to what he is asked. –
GOR.: Well, question him yourself, if you want.”
 Jean-Christophe Pitavy

In both cases, the person in discussion with Socrates focalises him in relation to
a previously established background that is maintained (examination of a result,
questions). In the following example, P1 is opposed to P3 as the essential piece of
information modified in relation to an identical predicate:

(11) Kakeînos adikeîn mèn hōmológei, ēntebólei dè kaì hikéteue mḕ apokteînai all’
argúrion práxasthai. Egṑ d’ eîpon hóti
“ouk ego se apoktenô, all’ ho póleōs nómos…”
neg P1N P2acc I-will-kill but art city Gsg law Nsg
 (Lysias, On the murder of Eratosthenes, 25–26)
“And recognising that he was guilty, he asked me, begged me not to kill him,
but only to ask him for money. I said to him “It is not I who will kill you,
but the law of the City.”

The speaker can also produce an extended utterance to emphasize the importance
of a role (played by the speaker himself, in the following example):

(12) Oîmai d’humôn mnēmoneúein eníous hóti, hēník’ebouleúesth’hupèr tôn

­basilikôn,
parelthon prôtos ego pareinesa, oîmai… eipeîn hóti


approaching first P1N I-advised I-think to say conj
moi sōphroneîn àn dokoîte, ei…
 (Demosthenes, On the Liberty of the ­Rhodians, 6, 2)
“I think some of you remember that when you were pondering what was
to be done about the king, I was the first to come forward with advice, and
I urged you, I think I said that I thought it wise to…”

These different examples are obviously carefully chosen; in many cases, the use of
the strong pronoun cannot be so easily interpreted as far as its focal or topical role
is concerned. One can adduce two sets of reasons for this:

–– firstly, in discursive speech, the informative status of the personal constituents

is very complex. It is implied in a large number of structures which make it pos-
sible, for example, to mark the responsibility of the speaker (Ducrot 1984: 200)
in the form of a comment clause.9 The P1 and P2 pronouns involved in these
structures are thus not on the same plane as the constituents representing the
speaker or the addressee in the predicative nucleus of the sentence;

.  See the expressions meaning “in my opinion, in your opinion” given earlier, Table 5 and
Example (2) and Pitavy 2006b: 233.
 Accented and unaccented pronouns in Ancient Greek 

–– strong pronouns are only one of the marks in the construction of an extended
utterance, and we must take account of the role played in the m
­ eta-informative
structure by other processes.

4.  Auxiliary means of expression besides pronominal forms

The meta-informative structuring of an utterance can be achieved in a number

of ways: over and above morphological markers, we may cite for example: word
order, syntactic constructions (dislocations), grammatical markers (pragmatic
particles), prosodical marks such as intonation, and even lexical choice.10
In the case of Greek pronouns, we are of course dealing with several mark-
ers at once. A recognition of this multi-factorial situation, linked to the use of
pronoun accented and unaccented series, opens up interesting perspectives, but
has not as yet been sufficiently examined. Not all of these factors merit the same
degree of examination as far as Ancient Greek pronouns are concerned. Here we
will look at some of the markers used in combination with pronominal forms.

4.1  Intonation
The choice of a pronoun from one or other series is morphological and the dif-
ference in accentuation (clitic forms) indicates a prosodic contrast. The use of
this criterion is extremely limited, and it is impossible, for example, to take into
account criteria which are either unknown or totally hypothetical (since we have
no available data). Strictly speaking, the only indication we have of intonation and
rhythmical organisation comes from the written accent, which involves the prob-
lems that have already been mentioned.

4.2  Syntax
Clitic forms, which are attached to another constituent, clearly do not behave in
the same way as Tonic forms (called separate). Accented Greek pronouns are used
as noun phrases, which means they are “autonomous” in relation to the verb and
that they can easily be displaced, to the left or to the right of the verbal nucleus.
It is reasonable to suppose that the use of the same accented form, as in any
noun phrase, would not have the same value if the pronoun preceded the verb
rather than followed it.

.  Cf. Lambrecht (1994: 6), Dik (1983) for Greek word order and Dik (2003)
 Jean-Christophe Pitavy

I will here deal with two problems as regards the syntax of pronouns: clitic
forms, and the place of accented forms.

4.2.1  The place of clitics

Personal clitics that are verb arguments are usually used immediately next to the
verb phrase.11 This principle should not be overlooked: for P2, tonic complement
forms are only differentiated from clitics by the accent (cf. Table 1). This is, as we
have seen, an untrustworthy characteristic from a textual point of view, since the
texts were given accents after they were first written. Working out the preferred
place of accented pronouns in relation to clitics for P1 forms could produce a basic
hypothesis for “textually neutral” P2 forms.
The syntactical criterion is also a limitation on the freedom of ­meta-informative
organisation available to the speaker.
This is obviously true in the case of constructions in which the contrast
between accented and unaccented forms is practically neutralised in discourse:
for example, after a preposition (with an argument that is the indirect object of the
verb) an accented pronoun is almost always used (Table 6):

Table 6.  P1 and P2 pronoun forms governed by prepositions

“by me, by you” hup’ emoû, hupò soû v. ? hupó mou,? hupó sou
“near me, near you” par’ emoí, parà soí v. ? pará moi,? pará soi

As a matter of fact, there can be found in Greek prepositional phrases in which

it is the preposition, usually proclitic, which bears the accent of the group that it
forms with a clitic pronoun:12

.  In Greek, as in a number of languages, clitics are considered to be excluded from first
place in the sentence and typically appear in second place – these are “Wackernagel” clitics, a
reference to the “law of position” set out in Wackernagel (1892).
.  This usage brings to mind phrases with prepositional accentuation to be found in certain
Slavonic languages (for example, in Russian из дому íz domu “outside one’s home”, гулять
по полю guljat’ pó polju “walk in the field”; Polish: na dół “downstairs”, wypić do dna “drink
in one go”, etc.). This kind of proclitic attraction is presented as being disappearing in certain
­descriptions, cf. among others Comrie and Stone (1978: 82–85) and Proeme (1994: 361). In
fact the situation in Ancient Greek is very different: whereas in Slavonic languages this place-
ment of the stress on the first clitic takes place in Prep+N sequences, in Greek, where the
phenomenon is extremely rare and is attested to in accentuated forms, this accentuation is
normal, since it characterizes a sequence of clitics (Prep+Pro). In this case it is the use of
an enclitic pronoun which is noteworthy. Being proclitic, the preposition governs, morpho-
phonologically, a form bearing the accent for the PP that it forms with it. As there is no reason
 Accented and unaccented pronouns in Ancient Greek 

(13) Theódōros polloùs dḕ prós me epainésas

Theodorus N many acc ptc in front P1acc having-praised
xénous te kaì astoùs…(Plato Theaetetus145 b 8)
foreigners acc as-well-as citizens acc
“… although Theodorus praised in front of me many foreigners and
­citizens.”

These cases are, however, rare and do not provide grounds for suggesting that the
(almost non-existent) choice between prep + accented form and prep + clitic is a
significant one.

4.2.2  Accented pronouns and word order

The order of “full” constituents, in so far as it can be discerned from the texts, is the
subject of a great deal of attention in current research on the informative structure
of Ancient Greek.13
The old, simplistic assumption that accented pronouns have an emphatic
value (Kühner et al. 1898–1904: 456; Gildersleeve et al. 1911: 35; Schwyzer et al.
1939–71: 186–187; Crespo et al. 2003: 44) has been strongly called into question in
recent years:14 the use of accented forms alone would not be sufficient to suggest
that they have a focus value (or indeed a topic value, since the “emphatic” use of
such forms can correspond to either pragmatic role).
As far as the use of accented pronouns is concerned, the question of syntacti-
cal order must also be looked at from the point of view of the position of the pro-
noun in relation to the verb (anteposition or postposition).
Thus certain examples, taken from both prose and classical dramatic poetry,
show that there is no emphasis in the case of accented pronouns placed after the
verb. Translated into the terminology used here, this means that the use of a post-
posed accented pronoun does not make an extended structure, but an unmarked
variant of a simple utterance.

to think that the simple use of a prepositional construction involves the use of an extended
structure, and as examples of post-prepositional clitics are extremely rare, we must conclude
that the speaker has no other choice of pronominal form and that, independently of the meta-
informative conditions, the morphological contrast is neutralised in the sentence and in the
utterance.
.  On word order in Ancient Greek, see Dik  (1995) for data taken from the works of
Herodotus.
.  Dik (2003), who proposes the absence of emphatic value in certain postposed personal
subject pronouns.
 Jean-Christophe Pitavy

Although this chapter is specifically devoted to the conditions of usage of

accented forms, as opposed to unaccented forms, it is important to clarify what
we can learn from a study of pronoun uses looked at from the point of view of
syntactical order. I will not here take account of all the data provided by the refer-
ence corpus on this question, but, as previously, I will give examples to illustrate
some types of uses. The incidence of syntactical order can be illustrated by noting
certain patterns (structures involving prototypical verbal expressions).

4.2.3  The anteposed accented pronoun

In different ways, the data dealt with here does not really undermine the hypoth-
esis that the postposed accented pronoun does not set up an extended utterance.
In the examples already quoted, it is in fact the anteposition which is dominant.15
There is one example of postposition, to which I will return later. We may of course
consider that the examples illustrating uses mentioned in this chapter are not
representative as regards the proportion of postposed forms used in texts. A full
examination of the corpus of texts by Plato, Lysias and Demosthenes provides
more exact data. We can quote some typical structures which illustrate the con-
trast between the two positions in relation to the verb.16 For the purposes of this
presentation, I will keep to the following structures: pron. subject (P1, P2) + the
verbs “to say” and “to believe, think”.
These constructions (an accented pronoun with a verb form) confirm the
great majority of anteposed accented pronouns.

4.2.3.1  The verbs “to say” and “to think” + pronoun.  The P1 pronoun egṓ pre-
cedes légō “I say” in 73 cases,17 whereas the reverse word order (légō egṓ) is only
found once.18
The P2 pronoun sú is used in Plato 85 times19 before légeis “you say”, whereas
it only appears after the same verb form 5 times.

.  In the Example (5) dokeîs sù truphân, we may consider the subject pronoun anteposed
to the modal auxiliary (“to seem + infin.”). In all the examples of sú with dokeîs + inf., the
pronoun always precedes the infinitive.
.  As we are only talking about the overall illustration of a corpus which merits a much
more developed study, I here consider anteposed any pronominal form placed in the p ­ reverbal
zone, whether it is immediately next to the verb or not, and the same is true for postposition
cases of the same forms, in the postverbal zone.
.  Plato: 56 instances; Lysias: 3 and Demosthenes: 14.
.  Plato, Letters.
.  Plato: 82 instances; Lysias: 1; Demosthenes: 2.
 Accented and unaccented pronouns in Ancient Greek 

The verb oîmai “to believe, think”, almost systematically used with a subject
pronoun, notably in discursive comments (see above), has the pronoun egṓ before
the verb 78 times in all of these authors,20 whereas the reverse order (egṓ oîmai) is
only found 15 times in Plato and only once in Demosthenes.
Expressions involving a verb of opinion, with the personal pronoun in the
dative (of the kind dokeî + P1/P2 “it seems to me / you”) confirm this quantitative
data in Plato and the orators.
Even if we take the null hypothesis, that is that the use of the (accented) sub-
ject pronoun together with the verb is not significant, the preverbal position is
dominant. If we now assume that, on the contrary, the accented form corresponds
to a marked usage, this value would be largely confirmed by the place of the pro-
nominal constituents. In other words, when it is used, the “strong” form of the
pronoun contributes to an extended structure, just as anteposition does in relation
to the verb.

4.2.3.2  Preverbal v. postverbal accented pronoun.  At the same time, in so far as

the corpus presents some examples in which both positions are possible, we may
ask whether the use of the accented pronoun allows us to hypothesise a difference
of value between forms in the preverbal zone and those in the postverbal zone. We
can find some answers in the following examples:

(14) «SŌ. Éti dḕ tríton skepsṓmetha ei sundokeî soí te ka hēmîn. – CRIT. Tò poîon
dḕ toûto? – SŌ. Egṑ dḕ phrásō.
Dokeî gàr dḕ emoí te kaì tôide,
it-seems in-fact ptc P2D as-well-as DEMD
perì hósōn estìn epistḗmē…»(Plato, Laches, 198 c 11)
“So. We still have to examine a third point in order to know whether you
agree with us. – critias: what point? – So. I will explain. He and I both
have the impression that everything for which there is knowledge…”

Here we have the one single occurrence in the entire corpus of the postposition of
the accented pronoun, in a construction in which the pronoun is usually placed in
the preverbal position: emoì (gàr) dokeî. As I mentioned earlier, subject pronoun
forms and accented forms are used in Greek as constituents. Here, therefore, we
must consider the personal pronoun coordinated with a demonstrative pronoun to
have the same syntactic function (tôide “to the latter”). The use of the unaccented
form of the personal pronoun is not possible: one cannot have *moi coordinated
with a NP in the dative or a demonstrative pronoun in the same case, both struc-

.  Plato: 39 instances; Lysias: 7 et Demosthenes: 32.

 Jean-Christophe Pitavy

tures having the status of constituent. In this case, the postverbal position can be
analysed as being independent of a possible choice of the accented form by the
speaker. The utterance is opened by the verb form placed at the start and u
­ nderlined
by gár (focalising value), and since the complement in the dative includes a demon-
strative pronoun (tôide), the personal pronoun is automatically accented.21
With the Example (9), given earlier, we can analyse a comparable situation
with the marker kaí:

(9) THEAI. skópei dè kaì sú.(Plato, Theaetetus, 147e4)

examine ptc also P2N
“theaetetus: see if you also agree.”

The corpus contains many examples of imperatives with a subject pronoun pre-
ceded by the inclusion marker kaí (“also”). In fact this adverbial marker always
precedes the constituent to which it refers, which is autonomous in relation to the
verb (NP, adverb, tonic forms). The unaccented forms me, se, moi, soi etc. and,
with all the more reason, the zero nominative pronoun, cannot therefore be accen-
tuated by kaí. In the Example (9), since the subject is focalised it is bound to be
marked by the tonic form, independently of its position in relation to the verb.

4.2.3.3  Syntax and clause type.  The corpus offers another, particularly interest-
ing example of the same syntactical construction carried out in two different or-
ders and presenting a contrast that is both formal and meta-informative:

(15) Tí légeis sú? éphē. – Hà sù légeis,

What you-say P2N? said-he what P2N you say
émoige dokô. (Plato, Republic, 339d3)
P1D I-seem
“thrasymacus: What are you saying? – SO.: What you yourself are
­saying, it seems to me.”

It is worthy of note that the second expression, which in formal terms includes
the same constituents as the first, shows an inverse word order. In Socrates’
reply, we clearly understand from the micro-context that it is the subject,
“you” which is focalised by the anteposed accented subject pronoun, whereas

.  The corpus has only one other example of P1D paired with other pronoun and noun
forms in the dative, a complement of the same verb (dokeî), but in the preverbal position: SO.
“ … Emoì kaì humîn kaì hólēi tê pólei oú moi dokeî…” “for me, for you, and for the whole City,
it doesn’t seem to me that…” (Plato, Apology, 34 e 2).
 Accented and unaccented pronouns in Ancient Greek 

in ­Thrasymacus’ question, we see postposition in an utterance which does not

as readily have a “strong” value: why would the speaker insist (focalisation) on
Socrates at that point?22 Thrasymacus’ question does not therefore seem to have
an extended (focalising) structure, unlike the reply. Indeed, the structure *tí sù
légeis? is not found in the corpus, whereas there are three examples of tí légeis sú?
We must furthermore take into account the fact that some of the occurrences of
the postpositon of the subject pronoun with this sort of verbal expression appear
in questions.23 This is probably because there is an aspect of informational hier-
archy dictating the word order: in the question in Example (15) the focalising
element, the interrogative tí (“what”) is anteposed. It would therefore be difficult
to have a sequence using anteposition of the kind: *tí sù légeis.24

4.3  Pragmatic particles

Greek is rich in particles which allow a constituent to be assigned a certain meta-
informative status. These particles can intensify the assertive value of the utterance
(dḗ: “really”, toínun “indeed, certainly”) or a restriction (ge “at least”), like certain
intensifiers in English or in French. Here too, a study of the association of a par-
ticle with one of the pronoun forms can teach us a great deal about the pragmatic
status assigned to the personal constituent in question.
Considerable work needs to be done simply on the description of particles
and an analysis of their pragmatic uses. As this chapter is more specifically devoted
to the role played by certain pronominal forms, I will only give a few examples of
the association between particles and pronouns.

4.3.1  Dḗ
In the corpus of Attic Greek prose writers, for example, the insistence particle dḗ
“really”, is above all used with subject pronouns (twenty occurrences evenly dis-
tributed between the two persons: 9 egṑ dḗ and 11 sù dḗ) while it is rarely used with
complement forms: 1 emè dḗ ; 1 emoû dḗ; 1 emoì dḗ; 3 soì dḗ; 1 sè dḗ. For example,
we can cite:

.  It would of course be possible in a different situation where the addressee is emphasised
in contrast to the other participants in the debate, but that is not the case here.
.  Out of the 5 occurrences of the sequence légeis sú, 4 are interrogative (Symposium,
­Charmides, Euthydemus, Republic).
.  This seems to be confirmed by the fact that in our corpus, sequences of accented, focalised
interrogative pronouns + accented personal pronoun of the type tí émoi, are not found, whereas
one can find without difficulty the variant with the unaccented personal pronoun (tí moi).
 Jean-Christophe Pitavy

(14) Taûta de autós te skopô kath’ hóson dúnamai,

dem ptc pro coord I-investigate as-much-as I-can
kaì toùs állous erōtô hoîs àn horô toùs néous ethélontas suggígnesthai.
Soì de ouk olígistoi plēsiázousi,
dem ptc neg very-few they-come
kaì dikaíōs.(Plato,Theaetetus, 143 d 7)
coord rightly
“These are the things I myself investigate, as far as I can, and about
which I question those others with whom I see that the young men like to
associate. Now a great many of them come to you, and rightly”

In this passage, in which Socrates addresses Theodorus who teaches geometry,

the same particle dḗ is used twice with pronominal constituents to topicalise old
centres of attention (anaphoric demonstrative taûta) associated with a comment
that brings in new elements. It is these meta-informative relationships between the
phrase in question and the preceding co-text which allows us to confirm the old
status of the centres of attention marked by this particle and the new status of the
piece of comment that follows. It is premature, in a work mainly given over to the
use of pronouns, to say whether this particle serves exclusively to construct topics.
It seems more to the point to propose a principle of cooperation between at least
three devices:

–– the use of a disjunct accented form;25

–– the use of an intensifying particle (dḗ);
–– the position  of the pronoun: initial and possibly separated from the verbal
core.

These three devices used together allow for the construction of an extended utter-
ance involving either a topic or a focus.

4.3.2  Ge
With the particle ge (cf. Denniston 1934: 114–115), the incidence of morpho-­
syntax (the pronoun form) with pragmatics (the use of ge) is even clearer,
although it must be specified that this particle is enclitic, that is to say it is

.  Here I am of course putting forward the hypothesis that the word soi in soì dḗ should be
interpreted as tonic (“strong”), not clitic, for a number of reasons. Syntactically a clitic cannot
be used in first place (this confirms Wackernagel’s law). From the point of view of the use of
pragmatic particles, accented forms are normally used with dḗ or ge (see the summary given
at the beginning of Section 4.3.1.). These parameters thus allow us to confirm the accentuation
given by the texts.
 Accented and unaccented pronouns in Ancient Greek 

associated with the preceding constituent in an accented group. If this preced-

ing constituent is itself clitic, it is given an accent, so that unaccented P2 + ge
sequences can no longer be distinguished from accented P2 + ge sequences.
There are very few occurrences of ge with unaccented forms (1 mé ge: 2 moí ge)
whereas accented forms are much more numerous with P1 (372): 9 emoû; 11
emé; principally in the dative (352 émoige). This data is obviously distorted: the
dative émoige “to me” is primarily used in formulae of agreement or confirma-
tion such as émoige dokeî “to me it-seems”, which are used for acquiescence (a
confirmation structure equivalent to yes, see Pitavy 2006a: 261) and are therefore
predominantly in the first person. For P2, there are 21 uses with “textually neu-
tral” forms (12 soí ge; 7 soû ge; 2 sé ge). We may surmise that these forms are in
fact accented, if we apply the principle of symmetry with P1 pronouns, which
was demonstrated and used earlier.
(14) … oúkoun soí ge… ónoma Hermogénēs,
is-it-not P2D ptc name NSg Hermogenes
oudè àn pántes kalôsin ánthrōpoi…(Plato, Cratylus, 383 b 5)
“Your name is not…Hermogenes even if all mankind call you so”

The speaker, Cratylus, is presenting a theory on the meaning of a person’s name

in relation to their personality. After going through several names, he comes to
the case of his addressee, Hermogenes (“as for you…”) whom he treats differently
using a play on words: Hermogenes means “son of Hermes” the god of merchants
and bankers, which does not suit the person in question since he has not suc-
ceeded in business. The second part of the utterance is thus a comment (a para-
doxical one in this case) on the topical addressee.

Conclusion

A great deal of data, a small sample of which has been presented here, seems to
confirm that the use of the accented personal pronoun at the same time as the
verb form enables the construction of an extended utterance. A verification of
this principle would demand a meticulous examination of the meta-informative
structure of a greater number of uses. There are obviously exceptions, and I have
quoted some of these, notably relating to certain formulaic structures or those that
are used at the level of meta-informative comment. P1 personal pronouns which
involve the speaker as a privileged, constituent reference point, are particularly
common in such uses.
Secondly, it is not enough to look to the meta-informative structure of each
usage in order to decide whether we are dealing with topicalisation or focalisation.
 Jean-Christophe Pitavy

A relationship must be more subtly established between the choice a­ pparently

made by the speaker and the devices assisting this structuring: certain purely syn-
tactical (word order) and pragmatic (particles) parameters have been brought into
play, linked to pronominal marking. Admittedly, an analysis of corpus data sys-
tematically enlarged at these formal levels would allow for a more accurate and
even clearer separation of the pragmatic strategy adopted.
But over and above the presentation of the complex elements of the situation in
Ancient Greek, (elements that result from both the language and the historical and
material conditions of access to the corpuses), the aim of this study was to look at
the advantages that the meta-informative centring theory presented in relation to
assessing the value of the use of certain strong pronoun forms. In cases where these
pronouns are shown to contribute to an extended utterance, it has proved possible
and easier to appreciate the distinction between focalisation and topicalisation.
 Translation from French by Rosemary Masters

Abbreviations used

acc: accusative inf: infinitive

art: article n: nominative
conj: conjunction neg: negative
coord: coordinating conjunction np: noun phrase
d: dative npr: proper noun
dem: demonstrative p1, p2 etc.: first, second person…
fut: future pl: plural
g: genitive ptc: particle
imp: imperative sg: singular
ind: indicative

References

Charaudeau, Patrick & Maingueneau,  Dominique.  2002. Dictionnaire d’analyse du discours.

Paris: Éditions du Seuil.
Combettes, Bernard. 1983. Pour une grammaire textuelle. La progression thématique. Brussels-
Paris: De Boeck.
Comrie, Bernard & Stone,  Gerald. 1978. The Russian Language since the Revolution. Oxford:
OUP.
Creissels, Denis. 1995. Éléments de syntaxe générale. Paris: Presses universitaires de France, Coll.
Linguistique nouvelle.
Crespo, Emilio, Conti, Luz & Maquieira, Helena. 2003. Sintaxis del griego clásico. Madrid:
­Editorial Gredos.
 Accented and unaccented pronouns in Ancient Greek 

Denniston, John Dewar. 1934. The Greek Particles. Oxford: The Clarendon Press.
Dik, Helma. 1995. Word Order in Ancient Greek. A Pragmatic Account of Word Order Variation
in Herodotus. Amsterdam: J.C. Gieben.
Dik, Helma. 2003. On unemphatic “emphatic” pronouns in Greek: Nominative pronouns in
Plato and Sophocles. Mnemosyne 56(5): 535–550. Leiden: Brill.
Dik, Simon Cornelis. 1978. Functional Grammar. Amsterdam: North-Holland.
Dik, Simon  Cornelis. 1989. The Theory of Functional Grammar. Part 1: The Structure of the
Clause. Dordrecht: Foris Publication.
Ducrot, Oswald. 1984. Le Dire et le Dit. Paris: Éditions de Minuit.
Gildersleeve, Basil Lanneau & Miller, Charles William Emil. 1911. Syntax of Classical Greek from
Homer to Demosthenes. New York NY: American Book Company.
Givón, T. 1984. Syntax. A Functional-Typological Introduction, 1. Amsterdam: John Benjamins.
Halliday, Michael A.K. & Hasan, Ruqaya. 1976. Cohesion in English. London: Longman.
Ildefonse, Frédérique. 1997. La naissance de la grammaire dans l’Antiquité grecque. Paris: Vrin.
Kühner, Raphael & Gerth, Bernhard. 1898–1904. Ausführliche Grammatik der griechischen
Sprache. Satzlehre. Hanover: Hahnsche Buchhandlung.
Lambrecht, Knud. 1994. Information Structure and Sentence Form. Topic, Focus and the Mental
Representations of Discourse Referents. Cambridge: CUP.
Langacker, Ronald. 1987. Foundations of Cognitive Grammar, Vol. 1: Theoretical Prerequisites.
Stanford CA: Stanford University Press.
Lyons, John. 1968. Introduction to Theoretical Linguistics. Cambridge: CUP.
Moeschler, Jacques & Reboul, Anne. 1994. Dictionnaire encyclopédique de pragmatique. Paris:
Éditions du Seuil.
Nølke, Henning. 2001. Le regard du locuteur 2. Pour une linguistique des traces énonciatives.
Paris: Kimé.
Perdicoyanni-Paléologou, Hélène. 2002. The endophoric Use of ὁ, ὅδε, ἐκεῖνος, αὐτός in Ori-
gen’s Homilies on Jeremy, as translated by St Jerome (is, hic, iste, ille). Belgisch Tijdschrift
voor Filologie en Geschiedenis 80: 35–50.
Pitavy, Jean-Christophe.  2006a. Cohérence discursive, ellipse: Du nouveau en questions
et réponses? In Cohérence et discours, Frédéric  Calas (ed.), 251–262. Paris: Presses de
l’Université Paris-Sorbonne.
Pitavy, Jean-Christophe.  2006b. Au sujet du pronom sujet: ἐγώ et la liberté du locuteur en
grec. In Ἐν κοινωνίᾳ πᾶσα φιλία. Mélanges offerts à Bernard Jacquinod, Jean-Luc Breuil,
­Christophe Cusset, Florence Garambois, Nicoletta Palmieri & Éric Perrin-Saminadayar
(eds), 227–242. Saint-Étienne: Presses universitaires.
Proeme, Henk. 1994. On word stress in Polish prepositional phrases with a monosyllabic pro-
noun or noun. In  Dutch Contributions to the Eleventh International Congress of Slavists,
Bratislava, August 30-Septembre 9, Adriaan Barentsen, B.M. Groen, Rob Sprenger (eds),
Amsterdam: Rodopi.
Riegel, Martin, Pellat Jean-Christophe & Rioul René. 1994. Grammaire méthodique du français,
Paris: Presses universitaires de France.
Schiering, René. 2006. Cliticisation and the Evolution of Morphology: A Cross-linguistic Study
on Phonology in Grammaticalisation. Ph.D. dissertation, University of Konstanz.
Schwyzer, Eduard & Debrunner, Albert. 1939–71. Griechische Grammatik, Munich: C.H. Beck.
Wackernagel, Jacob. 1892. Über ein Gesetz der indogermanischen Wortstellung. Indoger-
manische Forschungen I: 333–436.
 Jean-Christophe Pitavy

Włodarczyk, André & Włodarczyk Hélène. 2006. Focus in the meta-informative centering the-
ory. In La focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André. 2004. Centre d’intérêt et ordres communicatifs. In Énoncer, l’ordre infor-
matif dans les langues, Pierre Cotte, Martine Dalmas & Hélène Włodarczyk (eds), 13–32.
Paris: L’Harmattan, Coll. Sémantiques.
Zwicky, Arnold M. 1995. What is a clitic? In Clitics: A Comprehensive Bibliography 1892–1991
[Library and Information Sources in Linguistics 22], Joel A. Nevis, Brian D. Joseph,
Dieter Wanner & Arnold M. Zwicky (eds). Amsterdam: John Benjamins.
 Personal subject pronouns
and the meta-informative centering
of utterances in classical Latin

Perrine Vedrenne-Cloquet
Université Paris-Ouest

Personal subject pronouns are used as meta-informative markers in

classical Latin. Given that they are optional when the utterance contains a
personal-ending verb, we put forward the hypothesis that they inherently
emphasise the person already involved by the personal ending of the verb.
Careful observation of examples taken from Cicero’s Correspondence shows that
the pronouns ego and tu featured in discourse have a highly important role in the
construction of informative coherence, since they can be used as topicalisations,
but also as focalisations. Moreover, they can be used as additional markers: their
second position gives to the first position element an informative emphasis which
it would not otherwise have.

1.  Introduction

This article deals with the role played by the personal pronouns ego and tu in the
meta-informative organisation of utterances. In Latin, a verb with a personal end-
ing is sufficient for the grammaticality of an utterance.1 The first and second per-
son subject pronouns with the same reference as an accompanying personal verb
form thus constitute a redundant expression of the subject. Their presence in this
sort of context, where there is no ambiguity over identifying the subject, is usually
analysed in normative grammar in terms of its expressive or stylistic contribution.
It seems to us that this question should be studied as a problem of informative
structure (Pinkster 1990), and more precisely, from the angle of the meta-infor-
mative centering theory (Włodarczyk 2003, Włodarczyk & Włodarczyk 2006a,
2006b, 2008) the terminology of which we will employ here.2

.  See parallel usages in Italian or Spanish.

.  See the glossary at the end of this volume.
 Perrine Vedrenne-Cloquet

In a verb form with a personal ending, such as credo, the verbal predicate
is assigned to a personal entity which becomes the global centre of attention
(CA), and thus takes on the role of “subject”. This term is to be understood in its
­Aristotelian sense (“what we speak about”), since it is hardly possible to treat a
personal ending as a grammatical subject.
When the personal pronouns ego and tu are associated with a verb that has
an ending corresponding to the person in question, we may hypothesise that this
double reference to the person takes on a meta-informative value.3 Or rather, if in
an utterance with no pronoun the personal entity already has a ­meta-informative
value, then in an utterance containing ego or tu, this pronoun is the result of
a supplementary informative prominence (a “meta-meta-informative” value).
According to the MIC theory, two levels must therefore be distinguished:

–– Level one (credo): pronoun Ø + verb with a personal ending: subject + predi-
cate corresponding to a base utterance not divided into old and new,
–– Level two (ego credo): full pronoun + verb with a personal ending: extended
utterance in which the pronoun is either topicalised (old meta-informative
status) or focalised (new meta-informative status) and contrasts with the rest
of the utterance which has the opposite meta-informative status.

We will see that it is in fact impossible to attach just one meta-informative role to
the pronouns ego and tu: depending upon the context, they may appear as Topic
or Focus.
The Topic value of ego and tu is quite legitimate: it emphasises information
about the actors in the situation spoken about in the utterance. These pronouns
point to the participants of the speech act; the fact that they have an old meta-
informative status is therefore not surprising in theory. Moreover, in an extended
utterance, as an amplification of the subject of the base utterance, they fulfil as
Topic the same global CA role as the subject. We will thus see, first and foremost,
to what extent this Topic value is relevant to the construction of the informative
cohesion of the text. These pronouns may, however, also be used as Focus, whether

.  Thus, in parallel with meta-informative centering theory, Touratier gives these pronouns,
when they accompany a personal verb, the syntactic function of extraposition, a function
which acts as a bridge between the syntactic and informative levels, since according to him
it makes “du contenu du syntagme extraposé à gauche, le thème ou le support informatif du
reste de l’énoncé” (the content of the phrase moved to the left, into the theme or informative
basis for the rest of the utterance) (1994: 35). Other syntactical presentations take into account
the informative organisation of the sentence and more or less correspond to extraposition as
defined by Touratier: cf. extraposition to the left (topical extraposition ), or Top and Foc posi-
tions on the left periphery of the sentence (Rizzi 1997).
 Personal subject pronouns and the meta-informative centering of utterances in classical Latin 

the status of new information of this CA emerges from the text structure or is the
result of encyclopaedic knowledge.
These two values of Topic and Focus will be illustrated by a detailed analysis
of various occurrences which may seem fastidious to non-latinists, but is justified
by the importance of context for understanding the specific role of these pronouns
in informative coherence. Finally, the meta-informative role of ego and tu explains
certain usages in context, where these pronouns, rather like particle-connectors,
act as markers of the highlighting of other elements of the utterance which are
treated as topics. We will give examples of this usage, after listing the other means
available to Latin for marking informative salience.

2.  Topicalisation

The topicalising value of the personal pronouns ego and tu can be seen clearly in
a great number of usages. Detailed analysis, however, shows that ascertaining this
value is not in itself enough to explain fully the informative role of the pronouns
in question. In view of their particular enunciative dimension, the enunciative
structure must be integrated with the informative structure. This can be observed
in the following example, in which thematic progression relies on the topicalisa-
tions of Pansam and ego:

(1) 
Gratulor nobis Q. filium exisse: molestus non erit. Pansam bene loqui
credo; semper enim coniunctum esse cum Hirtio scio; amicissimum Bruto
et Cassio puto, si expediet – sed quando illos uidebit ? – Inimicum Antonio
quando aut cur ? Quousque ludemur ? Ego autem scripsi Sextum aduentare,
non quo iam adesset, sed quia certe id ageret ab armisque nullus discederet.
[Att. 15,22 ]
“I congratulate ourselves that young Quintus has gone out of town: he
won’t be a nuisance to us. I believe Pansa is using satisfactory language. For
I know that he has always been closely united with Hirtius. I think he will
be a very warm friend to Brutus and Cassius if it turns out to be expedient.
But when will he ever see them? And that he will be opposed to Antony –
but when and on what grounds? How long are we to be fooled? However,
I wrote you word that Sextus Pompeius was coming, not because he was
actually near, but because he was certainly contemplating that move and
because he was not showing any signs of abandoning arms.”4

.  The English translation of Latin examples are taken from http://perseus.uchicago.edu/ 
 Perrine Vedrenne-Cloquet

This first part of a letter addressed to Atticus, in which Cicero writes about young
Quintus, his nephew, then Pansa, sets out an initial predication (Pred1: gratulor
nobis; credo; scio; puto) and a second predication (Pred2: Quintum filium exisse;
Pansam bene loqui; coniunctum esse cum Hirtio; amicissimum Bruto et Cassio). In
the first utterance (gratulor nobis Q. filium exisse), the verb gratulor is placed first –
in the opening line the presence of the speaker is stated from the outset, in a sub-
jective affirmation.5 But subsequently the second predication is in the word-initial
position: here it is word order which plays its part in the topicalisation of Pansam
(“I believe Pansa is using satisfactory language”), and underlines the thematic pro-
gression. The verbs in Pred1 (credo, scio, puto) are placed at the end, after Pred2
(Pansam bene loqui; coniunctum esse cum Hirtio; amicissimum Bruto et Cassio),
benefiting informative cohesion.
Then after several interrogative propositions, the order is inverted: the utter-
ance ego autem scripsi Sextum aduentare follows the same syntactical model as
previous utterances, in which the first person verb governs an infinitive propo-
sition, but the word order and the relative importance are different: Pred1 (ego
autem scripsi) is at the head, and this time topicalisation concerns the first person
subject, in the form of the pronoun ego, the connecting word autem underlining
the change of Topic6 (“However, I wrote you word…”). The form ego scripsi is in
effect the opposite of the forms credo, scio, puto which have no personal subject
pronoun.
The distinction between the first person verb forms (credo, scio, puto) and
the expression of the first person reinforced by the pronoun ego in ego scripsi
therefore appears to be motivated by the choice of a different topicalisation: that
of the topic of Pred2 (Pansam) in the first case, and of the topic of Pred1 (ego) in
the second case.
But the parallel initially established between these forms is illusory: scripsi, in
the perfect tense refers to another utterance in the first person, temporally distinct
from the first person of credo, scio and puto. Furthermore, credo, scio and puto are
meta-informative markers: unlike scripsi, they belong to the plane of the speech
act. The first person occurrences are thus placed on two different speech levels: the

.  The letters of Cicero begin for the most part with an utterance in the first person: most
often, an indication of time and/or place is followed by the comment accepi litteras or perueni,
etc. Moreover we find in volume IX of Correspondence  at the very beginning: “Spero tibi”
(Att.  14,5,1), “O mi Attice, uereor” (Att.  14,12,1), “Quam uellem” (Att.  15,4a), “Gratulor”
(Att. 15,22), “Mirifice torqueor” (Att. 15,23,1), and “Gaudeo” (Att. 5,27,1), and with no subjec-
tive marking, “Narro tibi” placed in parataxis (Att. 15,16a,1, Att. 15,21,1).
.  Cf. pragmatic analyses of autem, and in particular the study by Kroon 1995.
 Personal subject pronouns and the meta-informative centering of utterances in classical Latin 

first person of credo concerns the speech act, whereas the first person of ego scripsi
refers to the situation spoken about in the utterance. Here, ego does not so much
refer to the speaker as to the person who in an earlier letter wrote “Sextus quem
advent are aiunt” (this is in fact the letter Att. 15,21,1). Because of the influence
of the pronoun ego, the first person is topicalised in the same register as Pansam:
“I say [with modalization] about Pansa that…” and “I say [without modalization]
about myself that in the past…” The topicalisation of the first person by the pro-
noun ego here gives prominence to what, in this first person subject, does not
stem from the speech act plane. It is important therefore to adjust both the syn-
tactical and the pragmatic organisation so as to explain as fully as possible the
­meta-informative organisation of the message.
The first person can thus be the object of topicalisation in order to distinguish
between different speech participant instances in an utterance. Through this use of
ego as contrasting topic, the speaker is contrasted with Pansa, by no longer simply
taking him as the mark of the person who says something, but as the mark of the
person about whom something is said, with referential identity between the two.
Topic value thus plays its part in discourse structure by enabling a distinction to
be made in the use of a first person form, between what belongs to the level of the
utterance and what concerns the level of the speech act.

3.  Focalisation

Studying the informative function of personal pronouns undoubtedly entails

paying attention to the speech act and its participants. But the presence of the
personal pronoun can also be justified by textual cohesion: considering the overt
expression of the speaker or of the person spoken to by a personal pronoun only
as topicalisation is not enough for understanding the communicative intention at
work in certain usages.
It seems that personal pronouns can in fact be used as either topicalisation or
focalisation. An example of this focalizing value can be seen in another of Cicero’s
letters, again addressed to his friend Atticus, with the pronoun tu.

(2) De meo itinere uariae sententiae; multi enim ad me. Sed tu incumbe,
quaeso, in eam curam; magna res est. [Att. 15,25]
“As to my journey various opinions are expressed: for I have a great number
of visitors. But pray throw yourself heart and soul into that question. It is a
serious matter.”

In line with what has been noted above, we postulate that the 2nd person is
emphasised by the personal pronoun at the start: as for you (in bold), I give an
 Perrine Vedrenne-Cloquet

order, incumbe in eam curam – in front of an utterance with no personal pronoun,

incumbe, quaeso, in eam curam, where the 2nd person is not the object neither of
topicalisation nor focalisation.
Thus it is important to take into account the coherence of the text, in so far as
the content of the order is not new information: eam is a repetition, referring back
to de meo itinere, and the process incumbere in eam curam is precisely that which
is attributed to multi in their uariae sententiae. What is new is the command form
and the person to whom the command is given, tu. Here the pronoun is thus a
focalisation of the 2nd person: the focus value presents it as new information as
opposed to the multi of the preceding utterance. In this particular example it can
be difficult to distinguish between a focus tu and a contrasting topic tu (this would
correspond to the English expression “as for you” or “for your part”). But it seems
to us that the choice of the discourse particle favours a focalising interpretation.
We can see in fact, as is shown by comparison with the example studied earlier,
that there is a difference in the use of the connectors sed and autem: autem, coming
in second place, emphasises a topic, whereas sed, in first place, prepares the way for
a focus element. The meta-informative function of connectors is, for that matter,
worthy of a specific in-depth study.
We can nonetheless find utterances in which the focus value of the personal
pronoun is unequivocal. This is the case, outside the corpus, in this very short
epigram by the poet Martial:

(3) Auriculam Mario grauiter miraris olere. Tu facis hoc: garris, Nestor, in
auriculam. [Mart., Ep. III, 28]
“You’re amazed that Marius’ ear smells nasty. It’s you that make it so: you
chatter, Nestor, into his ear.”

In the utterance Tu facis hoc, it is difficult to deny the New meta-informative status
of the pronoun tu as focalised CA, when the remainder of the utterance, as “back-
ground”, is made up of the anaphoric pronoun hoc, with no prominence at all, and
of the verb facis, which is in itself also anaphoric.
Here is another example, outside the corpus, taken from one of Ovid’s Heroi-
des, which shows that the distribution of meta-informative new/old information
can also depend upon encyclopaedic knowledge:

(4) Non ego cum Danais arma cruenta fero; [Ov., Her. V, v.156]
“I do not bring a bloody war with the Greeks.”

In this fictional epistle, Oenone is addressing Paris and setting herself more or less
implicitly in contrast to Helen, the woman for whom he has left her. Here, ego is
clearly focalised: from the first polemical negative non, the various participants –
at different levels of coding and de-coding, the author, the fictional speaker, the
 Personal subject pronouns and the meta-informative centering of utterances in classical Latin 

fictional listener, the readers – pick up the meaning X cum Danais arma cruenta
fert (“X does bring a bloody war with the Greeks”), which is given as old informa-
tion in so far as it refers this time not to an anaphora within the text, but to the
knowledge held by the participants outside the text itself.
If the pronouns ego and tu are used to put meta-informative emphasis on the
subject when they are associated with a verb that has a personal ending, the study
of their place in discourse forces us to distinguish a number of nuances: we can
thus recognise a topicalising and a focalising value, on condition that we take into
consideration the overall coherence of the discourse. As has already been said,
we must not, however, loose sight of the specific pragmatic (meta-informative)
usage of these pronouns if we want to give as accurate an account as possible of
their informative use. But the part played by what are called personal “subject”
pronouns in the informative coherence of a text is more complex than it appears,
because the forms ego and tu can be used to modify the meta-informative status
of an element by detaching it from the rest of the utterance (Amacker 2001, Griffe
2007). Only the fact that these forms, associated with an accompanying personal
verb form, are necessarily felt to carry a meta-informative emphasis, explains that
they can, in second place in the utterance, underline the first-placed element, and
thus give it an informative prominence that it would not ordinarily have.

4.  Ego and tu as additional meta-informative markers

In order to shed light on the defining role that ego and tu can play in an utterance,
it is useful to review the means available in Latin for emphasising prominence in
the absence of markers of prosody in the written form of the language.
Generally speaking, prominence is bestowed by dislocation variously alter-
ing the order of words. Let us look at the three examples of increasing syntactical
change proposed by Colette Bodelot:7

–– Word-initial position (left dislocation), outside the syntactic structure.

(5) Cic. Att. 14,10,2: Nausea iamne plane abiit?

“Is your sickness quite gone by this time?”

Nausea is placed before the adverb jam, hosting the enclitic interrogative particle -ne.

.  Bodelot 2007: 20. Cf. also Bortolussi 2007: 33–34.

 Perrine Vedrenne-Cloquet

–– Word-initial position (left dislocation) with anaphora, and agreement between

the element that has been syntactically moved and its anaphoric indicator:
(6) Cato Agr. 5,3: Amicos domini, eos habeat sibi amicos.
“The master’s friends, may he consider them his own.”

Amicos domini is detached and repeated by eos.

–– Absence of grammatical agreement: the element that has been moved to the
left is in the nominatiuus pendens, and its anaphoric indicator is in the case
demanded by the syntactical structure of the clause.
(7) Plaut. Men. Arg. 1–2: Mercator Siculus, quoi erant gemini filii,/Ei surrupto
altero mors optigit.
“A Sicilian merchant, who had twin sons, had one of them taken away and
died shortly afterwards.”

Mercator Siculus is in the nominatiuus pendens, and the indicator, carrying the
syntactical function, is in the dative: ei.
Topicalisation can therefore be signalled by different kinds of extraction:
word-initial position, word-initial position with anaphora and grammatical agree-
ment, and word-initial position with anaphora but no grammatical agreement.
Furthermore, connectors such as autem or quidem can also serve as markers. But
a study of certain collocations tends to show that the pronouns ego and tu are in
their turn able to signal the topicalisation of other elements: always supposing that
they are themselves sufficiently clearly prominent to become as a consequence
discriminating indicators of changes in word order.
We can thus see constructions which have an anaphoric pronoun in first place,
marked morphologically by a particular case (usually the accusative or the dative),
and a 1st or 2nd person personal pronoun in second place: this morpheme then
brings about the detachment of the element that precedes it from the rest of the
utterance. Without that detachment and in the absence of any intonation criterion,
the prominence of the anaphora would not be evident. In the same way, in the
examples taken from C. Bodelot, the particule -ne, normally attached to the first
word, is attached to the next and thus detached from Nausea.
(8) 
sed tamen idem σύνταγμα misi ad te retractatius, et quidem αρχέτυπον
­ipsum crebris locis inculcatum et refectum. Hunc tu tralatum in
­macrocollum lege arcano conuiuis tuis.8 [Att. 16,3,1]

.  Cf. also Huic ego locum in proximo conduxi et, ut possum, ex meis angustiis illius sustento
tenuitatem. [Fam. 16,21,4, letter from young M. Tullius to Tiro]; Huic ego litteras ipsius arbi-
tratu dabo; eae te ne mouerint. [Att. 16,1,6]; Eas ego oportet perspiciam, corrigam; [Att. 16,5,5].
 Personal subject pronouns and the meta-informative centering of utterances in classical Latin 

“but nevertheless I am sending you the same essay somewhat more

­carefully revised – and it is indeed the original copy itself with
­interlineations and corrections in many places. Get this copied on large
paper and read it privately to your guests.”
Literally: “This, read it copied on large paper privately to your guests.”

The anaphoric pronoun Hunc takes up αρχέτυπον (“the original text”), tralatum
is the attributive of hunc, the complement of the object of the verb lege. The per-
sonal pronoun, in the same person as the governing verb, separates the comple-
ment of this verb and its attributive. In other words, an element belonging to the
first predication (tu lege) separates hunc from the second predication element tra-
latum.9 The intercalation of the personal pronoun between hunc and tralatum in
macrocollum signals the topicalisation of hunc by detaching it from the rest of the
second predication: without this intercalation, the string hunc tralatum in macro-
collum is unremarkable and nothing in the written form marks any dislocation.
The sentence in question could be translated as follows, if one wanted to retain
the informative prominence: “This, (you) read it copied on large paper privately
to your guests.”
The same kind of collocation appears in more complex utterances, such as:
(9) quodque ex istis fructuosis rebus receptum est, id ego ad illud fanum
­sepositum putabam.10 [Att.15,15,3]
“while the receipts from those paying properties I thought had been put
aside for building the shrine.”
Literally: “and what has been received from those paying properties, that,
I thought had been put aside for building the shrine.”

Syntactically, utterance (9) is a correlation: it begins with a relative clause that is

then taken up by an anaphoric id followed by a personal pronoun. We therefore
find at the beginning of the utterance an occurence, as previously quoted, of the
dislocation of the topicalising value: the topicalised element is in word-initial
position and taken up by an anaphora. In fact this kind of dislocation does not

.  Attributive and predication since we have here a case of the attributive of the object where
“la caractérisation de l’objet N1 par son attribut est toujours le propos de l’énoncé.” “the char-
acterisation of the object N1 by its attributive is always the purpose of the utterance.” (Riegel
et al. 1994: 241)
.  Cf. also Quod sine ulla dubitatione apud consules obtinuimus propter summam boni-
tatem et aequitatem causae, id tu nos obtinuisse non modo facile patiare sed etiam gaudeas.
[Att. 16,16b,2] “what we have obtained from the consuls without any hesitation, owing to their
great kindness and the justice of our cause, you will not only acquiesce, but even rejoice, in
our having secured.”
 Perrine Vedrenne-Cloquet

apply exclusively to NPs, but also to clauses. However, after this first topicalisation
(of the relative clause),11 the meta-informative status of the anaphoric pronoun
id is in turn modified by ego: its position between id and the rest of the utter-
ance disturbs the normal word order id ad illum fanum sepositum putabam. The
intercalation of the personal pronoun between id and ad illum fanum sepositum
detaches id and signals its topicalisation: “what has been received, that, I thought
had been put aside.” Thus id takes up the topicalisation and is itself the object of
a second topicalisation: it is an informative marker which is in its turn pushed to
the front.
This analysis is furthermore corroborated by C. Touratier’s analysis of the
phenomenon of correlation, at a more syntactical level and using different
terminology:

Lesdites constructions sont des cas de subordination qui présentent deux

particularités notables: d’une part la proposition subordonnée est extraposée par
rapport au reste de l’énoncé, et d’autre part cette subordonnée est reprise par un
constituant anaphorique (le corrélatif) qui est lui-même extraposé par rapport au
nouveau reste de l’énoncé. (…) Au point de vue du sens en effet, la subordonnée,
en tant que constituant extraposé, est le support informatif de l’énoncé, et le
corrélatif, en tant que constituant extraposé de deuxième niveau, est aussi un
support informatif, mais le support informatif second de l’apport informatif de
l’énoncé.(Touratier 1994: 697–698)
“Such constructions are cases of subordination which present two notable
particularities: on the one hand the subordinate clause is extraposed from the
rest of the utterance, and on the other it is taken up by an anaphoric constituent
(the correlative) which is itself extraposed from the new rest of the utterance. (…)
In fact from the point of view of meaning, the subordinate clause, as an extraposed
constituent, is the informative basis of the utterance, and the correlative, as a
second level, extraposed constituent, is also an informative contributor, but is
second to the informative contribution of the utterance.”

We must note, however, that outside this general context, one could view id as
focalised. Indeed, taken separately, the utterance could equally well be translated as
“what has been received, it is that which I though had been put aside.” In this case,
id, taken from the preceding topicalisation, would then undergo focalisation and
would become the new meta-informative contributor of the rest of the u ­ tterance.
It would then serve as a pivot, a junction point between the ­topicalisation and focal-
isation of the same content, but perceived informatively in two different ways.

.  Bortolussi (2007: 37) has established that in Latin there can only be “thematization” of
the element in word-intial position when it is repeated, whereas word-initial position without
repetition can also be focalising.
 Personal subject pronouns and the meta-informative centering of utterances in classical Latin 

However, the wider context in no way justifies focalisation, which seems to

confirm this translation of the passage as a whole:
(10) Profectionem meam, ut video, Erotis dispensatio impedit. Nam cum ex
­reliquis, quae nonis april. fecit, vel abundare debeam; cogor mutuari;
quodque ex istis fructuosis rebus receptum est, id ego ad illum fanum
­sepositum putabam. Sed haec Tironi madavi, quem ob eam causam
­Romam misi. Te nolui impeditum impedire.”
“My leaving Italy is hindered by Eros’s way of doing business. For whereas
from the balances struck by him on the 5th of April I ought to be well off;
I am obliged to borrow, while the receipts from those paying properties
I think have been put aside for building the shrine. But I have charged Tiro
to see to all this, whom I am sending to Rome for the express purpose.
I did not wish to add to your existing embarrassments.”

There thus seems to be nothing here, other than the context, which allows us
to distinguish between the topicalisation and the focalisation of the anaphora.
More generally speaking, without the help of the intonation and interpretation
provided by the wider context, it is often difficult to distinguish clearly between
the meta-informative status of the pronouns ego and tu: topic, focus or additional
meta-informative marker.

5.  C
 onclusion

Personal pronouns do not stipulate what kind of informative prominence they

provide. Functioning as landmarks in syntagmatic order, they can, in second
place, give prominence to the element which comes before them. That at least is
what we have been able to observe from the “anaphora + ego/tu” constructions.
Latin “subject” pronouns thus play a specific role in the structuring of information
in discourse. Over and above any expressive or emphatic effects, they contribute
actively to the pragmatic structuring of the text or discourse. Rather than being
based on a simple morphological opposition (full/zero pronoun, or tonic/atonic
form), at the heart of the distinction between a base utterance and an extended
utterance, the meta-informative status of these pronouns is based on criteria of
position, the pragmatic value of which often tends to be neutralised by literary
translations.
 Translated from French by Rosemary Masters

Corpus

Cicéron, Correspondance, t. IX (de septembre 45 au 6 août 44 a. C.), texte établi, traduit et annoté
par Jean Beaujeu 1988. Paris: Les Belles Lettres.
 Perrine Vedrenne-Cloquet

References

Amacker, René. 2001. Indices de thématisation segmentée non initiale chez Cicéron. In De
lingua latina nouae quaestiones: Actes du Xe Colloque international de linguistique latine:
Paris-Sèvres, 19–23 avril 1999, Claude Moussy, Michèle Fruyt, Jacqueline Dangel, Lyliane
Sznajder & Léon Nadjo (eds), 185–200. Louvain: Peeters.
Bodelot, Colette. 2007. Syntaxe liée vs. syntaxe non liée ou de l’utilité de distinguer une macro-
syntaxe de la (micro-)syntaxe en latin. In Eléments ‘asyntaxiques’ ou hors structure dans
l’énoncé latin, Actes du colloque international de Clermont-Ferrand, 16–17 septembre 2005,
Colette Bodelot (ed.), 15–33. Clermont-Ferrand: Presses Universitaires Blaise Pascal.
Bortolussi, Bernard. 2007. Ordre des mots et thématisation en latin. LINX 55: 33–47.
Griffe, Michel. 2007. L’extraposition en latin. In Eléments ‘asyntaxiques’ ou hors structure dans
l’énoncé latin, Actes du colloque international de Clermont-Ferrand, 16–17 septembre 2005,
Colette Bodelot (ed.), 37–54. Clermont-Ferrand: Presses Universitaires Blaise Pascal.
Kroon, Caroline. 1995. Discourse particles in Latin. A study of nam, enim, autem, vero and at.
In Amsterdam Studies in Classical Philology, 4. Amsterdam: J.C. Gieben.
Pinkster, Harm. 1990. Ego, tu, nos. Opmerkingen over het gebruik van subjektpronomina in het
bejzonder in Cicero De Oratore II. Lampas 19: 309–322.
Riegel, Martin, Pellat, Jean-Christophe & Rioul, René. 1994. Grammaire méthodique du français.
Paris: Presses Universitaires de France.
Rizzi, Luigi. 1997. The fine structure of the left-periphery. In Elements of Grammar: Handbook of
Generative Syntax, Liliane Haegeman (ed.), 281–337. Dordrecht: Kluwer.
Touratier, Christian. 1994. Syntaxe Latine. Louvain: Peeters.
Włodarczyk, André. 2003. Les homotopies du topique et du focus. In Ordre et distinction dans
la langue et le discours. Actes du colloque international de Metz (18,19 et 20 mars 1999),
­Bernard Combettes, Catherine Schnedecker & Anne Theissen (eds), 513–526. Paris:
­Honoré Champion éditeur.
Włodarczyk, André & Włodarczyk, Hélène. 2006a. Focus in the meta-informative centering
theory. In La Focalisation dans les langues, André Włodarczyk & Hélène Włodarczyk (eds),
27–46. Paris: L’Harmattan, Coll. Sémantiques.
Włodarczyk, André & Włodarczyk, Hélène. 2006b. Subject in the meta-informative centering
theory. Études cognitives/Studia kognitywne 7: 39–64. Warszawa: SOW, PAN.
Włodarczyk, André & Włodarczyk, Hélène. 2008. The pragmatic validation of utterances.
Études Cognitives/Studia Kognitywne 8: 117–128. Warszawa: SOW, PAN.
 Glossary of defined terminology

Terms with an asterisk (*) in the explanatory notes are also glossary entries
Aboutness Two kinds of aboutness: (a) the spoken-aboutness (resemblance) consists of
(Speak- creating *Centres of Attention (CA) and (b) the speak-aboutness (contiguity)
Aboutness) consists of combining the created *Centres of Attention (CA) with everything
(Spoken- which is predicated about them.
Aboutness)
Agent An animate entity seen fit to play active roles in semantic situations as
expressed by utterances. In the Associative Semantics (AS) theory, it is possible
for an animate entity to play more than one active role. Moreover, agents can
also play non active roles (passive or median).
Anchor Spatio-temporal component (or operation) of semantic situations. One of
their three fundamental constitutive components (the two others being frames
and roles).
Associative The Associative Semantics (AS) theory is an ontology-based semantic theory
Semantics of information as expressed by natural language utterances with clear-cut
(AS) distinctions between (a) the universal (or at least common) ontology and
(b) the semantics of natural languages. In this theory, roles and anchors are
seen as elementary (unary) relations. Roles are defined ontologically as active
and passive unary relations of associated semantic situations. The combination
of their realizations gives rise to median roles such as “instruments” or
“means” among others. The main principle of AS is that one participant may
enact more than one role.
Attention- That phrase of a *Base or *Extended Utterance which can be assigned a
driven Phrase meta-informative *Old or *New Status. In order to communicate judgments,
(ADP) speakers need to select some information stored in their memory (thus
creating Centres of Attention (CA). While the Centre of Attention (CA) is
determined functionally as a mental concept which results in the selection
of a chunk of information, the Attention-driven Phrase (ADP) is defined
as that part of an *utterance (linguistic expression) which bears the selected
information with an explicit meta-informative pointer to it. The segment
of an expression is considered to be centered (corresponding to a Centre
of Attention) if it represents an entity, anchor or situation which has been
selected (see *Selection) among other entities, anchors or situations. An ADP
often contains (syntactic, morphological, prosodic) meta-informative markers.
If an utterance expresses more than one Centre of Attention (CA), one of them
is global (main) and the other one is local (secondary).
Example: The utterance “John and Mary dance” contains a sole (though
collective) ADP although the semantic information contains two participants
“John” and “Mary” both selected as Centres of Attention. Indeed, semantically
we have two associated (joint concurrent) situations: [John dances] & [Mary
dances] developing the schema with a dual existential quantifier: Exist2(x)
dance(x). Note that the utterance “John dances with Mary” contains only one
entity (“John”) which has been centered. See: *Global (or main) Attention-driven
Phrase (G-ADP) and *Local (or secondary) Attention-driven Phrase (L-ADP).
 Glossary of defined terminology

Background The background is the speak-about part of the *Extended Utterance the
spoken-about part of which is known as the *Focus. As a rule, the *Status of the
spoken-about part of an *Extended Utterance is opposite to its speak-about. In
the case of the Focus-Background relationship, the speak-about has an *Old
meta-informative *Status. Antonym of *Comment.
Centre of Attention-centered information is that piece of knowledge to which humans
Attention are attuned by periods of time. In neurology humans are able to pay attention
(CA) to more than one concept at the same time provided that the selected *Centres
of Attention concern concepts which belong to different levels (regions
of memory). In linguistic predication two *Centres of Attention (CA) are
expressed by two different *Attention-driven Phrases (ADP) belonging to
two different levels of the constituency tree (structure of meta-information):
*Subject and *Object.
Combination Higher level operation on two selected chunks of information between which a
relationship can be established.
Comment The comment is the speak-about part of the *Extended Utterance the
spoken-about part of which is known as the *Topic. As a rule, the *Status of the
spoken-about part of an *Extended Utterance is opposite to its speak-about. In
the case of the Topic-Comment relationship, the speak-about has a *New meta-
informative *Status. Antonym of *Background.
Default Role The Default Role Mapping principle establishes the default relationships
Mapping between *Centre of Attention (CA) and *Agent (or *Figure) as expressed by an
*Utterance.
In Accusative languages, there is a default relation linking the *Subject with the
*Active Role (enacted by an *Agent or *Figure) while in Ergative languages, the
default role mapping links the *Subject with the Passive Role enacted by an
*Agent or *Figure).
Figure Inanimate entity. This kind of entity can enact both active and passive roles.
Active roles played by *Figures are said to be *Quasi-active (*Q-Role).
Focus The Local *Attention-driven Phrase (the spoken-about part) of an *Extended
utterance, the meta-informative Status of which is *New, whilst its speak-about
part (*Background) has an opposite (contrary) Status (*Old). Antonym of
*Topic.
Frame Representation of the space of states and actions. There are four kinds of
(Situation frames: 1 static and 3 dynamic frames: states are semantic situation frames
Frame) delimited by [+Space] only, events are semantic situation frames delimited
by [+Space] and determined by [+Time] (N.B.: events lack a middle stage.
For example: “to cough” as in “John just coughed”). Ordinary processes are
semantic situation frames delimited by [+Space] and determined by [+Time]
and [+Progression] and refined processes are semantic situation frames
delimited by [+Space] and determined by [+Time], [+Progression] and
[+Granularity].
Global The main *Attention-driven phrase (ADP) of a *Base Utterance which
Attention- expresses the hierarchically uppermost *Centre of Attention (CA) is defined
driven Phrase as a Global ADP of that utterance (*Subject). The ADP of an *Extended
(Global ADP) Utterance which expresses the hierarchically uppermost *Centre of Attention
(CA) is defined as a Global ADP of that utterance (*Topic). N.B.: The *Centres
of Attention may coincide in the expression plane, i.e. they may correspond to
the same ADP.
 Glossary of defined terminology 

Information In the MIC theory, information is defined as a (spatio-temporally) situated

semantic relationship between agents and/or figures. Kernel information
contains only the situation frame, its roles (participants) and anchors
(spatio-temporal locations). It can be assigned a propositional value
(True/False).
Level of The three following levels of interpretation are distinguished: shallow, standard
interpretation and deep. When the *Subject of a default (active, for Accusative languages, or
passive, for Ergative languages) diathesis sentence does not correspond to the
default (active or passive respectively) semantic role of a given situation, the
shallow level of interpretation is used.
Local The secondary (dependent) *Attention-driven phrase (ADP) of a *Base
Attention- Utterance which expresses the hierarchically lower *Centre of Attention
driven Phrase (CA) is defined as a Local ADP of that utterance (*Object). The ADP of an
(Local ADP) *Extended Utterance which expresses the hierarchically lower *Centre of
Attention (CA) is defined as a Local ADP of that utterance (*Focus). N.B.:
The *Centres of Attention may coincide in the expression plane, i.e. they may
correspond to the same ADP.
Meta- Information about another information (not just abstraction). The semantic
information content of the utterance is information itself, and the different linguistic forms
that may be chosen to express information are meta-information markers.
Meta-information concerns the way information is presented: in order
to achieve the ordering of concepts in expressions (linguistic utterances),
speakers create *Centres of Attention (CA) and *predicate about them.
Meta- In the MIC theory no declarative semantic content can be used in
informative communication without its meta-informative status (i.e. without the pragmatic
status centering of information). The meta-informative *Status may take one of the
two values: *Old or *New. See also *Validation.
Motivation of The *Old or *New *Status of an utterance (either its whole or a part of it) is
Status motivated either by (a) communication (anaphoric/cataphoric), (b) cognition
(known/unknown) or (c) ontological knowledge (generic/specific, general/
particular, potential/actual and habitual/occasional).
New See *Meta-informative Status.
Object That *Attention-driven Phrase (ADP) of a *Base Utterance which expresses the
*Local (secondary) meta-informative *Centre of Attention (CA).
Old See *Meta-informative status.
Participant An entity playing a role in a situation. Semantic role filler.
Predicate That speak-about part of a *Base Utterance which says (predicates)
something about the Global *Centre of Attention as expressed by the *Subject
*Attention-driven Phrase (ADP). In Elementary Logic, predicate is defined
as a formula containing a relation name (literal) and one or more terms
(arguments). Hence in logic, predicate is nothing but form; it has no meaning.
Therefore, both run(Agent: “Peter”) and run(Subject: “Peter”) are well-formed
predicate formulae. Whereas in the MIC theory, predicate is not defined
formally. Instead, predicate is seen as the result of *Aboutness. It belongs to the
meta-informative (pragmatic) level of speech.
Predication Speech act aiming at pointing to the central part of information (See *Centre
of Attention) communicated by a *Base Utterance in order to speak/tell
something about it (see *Speak-aboutness and *Spoken-aboutness).
 Glossary of defined terminology

Q-Role Agents typically fit dynamic situations (actions) well, whereas figures fit the
Quasi-active static ones (states). When it is not so, shallow level (i.e. a partially specified
Role semantic level) is introduced. There is an important feature of situation
Q-active Role participants which characterizes the shallow level: in active roles, the
inanimate entities (figures as opposed to agents) are semantically interpreted as
if they were animate. Such figures are said to enact quasi-active roles (Q-active
roles or Q-roles). In the cases of agentivation or figuration, for example, the
Q-initiator will be said to designate the figure in an active role (which normally
fits agents) and Q-source will be said to designate the agent in an active role
(which normally fits figures). See also *Role.
Role One of the three fundamental constitutive components of semantic
*situations. Two elementary (active and passive) and one derived (median)
roles are distinguished and defined as unary relations of associated semantic
situations. N.B. Due to the definition of semantic situations as associations,
one participant (role filler) may play more than one role in the given semantic
situation. See also *Roles and Voice.
Roles and In Accusative languages, by default, in the active voice utterances the *Subject
Voice corresponds to the *Active or *Q-Active Role while in the passive voice
utterances, the *Subject corresponds to roles which are neither active nor
q-active.
Selection Lower level operation on that chunk of information which is chosen on the
lower constitutive level of pattern recognition or pattern production.
Situation Situations are defined regardless of their participants. Their three fundamental
components are: (1) Frame (their instances are relations), (2) Role (their
instances are Participants) and (3) Anchors (their instances are spatio-temporal
locations).
Status See: Meta-informative *Status.
Subject Main meta-informative “spoken-about” part of a *Base Utterance, i.e. the
Attention-driven phrase (ADP) pointing at the global Centre of Attention
(CA). The meta-informative *Status of the subject may be either *New or *Old
as is the meta-informative *Status of a whole *Base utterance.
Theme That part (or set of subsets) of a text/discourse which expresses the global or
local *Centre of Attention (Concern).
Topic The Global *Attention-driven Phrase (the spoken-about part) of an *Extended
utterance, the meta-informative Status of which is *Old while its speak-about
part (*Comment) has an opposite (contrary) Status (*New). Antonym of
*Focus.
Utterance An utterance is a proposition uttered in a context and validated as bearing
information with *Old or *New Status. When used in an utterance, the
objective (denotative) oldness or newness of information motivates (see
*Motivation) the meta-level segmentation of communicated information
giving rise to the creation of pragmatic (connotative) content. Consequently,
an utterance can be assigned either *Old or *New meta-informative *Status
regardless of its truth-propositional semantic valuations.
Utterance The *Base Utterance has either a *New or an *Old meta-informative Status. It
(Base is not divided into parts (segments) with respect to the *New or *Old *Status,
Utterance) i.e.: the *Status of a *Base utterance cannot but be either entirely *New or
entirely *Old.
 Glossary of defined terminology 

Utterance An utterance which is divided into two parts having contrary

(Extended meta-informative *Status (either *Old Spoken-about + *New Speak-About
Utterance) or *New Spoken-about + *Old Speak-About). There are two kinds of such
extended utterances: (a) with a Topic-Comment relationship – where one part
of the utterance has an *Old meta-informative *Status (called *Topic) and the
other has a *New meta-informative *Status (called *Comment) and (2) with a
Focus-Background relationship – where one part of the utterance has a *New
meta-informative *Status (called *Focus) and the other has an *Old
meta-informative *Status (called *Background).
Validation The content (semantic information) of an utterance can be validated as True or
False, its meta-informative *Status as *Old or *New.
 Index
A B
addressee  70, 88, 104–105,
167–169, 197, 268–270,
272, 281
adjacent utterances  50, 55, 61,
127, 212
agent  5, 6–8, 12–13, 15–19, 54,
68, 81, 125, 138, 213
active  81
Aktionsart  32, 195, 197–200,
203–204, 206–207, 210
see also modes of action
anaphora  50, 54–55, 74, 169,
176, 179, 217, 291, 292,
293, 295
anaphoric  48, 49, 50, 51, 61,
122, 126, 127, 212, 217, 218,
220–221, 280
animacy  171, 173–174
animate  7–8, 12, 52, 68, 171
agent  6
entity  5, 6, 7, 68
argument  4, 9, 43–44, 59, 77,
170–174, 176
places  44
argumentation  47, 70, 169
aspect  21–23, 26–29, 31,
193–214, 216–226
cluster  207–208
pair  207–208
aspect
external view of  202–206
internal view of  202–206
aspectual seme  23
association of semantic
situations  13–14, 17, 18
associative semantics  13, 17, 52
attention  53–55, 59–70, 60–63,
103–105, 111–116, 121–123,
125–128, 131–140, 217–218
centre of  43, 45–47, 65–69,
171, 183–185, 188, 193–194,
197, 205, 236, 244, 259–262
peripheral  183, 188–189
attention centering  59, 61
attentional attitude  43
belief  3, 8, 55
C 136–137, 184
canonical place  237–239, 243,
248–249, 253–254,
256–257
word order see word order
case  5–6, 68–69, 77, 81, 89,
91–94, 277, 292–293
grammar  8–9
marking languages  73
accusative  71, 73, 91
nominative  81, 89, 91
cataphor  147
cataphora  41, 50, 54–55, 104,
176, 217
centering  45–46, 55, 61–62, 64,
114, 134, 147–149
centre of attention
see attention
shared  183–185
centre
backward-looking  61, 90
forward-looking  61, 90
change  22, 24, 25, 35, 209
clause boundaries  66
cleaving  86
cleft  145–147, 149–163
clefting  145, 149, 151–153,
158–160, 162–163
cognitive  24–25, 48–50, 60,
62–63, 67, 107–109, 112,
114, 126–128, 132, 136
motivation  217–219
coherence  148, 167–169, 175,
179, 261, 263, 267–268,
290–291
discourse  47, 90–91, 179
lack of  112
text  72, 74, 77, 84, 91
coherent  61, 65, 78, 148,
169, 175
cohesion  73, 169, 286, 288–289
colloquial  47, 86
common ground  70
communication  3–5, 8, 42,
47–48, 50, 54–55, 94–95,
103–104, 111–112,
communicative  4, 26, 47–48,
50, 54–55, 64, 89, 111–112,
146, 149, 158, 212,
217–218, 289
composability  4
compositionality  4
constituency  43, 66, 95, 125
constituent  21–24, 62, 67, 70,
73, 77–78, 124, 147–151,
161–162, 263, 265–268,
278–279
constituents, dislocated  66
constitutive structure  42
construction, impersonal  59,
75, 77, 86–87, 89–92, 224,
232, 266–267
context-dependency  63
contrast  61, 71, 76–77,
145–147, 262–265, 267, 271,
273–276, 278
contrasting  65
contrasting status  70
course of events  22, 24
D
declension  71, 75, 80, 89
decode  130, 139
deep semantic level  125
deep semantics  18
deep structure  10
default  13, 68, 71, 77, 123–124,
222, 237–238, 257
dependency  42–43, 95
discourse  54, 55, 61–65,
148–149, 167–176, 183–184,
188–191, 217–218, 220–221,
263–268, 289–291
coherence see coherence
grounding  48, 50, 111
representation  32
strategy  49, 70, 76, 104, 217
topic  170, 175
 Index
discrete  35
dislocated constituents, left or
right  66
dislocation  66, 76, 86, 94, 126,
145–147, 154, 162, 291–293
distinguished  45, 62–63,
70, 149
dynamism  22
E memory  41, 50–51, 103–114,
ellipsis  64, 71, 78, 80, 91, 126,
167–170, 173–180
elliptical  231, 234, 248–250
emphasis  71, 74–75, 79, 236,
239–244, 247–248, 250,
253–254, 257–258, 266, 275,
285, 291
emphatic  240, 262, 275, 295
enclitic  233–235, 238–239, 248,
253–254, 257–258, 266,
274, 280, 291
end-focus  159–161
entity  7, 11–12, 48, 51–52
see also animate entity
event  7, 25–27, 29–30, 32–37,
200–201, 209, 211
structure
(of Winskel G.)  32, 36
existential  51–52
expressive  94, 211, 223–224,
235–236, 242, 255–257, 285
F 91, 124, 169, 173, 175–178,
felicitous  53, 121
first order logic  41, 52, 95
following (neighbouring)
situation  23
formal language  19, 32
(or symbolic)
representation  9–10, 14,
19, 28, 32, 36, 43
formalised pragmatic
theory  44
fronting  92, 126
G 167–170, 176–177, 217–221,
generative
grammar  8, 44
syntax  86
grammatical obligation  91
granularity  7, 25–26, 28–29, 31,
36, 201
grounding  48–52, 54–55, 105,
126–127, 130–131, 212–213
H lexical meaning  43, 207–209
hierarchical order of types of
situations  26
higher conceptual level  33
homonymy  71
human  7–8, 87–89, 184,
194–195, 224
hypercategory  199–200,
206–209
I 116, 118, 121–122,
imperfective partner  211
impersonal (utterance,
construction,
sentence)  59, 75, 77, 86-87,
89-92, 224, 232, 266-267
implicit  8–9, 59, 63–64, 71,
80–84, 89, 91, 187–189
implicit lexical
knowledge  167–168,
179–180
inanimate  7–8, 12, 52, 68, 171
indefinite pronoun  89, 91
information  4–6, 13, 19, 41–50
covert  89
informative  4–5, 16–17, 41–51
function  46, 139
instance  11, 14, 54
inter-situational
relationships  33, 36
interdisciplinary  42, 104
interpretation  17–18, 68, 71–72,
213, 216, 223–224, 234,
248, 290
intonation  63, 66, 72–73, 75,
82, 85, 93–94, 137, 157,
233–236, 247–248, 253,
256, 262–263, 273
intoneme  242–243, 256–258
inverse verbs  171
K microscopic (model of
knowledge  63–64, 68, 71, 73,
103–111, 126–131, 138–139,
290–291
acquisition  41, 50, 111
shared  48–51, 212
L exit  24, 28, 213
languages
ergative  66, 68, 78
nominative  68, 77–78
linear order  44, 66
time model  24
linearization  93, 127
M
macro-structural level  33
mapping  46, 55, 59, 95, 118,
126, 139–140, 195
126–140
dynamic memory
process  131, 133–134
episodic  51, 103, 106–114,
116, 118, 128, 130
explicit  129, 130
implicit  129, 130
long term  50, 108,
121–122, 126, 128, 130–131,
133–135
semantic  51, 103, 106–111,
114, 118, 128, 130, 133
short term  128, 132
working see working
memory
mental operations  42
reference  50
representation  42, 46, 55,
62, 63, 67
meta-informative
competence  93
function  46, 139, 290
layer  43
level  4, 16
level, first  60–61, 64,
86, 94
particle  85
status  47–51
structure  44
meta-language, semantic  19
micro-structural level  33
situation)  32
modal logic  32
modes of action  28, 32
modes of action see also
Aktionsart
moment (of a situation)
enter  24
finish  24, 28, 202, 205, 206
start  24, 28, 30, 31, 200, 202,
203, 207, 211, 213, 220
 Index 
multi-party conversation  calculus  44
183–186, 188, 191
multiple inheritance of
semantic features  6, 7
N primitives  22, 33
neighbouring situation  23
non human  7–8, 87, 89
O 200–201, 205–206, 209,
object
focalised  72–73
topicalised  74, 91–92
omission  64, 80, 170, 173
ontological knowledge  4, 6
ontological role see also role
ontology  4–7, 46, 50, 52, 195,
212–213, 216–218
oral  73, 75, 78–79, 94, 107, 115,
232–234
P prominent  59, 62, 70, 86, 114,
paradigmatic  5, 45, 54, 123
partial function  3–4
participant type  12
active  68, 71, 90
passive  71, 81
particle  75, 94, 179, 186–187,
244, 262, 279–280, 287,
290–291
particle ga, Japanese  82, 85,
172, 174–175
particle wa, Japanese  84–85,
172, 174–175
patient  68, 125
pause  63, 67, 82, 84–85, 245,
253, 256, 258
perfective partner  207–210
person  78–82, 87–88, 171,
173–174, 176, 265–266,
271–272, 281–282,
285–286, 288–290,
292–293
personal pronoun see also
pronoun
pertinence  136–138
Petri nets  22, 32–34
phrase structure tree  43
possible world  46, 139
postcondition  23, 37
postposed  90, 124–125, 238–
240, 246–258, 275–276
pragmatic force  77
predicate  9, 51, 60, 86
ontological  3
preposed  235, 238–245, passive  3, 7, 12–19, 68, 73,
249–258 74, 75
primacy effect  134–135 primitive  12, 16
primitive role see also role semantic  13, 17, 68, 71,
77–78
pro-drop  263, 266–268
process S
ordinary  26, 29–31, salience  63, 136, 171–174, 287
schizophasia  113, 140
211, 221 selection  45, 136
parallel  22, 24, 33 semantic
refined  24–28, 25–27, 29–31, knowledge  4
36, 200–201, 204, 209–211 nest  28
production  103–105, 107, role see role
113–115, 125, 130, 135, semantics
138, 140 universal  6
progression  24–26, 28–32, shallow level of  13
34–36 standard level of  13
prominence  71, 93, 123, 184, sentence, impersonal  59, 75,
187, 218, 286, 289–293, 295 77, 86–87, 89–92, 224, 232,
266–267
123, 292 set-theoretical  14, 33, 44, 52
pronoun sequential order  44, 66
accented personal  259–267, situation
270, 273–281 association  12 17, 18
anaphoric  76, 83, 290, view, external  194, 201–206
292–294 view, internal  23, 194,
atonic personal  231, 234, 295 201–206
enclitic  233–234, 253, 274 associated  12, 14, 15, 16
personal  49, 78–82, 234, dynamic  9, 13, 22–24,
239, 241, 245, 257, 26, 28, 30, 34–35,
260–261, 263–264, 268, 201–202, 205
277–279, 289–290, preceding
292–294 (neighbouring)  23
stressed  243–245, 247–254, static  9, 13, 22, 23, 26, 28,
256–257 200, 201, 209
tonic personal  234–235, whole  28, 63, 84, 206,
239–243, 256–258, 262 214–216
unaccented personal  space  4, 7, 9, 14, 22, 26, 28–29,
259–260, 262–267, 31, 45
273–274, 276–279, 281 stage
pseudo-cleft  146–147 after  24, 28, 31, 202, 205,
210, 220
Q before  23, 24, 28, 32, 202
quasi-active  3, 13, 92 begin  23, 24, 28, 31, 32,
quasi-subject  91 200, 204
run  23, 24, 28, 30, 32, 202,
R 220
recall  114, 116, 132–135, 137 stages of a situation  23–25,
recency effect  134–135, 137 28–33
role state  5, 7, 13, 22, 24–30,
active  3, 7, 12 –13, 17, 62, 68, 33–35, 37, 200–201, 206,
73, 77 209, 211
 Index
status Π-motivation
actual  50–51, 106, 126, 127,
212–213, 216, 218, 220–223
occasional  50–51, 106, 126,
212–213, 215–216, 220–221
particular  50–51, 106,
125–127, 212–214, 220
specific  50–51, 126, 212–216
status Σ-motivation
general  50–54, 84–85, 105,
212–215, 220, 223–225
generic  50–51, 84–85, 88–89,
105, 212–214, 216, 220,
223–224
habitual  50–51, 84–85, 105,
211–213, 215–216, 220,
222–225
potential  50–51, 84–85, 105,
212–213, 215–216, 222–224
stress  72–73, 234–235, 246–248,
256
structural linguistics  5, 46,
78, 196
stylistic  84, 89, 248–249, 285
subject
prominency  85
pronoun  67, 238, 241, 245,
263, 265–267, 269–271,
276–279, 288
subject-like  91
subject-prominent  59
subject-verb agreement  81, 82,
90, 168, 170, 176, 177
subject
absent  67
anonymous  74–75, 86–87,
89–92, 149
dative  91
dummy  87, 154
explicit  77–78, 83–84, 86–87,
89–91
focalised  73, 79
oblique  91–93
quirky  91
see also quasi-subject
topicalised  63, 85, 175
universal definition  77
subjectless  86
subsequent (neighbouring)
situation  23
suffix  210
suffixal verb  211
switch-reference systems  235–237, 239, 241–242, 250,
178–179
syntagmatic  5, 45, 54, 123
syntax  20, 42–45, 93, 95, 273–274
T free (FWO)  43, 63, 75, 93,
topic prominency  85
topic-prominent  59, 86
transition  22–23, 33–35
truth  10, 54
conditions  4, 146 154
type  11, 14–19, 26–28, 50–52,
195, 201, 220
U Z
unknown  48, 50–51, 55, 104,
212, 218
usage  10, 193–197
utterance, impersonal  59, 75,
77, 86–87, 89–92, 224, 232,
266–267
V
valence  14, 17, 19, 42–43, 55,
68, 71, 77
schemata  43
verbal prefix  197–199, 207–210
voice
active  69, 71
passive  59, 69, 74, 75, 89
W
word order  43, 63, 72–73,
75–76, 93–94, 123–126,
137–138, 145, 147, 231–232,
255–258, 262–263, 267,
273, 275–276, 278–279,
282, 288, 292, 294
canonical  93, 125–126
123, 125, 256
rigid (RWO)  43, 59, 86,
93-94
working memory  103–105,
109, 114, 121–122, 128,
130–137, 140
written language  76, 78, 121
zero form  67, 80, 89–92, 263
Π-situations  49–51, 106
Σ-situations  49–51, 105