Cognitive Paths Into The Slavic Domain (Eds. D.divjak&a.kochańska)

Cognitive Paths into the
Slavic Domain
Edited by
Dagmar Divjak
Agata Kochanska
Mouton de Gruyter
Cognitive Paths into the Slavic Domain
≥
Cognitive Linguistics Research
38
Editors
Dirk Geeraerts
René Dirven
John R. Taylor
Honorary editor
Ronald W. Langacker
Mouton de Gruyter
Berlin · New York
Cognitive Paths
into the Slavic Domain
Edited by
Dagmar Divjak
Agata Kochańska
Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin
앪
앝 Printed on acid-free paper
which falls within
the guidelines of the ANSI
to ensure permanence and durability.
Library of Congress Cataloging-in-Publication Data
Cognitive paths into the Slavic domain / edited by Dagmar Divjak,

Agata Kochańska.
p. cm. ⫺ (Cognitive linguistics research ; 38)
Includes bibliographical references and index.
ISBN 978-3-11-019620-7 (hardcover : alk. paper)
1. Slavic languages ⫺ Grammar. 2. Cognitive grammar. I. Divjak,
Dagmar. II. Kochańska, Agata. III. Title.
PG59.D58 2007
491.8104⫺dc22
2007035019
Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
ISBN 978-3-11-019620-7
ISSN 1861-4132
쑔 Copyright 2007 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin
All rights reserved, including those of translation into foreign languages. No part of this book
may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopy, recording, or any information storage and retrieval system, without
permission in writing from the publisher.
Printed in Germany
Preface
The present volume contains a selection of research on Slavic languages
carried out within the Cognitive Linguistics framework. Most of the papers
were originally presented at the 3rd Slavic Cognitive Linguistics
Conference, which took place in September 2004 in Leuven (Belgium).
Contributions by Barbara Dancygier, Laura Janda and Elżbieta
Tabakowska were later solicited specifically for this volume. All of the
papers have been written or revised with one goal in mind: presenting a
volume of interest to both Slavic linguists and cognitive linguists.
To the extent that we have succeeded in achieving our goal, we are
greatly indebted to the many linguists who have lent us their expertise, be
it on Slavic or non-Slavic languages, in cognitive or descriptive linguistics,
i.e. Neil Bermel, Mario Brdar, Alan Cienki, Steven Clancy, Hubert Cuy-
ckens, Östen Dahl, David Danaher, Barbara Dancygier, Larry Feinberg,
Dirk Geeraerts, Elżbieta Górska, Stefan Gries, Gaëtanelle Guilquin,
Tuomas Huumo, Laura Janda, Robert Kirsner, Wojciech Kubiński, Ron
Langacker, Geoffrey Nathan, Elena Petroska, Anna Siewierska, Michael
Smith, Elżbieta Tabakowska, Willy Van Langendonck and Margareth
Winters. We would also like to express our gratitude to René Dirven, Dirk
Geeraerts, Laura Janda and Ron Langacker for aiding us in taking our idea
from book proposal to final product. Torkel Uggla and Ines Van Houtte
prepared the manuscript for publication – a task that was financially
supported by the Science Foundation Flanders (Belgium) – while Birgit
Sievert walked us through the production process. Last but not least, warm
thanks go to our respective husbands, Torkel and Andrzej, for supporting
our work, both in theory and in practice, as well as to Agata’s mother
Elżbieta Horszczaruk for her assistance in matters of everyday life.
While working on this volume we were guided by the strongly-felt hope
that the wide spectrum of cognitively-oriented research on Slavic data
presented would contribute to the appreciation of both the beauty of the
notoriously complex Slavic languages and the power of the theoretical
tools developed within the cognitive framework. These tools allow a
linguist to tackle language phenomena in all their wonderful complexity
and to enjoy subtleties and intricacies without any need to disregard non-
conforming facts or to force natural language into artifical shapes.
Dagmar Divjak and Agata Kochańska

Sheffield and Warsaw, October 2007
Table of contents
Preface ..................................................................................................... v
Why cognitive linguists should care about the Slavic languages and
vice versa ……………………………………………………………… 1
Dagmar Divjak, Laura A. Janda and Agata Kochańska
Part one. The nominal system: the meaning of case

Nominative and instrumental variation of adjectival predicates with
the Russian copula byt': reference time, limitation, and focalization ... 21
Alina Israeli
Why double marking in the Macedonian dativus sympatheticus? ....... 55

Liljana Mitkovska
Part two. The verbal system: the meaning of tense, aspect and
mood
What makes Russian bi-aspectual verbs special? ................................. 83
Laura A. Janda
Perfectives, imperfectives and the Croatian present tense .................. 111

Renata Geld and Irena Zovko Dinković
Conflicting epistemic meanings of the Polish aspectual variants in

past and in future uses: are they a vagary of grammar? .................... 149
Agata Kochańska
Conjunctions, verb forms, and epistemic stance in Polish and

Serbian predictive conditionals .......................................................... 181
Barbara Dancygier and Radoslava Trnavac
Part three. The sentential system: non-archetypal event

conceptions
Degrees of event integration. A binding scale for [Vfin Vinf]
structures in Russian ........................................................................... 221
Dagmar Divjak
viii Table of contents
The ‘impersonal’ impersonal construction in Polish. A Cognitive

Grammar analysis ................................................................................ 257
Anna Słoń
Part four. Changing language: category shifting

A Frame Semantic account of morphosemantic change: the case of
Old Czech věřící .................................................................................. 291
Mirjam Fried
A prototype account of the development of delimitative po- in

Russian ……………………………………………………………… 329
Stephen M. Dickey
The rise of an epistemic pragmatic marker in Balkan Slavic:

an exploratory study of nešto .............................................................. 375
Eleni Bužarovska
Part five. Motivating language: iconicity in language

Iconicity and linear ordering of constituents within Polish NPs ......... 411
Elżbieta Tabakowska
Discourse-aspectual markers in Czech sound symbolic expressions:

Towards a systematic analysis of sound symbolism ........................... 431
Masako U. Fidler
Subject index ………………………………………………………... 459

Why cognitive linguists should care about the Slavic
languages and vice versa
Dagmar Divjak, Laura A. Janda and Agata Kochańska
1. The cognitive paradigm and Slavic linguistic research
From its early days, cognitive linguistics has attracted the attention of lin-
guists with research interests in Slavic languages (to name but a few,
Cienki 1989; Dąbrowska 1997; Janda 1993a; Rudzka-Ostyn 1992 and
1996). In recent years this interest has rapidly expanded, as can be wit-
nessed by the establishment of the Polish Cognitive Linguistics Associa-
tion, the Russian Cognitive Linguistics Association, and the Slavic Cogni-
tive Linguistics Association, as well as by the many Slavic Cognitive
Linguistics conferences held at various venues in Europe and North Amer-
ica over the last seven years.
This is not surprising, for at least two reasons. First, one of the founding
assumptions of cognitive linguistics has been present in Slavic linguistics
all along: Slavic linguists have always recognized the fundamentally sym-
bolic nature of language and hence the fact that diverse formal aspects of
language exist for the purpose of conveying meaning. One striking illustra-
tion of the close affinities between cognitive linguistics and ideas formu-
lated within traditional Slavic linguistics comes from the relatively early
days of modern linguistic research on Slavic languages. In a study devoted
to the nature of the contrast between the perfective and the imperfective
aspect in Polish, a German Slavicist, Erwin Koschmieder (1934), proposed
two conceptualizations of time which could easily be paraphrased as in-
volving either the MOVING TIME metaphor for the perfective or the
MOVING EGO metaphor for the imperfective (for a discussion of the two
time metaphors see Radden 1991: 17ff). Other examples abound. Tradi-
tional analyses of Polish case by Kempf (1978), Klemensiewicz (1926) and
Szober (1923 [1963]) aimed to provide a full-fledged semantic analysis of
Polish case. This type of work with its emphasis on psychologically realis-
tic explanations, has always been “a characteristic feature of Polish
(Slavic?) linguistics” (Tabakowska 2001:12; translation AK), and contin-
2 Dagmar Divjak, Laura A. Janda and Agata Kochańska
ues to constitute an important source of insight and inspiration for cogni-

tive research in the area of Polish case.
In order to appreciate properly how cognitive linguistics resonates in
the Czech context, it is necessary to outline some basic facts pertinent to
the history of the Czech language and the development of linguistic ideas
in Prague. The Czech language had been excluded from the public arena
for nearly two hundred years and seemed headed for extinction when Josef
Dobrovský published a grammar of the language in 1809. Unbeknownst to
him, the Czech national revival was to follow shortly thereafter, and his
grammar was used to revive Czech and to restore its use in official do-
mains. In order to achieve this goal the vocabulary of the language needed
to be enriched, and metaphorical extension and metonymy played an im-
portant role. Lexical creations attributed to Josef Jungmann (the central
figure in this process) include odstín ‘nuance, shade of meaning’ (a meta-
phorical extension from stín ‘shadow’) and savec ‘mammal’ (a metonymi-
cal creation with the literal meaning ‘one that sucks’). Thus, the idea that
metaphor and metonymy play an important role in language remains be-
yond doubt for Czech scholars. In 1928-1939 the Prague Linguistic Circle
boasted famous Russian and Czech linguists who collaborated on develop-
ing a structuralist framework that in the post-WWII era evolved into lin-
guistic functionalism. These linguistic models contained concepts similar
to category structure and center (a.k.a. prototype) vs. periphery distinctions
(Vaňková et al. 2005: 33–34; Janda 1993b). The recognition of the role of
pragmatics in linguistics is a consistent theme in the history of Czech lin-
guistics and likewise provides a point of contact for cognitive linguistics.
Close affinities between the ideas developed within traditional Slavic
linguistics and the assumptions of the cognitive paradigm are also clearly
visible in Russian linguistics, especially in writings by followers of the
Moscow Semantic School. Cognitive linguists study how the structure of
language is dependent on our physiology, and our interaction with the envi-
ronment. Langacker (1987a: 47) argues that language-specific semantic
structure, made up of “conventional imagery”, must be distinguished from
a universal conceptual structure: “Lexicon and grammar are storehouses of
conventional imagery, which differs substantially from language to lan-
guage. (…) It is therefore a central claim of cognitive grammar that mean-
ing is language-specific to a considerable extent. It is this imagery that has
to be described, not the presumably universal cognitive representations that
these conventional images construe”. In the Russian tradition a similar idea
is expressed by the term “anthropocentrism” (Rakhilina 2000: introduc-
tion): language is tailored by human beings to their needs. Followers of the
Why cognitive linguists should care about Slavic languages 3
Moscow Semantic School propound that language structures on all levels

reflect the collective experience of the speakers of a language, and thus
linguistic data provide a “linguistic world view” (Rakhilina 2000: 10–11),
shared by the speakers of that language.
Politics have played a crucial role in bringing the Slavic linguistic tradi-
tion and the cognitive paradigm close to each other. Political circumstances
in Slavic-speaking countries during the Cold War era forced many linguists
into exile. Among them was one individual who had an enormous impact
on Slavic linguistics: Roman Jakobson. Despite his own experiments with
formalist descriptions (such as the one-stem verb system), Jakobson was
sympathetic to many functionalist ideas that would later form the core of
the cognitive linguistic framework (cf. Janda 1993b). Jakobson’s presence
shielded Slavic linguistics in the West, especially in the US in the 1980s,
from being entirely consumed by mainstream formalism, which almost
eclipsed all other approaches. The Cold War era was the time when East-
ern European linguists in general and Russian linguists in particular were
largely isolated from theoretical discussions in the West, and the politically
unrestrained writings of Chomsky led to the censorship of his entire oeu-
vre. As a consequence, East-European linguists were never forced to ex-
periment with autonomous theories of language, but rather maintained
focus on the form-meaning relationship and how it is embedded in the lar-
ger reality of human experience. They turned their energies inward, devel-
oping their own home-grown traditions, some of which became known in
the West. These include the Russian Smysl↔Tekst framework, first devel-
oped by Mel’čuk (1995 and 1999) in Moscow and the Natural Semantic
Metalanguage theory formulated by Wierzbicka (see 1972 for the first
book-length treatment). Most of the work done in Eastern Europe, how-
ever, never made it to the other side of the Iron Curtain, which is all the
more regretful since analyses presented, for example, by followers of the
Moscow Semantic School focus on precisely those issues that are of inter-
est to cognitive linguistics. This is illustrated, among others, by the work
done on metaphor by Arutjunova (1999) or on polysemy and synonymy by
Apresjan (Apresjan 1974 and 1995). The data presented and the conclu-
sions drawn are so relevant to cognitive linguistics that it has been claimed
only a list of terminological equivalents is needed to bridge the gap (Ra-
khilina 1998).1
Given that the fundamentally symbolic nature of language has always
been recognized in the Slavic linguistic tradition, one might doubt that
cognitive linguistics would have something to offer researchers working on
Slavic languages. After all, trying to look at Slavic data from a cognitive
linguistic perspective could be considered as merely recasting old ideas,

revamping them using a perhaps more fashionable vocabulary, with no real
gain as far as depth of understanding or explanatory power is concerned.
We believe, however, that this line of reasoning is misguided in several
important respects. It is of course short-sighted to assume that every theo-
retical claim made or assumption put forward by cognitive linguists has the
character of a truly revolutionary insight that was entirely alien to and per-
haps even unthinkable in the “pre-cognitive” linguistic world. Quite the
contrary, it seems that when the evolution of linguistic thought is looked at
from a sufficient distance, one finds more continuity than expected (cf.
Geeraerts 1988). Progress in linguistic science seems to resemble an up-
ward spiral movement. In a sense, we move in circles and return to those
places we have visited before, albeit that, with each new lap, we reach a
higher level. The theoretical framework of the cognitive paradigm has the
potential to move research in the domain of Slavic languages a level up,
where precise and detailed descriptions of the conceptual import of multi-
ple linguistic structures can be offered, where numerous and diverse lin-
guistic phenomena can be characterized in terms of a limited number of
general well-attested cognitive mechanisms, where not only the workings
of languages can be meticulously described, but can also be seen as moti-
vated by things larger than language itself – by the general human cogni-
tive make-up, by our biological, social, and cultural experience of the
world.
In the remaining part of this introduction we would like, first, to con-
sider some of the attractions that Slavic languages hold for cognitively-
minded researchers (section 2). Then, in section 3, we will discuss some of
the main theoretical assumptions of the cognitive paradigm, with special
emphasis on those ideas that are particularly relevant to the research in the
domain of Slavic languages presented in this volume. Finally section 4 will
be devoted to an overview of the volume, which is meant as a representa-
tive selection of work, illustrating a wide array of research topics that are
currently on the Slavic cognitive linguistic agenda.
2. Slavic languages: an ideal laboratory for a cognitive linguist
Slavic languages have multiple attractions in store for a cognitive linguist,

in particular in terms of the range of linguistic phenomena available. They
have few, if any, peers worldwide in terms of the size of this family of
languages: by any count (and the counts vary with the political allegiances
of the counters) there are at least a dozen Slavic languages, spoken by

close to a half billion people across an area covering over 1/6 of the dry
land on Earth. Also, few languages can compete with the Slavic family as
far as the documentation of their characteristics, both diachronic and syn-
chronic is concerned. By a great stroke of luck, SS. Cyril and Methodius,
the “Apostles to the Slavs”, undertook their Moravian mission and thus
inaugurated the development of a Slavic literary language just in time to
capture a very near equivalent to Late Common Slavic, the shared language
of the Slavs prior to their further linguistic differentiation. In their late
ninth century translations of the gospels, these saints codified what is now
known as Old Church Slavonic, a language which, despite certain Greek
influences and artificial features, allows us to triangulate effectively be-
tween the modern languages and the Proto-Indo-European trunk. Thus the
Slavic languages have something that even English (and its Germanic sib-
lings) lack: a fully-documented mother tongue. Though the record is not
without gaps, we do have over a thousand years of Slavic texts, enabling us
to trace in detail the histories of the daughter languages, and new discover-
ies are still being made. Given this breath-taking affluence of the historical
data available to students of Slavic languages, it is not surprising that the
present volume contains papers which are explicitly concerned with issues
pertaining to diachronic language change.
The menu of potential objects for linguistic inquiry (both diachronic
and synchronic) among modern Slavic languages is quite rich, thanks to the
roster of linguistic categories exquisitely articulated by their inflectional
and derivational morphology. The two main courses are case and aspect
and some issues pertaining to both of these areas of empirical investigation
are addressed in the present volume. Selecting from a long list of appetiz-
ers and side dishes that Slavic languages have to offer their connoisseurs,
the volume further discusses the proliferation of impersonal constructions
in Slavic languages, with special emphasis on constructions used to convey
the idea of a highly diffuse and unspecified causer. It also reflects on the
way in which Slavic languages encode complex events and the means they
use to convey the speaker’s epistemic stance; it also deals with issues re-
lated to the relatively free word order in Slavic languages and, finally, it
considers sound symbolic expressions.
Obviously, the present volume merely touches upon the above-
mentioned topics, leaving aside a vast range of other and equally delicious
specialties in the Slavic cuisine. Let us mention just a few items to whet
the appetite. Bulgarian and Macedonian have retained all the inherited past
tenses without compromising the distribution of aspect, yielding unex-
pected combinations such as imperfective aorists and perfective imperfects.

The old perfect has matured in these languages into an evidential tense
with a fascinating array of uses, including the “admirative”. Czech is
probably in the process of developing its own set of articles, oddly enough
recapitulating the history of English, with the numeral jeden ‘one’ serving
as the source for the indefinite article and the demonstrative ten ‘this/that’
as the source for the definite article (Kresin 2001). All Slavic languages
have a three-way gender distinction of masculine vs. feminine vs. neuter,
usually with further distinctions within the masculine involving various
construals of animacy and virility. Slavs show evidence of an enduring
preoccupation with counting men, since most of their languages have spe-
cial numerals and plural desinences used only with reference to male hu-
man beings. In Polish there are even special syntactic constructions just for
reference to the “virile” category. Sorbian, which shares with Slovene the
maintenance of the dual number, further observes a virility distinction in
the dual, which is labeled in Sorbian textbooks as an opposition of “ra-
tional” (i.e., male human beings) vs. “irrational” (including everything else,
ranging from women over rabbits to books). Ergativity may be creeping
into Polish, where the logical subjects of reflexive verbs are marked with
the Accusative, not the Nominative, as in Brown ma doskonały styl i ksi-
ążkę się czyta szybko i przyjemnie [Brown.NOM has perfect style.ACC and
the book.ACC REFL reads quickly and pleasantly] ‘Brown has a perfect
style and the book reads (literally ‘is read’) quickly and pleasantly’. Of
course, one could go on listing numerous other entrees on the Slavic menu
and still remain far from being exhaustive. Perhaps it is an overstatement to
say “If it has happened in any language, it has happened in a Slavic lan-
guage”, but this claim is not far from the mark: most known linguistic phe-
nomena do indeed have Slavic parallels.
Importantly, none of the above-mentioned or a host of other fascinating
phenomena have been “overstudied” in the literature. Fortunately, several
of the Slavic-speaking countries have created on-line national corpora that
support searches for linguistic parameters, such as the Russian National
Corpus (http://www.ruscorpora.ru), the Czech National Corpus
(http://ucnk.ff.cuni.cz), the Polish National Corpus (http://www.pelcra.pl)
and the Croatian National Corpus (http://www.hnk.ffzg.hr). We hope that
the present volume will be instrumental in bringing the richness and beauty
of Slavic languages closer to the cognitive community at large as this rap-
prochement would be beneficial to both the study of Slavic languages and
the development of cognitive theory.
In the next section we will briefly discuss the main theoretical concepts
developed so far within the cognitive paradigm, with special emphasis on
those assumptions and ideas that are most directly relevant to the analyses
offered in the present volume.
3. The theoretical framework of cognitive linguistics
3.1. The prototype approach to categorization
Over the last three decades in mainstream linguistics the conviction has
grown that language is not a purely formal, algorithmic system processed
in a separate language faculty. Instead, our language capacity is considered
an integrated part of human cognition. The description of language is thus
a cognitive discipline, part of the interdisciplinary field of cognitive sci-
ences. One of the fundamental qualities of human cognition that is most
pervasively present in language is categorization.
Categorization is not a matter to be taken lightly. There is nothing more

basic than categorization to our thought, perception, action and speech
(…) An understanding of how we categorize is central to any understand-
ing of how we think and how we function, and therefore central to an un-
derstanding of what makes us human (Lakoff 1987a: 5–6).
Categorization, in other words, matters to the linguist in at least two ways,

i.e. “both in its methodology and in its substance” (Taylor 1989: 1). A lin-
guist needs categories to describe the object of investigation, while the
objects that linguists study also stand for categories.
The view on categorization that prevails in cognitive linguistics is no
doubt prototype theory, introduced now more than three decades ago by
Eleanor Rosch (for an overview of her main psychological writings as well
as diverse kinds of linguistic applications see Taylor 1995). In the proto-
type approach to categorization, concepts are categories comprising proto-
typical members (be they local or global), as well as more peripheral mem-
bers, which constitute diverse kinds of motivated extensions from that
prototype. Two such motivating mechanisms are conceptual metonymy
(the mechanism of mentally accessing one entity via another (salient) entity
co-occurring within the same conceptual domain – cf. e.g., Lakoff and
Johnson 1980 ch. 8; Langacker 1993: 29ff) and conceptual metaphor (par-
tially understanding one – typically more abstract – domain of experience
via another – typically more concrete – domain of experience – cf. e.g.,

Lakoff and Johnson 1980; Lakoff 1990; Lakoff and Johnson 1999: 45ff).
Often, category members are linked to one another just via the proto-
type, that is, if member A is the prototype, member B will be similar to A,
and member C will be similar to A, but B and C are not necessarily similar
to each other. The link that exists between members in a radial category
does not need to reflect any objective relatedness between the entities in
reality. Instead, their conceptual relatedness is a reflection of what the hu-
man conceptualizer experiences as a result of his biological and cognitive
make-up, as well as his bodily, social, and cultural baggage. Members of a
linguistic category, e.g., interrelated senses, are linked to each other by
categorizing relationships such as instantiation and extension (Langacker
1999: 101–103). Both involve an act of comparison in which a standard is
matched against a target. Instantiation is a limiting case of extension that
arises when the discrepancy is zero. Extension constitutes recognition ac-
complished only with a certain amount of “strain”. Extension does not
occur at random, however – it implies some abstract commonality. “[T]he
‘outward’ growth of a lexical network by extension from prototypes is
inherently associated with its ‘upward’ growth by extraction of schemas”
(Langacker 1987a: 373). Perceived similarities among sub-groups of mem-
bers of a conceptual category are captured by schemas at various levels of
abstraction, a schema being an abstract characterization that is fully com-
patible with all the members of the category it defines. Importantly, in the
schematic network model low-level schemas are claimed to be conceptu-
ally more salient than higher-level ones, and there is no necessity to postu-
late the existence of the highest-level schema capturing what is common to
all category members for each conceptual category. Hence, it is the norm
(rather than a deviation from the norm) that there are conceptual categories
with not even a single property shared by all category members.
Recognizing that linguistic categories (such as e.g., meanings of linguis-
tic expressions, grammatical constructions as well as categories, etc.) can
also have a prototype structure equips a linguist with the theoretical scaf-
folding on which to build a principled approach to synchronic polysemy
(or synonymy, for that matter), be it the polysemy of individual mor-
phemes, words, or grammatical constructions. The same scaffolding, when
considered from a slightly different perspective, is a fundamental part of
the theoretical apparatus that can open up new and revealing venues in the
investigation of diachronic language change (cf. Geeraerts 1997), which
may be viewed as a diachronic consequence of synchronic polysemy.
3.2. A conceptual and imagistic approach to meaning
Meaning in the cognitivist framework is no longer defined in terms of out-

side-world entities to which the expressions in question might refer, but
rather in terms of the conceptualizations they evoke in the minds of lan-
guage users (cf. e.g., Langacker 1987a: 116ff; 1988: 49f). Conceptualiza-
tion, in turn, should be understood as both the conceptual content and the
specific construal imposed on that content by the conceptualizer (cf. Lan-
gacker 1988: 58ff).
A conceptualist approach to meaning facilitates a systematic recogni-
tion and principled treatment of the subjective dimension of language:
when human beings conceptualize aspects of the world around them they
are often preoccupied with their own role in the conceptualization process
and their own relation to the entities they conceptualize. In other words,
human beings often do not merely conceive of outside entities, but also of
themselves conceptualizing the entities in question. This peculiarity finds
important reflections in language: linguistic expressions that speakers em-
ploy in discourse are used not only to comment on states of affairs in the
outside world, but also to convey the speakers’ epistemic evaluation of
what they are talking about, their assessment of their relation with their
interlocutors, comments pertaining to the development of the current dis-
course itself, etc. It is an explicitly conceptualist view of meaning that fa-
cilitates analyzing subjectivity in language in as systematic and detailed a
way as the phenomenon in question deserves.
Moreover, a truly conceptualist view of meaning allows us to construct
a comprehensive, principled framework for all instances of language use in
which conflicting characterizations are assigned to the “same” aspects of
the universe of discourse (cf. e.g., the traditional problems associated with
an analysis of the semantic behavior of expressions in the context of predi-
cates of propositional attitudes). For this purpose, cognitive linguistics has
developed mental space theory (cf. e.g., Fauconnier 1985). In this theory, it
is explicitly recognized that conceived situations in the universe of dis-
course may be conceptualized from multiple vantage points, each of them
having the potential to constitute a separate mental space. A change in
vantage point may bring about a change in how the observed parts of the
universe of discourse appear to the conceptualizing subject. As the growing
body of work in cognitive linguistics demonstrates (cf. e.g., Cutrer 1994;
Dancygier 1998; Dancygier and Sweetser 2005; Fauconnier 1997:95ff;
Sweetser 1990 ch.5), the theoretical constructs postulated within mental
space theory are of fundamental importance for a unified analysis of the
semantics of tense, aspect, and mood, to name but a few grammatical cate-
gories.
An important aspect of the conceptualistic view on meaning is the rec-
ognition of the imagistic component of semantics, that is, of the fundamen-
tal role construal plays in meaning. A precise characterization of its dimen-
sions allows an analyst to offer detailed and rigorous characterizations of
meaning contrasts among linguistic structures which are equivalent in
truth-conditional terms, but nevertheless exhibit subtle yet important dif-
ferences in meaning, resulting in otherwise unexplainable differences in
discourse behavior. A principled account of construal is a necessary pre-
requisite for developing a full-fledged symbolic approach to grammar:
grammatical meaning is by necessity abstract and can hardly be character-
ized in terms of specific conceptual content. It may, nevertheless, be in-
sightfully analyzed in terms of the type of construal it imposes on con-
ceived scenes, as demonstrated, for example, by the highly revealing
notional characterizations of nouns and verbs proposed by Langacker (cf.
e.g. 1987b).
3.3. A usage-based approach to language
The third theoretical assumption we would like to highlight here concerns

the motivation of linguistic phenomena. By rejecting the “autonomy of
language” principle, cognitive linguists abandoned any intention of formu-
lating generalizations with absolute predictability. Human behavior is not
governed by deterministic laws, and language cannot be separated from
other cognitive abilities, so absolute predictability cannot be achieved. This
turns out to be an advantage. The cognitive linguist, freed from the task of
looking for deterministic rules, is allowed to look for cognitive motivations
behind linguistic facts and to discover that these facts “make sense” within
a pattern larger than language itself – the pattern of how intelligent crea-
tures strive to understand the world around them and how they communi-
cate their insights to others of their kind.
Yet, if there are no deterministic rules to discover and learn, then how
do children acquire language and what are linguists looking for? In the
usage-based approach propounded by cognitive linguists, knowledge of a
language emerges from actual usage, i.e. as the result of the entrenchment
and abstraction of patterns that recur in multiple usage events. A usage-
based view of language structure offers a promising framework for a cogni-
tive approach to first language acquisition (cf. e.g. Dąbrowska 2004;
Tomasello 2003). At the same time, a usage-based view provides the right
perspective for the full appreciation of corpus studies in linguistic research
that no longer asks whether a certain phenomenon is possible or impossi-
ble, but instead focuses on how likely or unlikely the pattern is to occur
(see Gries and Stefanowitsch 2006). Last but not least, the adoption of the
usage-based model is important for the study of language change, as it lays
the ground for recognizing the role that is played in historical linguistic
evolution by factors such as frequency and mechanisms such as context-
bound pragmatic inferencing.
4. Why cognitive linguists should read this volume
The purpose of the present volume is twofold. On the one hand, we want to
investigate to what extent the theoretical framework and analytic tools
developed within cognitive linguistics can be insightfully applied to the
study of Slavic languages. As may be apparent from the brief discussion in
section 2 above, Slavic languages, with their rich inflectional morphology
in both the nominal and the verbal system, provide an important testing
ground for a linguistic theory that seeks conceptual motivation behind
grammatical phenomena. On the other hand, the specific observations and
insights arrived at in the course of cognitively-oriented analyses of diverse
phenomena in Slavic languages may enrich the understanding of already
established aspects of the cognitive model of language and serve as cata-
lysts for their further development and refinement.
This volume is important for a number of reasons. First, as far as its de-
scriptive range is concerned, the volume deals with a variety of empirical
phenomena that are of major interest to any linguistic theory. As men-
tioned above, the topics discussed include the semantics of case, tense, and
aspect, complex event conceptions, voice phenomena, word order, sound
symbolism, and language change. Secondly, the analyses address a variety
of theoretical issues that are important for cognitive linguistics in general.
Among them the reader will find: the role of virtual entities in language,
the importance of subjectification in motivating both synchronic polysemy
and diachronic language change, different ways of conveying the speaker’s
epistemic attitude, various kinds of non-prototypical event conceptions and
their grammatical reflections, the role of metaphor in grammaticalization,
and the influence exerted by local, contextual factors of pragmatic nature
in diachronic morphosyntactic change. Topics of general theoretical inter-
est also include the issue of iconicity in language and the idea that overtly
occurring language structures are “hints” helping language users construct

and manipulate complex configurations of mental spaces with differing
epistemic status. Finally, it should also be mentioned that the studies col-
lected in this volume incorporate insights from a variety of theoretical
frameworks that together form cognitive linguistics proper, such as e.g.,
cognitive grammar, mental space theory, construction grammar, frame
semantics, grammaticalization theory and prototype semantics with special
emphasis on its applicability to historical semantics. It is the diversity of
this volume on both the empirical and theoretical level that makes it ap-
pealing to the cognitive community at large.
The contributions we have selected offer a representative sample of cur-
rent research in cognitively oriented Slavic linguistics, touching upon five
areas of interest to both Slavic linguists and cognitive linguists in general:
(i) the highly developed nominal system with its extensive case morphol-
ogy; (ii) the rich verbal system with its aspectual markers and multiple
tense distinctions; (iii) clausal syntax as a reflection of how events are
construed for the purpose of linguistic communication; (iv) strategies of
change that illustrate how the current systems have come into existence
and how they are likely to change; and finally, (v) motivations for the
structure of the existing systems, as offered by principles such as iconicity.
Part One of this volume presents two articles devoted to case. Israeli’s
article is a detailed analysis of contextual factors that motivate the choice
of the Instrumental rather than the Nominative in Russian predicates with
the copula byt’. She argues that the decision of the speaker to represent
events as or as if observed triggers the Nominative in the predicate. Focus
on something other than the participant described, or on time comparison
and on time limitation, trigger the Instrumental. In turn, Mitkovska’s study
explores the conceptual motivation for the double marking of possession in
Macedonian constructions in which the Dative appears together with a
possessive pronoun. She argues that the double marking of possession is
motivated pragmatically, i.e. by the need of the speaker to present the pos-
sessive relationship from the perspective of the possessor and to highlight
the possessed.
Part Two deals with issues pertaining to the semantics of tense and as-
pect markers in Slavic languages. Janda presents an empirical study of the
aspectual behavior of borrowed verbs in Russian. She finds that 40% of
them are imperfective, whereas 60% are bi-aspectuals that do not exhibit
the traditional imperfective/perfective distinction. The strong correlation
between the aspectual profile of a borrowed verb and its tendency to form
po- prefixed perdurative verbs reveals the influence the lexical semantics
of a borrowed verb plays in determining whether it will be recognized as a

bi-aspectual or a more ordinary simplex imperfective verb. The article by
Geld and Zovko-Dinković is an analysis of the non-present uses of the
present tense in Croatian. The authors suggest that the link between these
uses and the prototypical present-time meaning is the notion of epistemic
immediacy. In turn, Kochańska’s paper considers the respective epistemic
values of the Polish perfective and imperfective aspect in the past and the
non-past tense. The epistemic meanings of the two aspectual variants are
analyzed as motivated extensions from their prototypical senses. The au-
thor’s claim is that although each of the two aspectual variants exhibits
conflicting epistemic behaviors in the past and in the non-past tense, this
may be accounted for by taking into consideration the prototypical mean-
ings of both aspects and how they interact with the epistemic values of
past, present and future time frames. The last study in this part, by Dancy-
gier and Trnavac, is a mental-space analysis of conditionality in Polish and
Serbian, with reference to English. Data from temporal, conditional, and
coordinate constructions in Polish and Serbian are used to establish the
basic formal and semantic parameters defining conditional meaning. In
contrast to English, Polish and Serbian rely less on conjunctions and clause
order, and more on tense, mood and aspectual forms, as well as on overt
markers of sequentiality.
The next section, Part Three, contains two articles dealing with ques-
tions of how clausal syntax reflects the way in which events are conceptu-
alized. Divjak’s article investigates degrees of verb integration as well as
factors motivating them in the case of the [VFINVINF] construction in Rus-
sian. Playing on the human capacity to impose alternate structurings on a
conceived phenomenon, she provides evidence for the existence of an expe-
rientially motivated binding scale in Russian, a cline of eight different de-
grees of integration between the events expressed by means of a [VFINVINF].
In turn, Słoń’s article deals with the use of a Polish impersonal construc-
tion, the 3rd SG NEUTR construction, that defocuses a non-human and in-
animate instigator. She shows that this construction is used when the insti-
gator is particularly diffuse and difficult to identify.
Part Four of the volume is concerned with issues pertaining to language
change. Fried’s study analyzes mechanisms of morphosyntactic change on
the basis of the diachronic evolution of the Old Czech “long” present ac-
tive participle věřící ‘(the one) believing’ in relation to the polysemous
verb věřiti, from which it is derived. She concludes that the relative sur-
vival rates of individual uses are determined by an equilibrium between
polysemy and isomorphism. Dickey’s paper applies principles of prototype
semantics to explain the development of the Russian prefix po- from a

primarily resultative prefix to a delimitative prefix. He argues that the de-
velopment of modern Russian delimitatives followed the development of
po- as a perfectivizing prefix for determinate motion verbs. The last study
in this part, by Bužarovska, focuses on the semantic change of the indefi-
nite pronoun nešto into an epistemic mitigation modal in Macedonian,
within a wider Balkan Slavic context. She suggests that the strengthening
of invited inferences and subjectification are the two cognitive mechanisms
that play a major role in this metonymically-based process.
Finally, Part Five addresses the issue of iconic motivation in language.
Tabakowska’s article investigates the ordering of multiple (mainly double)
adjectival modifiers within Polish nominal phrases. Although the structure
of these NPs is shown to depend on the traditional dichotomous division of
adjectives into two categories – the characterizing (attributive) and the
specifying (restrictive) – for prototypical cases, the article demonstrates
that the borderline between them is fuzzy: an adjective may be allotted to
either category depending on communicative needs, which are often dis-
course-sensitive. In turn, Fidler focuses on sound symbolic expressions
(SSEs) in Czech and investigates how SSEs relate to grammar. By analyz-
ing how SSEs develop into discourse-aspectual markers, she contributes to
our understanding of processes of word derivation and variation in lan-
guage.
It is our hope that this collection of diachronic and synchronic research
on a wide range of phenomena in Slavic languages, carried out within a
variety of cognitive linguistic frameworks, will invite to explore the Slavic
domain further along cognitive paths. We remain confident that any such
exploration will be a fruitful and exciting enterprise.
Notes
1. This is, of course, an oversimplification of the situation. Where cognitive

linguists recognize the crucial role the structure and functioning of the human
brain plays in language and strive both to implement cognitive findings in their
linguistic models as well as to inform cognitive science with their linguistic
findings, linguists belonging to the Moscow Semantic School do not show
great interest in the cognitive physiological and psychological side of lan-
guage. This difference in focus results in differences in heuristic methodology.
According to the Moscow Semantic School, combinability of words signals
combinability of concepts (Rakhilina 2000: 10–11). Russian cognitive lin-
guists therefore claim that a cognitive approach to language “should rely on

the experience of all native speakers, as it is consolidated in their language,
and that experience reveals itself in the linguistic behavior of the lexeme,
above all in its combinatorial possibilities” (Rakhilina 2000: 353). In other
words, as opposed to American and European cognitive linguists who more
and more frequently resort to psycholinguistic methods to investigate concep-
tual structure, Russian linguists believe it should suffice to rely on linguistic
evidence of conceptual structure (cf. Rakhilina 2000: 10–11).
References
Apresjan, Ju.D.
1995 Izbrannye Trudy. Tom 1, Leksičeskaja Semantika [Selected Writings,
Part 1, Lexical Semantics]. Moskva: Škola “Jazyki Russkoj
Kul’tury”.
Apresjan, Ju.D.
1974 Regular polysemy. Linguistics. An international review 142:5–32.
Arutjunova, Nina D.
1999 Jazyk i mir čeloveka [Language and the world of man]. Moskva:
Škola “Jazyki Russkoj Kul’tury”. [2nd corrected edition].
Cienki, Alan
1989 Spatial Cognition and the Semantics of Prepositions in English, Pol-
ish, and Russian. Munich: Verlag Otto Sagner.
Cutrer, Michelle
1994 Time and Tense in Narrative and in Everyday Language. Unpub-
lished Ph.D. dissertation, University of California, San Diego.
Dancygier, Barbara
1998 Conditionals and Prediction. Cambridge: Cambridge University
Press.
Dancygier, Barbara, and Eve Sweetser
2005 Mental Spaces in Grammar: Conditional Constructions. Cambridge:
Cambridge University Press.
Dąbrowska, Ewa
1997 Cognitive Semantics and the Polish Dative. Berlin/New York: Mou-
ton de Gruyter.
2004 Language, Mind and Brain: Some Psychological and Neurological
Constraints on Theories of Grammar. Edinburgh: Edinburgh Univer-
sity Press.
Fauconnier, Gilles
1985 Mental Spaces. Cambridge, Mass.: MIT Press.
1997 Mappings in Thought and Language. New York: Cambridge Univer-
sity Press.
Geeraerts, Dirk
1988 Cognitive Grammar and the History of Lexical Semantics. In Topics
in Cognitive Iinguistics, Brygida Rudzka-Ostyn (ed.), 647–677. Am-
sterdam/Philadelphia: John Benjamins.
1997 Diachronic Prototype Semantics. Oxford: Clarendon Press.
Gries, St.Th., and A. Stefanowitsch
2006 Corpora in Cognitive Linguistics: Corpus-based Approaches to Syn-
tax and Lexis. Berlin/New York: Mouton de Gruyter.
Janda, Laura A.
1986 A Semantic Analysis of the Russian Verbal Prefixes za-, pere-, do-,
and ot-. Munich: Verlag Otto Sagner.
1993a A Geography of Case Semantics: The Czech Dative and the Russian
Instrumental. Berlin/New York: Mouton de Gruyter.
1993b Cognitive Linguistics as a Continuation of the Jakobsonian Tradition:
The Semantics of Russian and Czech Reflexives. In American Con-
tributions to the Eleventh International Congress of Slavists in Brati-
slava, Robert A. Maguire and Alan Timberlake (eds.), 310–319. Co-
lumbus, Ohio: Slavica.
Janda, Laura A., and Steven J. Clancy
2002 The Case Book for Russian. Bloomington, IN: Slavica.
Kempf, Zdzisław.
1978. Próba teorii przypadków [An attempt at a theory of case]. Wrocław:
Zakład Narodowy im. Ossolińskich.
Klemensiewicz, Zenon
1926 Orzecznik przy formach osobowych być [The predicate with personal
forms of ‘be’]. Prace filologiczne [Philological works] XI: 123–181.
Koschmieder, Erwin
1934 Nauka o aspektach języka polskiego w zarysie. Próba syntezy. [An
attempt at a comprehensive theory of Polish aspect. An outline].
Wilno.
Lakoff, George
1987 Women, Fire, and Dangerous Things. What Categories Reveal about
the Mind. Chicago: University of Chicago Press.
1990 The Invariance Hypothesis: Is Abstract Reason Based on Image-
Schemas? Cognitive Linguistics 1: 39–74.
Lakoff, George, and Mark Johnson
1980 Metaphors We Live By. Chicago: Chicago University Press.
1999 Philosophy in the Flesh. The Embodied Mind and Its Challenge to
Western Thought. New York: Basic Books.
Langacker, Ronald W.
1987a Foundations of Cognitive Grammar. Vol. 1. Theoretical Prerequi-
sites. Stanford, California: Stanford University Press.
1987b Nouns and Verbs. Language 63: 53–94.
1988. A View of Linguistic Semantics. In Topics in Cognitive Linguistics,

Brygida Rudzka-Ostyn (ed.), 49–90. Amsterdam/Philadelphia: John
Benjamins.
1993 Reference-Point Constructions. Cognitive Linguistics 4: 1–38.
1999 Grammar and Conceptualization. Berlin/New York: Mouton de
Gruyter.
Mel’čuk, Igor’, A.
1995 Russkij jazyk v modeli“smysl ↔ tekst” [The Russian language in the
“meaning↔text” model]. Moskva: Škola “Jazyki Russkoj Kul’tury”
[Wiener Slawistische Almanach Sonderband 39.]
1999 Opyt teorii lingvističeskich modelej “smysl ↔ tekst” [An attempt at a
theory of linguistic “meaning↔text” models]. Moskva: Škola “Jazyki
Russkoj Kul’tury”.
Radden, Günter
1991 The Cognitive Approach to Natural Language. L.A.U.D. Papers Se-
ries A, No. 300. Linguistic Agency of the University of Duisburg.
Rakhilina, Ekaterina V.
1998 Kognitivnaja semantika: istorija, personalii, idei, rezul’taty [Cogni-
tive semantics: history, figures, ideas and results]. Semiotika i infor-
matika 36: 274–323.
2000 Kognitivnyj analiz predmetnych imen: semantika i sočetaemost [A
cognitive analysis of physical names: semantics and collocation].
Moskva: Russkie Slovari.
Rudzka-Ostyn, Brygida (ed.)
1988 Topics in Cognitive Linguistics. Amsterdam/Philadelphia: John Ben-
jamins.
Rudzka-Ostyn, Brygida
1992 Case Relations in Cognitive Grammar: Some Reflexive Uses of the
Polish Dative. Leuvense Bijdragen 81: 327–373.
1996 The Polish Dative. In The Dative. Vol. 1. Descriptive Studies, Wil-
liam Van Belle and Willy Van Langendonck (eds.), 341–394. Am-
sterdam/Philadelphia: John Benjamins.
Szober, Stanisław
1963 Reprint. Gramatyka języka polskiego [A grammar of Polish].
Warszawa: Państwowe Wydawnictwo Naukowe, 1923.
Tabakowska, Elżbieta
2001 Kognitywizm: Obrazki z polskiej sceny [Cognitivism. Images from
the Polish scene]. Glossos 1 at http://seelrc.org/glossos/.
Taylor, John, R.
1989 Linguistic Categorization: Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Tomasello, Michael
2003 Constructing a Language. A Usage Based Theory of Language Ac-
quisition. Cambridge, Mass.: Harvard University Press.
Vaňková, Irena, Iva Nebeská, Lucie Saicová Římalová, and Jasňa Šlédrová
2005 Co na srdci, to na jazyku: Kapitoly z kognitivní lingvistiky [What's
on the heart is on the tongue: Chapters from cognitive linguistics].
Prague: Karolinum.
Wierzbicka, Anna
1972 Semantic Primitives. Frankfurt: Athenäum.
ȱȱ
ȱȱ¢Ǳȱȱȱȱ
ȱ
Nominative and instrumental variation of adjectival
predicates with the Russian copula byt': reference
time, limitation, and focalization
Alina Israeli
Abstract
This article reexamines the nominative/instrumental variation of adjectival

predicates. The nominative suggests either the permanence of the feature or
the speaker’s presenting events as if witnessed. The instrumental suggests
change, comparison, time limitation, or scope limitation. There are two pos-
sible ways of describing a changed state: either as temporal sequencing or
via a retrospective perspective. The first mode of description calls for the
nominative of the adjectival predicate referring to the earlier state while the
second mode calls for the instrumental of the adjectival predicate referring
to the earlier state. Additionally, the feature of focalization explains the use
of nominative as opposed to non-focalized instrumental.
Keywords: Russian language, adjectival predicates, instrumental, limita-

tion, temporal sequencing, retrospective perspective, focalization.
1. Introduction
The nominative/instrumental variation of adjectival predicates, including

those with the copula byt', has been studied previously, in particular by
Nichols (1981 and 1985), Černov (1983), and most recently by Zel'dovič
(2005). A variety of meanings have been postulated for the two cases to
account for the observed differences in their distribution. Nichols (1985:
362–363) suggests that the nominative case expresses “simple aorist past”
while the instrumental means “past habitual”:
(1) Včera on byl veselyj.

yesterday he was.M1 cheerful.M.SG.NOM
‘Yesterday he was cheerful.’
22 Alina Israeli
(2) V detstve on byl veselym.

in childhood he was.M cheerful.M.SG.INSTR
(both from Nichols 1985: 362)
‘As a child he was cheerful.’
Černov (1983: 91), who analyzes the same constructions, observes that
in the case of the “actualization of the temporal plan”, particularly with
time expressions emphasizing distance in time, the instrumental is usually
used. Černov labels this usage ‘pluperfect’ (the following examples are
from Černov 1983: 91):
(3) a. Togda, šest' let nazad ètot mal'čik byl

then six years ago this boy was.M
malen'kim, a ženščina molodoj.
small.M.SG.INSTR and woman young.F.SG.INSTR
(K. Simonov)
‘At that time, six years ago, this boy was small and the woman
was young.’
b. Kogda on byl ešče molodym i

when he was.M still young.M.SG.INSTR and
partizanil v ètix mestax, ego
fought (as guerilla) in these places his
molodaja žena byla vmeste s nim.
young wife was.F together with him
(A. Fadeev)
‘When he was still young and fighting as a guerilla around
these places, his young wife was with him.’
c. Odnako, v tu poru, o kotoroj idet reč', ona

however in that time about which goes talk she
byla daleko ne takoj dobroj i laskovoj
was.F far not such kind.F.SG.INSTR and tender.F.SG.INSTR
k ljudjam.
to people
(M. Gor'kij)
‘However, at the time we are talking about, she was far from so
kind and nice towards people.’
Case variation of adjectival predicates 23
Recently, Gasparov (1996: 227), comparing the nominative long form

and the short form of predicates, suggested that by using the nominative
the speaker invites the interlocutor to share his immediate impression of
the state/situation. Zel'dovič (2005) in his examination of the nominative
vs. instrumental adjectival predicates follows Gasparov in describing the
nominative use as “nabljudennost'” ‘sharing an observation’. Zel'dovič
(2005: 150) provides a rule that the instrumental is appropriate where nei-
ther sharing an observation nor the isolated quality of the situation is desir-
able.
But what makes either of those conditions undesirable? We do find par-
allel examples, such as (4), and consequently we must look for factors that
motivate the speaker to choose one construction over the other.
(4) a. Da, ded byl staryj i xlopot s

yes grandfather was.M old.M.SG.NOM and hassles with
nim ne men'še, čem s malym rebenkom.
him not less than with small child
(zhurnal.lib.ru/b/borzow_a_a/letter.shtml)
‘Yes, grandfather was old and caused no less hassle than a
small child.’
b. No ded byl starym, i azbuku Morze

but grandfather was.M old.M.SG.INSTR and alphabet Morse
uže zabyl …
already forgot.M.SG
(zhurnal.lib.ru/c/cwirk_a/provbud.shtml)
‘But grandfather was old and had already forgotten the Morse
code …’
This article will examine different types of narrative time, comparison,

limitation and focalization as features that determine the case of the predi-
cate adjective. Their implications for the lexical meanings of nouns and
adjectives, which in turn affect the choice of case, will also be examined.
The availability of data bases and search engines enables the selection of
parallel examples much more easily than previously. Out of the vast num-
ber of parallel examples, thoses with the most clearly juxtaposing prag-
matic contexts were chosen.
24 Alina Israeli
2. Types of narrative time and comparison
The past form byl may represent three different types of narrative time, the
first two of which involve comparison:
1. The time of the narrated event tn is compared with the time of speech (or
writing) ts. I will call this use of byl tn / ts. This is the case in (5):
(5) a. Znanie rodnoj istorii bylo očen' vysokim v

knowledge native history was.N very high.N.SG.INSTR in
drevnej Rusi...
old Russia
(D. Lixačev/ Nichols 1985: 363)
‘Knowledge of native history was very high in old Russia …’
b. v te gody, kogda my ešče žili v Staroj Russe i

in those years when we still lived in Old Russa and
mama byla sovsem moloden'koj.
mama was.F quite young.F.SG.INSTR
(D. Granin. Obratnyj bilet)
‘… during those years when we were still living in Staraya
Russa and mother was very young.’
2. The time of one narrated event tn1 is compared with the time of another
narrated event tn2. I will call this use of byl tn1 / tn2. This is the case in (6),
where pre-war memories of seventh grade (togda ‘at that time’) are com-
pared with post-war ones of a class reunion (‘lived to have grey hair’):
(6) Iz soroka čelovek, zakončivšix kogda-to 7 “B”, do sedyx

from forty people had-finished once 7 “B” till grey
volos dožilo devjatnadcat'. ... Naša kompanija togda byla
hair lived nineteen our company then was.F
nebol'šoj: tri devočki i troe rebjat …
not-big.F.SG.INSTR three girls and threesome guys
(B. Vasil'ev. Zavtra byla vojna)
‘Out of forty people who once had finished seventh grade, nineteen
lived to have grey hair. … Our gang was small then: three girls and
three guys …’
3. The time of the narrated event tn is not compared with another moment
in time. I will call this use of byl tn0. This is the case in (4a) and (4b) above
as well as in (7) below. It differs from type 1 by having to present events as
frozen in the moment: unlike Russia and mother in (5), raspberry in (7) did
not exist at a later moment thus not allowing even an implicit comparison.
In (4), it is the choice of the narrator to present events as frozen in time and
defying comparison.
(7) a. Malina byla sadovaja, očen' krupnaja …

raspberry was.F garden.ADJ.F.SG.NOM very large.F.SG.NOM
(V. Solouxin. Dom i sad)
‘The raspberries were from the garden and very large …’
b. *Malina byla sadovoj, očen' krupnoj …

raspberry was.F garden.ADJ.F.SG.INSTR very large.F.SG.INSTR
In (7b) the permanent nature of sadovyj ‘garden variety’, similar to kir-

pičnyj ‘made of brick’ below, makes it incorrect.
The predicative features may be of various kinds, the main one being
permanence vs. non-permanence. In reality, the distinction is more com-
plex, and we will discuss this point later on. First, however, we must con-
sider in more detail the proposed dichotomy. The following is a discussion
of how the three possible narrative time frames correlate with these two
main types of features under consideration.
A permanent feature, such as a description of the material an object is
made of, personal characteristics, and many others, exists as long as the
object having that feature does. Comparing this feature at any two points
during the life time of the object (tn1 / tn2) yields no change, since the fea-
ture is perceived as constant. Consequently, describing the object at any
moment (tn0) yields the adjectival form that conveys no-change, i.e. the
nominative. This form suggests the meaning of ‘observation’ (cf. Zel'dovič
2005).
If we compare Dom byl kirpičnyj / kirpičnym ‘The house/apartment
building was made of brick’, clearly, being made of brick could not be
perceived as temporary or evolving. And indeed in such stative descrip-
tions the nominative predominates:
26 Alina Israeli
(8) Dom byl kirpičnyj i, vidimo, očen' drevnij.

house was.M brick.ADJ.M.SG.NOM and apparently very old
(Genocid. Čuvstvo opasnosti)
‘It was a brick house, and evidently a very old one.’
When an object ceases to exist, its feature ceases to exist along with it,
and the information conveyed by the instrumental is that the comparison tn1
/ tn2 indicates a change (cf. Filip 2001). A Google search <29.V.2006>
produced eighty-seven examples of the nominative dom byl kirpičnyj and
five examples of the instrumental dom byl kirpičnym, four of which per-
tained to one and the same event of the explosion and the subsequent col-
lapse of an apartment building in Moscow. So in order to motivate this use
of the instrumental, the apartment building had to cease to exist:
(9) Poisk ljudej osložnjalsja tem, čto vzorvannyj dom

search people complicated by-that that blown.M.SG.NOM house
byl kirpičnym: …
was.M brick.ADJ.M.SG.INSTR
(Nezavisimaja gazeta No 170/1999-9-14)
‘The search for survivors was complicated by the fact that the build-
ing that had exploded was made of brick: …’
When looking at a photograph, a speaker has no point of reference other

than that frozen instant in time, and comparison is impossible:
(10) Emu bol'še vsego ponravilas' fotografija, potomu čto glaza

him more of-all pleased photograph because eyes
molodogo čeloveka ulybalis', a lico bylo otkrytoe
young man smiled.PL and face was.N open.N.SG.NOM
i prijatnoe. Emu, konečno i v golovu ne
and pleasant.N.SG.NOM him of-course even in head not
prixodilo, čto èto ego otec.
came that this his father
(Èdgar Berrouz. Tarzan)
‘He most of all liked the photograph, because the face of the young
man was open and pleasant with smiling eyes. It of course did not
occur to him that this was his father.’
In (11), the person in question has died, and consequently the feature
ascribed to his face has gone with him:
(11) Ego lico bylo otkrytym, svetlym. Èto bylo lico

his face was.N open.N.SG.INSTR light. N.SG.INSTR this was.N face
Rossii. Ne mogu poverit', čto Georgij Stepanovič
of-Russia not can.1SG to-believe that Georgy Stepanovich
[Žženov] ot nas ušel.
Zhzhenov from us left
(Učitel'skaja gazeta No 05/ 2006-02-07)
‘His face was open, light. This was the face of Russia. I cannot be-
lieve that G.S. Zhzhenov has left us.’
With non-permanent features, the speaker / narrator can choose between

two modes of description: the [+observation] feature or the implied
[+change] feature. Example (12a) introduces the speaker as the observer,
while example (12b) implies a change or, in other words, an implicit com-
parison:
(12) a. Deduška byl staren'kij.

grandfather was.M old.M.SG.NOM
b. Deduška byl staren'kim.

grandfather was.M old.M.SG.INSTR
‘Grandfather was old.’
Similarly, in (13) the speaker sees the openness of the face as a fleeting
momentary feature:
(13) Ego černo-sinie glaza vnimatel'no nabljudali za nej, lico

his black-blue eyes carefully watched after her face
bylo otkrytym i uprjamym…
was.N open.N.SG.INSTR and stubborn.N.SG.INSTR
(zhurnal.lib.ru/h/hikaru_b/108.shtml)
‘His blue-black eyes carefully observed her, his face was open and
stubborn …’
As we have seen in the case of otkrytyj ‘open’, the same feature could
be a permanent characteristic or a non-permanent one. This is also true for
adjectives like bol'šoj ‘big’. Comparing two types of elephants, the African
ones vs. the Asian one, while explaining the nominative vs. instrumental
variation in the predicates, Ionin and Matushansky (2002: 4) suggest that in
(14) “Nominative requires the subject to vary from situation to situation
28 Alina Israeli
while with an Instrumental predicate, the (definite) subject must be the

same”:
(14) a. V Afrike slony byli bol'šie, a v Azii –

in Africa elephants were big.PL.NOM and in Asia
malen'kie.
small.PL.NOM
b. # V Afrike slony byli bol'šimi, a v Azii –

in Africa elephants were big.PL.INSTR and in Asia
malen'kimi.
small.PL.INSTR
‘In Africa the elephants were large, while in Asia they were
small.’
In fact, it is not that the nominative case that requires the subject to
vary, but the quality of the feature that the nominative case represents: this
is what renders (14b) incongruent. Co-reference of the subject in (14b), as
opposed to non-co-reference in (14a), can be explained via reference to
permanence/ non-permanence of the charecteristic and to temporal com-
parison. In (14b) v Afrike ‘in Africa’ and v Azii ‘in Asia’ mean ‘while in
Africa’ and ‘while in Asia’ respectively, thus involving a temporal com-
parison of the type tn1 / tn2 and yielding a pragmatically impossible sen-
tence with non-permanent feature. In (14a) bol'šoj ‘big’ is a permanent
characteristic (pertaining not to individual elephants but to a species) not
allowing for temporal comparison.
The narrative time tn1 / tn2 de-facto means change or comparison. This
type of narrative time, however, offers two possibilities. Analyzing evolv-
ing reference, Moeschler (1996: 21) suggests that it can be construed either
as temporal sequencing (TS) or via a retrospective perspective (RP). In TS
events are viewed sequentially, that is, tn1 is perceived and described as a
point in time earlier than tn2, and tn2 is compared to tn1; in other words, we
are dealing with tn1 / tn2 proper. In RP the narration can view previous
points in time via flashbacks, or view the preceding moment tn1 as com-
pared to the “present” (tn0) or to a later point tn2, so this is strictly speaking
a reversed narrative time tn2 / tn1. In terms of temporal perspective, TS
views the narration as “then and later”, while RP views it as “now and
before”. The TS perspective, as in (15), calls for the nominative in the ad-
jectival phrase describing the state at t1, whereas the RP, as in (16), calls
for the instrumental.
(15) – Kogda ja byl takoj, kak tvoj Aleša, –

when I was.M such.M.SG.NOM as your Alesha
zagovoril on, neskol'ko uspokojas', – mne nebo
started-speaking he somewhat having-calmed-down to-me sky
kazalos' takim vysokim, takim sinim. Potom ono dlja menja
seemed such tall such blue later it for me
pobleklo, no ved' èto ot vozrasta?
faded but PCL this from age
(Ju. Kazakov. Vo sne ty gor'ko plakal)
‘“When I was like your Alesha,” he began after calming down
somewhat, “the sky seemed to me so high, so blue. Later it faded for
me, but this is from age, right?”’
(16) a. Kogda ja byl malen'kim, risunki byli drugie

When I was.M little.M.SG.INSTR drawings were others
i pravila drugie. Ty mne ob''jasni, požalujsta,
and rules others you to-me explain.IMP please
vaši teperešnie pravila.
your now.ADJ rules
(I. Grekova. Kafedra)
‘“When I was little, the drawings were different and the rules
were different. Please explain to me your current rules.”’
b. Togda ja byl staršim; a čto ja teper'? Ja

then I was.M older.M.SG.INSTR and what I now I
značu men'še, čem prjažka na uzdečke.
mean less than buckle on reins
(O. Genri. Serdce i krest)
‘At that time I was the elder, and what am I now? I mean less
than a buckle on the reins.’
The RP represents a form of comparison since the period which has

ended is viewed from some later point in time. The TS views the earlier
period as open-ended, as if the narrator placed himself back at the earlier
moment as an observer avoiding any comparison.
30 Alina Israeli
The stative (non-evolved) view, as in (15), and RP, as in (16), can be com-
bined and juxtaposed, as in (17):
(17) Deduška u Nasti byl staren'kijt0. Nu, konečno,

grandfather at Nastya was.M old.M.SG.NOM well of-course
on ne vsegda byl staren'kimt2/t1. Kogda-to on byl
he not always was.M old.M.SG.INSTR some-time he was.M
sovsem molodymt2/t1 i učastvoval v vojne, no teper',
quite young.M.SG.INSTR and participated in war but now
nado priznat', on byl staren'kimt0/t1. Ešče kogda Nastja
must to-admit he was.M old.M.SG.INSTR still when Nastya
byla malen'kojt2/t1, deduška byl bodrymt2/t1,
was.F little.F.SG.INSTR grandfather was.M vigorous.M.SG.INSTR
no teper' on xodil, opirajas na poločku, i kuril
but now he walked.M leaning on cane and smoked
staruju trubku, kotoraja byla ego trofeem s voennyx
old pipe which was.F his trophy from war.ADJ
vremen.
times
(Deduška. http://www.foryou.kherson.ua/slovo.1083.2.html)
‘Nastya’s grandfather was old. But of course he hadn’t always been
old. Once he was really young and participated in the war, but now,
one must admit, he was old. When Nastya was still little, grandfather
was full of energy, but now he walked with a cane and smoked an
old pipe which was his trophy from the war times.’
The first occurence of staren'kij ‘old’ is stative, the kind that Nichols
describes as having descriptive force and Zel'dovič as observed, while the
second occurence with teper' ‘now’ (cf. Mel'čuk 1985) implies compari-
son, a change of state.
In (18), which refers to a meeting of two school friends some twenty
years after graduation and the war, there are two instances of the nomina-
tive case:
(18) Ja pokosilsja na Venju. On byl lysyj, glaza

I looked-slanted at Venya he was.M bald.M.SG.NOM eyes
ego vycveli, no ja vspomnil, kakie oni byli
his faded but I remembered what they were
nebesno-golubye i kak on nravilsja devčonkam. On

sky-blue.PL.NOM and how he pleased girls he
byl samym dobrym iz nas i
was.M most.M.SG.INSTR kind.M.SG.INSTR from us and
samym doverčivym. On svjato veril vsemu,
most.M.SG.INSTR gullible.M.SG.INSTR he holy believed everything
čto govorili, pečatali, učili.
that spoke.PL printed.PL taught.PL
(D. Granin. Dom na Fontanke)
‘I glanced at Venya. He was bald, his eyes faded, but I remembered
what a sky blue color they used to be and how girls liked him. He
was the kindest among us and the most trusting. He faithfully be-
lieved everything he heard, read or was taught.’
One instance pertains to the moment tn0 (on byl lysyj ‘he was bald’) and
the other to a moment twenty years earlier at tn1 (oni byli nebesno-golubye
‘they were sky blue’). By using the phrase “I remembered”, immediately
before the second instance, the narrator justifies the transposition into a
different time which allows him to present the quality of eye color as sta-
tive, despite the change. However, the features of kindness and gullibility
are presented as changed, and consequently the message is that Venya no
longer believed everything he heard or read. The speaker is thus able to
interweave two different narrative times.
While future tense examples are not discussed here in the same detail
due to their infrequency (relative to past tense examples), it is worthwhile
to mention an example discussed in Zel'dovič (2005: 141). The author
states that (19a) rather than (19b) suggests that the speaker intends to put
the sweater on or empathizes with someone who intends to wear the
sweater due to the observation factor:
(19) a. Zavtra sviter budet suxoj.

tomorrow sweater be.3SG.FUT dry.M.SG.NOM
b. Zavtra sviter budet suxim.

tomorrow sweater be.3SG.FUT dry.M.SG.INSTR
‘Tomorrow the sweater will be dry.’
I believe that the fact that the speaker (or someone else) is planning to
wear the sweater tomorrow makes the process of drying and implicit com-
32 Alina Israeli
parison between the current wet and future dry state of the sweater immate-
rial; what matters most is the state it is in when worn.
Conversely, in (20) it is the result of drying that is important, and thus
implicit comparison is brought to the fore:
(20) Naprimer, v moroznuju pogodu vo vremja progulki

for-example in frosty weather in time of-walk
povesit' na ulice mokroe polotence i posmotret', čto
to-hang on street wet towel and to-look what
s nim budet. Na sledujuščee utro … voda zamerzla
with him be.3SG.FUT on next morning water froze
i prevratilas' v led. No k večeru polotence budet
and turned in ice but to evening towel be.3SG.FUT
suxim – ne budet ni l'da ni vody…
dry.M.SG.INSTR not be.3SG.FUT neither ice neither water
(http://www.aseko.org/bull/08.htm)
‘For example, in frosty weather during a walk, hang a wet towel out-
side and see what happens to it. The next morning … the water will
have frozen and turned into ice. But by evening the towel will be
dry, there will be neither ice nor water.’
There exist nouns and attributes that cannot pertain to any particular
moment in time, but only to a period, yet the features that they describe
cannot be considered permanent and described statively. This is the case
with the nouns ‘searches’ and ‘influence’ in (21) below:
(21) a. Moj oxotničij azart usugubilsja, vidimo,

my hunter.ADJ excitement aggravated apparently
tem, čto dolgie poiski byli besplodnymi
by-that that long searches were fruitless.PL.INSTR
i ja daže terjal nadeždu.
and I even lost hope
(V. Solouxin. Trava)
‘My hunter’s excitement was aggravated, apparently by the fact
that the lengthy searches had been fruitless and I was even los-
ing hope.’
b. Vozdejstvie ital'janskogo futurizma bylo ves'ma

influence Italian futurism was.N quite
značitel'nym, osobenno na russkij avangardizm.
remarkable.N.SG.INSTR especially on Russian avant-garde
(E. Xersonskaja. Možno tol'ko stroit' gipotezy.“Znanie – sila”,
1989: 01)
‘The influence of Italian futurism was highly significant, espe-
cially on the Russian avant-garde.’
3. Non-temporal comparison
The use of the instrumental may signal not only a comparison between two
different states of a single entity at two different points in time, but also a
comparison between two different entities, or different parts of the same
entity. In (22), the author compares two generations, and his verdict (using
the instrumental) is that they are different, not alike:
(22) a. Oni uvereny v sebe točno tak že kak byli

they sure in self exactly thus PCL as were
uvereny my, no … ne tak rvutsja k komandnym
sure we but not thus strive to command
vysotam, kak ljudi našego pokolenija. … Èti
heights as people our generation these
byli, požaluj, drugimi.
were MODAL others.PL.INSTR
(L. Zorin. Krapivnica)
‘They are sure of themselves just as much as we were, but …
are not striving for the leadership positions as did the people of
our generation. … Come to think of it, these were different.’
b. Eti byli drugie.

these were others.PL.NOM
‘These were others.’
In (22a) Zorin is comparing the internal qualities of two generations that

in his opinion are different. Had he used (22b) instead, he would have sim-
ply implied that those were different people in the sense of being different
tokens, not the same entity.
34 Alina Israeli
The nominative use explicitly avoids comparison. The pseudo-

comparison in (23) metaphorically explains that any comparison between
the two women is as impossible as would be a comparison between a dog
and a goat, whose only commonality, in the speaker’s eyes, resides in the
fact that they both have four legs and a tail.
(23) Lora i Tanja byli raznye, kak naprimer

Lora and Tanya were.PL different.PL.NOM as for-example
sobaka i koza. Oni čem-to poxoži: primerno
dog and goat they something alike approximately
odinakovoj vysoty, obe na četyrex nogax i s
same height both on four legs and with
xvostom.
tail
(V. Tokareva. Odin kubik nadeždy)
‘Lora and Tanya were different, as for example a dog and a goat.
They have something in common: they are about the same height,
and both have four legs and a tail.’
4. Limitation
Zel'dovič (2005: 132), following Timberlake (1985: 278–282) points out

that the instrumental signals temporal limitation. Some instances of time
comparison could be viewed as time limitation. Yet there are clear in-
stances of time limitation where there is no explicit comparison (discussed
in 4.2).
Zel'dovič (2005: 132) also mentions that the nominative form interprets
the situation as isolated, “by itself” without correlation with anything else.
In other words, the instrumental signals limitation in scope.
4.1. Limitation in scope
Limitation in scope refers to a feature either not being consistent, not mani-
festing itself constantly, apparent only within a given frame of reference, or
not being assumed to have impact beyond the speaker. Consider (24), in
which an intelligent person behaves stupidly, thus setting a limit to his wit.
(24) On byl umnym, no vel sebja glupo.

he was.M smart.M.SG.INSTR but conducted self stupidly
On stesnjalsja, i čtoby skompensirovat' èto,
he was-bashful and in-order to-compensate this
pozvoljal sebe lišnee i inogda vel sebja
allowed self extra and sometimes conducted self
sliškom agressivno.
too aggressively
(cooler.irk.ru/hackers/hackers-69.html)
‘He was smart, but acted stupid. He was bashful and in order to
compensate for this went overboard and sometimes behaved too ag-
gressively.’
Other examples of limitation in scope are given in (25):
(25) a. Pervyj ètaž byl kirpičnym, vtoroj i tretij

first floor was.M brick.ADJ.M.SG.INSTR second and third
derevjannye.
wooden.PL.NOM
(Dombaj. Info)
‘The first floor was made of brick, the second and third ones
were made of wood.’
b. Dom naš dlja Tuly byl svetskim.

house our for Tula was.M worldly.ADJ.M.SG.INSTR
(Knjaz' G. E. L'vov i ego tul'skoe okruženie)
‘For Tula, our house was worldly.’
In (25a) the quality of being made of brick is limited to one floor (as
opposed to the rest of the house which is made of wood); in (25b), the limi-
tation is geographic (i.e. restricted to Tula) as well as in scope: the author
deliberately states the limit of their house’s worldliness, which in other
places, e.g. St. Petersburg, might not have been considered worldly at all.
Going back to (4a) and (4b), repeated below, we can see that in (4a) it is
the grandfather’s oldness and its ramifications that are the speaker’s con-
cern, while in (4b), only the implication of the oldness with respect to his
remembering the Morse code is of note:
36 Alina Israeli
(4) a. Da, ded byl staryj i xlopot

yes grandfather was.M old.M.SG.NOM and hassles
s nim ne men'še, čem s malym rebenkom.
with him not less than with small child
(zhurnal.lib.ru/b/borzow_a_a/letter.shtml)
‘Yes, grandfather was old and caused no less hassle than a
small child.’
b. No ded byl starym, i azbuku Morze

but grandfather was.M old.M.SG.INSTR and alphabet Morse
uže zabyl …
already forgot.M.SG
(zhurnal.lib.ru/c/cwirk_a/provbud.shtml)
‘But grandfather was old and had already forgotten the Morse
code …’
4.2. Limitation in time
As was mentioned earlier, following Timberlake (1982) and Zel'dovič

(2005), it is possible to speak of limitation in time in isolated instances
such as (26), which – according to Zel'dovič (2005: 132) – implies that the
feature is no longer present:
(26) On byl krasivym.

he was.M handsome.M.SG.INSTR
‘He used to be handsome.’
It is interesting, however, to examine parallel examples to compare

when the narrator chooses the device of presenting an event as if being
observed versus when he chooses to present the event as limited in time.
Example (27) is a flashback to a time the speaker explicitly remembers,
and even though this state does not exist anymore, the narrator describes it
as if observed; in (28) below, where the instrumental is used, the desig-
nated state of loving is construed as experienced by the sentential subject
(the girlfriend) only at those limited moments when the object of love (the
rabbit) was cheerful:
(27) Da, no ja pomnju, kogda on byl veselyj

yes but I remember when he was.M merry.M.SG.NOM
i bespečnyj.
and carefree.M.SG.NOM
(È. Xeminguèj [Hemingway]. Ostrova v okeane)
‘Yes, but I remember when he was cheerful and carefree.’
(28) Ego podružka, Pušanka, ljubila krolika Veselogo, kogda on byl

his girlfriend Pushanka loved rabbit Vesely when he was.M
veselym, i ne ljubila ego grustnym.
merry.M.SG.INSTR and not loved him sad. M.SG.INSTR
(A. Magidovič Obrečennost'. www.proza.ru/texts/2005/07/31-
204.html)
‘His girlfriend, Pushanka, loved the rabbit Vesely when he was
cheerful, and did not love him when he was sad.’
5. Focalization
Focalization is understood here in the sense of Genette (1972) and

Moeschler (1996). For Genette, focalization revolves around the relation-
ship between the narrator and the protagonists, whether the narrator is part
of the narrative or just an outside voice, in which case he may focus his
attention on one character or intermittently on a variety of characters.
In some contexts, an object or a person described with a predicative fea-
ture correlates in some way with other participants of the narrated event.
The object or person in question could be seen through the eyes of or have
some relation to another participant. I suggest that the nominative repre-
sents focalization on the object or person described by the nominative
predicate, while the instrumental represents focalization on some other
participant(s). Zel'dovič (2005: 144) clearly equates “observation” with
“involvement”. I believe the opposite is true: the speaker’s involvement as
a participant of a narrated event is clearly in opposition to his remaining in
the role of a non-participant observer.
In (29) the speaker is a non-involved observer looking at archival pho-
tographs, whereas in (30) the speaker is a participant of the narrated event.
(29) Pervyj fotosnimok ... Den' byl solnečnyj, no,

first photograph day was.M sunny.M.SG.NOM but
po-vidimomu, dostatočno proxladnyj, o čem možno sudit'
apparently rather cool about what possible to-judge
38 Alina Israeli
po odežde prisutstvujuščix.
by clothes present-people
(L.P. Korsakov, T. V. Kirpičenko. Redkie fotosnimki…
http://www.fessl.ru/publish/grodek/kor.shtml)
‘First snapshot… The day was sunny, but apparently rather cool,
judging by the clothing of those present.’
(30) I vot nakonec ja vyzvan na tak nazyvaemuju “mandatnuju”

and here finally I called on so called mandate
komissiju, … Den' byl solnečnym. Po pogode
commmittee day was.M sunny.M.SG.INSTR by weather
bylo i nastroenie vsex prisutstvujuščix na komissii, ...
was and mood of-all present-people on committee
(K. Veter. Orbity ispytatelja “gestapo” ja vyderžal...)
‘And here finally I am summoned by the so-called “mandate” com-
mittee, … The day was sunny. The committee members’ mood was
in accord with the weather.’
An involved observer (the instrumental use) either filters the informa-

tion, as in (31a) below where the grandfather’s kindness is known through
Leka’s perception as far as it is related to Leka, or implies the relevance of
the feature for the speaker or for the character who is the focus of the nar-
ration. This is the case in example (31b), where the emphasis is not just on
the quality of the letter, but on its impact:
(31) a. Deduška byl dobrym. Leka znal točno.

grandfather was.M kind.M.SG.INSTR Leka knew exactly
(A. Lixanov. Zvezdy v sentjabre)
‘Grandfather was kind. Leka knew this for a fact.’
b. Pis'mo bylo strannym, polnym zagadočnyx

letter was.N strange.N.SG.INSTR full of-enigmatic
namekov, ot nix stanovilos' trevožno.
hints from them became alarmingly
(I. Gerasimov. Otkrovennye proisšestvija/ Gustavsson-148)
‘The letter was strange, full of enigmatic hints; one became
alarmed by them.’
Information that is not filtered, as in (32a), and has no impact on the

speaker, as is the case with a letter in (32b), is given in the nominative:
(32) a. Vse ego očen' uvažali i ljubili, potomu čto on byl

all him very respected and loved because he was.M
dobryj i loxmatyj.
kind.M.SG.NOM and disheveled.M.SG.NOM
(V strane černogo i loxmatogo)
‘Everyone respected and loved him very much, because he was
kind and disheveled.’
b. Deneg, konečno, nikakix ne prišlo. Da i samo

money of-course no-kind not came PCL and itself
pis'mo bylo strannoe, otryvistoe, s erničeskimi
letter was.N strange.N.SG.NOM abrupt with jerky
stixami i pustjakovymi novostjami. A v konce
poems and trifle news and in end
posle podpisi, – postskriptum. Teper' nikto i
after signature post-scriptum now no-one and
ničego uže bolee ne uznaet. Da i nadobnosti
nothing already more not find-out PCL and need
osoboj net.
particular isn’t
(S. Vitvickij. Dvadcat' sed'maja teorema ètiki)
‘Of course, no money was sent. And even the letter itself was
strange and abrupt, with ridiculous poems and trifling news.
And at the end, after the signature, there was a postscript. Now
no one will find out anything anymore. And there is no special
need to.’
Similarly, the size of an object may either have an impact on the

speaker or not. In examples (33)–(34), all of which contain the phrase dom
byl bol'šoj / dom byl bol'šim ‘the house was big’, the narrators describe
either their own house/home or the one where they lived, i.e. they were
always involved with the house/home in question in some way. In (33), the
house is described objectively, that is its size and another attributes, or
features, by which the reader can identify the type of the house is given
(dvuxètažnyj ‘two-story’):
(33) Veter perestal, metelica končilas' – i my nakonec,

wind stopped blizzard ended and we finally
stukotja (sic) zubami, obsypannye snegom, vošli v dom.
chattering teeth sprinkled with-snow entered in house
40 Alina Israeli
A dom byl bol'šoj, dvuxètažnyj – znaete

and house was.M big.M.SG.NOM two-storied.M.SG.NOM you-know
èti severnye prostornye doma? – xozjajka naša, tetka
these northern spacious houses landlady our aunt
moego druga, spala vnizu, my raspolagalis' na
of-my friend slept downstairs we settled on
vtorom ètaže.
second floor
(Ju. Bondarev. Seans spiritizma)
‘The wind subsided, the blizzard ended, and finally with teeth chat-
tering and covered with snow we entered the house. And the house
was big, two-story – are you familiar with such spacious northern
houses? – our landlady, the aunt of my friend, slept downstairs, and
we had the second floor.’
In contrast, the example in (34) does not describe the house in objective
terms, but rather subjectively, and accordingly the case used is the instru-
mental.
(34) Ja edu i ne udivljajus' ničemu. Edu 25 let žizni,

I travel and not surprised at-nothing I-travel 25 years of-life
vse vremja k odnoj svetloj točke. Dom byl bol'šim,
all time to one light point house was.M big.M.SG.INSTR
nesmotrja na to, čto nazyvalsja kvartiroj, čto vokrug
despite on that that called apartment that around
bylo sovetskoe vremja i čto vremeni ne bylo voobšče,
was Soviet time and that time not was in-general
a bylo tol'ko bol'šoe prostranstvo Doma. Doma byl
and was only big space of-house at-home was
ja.
I
(http://www.kulichki.com/moshkow/akm/txt/kelt/otschet.html)
‘I am traveling and I am not surprised at anything. I’ve been travel-
ing for 25 years of my life, always towards the same point of light.
The House was big, even though it was called an apartment, and
even though around us was Soviet time and there was no time in
general but there was only the large space of the House. It was me
who was home.’
In (35)–(36) with the predicate vynuždennyj ‘compelled’, whenever the

nominative case is used the motivation for the action came from within the
participant in question; in (35a) the composer Bul-bul ogly decides to give
a concert to inaugurate his star (similar to those on Hollywood Boulevard),
in (35b) the famous actress Doronina explains her motivation to leave one
theater and to go to another. She was not forced to leave but felt compelled
to do so at one point:
(35) a. Znaete, tot koncert byl vynuždennyj. Vot

you-know that concert was.M compelled.M.SG.NOM here
rešili otkryt' moju zvezdu, a ja rešil ne
decided to-open my star and I decided not
otkazyvat'sja. Tam nikogo net s azerbajdžanskoj
to-refuse there nobody isn’t with Azerbaijani
familiej, a ja podumal, čto xorošo, esli by ona
name and I thought that good if COND she
pojavilas'. No dlja togo, čtoby otkryt' zvezdu, nado
appeared but for that in-order to-open star must
dat' koncert.
to-give concert
(Polad Bjul'-Bjul' ogly: Ne bojtes', ja s vami!)
‘You know, that concert was forced on me. They decided to in-
augurate my star, and I decided not to turn this down. There is
no one there with an Azerbaijani last name, and I thought that it
would be good if such a name appeared. But in order to inaugu-
rate one’s star one has to give a concert.’
b. Perexod v MXAT byl vynuždennyj…

transfer in MXAT was.M compelled.M.SG.NOM
(www.smotr.ru/inter/inter_doronina_0606.htm)
‘The transfer to MXAT was forced…’
In contrast to the choice of nominative in the above examples, the use of

the instrumental in (36) is triggered by the fact that the source of the com-
pelling force is an outside event or person – another player’s injury or the
government’s decisions – and hence the event is not focalized on the pro-
tagonist:
42 Alina Israeli
(36) a. Ego vyxod na pole byl vynuždennym, poskol'ku

his exit on field was.M compelled.M.SG.INSTR since
travmirovalsja Ašvetija.
injured Ashvetiya
(www.korrespondent.net/main/77081)
‘His going out on the field was forced since Ashvetiya got in-
jured…’
b. Vyezd byl vynuždennym dlja bol'šinstva

departure was.M compelled.M.SG.INSTR for majority
pokinuvšix rodinu ljudej, no otnjud' ne byl
left homeland people but at-all not was
vynuždennym istoričeski: v značitel'noj stepeni rešenija
compelled historically in significant degree decisions
èmigrantov opredeljala politika stalinskogo rukovodstva.
of-emigrants decided politics of-stalinist leadership
(http://www.hro.org/editions/press/0702/05/05070222.htm)
‘The departure was forced for the majority of those people who
left the homeland, but it was certainly not forced historically: to
a significant degree the decisions of the emigrants were trig-
gered by the policies of the Stalinist government.’
Another instance of focalization vs. non-focalization can be observed

with members of a group. If a person is viewed as part of a group, as one of
them, he is described with the instrumental, while if he is viewed as stand-
ing out from the group, with focalization on him as an individual, he is
described with the nominative. The larger context of (37) describes the
punishment that the first person suffers while no one else in the group
does. The group is described with my ‘we’, while the speaker ja ‘I’, being
the first is juxtaposed to everyone else with the conjunction a, which ex-
plains why he was punished.
(37) My begali po koridoru, i tam byla takaja komnata

we ran along hallway and there was such room
special'naja, gde prinimali v pionery, my tuda
special where accepted in pioneers we there
vorvalis'. S šumom otkryli dver', a ja byl pervyj.

rushed-in with noise opened door and I was.M first.M.SG.NOM
(www.newsinfo.ru/?a=radio&sa=view_new&id=61730&date=2004-
05-19)
‘We ran along the hallway, and there was a special room there where
the induction into the Pioneers took place, and we broke into that
room. We opened the door loudly, and I was the first.’
The nominative, as in (37), implies that the narrator is entirely focused

on himself, and is not being conceptualized as part of the group. In (38),
the instrumental expresses the focus on membership in a group. The
speaker’s fate was exactly the same as that of all other members of the
group, except that he was the first:
(38) Rovno v 09.00 nas nakonec-to zapustili v zdanie i

exactly in 9:00 us finally ushered in building and
preprovodili po auditorijam, pričem každuju gruppu
led to classrooms meanwhile each group
iz gde-to sta čelovek razdeljali na čelovek
from somewhere hundred people divided into persons
desjat' i raspredeljali ix uže po auditorijam. Ja
ten and distributed them already to classrooms I
byl v pervoj desjatke, malo togo, ja byl pervym!
was in first ten little of-that I was.M first.M.SG.INSTR
(http://www.dubinushka.ru/template.php?pod2rID=4)
‘At exactly 9:00 am we were finally let into the building and ushered
into the respective classrooms, meanwhile each group of somewhere
around a hundred was divided into groups of ten and then already
distributed into classrooms. I was in the first ten, moreover, I was
first!’
Examples (39)–(40) demonstrate the conceptual distinction between ja

byl staršij and ja byl staršim, both of which mean ‘I was the oldest’. Here
again the instrumental emphasizes the comparison, the relationship within
the group, while the nominative provides focalization on the individual.
(39) a. Kogda ja rodilsja, emu bylo 66 let, materi že

when I was-born he was 66 years mother PCL
moej – vsego 27–28. Ja byl staršim iz
my only 27–28 I was.M oldest.M.SG.INSTR from
44 Alina Israeli
detej. Posle menja rodilsja moj brat Boris,

children after me was-born my brother Boris
vsled za nim Aleksandr, kotoryj vskore umer, a
following after him Alexander who soon died and
poslednim brat Lev. Otec moj umer … Vsled za
last brother Lev father my died following after
otcom čerez neskol'ko mesjacev umerla ot čaxotki
father after several months died from consumption
i mat', i nas, vsex trex brat'ev vzjala na
and mother and us all three brothers took on
vospitanie naša tetka …
upbringing our aunt
(A.A. Brusilov. Vospominanija)
‘When I was born, he was 66 years old, and my mother only 27
or 28. I was the oldest of the children. After me, my brother Bo-
ris was born followed by Alexander who soon died, and brother
Lev was the last one. My father died … My mother afflicted
with consumption followed him to the grave a few months
later, and we, all three brothers were taken in by our aunt … ’
b. V našej sem'e bylo šestero detej, iz kotoryx

in our family was six children from which
dvoe umerli. Papa rabotal, mama byla domoxozjajkoj.
two died dad worked mama was housewife
Žili, kak i vse posle vojny, bedno, trudno. Ja
lived as and all after war poorly in-hardship I
byl staršim iz detej no roditeli
was.M oldest.M.SG.INSTR from children but parents
sumeli dat' mne vozmožnost' okončit' desjat' klassov.
managed to-give me possibility finish ten classes
(Optimist. http://drama.krasnoyarsk.ru/sangadgieva1.php3)
‘There were six children in our family, two of whom died. Dad
worked, mother was a housewife. We lived like everyone else
after the war, in poverty, with difficulties. I was the oldest of
the children, but my parents managed to let me finish high
school.’
In contrast to (39a), where the fate of the oldest child is the same as that
of the others, in (40) the elder is assuming responsibilities similar to those
of his parents; he is focalized in contrast to the other children. If we com-
pare the types of consequences of being the oldest in the contexts in which
the instrumental and the nominative are respectively employed, the instru-
mental identifies the oldest child as simply one of the children, as part of
the group, whereas the nominative sets him apart. So it is despite being the
oldest that the narrator gets an education in (39b), whereas in (40) it is
because of being the oldest that he assumes responsibilities closer to those
of his parents: he must take care of the other children, wash floors, acquire
a profession, and work along with his parents.
(40) a. Ja rodilsja v krest'janskoj sem'e. Otec i mat'

I was-born in peasant family father and mother
rabotali, kak katoržniki i, naverno, poètomu
worked like chain-gang and probably therefore
xodili v serednjakax. Četveryx detej nado bylo
walked in average-peasants four children must was
prokormit'. Ja byl staršij i, skol'ko
feed I was.M oldest.M.SG.NOM and as-much
pomnju sebja, vodilsja s mladšimi – èto
I-remember self kept-company with youngers this
byla moja pomošč' roditeljam.
was my help to-parents
(http://www.kuzbass.ru/nkz/stalinsk/chykomin.htm)
‘I was born to a peasant family. Father and mother worked like
prisoners and probably therefore were middle income peasants.
Four children had to be fed. I was the eldest and as far as I re-
member myself, I hung around the little ones – this is how I
helped my parents.’
b. Žili kak vse. V kommunalke na Maksima

lived like all in communal-apartment on Maxim
Gor'kogo: mama, papa, ja i brat s sestroj.
Gorky mama dad I and brother with sister
Vse v odnoj komnate. Mne čego-to večno ne
all in one room to-me something eternally not
xvatalo. Tem bolee ja byl staršij i mne
sufficed especially I was.M oldest.M.SG.NOM and to-me
«dostavalos'» bol'še drugix. S tret'ego klassa ja i
got more others from third class I and
46 Alina Israeli
tol'ko ja myl poly. Nikakix poblažek.

only I washed floors no-kind indulgences
(Andrej Sokolov: sorevnovanie s samim soboj)
‘We lived like everyone did. In a communal apartment on
Gorky Street: Mom, Dad, me and brother and sister. All in one
room. I was always short of something. Especially since I was
the oldest and «got it» more than the others. Ever since the third
grade I and I alone washed the floors. No excuses.’
c. Kogda mne bylo 14 let, tjaželo zabolela mat'. V

when me was 14 years heavily got-ill mother in
sem'je ja byl staršij, i potomu
family I was.M oldest.M.SG.NOM and therefore
neobxodimo bylo polučat' special'nost'.
necessary was to-receive profession
(Aktivy i passivy)
‘When I was 14 years old my mother got very ill. I was the old-
est in the family and that is why it was essential for me to ac-
quire a profession.’
Timberlake (2004: 284) analyzes a similar set of examples even though

he is not concerned with the copula byt'. He states that the instrumental
“characterizes an individual in a sequence of individuals”. In other words
the instrumental does not focalize on the individual, unlike the nominative:
(41) a. Imenno Ždanov pervyj vvel massovye

specifically Zhdanov first.M.SG.NOM introduced mass
aresty kommunistov.
arrests of-Communists
‘It was specifically Zhdanov who first introduced mass arrests
of Communists.’
b. Pervym deržal venec nad nevestoj ee

first.M.SG.INSTR held wreath over bride her
brat Saša, vtorym – ja.
brother Sasha second.M.SG.INSTR I.
‘The first holding the wreath over the bride was her brother Sa-
sha, the second was me’.
Zel'dovič (2005: 146) offers (42) without any commentary:
(42) Iz klounov Karandaš byl samym

from clowns Karandash was.M most.M.SG.INSTR
veselym / samyj veselyj.
merry.M.SG.INSTR / most.M.SG.NOM merry.M.SG.NOM
‘Among the clowns Karandash was the most cheerful.’
In fact, if the speaker sets Karandash apart from other clowns, he will
use the nominative, but if the message is that Karandash was comparable to
other clowns despite being more cheerful, the speaker will use the instru-
mental.
6. Implications for lexical analysis
As was shown earlier, in the case of permanent features the use of the in-
strumental case is possible only when the object itself ceases to exist, as in
(9) repeated below:
(9) … vzorvannyj dom byl kirpičnym.

exploded house was.M brick.ADJ.M.SG.INSTR
‘the building that exploded was made of brick.’
Some attributes are “observable”, so that the speaker can witness them,
and yet subject to change, such as bol'šoj ‘big’ or molodoj ‘young’. The
natural obstacle there is the nature of things or the “world order” (in the
sense of Cooper and Ross 1975): things and people progress from being
small to getting larger, from being young to growing old, and not in the
reverse direction. Consequently it is possible to find examples such as (43)
but not (44), except in some magic fiction:
(43) a. Kogda ja byl molodym.

when I was.M young.M.SG.INSTR
‘When I was young.’
b. Kogda ja byl malen’kim.
when I was.M little.M.SG.INSTR
‘When I was little.’
48 Alina Israeli
(44) a. #Kogda ja byl starym.

when I was.M old.M.SG.INSTR
‘When I was old.’
b. #Kogda ja byl bol'šim.

when I was.M big.M.SG.INSTR
‘When I was big.’
Some attributes can be used to describe an entity objectively, as in (45),

or subjectively, as pertaining to the person who is the center of the focal-
ized description, be it the speaker himself/herself involved with the task or
a protagonist whose point of view the speaker or the narrator is presenting,
as in (46):
(45) а. Rabota byla interesnaja.

work was.F interesting.F.SG.NOM
‘The work was interesting.’
b. Zadača byla trudnaja.

problem was.F difficult.F.SG.NOM
‘The problem was difficult.’
(46) а. Rabota byla interesnoj.

work was.F interesting.F.SG.INSTR
‘The work was interesting.’
b. Zadača byla trudnoj.

problem was.F difficult.F.SG.INSTR
‘The problem was difficult.’
And then there are adjectives that express qualities that are by defini-
tion available to and “knowable” only by the experiencer. Some such quali-
ties are purely interactive and inherently relative. A case in point is nesgo-
vorčivyj ‘intractable’. In order to be described as a person who does not
easily come to agreement with others, one has to be viewed through the
prism of those who perceive him as such. As a result, we do not find any
examples with the nominative, for example on Google <30.V.2006> byl
nesgovorčivyj, only with the instrumental case byl nesgovorčivym or byla
nesgovorčivoj, whether pertaining to a more general description as in (47a)

or to a moment in time, as in (47b).
(47) a. – On vsegda byl nesgovorčivym, – ulybnulsja

he always was.M intractable.M.SG.INSTR smiled
Rlu, – no v duše on trepetnyj, kak krylyško
Rlu but in soul he trembling like wing
motyl'ka…
of-moth
(È. Selivanova. Ja, Mak i Makolej,a s nami Veter, Alli i piraty)
‘“He was always intractable,” Rlu smiled, “but in his soul he is
trembling like the wing of a moth.”’
b. Starik byl nesgovorčivym i uprjamym,

old-man was.M intractable.M.SG.INSTR and stubborn
rasstat'sja s den'gami po-xorošemu ne zaxotel.
to-separate with money willingly not wanted
(www.vesti.ee/april01/04/est7.htm)
‘The old man was intractable and stubborn and did not want to
part with money willingly.’
Adjectives similar to nesgovorčivyj express qualities that can only be

ascribed to an entity by an interacting participant, thus necessarily involv-
ing a limitation in scope and requiring the instrumental.
Similarly, some nouns, such as for example dom ‘house’, denote entities
that are concrete and at least some of whose facets are not subject to inter-
pretation, while others, such as for example čelovek ‘person’, denote enti-
ties that are concrete but can be perceived differently by different observ-
ers. Then there are abstract nouns, for example pogoda ‘weather’ or den'
‘day’, where the speaker may be either an uninvolved observer, as in (29),
or one affected by the referent of the noun, as in (30). And finally there are
nouns representing events such as vosxoždenie ‘ascent’, that cannot exist
without an experiencer and thus require the instrumental:
(48) a. Vpečatlenija učastnikov vosxoždenija: vosxoždenie

impressions of-participants of-ascent ascent
50 Alina Israeli
bylo trudnym, složnym.

was.N difficult.N.SG.INSTR complicated.N.SG.INSTR
(www.mservice.ru/news/sub_club.htm)
‘The impressions of the participants in the ascent: the ascent
was difficult and complicated.’
b. Pogoda byla kak na zakaz, i vosxoždenie bylo ne

weather was like on order and ascent was.N not
složnym.
complicated.N.SG.INSTR
(www.perekrestok.az/russian/features2001/3107.htm)
‘The weather was made to order and the ascent was not compli-
cated.’
Among the examples with vosxoždenie found on the Internet, there are a
large number with the instrumental (24 with trudnym, 4 with očen' trud-
nym, 3 with dolgim, 26 with složnym, 24 with nesložnym, and so on). There
is only one example with the nominative:
(49) Čitaja knigu umom ja ponimal, čto vosxoždenie

reading book with-mind I understood that ascent
bylo nezaurjadnoe; no malo li krutyx
was.N unusual.N.SG.NOM but few Q-PCL steep
sten, nepristupnyx v ponimanii predyduščix pokolenij,
walls unaccessible in understanding of-previous generations
v konce koncov pokorjalos' čeloveku.
in end of-ends submit to-person
(S. Šibaev. Stena.)
‘Reading the book I understood in my mind that the ascent was ex-
ceptional, but there are so many steep walls that are inaccessible in
the understandings of previous generations that finally yielded to
man.’
This example involves a situation in which the author’s perception of

the ascent is a mental one, that is he is trying to take an objective view of
someone else’s exploits. He did not experience the ascent directly and only
reads about it in the book; the reading experience is an indirect experience
and thus allows the “objectivity” of the nominative.
In other words, adjectives range from those that describe objective

qualities of a noun (material, absolute size, absolute age and so on) to those
that describe strictly relative, subjective qualities of a noun (impression-
istic qualities, individually perceived qualities); there are also adjectives
that are in the middle, so to speak, that may represent relative qualities,
qualities limited to the speaker or to the speaker’s world. Similarly, entities
denoted by nouns range from those that are concrete, may have a large
number of features viewed as permanent, and lend themselves to objective
description, to those that exist only as part of one’s individual experience,
thus having only subjective features associated with them; there are also
nouns in the middle whose referents may have some objective features and
some individually perceived features. Those nouns and adjectives that per-
tain to the more objective end of the spectrum are more likely to be de-
scribed with the nominative case of the adjectival predicate; those nouns
and adjectives that belong to the individually experienced end of the spec-
trum are more likely to be described with the instrumental case of the ad-
jectival predicate.
7. Conclusion
The choice between the nominative and the instrumental case of adjectival
predicates is governed by a number of factors. The ability of the speaker to
represent events as or as if observed triggers the nominative of the predi-
cate; however, if the speaker is a participant or if he focalizes the narration
on one of the participants other than the one described with the adjectival
predicate, the use of the nominative is blocked. In addition to focalization
away from the participant described, time comparison and time limitation
trigger the instrumental.
This is an area where grammar and lexicon interact. Adjectives repre-
senting permanent features are more likely to be used in the nominative;
the instrumental is likely to be used when the object or person (and thus the
feature) no longer exist. Properties expressed by adjectives used with ab-
stract nouns designating events cannot be observed at a particular moment
in time; adjectives that describe qualities that affect the participant are
more likely to be used in the instrumental.
52 Alina Israeli
Notes
1. The following abbreviations are used in the glosses: M – masculine, F – femi-

nine, N – neuter, SG – singular, PL – plural, NOM – nominative, INSTR – in-
strumental, ADJ – adjective, 1 – first person, 3 – third person, IMP – impera-
tive, FUT – future, PCL – particle, Q – question, COND – conditional
particle.
References
Cooper, William E., and John Robert Ross

1975 World order. In Papers from Parasession on Functionalism, Robin E.
Grossman, L. James San, and Timothy J. Vance (eds.), 63–111. Chi-
cago: Chicago Linguistic Society.
Černov, V. I.
1983 Funkcional'nye i semantičeskie svojstva predikativnyx prilagatel'nyx v
forme imenitel'nogo i tvoritel'nogo padeža. [Functional and semantic
properties of predicate adjectives in nominative and instrumental
case.] Russkij jazyk v škole [Russian Language in School] 6: 90–96.
Filip, Hana
2001 The semantics of case in Russian secondary predication. In SALT XI
Proceedings, Rachel Hastings, Brendan Jackson, and Zsofia Zvolen-
sky (eds.), 192–211. Ithaca: CLC Publications.
Gasparov, B. M.
1996 Jazyk, pamjat', obraz: lingvistika jazykovogo suščestvovanija [Lan-
guage, memory, and image: the linguistics of language function]. Mos-
cow: Novoe literaturnoe obozrenie [New Literary Review].
Genette, Gérard
1972 Figures III. Paris: Éditions du Seuil.
Ionin, Tania, and Ora Matushansky
2002 Encasing the Time: Temporal Effects on Russian Predicative Case.
http://www-rcf.usc.edu/~ionin/MyDownloads/NELS_handout.pdf
Mel'čuk, Igor' A.
1985 Semantičeskie ètjudy I: ‘SEJČAS’ i ‘TEPER'’ v russkom jazyke [Seman-
tic studies I: ‘SEJČAS’ and ‘TEPER'’ in the Russian language]. Russian
Linguistics 9: 257–279.
Moeschler, Jacques.
1996 Time in evolving reference: Temporal sequencing, tenses and focalisa-
tion. http://www.unige.ch/lettres/linguistique/moeschler/
publication_pdf/time_evolving_reference.pdf
Nichols, Johanna
1981 Predicate Nominals: A Partial Surface Syntax of Russian. Berkeley:
University of California Press.
1985 Padežnye varianty predikativnyx imen i ix otraženie v russkoj gram-
matike [Case variants of predicate names and their reflection in Rus-
sian grammar]. In Novoe v zarubežnoj lingvistike XV [New Research
in Foreign Linguistics XV], T. V. Bulygina and A. E. Kibrik (eds.),
342–387. Moscow: Progress.
Timberlake, Alan
1982 Invariance and the syntax of Russian aspect. In Tense-Aspect: Between
Semantics and Pragmatics, Paul J. Hopper (ed.), 305–331. Amster-
dam/ Philadelphia: John Benjamins.
1985 Invariantnost' i sintaksičeskie svojstva vida v russkom jazyke [Invari-
ance and syntactic properties of aspect in the Russian language]. In
Novoe v zarubežnoj lingvistike XV [New Research in Foreign Linguis-
tics XV], 261–285. Moscow: Progress.
2004 A Reference Grammar of Russian. Cambridge: Cambridge University
Press.
Zel'dovič, G. M.
2005 Russkoe predikativnoe imja [Russian predicative name]. Toruń:
Wydawnictwo Uniwersytetu Mikołaja Kopernika.
Why double marking in the Macedonian dativus
sympatheticus?
Liljana Mitkovska
Abstract
This paper discusses the possible reasons for and implications of marking
possession explicitly when it is already marked by dativus sympatheticus
(DS) in Macedonian. The term ‘dativus sympatheticus’ denotes a dative
construction that implies a possessive relationship between the dative refer-
ent and another participant in the situation. The DS/explicit possessive
variation illuminates the specific construal that each of the two structures
imposes on the possessive relation expressed. Since the dative object al-
ready implies possession, the introduction of an explicit possessive marker
in the DS construction has been deemed as superfluous and anomalous.1
However, such constructions have been attested frequently, both in Mace-
donian and crosslinguistically.2 It is argued here that the double marking of
possession is motivated pragmatically, i.e. by the need of the speaker to pre-
sent the possessive relationship from the perspective of the possessor (thus
the dative object is chosen), and at the same time to highlight the possessed
(which triggers the possessive pronoun). It will be demonstrated that the es-
tablishment of the construction is supported by structural and semantic fac-
tors: with certain verbs that require a dative object the pragmatic goal to put
the focus on the possessed results in double marking, as the omission of the
dative either produces an ungrammatical clause or affects the meaning con-
siderably. The application of such doubling to contexts that would equally
accept either DS or a possessive pronoun produces an effect that is clearly
different from the meanings of either of the structures taken separately.
Keywords: dative/indirect object, possession, affectedness, construal,

communicative function.
56 Liljana Mitkovska
1. Introduction
This paper discusses the possibility of marking possession explicitly in the

indirect object construction in Macedonian when that construction already
has a possessive meaning, as illustrated in example (1). The main conten-
tion rests upon the tenet that the possessive indirect object functions as a
grammaticalized construction which conveys a specific possessive sense
and thus it does not have to be in collision with the possessive pronoun.
(1) Ti si mi ja zemal mojata

you.SG be.2SG IOCL.1SG DOCL.3SG.F take.PAST my.DEF
kniga.
book.F.SG
‘You have taken my book.’
The starting point for the study of this phenomenon is the Cognitive
Linguistic premise that people are able to construe a situation in various
ways and that different grammatical structures are usually available to en-
code the same situation (Langacker 1991). In particular, two significant
theoretical assumptions underlie the present analysis.
First, it is presumed, in accord with Cognitive Linguistics (Lakoff 1987;
Langacker 1991 and others), that language units are symbolic representa-
tions of image schemas at various degrees of abstraction associated with
the usage events they categorize. Used in a variety of contexts, these sche-
mas must often be expanded and modified to accommodate the usage
events that do not fully match the conventional specification, which gives
rise to new conventional units that constitute an expansion from more basic
ones. As a result, grammatical categories are complex, containing a net-
work of nodes united by a prototype principle. Consequently, the indirect
object construction in Macedonian is viewed as a polysemous category
exhibiting a number of related meanings (Rudzka-Ostyn 1996). The most
central among the meanings are the recipient of concrete and abstract ob-
jects and the recipient of effects (experiencer). ‘Possession’ is also under-
stood as a complex notion, involving a broad array of relationships clus-
tered around the prototypes: ownership, part whole, and kinship
relationships (Langacker 1991 and 2000; Topolinjska 1997).
The second assumption concerns the theory of grammaticalization. Ac-
cording to some findings (Sweetser 1990; Heine et al. 1991; Hopper and
Traugott 1993) a contextually inferred meaning can be spread to other con-
texts by analogy. What starts as a contextual and conversational implica-
Why double marking in the Macedonian dativus sympatheticus? 57
ture can be conventionalized by recurrent usage and give rise to new mean-
ings through contextual manipulation and metaphor. The implication of
possessive relationships in an IO construction is no surprise since the da-
tive and genitive are conceptually related, the former expressing transfer of
an object and the latter the locus of the resultant state. Thus a reanalysis in
appropriate contexts can be imagined. In fact, the conceptual transfer from
dative to genitive has been attested in many languages (Heine et al. 1991:
167).
The paper is organized as follows. In the second section the so called
dativus sympatheticus is defined in comparison with explicit possessive NP
constructions. Then the syntactic, semantic and pragmatic conditions under
which double marking occurs are examined, followed by a short analysis of
analogous developments in reflexive situations. Finally a short conclusion
is presented.
2. Dativus sympatheticus vs. possessive proper
The term dativus sympatheticus (DS) covers constructions in which the

dative (i.e. indirect object (IO))3 referent is interpreted as a possessor of
another participant in the predication, which may be expressed in the direct
object (2), intransitive subject (with both derived, passive and reflexive,
and lexical intransitives) (3), prepositional object (4) or some other syntac-
tic position. This construction is common in many Slavic languages, Ma-
cedonian not being an exception (Mitkovska 2000).
(2) Na Mirko mu ja zapalile

to Mirko.IO.3SG.M IOCL.3SG.M DOCL.3SG.F set.PAST on fire
kolata
car.DO.3F.DEF
‘They have set Mirko’s car on fire.’
(3) Ochite im svetea od radost.

eye.PL.DEF.SUB IOCL.3PL shine.PAST from joy
‘Their eyes glittered with joy.’
(4) Ana mi ja stavi

Ana. F.SG.SUB IOCL.1SG DOCL.3SG.F put.PAST
knigata vo ranecot.
book.DO.3F.DEF in backpack.DEF
‘Ana put the book in my backpack.’
Although pragmatically the dative enters into a possessive relation with

another noun, in dativus sympatheticus constructions it is syntactically
connected to the verb and does not belong to the same noun phrase as the
possessed entity. For example, the dative mu in (2) is connected syntacti-
cally to the verb zapalile. Such constructions are sometimes referred to as
external possessive constructions (Payne and Barshi 1999). A variety of
other terms are also in use in the linguistic literature4: possessive dative or
dative of possession5, dative of interest, dativus commodi/incommodi, inal-
ienable dative (Levin 1986). The traditional term dativus sympatheticus
seems less theoretically biased. Havers (1911) (see the discussion in Ku-
čanda 1996: 327) defines it in relation to six subgroups on the basis of
lexical classes of nouns or verbs that take the dative: (i) the whole body or
parts thereof; (ii) the human soul; (iii) persons or things that belong to a
person; (iv) verbs that denote acquisition or loss of possession; (v) nouns
denoting kinship terms and friendship, and (vi) verbs of motion.
What Havers describes is actually the prototype of these constructions:
situations in which the possessor is highly affected by a change of state or
place of an inalienably possessed entity. However, once established the
construction tends to extend its usage to various degrees, which results in
accommodation of its meaning to fit in those situations. Thus in Macedo-
nian the dative object can establish a possessive relation with possessed
other than body parts, as illustrated by examples (2) and (4) above, and
both close and distant alienable objects may be interpreted as possessed.
DS construction is also not limited to transitive verbs, but has come to
include other types of verbs, such as unergatives6 (example 5) and statives
(example 6), even the copula sum (be) (example 7), which do not express
affectedness per se. This has caused reinterpretation of the meaning of the
construction in those contexts. There the dative referent is viewed as the
affected party, the nature of the affectedness being subject to contextual
interpretation.
(5) Mirko ti sedna na stolčeto. (AC)

Mirko.SUB IOCL.2SG sit.PAST on chair.DEF
‘Mirko sat on your chair.’
(6) Ti ja vidov kolata na

IOCL.2SG DOCL.3SG.F see.PAST car.DO.3F.DEF in
parkingot. (AC)
carpark.DEF
‘I saw your car in the car park.’
(7) Ubava ti e frizurata. (AC)

nice IOCL.2SG be.3SG.PRES hairstyle.DEF
‘That’s a nice hairstyle you have.’
The dative in many Slavic languages has been shown to imply posses-
sive meaning. There is no consensus among linguists, however, as to the
status of DS, i.e. whether it is, actually, a separate construction. Kučanda
(1996: 330), for example, argues against it: “The distinction between sev-
eral subsets of the free dative is quite unnecessary because there is no con-
clusive syntactic and/or semantic evidence to support such a division. The
interpretation of the free dative depends on the semantics of the verb and
the meanings such as advantage, disadvantage and possession are addi-
tional layers of semantic interpretation.” Other linguists, on the other hand,
support the view that DS should be regarded as an independent construc-
tion. King (1998), rejecting the contentions of some linguists that the da-
tive of possession in Czech (‘dative of interest’ in her terminology) is only
a contextual variant of the free dative, insists that it differs functionally
from all other dative constructions. When the speaker chooses to use this
construction the speaker knows exactly what s/he wants to do, namely to
present the situation from the point of view of the possessor of the in-
volved entity. Fried (1999) also holds the view (within the Construction
Grammar framework) that there is a separate construction, which, besides
sharing properties with the other dative constructions, has its own specific
configuration. “In particular, it must be made explicit that EPRs (external
possessors) represent interest overlaid with possession and that the co-
occurrence of these two semantic features is subject to specific grammati-
cal constraints, rather than being a matter of pragmatic plausibility.” (ibid.:
494)
The analysis in this paper is based on the assumption that although for-
mally DS does not differ from other dative constructions it is associated
with a unique semantic-pragmatic construal, namely to underline the af-
fectedness of the possessor. In fact, this construal has been conventional-
ized, which is most evident in less prototypical examples with stative
verbs, where even though no real affectedness exists, the construction con-
strues the situation as affecting the possessor.7 Example (8) may be uttered
as a warning to the addressee that the situation described in the sentence
may be unpleasant for her. Example (9) suggests that the subject’s knowl-
edge of the hiding places of some people may be dangerous for those peo-
ple.8
(8) Ti se gleda postavata. (AC)

IOCL.2SG REFL.DO see.PRES slip.DEF.SUB
‘Your slip is showing.’ (It is unpleasant for you.)
(9) Im gi znaeše skrišnite mesta. (AC)

IOCL.3PL DOCL.3PLknow.PAST hiding.ADJ.DEF place.PL
‘He knew their hiding places.’ (This is not good for them.)
Possession is obviously an important component in DS. How is it re-

lated to constructions which express possession explicitly, i.e. the posses-
sive pronoun for pronominal possessors (negovata kniga – his book) and
the possessive adjective and the na-construction for nominal possessors
(Mirkovata kniga/ knigata na Mirko – Mirko’s book)? Even though they
may encode the same factual situations, they do not convey the same mean-
ing. Bally (1926/1996:55–56) was right when he claimed that the posses-
sive dative and the possessive pronoun allow the speaker of French (and
other Indo-European languages) to present a body part from two different
points of view: as an integral part of the person or as detached from him.
This viewpoint difference is metaphorically transferred when expanded to
inalienable possession and appropriately adapted according to the context.
Accordingly, DS and possessive proper are not in free variation, but
have distinct properties making it possible to present the same possessive
relation from different perspectives. Those properties can briefly be de-
scribed as follows:
(1) DS expresses ‘indivisibility’, i.e. a close connection between the
possessed and the possessor, since the possessor is presented as automati-
cally affected by whatever happens to the possessed. Thus, this construc-
tion presents a situation from the point of view of the possessor, indicating
that the event has a positive or negative effect on him or her. King (1998)
(among others) has argued convincingly that the distinct function of the
dative of possession is to express empathy with the possessor: “When the
speaker chooses the Dative over any other available construction, he ex-
presses a certain kind of knowledge. He shows that he is aware of a posses-
sive relationship between two referents, one of which is animate, and most
likely human. The speaker also knows that the referent, [A], is physically
or psychologically affected, either positively or negatively, by something
that happened to or with the other referent [B].” (ibid.: 250–251)
(2) Explicit possessive constructions, on the other hand, focus the atten-
tion on the possessed, hence the effect of detachment mentioned by Bally.
Examples (10) to (12) below illustrate this.
(10) (No vreme e da se zatvori za deneska izložbata. Kasierkata ja

zaklučuva masata, … )
‘But it’s time to close the exhibition for today. The cashier locks the
cash desk, …’
Koga izleguvaat, toj samo za malku može da

as go out.PRES.3PL he only for little can.PRES.2SG PART
gi prosledi nejzinite pravi noze
DOCL.3SG.M follow.PRES.3SG her.DEF shapely legs
‘As they go out, he can only for a moment follow her shapely legs
(što se gubat megju minuvačite.) (Izbor:74)

(which are disappearing among the passers by).’
(11) Crveniloto isčezna od negovoto lice,

blush.DEF disappear.PAST from his.DEF face
‘The blush disappeared from his face,
(no nelagodnosta sè ušte mu beše prisutna vo izrazot.) (ON:76)

(but embarrassment still showed in his expression) (lit.: but he still had
embarrassment present in his face.)’
(12) Čuvstvuvav kako se vpivaat Jelinite prsti

feel.PAST.1SG how REFL.DO press.PRES Jela’s.DEF fingers
vo mojata raka, … (Izbor:127)

in my.DEF hand
‘I felt Jela’s fingers press against my hand, …’
In example (10) the sentence pictures the legs of the woman abstracted
from the rest of her body. It resembles the image shown by a camera focus-
ing on them as she moves away. In (11) the first part of the sentence fo-
cuses on the man’s face and what happens to it. Notice how the attention
switches to the person in the second half with the dative. Explicit posses-
sive constructions are often used when the possessed is singled out by
some specific properties, as in example (10) or contrasted, as in (12).
In prototype situations the possessor is in reality also affected, but ex-
plicit possessive constructions provide information about the possessed and
what happens to it, rather than about the state of the possessor. Where the
verb does not express affectedness, this feature will not be present in the
adnominal possessive construction (as in example 10). In such situations
the affectedness can only be imposed by the dative, and the sentence with
DS may differ considerably in meaning from the one with the possessive,
as illustrated by the examples in (13). In (13a) DS ensures an interpretation
that the duke is aware and positively affected by the activity, while in (13b)
the activity is presented as independent of him. Notice that the latter may
be used in situations when the duke is not at all present, he may have been
abducted, while his boots were the only thing found and brought to the
castle.
(13) a. Slugite mu gi donesoa

servants.SUB IOCL.3SG.M DOCL.3PL bring.PAST
čizmite, …
boots.DEF.DO
‘The servants brought him his boots, …’
(Knezot se oblekuvaše pcuejќi i se rasprašuvaše što se slučuva.)

(DP:10)
‘The Duke was getting dressed while swearing and inquiring
what was going on.’
b. Slugite gi donesoa negovite čizmi, …

servants.SUB DOCL.3PL bring.PAST his.DEF boots.DO
‘The servants brought his boots, …’
To sum up, the two constructions expressing possessive relations dis-

cussed above construe a situation from different perspectives and cannot
easily replace each other in certain contexts. Each is associated with a clus-
ter of specific semantic-pragmatic features and in addition there are some-
times structural constraints.
3. Conditions for additional marking of possession in DS
The analysis so far clearly indicates that DS implies possession, albeit in

its own specific way. What motivates, then, the occurrence of the explicit
possessive marker in the same clause, such as exemplified in (1) above,
repeated here in (14)?
(14) Ti si mi ja zemal
you.SG be.2SG.PRES IOCL.1SG DOCL.3SG.F take.PAST
mojata kniga.
my.DEF book.F.SG
‘You have taken my book.’
Many scholars dealing with the dative of possession (DS) have noticed
this phenomenon, often characterizing it as awkward or not quite accept-
able. Sometimes this possibility is regarded as proof that dative and ex-
plicit possessive do not have the same function (see for example Kučanda
1996: 326). In the same vein, Janda (1993: 86) claims that: “[t]he conten-
tion that the dative expresses affectedness via possession rather than pos-
session per se is upheld by the fact that there is no rule preventing the da-
tive from co-occurring with other possessives.” On the other hand,
Velazquez-Castillo (1999: 92) maintains that such co-occurrence is “com-
pletely inadmissible” for standard Spanish, although it is possible for some
varieties of Spanish (colloquial Paraguayan Spanish, for example). She
assumes that “there must be something in the meaning of the dative that
motivates this incompatibility” and concludes that “[s]ince the dative al-
ready subsumes the concept of possession, the possessive adjective is not
necessary to establish the possessive relation.”
In Macedonian, sentences containing both the dative of possession and

the explicit possessive (mainly the possessive pronoun) can be encountered
both in spoken and in written language. There are, however, different
views regarding their status: Najceska-Sidorovska (1973) claims that the
possessive expression is necessary to ensure the possessive relation, as in
example (15a), because (15b) is not precise enough as to the ownership of
the boat. This is not quite correct, though, since the first interpretation, if
there is no counterevidence, is possessive. Moreover, the broader context
rarely allows ambiguity.
(15) a. Mu go zede negovoto kajče

IOCL.3SG.M DO.3M.DEF take.PAST his.DEF boat.DO
‘He1 took his2 boat from him2 …’
b. Mu go zede kajčeto.
IOCL.3SG.M DO.3M.DEF take.PAST boat.DEF.DO
‘He1 took his2 boat.’
Contrary to this, Minova-Gurkova (1982: 109) regards such construc-

tions as totally incorrect “because they contain unnecessary repetition of
information”. She suggests that either the dative or the possessive should
be used in one clause, but never both. Examples (16b) and (17b) illustrate
how she believes the attested ‘incorrect’ sentences (16a) and (17a) should
be corrected: in (16b) the dative is preferable, whereas in (17b) the posses-
sive is preferable. Although no explanation of the grounds for such a
judgment is given, it seems that the most probable construal for the predi-
cate-possessor-possessed combination has been chosen, i.e. in (16) the verb
se smali (‘shrink’) emphasizes affectedness which strongly suggests a da-
tive construction, whereas example (17) with stative verb sum (be) does
not have that effect. These sentences are taken out of context and that is the
reason why the affectedness in example (17) is not easily evoked.
(16) a. Ne možam da pretpostavam deka mojot čerep

not I-can.PRES PART I-imagine.PRES that my.DEF skull
mi se smalil.
IOCL.1SG REFL.DO shrunk.PAST
b. Ne možam da pretpostavam deka čerepot

not I-can.PRES PART I-imagine.PRES that skull.DEF
mi se smalil.
IOCL.1SG REFL.DO shrunk.PAST
‘I can’t imagine that my skull has shrunk.’
(17) a. Gore mu e samo ednata polovina

above IOCL.3SG.M be.3SG.PRES only one.DEF half
od negovoto telo, trupot bez noze.
of his.DEF body torso.DEF without legs.
b. Gore e samo ednata polovina od negovoto telo,

above is only one.DEF half of his.DEF body.N.SG
trupot bez noze.
torso.DEF without legs.
‘Only one half of his body is above, the torso without legs.’
These suggestions raise the question of what is actually doubled in the

sentences that contain both a dative of possession and the possessive
proper. It is argued here that it is the marking of possession that appears
twice, once in the dative and once in the possessive construction. The da-
tive retains the possessive component even though the relationship is ex-
plicitly marked in the possessive NP. This is in agreement with Kučanda’s
and Janda’s claim that the two constructions differ in the way they express
possession. Consequently, it is reasonable to assume that there are circum-
stances in which the speaker needs or chooses to use both. In what follows
I investigate the conditions under which double marking of possession
occurs in Macedonian.
First, there are contexts in which the dative is structurally required. At
that point there is overlapping between the possessive and the indirect ob-
ject function of the dative. In examples (18) and (19) the dative is part of
the semantic configuration of the predicate so it cannot be omitted if the
speaker wants to single out and stress the possessed entity by using the
possessive pronoun. Therefore the doubling here is structurally imposed.
Exactly for that reason they have an important role in setting up the pat-
tern.The dative alone would be enough to imply possession, but obviously
the speakers needed stronger emphasis. Example (18) is especially illustra-
tive. It is a slogan of an organization that helps to keep young girls away

from prostitution. The author felt the need to stress strongly the possessive
relation, although the presence of this relation is quite clear from the dative
itself (18b). The verb pripagja (belong) requires an indirect object, thus
this indirect object cannot be omitted, as it would make the clause incom-
plete (18c). The same analysis applies to the examples in (19). The sen-
tence in (19b) would be perfectly acceptable, though not evoking the same
effect as the one in (19a). On the other hand, (19c) is anomalous.
(18) a. Tvoeto telo ti pripagja tebe. (MD)

your.SG.DEF body.SUB IOCL.2SG belong.PRES you.SG.IO
‘Your body belongs to you.’
b. Teloto ti pripagja tebe.

body.DEF.SUB IOCL.2SG belong.PRES you.SG.IO
‘The body belongs to you.’
c. * Tvoeto telo pripagja.

your.SG.DEF body.SUB belong.PRES
*‘Your body belongs.’
(19) Sakam da mi dadeš prclinja

I-want.PRES PART IOCL.1SG give.PRES.2SG pigtails
‘I want you to give me pigtails
a. na dvete strani na mojata kosa. (D)

on both sides of my.DEF hair
on both sides of my hair.’
b. na dvete strani na kosata.

on both sides of hair.DEF
on both sides of the hair.’
c. * Sakam da dadeš prclinja na

I- want.PRES PART give.PRES.2SG pigtails on
dvete strani na mojata kosa.
both sides of my.DEF hair
*‘I want you to give pigtails on both sides of my hair.’
Next, there are contexts in which the dative argument is not obligatory,
but its omission may affect the meaning of the clause. In examples (20) and
(21), if the dative argument was dropped the elimination of the affected-
ness component would cause the situation to acquire a general interpreta-
tion which leaves the possessor out of the scene. The difference is also
obvious in the English translation. So in these sentences, in order to convey
the intended meaning and to emphasize the possessed, the speaker needs
both the dative and the possessive pronoun. Naturally, the dative itself will
be enough to imply possession, but that alone does not satisfy the speaker’s
wish to stress the possessed.
(20) (Jas navistina uživav igrajќi vo prodolženijata na 007 vo šeesettite, no)
‘(I really enjoyed playing in the 007 series in the sixties, but)
a. mi beše potreben nov predizvik,

IOCL.1SG be.3SG.PAST necessary new challenge,
nova nasoka vo mojata kariera. (V)
new direction in my.DEF career
I needed a new challenge, a new direction in my career.’
b. (?) beše potreben nov predizvik, nova nasoka

be.3SG.PAST necessary new challenge, new direction
vo mojata kariera.
in my.DEF career
(?) a new challenge was needed, a new direction in my career.’
(21) (Ќe se počuvstvuvate neprijatno so ogled na toa što postojano)

‘(You will feel ill at ease since constantly)
a. ќe vi se zabeležuva za
PART.FUT IOCL.2PL REFL.DO criticise.PRES for
vašeto odnesuvanje. (V)
your.PL.DEF behaviour
you will be criticized because of your behaviour.’
b. (?) ќe se zabeležuva za vašeto odnesuvanje.

PART.FUT REFL.DO criticise.PRES for your.PL.DEF behaviour
there will be criticism of your behaviour.’
Furthermore, there are contexts in which doubling of the possessive in-

formation sounds more natural, even desirable. In particular, if two pos-
sessed in a clause are related to the same dative referent, one of them is
easily modified by a possessive, as in examples (22) and (23). This reduces
the interpretation load and eliminates the potential ambiguity. It is also
much easier to add a possessive modifier if the possessed is already speci-
fied by a modifying adjective (example 24). In those situations the pos-
sessed is already individualized and receives special attention.9 Conse-
quently the possessive is a natural addition that intensifies the already
existing effect. The use of the possessive pronoun in DS is especially eas-
ily embraced when both conditions mentioned above are fulfilled, as in
example (23). However, it has to be pointed out that in Macedonian it is
not obligatory to use the possessive pronoun in such contexts, either by
itself or in combination with the dative. The speakers may choose to use
only the dative. Example (25) is quite a natural Macedonian sentence (it
was actually attested, as indicated in the parenthesis).10 It contains two
possessed belonging to the same possessor, both of which are especially
singled out with adjectival modifiers, and yet the possessive relation is
only indicated by the dative construction.
(22) Karierata rapidno mi se

career.DEF.SUB rapidly IOCL.1SG REFL.DO
uništuvaše pred moite oči. (V)
destroy.PAST in front of my.DEF eyes
‘My career was rapidly being destroyed in front of my eyes.’
(23) Negovata ubava rusa kosa sega mu

his.DEF nice blond hair.SUB now IOCL.3SG.M
pagjaše od dvete strani na liceto,
fall.PAST on both sides of face.DEF
(razdelena po temeto na ramen patec.) (ON:74)
(parted down the crown in a straight line.)
‘His nice blond hair was now falling on both sides of his face, parted
down the crown of his head in a straight line.’
(24) Temna senka mu go zamrači

dark shadow.SUB IOCL.3SG.M DOCL.3SG.M cloud.PAST
negovoto i taka potemneto lice. (N.-Sidorovska 1973:120)

his.DEF already tanned face.DO
‘A dark shadow clouded his already tanned face.’
(25) Iznemošteniot glas mu gi

faint.DEF voice.SUB IOCL..3SG.M DOCL.3PL
izdavaše duševnite maki niz koi
reveal.PAST emotional.DEF suffering.DO through which
pominal ovie dva dni. (ON:130)
go.PAST these two days
‘His faint voice revealed his torments, which he had gone through
for the last two days.’
Finally, there are constructions in which the dative and the possessive
express the same factual situation, the difference being only in emphasis
(cf. examples (26a) and (27a)). Examples (26b) and (26c) and (27b) and
(27c) show that clauses with either the dative or the possessive are per-
fectly well formed and express the same facts. However, the speaker de-
cides to use both markers, thus keeping the possessor’s perspective with
the dative, while highlighting the possessed at the same time by adding the
possessive pronoun.
(26) (Na kogo mu odgovara vo ovaa država)

‘(Who benefits in this country)
a. da ni go minira našeto
PART IOCL.1PL DOCL.3SG.M undermine.PRES our.DEF
rabotenje? (TV)
work
by undermining our work to us?’
b. da go minira našeto rabotenje?

part DOCL.3SG.M undermine.PRES our.DEF work
by undermining our work?’
c. da ni go minira rabotenjeto?
part IOCL.1PL DOCL.3SG.M undermine.PRES work.DEF
by undermining the work to us?’
(27) a. Ana, koj ti bil najgolemiot gaf vo

Ana which IOCL.2SG be.PAST biggest.DEF gaffe in
tvojata kariera kako voditelka? (V)
your.SG.DEF career as presenter
‘Ana, which was the biggest gaffe to you in your career as a pre-
senter?’
b. Ana, koj bil najgolemiot gaf vo

Ana which be.PAST biggest.DEF gaffe in
tvojata kariera kako voditelka?
your.SG.DEF career.DEF as presenter
‘Ana, which was the biggest gaffe in your career as a presenter?’
c. Ana, koj ti bil najgolemiot gaf

Ana which IOCL.2SG be.PAST biggest.DEF gaffe
vo karierata kako voditelka?
in career.DEF as presenter
‘Ana, which was to you the biggest gaffe in your career as a pre-
senter?’
The introduction of the possessive proper in DS, in contexts exempli-

fied in examples (26) and (27) above, seems to be the latest development.
One reason why this conclusion suggests itself is the fact that such con-
structions sound novel, yet some speakers reject them as erroneous, even if
they themselves might use them colloquially. Another reason that supports
this assumption is the effect this introduction produces on the functional
value of the whole construction. As a matter of fact, in the previously de-
scribed contexts, illustrated by examples (18) to (25), the presence of the
possessive in a DS construction is structurally or semantically justified and
it does not introduce a striking change in the meaning of the construction.
However, when the same structural combination is extended to contexts
which allow either DS or possessive proper without any structural or se-
mantic constraints, it produces a modified effect. Namely, when possession
is coded separately alongside DS in such contexts, the possessive compo-
nent in the dative is weakened, but the emphasis on the dative referent as
the affected party is strengthened. Such clauses have a double stress which
makes them sound excessive and intense, which presumably produces the
effect of awkwardness and incorrectness that is often associated with them.
This effect, in particular, is intentionally employed by the speaker to fulfill
certain communicative functions in notably marked emotional contexts.

For instance, the utterance in (28) serves to reproach the addressee for
wearing the speaker’s jacket. In (29) an older sister is warning the younger
one not to wear her shoes, and in (30) the mother reprimands her son for
not having his own roller blades on, but somebody else’s. In (31) the per-
son is being reassured that nothing is going to happen with his coffee mug.
(28) Blagoja, ti mi ja zanosi

Blagoja you.SUB IOCL.1SG DOCL.3SG.F start.PAST wearing
mojata jakna! (AC)
my.DEF anorak.DO
‘Blagoja, you started wearing my anorak regularly!’ (reproach)
(29) Nemoj da si mi gi
don’t.IMP PART be.2SG.PRES IOCL.1SG DOCL.3PL
oblekla moite čevli! (AC)
put.PARTC on my.DEF shoes.DO
‘Don’t you dare put on my shoes!’ (warning)
(30) Martin, kade ti se tvoite roleri? (AC)

Martin where IOCL.2SG be.3PL.PRES your.SG.DEF roller blades
‘Martin, where are your roller blades? (i.e., what have you done
with your roller blades?)’ (reprimand)
(31) Ne se sekiraj, ne ti pieme

not REFL.DO worry.IMP not IOCL.2SG drink.PRES.1SG
od tvojata šolja. (AC)
from your.SG.DEF mug
‘Don’t worry, we are not using your mug.’ (reassurance)
Generally, the analysis of the various contexts in which double marking

of possession in DS is encountered reveals different stages in the estab-
lishment of this complex syntactic construction and indicates some possi-
ble directions of its development. There are still a great number of various
effects that it evokes, contingent on the particular context. These effects
can be summed up under the general goal of the speaker to present the
possessor as affected even though it is the possessed which is in focus.
However, it remains an open question whether this new meaning evident in
certain contexts has become conventionalized.
4. Reflexive situations
So far, I have intentionally avoided reflexive situations in which the sub-

ject and the dative object are coreferential. The reflexive pronoun has spe-
cific semantics and has developed different subjectively colored uses (see
for example Janda 1993 for Czech and Rudzka-Ostyn 1996 for Polish), in
which the referents of the dative are “conceptualized as experiencers of
their own condition” (Rudzka-Ostyn 1996: 376–377). These nuances are
felt, to various degrees, in all functions of the reflexive dative, including
the possessive dative (compare example (32a.) below), overlaying them
with a tone of familiarity, closeness and informality.
(32) a. Si sedeše spokojno na stolot kako

REFL.IOCL sit.PAST calmly on chair.DEF as if
otsekogaš tamu da si sedela,… (ON:134)
forever there PART REFL.IOCL sit.PAST
‘She was sitting calmly on her chair as if she had always been sit-
ting there, …’
b. Si sedeše spokojno na svojot stol kako ...

REFL.IOCLsit.PAST calmly on her.REFL.DEF chair as if
‘She was sitting calmly on her chair as if …’
It seems that precisely for this reason the possessive component in si is

weakened and often freely allows the appearance of the reflexive posses-
sive pronoun (svoj, svoja, svoe, svoi) in the same clause, as illustrated in
example (32b). Because of the extra expressive component in the clitic the
doubling of the possessive marking in reflexive DS constructions is not felt
as something superfluous. In fact, the clause in (32b) does not evoke any
conflicting effects.
At the same time, there are reflexive constructions which exhibit similar
characteristics as the other DS constructions with double marking that were
discussed previously. The use of both markers is sometimes semantically
motivated. The desire to stress the possessed, but also to keep the posses-
sor’s perspective is the driving force. For example, without the dative clitic
the example in (33a) would sound distancing, as if it does not concern the
person advised to do that (compare (33b)). In (34) all options are possible,
but in (34b) the possessive relationship is not clearly stated without the
possessive adjective svoite.
(33) a. Povratete si ja vašata

get back.IMP REFL.IOCL DOCL.3SG.F your.PL.DEF
mladeška svežina. (Dn.)
youthful freshness.DO
‘Get back your youthful freshness.’
b. Povratete ja vašata mladeška svežina.

get back.IMP DOCL.3SG.F your.PL.DEF youthful freshness.DO
‘Get back your youthful freshness.’
(34) a. Toj neka si se griži za

he.SUB PART REFL.IOCL REFL.DO worry.PRES for
svoite deca, ... (AC)
his.REFL.DEF children
b. Toj neka si se griži za

he.SUB PART REFL.IOCL REFL.DO worry.PRES for
decata, ...
children.DEF
c. Toj neka se griži za svoite deca, ...

he.SUB PART REFL.DO worry.PRES for his.REFL.DEF children
‘He’d better take care of his own children, ...’
Situations involving parts of the body require neither the dative nor the
possessive pronoun for establishing the possessive relation. The sentence
in (35a) is a typical example, where the possessive relationship is clearly
conveyed without any marker –one cannot hold back somebody else’s
tears. In such contexts the introduction of the dative reflexive si is already
felt as emphatic, as in (35b). Still, we encounter examples such as the one
in (32c), where the presence of both the dative and the possessive intensi-
fies this feeling. It seems, however, that in the latter case the construal of
double marking as defined in the previous section dominates. Namely, they
stress the possessed, but also present the situation from the perspective of
the possessor as the affected entity.
(35) a. Majka mi veќe ne uspea

mother POSSCL.1SG longer not succeed.PAST
da gi zadrži solzite. (ON:26)
PART DOCL.3PL hold back.PRES tears.DEF
b. Majka mi veќe ne uspea da

mother POSSCL.1SG longer not succeed.PAST PART
si gi zadrži solzite.
REFL.IOCL DOCL.3PL hold back.PRES tears.DEF
c. Majka mi veќe ne uspea da

mother POSSCL.1SG longer not succeed.PAST PART
si gi zadrži svoite solzi.
REFL.IOCL DOCL.3PL hold back.PRES her.REF.DEF tears
‘My mother could no longer hold back her tears.’
The data from reflexive situations indicate parallel processes to those

observed in the other types of DS. Moreover, they show a stronger motiva-
tion for double marking in reflexive situations, which renders such con-
structions more frequent and less unusual. This suggests that the introduc-
tion of the possessive proper in DS may have been supported by the
situation in the reflexive constructions.
5. Conclusions
Looking back at the examples of double marking in DS we find no syntac-

tic or semantic preferences. The possessed can occupy any syntactic posi-
tion: direct object, subject or prepositional phrase. The type of the pos-
sessed does not seem to be an important factor either, as both closely
connected entities, such as body parts and kinship terms, as well as aliena-
bly possessed entities equally appear in these constructions. Neither are
there constrains on the dative referent or the predicate other than those
applying to the DS construction.
Indeed, it is never obligatory in Macedonian to mark possession twice,
since either the dative or the adnominal possessive pronoun can always
express the required meaning. It has been shown in this paper, however,
that double marking arises in some contexts, in which for structural or se-
mantic reasons the dative is retained even when the speaker needs or
chooses to emphasize the possessed with a possessive pronoun. As a result,

a new meaning is inferred, which diverges from the meanings of the indi-
vidual constructions entering into this component structure. This meaning
becomes even more pronounced when the possessive is added alongside
DS in contexts in which it is not necessary to preserve the grammaticality
or the specific meaning of the clause. We find a similar development in
reflexive situations, where this double marking is even more common due
to the specific nature of the reflexive dative.
It is not still clear whether we are dealing with a new structure here, but
this combination seems to be gaining a composite meaning and developing
into a separate construction. For the time being it functions mainly as a
modification of the existing DS construction, but the fact that it is em-
ployed for functions distinct from the ones characteristic for DS indicates
that it can develop in a slightly different direction.
The Cognitive Linguistics framework offers a suitable theoretical basis
for the analysis of this type of phenomena. The usage based model allows
for modification of language units in the course of their accommodation to
language contexts in the communication process. Deviations from regular
patterns as a result of the confluence of semantic configurations are ex-
plained, rather than rejected as anomalous. The investigation of double
marking in DS in Macedonian based on such premises reveals that it is a
natural outcome of the semantic properties of DS and could be expected to
develop in other related and unrelated languages that employ DS as a spe-
cific means of marking possession.
Notes
1. See for example Minova-Gurkova (1982).

2. For example in Czech: King (1998: 164) says that “[c]olloquially, the dative is
often modified by an adjective and a possessive pronoun, ...”; Fried (1999:
482) also states that the given example “where both IPR and EPR are present,
is admittedly awkward stylistically, but not ungrammatical ...”.
3. Macedonian has lost the nominal case markers on nouns. There are only dis-
tinct accusative and dative case pronoun forms, which are associated with the
direct and the indirect object respectively. For this reason, the terms dative and
indirect object, when referring to Macedonian, are used interchangeably in this
paper.
4. See Kučanda 1996 for illustration of the terminological variation.
5. This is the most common term, but it is sometimes confused with the adnomi-
nal dative with possessive function, such as the one used in Bulgarian for all
types of possession (e.g. knigata mu – literally: ‘the book to him’ (‘his book’))
and in Macedonian for kinship relations only (e.g. brat mi – literally: ‘brother
to me’ (‘my brother’)) .
6. The term unergative is used here to indicate intransitive predicates whose va-
lence contains an agentive argument, but no direct internal argument in con-
trast to unaccusative predicates, which contain an internal argument, but do not
introduce any agentive argument.
7. Langacker (1991: 294) notes that “linguistic structure embodies conventional
imagery and thus imposes a certain construal on the situation it codes”.
8. The comments here refer to the given examples, in which the affectedness is
negative. It is true, though, that in other situations it could also be understood
as positive.
9. The same tendency has been noticed by King (1998: 164) for Czech, especially
in situations where the modifier is an unusual one.
10. The source is cited fully in the list below.
References
Bally, Charles
1996 The expression of concepts of the personal domain and indivisibility
in Indo-European languages. In The Grammar of Inalienability: a ty-
pological perspective on body-part terms and part-whole relation,
Hilary Chappel and William McGregor (eds.), 31–64. Berlin/New
York: Mouton de Gruyter.
Chappel, Hilary, and William McGregor (eds.)
1996 The Grammar of Inalienability: a typological perspective on body-
part terms and part-whole relation. Berlin/New York: Mouton de
Gruyter.
Fried, Miriam
1999 From Interest to Ownership: A Constructional View of External
Possessors. In External Possession, L. Doris Payne and Immanuel
Barshi (eds.), 473–504. Amsterdam/Philadelphia: John Benjamins
Publishing Company.
Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer
1991 Grammaticalization: A Conceptual Framework. Chicago and Lon-
don: The University of Chicago Press.
Hopper, Paul, and Elizabeth Closs Traugott
1993 Grammaticalization. Cambridge University Press.
Janda, Laura A.
1993 A Geography of Case Semantics: The Czech Dative and Russian
Instrumental. Berlin/New York: Mouton de Gruyter.
King, P. Katerina
1998 The Czech Dative of Interest: The Hierarchical Organization of Pos-
session in Discourse and Pragmatics. Ph.D. diss., Harvard University.
Kučanda, Dubravko
1996 What is the Dative of Possession?. Suvremena lingvistika 41/42:
309–318.
Lakoff, George
1987 Women Fire and Dangerous Things: What Categories Reveil about
the Mind. Chicago/London: The University of Chicago Press.
Langacker, Roland W.
1991 Foundations of Cognitive Grammar. Vol. 1: Theoretical prerequi-
sites. Stanford University Press.
2000 Grammar and Conceptualization. (Cognitive Linguistics Research 14.)
Berlin/New York: Mouton de Gruyter.
Levine S. James
1986 Remarks on the Pragmatics of the ‘Inalienable Dative’ in Russian. In
Case in Slavic, Richard D. Brecht and James S. Levin (eds.), 437–
451. Columbus, OH: Slavica.
Minova-Gurkova, Liljana
1982 Kratkite zamenski formi za indirekten objekt i posvojnite zamenski pri-
davki [The pronominal clitics for indirect object and the possessive
pronominal adjectives]. Literaturen zbor XXIX, Kn. 2: 108–110.
Mitkovska, Liljana
2000 On the possessive interpretation of the indirect object in Macedonian.
Linguistica Silesiana 21: 85–101.
Najčeska-Sidorovska, Marija
1973 Sintagmi so Dativus Sympatheticus i negovoto razgraničuvanje od drugite
značenja vo ruskiot, makedonskiot, srpskohrvatskiot i bugarskiot jazik
[Constructions with Dativus Sympatheticus and its differentiation from
other types of meaning of the dative in Russian, Macedonian, Serbo-
Croation and Bulgarian]. Makedonski jazik XXIV: 119–130.
Payne L. Doris, and Immanuel Barshi (eds.)
1999a External Possession. Amsterdam/Philadelphia: John Benjamins Pub-
lishing Company.
Payne L. Doris, and Immanuel Barshi

1999b External Possession: What, Where, How and Why. In External Pos-
session, L. Doris Payne and Immanuel Barshi (eds.), 3–29. Amster-
dam/Philadelphia: John Benjamins Publishing Company.
Rudzka-Ostyn, Brygida
1996 The Polish dative. In The Dative, Vol. 1: Descriptive Studies, Wil-
liam van Belle and Willy van Langendonck (eds.), 341–394. Amster-
dam/Philadelphia: John Benjamins Publishing Company.
Sweetser, Eve
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects
of Semantic Structure. Cambridge: Cambridge University Press.
Topolinjska, Zuzana
1997 Makedonskite dijalekti vo Egejska Makedonija. Kniga 1, Sintaksa,
Tom II [Macedonian dialects in Aegean Macedonia, book 1, Syntax,
Vol. II]. Skopje: MANU.
Velasquez-Castillo, Maura
1999 Body-Part EP Constructions: A Cognitive/Functional Analysis. In
External Possession, L. Doris Payne and Immanuel Barshi (eds.),
77–108. Amsterdam/Philadelphia: John Benjamins Publishing Com-
pany.
Abbreviations for sources
(AC) – Attested in conversation

(D) – Denes, weekly magazine
(Dn) – Dnevnik, daily newpaper
(DP) – Uroševiќ, Vlada. 1996. Dvorskiot poet vo aparat za letanje.
Skopje: Kultura.
(Izbor)– Stalev, Georgi (ed.). 1990. Izbor. Skopje: Detska radost,
Kultura, Makedonska kniga, Misla, Naša kniga.
(MD) – Makedonija denes, daily newpaper.
(ON) – Nikolova, Olivera. 1993. Preminot ne e osvetlen. Skopje:
Detska radost.
(TV) – television
(V) – Vikend, weekly addition to the daily newspaper Dnevnik.
Abbreviations for glosses
ADJ – adjective DO – direct object

CL – clitic F – feminine
DEF – definite article FUT – future tense
IO – indirect object POSSCL– possessive clitic

IMP – imperative PRES – present tense
M – masculine PL – plural
N – neuter REF – reflexive
PART – particle SG – singular
PARTC – participle SUB – subject
PAST – past tense
ȱ ȱ
ȱȱ¢Ǳȱȱȱȱ
ǰȱȱȱȱ
What makes Russian bi-aspectual verbs special?
Laura A. Janda
Abstract
Nearly all Russian verbs are unambiguously either Perfective or Imperfec-

tive. A few hundred verbs are Bi-aspectual, and most of these verbs are for-
eign borrowings. Although scholars have often noted a correlation between
foreign origin and Bi-aspectuality, no one has ever considered the fact that
many borrowed verbs are not Bi-aspectual, nor has anyone ever compared
the behavior of Bi-aspectual and non-Biaspectual foreign verbs. An empiri-
cal study shows that nearly 40% of borrowed verbs are Imperfective, not Bi-
aspectual and that their behavior, in terms of formation of po- prefixed per-
durative verbs denoting ‘do X for a while’ is significantly different from the
behavior of Bi-aspectual verbs. Bi-aspectual borrowed verbs are less prone
to create po- forms than Imperfective borrowed verbs. This difference in be-
havior is predicted by Janda’s “cluster” model of Russian aspect, which has
been proposed as an alternative to the traditional “pair” model, and points to
a difference in the semantic profile of Bi-aspectual vs. Imperfective bor-
rowed verbs. This study thus sheds light on the interaction between lexical
semantics and aspect.
Keywords: aspect, Russian, Bi-aspectual, foreign borrowing
1. Introduction*
Russian is famous for its aspectual distinction between Perfective (the se-
mantically marked value1, in this paper signaled by a superscript “p”) and
Imperfective (signaled by a superscript “i”). Russian aspect is obligatorily
expressed by all verb forms, and furthermore is formally marked by a sys-
tem of aspectual affixes (prefixes and suffixes). For example, ‘write’ can
be rendered in Russian as an Imperfective simplex verb pisat’i, or can be
prefixed to yield the corresponding perfective napisat’p. A few hundred
“Bi-aspectual” verbs can express aspect in the absence of these affixes,
thus constituting exceptions to the overall system of morphological mark-
84 Laura A. Janda
ing of aspect in Russian. To illustrate, the Bi-aspectual verb likvidirovat’p/i

‘liquidate’ can express both Perfective and Imperfective aspects without
the addition of any affixes.
There are good reasons to associate Bi-aspectuality with the borrowing
of foreign verbs into Russian: most Bi-aspectual verbs are foreign and for-
eign verbs are often Bi-aspectual. However, crucial questions concerning
this association have never been addressed in the literature, such as: What
is the relationship between foreign verbs and Bi-aspectuality? and Why do
some foreign verbs become Bi-aspectual in Russian while others become
Imperfective simplex verbs? On a more theoretical level, it is also neces-
sary to ask What is the relationship between Bi-aspectuals and the Russian
aspectual system? This article will explore these issues in light of new
research on the metaphorical motivations for Russian aspect (Janda 2004
and forthcoming a), and will also present an empirical study of foreign
verbs and their behavior in Russian. I will show that many foreign verbs
become Imperfective simplex verbs, not Bi-aspectual verbs in Russian. I
will argue that the lexical semantics of the borrowed verb strongly influ-
ence its aspectual fate in Russian. A semantic distinction that is crucial to
the entire aspectual system is Completability (Janda forthcoming a). A
Completable action heads toward a result, where as a Non-Completable
action is something that can be engaged in without heading toward a result.
Completability is motivated by the metaphor: A COMPLETABLE ACTION IS
TRAVEL TO A DESTINATION. Bi-aspectual verbs are special because they
tend to lack a Non-Completable construal, a claim that is confirmed by
empirical data.
This article begins with what is known (or assumed to be known) about
the Russian aspectual system (section 2) and the status of borrowed and Bi-
aspectual verbs in Russian (section 3). Recent research on the Russian
aspectual system predicts that Bi-aspectual verbs should have a very char-
acteristic profile of aspectual behavior (section 4). An empirical study of
over 550 foreign verbs demonstrates that the profile of borrowed Bi-
aspectual verbs in that group clearly contrasts with the profile of borrowed
Imperfective verbs (section 5), supporting the predictions of the meta-
phorical model. In conclusion (section 6), I will assert that lexical seman-
tics do affect the grammatical category of aspect and its morphological
expression, and that these effects can yield measurable differences in as-
pectual behavior.
What makes Russian Bi-aspectual verbs special 85
2. Traditional assumptions about the Russian aspectual system
This section will describe the Russian aspectual system in its prototypical
instantiation, in other words, excluding discussion of Bi-aspectual verbs.
There is nothing unprecedented about discussing aspect in Russian while
ignoring the Bi-aspectuals; indeed Glovinskaja’s recent book (2001) on
Russian aspect makes no mention of Bi-aspectuals (either individually or
as a phenomenon), and aside from one article in 1998 by Čertkova and
Čang and another one in 1999 by Jászay, there have been little more than
perfunctory remarks on the topic since the 1960s. I will look first at the
morphology of Russian aspect and then give a brief overview of its seman-
tic expression.
A given verb in Russian is either entirely Perfective (semantically
marked) or entirely Imperfective (semantically unmarked) in all tenses and
forms. Both the semantic and the morphological expression of aspect are
obligatory for all forms of all verbs. Russian is thus unlike most other lan-
guages with Perfective and Imperfective aspect, for which aspect is ex-
pressed only in restricted contexts (for example only in the past tense), and
Imperfective is the marked value (as in French).2 In other words, Russian
verb forms are always either Perfective or Imperfective, regardless of tense
or other grammatical categories, and their aspectual status is also signaled
morphologically. As a consequence most (but not all) scholars consider
aspect to be a derivational, not an inflectional, phenomenon in Russian (cf.
Isačenko 1960; Dahl 1985; Zaliznjak and Šmelev 2001; see also discussion
in Janda forthcoming b).
The morphological expression of aspect is achieved primarily by means
of Perfectivizing prefixes (pro-, za-, s-, ot-, na-, po-, vy-, o-, raz-, ob-, u-)
and Imperfectivizing suffixes (-yva-/-iva-, -va-, -a-).3 There is also a Per-
fectivizing semelfactive suffix (-nu) which is more restricted in its use.
This morphology yields the following types of verbs: verbs with no as-
pectual affixes, which are typically Imperfective like pisat’i ‘writei’; verbs
with prefixes, which are typically Perfective like napisat’p ‘writep’,
perepisat’p ‘revisep’, popisat’p ‘writep (for a while)’; and verbs with suf-
fixes, which are typically Imperfective like perepisyvat’i ‘revisei’. There
are some deviations to this pattern, such as unaffixed Perfectives like dat’p
‘givep’ and prefixed Imperfectives like nasledovat’i ‘inheriti’, verbs with
multiple prefixes like poperepisyvat’p ‘revisep (for a while)’, and semelfac-
tive suffixed Perfectives like čixnut’p ‘sneezep (once)’, but generally there
are aspectually related verbs that morphologically disambiguate such cases,
such as davat’i ‘givei’, unasledovat’p ‘inheritp’, perepisyvat’i ‘revisei’, and
86 Laura A. Janda
čixat’i ‘sneezei’. There are also a few suppletive verbs that break these
rules, such as kupit’p ‘buyp’ vs. pokupat’i ‘buyi’. However, most Russian
verbs fit into the pattern of Imperfective simplex, to which one can add a
prefix to get a Perfective verb, to which it might be possible to add a suffix
to get an Imperfective verb. Overall this is a fairly robust and reliable sys-
tem (cf. Timberlake’s 2004: 401–7 “tripartite system”).
The meaning of Russian aspect is the subject of a vast literature which
cannot be adequately surveyed in this article (cf. Janda 2004). Suffice it to
say that this literature has primarily focused on describing aspect in terms
of semantic features, and since Perfective is taken to be the marked mem-
ber of the opposition, most features describe Perfective aspect, leaving
Imperfective as the default. The most common feature labels are “bound-
edness” (Forsyth 1970; Avilova 1976; Jakobson 1957/1971 and Padučeva
1996; cf. also “delimitation” in Bondarko 1971; “closure” in Timberlake
1982;, and “demarcatedness/dimensionality” in van Schooneveld 1978),
“totality” (Comrie 1976; Dickey 2000; Durst-Andersen 1992; Smith 1991;
Isačenko 1960 and Maslov 1965; cf. also “completion” in Vinogradov
1972), “definiteness” (Bondarko 1971 and Dickey 2000; cf. also “change”
and “sequencing” in Durst-Andersen 1992 and Galton 1976), “punctuality”
(Čertkova 1996; Mazon 1914), and “resultative” (Čertkova 1996 and Vi-
nogradov 1972). If the Imperfective aspect is assigned a feature at all, it is
most often “durativity” (Bondarko 1971; Padučeva 1996; Čertkova 1996).
What the tradition of feature analysis tells us is that a Perfective verb de-
scribes a single, unique event viewed in its entirety at a given point in time,
whereas the Imperfective describes all other events, especially those that
are extended in time. What this tradition doesn’t tell us is that the Perfec-
tive and Imperfective aspects have several dozen, often seemingly contra-
dictory uses that features are inadequate to account for. For example, the
general-factual and “polite” imperative uses require Imperfective verb
forms to describe unique whole events, whereas the Perfective is required
for habitually sequenced events, despite the fact that these events are re-
peated over long and indefinite periods of time. Although features do pin-
point the most significant facts about the semantics of aspect, they gloss
over the untidy realities of a very complex phenomenon.
Feature analysis has another relevant by-product, namely the notion of
the “aspectual pair”, consisting of one Perfective and one Imperfective
verb with the same denotation. The assumption that the “aspectual pair” is
the exclusive or dominant pattern observed in the Russian aspect system is
entrenched and pervasive in Russian linguistics (cf. Vinogradov 1938;
Šaxmatov 1941; Bondarko 1983; Čertkova 1996; Zaliznjak and Šmelev

2000; Glovinskaja 2001).
In this article I will not challenge the traditional assumptions about
Russian aspect concerning the marked value of Perfective and the “tripar-
tite” derivational system consisting of simplexes and affixed forms. I will
suggest that the concepts of semantic features and the “aspectual pair” are
in need of revision in order to account for the complex facts of Russian. In
a sense, the revisions I will suggest (in section 4) extend and contextualize
the traditional assumptions, giving them firmer semantic grounding, and it
is precisely semantic factors that will be crucial in the analysis in this arti-
cle. But before considering these semantic factors, one needs to examine
what is known about borrowed and Bi-aspectual verbs in section 3.
3. Traditional assumptions about borrowed and Bi-aspectual verbs in

Russian
This section will define borrowed and Bi-aspectual verbs, discuss the
status of borrowed and Bi-aspectual verbs in Russian, and outline the vari-
ous paradigmatic, semantic, syntactic, and morphological constraints on
Bi-aspectual verbs.
Borrowed verbs are verbs that cannot be traced to the common origins
of Slavic languages and have entered Russian since it commenced its de-
velopment largely independent of other Slavic languages. Although this is
a long period (reaching from the 12th century to the present), there was
very little activity in borrowing verbs until the Petrine period (late 17th
century), when contact with and knowledge of western European languages
became common among the Russian elite (Avilova 1968). All borrowed
verbs in Russian contain the suffix -ova-, which integrates them into the
inflectional system of Russian (without, however, providing any aspectual
designation). The suffix -ova- is itself Slavic, but is most often used with
various extensions in the presence of foreign stems, appearing as -izova-, -
irova-, -izirova-, -ficirova-. Although the -ova- suffix is indispensable –
without it a foreign word would lack inflection and also the ability to func-
tion as a verb in Russian – it is claimed that a desire to preserve the foreign
flavor of borrowed verbs caused them to resist further affixation (Mučnik
1966; Avilova 1968). This meant that foreign verbs would resist aspectual
prefixation and suffixation, supporting their recognition in Russian as Bi-
aspectual verbs (both Mučnik 1966 and Gladney 1982 compare this to the
importation of indeclinable foreign nouns such as kino ‘cinema’).
88 Laura A. Janda
Bi-aspectual verbs are verbs that can express both Perfective and Imper-
fective aspect with the same morphological form, without recourse to the
Perfectivizing and Imperfectivizing affixes described in section 1. Esti-
mates of the number of Bi-aspectual verbs vary. Mučnik (1966) Gladney
(1982) and Anderson (2002, based on combined listings in dictionaries)
suggest that there are approximately 600. Čertkova and Čang (1998) rec-
ognize 412 Bi-aspectual verbs, only 289 of which are “true Bi-aspectuals”
(the other 123 verbs are claimed to behave as both Bi-aspectuals and as
“paired” verbs). Wheeler 1972/1992 lists 348 verbs as Bi-aspectual. In-
deed, as Jászay (1999: 169–170) laments, dictionaries do not agree on the
identification of verbs as Bi-aspectual (cf. also Čertkova and Čang 1998:
24). Nor do scholars. Mučnik (1966: 69), for example, rejects Isačenko’s
(1960: 144) assignment of Bi-aspectuality to certain forms of or-
ganizovat’p/i ‘organizep/i’. And, as Jászay’s (1999) research shows, native
speakers also vary in their acceptance of Bi-aspectual forms, leading him to
coin the term “častičnaja dvuvidovost’” (“partial Bi-aspectuality”).
There is even less clear information on the status of borrowed verbs in
Russian and their relationship to Bi-aspectuality. Avilova (1968: 66) opens
her article with the statement “Принято считать, что глаголы с
заимствованной основой в русском языке являются двувидовыми” (“It
is commonly assumed that verbs with borrowed stems in Russian are Bi-
aspectual”), an assumption that she never challenges, although she repeat-
edly acknowledges the existence of borrowed verbs that are merely Imper-
fective (Avilova 1968: 66, 70, 71, 72, 73, 76). Čertkova and Čang (1998:
15) recognize 48 Imperfective verbs, nearly all of them foreign, among
verbs (erroneously in their view) listed as Bi-aspectuals in dictionaries. It is
common for scholars to note that some Bi-aspectual verbs are developing
prefixed Perfectives, which means that the borrowed simplex verb is Im-
perfective (Isačenko 1960; Mučnik 1966; Jászay 1999; Zaliznjak and Šme-
lev 2000). However, with the lone exception of Avilova (1968; and these
are merely parenthetic remarks), no one mentions the fact that some bor-
rowed verbs do not enter the Russian lexicon as Bi-aspectuals, and no one
has investigated the relationship between Bi-aspectual and Imperfective
verbs among borrowed verbs in Russian. Avilova (1968) furthermore
claims that the foreign verbs that do enter Russian as Imperfectives are
imperfectiva tantum (Imperfective isolates with no aspectually related
Perfective verbs listed in dictionaries).
Mučnik (1966: 64) challenges the prevailing assumption that foreign
Bi-aspectual verbs are predominantly scientific, technical, or professional
terms. He claims that only 35% of bi-aspectual verbs fall into those three
groups, and that the rest belong to the common lexicon of the literary lan-
guage. Unfortunately there is no corresponding data on the status of non-
Bi-aspectual foreign verbs.
It is true that there is some correlation between Bi-aspectuality and for-
eign origin among verbs, but this fact has always been examined only by
looking at the frequency of foreign borrowings among Bi-aspectuals.
Čertkova and Čang (1998: 13) state that only about 10% of Bi-aspectual
verbs are part of Russian’s Slavic heritage. Anderson’s (2002) combined
list shows that 95% of Bi-aspectual verbs in Russian are foreign. Isačenko
(1960: 144) points out that in addition to native and foreign Bi-aspectual
verbs, there is some marginal tendency to build new Bi-aspectual verbs
using Russian stems and the -ova- suffix: zvukoficirovat’p/i ‘equip with
soundp/i’. But no one has ever asked how many foreign verbs are Bi-
aspectual. Nor has anyone pointed out the significant role of foreign verbs
in building parts of the Russian lexicon other than the Bi-aspectuals. For
example, foreign borrowings account for 20%4 of Imperfective only verbs
and are the sole productive source of such verbs in Russian, but there is no
mention of foreign verbs in Zaliznjak and Šmelev’s (2000: 85–6) discus-
sion of Imperfective isolates.
One thing that most scholars do agree on is that Bi-aspectual verbs are
not ambiguous in their expression of aspect. In other words, Bi-aspectual
verbs do not express a neutral aspect or a lack of aspect. Every use of a Bi-
aspectual verb like likvidirovat’p/i ‘liquidatep/i’ is either Perfective or Im-
perfective, as disambiguated by context (Isačenko 1960: 143–44; Mučnik
1966: 61; Avilova 1968: 66; Galton 1976: 294; Gladney 1982: 202;
Čertkova 1996: 100–109; Jászay 1999: 169; Zaliznjak and Šmelev 2000:
10), just as number is disambiguated when English fish is used in context.
The one significant exception is Timberlake (2004: 407–9) who suggests
that Bi-aspectual verbs are “anaspectual”, and do not express aspect. Due
to the prevailing assumption that Russian verbs come in “aspectual pairs”
(see section 2), most scholars consider Bi-aspectual verbs to be syncretic
examples of “pairs” where the two verbs in each “pair” are homophonous.
This means that Bi-aspectual verbs are deviant in their formal morphologi-
cal marking of aspect, but not in their semantic expression of this linguistic
category.
Bi-aspectual verbs (both foreign and native) display a range of some-
what unusual phenomena, most of which can be stated as constraints on
their paradigmatic, syntactic, and semantic expression of aspect. Although
most forms in the paradigm of Bi-aspectual verbs can express both Perfec-
tive and Imperfective, some forms are of necessity monoaspectual, namely
90 Laura A. Janda
the periphrastic future (budu likvidirovat’i ‘I will be liquidatingi’) and the

various participles and gerunds (likvidirujuščiji ‘that which is liquidatingi’,
likvidiruemyji ‘that which is being liquidatedi’, likvidirujai ‘while liquidat-
ingi’, likvidirovavšijp ‘that which had liquidatedp’, likvidirovannyjp ‘that
which was liquidatedp’ and likvidirovavp ‘having liquidatedp’; cf. Mučnik
1966: 61). The non-past forms can express both aspects and both present
and future tense, but only in two combinations, either as future Perfective
(likvidirujup ‘I will liquidatep’) or as present Imperfective (likvidirujui ‘I am
liquidatingi’). Many Bi-aspectual verbs show a strong tendency to express
only Perfective aspect in past tense forms (Zaliznjak and Šmelev 2000: 75).
Some Bi-aspectuals cannot be used in the following syntactic contexts
(which would normally permit Imperfective verbs): with phasal verbs
(načnu + infinitivei ‘I will start X-ingi’), with temporal adverbs like dolgo
‘for a long time’, dva časa ‘two hours’ in the past tense, and with modal
verbs (ne sleduet + infinitivei ‘one shouldn’t Xi’; cf. Čertkova and Čang
1998: 23). It is not uncommon for a verb to be Bi-aspectual in only one
meaning, but “paired” in all other meanings; for example, Čertkova and
Čang (1998: 27) note that restavrirovat’i/p ‘restorei/p’ functions both as a
Bi-aspectual and as a simplex Imperfective (with the Perfective otrestavri-
rovat’p) in concrete meanings, but only as a Bi-aspectual in metaphorical
meanings. Although such constraints involve all types of Bi-aspectual
verbs, the “native” Russian Bi-aspectuals seem particularly vulnerable
(Zaliznjak and Šmelev 2000: 72–75). Overall, this means that Bi-aspectual
verbs do not typically show the full range of options that we would expect
from an Imperfective verb and its corresponding Perfective(s) .
There is an overall tendency to eliminate Bi-aspectuality by integrating
both foreign and native verbs into the system of Russian aspectual mor-
phology (Isačenko 1960: 146–48; Mučnik 1966; Avilova 1968; Gladney
1982: 212; Čertkova and Čang 1998; Zaliznjak and Šmelev 2000: 75). This
can happen either by: a) designating the simplex verb as an Imperfective
and accepting a prefixed form as the Perfective correlate; or b) by designat-
ing the simplex verb as a Perfective and accepting a suffixed form as the
Imperfective correlate. For example, analizovat’i ‘analyzei’ is now Imper-
fective and only the prefixed proanalizovat’p is acceptable in Perfective
contexts, whereas arestovat’p ‘arrestp’ is now a simplex Perfective, and
arestovyvat’i is required in Imperfective contexts (Jászay 1999: 170, 172).
The first solution, that of recognizing the borrowed verb as a simplex
Imperfective that can have prefixed Perfectives, is by far the most com-
mon, and there are probably two reasons for this. One is that this pattern
conforms best to the prototypical use of Russian aspectual morphology
(Timberlake’s 2004: 401–7 “tripartite system”, mentioned above). The

other reason for favoring the recognition of foreign verbs as simplex Im-
perfectives is that there are strong morphological constraints on the suf-
fixation of unprefixed verbs with the -ova- suffix. Imperfectivization
through suffixation in -yva-/-iva- is only possible when the stress falls on
the -a- vowel of the -ova- suffix, and only about 15% of foreign Bi-
aspectual verbs have that stress. The remainder (and this includes nearly all
verbs with the extended versions of the -ova- suffix) are ineligible to be
integrated into the aspectual system by means of suffixation (Mučnik 1966;
Avilova 1968; Jászay 1999). The historical trend toward elimination of Bi-
aspectuality is not perfectly uniform (Mučnik 1966: 65 notes that some
foreign verbs were “paired” in previous centuries but are now Bi-
aspectual), nor is its rate (which is definitely slower for the relic, native Bi-
aspectual verbs, and can differ widely among foreign verbs, cf. Zaliznjak
and Šmelev 2000; Jászay 1999). The outcome of this process is also vari-
able. Sometimes the acquisition of aspectual morphology is claimed to
eliminate Bi-aspectuality, and in other instances morphologically marked
forms coexist with their Bi-aspectual equivalents (Avilova 1968; Čertkova
and Čang 1998).
Here is a summary of what is known about borrowed and Bi-aspectual
verbs in Russian: They share the -ova- suffix (plus extensions), which lim-
its their integration as Perfective simplex verbs, favoring instead their rec-
ognition as simplex Imperfectives. We know relatively little about the
numbers and aspectual profiles of non-Bi-aspectual borrowed verbs, and
their status has not been compared to that of the Bi-aspectual borrowed
verbs. Though we know more about the numbers and status of Bi-aspectual
borrowed verbs, much of that information is hard to assess due to conflict-
ing claims concerning various issues, such as how many and which verbs
are Bi-aspectual in which forms. We do know that most Bi-aspectual verbs
are borrowed, that they unambiguously express aspect in context, that this
semantic expression of aspect is gradually being formally supported by the
acquisition of aspectual morphology, mostly in the form of prefixed Per-
fectives. We also know that there are various constraints on Bi-
aspectuality.
What we don’t know is how the numbers and status of borrowed Bi-
aspectual verbs compare with the numbers and status of non-Bi-aspectual
borrowed verbs. There seems to be an implicit assumption that all (or vir-
tually all) borrowed verbs are Bi-aspectual (cf. the quote from Avilova
above), and that if there are other borrowed verbs, they aren’t interesting.
The empirical study in section 5 will address these issues, and will show
92 Laura A. Janda
that the comparison between Bi-aspectual and non-Bi-aspectual borrowed

verbs is actually revealing, for it indicates that semantic factors may be
strong predictors of the aspectual status of borrowed verbs. Before under-
taking this comparison, however, it is necessary to examine the semantics
of aspect in Russian in more detail, focusing on issues of particular rele-
vance for Bi-aspectual and Imperfective verbs. This is the topic of section
4.
4. Implications of the cluster model for borrowed and Bi-aspectual

verbs
This section will present an alternative to the “aspectual pair” model that is
based on metaphorical motivations for aspectual behavior in Russian. This
alternative model, the cluster model, highlights the differences between Bi-
aspectual and Imperfective simplex verbs on the basis of both semantic
content and morphological behavior.
Given the predominance of the “aspectual pair” model in general (cited
in section 2), as well as the fact that Bi-aspectual verbs are assumed to be
syncretic “aspectual pairs” (cited in section 3), it is necessary to examine
the concept of the “aspectual pair” in some detail. On the basis of a large,
stratified sample representing the full range of the Russian verbal lexicon
(Janda forthcoming a), I have advanced the cluster model of Russian verbal
aspect. This model acknowledges the existence of the aspectual partner-
ships traditionally labeled “pairs”, but at the same time recognizes that
these partnerships are usually embedded in larger clusters of aspectually
related verbs. On the motivation of three metaphors, the Russian aspectual
system differentiates Imperfectives from not one, but four distinct types of
Perfective verbs in Russian, each of which has a specific semantic and
morphological profile. I will look first at the three metaphors and then at
the four types of Perfective verbs.
The three metaphors that govern the Russian aspect system compare the
temporal contours of events to the behavior of concrete objects in three
source domains involving physical matter, motion, and granularity. The
second of these metaphors, the one comparing events to motion, will prove
most important in the analysis of borrowed and Bi-aspectual verbs. The
metaphors (in keeping with traditions of cognitive linguistics, cf. Lakoff
1987) will be stated in capital letters:
1) A PERFECTIVE EVENT IS A SOLID OBJECT, AN IMPERFECTIVE EVENT IS
A FLUID SUBSTANCE. This metaphor distinguishes Perfective from Imper-
fective aspect in Russian, on the basis of fourteen parameters that show

isomorphism between properties of matter and uses of aspect, as explained
in detail in Janda 2004 (cf. also Mehlig 1994 and 1997). This metaphor
will not be discussed further here.
2) A COMPLETABLE ACTION IS TRAVEL TO A DESTINATION. This meta-
phor distinguishes Completable and Non-Completable actions. Motion can
involve either travel to a destination, as in Russian idtii ‘walki (toward a
destination)’, or motion in random or multiple directions, as in Russian
xodit’i ‘walki (without a unitary destination)’. Metaphorically any Com-
pletable goal-directed action, like pisat’i dissertaciju ‘writei (one’s) disser-
tation’ is a bit like a trip to a destination (here, the last page of the disserta-
tion). At all times when you are writing your dissertation you are making
progress, and when you reach the end, you are done. A Non-Completable
action is more diffuse, lacking a unitary goal, like rabotat’i v kabinete
‘worki in (one’s) office’. You don’t actually finish working, you just work
until it is time to go home and then do it some more the next day. In addi-
tion to distinguishing the two types of motion verbs in Russian, Com-
pletability distinguishes three types of Perfective verbs, the Natural Perfec-
tive, Specialized Perfective, and Complex Act Perfective, as detailed
below.
3) A SINGLE CYCLE OF REPEATED ACTION IS A SINGLE GRAIN OF SAND.
This metaphor distinguishes Singularizable actions from actions that are
not Singularizable, by comparing actions to sand. If the action, like sand,
has single identical grains, then the action can be Singularized. An example
is ščipat’i ‘pinch/plucki’, which refers to pinching or plucking in general,
from which a single pinch or pluck can be extracted. Singularizability dis-
tinguishes the fourth type of Perfective verb, the Single Act Perfective.
Metaphor 2) makes several crucial distinctions among Perfective verbs.
Three types of Perfectives are motivated by this metaphor: 1) Natural Per-
fective (the logical conclusion of an activity), 2) Specialized Perfective (a
specialized conclusion of the activity), and 3) Complex Act Perfective (a
combination of activity and external boundaries). These three Perfectives
are associated with different construals of Completability. If an activity can
be construed as Completable, it can have a Natural Perfective. If an activity
can be construed as Completable if it is performed in a specialized way, it
can have a Specialized Perfective. And if an activity can be construed as
Non-Completable, it can have a Complex Act Perfective. Examples of
these construals and Perfectives appear below.
Most simplex Imperfective verbs in Russian are ambiguous for Com-
pletability and can be construed either way depending upon context. The
94 Laura A. Janda
verb pisat’i ‘writei’ is construable as either Completable, as in Sestra pišeti

dissertaciju ‘(My) sister is writingi (her) dissertation’, or as Non-
Completable, as in Sestra pišeti naučnuju fantastiku ‘(My) sister writesi
science fiction/is a science fiction writer’. In the latter example the aspec-
tual function of pisat’i ‘writei’ is comparable to that of rabotat’i ‘worki’. If
a verb can be construed as Completable, then it can have a Natural Perfec-
tive which describes the logical completion of the activity. Thus pisat’i
‘writei’ has the Natural Perfective napisat’p ‘writep’. Although rabotat’i
‘worki’ cannot be construed as Completable and lacks a Natural Perfective,
if further semantic content is added, a specialized Completable meaning
can arise, as in the Specialized Perfective pererabotat’p ‘re-do, revisep’.
This is possible for some, but not all Non-Completable verbs. Skripet’i
‘squeaki’, for example, is exclusively Non-Completable and lacks both
Natural and Specialized Perfectives.
Only verbs with a Non-Completable construal can form Complex Act
Perfectives (also known as Aktionsarten or actionality, cf. Bertinetto and
Delfitto 2000; Tatevosov 2002) which combine the activity described by
the Imperfective with external boundaries. The best example of this in
Russian is the po- prefixed delimitative, as in popisat’p ‘writep (for a
while)’, porabotat’p ‘workp (for a while)’, poskripet’p ‘squeakp (for a
while)’. A verb that lacks a Non-Completable construal cannot have a
Complex Act Perfective. For example, krepnut’i ‘get strongeri’ is inexora-
bly Completable; one cannot engage in the activity of getting stronger
without making progress and actually getting stronger. A verb like krep-
nut’i ‘get strongeri’ cannot form a Complex Act Perfective.
For the remainder of this article, the most important distinction here is
between verbs that do have a Non-Completable construal (like pisat’i
‘writei’, rabotat’i ‘worki’, and skripet’i ‘squeaki’) and verbs that do not
have a Non-Completable construal (like krepnut’i ‘get strongeri’). I will
argue that typical borrowed Imperfective verbs are like most native Rus-
sian Imperfectives: they have a Non-Completable construal and can form
Complex Act Perfectives. However, typical borrowed Bi-aspectual verbs
lack a Non-Completable and hence also lack Complex Act Perfectives.
The fourth type of Perfective, the Single Act Perfective, is motivated by
Metaphor 3 and is available only to verbs that form Complex Act Perfec-
tives. If an activity is construable as both Non-Completable and Singulariz-
able, a Single Act Perfective can be formed, as in skripnut’p ‘squeakp
(once)’ and ščipnut’p ‘pinch/pluckp (once)’. The verb ščipat’i ‘pinch/plucki’
also has all the other types of Perfectives, namely the Natural Perfective
obščipat’p ‘pluckp’, Specialized Perfectives such as vyščipat’p ‘pluck outp’,
and Complex Act Perfectives such as poščipat’p ‘pinch/pluckp (for a

while)’. Thus ščipat’i ‘pinch/plucki’ illustrates the maximal cluster type
with all four types of Perfectives. This and all other cluster structures at-
tested in Russian conform to the following implicational hierarchy (items
to the left of the “>” are included in a cluster prior to items on the right,
and items in parentheses are optional and unordered):
Imperfective > (Natural Perfective/Specialized Perfective) > Complex Act

> Single Act
This implicational hierarchy was originally proposed on the basis of a

study (Janda forthcoming a) that did not specifically target borrowed and
Bi-aspectual verbs, but at the same time did not exclude them. That study
included eighteen borrowed verbs, seven of which serve as simplex Imper-
fectives, and eleven of which are Bi-aspectual. The seven borrowed Imper-
fectives behave much like any other Imperfectives. Collectively they repre-
sent five different cluster structures, which include various combinations
of all four types of Perfectives. In other words, the borrowed Imperfectives
are unremarkable. The eleven Bi-aspectual verbs (nine of which are bor-
rowed and two of which are native) are on the contrary conspicuous for
their inability to have Complex Act and Single Act Perfectives in their
clusters. The Bi-aspectual verbs form only two types of clusters: Imperfec-
tive + Natural Perfective and Imperfective + Natural Perfective + Special-
ized Perfective. In both cluster types the Imperfective and Natural Perfec-
tive are syncretic, although for some verbs there are alternative prefixed
Natural Perfectives. Thus, for example, the verb annulirovat’p/i ‘annulp/i’
serves as both Imperfective and Natural Perfective in its cluster, but klassi-
ficirovat’p/i ‘classifyp/i’ has in addition the alternative Natural Perfective
rasklassificirovat’p. In other words, whereas Imperfective borrowed verbs
seem unremarkable in their cluster structure, Bi-aspectual borrowed verbs
have a very limited repertoire of cluster structures.
The cluster model suggests a link between semantic content and cluster
structure. According to this model, only a verb that can be construed as
Non-Completable can have a Complex Act Perfective, and only verbs that
have a Complex Act Perfective are eligible to form a Single Act Perfective.
Borrowed Imperfectives seem to form Complex Act Perfectives at a rate
comparable to that of other Imperfectives. The cluster structure displayed
by Bi-aspectual verbs indicates that they cannot be construed as Non-
Completable, and this is correlated with a lack of Complex Act (and by
necessity also Single Act) Perfectives. Given the paucity of data on bor-
96 Laura A. Janda
rowed verbs in the original study, this correlation can only be stated as a
hypothesis, and this hypothesis will be tested empirically in section 5.
5. Empirical study of borrowed and Bi-aspectual verbs in Russian
The cluster model offers us a metric by which Bi-aspectual and non-Bi-

aspectual borrowed verbs can be meaningfully compared, namely the for-
mation of Complex Act Perfectives. This section will present an empirical
study of borrowed verbs and analyze its results. In any empirical study, a
number of practical decisions have to be made, some of which have conse-
quences for the data. Every attempt will be made to describe these deci-
sions and their effects in this section. Despite unavoidable imperfections,
the data provide compelling evidence that borrowed Bi-aspectual and non-
Bi-aspectual (Imperfective) verbs have very different profiles, both in
terms of their participation in derivational morphology and in terms of their
semantics.
The first task was to assemble a list of borrowed verbs, classified as Bi-
aspectual as opposed to non-Bi-aspectual. The goal was to have a list of
manageable size with clear classifications. Because dictionaries differ in
their identification of Bi-aspectual verbs (see section 3), it was decided to
initially cull verbs and classifications only from one dictionary to reduce
“noise” in the data (and on the assumption that any single dictionary
should be reasonably consistent). Wheeler 1972/1992 was selected for this
purpose because it is a modest, but fairly comprehensive dictionary, and
because it lists just one option (Bi-aspectual or Imperfective) for each entry
of a foreign verb. Because the use of search engines would later make it
impossible to automatically differentiate between reflexive (-sja) and non-
reflexive forms of verbs, all reflexive verbs are counted together with any
non-reflexive verb as a single verb. Thus rekomedovat’/sjap/i ‘recom-
mend/be recommendedp/i’ counts as a single verb in this study. This made it
impossible for us to detect differences in the behavior of non-reflexive and
reflexive verbs. However, verbs of this type (with both reflexive and non-
reflexive forms) constitute only 6% of our data, so the presence of such
verbs should not have a significant effect on the data, nor were there
enough such verbs to determine whether their behavior was actually differ-
ent. It was also impossible for us to disambiguate verbs that had two en-
tries, such as massirovat’p/i which is listed as both ‘amassp/i’ and ‘mas-
sagep/i’. There were only three such verbs in the entire data set (two among
the Bi-aspectuals and one among the non-Bi-aspectuals), constituting only
0.5% of the database. These three verbs were each treated as if they were a
single entry, despite multiple meanings. There was one verb (or group of
verbs) that had to be excluded from the study altogether due to the fact that
the differentiation between reflexive and non-reflexive uses as well as dif-
ferent meanings associated with different stress made it impossible to clas-
sify it as either Bi-aspectual or non-Bi-aspectual in such a way that it could
be distinguished by search engines: pikirovat’/sja. This verb is Bi-
aspectual when it is not reflexive, and has two meanings: ‘dive, swoop’
(aeronautical) when stressed on the second syllable; ‘thin out’ (agricul-
tural) when stressed on the last syllable. When reflexive, this verb is Im-
perfective only and means ‘exchange caustic remarks, cross swords’. Dis-
ambiguation of these uses proved to be impossible given the remaining
design of the study. There was also one unprefixed Perfective foreign verb,
atakovat’p ‘attackp’, which was not included in the study because it could
not be classified as either Bi-aspectual or non-Bi-aspectual.
555 relevant foreign verbs were culled from Wheeler 1972/1992, 349
(63%) of which were designated by that source as Bi-aspectual, and 206
(37%) of which were designated as Imperfective. These numbers alone
make a compelling case for comparing borrowed Bi-aspectual and non-Bi-
aspectual verbs, since if nearly 40% of the borrowed verbs are non-Bi-
aspectual, then one cannot assume that all (or nearly all) borrowed verbs
are Bi-aspectual. At this point I have established that Russian has at least
two types of borrowed verbs and they are attested in roughly similar num-
bers. This fact begs the question of why some borrowed verbs are Bi-
aspectual whereas others are not. It should also be noted that these figures
probably underreport the rate of non-Bi-aspectual verbs, since scholars
(cited in section 3) frequently comment that dictionaries list verbs that are
no longer Bi-aspectual as Bi-aspectual. It is likely therefore that some of
the items classified in our list as Bi-aspectual are actually Imperfective.
The hypothesis from section 4 is that one should expect a strong ten-
dency for the Imperfective borrowed verbs to form Complex Act Perfec-
tives, whereas these Perfectives should be rare or non-existent for the Bi-
aspectual borrowed verbs. To test this hypothesis I needed to document the
use of Complex Act Perfectives for all verbs in the study. Because the most
common prefix found with Complex Act Perfectives is po-, it was decided
that these forms would be searched (in Janda forthcoming a, only one verb
was found that formed a Complex Act Perfective with a prefix other than
po- in the absence of any po- form). Searching for po- forms of course
carried with it the liability that in addition to Complex Act Perfectives I
would collect po- Natural Perfectives. Descriptions in the literature indi-
98 Laura A. Janda
cate that the vast majority of po- forms associated with borrowed Bi-
Aspectuals would not be Natural Perfectives. For example, Avilova (1968:
67) lists po- among the prefixes used least frequently in prefixed Natural
Perfectives of borrowed Bi-aspectual verbs (she ranks po- eighth in a list of
ten prefixes used by these verbs). Overall use of po- to form Natural Per-
fectives in Russian is very high (cf. Dickey 2006: 3; Čertkova 1996), so it
was expected that confounding data would be located particularly among
the borrowed Imperfectives. To control this, examples of uses of po- forms
were manually checked and verified for the presence of delimitative mean-
ing (signaled by adverbials such as nekotoroe vremja ‘for a while’). When
po- forms were found, it was nearly always possible to verify the meaning
in at least some of the “hits” as delimitative.
Searches of the verbs were performed on www.yandex.ru in June 2006,
and data was collected on the number of hits turned up for each verb both
without a po- prefix and with a po- prefix. Table 1 gives basic data on the
numbers gathered in this study. Of course data collected on a search engine
is notorious for being unstable and unreliable. However, given the low
frequencies of many of the verbs involved, it was impossible to collect
sufficient data for all but a handful of these verbs from corpora, so a search
engine was the only option.
The most striking observation to be made in Table 1 is that the average
number of hits for po- prefixed verbs is more than five times higher for the
Imperfective borrowed verbs than that for the Bi-aspectual borrowed verbs.
This can be restated as the proportion of po- prefixed forms, which is
2.52% for Bi-aspectual verbs and 12.67% for Imperfective verbs. A logistic
regression model using Pearson’s statistic to adjust for hit rate heterogene-
ity among verbs – this is needed as a different number of examples was
collected for each verb and the probability that a verb will be prefixed with
po- differs even for verbs that belong to the same class – reveals that there
is a demonstrable positive correlation between Imperfective borrowed
verbs and the presence of po- prefixed forms [χ2= 107.37, df=1, p<.0001].
The two types of verbs give very different data, with the Bi-aspectual
verbs yielding obvious differences at the low end of the scale and the Im-
perfectives yielding interesting numbers at the high end of the scale.
Whereas only 10% of Imperfective verbs yield zero po- forms, the absence
of po- forms is more than three and a half times as frequent (36%) for Bi-
aspectual verbs. The presence of small numbers (≤ 25) of po- forms is very
common for Bi-aspectual verbs (44%), but thereafter it drops precipitously
and high numbers of po- forms are vanishingly rare. 35% of Imperfective
verbs also yielded low numbers of po- forms (≤ 25), but this may have
been due in part to the presence of many verbs with very low overall fre-
quency in this group (an effect corrected for in the discussion of Table 2
below).
Table 1. Unprefixed and po- prefixed Bi-aspectual and Imperfective borrowed

verbs
Bi-aspectual borrowed Imperfective borrowed
verbs verbs
unprefixed po- prefixed unprefixed po- prefixed
Average # 1,903 51 1,973 265
hits
Maximum # 77,799 2,444 25,784 1,697
hits
Minimum # 19 0 20 0
hits
At the other end of the scale, verbs listed as Imperfective were ten times
more likely (20%) to provide over 500 “hits” for po- forms than Bi-
aspectual verbs (with only 2% of verbs having over 500 po- “hits”). 80% of
Bi-aspectual verbs have 0–25 po- form attestations, but 54% of Imperfec-
tive verbs have 26 or more po- form attestations. At the low end of the
scale, cutting the scale of po- attestations between 0-to-5 and 6-and-up,
there is an even more revealing result. 63% of Bi-aspectual verbs have
between zero and 5 po- form attestations, whereas 67% of Imperfective
verbs have 6 or more po- form attestations. The two types of verbs, bor-
rowed Bi-aspectuals and borrowed Imperfectives, thus behave as mirror
images of each other in terms of their tendency to create po- prefixed
Complex Act Perfectives.
As mentioned above, there are some very low frequency verbs in the
data, particularly as pertains to Imperfective verbs that have only a few po-
attestations. For example, brasovat’i ‘bracei’ yields only 38 unprefixed
“hits” and 5 po- prefixed “hits”. In general, it is hard to say how much
small numbers (or small differences in numbers) of “hits” might mean on a
search engine, since they can result from typos or reduplicated returns. And
it is hard to say how significant effects are when dealing with low-
frequency verbs. In order to correct for these problems, the focus was nar-
rowed to verbs that are robustly attested by the search engine, namely
verbs with over 1000 “hits”. All of these verbs are well-established in Rus-
sian and should be familiar to any native speaker. To compensate for the
100 Laura A. Janda
vagaries of search-engine data, I looked only at the extremes of the scale

for verbs that have over 1000 attestations in order to show the largest pos-
sible differences. Tables 2 (a) and 2 (b) list all the high-frequency verbs
found with either zero or over 1000 po- form attestations.
Table 2(a). High-frequency Bi-aspectual verbs with either zero or over 1000 po-
prefixed hits
Bi-aspectual verbs with over 1000 unprefixed hits
Zero po- prefixed hits: 70 Over 1000 po-
prefixed hits: 3
abonirovat’/sja ‘subscribe’ inkrustirovat’ ‘encrust’ massirovat’ ‘amass;
massage’
angažirovat’ ‘book, engage’ intensificirovat’ rekomendovat’/sja
‘intensify’ ‘recommend’
anglizirovat’ ‘anglicize’ internacionalizirovat’ remontirovat’
‘internationalize’ ‘renovate’
assignovat’ ‘allocate, budget’ internirovat’ ‘place in
internment’
debetovat’ ‘debit, charge’ kanonizirovat’ ‘canonize’
deblokirovat’ ‘unblock’ kanonizovat’ ‘canonize’
decentralizovat’ ‘decentralize’ kapitalizirovat’ ‘reinvest’
dekretirovat’ ‘decree’ kollektivizirovat’
‘collectivize’
demaskirovat’ ‘unmask’ kolonizovat’ ‘colonize’
demilitarizovat’ ‘demilitarize’ kristallizovat’sja
‘crystallize’
demobilizovat’ ‘demobilize’ kvalificirovat’ ‘qualify,
test’
denacionalizirovat’ ‘privatize’ latinizirovat’ ‘latinize’
denaturirovat’ ‘denature’ litografirovat’
‘lithograph’
denonsirovat’ ‘denounce’ meblirovat’ ‘furnish,

upholster’
dešifrirovat’ ‘decipher’ metallizirovat’ ‘apply
metal coating’
detalizovat’ ‘work out in demilitarizovat’ ‘militarize’
tail’
dezertirovat’ ‘desert’ njuansirovat’ ‘supply
nuances’
dezavuirovat’ ‘disavow’ monopolizirovat’
‘monopolize’
Table 2(a) continued

Zero po- prefixed hits: 70 Over 1000 po-
prefixed hits: 3
dezorganizovat’ ‘disorganize’ orkestrovat’ ‘orchestrate’
dezorientirovat’/sja ‘disorient’ ornamentirovat’
‘ornament’
èšelonirovat’ ‘echelon, personificirovat’
designate’ ‘personify’
evropeizirovat’ ‘europeanize’ plakirovat’ ‘plate, coat’
flankirovat’ ‘flank’ radioficirovat’ ‘equip with
a radio’
frakcionirovat’ ‘sort, ratificirovat’ ‘ratify’
fractionate’
germanizirovat’ ‘germanize’ reducirovat’ ‘reduce’
gil’otinirovat’ ‘guillotine’ repatriirovat’/sja
‘repatriate’
gofrirovat’ ‘crimp’ satinirovat’ ‘glaze, polish’
gradirovat’ ‘boil out of saline’ servirovat’ ‘serve, dish
up’
granulirovat’ ‘granulate’ sindicirovat’ ‘syndicate’
gummirovat’ ‘coat with skandalizirovat’
rubber’ ‘scandalize’
immunizirovat’ ‘immunize’ tamponirovat’ ‘plug’
individualizirovat’ temperirovat’ ‘temper’
‘individualize’
indossirovat’ ‘endorse’ terrasirovat’ ‘terrace’
inkorporirovat’ ‘incorporate’ vulkanizirovat’
‘vulcanize’
inkriminirovat’ ‘incriminate’ vulkanizovat’ ‘vulcanize’
Again there is a difference between the Bi-aspectual and Imperfective

borrowed verbs in Russian. Looking only at well-attested (over 1000
“hits”) verbs, among Bi-aspectuals, 70 verbs yield zero po- forms, whereas
only 3 verbs have over 1000 po- forms. Among robustly attested Imperfec-
tive verbs there is the opposite pattern: whereas only 5 verbs yield zero po-
forms, 26 verbs return over 1000 po- form “hits”.
Given this data, I would like to argue that a prototypical Bi-aspectual
borrowed verb lacks Complex Act (and therefore Single Act) Perfectives,
whereas Complex Act Perfectives are commonly found among Imperfec-
tive borrowed verbs. More concretely, a typical Bi-aspectual verb like dez-
ertirovat’p/i ‘desertp/i’ cannot form a Complex Act Perfective; there is NO
*podezertirovat’p *‘desertp (for a while)’. But a typical Imperfective bor-
102 Laura A. Janda
rowed verb like flirtovat’i ‘flirti’ can form a Complex Act Perfective; pof-
lirtovat’p ‘flirtp (for a while)’ exists and is well-attested.
Table 2(b). High-frequency Imperfective verbs with with either zero or over 1000
po- prefixed hits
Imperfective verbs with over 1000 unprefixed hits
Zero po- prefixed hits: 5 Over 1000 po- prefixed hits: 26
fetišizirovat’ ‘fetishize’ aplodirovat’ ‘applaud’ manipulirovat’
‘manipulate’
fraxtovat’ ‘freight’ buksovat’ ‘skid’ marinovat’ ‘marinate’
legirovat’ ‘alloy’ èksperimentirovat’ masturbirovat’
‘experiment’ ‘masturbate’
vualirovat’ ‘veil’ èkspluatirovat’ ‘exploit’ praktikovat’/sja
‘practice’
èrodirovat’ ‘erode’ fantazirovat’ ‘fantasize’ psixovat’ ‘act flaky’
fextovat’ ‘fence’ redaktirovat’ ‘edit’
flirtovat’ ‘flirt’ regulirovat’
‘regulate’
improvizirovat’ šantažirovat’
‘improvise’ ‘blackmail’
interesovat’/sja šinkovat’ ‘chop’
‘interest’
intrigovat’ ‘scheme’ spekulirovat’
‘speculate, gamble’
ironizirovat’ ‘mock’ tancevat’ ‘dance’
komandovat’ ‘be in trenirovat’ ‘train’
charge’
kritikovat’ ‘criticize’ žonglirovat’ ‘juggle’
It would not be possible to discuss all 555 borrowed verbs in this sur-
vey, but Table 2 provides us with an opportunity to look more closely at a
more manageable (and probably more representative, due to the fact that all
these items are high-frequency, familiar verbs) fraction of the data. The
high-frequency Bi-aspectual verbs in the left column do not form Complex
Act Perfectives. It is fairly easy to see how this is motivated by the mean-
ings of these verbs. These Bi-aspectual verbs for the most part describe
actions that one cannot engage in without having an effect and/or making
progress toward a goal. One can’t just do these things for a while and stop
and have no result. These activities are strongly telic. For example, one
can’t spend some time guillotining (gil’otinirovat’p/i) or immunizing (im-
munizirovat’p/i) without having guillotined or immunized someone. There
are several discernable, yet overlapping, groups of verbs here, most of

which are transitive and all of which involve a goal-directed change:
Table 3. Semantic groupings of high-frequency Bi-aspectual verbs

Meanings High-frequency verbs
Covering a sur- gummirovat’p/i ‘coat with rubberp/i’, meblirovat’p/i ‘upholsterp/i’,
face metallizirovat’p/i ‘coat with metalp/i’, ornamentirovat’p/i ‘orna-
mentp/i’, plakirovatp/i ‘platep/i’, satinirovat’p/i ‘polishp/i’
Removal demaskirovat’p/i ‘unmaskp/i’, deblokirovat’p/i ‘unblockp/i’,
demilitarizirovat’p/i ‘demilitarizep/i’, demobilizirovat’p/i
‘demobilizep/i’, denacionalizirovat’p/i ‘privatizep/i’, dezertirovat’p/i
‘desertp/i’
Physical change denaturirovat’p/i ‘denaturep/i’, gofrirovat’p/i ‘crimpp/i’,
of state granulirovat’p/i ‘granulatep/i’, kristallizovat’sjap/i ‘crystallizep/i’,
temperirovat’p/i ‘temperp/i’, vulkaniz(ir)ovat’p/i ‘vulcanizep/i’
Cul- anglizirovat’p/i ‘anglicizep/i’, evropeizirovat’p/i ‘europeanizep/i’,
tural/Linguistic germanizirovat’p/i ‘germanifyp/i’, internacionalizirovat’p/i ‘inter-
change of state nationalizep/i’, latinizirovat’p/i ‘latinizep/i’, dešifrirovat’p/i ‘deci-
pherp/i’
Arrangement decentralizovat’p/i ‘decentralizep/i’, dezorganizovat’p/i
Organization ‘disorganizep/i’, èšelonirovat’p/i ‘echelonp/i’, flankirovat’p/i
‘flankp/i’, frakcionirovat’p/i ‘fractionatep/i’, kollektivizirovat’p/i
‘collectivizep/i’
Budgetary assignovat’p/i ‘allocatep/i’, debetovat’p/i ‘debitp/i’,
arrangement kapitalizirovat’p/i ‘reinvestp/i’, monopolizirovat’p/i ‘monopolizep/i’
Proclamation dekretirovat’p/i ‘decreep/i’, denonsirovat’p/i ‘denouncep/i’,
dezavuairovat’p/i ‘disavowp/i’, indossirovat’p/i ‘endorsep/i’,
inkriminirovat’p/i ‘incriminatep/i’, kanon(iz)irovat’p/i ‘canonizep/i’,
ratificirovat’p/i ‘ratifyp/i’
Getting something covered, removed, changed, or arranged – all these

are good examples of the metaphorA COMPLETABLE ACTION IS TRAVEL TO
A DESTINATION. It is hard (perhaps impossible) to construe these actions as
Non-Completable. The lexical semantics of these verbs gives them an as-
pectual profile that excludes the formation of Complex Act Perfectives, an
effect which can be measured.
All three Bi-aspectual verbs in Table 1 with over 1000 po- forms raise
suspicions. Massirovat’p/i has two meanings, ‘amassp/i’ and ‘massagep/i’.
Although both meanings are listed as Bi-aspectual by Wheeler 1972/1992,
the second one (which also appears to be vastly more numerous), is listed
as an Imperfective in Ožegov 1949/1989. It was not possible to manually
104 Laura A. Janda
verify all attestations, but it appears that the po- forms of massirovat’? are
all (or nearly all) based on the ‘massage’ meaning, which is probably not
Bi-aspectual (Ožegov 1949/1989 notes that the Bi-aspectual use is obso-
lete). Čertkova and Čang specifically cite rekomendovat’/sjai ‘recommendi’
as an example of a verb that is erroneously entered in dictionaries as Bi-
aspectual despite the fact that it is really an Imperfective. Remontirovat’i
‘renovatei’ is likewise listed by Ožegov1949/1989 as an Imperfective.
Among the Imperfective verbs in Table 2, two are listed as Bi-
aspectual by Ožegov 1949/1989: legirovat’p/i ‘alloy p/i’, which yielded no
po- forms, and improvizirovat’p/i ‘improvizep/i’, which yielded over 1000
po- forms. The Imperfective verbs with over 1000 po- forms behave much
like typical native Russian simplex Imperfectives. In addition to readily
forming Complex Act Perfectives, a number of them form other types of
Perfectives in various combinations:
Natural Perfectives: sfantazirovat’p ‘fantasizep’, skomandovat’p ‘give a
commandp’, zamarinovat’p ‘marinatep’, otredaktirovat’p ‘editp’, ureguli-
rovat’p (also ot-, za-) ‘regulatep’, stancevat’p ‘dancep’, natrenirovat’p
‘trainp’
Specialized Perfectives: podregulirovat’p ‘adjust, resetp’, raskritikovat’p
‘tear to piecesp’, vytancevat’p ‘obtain by dancingp’
Single Act Perfectives: psixanut’p ‘crack upp (once)’ (from psixovat’i
‘act flakyi’), spekul’nut’p ‘gamblep (once)’ (from spekulirovat’i ‘gamblei’)
Semantically the typical Imperfective verbs describe human behaviors
associated with various social and professional settings. Most of these
verbs can be used intransitively, and typically these are verbs that can be
used in expressions like on po-X-oval nemnogo i brosil ‘he X-ed for a
while and then stopped’, implying no result. The verbs in this group do not
fit into the subtypes listed for typical Bi-Aspectual verbs above. On the
contrary, these verbs are easily construable as Non-Completable.
The trends in this empirical study are strong and compelling enough to
outweigh whatever shortcomings there may be in the data. Clearly the Im-
perfective and Bi-aspectual borrowed verbs of Russian are very different
groups of verbs, both objectively (quantified in attestations of Complex
Act po- Perfectives), and subjectively (given the semantic groups of verbs
observed). Brief conclusions follow in section 6.
6. Conclusions
Borrowed verbs in Russian exist as both Bi-aspectual and Imperfective

verbs. Both types of borrowed verbs are well-attested in Russian and each
type has its own distinctive semantic and derivational profile. The bor-
rowed Imperfective verbs behave much like other Imperfective verbs in
Russian, and derive all four types of Perfectives. The borrowed Imperfec-
tives show a strong tendency to derive Complex Act Perfectives, in particu-
lar delimitatives prefixed in po-. Semantically, the borrowed Imperfectives
describe behaviors characteristic of people in various social settings. Most
Russian Bi-aspectual verbs are borrowed, and these verbs are restricted in
their derivational morphology because they do not form Complex Act and
Single Act Perfectives. They tend to refer to actions that are construable
only as Completable, involving changing the state or arrangement of some-
thing.
Given these facts, we can speculate on what role the lexical semantics
of a borrowed verb likely play in determining whether a borrowed verb
will be recognized as a Bi-aspectual or a more ordinary simplex Imperfec-
tive verb. Recall that when a foreign verb arrives in the Russian lexicon, it
starts out with no aspectual morphology, nor any accompanying cluster of
aspectually related verbs that might disambiguate its aspect. As a new im-
migrant to Russian, a pilgrim verb has only its lexical meaning at the out-
set. A critical test for a new verb is whether it can describe Completable
and Non-Completable actions, as motivated by the metaphor: A COMPLET-
ABLE ACTION IS TRAVEL TO A DESTINATION. Verbs with strongly telic
meanings can only describe Completable actions. The fact that these verbs
necessarily describe results (or progress toward results) makes it possible
to use them in Perfective contexts even without Perfectivizing morphology.
Such verbs are excellent candidates for Bi-aspectuality, although they may
later acquire the morphology of the Russian aspectual system. Verbs that
have a Non-Completable interpretation are appropriate in Imperfective
contexts, and will usually require Perfectivizing morphology for Perfective
use. Such verbs enter the lexicon as Imperfectives.
Notes
* The author would like to thank John Korba for collecting data and Chris
Wiesen for assistance with statistical analysis. Thanks are also due to the re-
viewers for this volume, to Tore Nesset who commented on an earlier version
106 Laura A. Janda
of this article, and to Alexander Berdichevsky who spotted an error in the data.
All remaining imperfections are to be attributed to the author.
1. Although nearly all scholars agree that Perfective is semantically marked in
Russian, Galton 1976 gives the opposite assignment and Padučeva 1996 con-
siders Perfective vs. Imperfective to be an equipollent rather than asymmetri-
cal relationship.
2. Dahl (1985: 71–72) and Smith (1991: 277) point out this typological correla-
tion concerning the markedness values of Perfective and Imperfective.
3. This list simplifies the picture a bit, citing only the affixes relevant to Bi-
aspectual and foreign verbs. Švedova et. al. 1980 lists 17 Perfectivizing
prefixes, but only the 11 given here are attested with Bi-aspectual verbs by
Avilova (1968) and Čertkova and Čang (1998), and they are given in descend-
ing order of frequency. These prefixes are those common to both Bi-aspectual
and non-Bi-aspectual verbs in Russian.
4. 85 of the 428 verbs listed as Imperfective only in Wheeler 1972/1992 have
foreign origins. This figure includes 29 verbs other than those in the study de-
scribed in section 4, since it includes items with suffixes other than -ovat’,
such as špionit’i ‘be a spyi’ (which is not strictly a borrowed verb, but is built
from a borrowed noun).
References
Anderson, Cori
2002 Biaspectual Verbs in Russian and their Implications on the Category
of Aspect. Honors Thesis, University of North Carolina-Chapel Hill.
Avilova, Natal’ja S.
1968 Dvuvidovye glagoly s zaimstvovannoj osnovoj v russkom literatur-
nom jazyke novogo vremeni [Bi-aspectual verbs with foreign stems
in the modern Russian literary language]. Voprosy jazykoznanija:
66–78.
1976 Vid glagola i semantika glagol’nogo slova [Verbal aspect and verbal
semantics]. Moscow: Akademija nauk SSSR.
Bertinetto, Pier Marco, and Denis Delfitto
2000 Aspect vs. actionality: Why they should be kept apart. In Tense and
Aspect in the Languages of Europe, Östen Dahl (ed.), 189–225. Ber-
lin: Mouton de Gruyter.
Bondarko, Aleksandr V.
1971 Vid i vremja russkogo glagola [Russian verbal aspect and tense].
Moscow: Prosveščenie.
1983 Principy funkcional’noj grammatiki i voprosy aspektologii [Princi-

ples of functional grammar and issues of aspectology]. Leningrad:
Nauka.
Comrie, Bernard
1976 Aspect. Cambridge: Cambridge University Press.
Čertkova, Marina Jurevna
1996 Grammatičeskaja kategorija vida v sovremennom russkom jazyke
[The grammatical category of apsect in modern Russian]. Moscow:
Moscow State University.
Čertkova, Marina Jurevna, and Pej-či Čang
1998 Evoljucija dvuvidovyx glagolov v sovremennom russkom jazyke
[The evolution of bi-aspectual verbs in modern Russian]. Russian
linguistics (22): 13–34.
Dahl, Östen
1985 Tense and aspect systems. Oxford: Basil Blackwell.
Dickey, Stephen M.
2000 Parameters of Slavic Aspect. A Cognitive Approach. Stanford: CSLI
Publications.
2006 Aspectual pairs, goal orientation and po- delimitatives in Russian.
Glossos (7): http://seelrc.org/glossos.
Durst-Andersen, Per
1992 Mental Grammar: Russian Aspect and Related Issues. Columbus,
OH: Slavica.
Forsyth, John
1970 A grammar of aspect: Usage and meaning in the Russian verb.
Cambridge: Cambridge University Press.
Galton, Herbert
1976 The main functions of the Slavic verbal aspect. Skopje: Macedonian
Academy of Sciences and Arts.
Gladney, Frank Y.
1982 Biaspectual verbs and the syntax of aspect in Russian. Slavic and
East European Journal (26): 202–215.
Glovinskaja, Marina Ja.
2001 Mnogoznačnost’ i sinonimija v vido-vremennoj sisteme russkogo
glagola [Polysemy and synonymy in the tense-aspect system of the
Russian verb]. Moscow: Russkie slovari.
Isačenko, Aleksandr V.
1960 Grammatičeskij stroj russkogo jazyka v sopostavlenii s slovackim.
Morfologija, čast vtoraja [The grammatical system of Russian in
comparison with Slovak. Morphology, part two]. Bratislava:
Vydavatel’stvo Slovenskej akadémie vied.
108 Laura A. Janda
Jakobson, Roman O.
1957/1971 Shifters, verbal categories, and the Russian verb. In Selected
Writings II, 130–147. The Hague: Mouton.
Janda, Laura A.
2004 A metaphor in search of a source domain: the categories of Slavic
aspect. Cognitive Linguistics (15): 471–527.
forthc. a Aspectual clusters of Russian verbs. Studies in Language.
forthc. b Inflectional Morphology. In Handbook of Cognitive Linguistics, Dirk
Geeraerts and Hubert Cuyckens (eds.). Oxford: Oxford University
Press.
Jászay, László
1999 Vidovye korreljaty pri dvuvidovyx glagolax [The aspectual
correlates of bi-aspectual verbs]. Studia Russica (17): 169–177.
Lakoff, George.
1987 Women, fire, and dangerous things. Chicago: University of Chicago
Press.
Maslov, Jurij S.
1965 Sistema osnovnyx ponjatij i terminov slavjanskoj aspektologii [The
system of basic concepts and terms in Slavic aspectology]. Voprosy
obščego jazykoznanija: 53–80.
Mazon, André
1914 Emplois des aspects du verbe russe. Paris: Librairie ancienne Honoré
Champion.
Mehlig, Hans Robert
1994 Gomogennost’ i geterogennost’ v prostranstve i vremeni
[Homogeneity and heterogeneity in space and time]. Revue des
etudes Slaves (66): 595–606.
1997 Nekotorye zamečanija po povodu opisanija kategorii vida v russkom
jazyke [Some notes on the description of the category of aspect in
Russian]. Russian Linguistics (21): 177–193.
Mučnik, I. P.
1966 Razvitie sistemy dvuvidovyx glagolov v sovremennom russkom
jazyke [The development of the system of bi-aspectual verbs in
modern Russian]. Voprosy jazykoznanija: 61–75.
Ožegov, Sergej I.
1989 Reprint. Slovar’ russkogo jazyka [Dictionary of Russian]. Moscow:
Russkij jazyk. Original edition, 1949.
Padučeva, Elena Viktorovna
1996 Semantičeskie issledovanija [Semantic investigations]. Moscow:
Jazyki russkoj kul’tury.
Schooneveld, Cornelis H. van

1978 Semantic transmutations: Prolegomena to a calculus of meaning.
Vol. 1: The cardinal semantic structure of prepositions, cases, and
paratactic conjunctions in contemporary standard Russian.
Bloomington: Physsardt.
Smith, Carlota S.
1991 The Parameter of Aspect. Dordrecht: Kluwer.
Šaxmatov, A. A.
1941 Sintaksis russkogo jayzka [Russian syntax]. Leningrad: Učpedgiz.
Švedova, Natal’ja Ju.
1980 Russkaja grammatika [Russian grammar]. Moscow: Nauka.
Tatevosov, Sergej
2002 The parameter of actionality. Linguistic Typology (6): 317–401.
Timberlake, Alan
1982 Invariance and the syntax of Russian aspect. In Tense-aspect: Be-
tween semantics and pragmatics, Paul Hopper (ed.), 305–331. Am-
sterdam: John Benjamins.
2004 A Reference Grammar of Russian. Cambridge: Cambridge Univer-
sity Press.
Vinogradov, Viktor V.
1938 Sovremennyj russkij jazyk. Grammatičeskoe učenie o slove [Modern
Russian. Grammatical studies of the word]. Moscow: Učpedgiz.
1972 Russkij jazyk [Russian], 2nd ed. Moscow: Vysšaja škola.
Wheeler, Marcus
1992 Reprint. The Oxford Russian-English Dictionary. Oxford: Clarendon
Press. Original edition, 1972.
Zaliznjak, Anna A.
1977 Grammatičeskij slovar’ russkogo jazyka [Grammatical dictionary of
Russian]. Moscow: Russkij jazyk.
Zaliznjak, Anna A., and Aleksej D. Šmelev
2000 Vvedenie v russkuju aspektologiju [Introduction to Russian
aspectology]. Moscow: Jazyki russkoj kul’tury.
Perfectives, imperfectives and the Croatian present
tense
Renata Geld and Irena Zovko Dinković
Abstract
This paper attempts to re-examine the nature of the Croatian present tense in
the light of the cognitive linguistic view of grammatical structure. In order to
account for what Croatian traditional grammars describe as the usage of the
present tense for the so-called ‘non-real present’, we will use the naive char-
acterization approach suggested by Langacker (1991: 250). We will con-
sider the following issues: tense as an epistemic category, the nature of per-
fective and imperfective processes (cf. e.g. Langacker 2001b) and the
relevance of this classification of process types for aspectual phenomena in
Croatian, as well as the role of extreme subjectification (Langacker 2003:
13) in the characterization of non-temporal uses of the Croatian present
tense. Rather than hypothesizing distinct meanings for the present tense
morpheme, we recognize the present tense morpheme as a consistent marker
of proximity, indicating immediacy of the process in relation to the speaker.
In doing so, we will discuss the high degree of subjectivity present in cases
where the speaker describes ‘virtual processes’ that occur as a result of her
conceptualizing activity. Finally, we will demonstrate how general cognitive
models proposed for examining the grammatical category of tense can yield
situation-specific models that incorporate process types, as well as aspects
of viewing arrangement. Having taken into consideration the nature of proc-
ess types, the epistemic value of the present tense marker and specific depar-
tures from the canonical viewing arrangement (Langacker, 2003:18), we ar-
gue that non-temporal usages of the Croatian imperfective and perfective
present are as ‘real’1 as their temporal usages, or more specifically, that, just
like in English, their non-temporal or non-present uses can be accounted for
in terms of the inherent value of indicating present time (cf. Langacker
2001b: 267).2
Keywords: present tense, Croatian, perfectives, imperfectives, extreme

subjectification, virtual reality
112 Renata Geld and Irena Zovko Dinković
1. Introduction
1.1. The problem of the Croatian present tense
Croatian grammars (Katičić 1991: 46–48; Barić et al. 2003: 408–409)

stress that ‘real’ present time can only be related to imperfective verbs and
their specific use in indicating events occurring at the time of speaking. All
other uses of both perfective and imperfective present are related to the so-
called ‘non-real’ present. Thus, the Croatian present tense is said to be
‘real’ only when used temporally; other uses of the present tense are classi-
fied as ‘non-real’ present. As Langacker observes in his characterization of
the English present tense,3 the traditional view suggests that “the present
tense can be used for anything but the present time” (2001b:251, original
emphasis). The same restrictive view is given in the above-mentioned de-
scription of the Croatian present tense.
Thus, the central argument concerning the nature of the Croatian pre-
sent tense, which closely corresponds to the one suggested for the English
present tense, is that its usage largely indicates ‘non-real’ present time. The
Croatian present tense indicating ‘real’ present time is restricted to imper-
fective verbs marked for the present tense and designating processes occur-
ring at the time of speaking. Examples (1)–(7) below illustrate the use of
both imperfective and perfective present.4 We will refer to the present
tense forms of perfective and imperfective verbs as perfective and imper-
fective present respectively. Let us consider the examples (1a-c):5
(1) a. Barbara gradi-Ø dvorac-Ø od pijesk-a.

Barbara build.IMPF.3SG.PRES castle.ACC of sand.GEN
‘Barbara is building a sand castle.’
b. *Barbara izgradi-Ø dvorac- Ø od pijesk-a.

Barbara build.PERF.3SG.PRES castle.ACC of sand.GEN
‘Barbara builds a sand castle.’
c. Svako ljeto Barbara izgradi-Ø

every summer Barbara build.PERF.3SG.PRES
dvorac- Ø od pijesk-a.
castle.ACC of sand.GEN
‘Every summer Barbara builds a sand castle.’
Perfectives, imperfectives and the Croatian present tense 113
d. Svako ljeto Barbara gradi-Ø

every summer Barbara build.IMPF.3SG.PRES
dvorac- Ø od pijesk-a.
The imperfective present (1a) is perfectly acceptable. The perfective

present (1b) cannot be used in an independent clause unless it is accompa-
nied by an adverbial of frequency that makes the event iterative or habitual,
as in (1c). If the same adverbial of frequency is used with the imperfective,
as in (1d), the event is also construed as habitual. This habitual meaning of
the present evident in (1c,d) seems to contradict the standard definition of
the present tense. Examples (1c) and (1d), together with others that follow,
illustrate what different authors have labeled, for their own analytical pur-
poses, non-temporal, non-canonical or relative usage of the present tense
(cf. Barić et al. 2003; Comrie 1985; Cutrer 1994; Katičić 1991; Langacker
1991). The difference in meaning between examples (1c) and (1d), as well
as other examples provided in this section, will be discussed in detail in
section 3.
Both the perfective and the imperfective present can be used in clauses
introduced by the subordinator kad ‘when’, either to express habitual ac-
tions (2a, b, c) or future events (2d):
(2) a. Kad ode-Ø u kin-o, Iris

when go.PERF.3SG.PRES to cinema.LOC Iris
jede-Ø kokic-e.
eat.IMPF.3SG.PRES popcorn.ACC
‘Whenever she goes to the cinema, Iris eats popcorn.’
b. Kad ode-Ø u kin-o, Iris

pojede-Ø kokic-e.
eat.PERF.3SG.PRES popcorn.ACC
c. Kad ide-Ø u kin-o, Iris

when go.IMPF.3SG.PRES to cinema.LOC Iris
jede-Ø kokic-e.
d. Kad ode-Ø u kin-o, Iris

će jesti6 kokic-e.
eat.IMPF.3SG.FUT popcorn.ACC
‘When she goes to the cinema, Iris will eat popcorn.’
Both the imperfective and the perfective present are possible in hypo-
thetical situations located at different points on the realis–irrealis contin-
uum, as illustrated by (3a, b) below. In an example such as (3a) all four
combinations are acceptable: (i) the perfective present tense in both the
main and the subordinate clause, (ii) the imperfective present tense in both
clauses, (iii) the perfective present tense in the subordinate and the imper-
fective present tense in the main clause, and finally (iv) the reverse. All
four examples will be discussed in greater detail in section 3.
(3) a. Ako uči-m/nauči-m,

if study.IMPF/PERF.1SG.PRES
prođe-m/prolazi-m na ispit-u.
pass.PERF/IMPF.1SG.PRES on exam.LOC
‘If I study, I pass the exam.’
b. Da uči-m/nauči-m,
prošao bih7 na ispit-u.
pass.PERF.1SG.M.COND on exam.LOC
‘If I studied, I would pass the exam.’
Both subordinators ako and da introduce hypothetical “mental spaces”

(cf. Fauconnier 1997; Sweetser 1996). The conditional meaning in the
protasis reflects the speaker’s epistemic stance (Fillmore 1990), or more
precisely, the degree to which the situation described in the protasis is, in
the speaker’s eyes, dissociated from the real state of affairs. Offering the
relevant arguments for our hypothesis later in this paper, we wish to sug-
gest that the prototypical value of the Croatian present tense marker is to
consistently “place” the process close to the base space of reality. Further-
more, we believe that in Croatian, just like in other languages (Fauconnier
1997), mental spaces in discourse may be evoked in two different ways:

either via an overt space builder or more tacitly. In both cases the result is a
virtual world8 in which the epistemic value of the present tense is realized
relative to that world. The space builders that overtly establish mental
spaces are, for example, the subordinators kad ‘when’, ako ‘if’ and da ‘if’
(see (2a,b,c,d) and (3a, b)), whereas tacitly indicated mental spaces are the
result of other clausal and/or contextual elements. For example, the perfec-
tive present in the relative clause in example (4) is possible because of the
tacit mental space indicated by this clause and its relation to the main
clause. The relative pronoun koju indicates a new space and the relative
clause seems to involve a degree of condition-like epistemic stance on the
part of the speaker.
(4) Nikada neću baciti knjig-u koj-u

never not-throw.PERF.1SG.FUT book.ACC which.ACC
kupi-m.
buy.PERF.1SG.PRES
‘I will never throw away the book I buy.’
The perfective present can be used to express future time but only when
it is preceded by a modal adverb:
(5) Možda kupi-m nov-u knjig-u.

maybe buy.PERF.1SG.PRES new.ACC book.ACC
‘I might buy a new book.’
The adverb možda establishes a particular mental space that indicates

the possibility that the action that follows will be performed. Conditions
that would need to be fulfilled for the potentiality to turn into reality re-
main tacit. The imperfective present, on the other hand, can be used to
indicate future without being accompanied by možda. It expresses a fixed
plan whose realization is certain to take place some time in future. How-
ever, the imperfective present is frequently accompanied by a time expres-
sion specifying the exact time of the anticipated event:
(6) Dolazi-m rano ujutro.

come.IMPF.1SG.PRES early in-the-morning
‘I am coming early in the morning.’
Both the imperfective and the perfective present can be interpreted as

the so-called ‘historical present’:
(7) Dolazi-m/Dođe-m kuć-i i

come.IMPF/PERF.1SG.PRES home.DAT and
vidi-m da nema-m ključev-e.
see.1SG.PRES that not-have.1SG.PRES keys.ACC
‘I come home and I see that I have no keys.’
The imperfective present is acceptable with performatives, whereas the

perfective present is not. Let us consider the last example:
(8) Obećava-m da ću uči-ti više.

promise.IMPF.1SG.PRES that study.IMPF.1SG.FUT more
‘I promise I will study more.’
The above examples reveal that the Croatian present is indeed used and
favoured in a number of contexts which are traditionally not identified as
expressing ‘real’ present time. However, we would like to propose that the
cognitive, non-objectivist, view of semantic structure provides a frame-
work within which we can attempt to reanalyze the instances of the so-
called ‘non-real’ Croatian present as designating present “conceptual oc-
currences” (Langacker 2003: 15, original emphasis), thereby defending the
idea that the Croatian present is consistently used to locate the process at
the time of speaking.9 In order to avoid terminological confusion, we
should mention that the term ‘non-real present’ refers to ‘non-temporal’
uses of the present tense.
In the following section, we will present some relevant features of the
Croatian tense and aspect, relying on the descriptions offered by Croatian
grammarians. Then, in section 2, we will give a short description of Lan-
gacker’s epistemic models, the notion of epistemic grounding, the classifi-
cation of process types, and the notions of canonical viewing arrangement
and subjectification, that is the fundamental models and ‘tools’ within the
theory of cognitive grammar necessary for a fresh analysis of the Croatian
present. Finally, in section 3, we will use the examples from the introduc-
tion to suggest a new characterization of the Croatian present tense guided
by relevant cognitive linguistic assumptions.
1.2. General remarks on Croatian tense and aspect – the traditional view
Croatian verbs are marked for person, number, gender, tense, aspect, mood
and voice. Since the categories of tense and aspect are crucial for our
analysis, we will give a brief overview of the traditional definition of these
two categories in Croatian grammars.
1.2.1. Tense
Croatian has three synthetic tenses, i.e. one present tense and two past
tenses (imperfekt and aorist), and four analytic ones, i.e. the perfect, plu-
perfect, future and perfective future. Tenses operate in a framework that
suggests a threefold division of time into past, present and future. Since the
Croatian term “vrijeme” is used to denote both time and tense, it is difficult
to tell the two apart in traditional grammatical descriptions. Katičić simply
states that the category of “vrijeme” can be grammatically marked as pre-
sent, past and future (Katičić 1991: 45), and makes a further opposition
between so-called completeness and incompleteness, a distinction that
differs from the aspectual distinction. Thus, he suggests the following sys-
tem: present, past, future; and complete present, complete past, complete
future.10
1.2.2. Aspect
In the introduction to the second volume of his contrastive study on Serbo-

Croatian and English grammar, Filipović (1978: 7) categorizes verbal as-
pect and word order as “the most difficult Serbo-Croatian grammatical
features for English-speaking students”. Perfective verbs are generally
described as verbs that express the action as a whole, and imperfective
verbs as verbs designating processes which can be divided into phases. The
basic meaning of the perfective is to “represent a single concrete action”,
whereas the basic function of the imperfective is to “denote a gradual, con-
crete process and to express duration” (Silić 1978: 43). Another factor that
is stressed as crucial in the interpretation of aspect is context. Basic aspec-
tual meanings are the least conditioned by context, while other meanings
are much more context-dependent. For example, the basic meaning of the
perfective, i.e. to represent a single concrete action, is least conditioned by
the context (Silić 1978: 43). In order to express, for instance, a repetitive
meaning, the perfective verb is accompanied by an appropriate adverbial.

The basic function of the imperfective is to denote a stable process. Its
other meanings, e.g. iterativity, are determined by the context.
Perfectivization and imperfectivization are realized by suppletion, suf-
fixation or prefixation and the three processes result in the change of ver-
bal aspect. Suppletion is evident in aspectual pairs such as reći ‘say’ and
govoriti ‘talk’. The same mechanism is claimed to be present in aspectual
pairs in which perfective and imperfective verbal aspect is formed from
different root morphemes, as in prići/prilaziti ‘approach’ or otići/odlaziti
‘leave’ (Barić et al. 1995). It is important to stress that suppletion is not
productive and can be found in only a few verbal pairs. On the other hand,
suffixation and prefixation are very productive patterns. The opposition
expressed by suffixation (-ije/-a, -i /-ava, Ø/-a, -e/-a, -je/-a, -je/ijeva, -i/-a,
-i/-ja, -i/-iva, -i/-jiva, -i/-ava, etc.) serves to indicate the change of perfec-
tives into imperfectives, as in, for example, vidjeti/viđati ‘see’ or ba-
citi/bacati ‘throw’. It is important to bear in mind, however, that in such
pairs aspectual difference is often coupled with the change in the manner
the action is performed. The mechanism of prefixation is also very produc-
tive. Imperfective verbs are perfectivized by a number of perfectivizing
prefixes: bosti/ubosti ‘stab’, čitati/pročitati ‘read’, govoriti/progovoriti
‘speak’, and others.11 Which prefix is used depends on the semantics of the
verb. Similarly to suffixation, prefixation is responsible for other changes
besides aspectual ones. For example, in the aspectual pair bacati/nabacati
perfectivity is signaled by the prefix na-, but the same prefix adds the nu-
ance of determining the direction of the action (Barić et al 1995). In the
same way the prefix pro- in the pair govoriti/progovoriti makes the latter
verb ingressive.
Both perfectives and imperfectives are classified in terms of ‘aktion-
sarten’, which typically results in a long list of verb groups such as ingres-
sive, completive, serial, resultative, distributive, relational, intensive, in-
choative, and so on (Silić 1978; Weber 1978; Barić et. al 1995). In section
2.3. some of these notions will be further described in terms of their rela-
tion to Langacker’s (1987) classification of process types.
Before we proceed with a more detailed analysis of the Croatian present
tense examples, let us, in the next section, consider those fundamentals of
cognitive grammar that are essential for re-examining the meaning of the
Croatian present, as well as the basic framework within which the English
present tense has been redefined.
2. Conceptual content and construal
According to cognitive grammar, lexicon and grammar form a continuum,

and all grammatical and lexical elements are tools for construing concep-
tual content. Therefore, an analysis of grammar should start with a ques-
tion of what a certain element does, that is, in what way it structures a par-
ticular content. We shall consider the following: a) the role of the present
tense morpheme, b) the nature of the Croatian verbal aspect, as well as its
relation to process types and c) the importance of the speaker as the con-
ceptualizer.
2.1. Epistemic value of the present tense morpheme
In discussing finite clauses Langacker (1991: 244–251) pays special atten-

tion to the notion of clausal g r o u n d i n g , that is, situating the profiled
event relative to the ground. The g r o u n d is defined as the speech event,
its participants and its immediate circumstances. Elements that effect
clausal grounding in English are those indicating tense and modality. There
are three cognitive models that are used to describe the conceptual import
of tense and modality: the basic epistemic model, the elaborated epistemic
model and the time-line model (Langacker 1991: 242–245).
The b a s i c e p i s t e m i c m o d e l (see Figure 1a) pertains to what a per-
son knows or thinks she knows. A particular conceptualizer accepts certain
situations as real. These situations constitute her k n o w n r e a l i t y . Every-
thing other than known reality is i r r e a l i t y . The basic epistemic model is
extended into the e l a b o r a t e d e p i s t e m i c m o d e l (see Figure 1b)12 in
which the known reality is surrounded by a large region of u n k n o w n
r e a l i t y . Unknown reality incorporates the situations whose reality the
conceptualizer “suspects or contemplates but does not accept as having
been established; and those of which she is entirely ignorant” (Langacker
1991: 243). The region closest to the conceptualizer’s vantage point is
called i m m e d i a t e r e a l i t y . The other relevant model that originates in
the basic epistemic model is t h e t i m e – l i n e m o d e l (see Figure 1c). The
time–line model incorporates time (t), i.e., the axis representing the dynam-
ics of reality, and the ground (G), which is determined by the speech event.
The speech event is the central point of immediate reality and the concep-
tualizer’s vantage point. The squiggly line represents the speech event co-
inciding with the present.
(a)
Irreality
(Known) reality Immediate
reality
(b) (c)
Non-Reality
Unknown reality
Known reality
Immediate
reality
C G
Past Present Future
Figure 1. The basic epistemic model, the elaborated epistemic model and the time
line model (taken from Langacker 1991: 242–244)
Within the basic epistemic model, the English tense morphology is de-
scribed in terms of the notion of proximity. Thus, instead of “present” vs.
“past” we speak of a “p r o x i m a l / d i s t a l contrast in the epistemic sphere”
(Langacker 1991:245). The conceptual import of the zero tense morpheme
is i m m e d i a c y and the import of its marked counterpart is n o n -
i m m e d i a c y . Thus, the proximal morpheme contributes to the process
being construed as immediate to the speaker, whereas the distal morpheme
always conveys non-immediacy.
The proposed models offer three notions that will prove to be instru-
mental in the description of the Croatian present tense: ground(ing) and the
conceptualizer’s vantage point, the notion of proximity in the epistemic
sphere and the notion of reality. In the course of the work these notions
will be used and contrasted in terms of their relation to other relevant no-
tions: ground and its relation to the profiled process (Langacker 1990,
2003) in section 2.5., the notion of actual reality and its relation to its vir-
tual counterpart in section 3.2., and proximity as a prototypical value of the
Croatian present tense morpheme throughout section 3., especially in

3.2.3., 3.2.4., and 3.2.5.
2.2. Epistemic grounding
Epistemic grounding distinguishes finite clauses from non-finite ones

(Langacker 1987). An entity, a thing or a process is grounded when its
location is determined in relation to the participants in the speech event
and their spheres of knowledge. At the schematic level (see Figure 1a),
time per se is not a factor. In English, there are two relevant oppositions:
the presence vs. the absence of a modal and the presence vs. the absence of
the “past tense” morpheme. The absence of modals signals the speech
event participants’ recognition of what is profiled as part of established
reality, and, as already mentioned before, the conceptual import of the zero
tense morpheme is immediacy. Therefore, we can conclude that the zero-
marked grounding predication derives its primary value with reference to
the basic epistemic model. Immediate reality coincides with the time of
speaking (see Figure 1c), so present time is conceived as an indissociable
aspect of immediate reality and the predication of immediate reality is in-
terpreted as one of present time (Langacker 1991: 246). Analogous charac-
terization can be used for Croatian. The Croatian “present tense” mor-
pheme represents the option in which the profiled process is directly
accessible to the speaker, whereas the perfekt13 marks the separation of the
process from the speaker, which happens to be an obvious case of iconic-
ity. In short, even though in both English and Croatian the fundamental
semantic characterization of the zero or “present tense” morpheme is re-
lated to epistemic immediacy, it is prototypically construed with reference
to the time-line model (see Figure 1c). Thus, we speak of a present tense.
Generally, if the content verb within the clausal head stands alone, it is
inflected for tense and it profiles a process which is grounded. An example
is given in (9):
(9) a. She swims very well. (English)

b. Jako dobro pliva. (Croatian)
The semantic content of the verb plivati ‘swim’ can be further con-
strued in different ways depending on the auxiliary elements that can be
‘built into it’, such as aspectual markers. As far as aspectual marking is
concerned, it is needless to stress that the English and Croatian aspectual
systems cannot be compared at a formal level. Croatian, for instance, just

like other Slavic languages, displays an inventory of morphologically
linked verbal groups (cf. Dickey 1997; Silić 1978; Weber 1978), whereas
English verbs demonstrate their aspectual nature by the (in)capacity to take
the progressive form (cf. Comrie 1976; Langacker 1991). However, we
claim that, despite the different aspectual inventories of English and Croa-
tian, the cognitive semantic characterization of the English present tense
can be used in re-examining the Croatian present tense.14
2.3. Process types and the Croatian aspect
Langacker (1987: 258–261, 2001b: 255–258, 2003: 17) argues that the
aspectual distinction for verbs is easily compared to the count/mass distinc-
tion for nouns. Let us imagine a stone painted against a plain white back-
ground. It is bounded within our immediate spatial scope (Figure 2a). The
basic difference between count and mass nouns is that with a count noun
the boundaries fall within the conceptualizer’s immediate scope, whereas a
mass noun profiles a thing whose substance leaves the boundaries out of
the immediate scope (Figure 2b). Furthermore, a mass is internally con-
tractible/expandable in the sense that a portion of a mass of whatever size
constitutes a valid instance of this mass. Thus a handful of sand remains an
instance of sand. Masses are construed as internally homogeneous and they
cannot be replicated. Count nouns, on the other hand, exhibit at least a
minimal degree of internal heterogeneity and they profile things which
cannot be further divided and continue to represent a valid instance of the
category.
Langacker’s comparison of the count and mass distinction with the as-
pectual characteristics of verbs is straightforward: a perfective verb pro-
files a process construed as being bounded within the scope of predication,
whereas an imperfective verb profiles a process that can completely “oc-
cupy the scope of predication and even overflow its boundaries” (Lan-
gacker 1987: 261).15 Furthermore, imperfective processes represent the
perception of constancy through time and all the component states of the
imperfective process are identical. Thus, imperfective processes exhibit
homogeneity, contractibility and non-replicability (Langacker 1991: 21).
On the other hand, perfective processes always involve a change through
time and it is this change that contributes to their prototypical character.16
(a) Count (b) Mass
IS IS
space space
stone sand
Figure 2. Count/mass nouns (adapted from Langacker 2001b: 256)
There are three classes of Croatian verbs that are traditionally classified
exclusively as imperfectives: these are the so-called stative verbs, rela-
tional verbs and evolutive verbs (Silić 1978: 60–62). Stative verbs are said
to denote static duration, like spavati ‘sleep’ or nadati se ‘hope’, whereas
relational verbs denote relationships, e.g. sadržavati ‘contain’, znati
‘know’, imati ‘have’. Evolutive verbs differ in that they involve a quantita-
tive or qualitative change (Silić 1978: 61). Typical evolutive verbs would
be šetati ‘walk’, pušiti ‘smoke’, živjeti ‘live’, and so on. These three groups
are said to occur only in the imperfective form because their aspect cannot
be changed morphologically without a change in meaning (e.g. znati
‘know’ when prefixed changes into doznati ‘find out’) (Silić 1978: 61).
Generally, it seems that the central notion in the above described classi-
fication of Croatian imperfectives is the notion of homogeneity that is ob-
servable if the process is viewed from a considerable distance.17 It over-
rides the presence of change that we believe is inherent18 in a great number
of imperfective verbs such as plakati ‘cry’, šetati ‘walk’ or spavati ‘sleep’.
In terms of process types, these processes are internally homogenous but
they occur in bounded episodes and thus involve change with regard to
their initiation and termination (Langacker 1990a). This same “inherent
perfectivity” (Langacker 1987: 259) is fundamental in classifying such
verbs in English as perfectives (Figure 3a). This classification is related to
the fact that the English grammatical inventory contains the imperfectiviz-
ing progressive construction that neutralizes this change. The endpoints of
a bounded process are excluded from the immediate scope, which conse-
quently encompasses only a homogeneous internal portion of the process.
It is only this portion that is profiled (cf. Figure 3b). The same ‘internal
view’ (Langacker 2001b: 259) of the event is taken in the following Croa-
tian example:
(10) Ne ulaz-i, Barbara spava-Ø.

not enter.IMPF.2SG.IMP Barbara sleep.IMPF.3SG.PRES
‘Don’t go in, Barbara is sleeping.’
(a) Perfective (b) Progressive (c) Imperfective
MS/IS MS MS/IS
IS
t t t
Figure 3. Basic aspectual classes (taken from Langacker 2001a: 12)
It seems that the verb spavati can easily instantiate both aspectual cate-
gories. In (10) it profiles an imperfective process; that is a stable situation
within the immediate temporal scope (Figure 4a). The internal view on the
situation is reinforced by the imperative in the preceding clause. It indi-
cates that we are changing our perspective and narrowing down our view to
a temporary situation coinciding with the speech event. Sleeping normally
occurs in bounded episodes, and in this particular case the boundaries of
one of those episodes are cancelled and the immediate temporal scope co-
incides with the speech event. However, it is important to stress that such
construal does not depend on the imperative used in the preceding clause.
It is the interplay between the aspect of the verb and the present tense mor-
pheme that allows for the construal of a stable situation within the immedi-
ate temporal scope.
Although the verb spavati ‘sleep’ does not imply internal variation, it
designates a process that implies limits in terms of its beginning and its
end. Thus, if we use the same imperfective present of the verb spavati with
an adverbial of frequency we construe each event as an episode coinciding
with the immediate temporal scope, as in example (11):
(11) Iris spava-Ø i svako popodne.

Iris sleep.IMPF.3SG.PRES and every afternoon
‘Iris sleeps in the afternoons, too.’
Such inherent perfectivity that enables us to view the event in its totality
is present in a large number of Croatian imperfectives and is evident in
their iterative usage. Therefore, we will call them non-canonical imperfec-
tives. An analogous example of a non-canonical imperfective is given in
(12):
(12) Barbara redovito piše-Ø zadać-u.

Barbara regularly write.IMPF.3SG.PRES homework.ACC
‘Barbara does homework regularly.’
However, a relational imperfective, such as znati ‘know’ (Figure 3c) is

a canonical imperfective. It always profiles an instance of a temporally
unbounded process that is stable and constant, and does not involve any
change. Just like a mass noun, it is indefinitely contractible/expansible and
any portion of its component states is a valid instance of the category. In
example (13) the verb is used with the present tense morpheme and the
profiled segment of the process coincides with the time of speaking (Figure
4b).
(13) Iris zna-Ø što želi-Ø.

Iris know.IMPF.3SG.PRES what want.IMPF.3SG.PRES
‘Iris knows what she wants.’
(a) Non-canonical (b) Canonical imperfectives (c) Perfectives

Imperfectives
MS MS MS
IS IS IS
t t t
Figure 4. Aspectual classes and the Croatian present tense

Finally, the morphologically marked perfective always profiles an in-

stance of a temporally bounded process implying substantial change. Fig-
ure 4c represents the conceptual configuration that has to be satisfied for
the present perfective to be acceptable. The example (14) containing the
perfective napisati ‘write’ or ‘finish writing’ does not satisfy such configu-
ration. As suggested in section 2.1, the present tense morpheme contributes
to the process being immediate to the speaker. Thus, in the present tense,
the immediate scope of predication is a segment of immediate reality coin-
ciding with the speech event (Figure 4).
(14) *Ne ulaz-i, Barbara sada

not enter.IMPF.2SG.IMP Barbara now
napiše-Ø zadać-u.
write.PERF.3SG.PRES homework.ACC
*‘Don’t go in, Barbara writes her homework now.’
The perfective present (Figure 4c) is unacceptable because it profiles an

event, which, due to its length, cannot fully overlap with the speech event.
The action of writing, expressed here by the so-called totalizing perfective,
is simply longer than the one of uttering the sentence. This problem of
duration and its implications will be discussed in the section that follows.
Before we proceed, let us sum up the key points related to the Croatian
present tense that are relevant for the remainder of the paper. Traditionally,
the present tense has been mostly described as indicating ‘non-real’ pre-
sent. This rather paradoxical description is related to the fact that all non-
temporal uses of both perfective and imperfective present tense have been
labeled as ‘non-real’. The real (or temporal) usage of the present tense is
mentioned only in relation to the imperfectives used to describe actual,
ongoing events occurring at the time of speaking. However, it is our inten-
tion to provide evidence for the following:
1) The perfective present indicates processes that are bounded within the
immediate temporal scope but still coincide with the time of speaking.
Such temporally bounded processes imply substantial change and they are
always profiled by the morphologically marked perfectives.
2) The Croatian imperfective present indicates two distinct, but equally
‘real’, kinds of processes: a) processes that are coincident with the speech
event and construed as unbounded and internally homogeneous within our
immediate temporal scope, and b) processes coinciding with the speech
event, so that the boundaries of the process overlap with the boundaries of
the immediate temporal scope.
3) The Croatian present tense is a consistent conceptual tool for describing

the process coincident with the speech event. It is the speaker’s point of
view that changes. Most cases of the Croatian present tense involve special
viewing arrangements whereby the speaker is in control of the events being
described.
2.4. Epistemic problem with present perfectives in English and Croatian
Both English and Croatian perfectives are largely unacceptable in the pre-
sent tense with the present-time meaning because the conceptual configura-
tion their felicitous use requires is simply difficult to achieve. There are
two problems with the profiled perfective event coinciding with the speech
event: the durational problem and the epistemic one (Langacker 2001b:
263). The durational problem was already presented in example (14): the
present tense morpheme should signal that the profiled perfective process
coincides with the speech event. The event of writing, however, is longer
than the speech event describing it. The epistemic problem resides in the
fact that in order to be described, an event first needs to be observed.
These problems do not arise with imperfectives such as know ‘znati’
(see fig. 4b) because any portion of the process counts as an instance of the
process type. As far as present perfectives are concerned, the situation in
which none of the above-mentioned problems arises are cases in which the
speaker is “in control of the event” (Langacker 2001b: 263) rather than
simply reporting it. In the case of English, typical examples of control are
performatives. The speech event and the profiled bounded process are one
and the same, as in (15):
(15) I promise I will study more.
In (15) the viewing arrangement is such that the speech event is put on-
stage and profiled, as represented in Figure 5. In Croatian, the use of the
imperfective in contexts involving this particular viewing arrangement
supports the idea of the episodic nature of non-canonical imperfectives.
The non-canonical imperfective obećavati ‘promise’ is construed as a se-
ries of bounded occurrences. The speaker’s control is evident in the fact
that the speaker is capable of simultaneously performing and reporting a
single episode of an imperfective process. The episode coinciding with the
speech event is internally homogeneous, which contributes to the speaker’s
promise being construed as equal at any point during the speech event. The
boundaries of the process overlap with the boundaries of the immediate

temporal scope. The situation is exemplified in (8), repeated below as (16):
(16) Obećava-m da ću učiti više.

promise.IMPF.1SG.PRES that study.IMPF.1SG.FUT more
‘I promise I will study more.’
The perfective form of the same verb, that is obećati, cannot fulfill the
performative function because it cannot profile the process identical with
the speech event. It involves substantial change and it is always bounded
within the immediate temporal scope.
MS
IS
Figure 5. Performatives (taken from Langacker 2001a: 27)
Generally, the issue of the speaker’s control over the event is relevant
for the Croatian perfective present, that is, cases involving morphologically
marked perfectives. As already mentioned, Croatian grammars claim that
the morphologically marked perfectives do not express present time.
Again, the problem resides in the default-viewing arrangement whereby the
speaker simply reports on what happens. However, we can easily imagine
the speaker actually doing what she is saying and thus making each event
coincide with the sentence that describes it, as in (17):
(17) Prvo sjedne-m pa ustane-m.

first sit.PERF.1SG.PRES then get up.PERF.1SG.PRES
‘First I sit down and then I get up.’
A heterogeneous process bounded within the immediate temporal scope

is what sanctions the usage of a morphologically marked perfective as a
performative. On the other hand, in cases where there is no need for a pre-
cise temporal overlap between the speech event and the action being per-
formed, the usage of the present tense perfectives is felicitous, as exempli-

fied in (17).
The last three examples prove that the durational and epistemic prob-
lems are not caused by the present tense perfectives as such. Rather, they
originate in the canonical or default viewing arrangement (Langacker
2001a; 2001b, 2003). The following section deals with departures from this
kind of viewing arrangement and their role in the characterization of the
Croatian present tense.
2.5. Departures from the canonical viewing arrangement
Langacker argues for a naïve characterization of the English simple present

tense, whereby the present tense “indicates the occurrence of a full instan-
tiation of the profiled process that precisely coincides with the time of
speaking” (1987: 250). The same characterization will be instrumental for
our discussion of the Croatian present tense. In order to account for such
characterization of the present tense, we need to consider the nature of
specific semantic change called subjectification (Langacker 1990b; 2003).
As previously stated, the prototypical role of tense is to specify the lo-
cation of the profiled event relative to the time of speaking. As a grounding
element, it construes the grounded process with a high degree of objectiv-
ity whereas the ground (G) itself is construed subjectively (Langacker
1990b; 2003) (Figure 6).
Figure 6. Grounding predication (adapted from Langacker 2003: 8)

The speaker (S) and the hearer (H) are viewers or conceptualizers view-
ing the profiled process from offstage. In canonical viewing arrangements,
the speaker simply reports on what happens in the world. However, in

cases of non-temporal uses of the present tense conceptualizers are not
simple observers of events in the actual world. They describe processes
whose occurrence is conceptual and their conceptualizing activity does not
depend on the actual observation and time-of-speaking input from the out-
side world. Therefore, we can say that the occurrence of such processes is
construed with extreme subjectivity (Langacker 2003: 16). Such conceptual
occurrences belong to virtual reality that is based on the speaker’s knowl-
edge about the world and its structure (Langacker 1991; 2001a; 2001b;
2003). As already defined in section 2.1., the conceptualizer’s known real-
ity pertains to what she knows or thinks she knows. This reality extends to
the moment of speaking and represents whatever has happened up to that
moment and is known to the conceptualizer. At the moment of speaking,
the conceptualizer employs the present tense either to report on what she
construes objectively in the actual outside world or whatever unfolds con-
ceptually in her virtual reality. A more detailed explanation, based on the
examples introduced in section 1, will be given in section 3.
3. The Croatian present tense
3.1. The actual plane
Prototypically, the speaker uses the Croatian imperfective present tense to

report ongoing events occurring in the actual world around her. This ca-
nonical situation is exemplified in (1a), here repeated as (18):
(18) Barbara gradi-Ø dvorac-Ø od pijesk-a.

Barbara build.IMPF.3SG.PRES castle.ACC of sand.GEN
‘Barbara is building a sand castle.’
This particular usage is typical of Croatian imperfectives whose aktion-

sart involves some change. The verb graditi ‘build’ is a resultative imper-
fective. As far as the nature of the process is concerned, we need to stress
that this process involves some change, but the change is not substantial, as
is the case with the morphologically marked perfective izgraditi ‘finish
building’. Rather, the process designated by graditi involves bounded epi-
sodes each of which has its limits in terms of initiation and termination of
the process (Figure 7).
Figure 7. Non-canonical imperfectives
When a resultative imperfective verb is used in the present tense indi-

cating the ongoing event, the endpoints of the bounded episode are left
outside the immediate temporal scope (Figure 8).
Actual
plane
MS
IS
Figure 8. The present tense
However, as previously mentioned, not all situations the speaker is

likely to describe are immediate and ongoing. Cognitive processing enables
us to construct meanings that imply situations which are not present or
verifiable in our immediate reality. The conceptualizer’s viewing arrange-
ment changes whenever she describes an event whose occurrence is the
result of her mental activity rather than the actual state of affairs in the
outside world. Thus, when the present tense is used, the value of the pre-
sent time is built into the conceptual content by the speaker’s subjective
moment-of-speaking access to a variety of relationships in virtual worlds of
various kinds. The following section describes and explores the role of
such conceptual worlds in the characterization of the Croatian present

tense.
3.2. Virtuality
Virtual worlds are conceptual worlds distinct from the actual world.19 They
include a number of mental constructions (cf. Langacker 2001a: 33) such
as, for example, schedules, scripts, mental replay, and the so-called struc-
tural plane (Langacker 2003: 19), which comprises generalizations repre-
senting the world’s structure. Virtual worlds represent conceptual reality
consisting of virtual events that “occur” in them by virtue of being men-
tally accessed by the conceptualizer whenever she wishes to do so. The
virtual events are ‘fitted’ into the immediate scope imposed by the present
tense morpheme. The term virtual world will be used to refer to any kind
of virtual plane in Langackerian sense. Thus, it is going to host various
virtual events: those representing particular facets of the real world’s struc-
ture and those pertaining to virtual documents and schedules related to
specific events and situations in the actual world.
3.2.1. Virtuality and habituals
Let us now reconsider example (1c) from the introduction, here repeated as
(19):
(19) Svako ljeto Barbara izgradi-Ø

every summer Barbara build.PERF.3SG.PRES
dvorac-Ø od pijesk-a.
The point of view assumed in this situation is not that of a direct de-
scription of what is given in the actual world at the present moment: the
process of building a castle can be observed and described in its totality
because it does not actually occur at the time of speaking. The virtual
world holds the event representation corresponding to an indefinite number
of actual instantiations and representing the generalization pertaining to
their habitual nature (Figure 9).
Structural
plane
Actual world
t
Figure 9. Virtuality and Croatian perfective habituals (based on Langacker 2003:

19)
The profiled event coinciding with the speech event is a virtual instance
of the perfective process of building a sand castle. With the change in-
volved in the process and the noun dvorac ‘castle’ used in the singular, the
clause designates Barbara’s habitual activity of building (at least) one sand
castle every summer. This activity is a part of the regular course of events
stored in the structural plane.
If, instead of the morphologically marked perfective, we use the non-
canonical imperfective graditi ‘build’ and keep the time expression svako
ljeto ‘every summer’, as in example (1d) repeated below as (20), the qual-
ity of the process within our immediate temporal scope is changed, but the
occurrence of the process is again conceptual rather than actual (Figure
10).
(20) Svako ljeto Barbara gradi-Ø

every summer Barbara build.IMPF.3SG.PRES
dvorac-Ø od pijesk-a.
‘Every summer Barbara engages in building a sand castle.’
Structural
plane
Actual world
t
Figure 10. Virtuality and Croatian imperfective habituals
As was suggested in section 2.3., non-canonical imperfectives do not

simply continue indefinitely through time but occur in bounded episodes.
The end points of an episode coincide with the boundaries of the immedi-
ate scope of predication. There is no specific variation within the constitu-
tive stages of the process, except for its beginning and its end. We should
certainly bear in mind that the notion of a bounded episode in the case of
graditi does not necessarily imply that the castle is (always) fully finished.
It simply implies that the subject initiates the process and brings it to some
sort of an end, and this process happens regularly (or habitually) every
summer.
To conclude, both perfective and imperfective habituals profile virtual
events occupying the structural plane. The speaker describes the events
occurring in the virtual world, just like those occurring in the actual world
around her. Therefore, we wish to stress that the central contrast between
the temporal and non-temporal usage of the present tense resides in the
perspective taken on the event. Once we accept a conceptualist approach to
meaning, there is no ground to support the traditional description of dis-
tinct values of the Croatian present tense morpheme. We claim that its role
is invariable and consistent, and that various uses of the present tense
should be explained by a closer examination of other facets of construal,
primarily perspective and its two closely related aspects: the degree of
subjectivity and vantage point. Let us remind ourselves that the tense
specifies the location of the event with reference to the time of speaking
and that the schematic meaning of the present tense morpheme is immedi-
acy. This general characterization of the present tense is evident in the
above mentioned examples and will be supported by the examples that

follow.
3.2.2. Virtuality and parallel habituals
At this point let us re-examine examples (2a), (2b) and (2c) repeated here
as (21a), (21b) and (21c) respectively:
(21) a. Kad ode-Ø u kin-o, Iris

jede-Ø kokic-e.
‘When she goes to the cinema, Iris eats popcorn.’
(‘When she gets to go to the cinema, Iris spends her time eating
popcorn.’)
b. Kad ode-Ø u kin-o, Iris

pojede-Ø kokic-e.
eat.PERF.3SG.PRES popcorn.ACC
(‘When she goes to the cinema, Iris treats herself with popcorn.’)
c. Kad ide-Ø u kin-o, Iris

when go.IMPF.3SG.PRES to cinema.LOC Iris
jede-Ø kokic-e.
(‘On the way to the cinema, Iris eats popcorn.’)
These three examples are instances of what we might call parallel

habituals. Such parallel habituals involve two (or more) clauses, which are
sequentially and conditionally related (Figure 11). In all examples, the
subordinate clause serves as a precondition20 for the second clause. If the
process in the subordinate clause is profiled by a morphologically marked
perfective, as in (21a) and (21b), the event in the second clause is con-
strued as something that necessarily follows. Although the perfective in the
subordinate clause in (21a) seems to indicate the inception of the event,
this starting point of the event metonymically refers to the whole frame of
going to the cinema. Thus, the imperfective in the main clause can be eas-
ily “fitted” into the frame. The perfective present in the subordinate clause
in (21a) might indicate that Iris does not go to the cinema so often, and that
when she does, she spends all her time eating popcorn instead of watching
the film. By designating a process involving substantial change, perfectives
in this kind of subordinate clauses often contribute to the meaning con-
strual involving something less ordinary and actually not very frequent, but
habitually related to what follows. Thus, the perfectives in (21a) and (21b)
do not express the habitual actions of going to the cinema, but precondi-
tions for the event that necessarily follows. If a morphologically marked
perfective is used in both clauses, as in (21b), the sentence expresses al-
most a ritualistic course of events whereby the two perfectives designate a
habitual sequence. The sequence in (21c) is a kind of neutral habitual. The
two imperfectives simply state that both going to the cinema and eating
popcorn while in the cinema are something Iris does from time to time.
However, due to the imperfective present in the subordinate clause, the
sentence has both a literal, i.e. a non-metonymical, and a metonymical
reading. Thus, going to the cinema can be interpreted either as the whole
frame of going to the cinema or the action of being on the way to the cin-
ema.
Process Process
Structural
plane
t Actual world
Figure 11. Virtuality and Croatian parallel habituals
The arrow between Process1 and Process2 in Figure 11 represents the

dependence of the second event on the first one. The dotted correspon-
dence lines indicate that the representation of the sequence in the plane of
the virtual world corresponds to an indefinite number of actual instantia-
tions of such a sequence. These events occur in a virtual way every time
the speaker ‘reads’ the generalization available in the structural plane. This
special viewing arrangement excludes the durational and epistemic prob-
lems and both clauses coincide with the respective speech events, no mat-
ter what type of process is profiled by the clause.
There is another non-temporal use of the Croatian present tense that can
be analysed in terms of the conceptual configuration containing a virtual
sequence, as sketched in Figure 12. This use is illustrated by example (2d),
repeated below as (22):
(22) Kad ode-Ø u kin-o, Iris

će jesti kokic-e.
eat.IMPF.3SG.FUT popcorn.ACC
‘When she goes to the cinema, Iris will eat popcorn.’
Process Process
Virtual
world
t Actual world
Figure 12. The virtual sequence indicating future
This sentence is interpreted as indicating future. The event profiled by

the subordinate clause is a present time virtual occurrence coextensive with
the speech event. This is consistent with the idea that the present tense
morpheme always indicates present time and immediacy to the speaker.
The future tense verb in the main clause profiles a virtual process viewed
as both temporally and causally following the first process.21 These two
virtual processes are specifically construed as having single actual coun-
terparts located in the future relative to the speech event.
3.2.3. Virtuality and conditionals
Examples (23a) and (23b) below – which are, respectively, repeated exam-
ples (3a) and (3b), represent the use of the Croatian present tense in condi-
tional clauses. The causal relationship between the two processes is analo-
gous to the relationship sketched in Figure 11. The situation in the main
clause results from the event given in the subordinate clause.
As already mentioned in the introduction, ako ‘if’ and da ‘if’, as well as
kad ‘when’, prije ‘before’, nakon ‘after’, and others, are space builders that
overtly establish mental spaces and shift the viewpoint to the space they
establish. Thus, the content of the protasis is viewed and apprehended from
a fictive vantage point (Figure 13). When used with the present tense, the
space builders delimit spaces within the virtual world in which the epis-
temic value of the present tense is realized relative to that world.22 The
time of speaking is fictively located in the mental space established by the
subordinator and the corresponding clause, hence the conceptual occur-
rence of the process in that clause coincides with the speech event.
(23) a. Ako uči-m/nauči-m,

prođe-m/prolazi-m na ispit-u.
pass.PERF/IMPF.1SG.PRES on exam.LOC
‘If I study, I pass the exam.’
b. Da uči-m/nauči-m,
prošao bih na ispi-tu.
pass.PERF.1SG.M.COND on exam.LOC
‘If I studied, I would pass the exam.’
This epistemic immediacy is central to understanding the conceptual

import of the Croatian present tense. The tense in question expresses what
is immediately given either by direct perception of actual events at the time
of speaking, or by immediate access to whatever has been accumulated in
terms of real-world knowledge. In all cases of non-temporal usage of the
present tense, what is expressed belongs to the epistemic background.23
Therefore, the condition given in the protasis is based on our knowledge
about how the world is structured, and the present tense in the hypothetical
space of the protasis indicates general ‘proximity’ of the process to the
speaker in terms of her known reality.
The cognitive distance, on the other hand, is marked by the use of sub-
ordinators. We believe that da marks a higher degree of cognitive distance
than ako. The role of the subordinators is twofold: they mentally frame the
event and they reflect the speaker’s relationship to whatever is described
by the present tense event that follows. The mentioned relationship as-
sumes the character of neutral, positive or negative epistemic stance (Fill-
more 1990). The present tense itself is not used to express any cognitive
distance. On the contrary, we wish to propose that the ‘presence’ of the
present tense stands in a strong conceptual opposition to the hypothetical
subordinators, especially da. The subordinator da creates a mental space
that is construed as epistemically more distanced from the speaker’s con-
ception of reality than the one established by ako, but the present tense
morpheme consistently places the event proximal to the speaker’s vantage
point (be it real or fictive).
Process
Virtual
t world
G'
Actual world
t
G
Figure 13. Shifted ground for viewing the process

As far as the aspectual nature of the processes is concerned, the epis-
temic and durational problems are avoided again by fictitious, moment-of-
speaking viewing of virtual events as they “appear” before the conceptual-
izer’s eyes. The location of these events coincides with the speech event
and does not differ from any other conceptual occurrence profiled by the
present tense. The morphologically marked perfective in the subordinate
clauses in both (23a) and (23b) involves the change typical of such perfec-
tives. In this particular case, it implies that the process of studying results
is a considerable change in the amount of knowledge, or, more precisely,
that everything necessary is well learnt and memorized. The non-canonical
imperfective učiti does not involve such thoroughness. It simply indicates
that studying itself is a necessary precondition for passing exams.
3.2.4. Virtuality, schedules and tacit mental spaces
The scheduled future in Croatian, just like in English, favours a time ex-
pression which indicates the expected occurrence of the actual event, as in
(6), repeated below as (24):
(24) Dolazi-m rano ujutro.

come.IMPF.1SG.PRES early in-the-morning
‘I am coming early in the morning.’
The imperfective present in (24) involves a subjectively construed

process whose temporal value is realized by the speaker’s ability to ‘read’
at the speech time the virtual schedule (cf. Langacker 2001a, 2001b, 2003)
consisting of mental representations of anticipated events. A single event is
read off and its virtual occurrence coincides with the speech event (Figure
14).
Virtual
world
Actual
t world
Figure 14. The scheduled future (adapted from Langacker 2001a: 31)
This particular use of the present tense strongly supports the idea of an
epistemic background status for present tense meanings as suggested by
Brisard (1999). The nature of the present tense is such that it can never be
used to describe experiences that cannot be anticipated on the basis of what
we have learnt about reality. Cases of the scheduled future represent an
expected extension from generics and habituals. In the virtual world based
on the speaker’s experience and knowledge about particular events, future
events are scheduled and read off before the moment of their actualization.
Since canonical imperfectives in the present tense profile processes that
are stable and continue indefinitely, the use of the present tense for sched-
uled future activities is restricted to non-canonical imperfectives that occur

in bounded episodes. Morphologically marked perfectives, on the other
hand, can be used to express futurity only if the verb is modified by the
modal adverb možda ‘maybe’, as in example (5), here repeated as (25):
(25) Možda kupi-m nov-u knjig-u.

maybe buy.PERF.1SG.PRES new.ACC book.ACC
‘I might buy a new book.’
Možda evokes a mental of a potential course of events. Conditions that

would need to be fulfilled for the potentiality to turn into reality remain
tacit. Tacit conditions and the resulting situation are represented by the
morphologically marked perfective that involves substantial change. Simi-
larly, the perfective present in the relative clause in example (4), repeated
below as (26), is possible due to the tacit mental space indicated by the
that-clause and its relation to the main clause.
(26) Nikada neću baciti knjig-u koj-u

never not-throw.PERF.1SG.FUT book.ACC that.ACC
kupi-m.
buy.PERF.1SG.PRES
‘I will never throw away the book that I buy.’
Example (26) is conceptually quite complex. It gives the resulting situa-

tion first and then tacitly expresses the condition in the relative clause.
Furthermore, it can be interpreted as referring to future and as habitual. In
both cases, however, the acceptability of the perfective present can be ex-
plained by going back to the nature of the virtual sequence described in
Figure 11 and the fictive vantage point sketched in Figure 13. The two
events form a sequence whereby the process of (not)-throwing the book
away depends on the process of buying one (as, for example, opposed to
getting one for free). Thus, the sequence is interpreted as follows: “If I buy
a book, I’ll never throw it away.” The process, given as the condition, is
viewed from the fictive vantage point, placed in the mental space tacitly
established by the relative pronoun (and overtly established by if in our
explicit interpretation of the sentence). Thus, the process, conceived as
occurring in this space, is viewed as coinciding with the speech event. The
presupposed viewing arrangement avoids durational and epistemic prob-
lems and allows for the use of the morphologically marked perfective in
the present tense.
3.2.5. Virtuality and mental replay
The last case we will consider exemplifies the historical present. In the
case of historical present the relevant virtual world is built from the knowl-
edge of what we have seen and experienced in reality, and the knowledge
of what we have learnt to believe the world is made of. The mental replay
is a unique mental construct that makes possible immediate access to spe-
cific past occurrences that are conceptually occurring at the moment of
speaking. Thus, in example (7), repeated below as (27), the speaker uses
the present tense to revive actual events from the past:
(27) Dolazi-m/Dođe-m kuć-i i

come.IMPF/PERF.1SG.PRES home.DAT and
vidi-m da nema-m ključev-e.
see.1SG.PRES that not-have.1SG.PRES keys.ACC
‘I come home and I see that I have no keys.’
Following Langacker’s proposal for English (2003: 23), we wish to

suggest that in cases of the historical present in Croatian, the viewpoint
does not shift into the past. Instead, the speaker’s vantage point remains
stable, as at the speech time she is fictively viewing the event by reading it
off from a virtual document (see Fig. 15). This virtual document is a series
of recalled events that correspond to actual occurrences of specific events
in the past. Each reading from the virtual document results in a virtual oc-
currence of the event and its precise coincidence with the moment of
speaking. Figure 15 represents a virtual occurrence of the perfective proc-
ess. The document of virtual events is stable and accessible at any time
subsequent to the actual occurrence of the events.
It seems that the central role of the Croatian present tense is indeed
based on its epistemic value of immediacy. It calls upon two modes of
experience, perception and mental access to conceptual representations,
and consistently indicates immediate givenness of a proposition (Brisard
1999). Although it is probably too early to speculate, we would like to
propose that there might be another important piece of evidence supporting
this kind of characterization of the present tense in Croatian, and that is the
sequence of tenses in reported speech. We will not attempt to discuss the
phenomenon in this paper. However, we would like to say that the actual
speaker’s reports of what was said in the past seem to be ‘very faithful’ to
the epistemic status of the situation described by the original speaker.
Thus, whereas the reporting verb in the main clause is in the past tense
indicating that the utterance is an event situated in the past reality, the con-
tent of the utterance is in the same tense as it was in the original statement.
In cases where the content of the utterance is in the present tense, the re-
ported statement is ambiguous as to the temporal duration of the situation
described. It might or might not extend through the actual speech event.
Such ambiguities are resolved by examining the nature of mental spaces
and other facets of construal. Yet, the present tense will always function as
“the most unmediated way of referring to the ground” (Brisard 1999: 365).
Virtual
world
t
Actual world
Figure 15. The historical present and perfectives
4. Conclusion
The aim of this paper was to demonstrate that the analytical models and
tools developed within Cognitive Grammar can be employed to analyse the
nature of the Croatian present tense. More specifically, we have applied
Langacker’s naïve characterization of the English present tense to the
analysis of the Croatian present tense and have provided evidence for the
claim that the Croatian present profiles a process that fully coincides with
the time of speaking.
First, we have re-examined some elements of Croatian verbal aspect and
its relation to process types. Then, we have emphasized the importance of a
particular, non-canonical viewing arrangement, which implies that what is
being viewed is not the actual occurrence of events in the world but their
virtual occurrence in the virtual world. We have proposed that all non-
temporal uses of the Croatian present tense are cases involving extreme
subjectification as a special kind of semantic shift. Furthermore, we have
insisted on the consistent conceptual import of the present tense as an indi-

cator of immediacy and claimed that the inherent value of the tense does
not change. Rather, what changes is the point of view, or more precisely,
the characterization of the nature of events being viewed. Finally, we have
examined several examples of the present tense in terms of mental spaces.
It is important to stress that the presented re-analysis of the Croatian
present tense is far from complete. Yet, we hope to have tackled some
relevant questions and offered some tentative answers. In doing so, we
have been guided by the fundamental principle of cognitive semantics
which says that meaning is to be identified with conceptualization and
should be treated as “the product of mental activity on the part of physi-
cally embodied, socio-culturally grounded human minds” (Langacker
2000: 26). Traditional treatment of linguistic meaning was based on the
idea that there is an objective relation of language to mind-independent
reality. However, the complexity of language simply demands an approach
based on a different perspective. Using the metalanguage of cognitive se-
mantics itself, we can say that a new approach needs a new orientation and
a new vantage point. The orientation should certainly be one of accepting a
subjective basis for meaning and the human mind itself as the vantage
point for a more revealing description of language.
Notes
1. This specific term is used here to indicate the label used in Croatian traditional
grammars.
2. We sincerely thank Agata Kochańska, Dagmar Divjak and two anonymous
reviewers for their truly generous comments on different aspects of the paper.
3. This refers to the English simple present tense.
4. The abbreviations used in the glossing are as follows: PERF = perfective,
IMPF = imperfective, PRES = present, FUT = future, COND = conditional,
IMP = imperative, GEN = genitive, LOC = locative, ACC = accusative, DAT
= dative. Croatian inflectional morphemes always carry more than a single
meaning, i.e. they are a combination of inseparable features. With certain
nouns (eg. dvorac 'castle') these features are expressed by means of a zero
morpheme which marks not only case, but also gender and number.
5. A short description of the relationship between Langacker’s classification of
process types and Croatian aspectual phenomena will be offered in section 2.2.
6. The Croatian future tense is analytic and formed with the present tense of the
unstressed form of the auxiliary htjeti ‘want’ and the infinitive of the main
verb.
7. The present conditional of Croatian verbs is formed with the aorist of the aux-
iliary biti 'be' and the participle form of the main verb.
8. It is important to point out that not all mental spaces in Fauconnier's sense are
virtual worlds in Langacker's sense. For example, temporal expressions evoke
mental spaces which have the epistemic status of past reality.
9. The terms ‘real’ and ‘non-real’ should not be confused (or identified) with
similar notions often used to describe real as opposed to hypothetical situa-
tions.
10. The Croatian term for completeness (cf. Katičić 1991: 45) is gotovost, and it is
described as a relevant grammatical category, which is claimed to be morpho-
logically marked. Croatian grammarians (Katičić 1991; Barić et al. 1995) put
forward the existence of the six conceived time categories and the tense system
is neatly fitted into these categories. For example, Katičić (1991:50) interprets
the perfekt form of the perfective verb vratiti (‘return’) as expressing the com-
pleted present. According to his interpretation, the action was completed in the
past, but it has certain ramifications for the present (for a cognitive reanalysis
cf. Stanojević and Geld 2005).
11. If we were to approximate the meaning of Croatian aspectual pairs we might
consider the following translations into English: for example, for umri-
jeti/umirati ‘die’ and ‘be dying’, for čitati/pročitati ‘read’ and ‘finish reading’,
for govoriti/progovoriti ‘speak’ and ‘start speaking’ or ‘speak out’, and so on.
However, in most cases the translation would be misleading because more ac-
curate translations of aspectual nuances are only possible in context.
12. The elaborated epistemic model plays a central role in describing various uses
of modals (Langacker 1991: 269–281). Thus, there will be no direct reference
to this particular model. However, since it is one of the three fundamental
models that enable us to grasp and describe the tense-modality system, we have
decided to include it here.
13. The Coratian perfekt is formed using the imperfective present tense of the verb
biti (‘to be’) and the l-participle of the content verb.
14. It is important to point out that we are not attempting to simplify the semantic
complexity of Croatian aspect. However, we argue that Langacker’s “quasi-
universal definitions of the perfective and imperfective” (Dickey 1997: 34) can
be easily applied crosslinguistically in the analysis of specific aspectual phe-
nomena.
15. For a discussion of the count/mass distinction and its relation to Slavic aspect
see Janda (2004).
16. Some perfective processes involve only a limiting case of change, namely
change residing only in the fact that they are temporally bounded.
17. The term distance employed here is related to Langacker’s (cf. 2006) discus-
sion of the notion of discreteness. An important aspect of our everyday experi-
ence is that a large group of discrete entities appears to us as a continuous sub-
stance when viewed from a distance. We would like to argue that the
traditional classification of Croatian imperfectives indicates viewing these

processes from a considerable distance. By “moving away” from the process
we impose continuity whereas by “getting closer” to the process we see that it
consists of bounded episodes. Finally, by “zooming in” we exclude the end-
points of an episode from the viewing frame. Thus, by changing our perspec-
tive we are constantly involved in the process of continuization/discretization
(cf. Langacker 2006:121).
18. It is important to stress that this inherent change is not something that is objec-
tively given. Homogeneity and heterogeneity is always a matter of how the
process is construed.
19. In the case of virtual entities, unlike in the case of actual ones, their reality
status is not at issue. They are mental constructs conjured up by the conceptu-
alizer for the purpose of doing some mental work.
20. This phenomenon is related to figure-ground alignment (Talmy 2000). The
event in the ground/subordinate clause is conceptualized as a basis for the
event in the figure/main clause.
21. The general idea of virtual sequences has been taken from Langacker (2003:
35).
22. We wish to thank Frank Brisard for his insightful and generous comments on
this matter.
23. The term “epistemic background” has been taken from Brisard (1999: 366). It
relates to the understanding of the present tense’s primary function as “the
most unmediated way” of referring to the ground, whereby the ground does not
only represent spatio-temporal relations at the moment of speaking, but in-
cludes our generic knowledge and “those elements of our conception of reality
which are held to be universally valid” (Brisard 1999: 365).
References
Barić, Eugenija, Mijo Lončarić, Dragica Malić, Slavko Pavešić, Mirko Peti, Vesna
Zečević, and Marija Znika
1995 Hrvatska gramatika [Croatian Grammar]. Zagreb: Školska knjiga.
Brisard, Frank
1999 A Critique of Localism in and about Tense Theory. Ph.D. diss., Uni-
versity of Antwerp.
Comrie, Bernard
1985 Tense. Cambridge: Cambridge University Press.
1976 Aspect. Cambridge: Cambridge University Press.
Cutrer, Michelle L.
1994 Time and Tense in Narratives and Everyday Language. Ph.D. diss.,
University of California, San Diego.
Dickey, M. Stephen
2000 Parameters of Slavic Aspect: a cognitive approach. Stanford: CSLI
Publications.
Fauconnier, Gilles
1997 Mappings in Thought and Language. Cambridge: Cambridge Uni-
versity Press.
Filipović, Rudolf (ed.)
1978 Kontrastivna analiza engleskog i hrvatskog ili srpskog jezika [Con-
trastive Analysis of English and Serbo-Croatian] Volume 2. Zagreb:
Institute of Linguistics, Faculty of Philosophy, University of Zagreb.
Fillmore, Charles J.
1990 Epistemic Stance and Grammatical Form in English Conditional
Sentences. In Papers from the 26th Regional Meeting of the Chicago
Linguistic Society. Volume 1, 137–162. Chicago Linguistic Society.
Janda, Laura
2004 A metaphor in search of a source domain: the categories of Slavic
aspect. Cognitive Linguistics 15 (4): 471–527.
Kašić, Bartol
2002 Reprint. Osnove ilirskoga jezika u dvije knjige [Foundations of the
Illyrian Language in Two Volumes]. Zagreb: Institut za hrvatski jezik
i jezikoslovlje. Original edition, Rome, 1604.
Katičić, Radoslav
1991 Sintaksa hrvatskoga književnog jezika [Syntax of the Croatian Liter-
ary Language]. Zagreb: Jugoslavenska akademija znanosti i umjet-
nosti.
1987 Foundations of Cognitive Grammar. Volume 1: Theoretical Prereq-
uisites. Stanford: Stanford University Press.
1990 a Concept, Image and Symbol. Berlin/New York: Mouton de Gruyter.
1990 b Subjectification. Cognitive Linguistics 1: 5–38.
1991 Foundations of Cognitive Grammar. Volume 2: Descriptive Applica-
tion. Stanford: Stanford University Press.
2000 Why a mind is necessary: Conceptualization, grammar and linguistic
semantics. In Meaning and Cognition: A multidisciplinary approach,
Liliana Albertazzi (ed.), 25–38. Amsterdam/Philadelphia: John Be-
jamins.
2001 a Cognitive linguistics, language pedagogy, and the English present
tense. In Applied Cognitive Linguistics I: Theory and language Ac-
quisition, Martin Pütz, Susanne Niemeier, and René Dirven (eds.), 3–
37. Berlin/New York: Mouton de Gruyter.
2001 b The English present tense. English Language and Linguistics 5 (2):
251–272.
2003 Extreme Subjectification: English Tense and Modals. In Motivation

in Language, Herbert Cuyckens, Thomas Berg, René Dirven, and
Klaus-Uwe Panther (eds.), 3–24. Amsterdam/Philadelphia: John
Benjamins.
2006 On the continuous debate about discreteness. Cognitive Linguistics
17 (1): 107–151.
Silić, Josip
1978 An Approach to the Study of Aspectuality in the Croatian Literary
Language. In Kontrastivna analiza engleskog i hrvatskog ili srpskog
jezika [Contrastive Analysis of English and Serbo-Croatian]. Volume
2. Rudolf Filipović (ed.), 42–70. Zagreb: Institute of Linguistics,
Faculty of Philosophy, University of Zagreb.
Stanojević, Mateusz-Milan, and Renata Geld
2005 Current relevance in Croatian: a cognitive account. Glossos Issue 6.
Sweetser, Eve
1996 Mental Spaces and the Grammar of Conditional Constructions. In
Spaces, Worlds, and Grammar, Gilles Fauconnier and Eve Sweetser
(eds.), 318–333. Chicago/London: The University of Chicago Press.
Talmy, Leonard
2000 Toward a Cognitive Semantics. Vol. 1. Concept Structuring Systems.
Cambridge, Massachusetts/London, England: MIT Press.
Weber, Ralph
1978 Aspect in Serbo-Croatian and English. In Kontrastivna analiza
engleskog i hrvatskog ili srpskog jezika [Contrastive Analysis of
English and Serbo-Croatian]. Volume 2. Rudolf Filipović (ed.), 42–
70. Zagreb: Institute of Linguistics, Faculty of Philosophy, Univer-
sity of Zagreb.
Conflicting epistemic meanings of the Polish
aspectual variants in past and in future uses:
are they a vagary of grammar?
Agata Kochańska
Abstract
Prototypically the Polish perfective signals the temporal bounding of the

profiled process, and the imperfective – the absence of such bounding.
However, in some cases the choice between the perfective and the imperfec-
tive has nothing to do with temporal bounding, but is motivated by other
factors, including the epistemic ones. Interestingly, the past and the future
epistemic uses of the two aspectual variants exhibit a seemingly inconsistent
semantic behavior. In past uses, the perfective and the imperfective signal,
respectively, the speaker’s epistemic certainty and its lack. On the other
hand, in future uses, the perfective designates future processes which are
construed canonically, that is, as epistemically uncertain, while the imper-
fective refers to those future events which are viewed as nearly certain to
occur. The aims of the paper are threefold: (i) to provide a conceptual analy-
sis of the past and future epistemic uses of the perfective and the imperfec-
tive; (ii) to analyze how these epistemic meanings relate to the prototypical
values of the aspectual variants under consideration; and finally (iii) to con-
sider how the epistemic values of the two aspectual variants in past uses re-
late to their epistemic meanings in future uses. I will argue that the epistemic
behavior of Polish aspectual variants – instead of being a vagary of grammar
– is motivated by the respective prototypical meanings of the Polish perfec-
tive and imperfective, as well as by the differences in the typical epistemic
status of events manifested in the temporal frames of the past, the present,
and the future.
Keywords: cognitive grammar, control cycle, epistemic attitude, Polish,

subjectivity, verbal aspect, virtuality
150 Agata Kochańska
1. Introduction*
Verbs in Polish, just like verbs in other Slavic languages, are either perfec-
tive or imperfective. Prototypically, perfective verbs profile processes
which are conceptualized in their totality, that is, as temporally bounded
within their immediate scope of predication. On the other hand, imperfec-
tive verbs designate processes which are conceptualized as ongoing, that is,
temporally unbounded within the adopted immediate scope. Such proto-
typical uses of, respectively, the perfective and the imperfective are illus-
trated in (1) below.
(1) a. Piotr zapisał mój adres.

Piotr wrote-down.PERF my address
‘Piotr wrote down my address.’
b. Piotr zapisywał mój adres, gdy

Piotr wrote-down.IMPERF my address when
nadjechał autobus.
came bus
‘Piotr was writing down my address when the bus came.’
In (1a), the complete event of writing down an address, together with its
beginning and end, is profiled. On the other hand, in (1b) the immediate
scope of predication (whose temporal expanse is specified by the adverbial
clause of time gdy nadjechał autobus ‘when the bus came’) encompasses
only a portion of internal phases of the process in question, i.e. the portion
which is coextensive with the temporal boundaries of the immediate scope.
It is only these internal phases that are actually designated, although the
notion of the complete event figures in the conceptual base of the predica-
tion. In other words, the profiled phases are specifically construed as inter-
nal stages of the process in question, which is evoked in the base of the
conceptualization as evolving from its start towards its desired completion
point. Hence, even though in (1b) the temporal boundaries of the event are
excluded from profiling, they form a part of the maximal scope.
However, it is a well-known – and perhaps notorious – fact about Slavic
aspect that the choice between the perfective and the imperfective may be
motivated by many other factors, in addition to the temporally bounded or
unbounded nature of the conceived event. The motivating factors are often
subjective in character: they may have nothing to do with the objective
properties of conceived events, but rather with their subjective construal on
Conflicting epistemic meanings of the Polish aspectual variants 151
the part of the speaker. I believe that the speaker’s epistemic attitude with
respect to what he is talking about is one of such subjective factors that
motivate aspectual usage in Polish. In the present paper I would like to
consider precisely those uses of aspectual forms in Polish in which the
speaker’s choice of either the perfective or the imperfective is meant to
convey an epistemic meaning.
Specifically, I would like to consider and compare the respective epis-
temic meanings of the perfective and the imperfective in the past tense and
in the non-past tense used in reference to future events.1 My aims in the
present analysis are: first, to characterize the epistemic meanings of the
perfective and the imperfective in the past and in the non-past tense; sec-
ondly, to investigate how the epistemic meanings in each group of exam-
ples relate to the prototypical values of the aspectual variants under con-
sideration; and finally, to consider how the epistemic values of the two
aspectual variants in the past tense relate to their epistemic meanings in the
non-past tense.
The overall discussion will be organized in four sections. Section 2 will
present an analysis of the perfective/imperfective contrast in non-past tense
sentences referring to future events. The reason for starting the discussion
with perfectives and imperfectives used in reference to future events is that
in examples of this kind the aspectual choice seems to have a relatively
obvious epistemic import. The epistemic motivation behind choosing either
the perfective or the imperfective in past-tense sentences is less straight-
forward and hence, the discussion of this topic will be relegated to section
3. In both sections, I will start with describing the prototypical meanings of
the perfective and the imperfective in the relevant tense. Then I will turn to
a characterization of their respective epistemic meanings and a discussion
of how the proposed epistemic values may be related to the prototypical
meanings. Section 4 will be devoted to a discussion of the relationship
between the epistemic meanings of the perfective and the imperfective in
the past tense and their respective epistemic values in the non-past tense.
Finally, section 5 will present a brief summary and conclusions from the
overall discussion.
The theoretical framework of the overall analysis will be that of cogni-
tive grammar, with special emphasis on its idea of meaning understood in
terms of both conceptual content and construal imposed upon this content
(cf. e.g. Langacker 1987a, 1991). Equally important will be cognitive
grammar’s recognition of the relevance of subjectively construed aspects
of the conceived scene in language (cf. e.g. Langacker 1990), its idea of
virtuality (cf. e.g. Langacker 1999, 2001), and its characterization of the
cognitive model of the control cycle, especially the control cycle mani-
fested in the epistemic sphere (cf. e.g Langacker 2002, 2004, ms)
Let me finish the introductory part of the present paper with a few re-
marks concerning the data to be considered. First, all examples under dis-
cussion are sentences referring only to non-repeated, internally heteroge-
neous, and complete events.2 The aim of limiting my considerations to data
of this kind is to exclude non-completion or repetitiveness as possible fac-
tors motivating the choice of the imperfective. Since in Polish non-repeated
complete events are prototypically designated by perfective verb forms,
when the speaker decides to refer to such events by means of the imperfec-
tive, his aspectual choice is not motivated by the objective properties of the
events in question, but rather by more subjective factors. Epistemic evalua-
tion of what is being talked about seems to be one such factor.
Secondly, all the examples under consideration are made-up sentences.
The reason for this is that by using made-up examples I could control the
parameters of contrast between the perfective and the imperfective. In this
way I could create contexts in which the choice between the perfective and
the imperfective clearly has a purely epistemic import. This, in turn,
greatly facilitates the ease of exposition. It should be stressed, however,
that the epistemic uses of the Polish perfective and imperfective, which are
illustrated in the present work by made-up examples, have all been attested
in the corpus that I have collected for a more extensive study of the seman-
tics of aspect in Polish (cf. Kochańska 2002a). Moreover, my own intui-
tions concerning the epistemic import of the two aspectual variants have all
been checked with at least twenty other native speakers of Polish (mostly
students at the courses in cognitive grammar that I have taught) and none
of them have expressed any reservations about the proposed semantic char-
acterizations.
2. The epistemic import of the perfective and the imperfective in the

non-past tense
Let me now turn to a specific discussion of the epistemic motivation be-

hind choosing either the perfective or the imperfective to refer to a future
event. In section 2.1 I will consider in some detail the prototypical mean-
ings of the perfective and the imperfective in the non-past tense. Then, in
section 2.2, I will turn to analyzing the respective epistemic values of the
non-past perfective and the non-past imperfective used in reference to a
future event.
2.1. The prototypical meanings of the perfective and the imperfective in

the non-past tense
The prototypical uses of the non-past perfective and imperfective are ex-
emplified in (2) below:
(2) a. Piotr zje śniadanie.

Piotr eat.PERF breakfast
‘Piotr will have his breakfast.’
b. Piotr je śniadanie.
Piotr eat.IMPERF breakfast
‘Piotr is having his breakfast.’
As illustrated by (2a), the non-past perfective profiles a future process,

that is, a process wholly following the time of speaking. On the other hand,
the non-past imperfective in (2b) is construed as profiling the portion of
component states of a process which is coextensive with the time of speak-
ing. The portion in question is unbounded within the immediate scope of
predication, although it is understood as part of a whole process, which
itself may be conceptualized as temporally bounded.
Let me note at this point that there seem to be good reasons for the lack
of a present-time interpretation for the Polish non-past perfective. Perfec-
tives profile full instances of temporally bounded processes. In turn, in the
case of a truly present-time interpretation of a clause, the profiled process
should be construed as temporally coinciding with the speech event that
describes it. As noted by Langacker in his analysis of the English simple
present tense, there are two problems with conceptualizing full instances of
temporally bounded processes that are coextensive with speech events
describing them. First of all, “the span of time required for a bounded
process to occur has no inherent connection with the time required for a
speech event describing it” (Langacker 1987b: 83; see also 2001: 263).
This is what may be called a durational problem. There is also another
problem, which is epistemic in nature. As Langacker (1987b: 83) notes,
[e]ven if the profiled process were the right length, the speaker could hardly
describe it with a precisely coincident speech event: to do so, he would have
to begin his description at exactly the instant when the process was initiated,
before he had a chance to observe its occurrence and identify it. Once he ob-
serves a full instantiation of the process (including its endpoint), it is too late
to initiate a temporally coincident description.
Even if the durational and the epistemic problem mentioned above are
disregarded, the viewing arrangement involved in conceptualizing proc-
esses coextensive with the speech time is still far from optimal, at least in
the temporal dimension. When the profiled process is coextensive with the
time of speaking, the object of observation is maximally close in time to
the observing subject. In turn, as Langacker (1993: 456) notes, the viewing
distance
shows a positive correlation with the size of the field of view. When the fo-
cus of visual attention is on a distant object, both the maximal field of view
and the viewing frame (i.e. the general locus of attention) subtend large por-
tions of the surrounding world. But if [we – A.K.] look at something very
close – e.g. the palm of [our] hand – [our] visual horizons shrink drastically
and [we] see but a limited portion of [our] immediate environment. This cor-
relation is mirrored in general conception.
Furthermore, “a decrease in [viewing – A.K.] distance does not invariably
enhance perception. As we approach an object, there is a point beyond
which any further approximation actually makes it harder to observe – we
are just too close to see it well” (Langacker 1993: 457).
One of the contexts in which Langacker’s durational and epistemic
problems do not arise (for a discussion of this point see Langacker 1987b:
83) is that of explicit performatives, whose temporal expanse – by defini-
tion – coincides with that of the speech event and whose occurrence is
initiated and controlled by the speaker. However, even in this case the ob-
ject of conception, that is, a process identified with the speech event, is so
close to the observing subject that its endpoints coincide with the bounda-
ries of the immediate scope, instead of falling properly within them. Thus,
the profiled process completely fills the conceptual onstage region. Since
in a configuration of this kind the processual endpoints fall at the fringes of
the onstage region, they are beyond the area of maximal conceptual acuity
and are not clearly “visible” to the conceptualizer. This situation is analo-
gous to the one in which we are approaching a cow and at some point we
come so close to the animal that its spatial expanse completely fills our
visual field. At that point the animal’s spatial boundaries are no longer
clearly visible to us – in such a case our viewing experience should be
more appropriately described as seeing cowhide, rather than seeing a cow.
It may thus be suggested that explicit performatives involve a concep-
tual configuration that is ambivalent in certain respects. On the one hand,
the profiled process is identified with the speech event and is by definition
conceptualized as temporally bounded. On the other hand, its endpoints –
also by definition – are conceptualized as coinciding with the boundaries

of the immediate scope, instead of properly falling within them. Therefore,
they are at the fringes of the conceptualizer’s immediate scope and beyond
the region of the maximal conceptual viewing acuity. This kind of configu-
ration is ambivalent in the sense that it simultaneously provides motivation
for coding both performatives’ bounded character and the fact that their
temporal boundaries are not clearly visible to the conceptualizer.
As illustrated by the examples in (3) below, languages such as e.g. Eng-
lish conventionally focus on the bounded character of the profiled process.
(3) a. I am reading/*read a book just now

b. I hereby *am promising/promise to do this
According to Langacker, the meaning of the English present tense

predication is that “a full instantiation of the profiled process occurs and
precisely coincides with the time of speaking” (1987b: 82). Due to the
durational and epistemic problems discussed above, English verbs profiling
temporally bounded processes do not normally occur in the simple present
tense form with the prototypical present-time meaning. One motivated
exception to this pattern are explicit performatives, which in English are
coded by verbs in the simple present tense. What sanctions the use of such
verbs in explicitly performative sentences is that, in such sentences, the
endpoints of the profiled process by definition coincide with the bounda-
ries of the temporal viewing frame delimited by the time of speaking. Thus
a full instance of a temporally bounded process is wholly contained in the
onstage region (although not properly within it). Therefore, in English it is
the bounded character of the process instance that is in focus.
The situation is different in Polish. To illustrate this, let us consider the
examples in (4) and (5) below:
(4) Przyrzekam, że nigdy cię nie opuszczę.

Promise-I.IMPERF that never you not leave-I
‘I promise never to leave you.’
(5) Przyrzeknę, że nigdy cię nie opuszczę.

Promise-I.PERF that never you not leave-I
a.*‘I promise never to leave you.’
b. ‘I will promise never to leave you.’
Sentence (4) is an explicit performative and it is an imperfective verb

phrase that is used in it. On the other hand, sentence (5), with the perfec-
tive, is unacceptable under the explicit performative interpretation given in
(a). It is only acceptable under reading (b), in which the profiled process
instance fully follows the time of speaking.
It seems, therefore, that Polish does not conventionally focus on the
temporally bounded character of the profiled speech act, but rather on the
poor “visibility” of its endpoints, which results from observing the event
from a “close-up” perspective. Therefore, present-time processes in Polish
are always designated by imperfectives, even in the case of explicit per-
formatives, whose endpoints coincide with the boundaries of the immedi-
ate scope of predication defined by the temporal expanse of the speech
event. Apparently, the use of the perfective in Polish is only sanctioned
when the endpoints of the profiled process fall properly within the concep-
tual viewing frame and are therefore clearly “visible” to the conceptualizer.
Further support for this claim comes from considering examples in the
past tense analogous to these in (6) and (7) below:
(6) a. Mył się przez pół godziny.

Washed-he.IMPERF himself across half hour
‘He was washing up for half an hour.’
‘He washed up for half an hour.’
b. *Umył się przez pół godziny.

Washed-he.PERF himself across half hour
‘He washed up for half an hour.’
(7) a. Umył się w pół godziny.

Washed-he.PERF himself in half hour
‘He washed up in half an hour’
b. *Mył się w pół godziny.

Washed-he.IMPERF himself in half hour
‘He washed up in half an hour.’
In (6), the temporal extension of the profiled process is specified by the

phrase containing the path preposition przez ‘across’. The import of using a
temporal przez-phrase is that the profiled process is construed as extending
from the beginning to the end of the time span specified by the preposi-
tional phrase ‘for half an hour’, thereby “filling” it completely. One inter-
pretation of (6a) which satisfies this conceptual configuration is that the
clause designates internal phases of a process “cut” from a temporally
more extended complete event by the temporal przez-phrase. Under the
second interpretation – which is more relevant to the present discussion –
the profiled process is construed as starting at the beginning of the period
specified by the prepositional phrase and being completed at the end of this
period. In other words, (6a) may be understood as profiling a full instance
of a temporally bounded process whose endpoints coincide with the
boundaries of the immediate scope adopted by the conceptualizer and
specified by the prepositional phrase of time. Examples analogous to (6a)
are used with this kind of meaning when the conceptualizer specifically
focuses on the extended duration of the profiled process, especially if this
duration is longer than expected. It may be argued that focusing on the
longer-than-expected duration of the profiled process involves conceptual-
izing its temporal expanse as completely filling the adopted temporal view-
ing frame. In other words, in a conceptualization of this kind, the proces-
sual endpoints have to be construed as coinciding with the boundaries of
the immediate scope, rather than falling properly within them. The impor-
tant thing is that, under such an interpretation, it is still imperfective verb
phrases that have to be used in reference to temporally bounded processes.
As illustrated by (6b), the use of perfectives is not felicitous when the end-
points of the profiled process are specifically construed as coinciding with
the boundaries of the temporal immediate scope.
On the other hand, when – as in (7) above – the temporal extension of
the profiled process is indicated by a phrase containing the preposition w
‘in’, instead of przez ‘across’, the process in question is construed as fal-
ling within the specified time frame together with its endpoints, as the im-
port of using a temporal w-phrase is specifically that the profiled process
was completed within the specified time span. When a conceptual configu-
ration of this kind is evoked, the use of the perfective is not only permissi-
ble, but in fact the only possible option: the imperfective in (7b) is unac-
ceptable if interpreted as profiling a one-time complete event (although, of
course, its use is fully sanctioned in the case of the habitual interpreta-
tion).3
In cases analogous to those in (6a) above it is past processes that are put
in profile. Since they are conceptually viewed from a relatively distant
present-time vantage point, the reason for the coincidence between the
endpoints of the profiled event and the boundaries of the temporal immedi-
ate scope is not the “close-up” perspective, which brings about the shrink-
ing of the locus of viewing attention in the case of explicit performatives.

Instead, the responsible factor is conceptualizing the profiled event as
maximally extended within the adopted immediate scope. In both cases,
however, the result is the coincidence of the processual endpoints with the
boundaries of the immediate scope and hence, their poor “visibility” to the
conceptualizer. The Polish perfective is unacceptable both in explicit per-
formatives (present time usage) and in cases analogous to (6) (past time
usage). I suggest that this is because the perfective in Polish specifically
requires that the endpoints of the profiled event fall within the region of
maximal conceptual acuity and are thus clearly “visible” to the conceptual-
izer.
In the case of clauses in the non-past tense, which are my main concern
in the present and the following section, such good “visibility” of the proc-
essual endpoints can only be achieved when the profiled event is located in
the future relative to the speech time, so that it can be mentally “viewed”
from a sufficiently distant perspective. Hence, it seems to be a well-
motivated fact about Polish that the present-time meaning may only be
conveyed by the non-past imperfective. The non-past perfective, on the
other hand, has the future-time meaning.
1.2. The epistemic import of the non-past perfective and imperfective
Let me start this part of the present discussion with the observation that the
present-time meaning is not the only semantic value that the non-past im-
perfective can assume. It may also be used in reference to future complete
events, just like the perfective. This is illustrated in (8) below.
(8) Samolot odleci /odlatuje za godzinę.

Plane will-depart.PERF /departs.IMPERF in hour
‘The plane will depart / departs in an hour.’
I would like to argue that in uses of this kind the perfective and the im-
perfective are not fully equivalent and that the semantic contrast between
them is epistemic in nature. The analysis that I will propose for the Polish
perfective/imperfective contrast in the future-time uses is analogous to the
analysis proposed by Langacker (cf. e.g. 1999, 2001) for the contrast in
English between the will future construction and the simple present tense
in its future use. My suggestion is that the perfective simply designates a
future event, whose occurrence is – by its very nature – only potential and
therefore, epistemically uncertain. On the other hand, the imperfective

seems to be applicable only to events which are under the speaker’s epis-
temic control as a result of – for example – being scheduled or planned to
occur. This is confirmed by the data in (9) below.
(9) a. Samolot odleci /*odlatuje, jeśli mgła

Plane will-depart.PERF /departs.IMPERF if fog
opadnie
falls-down
‘The plane will depart if the fog falls down.’
b. W rozkładzie samolot *odleci /odlatuje

In timetable plane will-depart.PERF /departs.IMPERF
za godzinę.
in hour
‘According to the timetable, the plane departs in an hour.’
In (9a) the profiled future event is conditional on another future situa-

tion, whose occurrence is only potential and epistemically uncertain. Obvi-
ously, an event of this kind cannot be planned or scheduled with absolute
certainty, because the realization of the conditions for its occurrence is not
guaranteed. Hence, the process profiled in (9a) is beyond the speaker’s
epistemic control. If the future-time use of the imperfective really is re-
stricted to future events over which the speaker can exercise epistemic
control, this aspectual variant should be unacceptable in contexts of this
kind – and, in fact, it is infelicitous in (9a). On the other hand, sentence
(9b) does not seem to describe a future event per se. Instead, it refers di-
rectly to an entry in a schedule for the future. The actual future event cor-
responding to that entry is invoked only indirectly, as a future realization of
the currently existing schedule. Not surprisingly, the perfective sounds odd
here, in contrast to the imperfective, which is perfectly natural.
Although typically (and quite reasonably) we consider future events as
belonging to the realm of what is as yet unrealized and thus epistemically
uncertain, we nevertheless feel that in some cases we can exercise a meas-
ure of epistemic control over what is going to happen. For example, it is a
deeply ingrained (although apparently quite weakly founded) belief in
Western culture that our plans and schedules have not only a causal, but
also a determining effect on the future evolution of reality. To put it differ-
ently, we view our schedules and plans as something comparable to scripts
for the future.4 Those scripts give us the feeling of epistemic control over
the future evolution of reality – if you want to know what is going to hap-
pen in the future, the only thing that you need to do is to have a look at the
relevant script and see what is there. I would like to suggest, therefore, that
the epistemic value of the perfective is the one associated with futurity by
default: from the present vantage point future events are construed by the
conceptualizer as only potential, uncertain, and beyond control. On the
other hand, the imperfective may be used in reference to those future
events which are specifically conceptualized as currently being under the
speaker’s epistemic control, for example due to being scheduled or
planned.
In (9b) above, we are dealing with the script of a schedule for future
events. However, as illustrated by the examples in (10) below, different
kinds of scripts for the future evolution of reality may be involved on dif-
ferent occasions.
(10) a. Wyjeżdżam dziś do Krakowa.

Leave-I.IMPERF today for Kraków
‘I’m leaving for Kraków today.’
b. Śniadanie jem za godzinę.

Breakfast eat-I.IMPERF in hour
‘I have my breakfast an hour later.’5
c. Wiosna zaczyna się za mniej więcej miesiąc.

Spring starts.IMPERF itself in less more month
‘Spring starts in about a month.’
The example in (10a) involves the script of a personal plan for future
actions. The scripts invoked in sentences (10b) and (10c) involve different
aspects of our conception of how the world generally works or is struc-
tured, rather than of what happens in it as a matter of contingent occur-
rence.6 In (10b), under its habitual reading, the relevant part of our idea of
how the world is structured is our conception of what is done habitually.
Sentence (10c), on the other hand, seems to invoke our mental representa-
tion of the calendrical cycle.
As has already been mentioned, the use of the non-past perfective is a
default way of referring to future events. Since it is a perfectly sensible
default attitude to view a future event as potential, uncertain, and beyond
control, this epistemic import of the perfective may reasonably be consid-
ered part and parcel of its prototypical meaning. The question, however, is
how the epistemic value of the non-past imperfective may be related to its
prototypical use, which profiles an ongoing present process. My suggestion
in this respect is that the apparent future-time use of the imperfective is, in
fact, a special type of the present-time use. When the perfective is used to
refer to a future event, it is this event that is actually designated. However,
in the case of the imperfective the profile is shifted to an entry in the con-
ceptualizer’s mental script for the future. At the speech time, the conceptu-
alizer consults his mental script and activates the relevant entry within it,
as a means of seemingly exercising epistemic control over the future evolu-
tion of reality. My claim is that when the imperfective is used to refer to a
future event, what is put in profile is such an activation, that is, a present-
time mental occurrence of an event, rather than its future actual occur-
rence.7 The corresponding actual future event is part of the overall con-
ceived scene, but is relegated to the conceptual base. Of course, despite
directly profiling entries in mental scripts which are activated at the time of
speaking, the uses of the imperfective illustrated by the examples in (9b)
and (10) are taken as pertaining to future events. This results from our ide-
alized cognitive models concerning the role of such mental scripts in shap-
ing the future course of events. We often seem to conceive of plans, sched-
ules, or what we view as patterns of how things are in the world as having
a causal or even a determining force with respect to future events. The
future occurrence of actions that are planned is commonly taken as being
nearly guaranteed by the very fact that we have made a plan to perform
them. Scheduled events are also taken as being bound to occur. Finally, our
experience that certain patterns have so far been regularly recurring in real-
ity leads us to the extrapolation that they are also bound to recur in the
future.
To sum up the main points of the analysis offered above, I suggest that
when the non-past imperfective is used in reference to a future event, the
profile is restricted to the present mental activation of an entry in a script
for future events. However, the overall conceived scene incorporates in its
maximal scope also the conception of the actual future event. The occur-
rence of this event is viewed as being almost guaranteed by the existence
of the mental script under consideration. This mental script is viewed as
providing the conceptualizer with a measure of “epistemic control” over
future events and it is the meaning of “epistemic control” that all uses of
the imperfective in reference to future processes seem to share.8
It may also be noted at this point that one aspect of the conceived scene
that the future-time uses of the non-past imperfective seem to share with
the prototypical present-time meaning is that in both cases the profiled
process is specifically conceptualized as part of a temporally more ex-

tended configuration. In the case of the prototypical present-time use, the
profiled phases, which are coextensive with the time of speaking, are con-
ceived of as internal phases of a complete process. In turn, in the case of
the future-time use, the profiled entry in a mental script is viewed as a vir-
tual analog of the corresponding future event. In both cases, an additional
mental operation is required: not only does the conceptualizer mentally
view the process which is the focus of his attention, but also he conceives
of this process specifically qua part of a temporally more extended back-
ground configuration. In this respect, both uses of the non-past imperfec-
tive differ from the non-past perfective, which profiles a process conceptu-
alized in and by itself, as a self-contained entity.9
Elsewhere (cf. Kochańska 2002a and 2002b) I analyze the non-past per-
fective and the non-past imperfective as clausal grounding predications (in
Langacker’s sense of this term – see e.g. Langacker 1990 section 2). The
reason is that – as suggested above – it is only the combination of the non-
past tense with either the perfective or the imperfective predication that
fulfills the role of locating the profiled process relative to the ground with
respect to such basic epistemic issues as time, reality, etc. Since the non-
past perfective and the non-past imperfective are grounding predications,
they may be analyzed – following Langacker (cf. 2004 section 2 and ms) –
as manifestations of what he calls the control cycle in the conceptualizer’s
epistemic domain. To elucidate this point, let me first characterize what
Langacker refers to as the control cycle model (cf. e.g. 2002: 193–194).
Langacker’s starting point is the observation that every living creature ex-
ercises different kinds of control over particular sets of entities. To give a
few examples, when you are holding an object in your hands, you are con-
trolling it physically. When you see an object, that object is under your
perceptual control. When you accept an event as part of your conception of
reality, you exercise epistemic control over the event in question. The crea-
ture exercising control is called the actor. The entities under the actor’s
control constitute the actor's dominion. The phase of the cycle in which the
actor statically exercises control over entities in his dominion is referred to
as the baseline.
In addition to entities inside the actor’s dominion (under the actor’s
control), there are, of course, also countless entities which are outside the
dominion (beyond the actor’s control). Some of the entities outside the
dominion may become more directly relevant to the actor in the sense that
the actor may gain the potential to interact with them. These entities are
said to be entering the actor’s field (defined in terms of potential interac-
tion). The very fact that the actor can potentially interact with an entity
creates a situation of tension. Any entity that appears in the actor’s field
has to be dealt with in some way. In turn, since any actual interaction with
an entity may be viewed as leading to the actor achieving at least some sort
of control over it, the potential for interaction also creates the potential for
gaining control. This potential is directed at the entity within the actor’s
field (the target). Hence, in this phase of the control cycle (called the po-
tential phase) the actor is being “pushed” towards gaining control over the
target which happens to appear within the field. In the case of the control
cycle on the epistemic level – which is my present concern – we may fur-
ther distinguish the following three stages within the potential phase: (i)
the formulation stage, when an event first presents itself to the actor’s
mind; (ii) the assessment stage, when the status of the event relative to
reality becomes an active issue; and (iii) the inclination stage, when the
actor inclines towards either accepting or rejecting the event as part of his
conception of reality, but is not yet able to resolve the matter in a definite
fashion (cf. Langacker 2002: 200).
In the control cycle manifested at any level (be it physical, perceptual,
or epistemic), the tension in the potential phase leads to the action phase,
when the actor actually does something to gain control over the target in
his field. Finally, in the result phase of the cycle, as a result of the preced-
ing action, the actor’s dominion is expanded to incorporate the target,
which is now – in one way or another – under the actor’s control. It should
be emphasized at this point that the notion of control seems to be under-
stood by Langacker in a maximally general way: in a sense, you gain
physical control over another living creature in your field both when you
catch this creature and when it irrevocably escapes you, you gain percep-
tual control over a thing in your visual field both when you manage to see
that thing clearly and when you lose sight of it for good, you gain epis-
temic control over an event both when you accept this event and when you
definitely reject it from your conception of reality. In all those cases we
have to do with the actor’s gaining control over the target in the sense that
the target in the actor’s field has been “dealt with” in some manner, the
tension of the potential phase has been resolved in one way or another, and
the target has been definitely placed with respect to the actor’s dominion
(of course, it remains definitely placed either inside or outside the actor’s
dominion only until its status becomes an active issue again).
In the case of clausal grounding predications, the relevant epistemic
dominion is the speaker’s conception of basic reality, that is, of “what the
conceptualizer accepts as having happened or obtained up to the present
moment” (Langacker ms). The relevant target is the process profiled by a

clause. The import of the Polish non-past imperfective is that the profiled
process is located in the speaker’s epistemic dominion as an either actual
or at least mental occurrence currently being realized – hence, the ground-
ing relationship pertains to the result phase of the epistemic control cycle.
This explains why examples analogous to both (2b) (Piotr je śniadanie
‘Piotr is having his breakfast’) and e.g. (9b) (W rozkładzie samolot odlatuje
za godzinę ‘According to the timetable, the plane departs in an hour’) con-
tain non-past imperfective verb phrases. In the former case, it is a present-
time event which is profiled. In the latter, the conceptualizer refers to an
actual occurrence located in the future. In both cases, however, the proc-
esses in question are under the conceptualizer’s epistemic control as a re-
sult of being immediately accessible to him. In cases analogous to (2b) they
are accessible because they can be directly apprehended in actuality. On
the other hand, in cases parallel to (9b), they are accessible in virtue of the
existence of the relevant mental script in which the virtual analog of the
future event exists and can be mentally observed at the speech time.
In turn, in the case of the non-past perfective the grounding relationship
pertains to the inclination phase of the cycle – the speaker inclines towards
accepting the profiled process into his conception of basic reality, but is
unable to do this in a definite fashion. This is because the process in ques-
tion is located in the future as a not-yet-realized occurrence. For this reason
Polish examples analogous to (9a) (Samolot odleci, jeśli mgła opadnie
‘The plane will depart if the fog falls down’) are only acceptable with non-
past perfective verb phrases. In cases similar to (9a) the conceptualizer
conceives of the relevant actual event as an occurrence with some likeli-
hood of being realized in the future. Hence, he inclines towards accepting
this occurrence into his conception of reality, but is – as yet – unable to do
so in a definite fashion. Such an occurrence cannot be directly apprehended
in actuality. Moreover, due to the very nature of the event in question, the
conceptualizer does not have any mental script at his disposal in which the
virtual analog of the event could be accessed. Hence, the event cannot be
included into the conceptualizer’s epistemic dominion, as it is not yet im-
mediately accessible to him, either in the actual or in the mental sense.
3. The epistemic meanings of the perfective and the imperfective in

the past tense
The next step in the present discussion is an analysis of the respective epis-
temic values of the perfective and the imperfective in the past tense. I will
start with a characterization of the prototypical meanings of the aspectual
forms under consideration in section 3.1. Then, in section 3.2, I will ana-
lyze the respective epistemic meanings of the two tense-aspectual forms
under consideration and how they are related to the prototypical meanings.
3.1. The prototypical meanings of the perfective and the imperfective in

the past tense
The prototypical meanings of the past perfective and the past imperfective
are illustrated by the examples in (11) below.
(11) a. Rano zjadłam śniadanie.

In-morning ate-I.PERF breakfast
‘In the morning I ate my breakfast.’
b. Gdy zadzwonił telefon, jadłam śniadanie.

When rang phone ate-I.IMPERF breakfast
‘When the phone rang, I was eating my breakfast.’
In (11a), where the perfective is used, the temporal immediate scope lo-
cated in the past contains all the component phases of the event eat, with
its beginning and ending. Hence, a complete process is designated. In
(11b), on the other hand, only selected internal phases of the event fall
within the temporal immediate scope and only those internal phases are
actually profiled. The profiled internal phases, however, are conceptualized
specifically as phases of a complete event. Hence, the profiled portion of
the event eat is conceptualized against the background of a temporally
more extended configuration, the idea of the entire event of eating, which
falls within the maximal scope of the overall conceptualization. In other
words, the past perfective prototypically profiles a temporally bounded
past event, which is conceptualized as a self-contained entity, without ref-
erence to anything else. On the other hand, the past imperfective prototypi-
cally designates only those internal phases of a complete process which fall
within the adopted temporal immediate scope, with the idea of the com-
plete process figuring in the conceptual base.10
3.2. The epistemic import of the past perfective and imperfective
Let me now turn to the epistemic meanings of the perfective and the imper-
fective in the past tense. Apparently, cases of a purely epistemic contrast
between the two aspectual variants in the past tense are not easy to find.
The group of uses in which epistemic meanings seem to surface and which
I am going to discuss in the present paper are certain sentences referring to
past events of verbal communication. Example of the uses in question are
given in (12) below.
(12) A: Czy ten artykuł będzie gotowy na piątek?

[Q] this article will-be ready for Friday?
‘Will this article be ready by Friday?’
B: a. Na pewno. Piotr powiedział /?mówił, że

For sure. Piotr said.PERF /?said.IMPERF that
go skończy.
it will-finish-he
‘Sure. Piotr said that he will finish it (by then).’
b. Nie wiem. Piotr ?powiedział /mówił, że go

Not know-I. Piotr ?said.PERF /said.IMPERF that it
skończy.
will-finish-he
‘I don’t know. Piotr said that he will finish it (by then).’
Let me note first that the semantic contrast illustrated in (12) is indeed
very subtle. In fact, both the perfective and the imperfective are actually
possible in both (a) and (b). Moreover, when the speaker wishes to refer to
a past event of verbal communication, there are multiple factors beside the
conceptualizer’s epistemic attitude which may motivate the choice of one
aspectual variant over the other. Hence, the question marks in (12) merely
signal that the aspectual variant not marked by them will be preferred by
the speaker who specifically wishes to emphasize his epistemic attitude
expressed in the first clause of the utterance and who conceives of this
attitude as being directly shaped by the profiled past event of verbal com-
munication. To put it differently, in contexts similar to (12) numerous fac-

tors may come into play and influence the conceptualizer’s aspectual
choice. Most of them have nothing to do with epistemic judgment. One
possibility, however, is that the conceptualizer makes reference to a past
event of verbal communication and at the same time wishes to convey his
own epistemic attitude with respect to what was communicated in the pro-
filed verbal action. Moreover, the conveyed epistemic attitude is specifi-
cally presented as in some way causally linked to the designated act of
communication. It is this kind of motivation behind the use of either the
perfective or the imperfective that is relevant to the present analysis.
Secondly, I would like to observe that, unlike in the case of the exam-
ples in the non-past tense which were discussed before, in (12) the
speaker’s choice of either the perfective or the imperfective has nothing to
do with the epistemic status of the event profiled by the perfective or im-
perfective verb phrase in bold. In either case, the profiled event of verbal
communication is treated by the speaker as part of known past reality. In-
stead, what is at issue in (12) is the conceptualizer’s epistemic assessment
of what was communicated in the profiled past event of verbal communica-
tion. In brief, it seems that the choice of the perfective signals the concep-
tualizer’s epistemic commitment with respect to what was communicated
in the profiled past event of verbal communication11 and the fact that this
commitment is a consequence of the profiled verbal action. On the other
hand, the choice of the imperfective specifically conveys the meaning that
at the time of speaking the conceptualizer is still unable to resolve his epis-
temic attitude with respect to what was communicated in the past, that he
can neither commit himself to what was said nor totally reject it, that he
still has reservations blocking his full epistemic acceptance (or rejection)
of the contents of the profiled past speech event. Moreover, the conceptual-
izer’s epistemic reservations are construed as caused by the fact that the
profiled past verbal action failed to establish the epistemic status of what
was communicated.
Having characterized the possible epistemic values of the perfective and
the imperfective in the past tense, I will now turn to the question of how
they are related to the basic meanings of the two aspectual variants. My
suggestion in this respect is that while the prototypical past perfective pro-
files an event which was objectively fully completed prior to the time of
speaking, in the case of the epistemic use, the notion of completion may be
understood as pertaining not only to objective, but also to subjective fac-
tors (note that I am using the notions of ‘objectivity’ and ‘subjectivity’ in
Langacker’s sense of these terms (cf. e.g. 1990: 6–9) – that is, as relating
either to the object of conceptualization or to its subject). In other words,

in the case of epistemic uses of past perfective verbs of verbal communica-
tion, not only was the profiled objective event of verbal communication
fully completed, but the conceptualizer accepted its contents into his own
conception of reality prior to the time of speaking. Therefore, the idea of
past completion applies to both the designated objective event of commu-
nication and the resolution of the conceptualizer’s own subjective epis-
temic attitude with respect to what was communicated in the profiled ob-
jective event.
On the other hand, as far as the past imperfective is concerned, its basic
use designates an event which is not completed, but rather in progress
within a particular temporal immediate scope in the past. The designated
phases are conceptualized qua component internal phases of the complete
event. Hence, the profiled portion of the event is viewed specifically as
part of a temporally more extended configuration, that is, of the complete
event. If we now turn to the epistemic use of past imperfective verbs of
verbal communication, we may observe that this use evokes a conceptuali-
zation in which the past event of communication is completed in the objec-
tive, but not in the subjective sense. The trajector in the profiled event of
communication successfully completed his verbal actions in the past –
hence the event is conceptualized as completed objectively. Nevertheless,
at the time of speaking the conceptualizer still has reservations whether or
not to accept the communicated contents as part of his own conception of
reality – hence the event is viewed as being “in progress” in the subjective
sense. Although objectively completed, the profiled past event of verbal
communication is nevertheless conceptualized specifically as being still “in
progress” along the subjective axis in the sense that the conceptualizer’s
epistemic attitude with respect to what was communicated in the profiled
event still remains unsettled at the present moment.
I would like to suggest at this point that – just like the contrast between
the non-past perfective and the non-past imperfective in section 2 above –
also the contrast between the perfective and the imperfective in the past
tense, which was illustrated by (12), can be characterized in terms of Lan-
gacker’s epistemic control cycle model (see the discussion of the model in
section 2 above). The main difference is that in the past tense the perfec-
tive and the imperfective do not constitute parts of grounding predications.
Hence, the relevant targets in the speaker’s epistemic control cycle are not
just the profiled processes. Also, the relevant epistemic dominion is not the
speaker’s conception of basic reality, that is, the set of realized occur-
rences. Instead, the targets involved are propositions expressed in the pro-
filed events of verbal communication, where the term ‘proposition’ is un-

derstood as referring to a process, or occurrence, plus the assessment of its
epistemic status (in other words, the term ‘proposition’ refers to the mean-
ing of a declarative finite clause, which designates an epistemically
grounded process). The relevant epistemic dominion is the speaker’s con-
ception of elaborated reality, which “consists of propositions pertaining to
occurrences” (Langacker ms), rather than of realized occurrences per se.
In an attempt at rephrasing the proposed analysis of the examples in
(12) in terms of the control cycle model, I would like to suggest that the
conceptualizer’s choice of the perfective in (12a) (Piotr powiedział, że go
[artykuł] skończy ‘Piotr said that he will finish the paper’) indicates that
the subjectively construed epistemic control cycle pertaining to the concep-
tualizer’s own conception of reality was completed and the result phase
was reached. The proposition expressed in the profiled past verbal action
was introduced into the conceptualizer’s epistemic field via that action and
the conceptualizer gained epistemic control over the proposition in ques-
tion by accepting it as part of his conception of reality. In the case of the
perfective, the conceptualizer’s epistemic control cycle remains, in a sense,
“invisible”, as the entire transition from the potential to the result phase is
construed as effortless and automatic.
On the other hand, the import of the imperfective in (12b) (Piotr mówił,
że go [artykuł] skończy ‘Piotr said that he will finish the paper’) is that the
subjectively construed epistemic control cycle is still “in progress” at pre-
sent: the proposition was presented to the conceptualizer’s mind in the
profiled past process of verbal communication, but its definite epistemic
status in the conceptualizer’s mind remains unresolved at the speech time.
The conceptualizer may incline towards either accepting or rejecting the
proposition, but he is unable to make a definite epistemic judgment. Being
unable to move beyond the inclination stage, the conceptualizer remains in
the state of tension and it is that tension which makes the operation of the
control cycle more prominent when the imperfective – instead of the per-
fective – is chosen to refer to a past complete process of verbal communi-
cation.
I would like to suggest at this point that our epistemic behavior with re-
spect to propositions that are presented to our minds in various ways seems
to be regulated by a particular idealized cognitive model which specifies
under what conditions we are ready to accept a proposition as part of our
conception of reality. It may be observed first that this conception is not
static, but inherently dynamic, as it changes with new experiences. The
already accumulated knowledge has a particular evolutionary momentum,
that is, a tendency “to continue its evolution along certain paths in prefer-
ence to others” (Langacker 1991: 277) – given what we already know
about the world, we expect to learn about it some new things, but not oth-
ers. One consequence of this is that we are not equally ready to accept as
true just any proposition entering our epistemic field at the formulation
stage – for example, we do not automatically take for a fact just any state-
ment about reality that a stranger might wish to utter. Propositions aug-
menting our conception of reality in ways which are compatible with its
evolutionary momentum seem to be accepted as true in a virtually auto-
matic fashion. However, our minds may also be confronted with proposi-
tions which in one way or another conflict with the inherent dynamism of
our idea of reality. To be accepted as true, such propositions have to be
invested with an epistemic force which is sufficient to counteract the in-
herent evolutionary momentum of our conception of reality and push its
development along previously unexpected paths.
The notion of an epistemic force of a proposition is meant to capture
all kinds of contextual factors which accompany the presentation of a
proposition to the conceptualizer’s mind and which contribute to the likeli-
hood of the conceptualizer accepting the proposition in question into his
idea of reality. One of the ways in which propositions may present them-
selves to the conceptualizer’s mind is via other people’s verbal actions. In
the case of propositions presented to the conceptualizer’s mind in this par-
ticular way the relevant aspects of the proposition’s epistemic force in-
clude, for example, the reliability of the speaker who presents the proposi-
tion to the conceptualizer’s mind or the degree of his own epistemic
commitment to what he is talking about. To put it differently, people differ
with respect to the degree to which they are recognized as authorities on
particular subject matters. In turn, a proposition expressed by an authority
in the relevant domain will naturally have a greater epistemic force than
the one conveyed by a layman. Also, the person expressing a proposition in
an utterance may invest his utterance with varying degrees of conviction
with respect to what he is saying, thereby endowing the expressed proposi-
tion with varying degrees of epistemic force.
The evolutionary momentum of our conception of reality may either
work in tandem with the epistemic force of a proposition presented to our
mind or counteract that force. In any case, a proposition becomes accepted
into the conceptualizer’s epistemic dominion only when its epistemic force
manages to overcome any potential resistance caused by the evolutionary
momentum of the conceptualizer’s dynamically changing knowledge sys-
tem. It may also happen that the proposition presents itself to the conceptu-
alizer’s mind with an epistemic force that is too weak to overcome the
evolutionary momentum of the existing knowledge. In this case, the evolu-
tionary momentum of our reality conception may make us reject the propo-
sition which is presented to our mind. Another possibility is that this evolu-
tionary momentum may be too weak in the relevant direction to counteract
fully the epistemic force of the proposition. The two forces may thus re-
main in a sort of dynamic equilibrium, a state of tension in which the epis-
temic status of the proposition remains unsettled in the conceptualizer’s
mind.12
In the context of the present analysis of the epistemic uses of Polish
past perfective and imperfective verbs of verbal communication, it may be
interesting to note that the way in which the above-mentioned aspects of
the proposition’s epistemic force figure in the conception of the profiled
past verbal action seems to correlate with the level of acceptability of using
either the perfective or the imperfective. This is illustrated by the examples
in (13) and (14) below:
(13) a. Lekarz powiedział /?mówił, że w takich przypadkach

Doctor said.PERF /?said.IMPERF that in such cases
to lekarstwo zdecydowanie pomaga.
this medicine decidedly helps
‘The doctor said that in such cases this medicine decidedly helps.’
b. Lekarz ?powiedział /mówił, że w takich

Doctor ?said.PERF /said.IMPERF that in such
przypadkach to lekarstwo chyba pomaga.
cases this medicine probably helps
‘The doctor said that in such cases this medicine probably helps.’
(14) a. Lekarz powiedział /?mówił, że w takich

Doctor said.PERF /?said.IMPERF that in such
przypadkach to lekarstwo pomaga.
cases this medicine helps
‘The doctor said that in such cases this medicine helps.’
b. Sąsiadka ?powiedziała /mówiła, że w takich
Neighbor-she ?said.PERF /said.IMPERF that in such
przypadkach to lekarstwo pomaga.
cases this medicine helps
‘A neighbor said that in such cases this medicine helps.’
Apparently, the presence of factors contributing to the greater epistemic

force of the proposition and thereby, to its immediate and automatic ac-
ceptance (factors such as a strong conviction on the part of the trajector of
the process of verbal communication in [13a] or his authority status in
[14a]) makes the use of the perfective preferable. On the other hand, when
contextual factors (such as the trajector’s weak conviction in [13b] or the
lack of authority in the relevant domain in [14b]) facilitate the construal of
the proposition’s epistemic force as weak, the imperfective becomes more
acceptable. Obviously, when a proposition presented to the conceptual-
izer’s mind is – for whatever reason – endowed with considerable epis-
temic force, it is much more likely that the epistemic control cycle will
reach its result phase smoothly and effortlessly, thereby becoming com-
pleted immediately in the past, when the proposition was presented to the
conceptualizer’s mind in the profiled event of verbal communication. Con-
versely, when the epistemic force of the proposition is relatively weak, the
epistemic control cycle may get and remain “jammed” up to the present at
its inclination stage. In this case, the conceptualizer’s mind can neither
fully accept nor reject the proposition, despite inclining towards one possi-
bility or the other.
4. Conflicting epistemic meanings of the Polish aspectual variants in

the two tenses
The final issue in the present analysis is the question of the relation be-
tween the epistemic value of the perfective/imperfective contrast in the
past tense and its counterpart in the non-past tense. It turns out that in the
past tense the perfective tends to be associated with epistemic certainty and
commitment, and the imperfective with the lack thereof. On the other hand,
in the non-past tense the epistemic values of the two aspectual variants
seem to be reversed: here the imperfective is correlated with a greater de-
gree of epistemic commitment, while the perfective may go together with
the epistemic uncertainty typically associated with the future. The epis-
temic behavior of the Polish perfective and imperfective in the past tense
seems to support the hypothesis put forward by Fleischman (cf. 1995: 539),
who claims that there is a cross-linguistic “attraction” between the imper-
fective and the irrealis, since the notions of irrealis and the lack of epis-
temic certainty seem to be closely related. However, the epistemic values
associated with the perfective and the imperfective in their future-time uses
directly contradict Fleischman’s suggestion: it is the imperfective which
portrays the designated future event as almost bound to occur, and there-
fore – as nearly part of reality. On the other hand, when the perfective is
used, the future event is conceptualized as potential, not yet real, and there-
fore epistemically uncertain.
The question therefore is: Is this seemingly inconsistent epistemic be-
havior of the Polish perfective and imperfective a vagary of grammar? Is
this an illustration of the fact that grammar is extravagant and whimsical, at
least as far as its semantic contribution is concerned?
I believe that the answer to these questions should be in the negative. In
my view, there is nothing extravagant or whimsical about the epistemic
behavior of the Polish perfective and imperfective, and any inconsistency
in that behavior is only apparent. However, to see this one needs to take
into consideration two factors. The first is that a common property of vari-
ous uses of the Polish perfective is that it presents the profiled event as a
self-contained whole, which is complete and independent of anything else.
The imperfective, on the other hand, evokes a conceptualization in which
the profiled phases of the process are embedded within a temporally more
extended configuration. In the case of the prototypical use of the imperfec-
tive, the profiled phases are conceptualized specifically as phases of a
complete process (as e.g. in Piotr zapisywał mój adres, gdy nadjechał
autobus ‘Piotr was writing down my address when the bus came’ (cf. (1b)
above). In turn, when the imperfective is used in reference to complete
events (as, e.g. in (12b) above Piotr mówił, że go [artykuł] skończy ‘Piotr
said that he will finish the paper’), these events are often thought of as
being conceptually linked to the time of speaking, thereby being part of a
temporally more extended configuration – if not in the objective, then at
least in the subjective sense.13
The second factor is that the past, the present and the future are not
simply different slices of the time line. On the contrary, as suggested by
Brisard (1999: 11–12, fn.9), they are experienced by the human conceptu-
alizer as having quite different epistemic properties. Events in the past
have already happened and thus, can be known to us. Their epistemic
status has already been decided upon. Events in the present are in the proc-
ess of happening – hence, the present is where potentiality turns into reality
and where unknown turns into known. The epistemic status of events in the
present is in the process of becoming resolved. Finally, events in the future
are as yet only potential and awaiting possible future realization. Thus,
they are not yet real and cannot be known, but only anticipated or imag-
ined.
What kind of epistemic attitude on the part of the speaker is signaled

by, respectively, the past and the non-past perfective and imperfective fol-
lows directly from the respective prototypical meanings of the two aspec-
tual variants, as well as from the epistemic properties of the past, the pre-
sent and the future time frames, in which the conceptual configurations
evoked by either the perfective or the imperfective are embedded. Since the
perfective evokes the conceptualization of an event construed as a self-
contained whole, it restricts this event solely to the temporal frame in
which it is placed. The past perfective locates an event wholly in the past
and is thus associated with the epistemic attitude prototypically connected
with the past – with the attitude of epistemic certainty and commitment.
Therefore, examples analogous to e.g. (12a) above (Piotr powiedział, że go
[artykuł] skończy ‘Piotr said that he will finish the paper’) convey the con-
strual of not only the profiled past event of verbal communication, but also
of the contents communicated in that event as entities whose epistemic
status was fully settled before the speech time (they were both included in
the conceptualizer’s idea of reality).
On the other hand, the non-past perfective places an event wholly in the
future. Hence, it signals an attitude of epistemic uncertainty, which is –
quite sensibly – associated with the future. Therefore, examples analogous
to e.g. (9a) (Samolot odleci, jeśli mgła opadnie ‘The plane will depart if
the fog falls down’) convey a particular degree of the conceptualizer’s
inclination to accept the profiled event as part of his conception of reality,
but not yet his full epistemic acceptance.
Turning now to the imperfective, I would like to suggest that its past
variant links a past event to the present in the epistemic domain. By linking
the profiled past event to the present, the imperfective endows the concep-
tualization of a past event of verbal communication with the epistemic
value prototypically associated with the present, rather than the past: the
epistemic status of what was communicated in the profiled past action,
instead of having already been settled, is still in the process of becoming
resolved at the time of speaking. Hence, although examples analogous to
(12b) above (Piotr mówił, że go [artykuł] skończy ‘Piotr said that he will
finish the paper’) present the profiled event of verbal communication as
part of the conceptualizer’s idea of reality, they nevertheless construe the
content communicated in that event as still epistemically uncertain at the
speech time.
In the case of the future use of the non-past imperfective, a future event
is again linked to the present. This time, the result of endowing the concep-
tualization of a future event with the epistemic value of the present is that
the profiled future process is viewed as – in a sense – already on its path

from potentiality to reality, from being unknown to being known. There-
fore, examples analogous to e.g (9b) above (W rozkładzie samolot odlatuje
za godzinę ‘According to the timetable, the plane departs in an hour’) pre-
sent the profiled event as immediately accessible to the conceptualizer – if
not in the sense of a direct actual apprehension, then at least in the sense of
accessing a virtual analog of a future actual event in a relevant mental
script.
5. Concluding remarks
In the present paper, I have tried to characterize the epistemic values of the
Polish perfective and the imperfective in the past and in the non-past tense.
I have argued that the epistemic values under consideration may be viewed
as relatively straightforward extensions from the prototypical meanings of
the two aspectual variants in the two tenses, once it is recognized that the
objectively construed future- vs. present-time realization or past comple-
tion vs. non-completion may have subjective counterparts in the conceptu-
alizer’s epistemic sphere. The conceptualization evoked by the epistemic
use of the non-past imperfective is that at the time of speaking the concep-
tualizer mentally activates a virtual analog of an actual future event and
thereby exercises a measure of epistemic control over that event. On the
other hand, the absence of such present-time mental activation, which is
indicated by the use of the non-past perfective, conveys the construal in
which the future event in question remains beyond the conceptualizer’s
epistemic control. In turn, the choice of the past perfective may evoke a
conceptualization in which not only the objectively construed profiled
process, but also the subjective epistemic control cycle is viewed as com-
pleted, with the relevant proposition placed under the speaker’s epistemic
control. On the other hand, the notion of non-completion evoked by the
past imperfective may be construed as pertaining not to the objective non-
completion within the immediate scope of predication, but rather to the
non-completion of the subjective epistemic control cycle, in which the
epistemic status of a proposition remains an open issue. The final claim of
the present paper has been that the conflicting epistemic values of the per-
fective/imperfective contrast, respectively, in the past and in the non-past
tense – far from being a whim of the Polish grammar – are in fact strongly
motivated by the interaction between the conceptual configurations evoked
by the aspectual variants under consideration and the epistemic properties

of the time-frames in which these conceptual configurations are embedded.
In conclusion, I would like to emphasize that – as observed by Lan-
gacker (1987a: 15) – “[w]hat one finds in language depends in a large
measure on what one expects to find”: to the extent to which the analysis
proposed above is considered successful, its potential success is attribut-
able to the theoretical framework which allows an analyst to see that mean-
ing in language involves not only conceptual content but also its construal,
to recognize the fact that our conceptualizations conveyed by linguistic
expressions may pertain not only to things “out there” in the world, but
also to subjective factors and virtual entities of various kinds, or to become
aware of the role idealized cognitive models of various sorts play in shap-
ing our conceptual world and linguistic behavior. What cognitive grammar
offers to a linguist is a theoretical framework of precisely this kind.
Let me finish with a kind of credo. It is my deep conviction that – for
the most part – grammar is not at all extravagant or whimsical in its role of
a tool for conveying meaning. On the contrary, it is often quite strongly
motivated, well-organized, sensible, and it gets its job done. It is sometimes
believed that properties of this kind make the things characterized by them
dull. Leaving open the issue of the applicability of this claim to other areas,
I strongly believe that at least in the case of grammar this is what makes it
all the more interesting. Yet, one needs the apparatus of cognitive linguis-
tics to be able to conduct a detailed conceptual analysis of grammatical
structures and to see all the beauty in their motivated, reliable functioning.
Notes
* I would like to thank Elżbieta Górska and Ronald W. Langacker, who gener-
ously offered me their time and commented on earlier versions of the present
analysis. Both my analysis and the way it is presented benefited greatly from
their wise guidance. My gratitude is also due to Dagmar Divjak for all her
helpful comments. Needless to say, all the remaining flaws and errors are en-
tirely my own.
1. As will be illustrated later on, the Polish non-past perfective designates a
future event, while the non-past imperfective may profile either a present
process (the prototypical use) or a future one. The present study will be con-
cerned with the respective epistemic values of the two aspectual variants in
the non-past tense, when they are both used in reference to future processes.
2. Note that in what follows I will use the terms ‘complete events’ or ‘complete
processes’ in reference to what is profiled by the examples under considera-
tion. These terms are meant to capture the fact that the examples in question
evoke the idea of a full instance of temporally bounded and internally hetero-
geneous processes. I decided to use the term ‘complete process/event’ instead
of ‘completed process/event’, as the latter term seems to bring about connota-
tions with the past (it is only a past process, a process that already occurred
that might or might not have been completed). In turn, the examples consid-
ered in the present paper evoke the idea of both past and future processes
which are, in both cases, conceptualized in their totality.
3. For a more detailed discussion of the semantics of przez- and w-phrases in
their temporal uses and their relation to the spatial ones, see Kochańska
(1996).
4. Such scripts are instances of what Langacker (cf. e.g. 1999) calls virtual
documents.
5. Note that, in addition to the habitual reading, sentence (10b) may also have an
interpretation pertaining to the speaker’s plan for future actions (‘I’m having
my breakfast in an hour’). However, it is only habitual reading that is relevant
at this point.
6. This conception is what Langacker calls the structural plane (cf. Langacker
1999: 95–97 and 2001:270, as well as the discussion of the struc-
tural/phenomenal distinction in Goldsmith and Woisetschlaeger 1982).
7. Note that the characterization proposed here involves just a slightly different
phrasing of how Langacker characterizes the English contrast between the
will-future and the future use of the simple present tense in terms of the actu-
ality/virtuality distinction (cf. e.g. 1999, 2001). The idea is that both the will-
future in English and the non-past perfective in Polish profile a future event in
the actual plane. On the other hand, the future uses of both the English simple
present tense and the Polish non-past imperfective designate a present-time ac-
tivation of an entry in a mental script, that is, a present occurrence of the rele-
vant event in a virtual plane. In this context it might be observed that, for ex-
ample, in (9a) above the Polish non-past perfective, which designates an actual
future event (Samolot odleci, jeśli mgła opadnie), is translated into English
with the will-future (‘The plane will depart if the fog falls down’). On the
other hand, the Polish non-past imperfective in (9b), which profiles a present
virtual occurrence (W rozkładzie samolot odlatuje za godzinę) is translated
with the English simple present tense (‘According to the timetable, the plane
departs in an hour’). It may also be noted that the English translations of the
remaining Polish examples considered in this section (the examples in [10])
contain present-tense verbs, instead of the will-future. Both (10b) and (c) are
translated with simple present verb forms, while (10a) is translated with the
present progressive. A discussion of why the English equivalent of (10a) be-
haves differently than the translations of (10c) and (d) is beyond the scope of
this paper. What is important for the present analysis is that all the English
sentences in question contain present-tense verb forms, instead of the will-
future, as all of them profile present-time activations of entries in a mental
script, rather than a corresponding future actual event.
8. This last way of thinking about the uses of the imperfective under considera-
tion has been suggested to me by Ronald Langacker (personal communica-
tion).
9. This way of thinking has been suggested to me – independently and on differ-
ent occasions – by Elżbieta Górska and Ronald Langacker. See also the dis-
cussion on the same point in section 3.1 and section 4 below.
10. The same point is made for non-past perfectives and imperfectives in section
2.2 above, and in reference to the Polish perfective and imperfective in general
in section 4 below.
11. It has been suggested to me by Ronald Langacker that the use of the perfective
in (12a) shows an affinity with the so-called factive predicates in that it also
involves strong identification of the actual ground with the virtual ground in-
voked by the complement clause (for a discussion of factivity cf. Langacker
2004 section 5). The term strong identification is meant to convey the idea
that the virtual ground and grounding relation conjured up mentally in the
conceptualization evoked by the complement clause are fully identified with
the actual ground and the actual grounding relation, with the consequence that
the conceptualizer not only formulates in his mind the proposition in the com-
plement clause, but also embraces it as part of his own conception of reality
(cf. Langacker 2004: 548–549). One difference between the epistemic use of
the perfective and factive predicates seems to be that factive predicates just
presuppose the conceptualizer’s commitment to their truth, while the epistemi-
cally used perfective seems to specifically convey this commitment.
12. I would like to note at this point that my use of terms such as ‘resistance’ or
‘force’ in the above description of the cognitive model governing human func-
tioning in the epistemic domain is by no means accidental. It was first ob-
served by Talmy (1988) that there are numerous abstract concepts which are
understood metaphorically in terms of forces, counterforces, presence or ab-
sence of barriers, etc. Such force-dynamic elements present in our understand-
ing of abstract notions also have linguistic manifestations. It should come as
no surprise that the ideas of force and resistance are present in the characteri-
zation of a cognitive model of how humans function with respect to issues of
knowledge – in fact, this conceptual area was one of the clearest cases in
which the existence and importance of force-dynamics in cognition and lan-
guage was first demonstrated (cf. the analysis of English epistemic modals in
Talmy [1988], as well as in Sweetser [1990]).
13. As mentioned above, this way of characterizing the Polish perfective and im-
perfective in schematic terms has been suggested to me – independently and
on different occasions – by Elżbieta Górska and Ronald W. Langacker. See

also the comments to the same effect in sections 2.2 and 3.1 above.
References
Brisard, Frank
1999 A critique of localism in and about tense theory. Ph.D. diss., Univer-
sity of Antwerp.
Fleischman, Suzanne
1995 Imperfective and irrealis. In Modality in Grammar and Discourse,
Joan Bybee and Suzanne Fleischman (eds.), 519–551. Amster-
dam/Philadelphia: John Benjamins.
Goldsmith, John, and Erich Woisetschlaeger
1982 The logic of the English progressive. Linguistic Inquiry 13 (1): 79–
89.
Kochańska, Agata
1996 Temporal meanings of spatial prepositions in Polish: The case of
przez and w. In The Construal of Space in Language and Thought,
Martin Puetz and René Dirven (eds.), 491–508. Berlin/New York:
Mouton de Gruyter.
2002a Selected issues in the semantics of the Polish imperfective. A cogni-
tive grammar account. Ph. D. diss., Warsaw University.
2002b A cognitive grammar analysis of Polish non-past perfectives and
imperfectives: How virtual events differ from actual ones. In
Grounding. The Epistemic Footing of Deixis and Reference, Frank
Brisard (ed.), 349–390. Berlin/New York: Mouton de Gruyter.
1987a Nouns and verbs. Language 63: 53–94.
1987b Foundations of Cognitive Grammar. Vol. 1. Theoretical Prerequi-
sites. Stanford, California: Stanford University Press.
1990 Subjectification. Cognitive Linguistics 1 (1): 5–38.
1991 Foundations of Cognitive Grammar. Vol. 2. Descriptive Application.
Stanford, California: Stanford University Press.
1993 Universals of construal. Proceedings of the Annual Meeting of the
Berkeley Linguistic Society 19: 447–463.
1999 Virtual reality. Studies in the Linguistic Sciences 29 (2): 77–103.
2001 The English present tense. English Language and Linguistics 5 (2):
251–272.
2002 The control cycle: why grammar is a matter of life and death. Pro-
ceedings of the Second Annual Meeting of the Japanese Cognitive
Linguistics Association: 193–220.
2004 Aspects of grammar of finite clauses. In Language, Culture, and

Mind, Michel Achard and Suzanne Kemmer (eds.), 535–577. Stan-
ford: CSLI Publications.
ms Enunciating the parallelism of nominal and clausal grounding.
Sweetser, Eve E.
1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects
of Semantic Structure. Cambridge: Cambridge University Press.
Talmy, Leonard
1988 Force dynamics in language and cognition. Cognitive Science 12:
49–100.
Conjunctions, verb forms, and epistemic stance in
Polish and Serbian predictive conditionals
Barbara Dancygier and Radoslava Trnavac
Abstract
In this paper, we use data from temporal, conditional, and coordinate con-
structions in Polish and Serbian to establish basic formal and semantic pa-
rameters defining conditional meaning in the two languages. Relying on the
mental spaces analysis of conditional constructionsin English (Dancygier
and Sweetser 2005), we describe relative salience of different constructional
parameters in Polish and Serbian. Specifically, we establish, through an
analysis of a broad range of conjunctions and verb forms, that, similarly to
English, predictive meaning is central to conditionality in both Polish and
Serbian. However, in contrast to English, the Slavic languages under scru-
tiny rely to a lesser degree on specific conditional conjunctions and on
clause order iconicity, and in fact favor different parameters (e.g., tense,
mood, and aspectual forms, or overt markers of sequentiality) as central to
predictiveness at a constructional level. At the same time, compositional
mechanisms which are responsible for the emergence of predictive condi-
tional meanings, even in absence of overt markers of conditionality, remain
very similar to English, thus giving support to the general concept of con-
structional compositionality.
Keywords: conditionals, coordinate constructions, predictive constructions,

Polish, Serbian
1. Constructional compositionality and conditional interpretation
Recent work within the broad area of cognitive linguistics devotes signifi-
cant attention to the correlation between syntactic form and meaning. More
specifically, work in Construction Grammar seeks new theoretical solu-
tions to numerous questions related to the correlation between meaning and
form on the sentential level.1 However, recent work within the Mental
182 Barbara Dancygier and Radoslava Trnavac
Spaces Theory (Fauconnier [1985] 1994, 1997, Fauconnier and Sweetser

1996, Dancygier and Sweetser 2005, Verhagen 2005) also addresses spe-
cific questions related to the mechanisms of sentence interpretation.
One group of constructions seems a particularly interesting field of en-
quiry: the constructions which express conditional meanings. Conditionals
are a historically fruitful area of linguistic research (see, for example,
Traugott et al. 1986, Athanasiadou and Dirven 1997, Declerck and Reed
2001), since they represent a particularly intricate set of relationships be-
tween form and meaning. Recent work on these constructions, especially
the analyses proposed in Sweetser (1990), Dancygier (1998) and Dancygier
and Sweetser (2005), develop an approach to constructional analysis which
focuses to an equal degree on the overall resulting interpretation and on the
specific grammatical and lexical components which may contribute to the
meaning of a range of constructions, rather than one construction in par-
ticular. Dancygier and Sweetser (2005) discuss a whole range of construc-
tions in English in which conditional meaning is expressed, even though
conditional conjunctions may not be present (as in coordinate or paratactic
constructions), or are combined with other expressions, such as the focus-
ing adverbs even and only. While not being ‘conditionals’ in the general
sense of the term, these constructions are interpreted conditionally because
of the presence of formal and lexical signals which prompt important com-
ponents of conditional meaning. The idea that constructional meaning may
emerge out of specific forms, rather than be correlated with a complete
syntactic/lexical pattern, is discussed in terms of the mechanisms of con-
structional compositionality. The term, first introduced in Sweetser (1999)
and further substantiated in Dancygier and Sweetser (2005), suggests that
any form expected in a specific type of construction can, in the right con-
text, ‘single-handedly’ prompt an interpretation which would otherwise be
suggested by the construction as a whole.
In the analyses we will rely on below, conditionality is understood as an
interpretation involving two mental spaces: the space set up in the protasis
p (typically, the first clause, introduced by a conditional conjunction simi-
lar to if), and its extension space q (the apodosis, often not introduced by a
conjunction; in some cases introduced by a conjunction similar to then).
The p space structures a situation which is a condition for the situation q,
although the conditioning relation may be established in different cognitive
domains: content, epistemic, speech act, or metalinguistic (see Sweetser
[1990] for an exhaustive discussion of the nature of cognitive domains and
clausal relations they appear in; also Traugott [1989] for a discussion of
the rise of epistemic meanings). The central group of constructions are
Polish and Serbian predictive conditionals 183
those functioning in the content domain – those which establish a condi-

tional relationship between facts or states of affairs. As Dancygier (1998)
argues, they are typically responsible for a conditional reasoning which is
predictive in meaning. In the analysis proposed in Dancygier (1993) and
Dancygier (1998) (then developed in Dancygier and Sweetser 2005), pre-
dictive meaning is understood as consisting of the following components:
non-assertiveness of the p-clause (that is, non-factuality of the p space or
its non-positive epistemic stance), sequential and causal relationship be-
tween p and q (q-space is considered to follow the p-space temporally and
causally, and inherits p’s non-factual interpretation), and overall future
reference. In the construction as a whole, p is not predicted, while q is pre-
sented as predictable from it, should it become factual.
The distinctions among different types of conditionals, and also be-
tween conditional and temporal conjunctions have been discussed by Fill-
more (1986, 1990a, b) in terms of epistemic stance. As Fillmore argues,
English temporal conjunctions like when represent the speaker’s positive
stance to the future event (so that the speaker expects the future situation to
occur as described), while a conditional conjunction like if does not ex-
press such a commitment and is thus neutral or (when the speaker has rea-
sons to believe the situation is unlikely) negative in stance. The concept of
negative stance as applied to past tense verb forms which are not used to
refer to past situations has been further developed in Fleischman (1989),
and then applied to an explanation of counterfactual and other distanced
conditionals in Dancygier (2002) and Dancygier and Sweetser (2005). Fur-
thermore, the correlation between various conditional, causal and conces-
sive conjunctions and their epistemic stance has also been explored in
Dancygier and Sweetser (2000).
Dancygier and Sweetser (2005) consider different groups of condition-
als, such as predictive, epistemic, speech act, and metalinguistic condition-
als. Our analysis will focus primarily on predictive conditionals as the
category with salient constructional features, also appearing in other re-
lated constructions. In English, a typical example of a predictive construc-
tion is sentence (1):
(1) If he is elected president next year, he will travel a lot.
In (1), the conjunction if is a space builder setting up the non-factual,

non-predictive future space p, while the q-clause represents the space
which can be predicted to follow p temporally and causally. The present
tense in p, which is used to set up a future space, is an important element of
the construction’s form, and, as Dancygier and Sweetser (2005) argue, is

an important compositional feature of a predictive conditional, marking the
non-assertive and non-predictive meaning of p, and at the same time de-
termining p’s role and place in a predictive construction. The fact that q
then uses the predictive modal will completes the interpretation of the
whole construction as predictive.
These formal features, especially the conjunction if and the verb forms,
mark the construction as expressing the speaker’s neutral stance to the
mental spaces described – there is no commitment on the part of the
speaker as regards the likelihood of p and q actually happening. The lack
of a predictive modal in p marks the space as non-predictive, and thus as
representing the speaker’s neutral epistemic stance. As has been argued
broadly in the literature on conditionals, past tense and aspectual forms
imposed onto the pattern represented by (1) change the stance to negative
(as in If he were/had been elected president, he would/would have traveled
a lot, where the events are presented as unlikely or counterfactual), without
affecting the general interpretation of the construction as predictive.
The above description, which is based on predictive constructions in
English, focuses on the use of tense and modality, but the role aspectual
forms can play in constructing conditionality in other languages is much
less clear. For example, a correlation between imperfective past tense and
conditionals has been observed across different Germanic, Romance and
Balkanic languages (see Fleischman 1995, Boogaart 1999, Iatridou 2000,
Giannakidou and Zwarts fc., Brisard and De Mulder ms.). However, as we
show below, the postulated link between imperfective aspect and condi-
tionals is not clearly present in Polish and Serbian.2 In fact, there are nu-
merous cases where a sequence of perfective (rather than imperfective)
forms gives rise to a conditional reading. We hypothesize that the capacity
for the perfective aspect to participate in building conditionality in Polish
and Serbian can be related to a commonly recognized property which Bar-
entsen (1998) calls ‘sekventnaja svjaz’ (‘sequential connection’), which is
absent from the semantics of imperfective aspect. According to Barentsen
(1998), in some languages (and more conspicuously in Russian) the event
represented by a perfective verb has to be linked to an external situation,
located before or after the event itself. A well-known manifestation of the
property of ‘sequential connection’ is the aoristic use of perfective forms
when the sentence uses a chain of verbs occurring in succession, and where
the links in the chain are not only temporal but also resultative (so that the
situation mentioned later in the chain is seen as resulting from the one de-
scribed by the directly preceding verb)3. For example, a Russian equivalent
of the sentence He got up, went up to the window, and waved, discussed by
Barentsen (1998:52), requires all three verbs to be used in the past form.
Also, observations made in Trnavac (2006) confirm the link between per-
fective aspect and ‘sequenced connection’ in the cases of Serbian and Rus-
sian. A more thorough discussion of the role perfective aspect plays as a
marker of ‘sequenced connection’ in the construction of conditional mean-
ings in Polish and Serbian requires further investigation, but the examples
we discuss below support the hypothesis.
In contrast to Polish and Serbian, English seems to rely primarily on
tense and modality in its construction of conditional meaning. Among oth-
ers, the contrast between the present tense and the past tense emerges as a
primary marker of epistemic stance. Interestingly, while past tense can also
be used to mark negative stance outside of conditionals (as in I thought you
were married, when the speaker is expressing uncertainty, rather than de-
scribing a past belief), the present tense may be used as a marker of neutral
stance in paratactic constructions such as (2) and (3),4 but it also comple-
ments other features of the sentences (such as the possibility of sequen-
tial/causal link between p and q) in suggesting a predictive interpretation:
(2) He makes one mistake and he is out. (Davies 1986)
(3) He makes one mistake and he’ll be fired.
The constructions in (2) and (3) can thus be interpreted similarly to a

conditional, and the interpretation relies on their constructional features
prompting for the same meanings that they trigger in a conditional. The
present tense presents the first conjunct as a neutral-stance, non-predictive
future mental space, while the future event described in q is interpreted as
a result predicted from p. In this way, a construction which does not use
any explicit markers of conditionality can compositionally prompt the con-
ditional meaning: the verb forms and the sequential/causal relation be-
tween p and q are enough to suggest a conditional interpretation.
The analysis proposed by Dancygier and Sweetser (2005) for English
constructions with conditional meaning relies on formal parameters spe-
cific to English. In what follows, we consider data from Serbian and Polish
to consider the range of constructions marking conditionality in the two
languages. More specifically, we will focus on formal features of Polish
and Serbian conditionals to determine the compositional contribution of
the conjunctions and verb forms used. As we show, the parameters of in-
terpretation such as epistemic stance or sequentiality play a role similar to
that in English, but the formal means establishing these meanings are dif-
ferent in interesting ways. Furthermore, both of the languages we consider
confirm that a whole range of conditionally interpreted constructions is
prompted by mechanisms of constructional compositionality.
2. Temporal conjunctions and conditional meaning
In the present section, we will outline the use of major temporal conjunc-
tions in terms of the stance they represent and the typical verb form pat-
terns they correlate with. While Polish and Serbian conjunctions are rela-
tively flexible in their co-occurrence with verb form patterns, there are
some restrictions which point to epistemic stance as a major distinguishing
category. In our examples below, we will highlight the most common pat-
terns and the important exceptions, but an exhaustive discussion of all the
possible combinations and their meanings would exceed the limits of this
paper.
Epistemic stance, as first introduced by Fillmore (1986, 1990a, b), is a
category which also applies to temporal conjunctions. In Fillmore’s origi-
nal discussion, a temporal conjunction like when marks the speaker’s posi-
tive stance – according to what the speaker knows at the moment of
speech, the future space being set-up is likely to become factual. The verb
forms used in a futurate construction like When they get married, they will
move to France, the same as the ones in (1), also indicate the predictive
interpretation of the whole construction. The similarity between if and
when constructions in English is primarily apparent among futurate con-
structions, but the difference in stance distinguishes them well: while when
typically marks positive stance, if does not, correlating with neutral and
negative stance instead. In Polish and Serbian, for comparison, the bound-
ary between temporal and conditional conjunctions is less clear, and fur-
ther complicated by the possibility for temporal conjunctions to be used in
constructions marking negative stance through conditional mood. Conse-
quently, the primary stance contrast is not between positive stance on the
one hand, and neutral and negative on the other, but rather between the
negative stance on the one hand, and the neutral and positive ones on the
other, with the latter two not being distinguished formally.
Kada is the primary temporal conjunction in Serbian. As example (4) il-
lustrates, it introduces a predictive futurate construction with a positive
stance, very much in the same way as its English equivalent.5 The contrast
between the non-predictive protasis and the predictive apodosis is also
marked similarly to English – the present tense marks p, while the future
tense marks q.6 However, the temporal clause itself usually requires perfec-
tive aspect – a grammatical feature which, as we show below, plays an
important role in predictive constructions in the languages under investiga-
tion here.
(4) Kada dođeš, daću ti knjigu.

when come.PRES.PF.2SG give.FUT.PF.1SG you.2SG.DAT book
‘When you come, I will give you a book.’
Interestingly, a temporal construction with kada can receive a condi-

tional interpretation with the addition of the conditional mood marker bi
(in contrast to English, where when is aligned with positive stance). The
typical position of bi in the protasis is immediately following the conjunc-
tion, while in the apodosis it appears after the verb; the person/number
inflection is attached to bi, rather than to any other constituent.
(5) Kada bi došao, dala bih

when IRR come.PRET.PF give.PRET.PF IRR
ti knjigu.
you.2SG.DAT book
‘If you came, I would give you a book.’
The Serbian conjunction kada thus seems particularly useful as a

marker of futurate predictive constructions, and, in contrast to English
when, is not in itself restricted in its correlation with stance. The verb
forms are thus the main source of information on the stance – present and
future can receive a positive interpretation, while the conditional mood
establishes the speaker’s negative stance to the reasoning. We should also
note that the perfective aspect seems to be an important contribution to the
constructional set-up of (4), where kada refers to the future, but the nega-
tive stance examples can use imperfective and perfective forms. This
seems to support our tentative suggestion made earlier that the choice of
perfective aspect correlates with other markers of a predictive construction
with a neutral or positive stance. We will return to the issue below.
Furthermore, Serbian temporal constructions suggest that the verb
forms chosen can override the stance typically associated with the conjunc-
tion. Unlike in English, where negative stance is primarily marked by the
past tense, negative stance in Serbian (and Polish) is marked by the forms
of conditional mood. Even though the functions of conditional mood are
varied, they generally correlate with functions very similar to those of

negative stance in English – unlikelihood, lack of commitment, non-
assertiveness, deference, politeness, etc. However, the negative marking
brought into the construction by mood may alter the standard interpretation
of the conjunction itself – as is the case with kada. Clearly, the stance
status of a conjunction is generally weaker than the status of a morphologi-
cal form specializing in non-factual meanings.
In Polish, the use of temporal conjunctions is different in several ways.
There are two basic temporal conjunctions, kiedy and gdy. Though gram-
mars are often vague in this respect, it seems possible to claim that kiedy
marks positive stance, while gdy marks neutral stance.7 Characteristically,
kiedy is preferred when a specific point in time is being referred to,
whether in the past or the future, while gdy is more common in reference to
the future, especially when the time of the expected realization of the p
space is not specified (see Bielec 1998). In fact, gdy is sometimes trans-
lated into English as if, while kiedy is typically rendered as when.
(6) Kiedy/?Gdy skończysz osiemnaście lat,

when finish.PRES.PF.2SG eighteen
dostaniesz samochód.
get.PRES.PF.2SG car
‘When/*If you turn eighteen, you’ll get a car.’
(7) Gdy pogoda się poprawi,

when weather REFL improve.PRES.PF.3SG
pojadę na wakacje.
go.PRES.PF.1SG for vacation
‘When/If the weather improves, I’ll take a vacation.’
Both (6) and (7) represent standard predictive reasonings, where the
prediction of q depends on the prior occurrence of the non-predicted p, and
both are significantly more common with the perfective present form in
both clauses. This is not surprising in view of the fact that the perfective
present form in Polish is the only available expression of future perfective
events (the only other verb form with future meaning is the periphrastic
future imperfective).8
In Polish, contrary to the case of Serbian kada, the association of kiedy
with positive stance is strong enough that conditional mood is not possible
in predictive constructions with kiedy as a space builder. This makes kiedy
more similar to English when than to Serbian kada. However, the restric-
tion on conditional mood in combination with kiedy does not seem to be a

general restriction on temporal conjunctions, since the temporal conjunc-
tion associated with weaker stance, gdy, is different in this respect. Not
only does it frequently appear with negative stance forms, but the combina-
tion of gdy and the conditional mood marker by is also considered an inde-
pendent conjunction, gdyby, which obligatorily marks negative stance,
whether for reasons related to politeness, various degrees of epistemic dis-
tance, or counterfactuality.9 Gdyby is thus a standard conditional conjunc-
tion10, associated uniformly with negative stance:
(8a) Gdybyś chciała, zabrałabym cię do kina.

if.IRR.2SG want.PRET.IPF take.PERF.IRR.1SG you to the movies
‘If you (had) wanted to, I would take you (have taken you) to the
movies.’
(8b) Gdybyś zadzwoniła przed piątą, zabrałabym

if.IRR.2SG call.PRET.PF before five take.PERF.IRR.1SG
cię do kina.
you to the movies
‘If you (had) called before five, I would take you (have taken you)
to the movies.’
The verb forms in a gdyby construction such as (8) do not clearly mark
temporal reference, so the sentence can be interpreted, depending on the
context, as a polite offer of a future action or a regret about a past missed
chance. Although there are no clear-cut distinctions, sentence (8a), with an
imperfective verb in p, is more likely to be interpreted as present or future,
while the perfective aspect in the protasis of (8b) is more readily under-
stood as referring to the past; still, a time adverb such as jutro (tomorrow)
or wczoraj (yesterday) could be used in both versions and would push the
interpretation towards the future or the past respectively (the future inter-
pretation is then more strongly marked for politeness). It seems, then, that
the presence of conditional mood (negative stance) is the overriding con-
structional feature in these cases.
Another conjunction in Polish is jak, whose temporal sense is best de-
scribed as when and after put together (see Bielec 1998) – it does imply
completion of p prior to q.11 Jak is thus most common in futurate sen-
tences, occasionally used in past contexts (often with tylko (‘only’), which
then stresses the immediacy of q following p), and practically unacceptable
in construction referring to the present (unless they are generic). The as-
sumption of the jak-space event being completed before the main clause or
q-clause event makes jak an ideal space builder of predictive constructions,
mainly with neutral stance:
(9) Jak przyjdziesz, pogadamy.

when/as come.PRES.PF.2SG talk.PRES.PF.1PL
‘When/After you come, we’ll talk. /If you come, we’ll talk.’
However, perhaps as a result of its strong association with predictive

constructions, jak is often considered a conditional conjunction (see Pis-
arkowa 1984, Bielec 1998). Indeed, jak can naturally be understood condi-
tionally, perhaps as a result of the weakness of the contrast between neutral
and positive stance, which is the sole parameter distinguishing predictive
uses of when from the predictive uses of if. Furthermore, jak-constructions
can be marked with conditional mood (so that the negative stance equiva-
lent of [9] is Jakbyś przyszedł, pogadalibyśmy), and then the conditional
meaning of jak is clear. Interestingly enough, possibly because of its ‘tem-
poral’ source meaning, jak seems to be restricted to predictive construction
(futurate, distanced, or generic), but is not acceptable in epistemic condi-
tionals like (10), where the sentence cannot be understood temporally:
(10) *Jak jest prezydentem, to

when/as is.PRES president, it
(znaczy że) również dowodzi armią.
(mean.PRES that) also lead.PRES army
‘If he is the President, then (it means that) he is also the Com-
mander-in-Chief.’
To conclude, both Polish and Serbian have predictive constructions,

whether temporal or conditional in meaning. However, the stance contrasts
are realized differently in Polish and Serbian. While English conjunctions
when and if can be claimed to distinguish positive from neutral stance, verb
forms override such distinctions in Polish and Serbian, where the main
contrast observed is thus between the positive/neutral stance on the one
hand, and negative stance on the other. Because negative stance is typically
marked by the presence of -bi or -by (the marker of conditional mood),
while there are no clear markers of the contrast between neutral and posi-
tive stance, the boundary between temporal and conditional future predic-
tions is not really clear. The importance of negative stance is also observed
in the way in which temporal conjunctions obtain conditional meaning as
soon as the mood marker is added. In rare cases, as in the case of Polish
gdy/gdyby, the rise of the conditional conjunction eventually obliterates the
underlying commonality of the temporal and conditional uses. As regards
gdy, its underlying neutral stance seems to have supported an emergence of
a conjunction which further increases its distance from positive stance,
especially since the salience of the neutral versus negative contrast typi-
cally exceeds that of the contrast between neutral and positive.
In conditional sentences with temporal conjunctions aspect seems to
play a secondary role. However, it does support the positive/neutral versus
negative contrast mentioned above. There is a general tendency in both
languages for sentences with the positive/neutral epistemic stance and fu-
ture reference to occur with perfective aspect (see use of kada with the
future events in Serbian, as well the occurrence of the perfective present in
the case of kiedy and gdy with positive/neutral stance in Polish), while
sentences with negative epistemic stance seem to freely occur with both
aspectual forms in both languages.
It could be concluded that in Polish and Serbian the joint contribution
of temporal conjunctions and aspectual forms is overridden by the role of
the contrast between indicative and conditional mood. By the same token,
negative stance emerges as the central parameter of conditionality in the
two languages.
3. Conditional conjunctions
The categories of epistemic stance apply similarly to the conjunctions spe-

cializing in conditional meaning. However, since positive stance is not
marked by conditional conjunctions proper, the main contrast in this area
of adverbial clauses is between neutral and negative stance. As in the sec-
tions above, our review of conditional conjunctions will focus on futurate
predictive constructions.12
In Serbian, the major conditional conjunction correlated with predictive
reasonings is ako.13 It marks the construction as representing neutral epis-
temic stance, as in (11):
(11) Ako dođeš, daću ti knjigu.

if come.PRES.PF.2SG give.FUT.PF.1SG you.2SG.DAT book
‘If you come, I will give you a book.’
For comparison, the conjunction exclusively connected to counterfactu-

ality and negative epistemic stance, whether in relation to past or present,
is da:
(12) Da sam na tvom mestu, ja bih predlog prihvatila.

if AUX on your place I IRR suggestion accept.PRET.PF
‘If I were you, I would accept the suggestion.’
(13) Da si došao, dala bih ti knjigu.

if AUX come.PRET.PF give.PRET.PF IRR you.2SG.DAT book
‘If you had come, I would have given you a book.’
Interestingly, Serbian conjunctions make a specific distinction between

negative stance in general and counterfactual meaning specifically, and can
thus express negative stance in ways which are constructionally independ-
ent of the conditional mood forms. Comparing our da examples with an
ako-construction using conditional mood reveals interesting contrasts:
(14) Ako bi ti nabavio novac za kuću, ja

if IRR you get.PRET money for house I
bih bila srećna.
IRR was happy
‘If you got (would get) the money for the house, I would be happy.’
While da (as in [12] and [13]) marks counterfactuality as the only avail-
able interpretation of the construction’s protasis, it also does not require
the use of the conditional mood in the protasis; consequently, bi, the pri-
mary marker of negative epistemic stance, appears in the apodosis only.
Also, sentences with da are partially sensitive to the preference for the
imperfective aspect – the protases referring to the present require an imper-
fective form. The case of da as a marker of counterfactuality also raises
interesting questions about the nature of counterfactual meanings as such.
While earlier analyses of English, such as Dancygier (2002), Dancygier
and Sweetser (2005), make a very strong argument for counterfactuality
being constructed on the basis of negative stance and the structure of the
base space, Serbian seems to have a construction specializing in counter-
factuality as such. For comparison, the construction with ako in (14),
which refers to a future situation (and is thus distanced, but not counterfac-
tual), employs bi, the marker of conditional mood/negative stance both in
the protasis and in the apodosis.
In Polish, the primary conditional conjunction is jeśli, with its variant

jeżeli. It is a conjunction which clearly marks neutral stance, and seems to
cover a range of meanings very similar to ako. Its neutral stance function
also seems to be related to its etymology, with the present tense form of the
verb być (to be) accompanied by a question marker li. It is thus used in
predictive constructions, but is also a standard choice in all the non-
predictive ones. An example of the predictive use is (15):
(15) Jeśli Janek przyjdzie, na pewno ci pomoże.

if John come.PRES.PF surely you help.PRES.PF.3SG
‘If John comes, he will surely help you.’
Similarly to other conjunctions, in Polish, jeśli appears in a conditional

mood conjunction /negative stance variant with by and past tense form, as
in (16):
(16) Jeśliby Janek przyszedł, na pewno by ci pomógł.

if.IRR John come.PRET.PF surely IRR you help.PRET.PF
‘If John came, he would surely help you.’
Counterfactual meanings are more commonly expressed through the

conjunction gdyby (discussed above), although gdyby is not restricted to
this use the way da is.14
4. Predictiveness and verb forms
The basic predictive construction as defined by Dancygier (1998) uses the

protasis to set up an unpredicted future space, and then describes its pre-
dicted consequences in the apodosis. English predictive constructions mark
the specific role of the protasis by restricting the use of the predictive mo-
dal will. A typical protasis of a futurate predictive construction in English
uses the present tense, rather then the modal, while referring to a future
situation – the phenomenon termed backshifting in Dancygier (1998),
Dancygier and Sweetser (2005). Backshifting has often been assumed to be
a peculiarity of English, resulting primarily from the fact that future mean-
ings are primarily expressed through predictive modality represented by
will. A closer look at Serbian and Polish, however, suggests that tense and
aspect play an important role in distinguishing the protases of predictive
constructions.
Serbian is an interesting example in this respect. Future events in Ser-

bian can be described with the use of one of two verb forms. Future I
(where a clitic ću or će accompanies the infinitive form of the main verb)
represents a probable (predicted) event in the future, while Future II (pre-
sent tense of the auxiliary verb biti plus past participle) denotes a future
event which precedes, or serves as background to some other future event.
Future II is used only in subordinate clauses, such as temporal, conditional
or relative clauses, though in contemporary Serbian Future II of perfective
verbs is often replaced with the perfective present. Consider the future
forms in (17), with Future II in the protasis and Future I in the apodosis:
(17) Ako budem dobila poziv, otići

if be.PRES.1SG get.PRET.PF invitation go.FUT
ću na prijem.
FUT CLIT on reception
‘If I get the invitation, I will go to the reception.’
The periphrastic Future II verb form (budem dobila) in the protasis of

(17) does not in fact employ a specific future tense form of any kind, and is
appropriate only in those clauses which present an event as background to
another event. In a predictive construction like (17), the verb form in the
protasis plays a role which is very similar to the role the present plays in
English predictives – it marks the mental space as not predicted, but neces-
sarily prior to the space in the apodosis, which is set up as an ordinary,
predicted future. In other words, Serbian seems to have a verb form which
specializes in marking the ‘background’ clauses of bi-clausal constructions,
with predictive constructions playing a prominent role among them. Con-
structions like (17) thus give very strong support to the cross-linguistic
validity of the general description of predictive constructions given in ear-
lier research and further support the constructional role of backshifted pre-
sent in English predictives.
Serbian ako is also used with two types of present tense forms in the
protasis p: imperfective and perfective, as in (18) and (19), respectively. In
this case, aspectual usage affects the interpretation in an important way.
The imperfective form, as in (18a), typically refers to a present situation,
although it can be used to mark a future approved plan similarly to English,
as in (18b); neither example is thus a standard predictive construction.
However, constructions like (19), with the present perfective form in the
protasis p and Future I in the apodosis q, refer to the future, and thus repre-
sent a typical predictive scenario.
(18a) Ako me pažljivo slušaš, sve

if me carefully listen.PRES.IPF.2SG all
ćeš razumeti.
FUT CLIT understand.INF.IPF
‘If you are listening to me carefully (now), you will understand
everything.’
(18b) Ako sutra putuješ, javi mi se.

if tomorrow travel.PRES.IPF.2SG call.IMPER.PF 1SG.DAT REF
‘If you are traveling tomorrow, call me.’
(19) Ako on sutra navrati, pričaćemo

if he tomorrow drop by.INF.PF talk.FUT.IPF
o tome.
about that
‘If he drops by tomorrow, we will talk about that.’
The difference between the imperfective protases of (18 a and b) and

the perfective protasis of (19) seems to rely on other, more specific com-
ponents of predictiveness. In Dancygier’s original description (1998), pre-
dictive meanings rely crucially on the sequential relationship between the
mental space set up by p and the one set up by q. Given that the q event can
only be predicted on the basis of an assumed occurrence of p, p has to be
understood as also preceding q temporally.15 In her explanation of the cor-
relation between sequentiality and predictiveness, Dancygier (1998) argues
that p has to conceptually precede q even if the clauses represent states or
iterative meanings, since in a predictive meaning the inception of the p
state will still have to precede the inception of the q state. The use of a
perfective form in the protasis of a conditional like (19) supports the se-
quential relation between p and q, by profiling p as a bounded event, com-
pleted before the inception or occurrence of q. The preference for perfec-
tive forms in predictive constructions thus seems to arise out of the need to
establish a sequential relationship between p and q. At the same time, the
present tense, in contrast to Future I, weakens the predictive role of the
clause. In the constructions which use Future II the non-predictive role of p
is clearly marked, while in the ones with present tense the non-
predictiveness is established by failure to use a future form, which is a
strategy very much like backshifting in English. The fact that Future I does
not appear in predictive protases is, however, the most significant regular-
ity, presenting Serbian predictive constructions as a very coherent class.
Furthermore, this confirms the cross-linguistic validity of Dancygier’s

(1998) definition of predictiveness, where p and q are marked as having
different roles.
We should add here that while Future I is indeed barred from predictive
protases, and thus cannot mark an assumption as an unpredicted back-
ground to a prediction, it does appear in the protases of epistemic condi-
tionals. This observation further supports the contrast between predictive
and non-predictive constructions postulated earlier (Dancygier 1993, 1998,
Dancygier and Sweetser 2005) in the description of English. While will is
typically not used in predictive protases in English (see Close 1980), it can
appear as a contextually available prediction in the protasis of a non-
predictive conditional, such as If he will be late, let’s have dinner now. The
protasis of such a sentence picks up a prediction made by another speaker
in the available context to use it in the reasoning, but the reasoning itself is
not predictive, since the apodosis does not predict having dinner now, but
rather performs a speech act of suggesting. The fact that the verb forms
mark similar distinctions in Serbian, as in Ako će on kasniti, mi možemo da
počnemo da večeramo, confirms the general validity of the criteria pro-
posed on the basis of English.
In Polish, the use of tenses in predictive protases is not restricted in any
significant way. The observations that can be made rely primarily on the
fact that the selection of available tenses in Polish is narrower than in Ser-
bian. There are two expressions of the future: the Future Imperfective,
consisting of the future form of the auxiliary być (be) and the past tense
imperfective form of the main verb, alternating with the imperfective in-
finitive (będę czytała/będę czytać – I will be reading), and the perfective
present form, przeczytam (I will read / will have read). None of these
forms is barred from appearing in predictive conditional protases and apo-
doses, but the majority of examples encountered in natural discourse rely
on the perfective present in both clauses, as in (20):
(20) Jeśli się postarasz, zdasz ten egzamin.

if REFL try.PRES.PF.2SG pass.PRES.PF.2SG that exam
‘If you try hard, you’ll pass that exam.’
Polish has no verb forms which are clearly associated with predictive-
ness. At the same time, the perfective present, as in Serbian, suggests two
meaning components crucial to predictive meanings – boundedness and
sequentiality. One can thus conclude that futurate conditionals in Polish
rely to a lesser degree on the contrast between the unpredicted p space and
the predicted q space, and to a more significant degree on the sequential

relation between p and q. However, as in sentences with temporal conjunc-
tions, aspect is not a primary element of conditionality, though it supports
the distinction between neutral and negative epistemic stance in the way
described for temporal constructions. Serbian counterfactual constructions
with conjunction da are the case in point here, since they show a prefer-
ence for imperfective aspect with the present tense in the protasis.
5. The apodosis marker to
In the preceding sections we have focused on the conjunctions introducing

the protasis and the specific verb forms in p and q. Most analyses of condi-
tionals also note that a conditional construction may be additionally sup-
ported by a conjunction introducing the apodosis. In most cases the con-
junction, such as the English then, is considered a formal marker rather
than a meaningful addition to the construction’s interpretation, especially
in view of the fact that in many constructions then (or its equivalent) is
optional, even though there are also clear restrictions on its application (see
Iatridou 1994, von Fintel 1991, 1994). There are, however, at least two
analyses which acknowledge a meaningful and constructionally relevant
status of a marker like then.
In the analysis of English conditionals proposed by Dancygier and
Sweetser (1997, 2005), then is treated as referring anaphorically to the
space set-up in the protasis, assuming its predicted or asserted status
(which p as such does not have), and continuing its elaboration with the
content of q. Then is often assumed to naturally co-occur with if, but it is
not typically used with even if conditionals or generic ones. The discussion
in Dancygier and Sweetser (1997) explains the restrictions, as well as the
exceptions to such restrictions. English then, however, is a rather specific
case in that its interpretation is also influenced by its relationship to the
temporal conjunction (and multi-functional wh-word) when. For compari-
son, in the case of to, the Polish equivalent of then, there is no tempo-
ral/conditional correlation. As Tabakowska (1997) observes, to is related to
deictic uses similar to it or this, and thus lacks a temporal dimension.16 In
the interpretation proposed by Tabakowska, the presence of to instructs the
hearer to treat the construction as a figure/ground sentence (see Langacker
1987 and 1991), such that q is the figure which can be conceptually located
within the ground search domain designated by p. The advantage of Taba-
kowska’s interpretation of to as a radial category is, among other things,
that it explains the usage of to in a number of different constructions, while

also accounting for the restrictions in its use (for example, to cannot mark
the apodosis if the order of clauses is q, if p, and not if p, q, as in *To idź,
jeśli chcesz).
Within conditional constructions, to also seems to play an interesting
role in the cases in which epistemic stance is not consistently expressed
throughout the construction, as in (21):
(21) Jeślibyś/Gdybyś się postarał, to zdasz.

if.IRR.2SG REFL try.PRET.PF then pass.PRES.PF.2SG
‘If you tried hard, you would pass.’
The shift from negative stance in p to neutral stance in q is not accept-

able without to, which suggests that to plays a role in establishing the rela-
tionship between the protasis space and the apodosis space. The observa-
tion coincides with the treatment proposed by Tabakowska (1997) for
topicalizing constructions or constructions which call for an interpretation
similar to a conditional without marking conditionality explicitly. The
framing of the space preceding to as the search domain, as Tabakowska
argues, also seems to frame the whole construction in a particular way. In
the case of (21), the use of to re-affirms the predictive power of the con-
struction as a whole, while helping the hearer reject the use of negative
stance in p as an expression of politeness or wish. The construction as a
whole thus draws a prediction from a background assumption which is
presented as doubtful.
Interestingly, even the use of to does not make the construction in (22)
below acceptable – a hesitant expectation cannot be claimed to follow from
a neutral stance assumption about a future event. In a sense, once p is pre-
sented as a background to prediction, q is expected to make that prediction.
In fact, the use of to strengthens the expectation that the construction will
be understood predictively, so the incoherence of (22) is increased when to
is present.
(22) *Jeśli się postarasz, (to) zdałbyś.

if REFL try.PRES.PF.2SG (then) pass.PRET.PF.IRR.2SG
‘If you try hard, you would pass.’
We are focusing our discussion here on predictive conditionals, but we

should note that to is also common in other conditionals, especially epis-
temic ones. Its general role seems to be the establishment of a
(con)sequential link between the clauses – when to is used, p is unambigu-

ously framed as the space which precedes q in the reasoning and, at an
appropriate level, also causes it – whether q happens to be a prediction or a
conclusion. This is coherent with Tabakowska’s observation that to can
frame constructions without an overt protasis, by presenting q as derivable
from the physical or social context. In other words, what to contributes to a
construction is the framing of the relationship between the clauses in such
a way that q follows from p.
The above review of data from Polish and Serbian focused on the mean-
ing contribution of individual constructional markers of conditionality:
conjunctions, verb forms, the use of to. The observation which emerges out
of the discussion is that, in spite of significant similarities, Serbian and
Polish differ in their treatment of predictive constructions in interesting
ways. Both languages profile a predictive construction with respect to the
criteria established earlier for English (Dancygier 1993, 1998): neutral or
negative epistemic stance, the marking of the difference between the back-
ground p clause and the predictive q clause, and the concept of sequential-
ity and causality, as applied to different cognitive domains. However, Ser-
bian and Polish are not entirely aligned in that they give prominence to
different means of establishing the predictive meaning of a construction.
Serbian relies primarily on verb forms like Future II and Future I to estab-
lish the difference between p and q in the reasoning and prompt the
(con)sequential relation between them. The restriction against Future I
being used in protases plays a significant role here, while the Serbian
equivalent of then, onda, seems to play a minimal role in conditional con-
structions and does not appear often. For comparison, the limited reliance
on the boundedness of the perfective present points to a much weaker role
of the verb forms in Polish. Instead, Polish gives significantly more weight
to the apodosis marker to, which, while not obligatory, has the power of
framing a construction as having a (con)sequential meaning, even if other
expressions are not unambiguously suggesting it. One might claim that the
predictive reasoning as such is in general less salient in Polish, while the
sequential (or figure/ground) interpretation of the construction is a more
central aspect of conditionality.
In the next section we will consider constructions which are interpreted
as predictive without overtly using a conditional or temporal conjunction.
We will focus on the means used in Serbian and Polish to establish a pre-
dictive meaning in such contexts, in order to confirm the initial observa-
tions suggested by the data reviewed so far.
6. Coordinate and conjunctionless conditionals
In our overview above, we focused on standard conditional constructions,

and the conjunctions which are used in them. However, a number of analy-
ses of conditionals also consider constructions which express conditional
meaning in absence of a specific conjunction (cf. Haiman 1978, 1980,
1983, 1986, Davies 1986).
Among English constructions, the ones discussed most often are coor-
dinate sentences with conjunctions and and or, especially those using im-
perative forms in the protasis:
(23) a. We finish this one paper and we are done.

b. Finish this one paper and you are done.
c. Finish this one paper, and you’ll get tenure.
d. Finish this one paper or you won’t get tenure.
In English, such constructions are interpreted predictively. From the

point of view of verb forms, they share the tendency not to use will in p,
while the sequence of clauses is iconic of the sequence of events. As
Dancygier and Sweetser (2005) have argued, these two features – the order
of p and q iconically marking the sequential interpretation and the verb
form of the first conjunct marking non-predictiveness and future reference
– are responsible for the interpretation of such sentences as expressing a
predictive conditional meaning. Furthermore, Dancygier and Sweetser
(2005) propose a similar interpretation to account for the predictive condi-
tional meaning of constructions which use verbless p clauses or do not use
any conjunctions:
(24) One more paper, and you get/you’ll get tenure.

Another book, another good review.
While some of these constructions additionally support the sequential

interpretation by suggesting that p constitutes a step in a sequence of simi-
lar events (consider the expressions one more or another), the general pat-
tern is quite similar to the one represented in the sentences under (23).
Coordinate and conjunctionless constructions with conditional meaning
constitute a very broad and varied class and an exhaustive discussion of
similar constructions in Polish and Serbian would exceed the limits of the
present paper. We will thus focus the discussion here on a brief review of
linguistic means which establish the sequential relationship between the

clauses and the non-predictive status of p.
6.1. Verb forms
In establishing predictive meanings of coordinate and conjunctionless con-

structions, Serbian and Polish rely on verb forms to a limited degree. How-
ever, the limitations seem to have different sources in the two languages. In
Serbian, the use of Future II in the protasis, which would unambiguously
establish the desired relation between p and q, is restricted to subordinate
clauses and is thus barred from appearing in the coordinate or conjunc-
tionless constructions. At the same time, no other forms really take over. It
is in fact a possibly significant comparative observation – the forms mark-
ing predictiveness are quite weak in English, since they also have other
functions, but their contribution to the overall meaning of the construction
is significantly stronger. In other words, the verb forms used in English
conditionals, while not specialized, do contribute compositionally to the
construction’s interpretation. In Serbian, for comparison, the existence of a
specialized form in fact limits the compositional contribution of other
forms, such as the much less specific contribution of the perfective present,
which is acceptable in some constructions, but certainly not very common.
The presence of a clear ‘subordinate clause/main clause’ verb form pattern
in Serbian, with a predictive function, prevents the coordinate construc-
tional frame from being elegantly re-framed as p/q predictive pattern, in
spite of possible sequentiality. Furthermore, coordinate constructions in
Serbian are more commonly interpreted as expressing correlations, and are
thus often understood as expressions of generic correlative or circumstan-
tial patterns, rather than as conditionals.
Interestingly, constructions with the imperative form in p (whether per-
fective or imperfective) and the coordinate conjunction i (and) seem to
dominate:
(25) Pročitaj ovu knjigu i videćeš da sam u pravu.

read.IMPER.PF.2SG this book and see.FUT that AUX in right
‘Read this book and you will see that I was right.’
(26) Čitaj ovu knjigu i uživaćeš u njoj!

read.IMPER.IPF.2SG this book and enjoy.FUT.IPF.2SG in her
‘Read this book and you will enjoy it.’
Although, as (25) and (26) suggest, both aspects can occur in the prota-
sis and apodosis of this construction, conditional reading is more salient in
(25), with two perfective sequenced events. Sentence (26), for comparison,
represents two parallel (non-sequential) actions and its reading is better
described as temporal. Furthermore, the verb in the apodosis of (25) has to
be marked by the future tense, since perfective present cannot mark future
reference in a coordinate construction Presumably, then, coordinate con-
structions like (25) can receive a conditional interpretation because they
rely heavily on the verb forms – the imperative in p and the future in q.
This constructional pattern is thus quite similar to English, where the im-
perative, tenseless, but suggesting a future non-predicted situation desired
by the speaker, plays the role of a protasis marker in predictive reasonings.
The imperative is thus a verb form which, when used in protases, composi-
tionally prompts for a predictive interpretation of the construction both in
Serbian and in English, but in Serbian it also has to be matched by a future
form in the apodosis.
In Polish the verb form selection in coordinate constructions is less sig-
nificant. As could be expected, imperfectives are extremely rare, whether
with future or present reference, and express a correlation or a generic
statement. Perfectives, on the other hand, which play an important role in
conditionals, are also common in coordinate constructions, and play a simi-
lar role of establishing a sequence of two causally related events.
(27) Przeczytasz tę książkę, a zobaczysz

read.PRES.PF.2SG this book, and see.PRES.PF.2SG
że miałam rację.
that have.PRET.1SG right
‘You read this book and you’ll see that I was right.’
Such constructions in Polish typically require that both p and q use the
perfective present forms, which refer to the future while profiling a se-
quence of two bounded events. In many cases, however, the perfectives in
p and q do not ensure the construction’s acceptability if a conjunction is
absent, as in (28):
???
(28) Przeczytasz tę książkę, zobaczysz że miałam rację.
read.PRES.PF.2SG this book see.PRES.PF.2SG that have.PRET.1SG right
‘You read this book, you’ll see that I was right.’
The sentence is somewhat more acceptable if a phrase suggesting an-

other step in a sequence is added to p – ?Przeczytasz jeszcze jedną książkę,
zobaczysz że miałam rację/ Read one more book, you’ll see that I was
right, but the sentence is still more acceptable with a coordinate conjunc-
tion. Clearly, the perfective present in Polish is not a salient enough marker
of predictiveness, and, while it may contribute the meaning of sequential-
ity, it does not establish it unambiguously, even if the sequence of clauses
is also iconic of the sequence of events. A possible conclusion, in agree-
ment with the observations made in the discussion of conditional and tem-
poral constructions above, is that predictiveness as a constructional feature
is less salient in Polish than a specific relationship between p and q, as
determined by the conjunction used.
The role of expressions like one more or another in supporting a predic-
tive interpretation of coordinate or conjunctionles constructions has been
noticed with respect to English in Haiman (1986), and further discussed in
Dancygier and Sweetser (2005), especially with respect to the scalar NP,
NP constructions, such as Another day, another dollar, which would be
unacceptable without them. The minimal unit semantics such expressions
evoke is a constitutive feature of coordinate and conjunctionless construc-
tions, although it is not restricted to them. However, in other constructions,
such as The X-er, the Y-er, the schema of paired scales is also readily in-
terpreted predictively, since subsequent steps along one scale are seen as
linked to subsequent steps along another scale, and they are naturally con-
ceived as linked causally.17 The fact that a conjunctionless construction
such as (28) in Polish can become acceptable when one of these expres-
sions is added further supports the observation that paired scales are con-
ceptually linked to predictiveness. What is more, this is also confirmed in
Serbian, in examples like (29):
(29) Napravi li još jednu grešku, leteće napolje!

make.PRES.PF (PRT IRR) more one mistake fly.FUT.IPF outside
‘(He) makes one more mistake, (he) is out.’
The protasis of (29) receives a scalar interpretation based on the expres-

sion još jednu gresku, (one more mistake). Moreover, the particle li re-
quires scalar markers like only, one more, etc. to be present. The construc-
tion overall is interpreted as a future prediction, but it relies
compositionally on several forms. The verb form pattern requires perfec-
tive present tense in p, and Future I in q, but the role of the protasis is addi-
tionally marked for neutral stance with the use of the particle li, which is a
common marker of unassertiveness or interrogative meaning in Slavic lan-

guages. It is also one of the few predictive constructions in Serbian in
which conjunctions do not occur, and the predictive interpretation emerges
compositionally out of three different factors: the scalar expression, the
particle li, and the verb form pattern. The construction, though not very
common, strongly supports the understanding of predictiveness as compo-
sitionally constructed.
Even if p uses the imperative form, which seemed the closest to being a
marker of an unpredicted future situation when no specialized tense form is
available, conjunctionless constructions in Polish are not readily accept-
able, as in (30a). However, the addition of a coordinate conjunction a, as in
(30b), not only makes the sentence acceptable, but also allows it to be in-
terpreted as a predictive construction, with the imperative clause taking the
function of p and the second conjunct functioning as q.
???
(30a) Przeczytaj tę książkę, zobaczysz że miałam rację.
read.IMPER.PF this book, see.PRES.PF.2SG that have.PRET.1SG right
‘Read this book, you’ll see that I was right.’
(30b) Przeczytaj tę książkę, a zobaczysz że miałam rację.

read.IMPER.PF this book, and see.PRES.PF.2SG that have.PRET.1SG right
‘Read this book, and you’ll see that I was right.’
What these examples show, then, is that in Polish the verb forms and
the iconic sequence of p and q play a role in the interpretation of construc-
tions as predictive, but they are not sufficient. The presence of a conjunc-
tion is required to help establish a link between the clauses. Without it, as
in (30a), the clauses are treated as independent utterances.
The examples in this section suggest that coordinate and conjunc-
tionless constructions in Serbian and Polish rely on very much the same
constructional features as their standard conditional counterparts. While
Serbian is more dependent on its verb forms pattern, Polish relies to a
higher degree on the selection of an appropriate conjunction. In both lan-
guages, the sequenced perfective forms prompt conditional reading and
neutral epistemic stance, as in the previously analyzed types of construc-
tions. However, as in the cases of other constructions, the perfective forms
in coordinate/conjunctionless sentences are not sufficient to mark condi-
tionality without the presence of other constructional features. In both lan-
guages these sentences need the support of a coordinate conjunction or a
particle like -li in order to be interpreted conditionally. Aspect again plays

a secondary role in building conditionality.
In the next section, we will review the conditional uses of coordinate
conjunctions in Serbian and Polish.
6.2. Coordinate conjunctions versus to
In the literature on English, two coordinate conjunctions are typically men-

tioned in the context of conditional meaning: and and or. Constructions
with or, such as Give me the money or I’ll shoot, have been widely dis-
cussed as examples of speech acts such as a threat or a warning, based on
the interpretation where the refusal to follow the order in p is presented as
bringing about the unwelcome q as its result (see Fillenbaum 1976, 1986).
The mental space set-up of such sentences has also been discussed in
Dancygier and Sweetser (2005). The equivalents of or in Polish and Ser-
bian, albo and ili respectively, seem to structure constructions which are
very similar to the English constructions with or, so we will not explore
these constructions here in any more detail. The equivalents of and, on the
other hand, seem to raise interesting questions.
In Serbian, the use of i (and) is restricted to constructions with impera-
tive protases, as in (25) and (26) above, and not acceptable in coordinate
constructions which are quite standard in English. For example, a pattern
like (31) is not often used in Serbian, but quite common in English:
?
(31) Devojka uradi to i
girl do.PRES.PF.3SG this and
smatraju je lošom.
consider.PRES.IPF.3PL AUX bad
‘A girl does that and they consider her bad.’
In Polish, coordinate conjunctions are common in constructions with a

conditional meaning. There are two such conjunctions, i and a, with the
former primarily used to mark the standard additive sense, coexistence, or
correlation, and the latter often used to contrast the two conjuncts in some
respect (quite often, the clauses connected with a have different subjects).
Interestingly, the use of the conjunction i in Polish rarely correlates
with a conditional interpretation of the relation between the clauses. With a
symmetric use of the perfective present in both clauses the correlative or
additive function of and still seems to prevail (as in [32a]), and the con-
structional contribution of the imperative (in [32b]) does not override it:
(32a) ?Przeczytasz tę książkę, i zobaczysz że miałam rację.

read.PRES.PF.2SG this book, and see.PRES.PF.2SG that have.PRET.1SG right
‘You read this book, and you’ll see that I was right.’
(32b) ?Przeczytaj tę książkę, i zobaczysz

read.IMPER.PF.2SG this book, and see.PRES.PF.2SG
że miałam rację.
This suggests an interesting pattern, especially in comparison with Ser-

bian. It appears that both languages cannot allow a correlative coordinate
conjunction like i to be re-interpreted sequentially and, as a consequence,
also conditionally, as can be seen in (31) and (32a). However, while the
imperative in p can occur in an i construction in Serbian (see [25] and
[26]), it still is unlikely in Polish, as in (32b). As a result, neither (32a) nor
(32b) are naturally read conditionally.18 Both sentences would, however, be
fully acceptable with the conjunction a, which is coordinate, but also con-
trastive, even though its contrastive function is not central then.
The overall pattern in Polish seems to present coordinate conjunctions
as primarily dedicated to correlative and additive meanings, with limited
potential for profiling sequentiality or predictiveness. It is possible that the
symmetry of verb forms is an additional factor acting against using coordi-
nate constructions predictively, suggesting symmetry in the roles played by
p and q.
As we could expect, the asymmetrical relationship between p and q in
Polish constructions is most typically marked by to. To is the most natural
choice for a number of verb form patterns, and is a preferred form in the
types of sentences investigated above, as in (33a and 33b):
(33a) Przeczytasz tę książkę, to zobaczysz

read.PRES.PF.2SG this book, then see.PRES.PF.2SG
że mialam rację.
‘You read this book, and you’ll see that I was right.’
(33b) Przeczytaj tę książkę, to zobaczysz

read.IMPER.PF.2SG this book, then see.PRES.PF.2SG
że miałam rację.
To sum up, conjunctionless constructions are not common in Polish.

Among coordinate conjunctions, only a is used in predictive constructions,
while to is the most common form used to link p and q. Given that to also
occurs in all conditionals, predictive as well as non-predictive, it emerges
as the most versatile and widespread marker of conditionality. More spe-
cifically, because of its deictic, rather than temporal, origin, it connects p
and q in a way that suggests that before q-situation is considered, the p-
situation has to occur or become a fact, depending on the specific domain
of conditional meaning. At the same time, being essentially a marker which
introduces q into the construction, it does not participate in or interact with
the marking of the epistemic stance of p, other than indicating it as concep-
tually prior to q. This accounts for the fact that to appears in all conditional
constructions, as well as in other ones (see the review in Tabakowska
[1997]), and that, when other constructional features are present, it may
also contribute to the construction’s predictive interpretation.
7. Conclusions
The picture which emerges from the contrastive review of major parame-
ters of predictive conditionality suggests many similarities between the
constructional and semantic features of the English sentences and those of
Serbian and Polish. Not surprisingly, all three languages are sensitive to
the parameters of epistemic stance and express it through the choice of
conjunctions and verb forms. Also, all three languages use verb forms to
mark the respective roles of p and q clauses in prompting for the predictive
reading and to signal the speaker’s negative stance. Furthermore, all three
select constructional forms which best express aspects of predictiveness
such as sequentiality or unpredicted status of p. However, the differences
in the degree to which a language relies on any of the parameters men-
tioned is what distinguishes them in significant ways.
In English, if is the primary conditional conjunction; it correlates with
the neutral or negative stance of constructions, while when is primarily
used as a positive stance space builder. In Polish and Serbian, for compari-
son, conjunctions as markers of stance are compositionally less salient than

the conjunctions in English. While conditional conjunctions are primarily
neutral or negative, the temporal ones do not clearly distinguish between
positive and neutral stance, and can also align with negative stance. Re-
gardless of whether they are primarily temporal or conditional, conjunc-
tions which mark negative stance do so via forms containing a mood
marker (-by or -bi). Even though conjunctions like Polish gdyby or jeśliby
have been grammaticalized as indivisible wholes, the ensuing protasis then
lacks a mood marker, while the apodosis requires it, so that each of the
clauses is marked for negative stance and the patterns of mental space em-
bedding are maintained. As a result, conditional predictive constructions
are introduced either by temporal or by conditional conjunctions, and the
presence or absence of conditional mood markers is the most salient com-
positional feature. The only notable exception is the Serbian da, which
prompts for a counterfactual interpretation while not marking mood in the
protasis.
In comparison to English, Polish and Serbian constructions with tempo-
ral meaning are further distinguished by the parameter of aspect. Contrary
to what has been suggested in earlier research, perfective aspect in Polish
and Serbian often supports predictive sentences with neutral and positive
epistemic stance. Even though aspect plays a secondary role in establishing
epistemic stance, it does compositionally contribute to a predictive inter-
pretation of the construction as a whole, by highlighting the sequential
relationship between the protasis and the apodosis. As we have suggested,
the precise nature of the semantics of sequentiality as prompted for by the
perfective aspect is not entirely clear, but concepts such as boundedness or
sequential connection (a link to an external situation, located before or
after the event itself) suggest a sufficiently salient explanation to call for
further research, especially on languages in which aspectual forms play a
more significant role in the grammar than they do in English.
Conjunctions are the most commonly used markers of the different
status of the protasis and the apodosis, but they are not the only ones. In-
terestingly, the contrast between the non-predicted protasis and the predic-
tive apodosis in Serbian is marked similarly to what was observed in Eng-
lish – via a restriction on tense selection (effectively equivalent to what
Dancygier [1998] describes as backshifting), thus offering an additional
confirmation of the validity of the parameter which has often been seen as
a peculiarity of English. The specialized marker of Serbian protases of
predictive constructions is the Future II tense. Polish differs in this respect,
as it uses the futurate form (the perfective present) in both protases and
apodoses. Futurate conditionals in Polish thus rely to a lesser degree on the

verb form contrast between the unpredicted p space and the predicted q
space, apparently because the selection of tenses with future reference in
Polish is narrower than in Serbian. However, Polish conditional construc-
tions of all kinds, whether with temporal, conditional, or coordinate con-
junctions, or no conjunction at all, regularly highlight the relationship be-
tween the clause playing the role of the protasis and the one playing the
role of apodosis by additionally marking the apodoses with to – the deictic
non-temporal marker. To seems to compositionally compensate for the lack
of backshifting or specific verb form patterns.
Both Polish and Serbian have coordinate conditionals, though they rely
to a different degree on two major constructional features of such condi-
tionals – the order of the protasis and the apodosis iconically marking the
sequential interpretation, and the verb form of the first conjunct marking
non-predictiveness of p and the future reference of the construction. In
Serbian, the most common coordinate constructions are ones with the im-
perative form in the protasis and the coordinate conjunction i. However,
constructions without conjunctions, with two sequenced verbal forms, are
also possible, though they have to be supported with the irre-
alis/interrogative particle li. One can conclude that in absence of an appro-
priate construction, Serbian requires that the neutral (non-asserted or non-
predicted) status of p be compositionally profiled by an imperative form or
the particle li. In Polish, perfective aspect paired with the iconic order of
the protasis and apodosis is not sufficient to prompt the conditional inter-
pretation, so to appears as a constructionally salient marker as well.
The above review of constructional parameters of conditionals in Polish
and Serbian confirms two important observations made earlier on the basis
of data from English (cf. Dancygier and Sweetser 2005). On the one hand,
the observations about Polish and Serbian confirm the crosslinguistic valid-
ity of the category of predictive constructions, with conditionals being
central members of the class, and temporals as well as coordinate and con-
junctionless constructions diverging in some respects, while preserving the
most central aspects of meaning. It thus seems likely that further studies of
different languages will provide more substantial confirmation of the ap-
proach to conditionality instantiated here, or point to typological differ-
ences which will use the presence or absence of predictive meanings as the
core criterion. On the other hand, Polish and Serbian show with rather un-
expected clarity that the approach to constructions in which specific as-
pects of a construction’s meaning are seen as correlated with specific for-
mal parameters brings new and crosslinguistically non-trivial results.
While expressing predictiveness in ways strikingly parallel to English, the

two languages investigated here realize individual parameters such as epis-
temic stance, sequentiality, or the overall schema of predictive reasoning in
ways which are tailored to the formal means available and their composi-
tional potential.
All three languages thus rely on constructional compositionality, but in
ways appropriate to the strength of the compositional contribution and the
scope of applicability of a given form. To give just one example, let us
consider the issue of the (sequential) connection between p and q and their
different roles in the reasoning. In English conditionals, then makes that
specific meaning contribution, but in many cases seems superfluous, and is
very rarely required. At the same time, and, being an ‘all-purpose’ coordi-
nate conjunction, can easily link the steps in a predictive reasoning, thanks
to its availability in sequential contexts. In Serbian, onda (then) is very
rare, and i (and) can only appear if supported by other means of distin-
guishing the roles of p and q (such as imperatives or li). In Polish, i is typi-
cally not possible, but a, which is coordinate and contrastive at the same
time, is common, apparently because contrasted mental spaces will not be
confused with co-temporal ones; to is overwhelmingly more common, be-
cause it announces p and q as linked in a sequential manner, without im-
posing a strict temporal meaning. Each language thus selects a parameter
which will contribute to the meaning most saliently. English relies on ico-
nicity and an ‘unpredicted’ sense of the present tense (although the form
has other uses, obviously), Serbian requires that the unpredicted status of p
is marked unambiguously and puts little emphasis on the sequence, while
Polish practically ignores the ‘unpredicted’ p, but requires a clear marking
of the (con)sequential relation between p and q. In each case a non-central
conditional construction is available, with a similar meaning, but prompted
by different forms. Most importantly, though, the form selected in each
case further prompts for an understanding of the whole construction as
predictive. In Dancygier and Sweetser (2005) the present tense in p was
presented as a formal parameter of great compositional salience in English.
The facts from Polish and Serbian assign a similar compositional role to
other formal parameters, but with the same effect for the construction as a
whole. The concept of constructional compositionality thus receives inde-
pendent confirmation.
Notes
1. We are referring here to the approaches represented primarily in Goldberg

(1995), Fillmore and Kay (1999), Fillmore, Kay and O’Connor (1988).
2. For more counterexamples to the assumed general connection between imper-
fective aspect and modal forms in Germanic and Slavic languages see Trnavac
(2006) and Boogaart and Trnavac (ms).
3. This ‘anaphoric’ capacity of the aoristic use of perfective aspect is also used in
Kamp and Reyle’s (1993) Discourse Representation Theory.
4. For further discussion of related constructions see Haiman (1978, 1980, 1983,
1986), Blakemore (1987), Davies (1986).
5. Kada also appears in generic constructions, but these remain outside the scope
of the present paper.
6. Perfective present is one of the forms of expressing futurity in Serbian; we will
return to a fuller description below. For a more detailed discussion on the per-
fective present in Serbian see Stevanović (1967), Vojvodić (1989), Ašić
(2000) and Trnavac (2006).
7. Kiedy is also the standard wh-word, parallel to when; for comparison, gdy is
only used as a conjunction.
8. From the point of view of the focus of this paper the actual classification of the
perfective present form as Future or Present does not seem to be crucial (see
Fisiak et al 1978, Dickey 2000 for some discussion). What really matters is
that the form effectively functions as the perfective future form. In our glosses,
we will mark it as perfective present (PRES-PF).
9. A less prominent –by conjunction in Polish is żeby, which primarily expresses
a wish. It is thus related to conditional conjunctions, but does not contribute
much to our analysis.
10. For reviews of Polish conjunctions see, among others, Bielec (1998), Gro-
chowski et al. (1984), Jodłowski (1976).
11. Jak is etymologically related to an adverbial pronoun of manner, and as sug-
gested by Pisarkowa (1984), it is the most frequent marker of conditionality in
contemporary colloquial Polish. We are presenting it here as a temporal con-
junction, but – mainly because of its consistently marking neutral stance – it
can also be seen as primarily conditional. Only a thorough investigation of jak
can clarify the priority among the meanings, but it is possible to assume that
jak, as other conjunctions in Polish, has both a temporal and a conditional
sense. A similar conjunction, kako, also exists in Serbian. It is primarily an ad-
verbial pronoun of manner, but when employed as a conjunction, it projects a
temporal interpretation with positive epistemic stance, where two events follow
in immediate succession. The immediacy sense correlates with a temporal
rather than a conditional interpretation, and with the occurrence of perfective
aspect, as in example (i):
(i) Kako dodješ tamo, odmah

when/after come.PRES.PF.2SG there immediately
mu se obrati.
3SG.DAT REFL turn to somebody.IMPER.PF.2SG
‘As soon as you get there, speak to him.’
12. In this review, we will not refer to concessive conjunctions or conjunctions

equivalent to unless. Unless is a conditional conjunction in a rather specific
sense (see Declerck and Reed 2000, Traugott 1997, Dancygier 2002), and thus
its equivalents call for a separate discussion.
13. For a more detailed discussion of the conditional conjunctions in Serbian see
Gojmerac (1977), Ivić (1983), Katičić (1984–85) and Mrazović and Vukadi-
nović (1990).
14. In Polish, counterfactual meaning can be emphasized with the negative particle
nie, used in the protasis and in the apodosis, which seems to overtly mark the
contrast between the real and the unreal. Thus a sentence Gdyby Janek nie
przyszedł, na pewno by ci nie pomógł (If John hadn’t come, he surely wouldn’t
have helped you) is unambiguously counterfactual. The constructional role of
such a use of negation is related to epistemic stance issues in an interesting
way, but it cannot be further explored in this paper. For a possible connection
between negation and epistemic stance see Verhagen (forthc.).
15. As regards other types of conditionals, such as epistemic or speech act condi-
tionals, the sequentiality is postulated in their respective cognitive domains,
but predictive conditionals rely on a sequence of events.
16. The temporal equivalent of then, wtedy, does not appear as a constructional
correlate of either temporal or conditional conjunctions.
17. For an in-depth discussion of the correlation between conditional and scalar
meanings see Schwenter (1999).
18. Although an additional stress on the second verb may prompt it in some con-
texts.
References
Ašić, Tijana
2000 Le présent perfectif en Serbe: temps, mode ou puzzle? Cahiers de
Linguistique Française 22: 275–294.
Athanasiadou, Angeliki, and René Dirven (eds.)
1997 On conditionals again. Amsterdam: John Benjamins.
Barentsen, Adrian
1998 Priznak “sekventnaja svaz’’’ i vidovoe protivopostavlenie v
russkom jazyke [The property of ”sequential connection” and as-
pectual opposition in Russian]. In Tipologija vida: problemy,

poiski, rešenija [Typology of aspect: problems, questions, solu-
tions], M. Ju. Čertkova (ed.), 43–58. Moskva: Škola “Jazyki
russkoj kul’tury”.
Bielec, Dana
1998 Polish: an essential grammar. London/New York: Routledge.
Boogaart, Ronny
1999 Aspect and temporal ordering: a contrastive analysis of Dutch
and English. The Hague: Holland Academic Graphics.
Boogaart, Ronny, and Radoslava Trnavac
(ms.). Imperfective aspect and epistemic modality.
Brisard, Frank, and Walter de Mulder
(ms.). Aspects of subjectivity and virtuality in the meaning of the French
imparfait. Submitted to Linguistics.
Close, R. A.
1980 Will in if-clauses. In Studies in English Linguistics for Randolph
Quirk, Sydney Greenbaum, Geoffrey Leech, and Jan Svartvik
(eds.), 100–109. London: Longman.
Couper-Kuhlen, Elizabeth, and Berndt Kortmann
2000 Cause, condition, concession, contrast. Berlin: Mouton de
Gruyter.
Dancygier, Barbara
1993 Interpreting conditionals: time, knowledge and causation. Journal
of Pragmatics 19: 403–434.
1998 Conditionals and prediction. Cambridge: Cambridge University
Press.
2002 Mental space embeddings, counterfactuality, and the use of
unless. English Language and Linguistics 6 (2): 347–377.
Dancygier, Barbara, and Eve Sweetser
1997 “Then” in conditional constructions. Cognitive Linguistics 8 (2):
109–136.
2000 Constructions with if, since and because: causality, epistemic
stance, and clause order. In Cause, condition, concession, con-
trast, Elizabeth Couper-Kuhlen and Bernd Kortmann (eds.), 111–
142. Berlin: Mouton de Gruyter.
2005 Mental spaces in grammar: conditional constructions. Cam-
bridge: Cambridge University Press.
Davies, Eirlys
1986 The English imperative. London: Croom Helm.
Declerck, Renaat, and Susan Reed

2000 The semantics and pragmatics of unless. English Language and
Linguistics 4 (2): 205–241.
2001 Conditionals: a comprehensive empirical analysis. Berlin/New
York: Mouton de Gruyter.
Dickey, Stephen M.
2000 Parameters of Slavic aspect: a cognitive approach. Stanford,
CA: CSLI Publications.
Fauconnier, Gilles
1994 Mental spaces. 2nd ed. Cambridge: Cambridge University Press
[1st edition 1985].
1997 Mappings in thought and language. Cambridge: Cambridge Uni-
versity Press.
Fauconnier, Gilles, and Eve Sweetser (eds.)
1996 Spaces, worlds, and grammar. Chicago: Chicago University
Press.
Fauconnier, Gilles, and Mark Turner
1996 Blending as a central process of grammar. In Conceptual struc-
ture, discourse, and language, Adele Goldberg (ed.), 113–130.
Stanford: Center for the Study of Language and Information
(CSLI).
1998a Conceptual integration networks. Cognitive Science 22 (2): 133–
187.
1998b Principles of conceptual integration. In Discourse and cognition,
Jean-Pierre Koenig (ed.), 269–283. Stanford: CSLI.
2002 The way we think: conceptual blending and the mind's hidden
complexities. New York: Basic Books.
Fillenbaum, Samuel
1986 The use of conditionals in inducements and deterrents. In On
Conditionals, Elizabeth Traugott, Charles Ferguson, Judy S.
Reilly, Alice ter Meulen (eds.), 179–195. Cambridge University
Press.
1986 Varieties of conditional sentences. ESCOL 3 (Eastern States Con-
ference on Linguistics): 163–182.
1990(a) Epistemic stance and grammatical form in English conditional
sentences. CLS 26: 137–162.
1990(b) The contribution of linguistics to language understanding. In Pro-
ceedings of the 1st Symposium on Cognition, Language, and Cul-
ture, Aura Bocaz (ed.), 109–128. Universidad de Chile.
Fillmore, Charles J., and Paul Kay

1999 Grammatical constructions and linguistic generalizations: the
what's X doing Y? construction. Language 75 (1): 1–33.
Fillmore, Charles J., Paul Kay, and Mary C. O'Connor
1988 Regularity and idiomaticity in grammatical constructions: the case
of let alone. Language 63 (3): 501–538.
von Fintel, Kai
1991 Exceptive conditionals: the meaning of unless. NELS 22: 135–
148.
1994 Restrictions on quantifier domains. Ph.D. diss., University of
Massachusetts at Amherst.
Fisiak, Jacek, Maria Lipińska-Grzegorek, and Tadeusz Zabrocki
1978 An introductory English-Polish contrastive grammar. Warszawa:
Panstwowe Wydawnictwo naukowe.
Fleischman, Suzanne
1989 Temporal distance: a basic linguistic metaphor. Studies in Lan-
guage 13 (1): 1–50.
1995 Imperfective and irrealis. In Modality in grammar and discourse,
Joanna Bybee, and Suzanne Fleischman (eds.), 519–551. Amster-
dam: John Benjamins.
Giannakidou, Anastasia, and Frans Zwarts
(forthc.) Temporal, aspectual structure and (non)veridicality. In Tense and
mood selection, Alessandra Giorgi, James Higginbotham, and
Fabio Pianesi (eds.). Oxford: Oxford University Press.
Gojmerac, Mirko
1977 Pokušaj klasifikacije pogodbenih rečenica [On classifying condi-
tional sentences]. Strani jezici VI (3–4): 167–175.
Goldberg, Adele
1995 Constructions: A Construction Grammar approach to argument
structure. Chicago: University of Chicago Press.
Grochowski, Maciej, Stanisław Karolak, and Zuzanna Topolińska
1984 Gramatyka wspólczesnego języka polskiego: Składnia. [A
grammar of contemporary Polish – Syntax] Warszawa:
Państwowe Wydawnictwo Naukowe.
Haiman, John
1978 Conditionals are topics. Language 54: 512–540.
1980 The iconicity of grammar: isomorphism and motivation. Lan-
guage 56: 515–540.
1983 Paratactic if-clauses. Journal of Pragmatics 7: 263–281.
Haiman, John (ed.)

1985 Iconicity in syntax. Amsterdam: John Benjamins.
1986 Constraints on the form and meaning of the protasis. In On Con-
ditionals, Elizabeth Traugott, Charles Ferguson, Judy S. Reilly,
Alice ter Meulen (eds.), 215–227. Cambridge University Press.
Hopper, Paul J. (ed.)
1982 Tense-aspect: Between semantics and pragmatics. Amsterdam:
John Benjamins.
Iatridou, Sabine
1994 On the contribution of conditional “then”. Natural Language Se-
mantics 2: 171–199.
2000 The grammatical ingredients of counterfactuality. Linguistic In-
quiry 31 (2): 231–270.
Ivić, Milka
1970 O srpskohrvatskim pogodbenim rečenicama [On conditional sen-
tences in Serbocoratian]. Lingvistički ogledi: 145–153.
Jodłowski, Stanisław
1976 Podstawy polskiej składni [Foundations of Polish syntax].
Warszawa: Państwowe Wydawnictwo Naukowe.
Kamp, Hans, and Uwe Reyle
1993 From discourse to logic. Dordrecht: Kluwer.
Katičić, Radoslav
1984–85 Vrste pogodbenih rečenica u standardnom jeziku srpskom ili
hrvatskome [Types of conditional sentences in standard Serbian
and Croatian]. Zbornik za filologiju i lingvistiku XXVII–XXVIII:
339–342.
Langacker, Ronald
1987 Foundations of cognitive grammar: Theoretical prerequisites.
Vol. 1. Stanford CA: Stanford University Press.
1991 Foundations of cognitive grammar: Descriptive application. Vol
2. Stanford CA: Stanford University Press.
Michaelis, Laura
1993 Towards a grammar of aspect: The case of the English Present
Perfect construction. Ph.D. diss., Department of Linguistics, Uni-
versity of California, Berkeley.
1994a The ambiguity of the English Present Perfect. Journal of Linguis-
tics 30: 111–157.
Mrazović, Pavica, and Zora Vukadinović

1990 Gramatika srpskohrvatskog jezika za strance [Serbian grammar
for foreign learners]. Novi Sad.
Pisarkowa, Krystyna
1984 Historia składni języka polskiego [History of Polish syntax].
Wrocław: Ossolineum.
Schwenter, Scott
1994 “Hot news” and the grammaticalization of perfects. Linguistics
32: 995–1028.
1999 Pragmatics of conditional marking: implicature, scalarity and
exclusivity. (Outstanding Dissertations in Linguistics.) New York:
Garland.
Stevanović, Mihailo
1967 Funkcija i značenje glagolskih vremena [Function and meaning
of verb tenses]. Beograd: Srpska akademija nauka i umetnosti.
Sweetser, Eve
1990 From etymology to pragmatics. Cambridge: Cambridge Univer-
sity Press.
1996a Reasoning, mappings, and meta-metaphorical conditionals. In
Grammatical constructions: their form and meaning, Masayoshi
Shibatani and Sandra Thompson (eds.), 221–333. Oxford: Oxford
University Press.
1996b Mental spaces and the grammar of conditional constructions. In
Spaces, worlds, and grammar, Gilles Fauconnier and Eve Sweet-
ser (eds.), 318–333. Chicago: Chicago University Press.
1999 Compositionality and blending: semantic composition in a cogni-
tively realistic framework. In Cognitive Linguistics: Foundations,
scope and methodology, Theo Janssen and Gisela Redeker (eds.),
129–162. Berlin: Mouton de Gruyter.
1997 Conceptualization: conditionals as an instance of figure/ground
alignment. In On conditionals again, Angeliki Athanasiadou and
René Dirven (eds.), 273–289. Amsterdam: John Benjamins.
Traugott, Elizabeth C.
1989 On the rise of epistemic meanings in English: an example of sub-
jectification in semantic change. Language 57: 33–65.
1997 Unless and but conditionals: a historical perspective. In On Con-
ditionals again, Angeliki Athanasiadou and René Dirven (eds.),
169–190. Amsterdam/Philadelphia: John Benjamins.
Traugott. Elizabeth C., and Richard B. Dasher

2002 Regularity in semantic change. Cambridge: Cambridge Univer-
sity Press.
Traugott, Elizabeth C., Alice ter Meulen, Judy Snitzer Reilly, and Charles A.
Ferguson (eds.)
1986 On conditionals. Cambridge: Cambridge University Press.
Trnavac, Radoslava
2006 Aspect and subjectivity in modal constructions. Ph.D. diss., Uni-
versity of Leiden.
Verhagen, Arie
2005 Constructions of intersubjectivity: Discourse, syntax, and cogni-
tion. Oxford: Oxford University Press.
(forthc.) Construal and perspectivisation. In Handbook of cognitive lin-
guistics, Dirk Geeraerts and Hubert Cuyckens (eds.). Oxford:
Oxford University Press.
Vojvodić, Dojčil
1989 Modalna upotreba ruskog i hrvatskosrpskog perfektivnog
prezenta u uvjetnim rečenicama [Modal use of the Russian and
Serbocroatian perfective present in conditional sentences]. Živi
jezici XXXI (1–4): 24–31.
ȱȱ
ȱȱ¢ǱȱȬ¢ȱ
ȱȱ
Degrees of event integration. A binding scale for
[VFIN VINF] structures in Russian
Dagmar Divjak
Abstract*
In this paper I merge insights from cognitive and functional approaches to

complementation to present a comprehensive model, a binding scale, for the
293 verbs that combine with an infinitive in Russian. Assuming a strong cor-
relation between syntactic and semantic structure, and playing on the human
capacity to impose alternate structurings on a conceived phenomenon, I use
three constructions to reveal how the events that fit the [VFIN VINF] pattern
may be conceptualized in Russian. These three constructions highlight dif-
ferent facets of the event as encoded in the argument structure and event-
temporal properties of the verbs that appear as finite verbs in the [VFIN VINF]
structure. The results provide evidence for a cline of eight different degrees
of integration between the events expressed by means of a [VFIN VINF]. These
eight marks on the scale visualize the degree of independence both verbs,
and the events they render, have with respect to each other: at the one ex-
treme there are verbs that exist independently of the second verb, at the
other end verbs that are integrated with the second verb. In Russian, the lat-
ter verbs express modality, intention, attempt, result and phase. Comparable
findings have been proposed for English (Givón 2001: 54–58) as well as for
non-Indo-European language systems (Dixon 1996). The verbs’ argument
and temporal event structure and the proposed cognitive-semantic interpre-
tations for each of those structures result in a cognitively plausible verb
classification that sheds light on what might have motivated the grammatical
encoding of multiple and complex events in Russian.
Keywords: alternate construal, argument structure, binding scale, comple-

mentation, temporal event structure, reification, Russian
222 Dagmar Divjak
1. Introduction
Broadly defined, verbs express actions, processes and states, often sub-
sumed under the cover term “events” (Pustejovsky 1991) or “situations”
(Smith 1997).1 When two events or situations need to be expressed, typi-
cally, two clauses are used, be it two independent clauses, two coordinated
clauses or a main and subordinated clause (Talmy 2000 ch. 6; Cristofaro
2003 ch 5–7). Yet, the Russian language includes 293 verbs,2 such as de-
cide or urge that combine with another event or situation in the form of an
infinitive, yielding She decided to study Russian or He urged her to par-
ticipate in that discussion. These verbs, for which the [VFIN VINF] structure
is a point of intersection, are the subject of this paper.
On a motivational iconic view the [VFIN VINF] pattern signals a high de-
gree of interrelatedness of the events involved. I will investigate whether
two events that are captured in a [VFIN VINF] construction are indeed interre-
lated more closely than events that are divided over two full fledged
clauses. The degree of clause integration will be taken as indicative for the
degree of event integration (cf. Givón 2001 ch. 12); clause integration will
be measured by analyzing properties from the argument- and temporal
event structure of the verbs that combine with an infinitive (section 3).
Each of these properties will be interpreted in terms of its cognitive and/or
functional import. The result is a “binding scale” (cf. Givón 2001: 41) from
looser to tighter integration, that comprises eight degrees (section 4). This
scale contributes to our understanding of why certain categories of verbs
might be more prone to grammaticalization than others (section 5).
2. Theory and methodology
2.1. Theoretical assumptions
In cognitively inspired approaches to language, grammar and lexicon are

seen as forming a continuum and encoding meaning on different levels that
are progressively characterized by a higher degree of specificity (Goldberg
1995: 32). The meaning conveyed by both grammatical and lexical struc-
tures is considered to be conceptual, i.e. “it represents a way of conceptual-
izing experience in the process of encoding it and expressing it in lan-
guage” (Croft 1999: 77). In addition, human beings are said to have the
capacity to construe a situation in alternate ways as “the lexical and gram-
matical conventions of a language provide an array of alternative expres-
A binding scale for [VFIN VINF] structures in Russian 223
sions for coding a conceptualization” (Langacker 1987b: 110–111). If both

constructions and lexemes have (conceptual) meaning, it seems reasonable
to assume that these meanings must interact when they are put in contact,
and hence the meaning of the lexeme has to be compatible with the mean-
ing assigned by the construction to the slot the lexeme occupies (Goldberg
1995: 24, 50). A distinction can then be made between verbs that are re-
stricted to a specific construction and hence allow only one view on a
scene, and verbs that can alternate between different constructions and thus
allow different views (Langacker 1987b: 107, 111, 138). Differences in
networks of constructions in which certain (groups of) verbs occur can thus
be expected to reveal systematic semantic and related conceptual differ-
ences between those (groups of) verbs that do share constructional charac-
teristics and those (groups of) verbs that do not display any of these proper-
ties (Levin 1993; Atkins and Levin 1995: 96).
2.2. Data collection
Data on the 293 verbs’ argument and temporal event structure was col-
lected by means of elicitation tests with 15 native speakers of Russian.3
The elicitation experiment with these native speakers of Russian was set up
as a small number design. Five native speakers judged the constructional
possibilities of the 293 verbs on a three-point scale.4 To guard against lexi-
cal effects, the tests were carried using pro-forms (cf. Smessaert et al.
2005),5 which ensures that the mutual effect of lexical items in a construc-
tion is minimized as much as possible; as a result, the acceptability or un-
acceptability of a particular construction is very unlikely to be influenced
by the lexical compatibility or incompatibility of the words used. More-
over, to check for repetition effects in judgments of grammaticality 10
control judgments were collected for every verb in every construction type
from an ever varying pool of native speakers.6 Finally, the results obtained
were systematically compared against information contained in dictionar-
ies, against utterances found in the Amsterdam Corpus (Barentsen, s.d.) as
well as on the Internet.
Systematization of the results was aided by the surprising consistency
of the elicitation results. Radically opposed judgments did not occur and a
verb was considered to be acceptable in a construction if 10 out 15 judg-
ments were positive and the construction was attested in corpora and/or on
the Internet. If, on the contrary, 10 out 15 judgments were negative and the
224 Dagmar Divjak
construction was not attested in corpora and/or on the Internet, it was con-
sidered infelicitous. Less clear cut cases were assigned a “?”.
2.3. Parameters
There are three important parameters along which verbs that combine with
an infinitive vary. These parameters relate to the function verbs typically
fulfill: verbs express events or situations that have participants and take
place at a certain moment in time. The first parameter zooms in on the fact
that the main participants of events or situations are encoded in the verb’s
argument structure that is the basis of the simplex sentence (section 3.1).
The second parameter checks whether the two verbs expressing these
events can be linked alternatively using a complex sentence consisting of a
main clause and a subordinate clause (section 3.2). The third parameter
investigates how the events interact with time, as the events each of these
verbs expresses take place at a specific moment in time and have a specific
temporal contour, i.e. an imperfective and/or perfective aspect (section
3.3). These three syntactic parameters have cognitive-semantic dimensions
that can be interpreted as encoding the degree of integration between the
finite verb event and the infinitive event.
In the following section I will sketch how the three parameters differen-
tiate between the 293 verbs studied.
3. Puzzling constructional differences
The first diagnostic, henceforth the thing-test, measures the direction of

integration between two verbs used in what looks like a simplex sentence.
This parameter is concerned with the cognitive status of the event ex-
pressed by the infinitive, seen from the point of view of the finite verb
event (cf. Cristofaro 2003: 256). Within the confines of a simplex sentence,
the majority of (finite) verbs link up with nouns [VFIN N], a minority with
an infinitive [VFIN VINF]. Nouns and infinitives differ as to the type of con-
ceived scenes they render as well as to how the scenes they designate are
construed (Langacker 1987b: ch. 5, 6, 7). Bonding with both kinds of enti-
ties or only with one of them then reveals the conceptualization-type typi-
cal of a particular verb.
3.1. On Things and (Atemporal) Relations: something versus do

something
In Cognitive Grammar, verbs represent relational predications whereas

nouns are non-relational predications (Langacker 1987b: 217, 249 and ch.
5, 6, 7). A relational predication like a verb profiles interconnections and is
in this respect “conceptually dependent”, i.e. “one cannot conceptualize
interconnections without also conceptualizing the entities that they inter-
connect” (Langacker 1987b: 215). A non-relational predication like a noun
profiles sets of interconnected entities and focuses on the (resulting) ab-
stract entity. Apart from instantiating diverging kinds of predication, nouns
and verbs differ in terms of the type of entities they designate and the sort
of scanning required (for a critique, see Broccias and Hollmann 2007).
Nouns are symbolic structures whose semantic poles profile things, i.e.
scenes that are conceived as being unrelated to time and are scanned sum-
marily (Langacker 1987b: 491, 493, 1991: 22 ff and 148 ff). As a conse-
quence, the conceived scenes are available to the conceptualizer at a single
point in processing time. Verbs, on the other hand, profile processes or
series of component states distributed through a continuous span of con-
ceived time and are scanned sequentially, i.e. the conceived scene evoked
by a verb is accessed at consecutive points in processing time (Langacker
1987b: 143–144, 491). Infinitives, now, are intermediary between nouns
and verbs as they profile atemporal relations. Langacker (1987a: 76,
1987b: 220) claims that the effect of the infinitival morpheme is to suspend
the sequential scanning of the verb stem; this imposes an atemporal profile
on the processual profile provided by the stem.
The verb planirovat’ ‘plan’ from examples (1) and (2) (taken from
Apresjan and Pall 1982: 58, volume 2) combines both with nouns, which
designate non-relational things and with infinitives, which have an atempo-
ral relational profile. On the Cognitive Grammar approach, a difference in
construction is always accompanied by a difference in meaning (Langacker
1987b: 294): different constructions represent different ways of construing
an event. Yet, combining with nouns or verbs does not affect the meaning
of planirovat’ ‘plan’ itself; this can be gleaned from dictionaries (Apresjan
and Pall 1982: 58) that list the combination with both a noun and a verb as
instantiating one and the same sense of the verb planirovat’ ‘plan’.
(1) Он планирует поехать в Москву.

he plan.IP.IND.3SG travel.PF.INF to.PREP Moscow.ACC.F.SG
‘He is planning to travel to Moscow.’7
226 Dagmar Divjak
(2) Он планирует поездку в Москву.

he plan.IP.IND.3SG trip.ACC.F.SG to.PREP Moscow.ACC.F.SG
‘He is planning a trip to Moscow.’
Support for considering both (1) and (2) instantiations of one and the
same ‘plan’ sense can be found at the more abstract level. The something
versus do something test can be seen as an operationalization of the distinc-
tion between nominal things and atemporal verbal relations: infinitives can
be referred to as doing something, whereas, nouns can typically be cap-
tured under the cover-term something as (1’) and (2’) show. Example (2’)
then reveals that both a noun designating a thing and an infinitive profiling
an atemporally construed process are possible answers to the question
‘What is he planning?’.
(1’) Что он планирует сделать?

what he.NOM plan.IP.IND.3SG do.PF.INF
‘What is he is planning to do?’
Поехать в Москву.
travel.PF.INF to.PREP Moscow.ACC.F.SG
‘To travel to Moscow.’
(2’) Что он планирует?

what he.NOM plan.IP.IND.3SG
‘What is he planning?’
Поездку в Москву.
trip.ACC.F.SG to.PREP Moscow.ACC.F.SG
‘A trip to Moscow.’
Поехать в Москву.
travel.PF.INF to.PREP Moscow.ACC.F.SG
‘To travel to Moscow.’
This phenomenon of referring to an atemporal verbal relation as if it

were a thing8 occurs naturally as example (1’’) illustrates: with planirovat’
‘plan’, the pro-noun ėto ‘this’, i.e. a largely delexicalized item that stands
for and abstracts over all nouns, including poezdka ‘trip’ in (2), can be
used to refer to the verb zaderživat’ ‘arrest’ (cf. Smessaert et al. 2005 and
references therein).
(1’’) (…) и нам пришлось задерживать

and we.DAT.PL happen.PF.IND.PAST.N. 3SG arrest.IP.INF
вас раньше, чем мы это планировали.
you earlier, than we.NOM that.ACC.N.SG plan.IP.IND. PAST.3PL
[A. Mаринина. Стилист.]
‘(…) and we had to arrest you earlier, than we were planning that.’
The situation is quite different with verbs that are often named as near-
synonyms to planirovat’ ‘plan’, i.e. dumat’ ‘intend, think (of)’, namere-
vat’sja ‘intend, mean’ and sobirat’sja ‘intend, be about’ (for a full account
see Divjak 2006). Namerevat’sja ‘intend, mean’ lacks the possibility of
combining with non-verbal entities altogether, thus being restricted to
combinations with an infinitive. For dumat’ ‘intend, think (of)’ and sobi-
rat’sja ‘intend, be about’ a difference in construction, i.e. a combination
with an infinitive or with a noun, goes together with a sharp difference in
the meaning of dumat’ and sobirat’sja themselves. Take the example of
dumat’ (taken from Apresjan and Pall 1982: 389–390, vol. 1). With a
preposition and a nominal object dumat’ roughly means ‘think’, as is ex-
emplified in (3); in this sense, dumat’ has a delimitative perfective coun-
terpart podumat’. The other lemma, illustrated in (4), is restricted to com-
binations with an infinitive, and can be translated ‘intend’. In this sense,
dumat’ does not have a perfective counterpart.
(3) Они (по)думали о будущем.

they think.IP (P).PAST.3PL about.PREP future.PREP.N.SG
‘They were thinking about the future.’
(4) Они думали остаться дома.

they think.IP.PAST.3PL stay.PF.INF home.ADV
‘They intended to stay at home.’
In contrast to planirovat’ ‘plan’ in (2), referring to the infinitive by

means of a pro-noun that abstracts over the noun buduščee ‘future’ in (3)
causes a change in the meaning of dumat’; in (3’) dumat’ translates as
‘think’, in (4’) as ‘intend’.
?
(3’) (…) и нам пришлось задерживать вас раньше, чем мы об
этом думали.
‘(…) and we had to arrest you earlier, than we were think-
ing.IP.IND.PAST 3PL about.PREP that.PREP.N.SG.’
228 Dagmar Divjak
(4’) (…) и нам пришлось задерживать вас раньше, чем мы думали

это сделать.
‘(…) and we had to arrest you earlier, than we had in-
tended.IP.IND.PAST.3PL to do.PF.INF that.ACC.N.SG.’
Obviously, the meaning of the [VFIN VINF] construction interacts in non-

trivial ways with the meaning of the lexeme that occupies its finite verb
slot; some concepts such as think seem to be more prone to semantic shifts
than others, e.g. plan when no second event is present.
The proposed something versus do something test can be used as a lan-
guage-based diagnostic for the “cognitive status” of the infinitive (Cristo-
faro 2003: 256) seen from the point of view of the finite verb. As men-
tioned above, things and relations differ as to how the entities they
designate are conceptualized; bonding with both things and relations or
only with one of them reveals the conceptualization-type typical of the
(finite) verbs under study. Although all 293 verbs combine with infinitives,
different pro-structures abstract over the infinitive, depending on which
verb is used as the finite verb. The pro-structures reveal how the infinitives
relate to the finite verbs. As the examples (1) through (4) in this section
show, there are at least two possibilities:9 1) the finite verb combines both
with a noun and with a verb and the verb occupies the same argument
structure slot as the noun and 2) the finite verb combines both with a noun
and with a verb but the verb does not occupy the same slot as the noun.
The verb planirovat’ ‘plan’ from example (1) expresses a process, i.e. is
a relational entity, and combines with infinitives, i.e. entities that just like
processes have an own relational profile, albeit an atemporal relational
profile. Yet, do something subsumes under the pro-noun something: at a
more abstract, non-lexicalized level it is reified, i.e. conceptualized as a
thing. Pronominalization can be seen as an extreme form of reification: a
relation is conceptualized as a pro-noun that overarches all things (Quine
1964: 13). If do something subsumes under something for a particular
(lemma of a given) verb, do something is basically conceptualized as a
thing, despite its relational appearance. In other words, the thing-test shows
that verbs like planirovat’ ‘plan’ do not need another relational profile: the
infinitive can be subsumed under a pro-noun. Thus, conceptually, plani-
rovat’ ‘plan’ treats the infinitive as any other non-relational entity it com-
bines with.10 One could say that a verb like planirovat’ ‘plan’ evokes con-
ceptualization of the conceived scene expressed by the infinitive like any
non-relational thing in that position, more precisely, that of direct object.
At a more abstract level, the infinitival relation is presented as a thing, i.e.
an entity that is scanned as a unitary whole and is made conceptually sub-

ordinate to the process expressed by planirovat’ ‘plan’. A selection of
verbs that behave like planirovat’ ‘plan’ and score positively on the thing-
test can be found in Table 2 in the Appendix; these verbs are marked 1
through 4 in the first column.
The situation is quite different with a finite verb like dumat’
‘think/intend’ that exemplifies the second scenario. The infinitive that fol-
lows this verb cannot be subsumed under a pro-noun that is within the ar-
gument structure of the finite verb. The absence of the possibility to sub-
sume the infinitive under a pro-noun that fills up a slot in the finite verb’s
argument structure indicates that, with certain verbs, the infinitival rela-
tional profile cannot be backgrounded or made conceptually subordinate.11
The finite verb necessarily evokes the idea of another verbal relation, albeit
an atemporal relation. Some of the verbs that score negatively on the thing-
test are listed in Table 2 and are marked 5 through 8 in the first column.
Of all Russian verbs, only 293 verbs combine with an infinitive and of
those 293 verbs only one third does not allow reifying and conceptually
subordinating the relation expressed by the infinitive. In other words, a
situation in which the infinitive cannot be viewed as a thing that plays a
role in the argument structure of the finite verb is rather unusual in the
Russian verbal system. In the majority of constructions, it is sufficient to
have one entity that profiles interconnections; that verb typically is (the)
finite (verb). With a minority of verbs two such entities seem to be re-
quired, signaling that the finite verb behaves differently, and hence has a
slightly different function than a finite verb typically has. It seems there-
fore likely to assume that the relation between the events expressed by the
finite and infinite verbs is changed and the anchor point of the construction
is removed from the finite verb12 (for discussion see section 6).
In the next sections I will focus on constructions that make explicit two
more properties of the finite verb that remain implicit in [VFIN VINF] struc-
tures. More specifically, I will explore how “close” the second verb needs
to be to the finite verb, “spatially” and temporally; this will shed light on
the strength and independence of the (finite) verb and the concept it ex-
presses. In section 3.2, I will concentrate on “spatial” distance. In order to
do this, I will check which verbs that combine with an infinitive are re-
stricted to the [VFIN VINF] pattern and which finite verbs can appear in that-
constructions as well. Langacker (1991: 440–442) describes complementa-
tion in terms of conceptual subordination that implies conceptual distanc-
ing, summary scanning and possibly reification, the latter two being basic
characteristics of things.
230 Dagmar Divjak
3.2. Clausal remnants?
Dogovorit´sja ‘agree, come to an understanding about’ can introduce that-

complement clauses13 and can use these complement constructions to ex-
press the infinitive content alternatively: example (5) can be (partially)
paraphrased using the pattern of (5’). Both examples are presented in the
form they were taken from the RNC.
(5) Во вторник после долгих переговоров представители "Радона"

и столичные чиновники договорились вернуться к вопросу о
дезактивации леса. [RNC // "Известия", 2002.12.18]
‘On Tuesday, (...), the representatives and the functionaries
agreed.PF.IND.PAST.3PL to come back.PF.INF to the question of decon-
taminating the forest.’
(5’) Тогда правительство нам отказало, сославшись на нехватку

денег, но мы договорились, что вернемся к этой теме год
спустя. [RNC // "Известия", 2002.04.15]
‘(...) but we agreed.PF.IND.PAST.1PL, that.CONJ we would come
back.PF.IND.NON-PAST.1PL to that topic in a year’s time.’
Unlike dogovorit’sja ‘agree’, verbs such as izbegat’ ‘avoid’ or uspet’

‘manage’ do not occur with a that-complement clause at all.
(6) Вчера доллар почти сравнялся с евро и иностранные

инвесторы избегают вкладывать в американские акции в
таких обстоятельствах. [RNC // "Известия", 2002.07.16]
‘Yesterday, the dollar was nearly equivalent to the euro and under
such circumstances foreign investors avoid.IP.IND.PRES.3PL invest-
ing.IP.INF in American stocks.’
(6’) *Вчера доллар почти сравнялся с евро и иностранные

инвесторы избегают, чтобы вкладывать в американские
акции в таких обстоятельствах.
‘Yesterday, the dollar was nearly equivalent to the euro and for-
eign investors avoid.IP.IND.PRES.3PL that.CONJ to invest.IP.INF in
American stocks under such circumstances.’
(7) Мы успели отработать для Москвы правила

градостроительного регулирования. [RNC // "Вечерняя
Москва", 2002.04.11]
‘We managed.PF.IND.PAST.1PL to work.PF INF out urban construction
regulations for Moscow.’
(7’) *Мы успели, чтобы отработать для Москвы правила

градостроительного регулирования.
‘We managed.PF.IND.PAST.1PL that.CONJ to work.PF.INF out urban con-
struction regulations for Moscow.’
Complementation can be described in terms of conceptual subordina-

tion and dependence.14 “[T]he process a complement clause describes un-
dergoes a kind of conceptual subordination: rather than being viewed in its
own terms as an independent object of thought, it is primarily considered
for the role it plays within the superordinate relationship expressed by the
main clause” (Langacker 1991: 440–442). This proposal can be transposed
to Russian; indeed, in the majority of cases that-complement clauses fill up
one of the finite verb’s argument structure slots (compare (8)); if the finite
verb opens up a slot with a preposition, that preposition will be visible in
the main clause as a space holder, as is shown in examples (9) and (9’).
(8) На самом деле пока вопрос не закрыт – это была первая

встреча, и мы договорились о том, что вернемся к обсуждению
данного вопроса. [www.au92.ru/msg/20051010_ycu1vok.html]
‘(...) and we agreed.PF.IND.PAST.3PL on.PREP this.PREP.N.SG, that.CONJ we
will come back.PF.IND.NON-PAST.1PL to the question of decontaminating
the forest.’
(9) Однако Ринцетти все же стремился показать выходы

новаторской партитуры Пуччини к открытиям музыкального
будущего, в частности, в предвосхищении импрессинистических
звучностей. [RNC // "Российская музыкальная газета",
2003.06.11]
‘(...) Nevertheless Rincetti strove.IP.IND.PAST.3SG to show.PF.INF the
effects of Puccini’s innovative score […]’
(9’) Автор стремился не к тому, чтобы предложить новую

интерпретацию этих произведений (амбивалентность Гоголя –
давно установленный факт), а к тому, чтобы показать
232 Dagmar Divjak
конкретно, в чем эта двуединая природа проявляется и как она

выражается художественно. [RNC // Феликс Раскольников
(1986–2000)]
‘The author did not strive.IP.IND.PAST.3SG to.PREP this.DAT.N.SG, to pre-
sent.PF.INF a new interpretation of these writings (…), but to.PREP
this.DAT.N.SG, that.CONJ to show.PF.INF concretely how that one-in-two
nature manifests itself (…)’
Yet, complement clause constructions are not only a (complementary)

diagnostic for the status of the finite verb and the infinitive; in addition,
that-complementation is an explicit measure for the degree of integration
between the two verbs in the construction. “[V]iewing the subordinate
clause as a main-clause participant implies a conceptual distancing
whereby this process is construed holistically and manipulated as a unitary
entity. It therefore encourages summary scanning of the component states
if not their reification as an abstract region” (Langacker 1991: 440). In
other words, construing the second verb’s content as a full-fledged com-
plement clause equals imposing a nominal construal on the second verb
and detaching that structure conceptually from the finite verb. The con-
struction with that-complementizer can be seen as iconically signaling the
distance between the two verbs and the events they express. “[D]istance
and objectivity, specifically in regard to the subject’s construal of the sub-
ordinate process” are central to the semantics of that-complement construc-
tions in English (Langacker 1991: 448). In Russian too, many of the verbs
that check positive for that-complementation allow a non-coreferential
subordinate clause and many other Russian verbs that take complement
clauses do not combine with infinitives at all. Contrary to the situation for
[VFIN VINF] sequences to which co-reference restrictions apply, no co-
reference restrictions hold between a main clause and a full-fledged com-
plement clause. Indeed, in the same ‘agree’-sense, dogovorit’sja could
introduce a that-complement clause with a non-coreferential subject; the
action that is being agreed upon is not necessarily carried out by the per-
sons who are agreeing on the action, as is illustrated in (10).
(10) При этом стороны договорились, что Молдавия будет

оплачивать при получении газа $60 за тысячу кубометров.
[RNC // "Известия", 2001.10.07]
‘In addition the parties agreed.PF.IND.PAST.3PL thatǯCONJ Moldova
would.IP.IND.PRES.3SG pay.IP.INF 60$ for 100 cubic meters of gas
(...)’
The parameter of objectivity is also present in Wierzbicka’s (1988:

132–141) analysis of that-complementation in English. Wierzbicka basi-
cally states that want-type verbs, i.e. verbs the semantic definition of which
includes reference to something that is wanted, combine with infinitives,
whereas know-type verbs or verbs in the semantic structure of which the
primitive “know” is present, take that-complement clauses. If a verb allows
both the infinitive and the complement clause option, it stresses its want-
ing-component when combined with an infinitive, but foregrounds its
knowing-component when followed by a that-complement clause. Al-
though it becomes clear from Wierzbicka’s contrastive analysis that Eng-
lish, French, Polish and Russian each divide the tasks differently between
[VFIN VINF] sequences and that-complement clauses,15 the prototypical, basic
distinction made, i.e. that a that-complement clause signals distance and
objectivity, seems to hold for all four languages.
Finally, a complement clause – like any other element that fills up an
argument structure slot – is scanned summarily, i.e. “all facets of a com-
plex structure are coactivated and simultaneously accessible” (Langacker
1987b: 493). Ultimately, a complement clause can be seen as a reified
structure. The resulting non-relational predication does not focus on the
entities and their interrelation but on the resulting abstract entity or thing; a
thing is “the greatest departure from the processual nature of a verb or
clause” (Langacker 1987b: 215, 1991: 423).
The verbs in (6) and (7), on the other hand, cannot be used in a con-
struction in which the second verb process is at the same time reified,
viewed objectively, subordinated to and distanced from the finite verb; the
events reported on can only be construed in one way, i.e. as a [VFIN VINF]
sequence. Verbs like izbegat’ ‘avoid’ are incompatible with (objective)
distance, whereas verbs such as uspet’ ‘manage, find the time’ reject both
reification and subordination as the thing-test had already revealed. What
does this blocking of reification, subordination, distancing and objectivity
signal? A complementary account of complementation that explicitly refers
to the degree of integration between two verbs has been proposed by
Givón. Givón (2001 ch. 12) accounts for the syntax and semantics of com-
plement clauses in terms of the cross-linguistically supported (semantic)
notion of binding. According to him, the degree of morphosyntactic inte-
gration between the finite and the infinite verb can be seen as iconically
coding the degree of semantic integration of two single events into a single
complex event structure. Givón’s (2001: 50) “referential cohesion and
event integration” predicts that “the more the two events coded in the main
234 Dagmar Divjak
and complement clauses share their referents, the more likely they are to be
semantically integrated as a single event; and the less likely the comple-
ment clause is to be coded as an independent finite clause”. Thus, finite
verbs that are restricted to combinations with infinitives in which morpho-
logical and syntactic information is to a large extent shared across the
verbs and in which strict co-reference rules apply, signal a higher degree of
dependence on a second verb than finite verbs that can combine with both
infinitive and full-fledged complement clause. Although the latter con-
structions also consist of two events, both events exist to a certain extent
independently of one another and the infinitive event can be made subordi-
nate to the finite verb event.
Apart from morphosyntactic integration, Givón (2001: 45) identifies co-
temporality as a necessary cognitive pre-condition for considering two
events a single (though complex) event. Cristofaro (2003: 120) takes the
comparable “degree to which the boundaries between these SoAs [States of
Affairs] are eroded or kept intact” as the basic component of semantic in-
tegration. The next section will look into how boundaries can be imposed
on Russian verbs that combine with an infinitive and how these verbs deal
with the co-temporality requirement.
3.3. On time
In this section I will explore the (im)possibility of modifying both verbs in

a [VFIN VINF] structure with conflicting time adverbials or adverbial expres-
sions of time. I will relate the findings to the aspectual behavior of the
verbs, aspect being the most natural way to impose boundaries on events in
Russian. In so doing I obtain a second measure for the degree of integration
between the finite verb and the infinitive, a measure that is moreover inde-
pendent of the verb’s argument structure and conceptual subordination of
one event to the other.
The verb postanovit’ ‘decide’, could be used in a construction that lo-
cates the finite verb and the infinitive in two different and not necessarily
tightly sequential moments in time. More concretely, in example (11) the
finite verb postanovit’ ‘decide’ could take a specification as to when ex-
actly the management of BMW undertook the act of setting the sales tar-
gets.
(11) На собрании акционеров 21ого декабря 2002ого года BMW

постановил продать в 2003 году 190 тыс. автомобилей.
[adapted from www.lenta.ru/auto/2000/10/16/bmw/_Printed.htm]
‘At the shareholders meeting on December 21st 2002 BMW de-
cided.PF.PAST.M.3SG to sell.PF.INF 190 000 cars in 2003.’
The verbs ljubit’ ‘love, like’ and osmelit´sja ‘dare’ demand overlap in
or tight sequentiality of time. This requirement is illustrated in examples
(12) and (13).
(12) *В молодости он любил разгадывать кроссворды когда ему

было 50 лет, и делал это виртуозно. [adapted from RNC //
"Известия", 2003.01.27]
‘When he was young he loved.IP.PAST.M.3SG to solve.PF.INF cross-
word puzzles when he was 50 years old and he excelled at it.’
(13) */??Вчера офицер осмелился завтра применить решительные

меры.
‘Yesterday the officer.NOM.M.SG dared.PF.PAST.M.3SG to take.PF.INF
decisive measures tomorrow.’
Unlike that-complement clauses that unite conceptual distancing with

conceptual subordination, temporal distancing does not imply conceptual
subordination. Inserting conflicting temporal specifications is a way to
measure the degree of distance or integration between the two verbs in
[VFIN VINF] structures, independent from their argument structure. The oc-
currence of temporal distance between two events merely entails their con-
ceptual distance. The two events take place at two different moments in
time; they are construed as distinct (though related) events (Wierzbicka
1975: 497–499; Lakoff and Johnson 1980: 131; Langacker 1991: 299 fn.
11).
Givón (2001: 44) treats co-temporality as one necessary cognitive pre-
condition for considering two events a single (though complex) event. His
already quoted “referential cohesion and event integration” could include
temporal and locational references: “the more the two events coded in the
main and complement clauses share their referents, the more likely they are
to be semantically integrated as a single event; and the less likely the com-
plement clause is to be coded as an independent finite clause” (see also
section 4.2).
236 Dagmar Divjak
What supports native speakers of Russian in their decision to accept or

reject a particular construction in which the finite verb and the infinitive
are modified with separate time adverbials? Inspection of the data reveals a
connection between finite verbs that do occur in [VFIN VINF] structures with
conflicting temporal specifications and the aspectual properties of those
finite verbs.16
Aspect, Janda (2004) argues, implies viewing events as physical enti-
ties. Aspect interacts with the TIME IS SPACE metaphor: events occupy time
in the way that matter occupies space, which yields the inference that PER-
FECTIVES ARE DISCRETE SOLID OBJECTS and IMPERFECTIVES ARE FLUID
SUBSTANCES. In a language where aspect is fully grammaticalized, this
notion is an organizing principle (Janda 2004): aspect in this grammatical
context prepackages all items in the verbal lexicon. Speakers of Russian
must choose between two grammaticalized aspectual forms to express an
event.
Generalizing the data it can be said that the metaphor PERFECTIVE IS A
SOLID OBJECT versus IMPERFECTIVE IS A FLUID SUBSTANCE provides em-
bodied support for the (im)possibility of modifying the finite verb and the
infinitive with separate time adverbials. If both the finite verb and the in-
finitive can be modified with separate time adverbials the finite verb has a
perfective counterpart or, in metaphorical terms, can be thought of as a
solid object that has boundaries.17 However, not all verbs that combine
with an infinitive and have a perfective form allow modifying the finite
verb and the infinitive with separate time adverbials (as with osmelit’sja
‘dare’ in example (13)). Having a perfective counterpart is a necessary but
not a sufficient condition. Finite verbs that cannot be pinned down to a
particular point in time express events that have affinity with fluid sub-
stances without clear boundaries; without having the perfective counterpart
to rely on, it is impossible to ascertain whether a particular event, ex-
pressed by the first verb, is bounded,18 i.e. has ended before the second
event, expressed by the infinitive, starts unfolding (compare ljubit’ ‘like’ in
example (12)).
In the preceding three subsections I have demonstrated how the behav-
ior of verbs that combine with an infinitive can be captured under at least
three different parameters involving the finite verb’s argument and event
structure: the possibility to refer to an event as if it were a thing (section
3.1), that-complementation (section 3.2) and temporal separability (section
3.3). So far I have suggested interpretations for each of the constructions
separately. It is the purpose of section 4 to elucidate the interaction of the
three parameters and its implications for the meaning of the verbs that oc-
cur as finite verbs in the [VFIN VINF] pattern.
4. Degrees of verb integration
From a logical point of view, there are eight possible ways to combine the
positive and negative results of the three tests, the thing-test, the that-
complementation test and the time test. These eight logical combination
options are illustrated in Table 1, an “idealized” model that – for ease of
presentation – imposes clear-cut boundaries on a phenomenon, graded in
nature (cf. Divjak 2007). In addition, Table 1 abstracts away from the for-
mal details distinguished for each parameter, e.g. the case marking on the
pro-nominal paradigm or the type of complementizer.19 The verbs that
populate each column are presented in Table 2 in the Appendix
4.1. Eight logical combination possibilities
On the first row in Table 1 the number of verbs that qualify for a certain
column is provided in square brackets. Rows two, three and four provide
the exact value of each parameter. The shorthand “thing” refers to the
something versus do something test presented in section 3.1. If a finite verb
opens up an argument structure slot in which both a noun and a verb can be
used, the value “+” is assigned. If infinitives cannot be referred to by
means of a pro-noun, the finite verb gets a “-” for the thing-test. The results
of testing for that-complementation, explained in section 3.2, are summa-
rized under “that”: if the infinitive can be distanced from the finite verb by
means of a that-complementizer, the finite verb gets “+that”. If the finite
verb – in a particular sense – is restricted to infinitives, the value “-that” is
assigned. Finally, “time” summarizes the results for the test with conflict-
ing temporal specifications from section 3.3. Combinations of finite verb
and infinitive that do allow two conflicting temporal modifiers in the con-
struction get “+time” whereas combinations of finite verb and infinitive
that reject two conflicting temporal modifiers in one construction are
marked “-time”.
The thing-test, having the largest coverage (70,6% of all verbs checks
positively for this test) is listed first. Besides, this parameter gives a good
view on the direction of the integration: it makes explicit whether the in-
finitive is drawn into the finite verb’s argument structure or whether the
238 Dagmar Divjak
finite verb is attracted to the infinitive. Temporal separability is the other

strong variable for testing the mutual independence of both verbs, yet it
occupies the last rank in the table (as only 38,2% of all verbs check posi-
tively for this test). That-complementation (for which 57,3% of all verbs
check positively) links the two rather independent parameters, reification
and temporal separability. Because complementation can – for this set of
verbs – by and large be considered a weaker version of both referential
inclusion and temporal separability, its values vary from column to col-
umn. In this way, Table 1 reflects both the degree of integration (horizon-
tally, on rows one and five) and the decrease in strength of correlation be-
tween the variables (vertically, on rows two through four), the higher
correlations holding between the thing-test and that-complementation, next
between that-complementation and temporal separability, and finally be-
tween the thing-test and temporal separability.
Table 1. Eight patterns for verbs that combine with čto/ėto (s)delat’ ‘do what/that’
1 [80] 2 [10] 3 [80] 4 [37] 5 [8] 6 [14] 7 [0] 8 [64]
+ thing + thing + thing + thing - thing - thing - thing - thing
+ that - that + that - that + that - that + that - that
+ time + time - time - time + time + time - time - time
P-verbs M-verbs
Although there are eight combinatory possibilities, 89% of all verbs that
combine with an infinitive find a place in columns one, three, four or eight.
These columns are the centers of gravity in this table, and I will discuss
them in turn: one and three each contain about 27% of all 293 verbs that
combine with an infinitive. Column eight collects 21.8% and column four
represents 12.6% of all verbs. On the basis of the cognitive-semantic inter-
pretations given for each parameter in the previous sections, these four
columns can be said to represent the four major degrees of verb integration
for [VFIN VINF] patterns. Column 1 verbs are strong and form [VFIN VINF] se-
quences in which two independent verbs are put in contact, whereas the
weak column 8 verbs form [VFIN VINF] patterns the two verbs of which are
characterized by the highest degree of integration. The other four columns
– two, five, six and seven – each represent less than 5% of all verbs that
combine with an infinitive.
4.2. Reflecting the degrees of verb integration
Column one in Table 1 consists of verbs that check positively on every

test; examples are planirovat’ ‘plan’ from (1) and (2), dogovorit’sja
‘agree’ in (5) or postanovit’ ‘decide’ in (11), and all verbs labeled 1 in the
first column of Table 2. When a [VFIN VINF] construction is filled up with a
finite verb of this type, the infinitive relation is conceptualized as a thing
and enters the argument structure of the finite verb. Verbs that belong to
column one and incorporate the infinitive to fill up one of their argument
structure slots thus behave like verbs that combine with nouns. In other
words, column one (finite) verbs function by themselves as construction
kernel; this type of (finite) verbs will be termed P-VERB.20 Moreover, the
independent events represented by both finite verb and infinitive can be
distanced from each other in a main and complement clause and in addi-
tion, the verbs in the [VFIN VINF] sequence can be modified with conflicting
temporal specifications. The first verb is not only strong enough to down-
grade the second verb and pull it into its argument structure; it does not
need the second verb to be contiguous, neither (spatially) within the con-
struction nor temporally within the event space. Events in [VFIN VINF] se-
quences, whose finite verb is one of the verbs that belong to this first col-
umn, are conceptually independent of each other: both verbs represent
independent events and together they form what I will call MULTIPLE
EVENT structures.
Column eight contains verbs such as dumat’ ‘intend’ from (4) or uspet’
‘manage’ from (7) and all verbs labeled 8 in the first column of Table 2.
They display the opposite behavior of P-verbs; [VFIN VINF] constructions
with finite verbs from column eight check negatively on every test. Impos-
sibility of referential inclusion of do something under something shows
that the infinitive relation cannot be downgraded and conceptualized as a
thing, a precondition for occupying a regular slot in the argument structure
of the finite verb. Neither can the infinitive event be distanced from the
finite verb event by means of that-complementation. At the same time, the
two verbs cannot be modified by conflicting temporal specifications and
thus do not express (temporally) independent events. [VFIN VINF] patterns
built on finite verbs from column eight display the highest degree of verb
integration. Column eight (finite) verbs do not function as full-fledged
240 Dagmar Divjak
construction kernels themselves; they are not strong enough to downgrade

the infinitival atemporal relation and incorporate it into their argument
structure. Instead, it can be said that the finite verbs themselves are down-
graded and form a complex construction kernel together with the infinitive.
These “downgraded” finite verbs will be labeled M-VERBS:21 they modify
the infinitive event and melt together into a single COMPLEX EVENT struc-
ture (Givón 2001: 43).
Between the two extremes, P-verbs and M-verbs, but still on the left-
hand side of Table 1, two other important columns, three and four, are
situated. Column three with verbs like stremit’sja ‘strive’ from example (9)
and osmelit’sja ‘dare’ from example (13) and all verbs labeled 3 in the first
column of Table 2, differs from the first and most independent type of P-
verbs in temporal event structure. Finite verbs from column three are not
conceptualized as having temporal independence from the infinitive they
combine with. Nevertheless, the infinitive event that enters the argument
structure of the finite verb as a regular object can be (objectively) dis-
tanced from the finite verb in a subordinated construction. This possibility
is significant for a degree of independence for both verbs, be it in a relation
in which the finite verb dominates the infinitive. Column four with verbs
like izbegat’ ‘avoid’ in example (6) and all verbs labeled 4 in the first col-
umn of Table 2, lacks not only temporal independence for both verbs, but
also the that-complementation option. In constructions with finite verbs
from column four, infinitival relations are restricted to occurring in a rei-
fied form as fillers of one of the finite verb’s argument structure slots.
Thus, for both column three and four verbs the infinitive fills up an argu-
ment structure slot. Therefore, the finite verbs can still be said to function
as construction kernels or P-verbs. Yet, the finite verbs differ from each
other and from column one P-verbs in the various degrees of integration
they engage in with their infinitives, more specifically, in the type of
closeness typical of them.
Apart from the four columns of Table 1 that together represent 89% of
finite verbs there are four columns that contain between 0% and 5% of all
verbs studied. Column two, the mirror image of the empty column seven,
displays an exceptional configuration of variable values: it combines the
presence of referential inclusion of do something under something and a
temporally separable finite verb with a lack of that-complementation alter-
natives. I hypothesize that the impulsiveness of many verbs that belong to
this column, e.g. vydumat’ ‘invent, make up’, pridumat’ ‘invent, think of’
(for more examples see verbs labeled 2 in Table 2) does not combine well
with an objective, contemplative construal. This category would be an in-
teresting starting point for an elaborate and precise study of the meaning of
that-complementation in Russian.
In columns five and six the finite verb does not have the strength to
pull the infinitive event into its argument structure.22 This indicates a
changed balance between the two events expressed, as is claimed for col-
umn eight. Both in column five and six the events expressed are temporally
independent, but the infinitive event with column five verbs can enter into
a subordinate relation with the finite verb, whereas it cannot do so with
column six verbs. Infinitive events combined with column six finite verbs
are temporally independent and equally important: the finite verb event
cannot dominate the infinitive event, but it does not need that event to be
temporally contiguous. Table 2 in the Appendix displays some of the verbs
belonging to these columns; they contain an interactive element of thinking
or speaking that can be confined to a particular moment in time, compare
here zastavit’ ‘force’ from category five and objazat’sja ‘bind oneself,
pledge, undertake’ from category six.
In sum, the four columns in Tables (1) and (2) that contain most ele-
ments check positively or negatively for either referential inclusion and
that-complementation or that-complementation and conflicting temporal
specifications or for all three parameters.
The aim of the last section is two-fold. On the one hand, I will summa-
rize the main conclusions that can be drawn from research into the con-
structional behavior of the 293 verbs that combine with an infinitive in
Russian; on the other hand I will present typological support for the de-
grees of integration between two verbs in the [VFIN VINF] pattern and the
semantic type of verbs that displays the highest integration.
5. Implications for a theory of grammaticalization
My findings highlight the distributional complexity of the [VFIN VINF]

structure. Although all [VFIN VINF] sequences arguably represent two events
as closely interrelated, the individual verbs that instantiate the sequence
nuance that view. Research on the argument and event structures character-
izing the lexemes that function as finite verb facilitates plotting the [VFIN
VINF] on a scale from looser to tighter integration between the two events
depicted, depending on the finite verbs used. M-verbs in Russian are at
present the type of verbs that exclusively combines with an infinitive, re-
quiring the highest degree of integration with that infinitive. In [VFIN VINF]
242 Dagmar Divjak
constructions with M-verbs, the infinitive can no longer be staged as the

finite verb’s (prepositional) object. M-verbs are being “downgraded” and
the infinitives that combine with these M-verbs are being “upgraded” from
subordinated thing to full-fledged process. The finite verb and the infini-
tive are integrated and form a complex event. These verbs decide together
on the contents of shared referents such as the subject, adverbs, and other
specifications (cf. Kemmer and Verhagen 1994: 117–119).
Constructions, too, seem to be categories with prototypical and periph-
eral members (Croft 2001: 51–53). Prototypicality in this context refers to
the function of both verbs in [VFIN VINF] sequences. The findings presented
in this paper suggests that the (interrelated) categories “main verb” and
“complement” might not be as monolithic as they are usually presented. It
is quite common to view complementation in general, i.e. not restricted to
the so-called infinitival complements, and even the space covered by coor-
dination and subordination in terms of a continuum or a cline (compare
Givón 2001 ch. 12 for complementation and Croft 2001 ch. 9 and Cristo-
faro 2003 for subordination). Like the phenomenon of clause combining
(Smessaert et al. 2005), verb combining seems to display a continuum “of
looser-to-tighter-integration” (Hopper and Traugott 1993: 177). Yet, in
Givón’s account (2001: 40–43) the semantic-cognitive dimensions of bind-
ing that are expressed in terms of the semantic properties of the main verb,
do not seem to affect the “syntactic status” of that verb: even for [VFIN VINF]
sequences in which the events are most integrated the finite verb remains
the “main” verb and the infinitive the “complement”. The parameter of
referential inclusion of do something under something visualizes, however,
that the prototypical relation of a main verb incorporating a complement
does not hold between a number of finite verbs and the infinitives they
combine with, e.g. (3’) and (4’).
The continuum of event integration can be laid bare by referring to the
interaction between three parameters, two parameters from argument struc-
ture and one parameter from temporal event structure. These parameters
facilitate locating each verb on the continuum from fully independent mul-
tiple event structures to highly integrated complex events. On the horizon-
tal axis in Table 1, eight different constellations have been distinguished
and defined in terms of the three parameters. The majority of verbs that
combine with an infinitive find a place in four of these eight categories.
The finite verbs in three of these four categories share the possibility of
accommodating the infinitive in one of their argument structure slots.
These verbs are P-verbs that, together with an infinitive, form MULTIPLE
EVENT structures. M-verbs, on the contrary, do not have the power to pull
the infinitive into one of their argument structure slots. The events ex-
pressed by this type of finite verb and the infinitive are linked up to the
extent that, together, they are presented as a COMPLEX EVENT. The overall
import of the finite verb is weakened whereas at the same time, the infini-
tive is foregrounded.
The distinction in the value of the semantic import made by each verb is
stressed by Dixon (1996: 176–179) who works with Primary A and B
verbs, and Secondary verbs. Primary verbs do not need to relate to any
other verb and are always expressed as verbal lexemes. Secondary verbs,
on the other hand, “encode meanings that relate to some other verb, and
which can be expressed as verbal lexemes or as affixes or verbal modifiers
or clausal particles. (…) Semantically, the infinitives are the core concepts
of the sentences with the [secondary] finite verbs providing semantic modi-
fication.” Dixon (1996: 187–188) enumerates as “secondary items” nega-
tion, modals (such as must and can) and aspectuals (like begin, finish, but
also try and attempt), as well as the so-called wanting- and making-type
verbs. The Russian counterparts can be found in category eight that con-
tains M-verbs.
There is a critical difference between Primary and Secondary verbs:
Primary verbs – which do not need to relate to any other verb – are always
expressed as verbal lexemes. Secondary verbs encode meanings that relate
to some other verb, and there are four main ways in which languages ex-
press secondary concepts. This can be done as Secondary verbs which have
essentially the same array of derivational and inflectional possibilities as
Primary verbs, e.g. take complement clauses. Other possibilities for ex-
pressing secondary concepts are offered by verbal affixes, modifiers to a
verb (including both adverbs and modal verbs) and non-inflecting particles
within a clause (Dixon 1996: 178).
Dixon (1996: 177) argues that the finite verbs are syntactically the main
verbs, with the infinitives functioning as verbs of embedded clauses, but
claims at the same time that the infinitives are the core concepts of the
sentences with the finite verbs providing semantic modification. According
to Dixon, “these sentences have the same surface syntax but different se-
mantic interpretations. (…) Here, as at other places in the grammar, we
find one syntactic mechanism being used for tasks which are semantically
rather different”. Yet, would a language deceive its speakers? All [VFIN
VINF] sequences may look similar, i.e. consist of a finite verb and an infini-
tive, yet depending on the finite verb, they take part in fundamentally dif-
ferent argument and temporal event structures, as was revealed by the three
parameters examined above (section 3). And it is the network of construc-
244 Dagmar Divjak
tions as a whole that defines the function and meaning of an element (van
den Eynde 1995: 116ff; Croft 2001: 27–28). In other words, verbs that
belong to different networks of constructions fulfill different functions and
signal differences in meaning, even though the individual constructions
they take part in may look similar.
The difference in degree of binding and its relation to semantics is
prominent in Givón’s (2001: ch. 12) analysis of complementation. Verbs
that display the strongest degree of binding are “modality verbs”. For
Givón modality verbs code “inception and termination, persistence, suc-
cess and failure, attempt, intent, obligation or ability vis-à-vis the comple-
ment state or event”. Although I object to using the term “complement” in
this context for the reasons just mentioned, the Russian category eight
verbs in Tables (1) and (2) that show the highest degree of integration with
the infinitive, express exactly Givón’s “modality” concepts, i.e. they are
“Modal” and “Intentional”, “Tentative”, “Resultative” and “Phasal” verbs.
A third source of support for my position on the weakened syntactic and
semantic import made by the finite verb can be found in Croft (2001: 216–
220 and 254–259). Croft refers to this “non-iconic mapping of arguments”
as a grammaticalization process called “clause collapsing”: a complex sen-
tence structure with a main verb and a complement verb is being reana-
lyzed as a single clause with a tense, aspect, and/or mood indicating form
(the former main verb) and a main verb (the former complement verb).”
The infinitive is chosen as main verb because it is the Primary Information
Bearing Unit (PIBU), the semantic head or “most contentful item that most
closely profiles the same kind of thing that the whole clause constituent
profiles”. In the process of grammaticalization head status gradually shifts.
“The construction’s profile will shift towards that of the PIBU, and the
profile equivalent that is not the PIBU may expand its extension – that is,
bear even less information than it did before. At this point, the formerly
dependent structure becomes the head. If the grammaticalization process
continues, the functional elements end up losing their status as autonomous
syntactic units, becoming affixed to the lexical head. (…) This diachronic
change is a gradual process (…) for some languages, one of the last steps
in this process is the reassignment of syntactic arguments to the former
complement verb.” It is this step we currently seem to be witnessing in
Russian.
Appendix
Table 2. Productively used, stylistically neutral verbs that combine with an infini-
tive in Russian
Cat. # Thing- That- Time- Verb Translation
test test test
1 yes yes yes велетьip&pf Order
1 yes yes yes вынуждатьip- Force
вынудитьpf
1 yes yes yes (приpf)готовитьсяip Get ready, prepare
1 yes yes yes (приpf)грозитьip Threaten
1 yes yes yes допускатьip- Allow, permit
допуститьpf
1 yes yes yes (поpf)желатьip 2 Desire, want, wish
for someone else
1 yes yes yes запрещатьip- Forbid, prohibit
запретитьpf
1 yes yes yes (поpf)клястьсяip Swear, vow
1 yes yes yes обещатьip&pf Promise
1 yes yes yes обязыватьip-обязатьpf Oblige
1 yes yes yes отказыватьсяip- Refuse, denounce
отказатьсяpf
1 yes yes yes подговариватьip- Put up, instigate
подговоритьpf
1 yes yes yes позволятьip- Allow, permit,
позволитьpf afford
1 yes yes yes поручатьip-поручитьpf Guarantee, vouch
for
1 yes yes yes постановлятьip- Decide, resolve
постановитьpf
1 yes yes yes предлагатьip- Offer
предложитьpf 1
1 yes yes yes предлагатьip- Suggest, propose,
предложитьpf 2 ask, order, tell,
invite
1 yes yes yes принуждатьip- Force

принудитьpf
1 yes yes yes (заpf)проектироватьip Project
1 yes yes yes (поpf)проситьip Ask, request
246 Dagmar Divjak
Catego- Thing- That- Time- Verb Translation

ry # test test test
1 yes yes yes разрешатьip- Allow, authorize
разрешитьpf
1 yes yes yes решатьip-решитьpf Decide
1 yes yes yes решатьсяip-решитьсяpf Decide
1 yes yes yes сговариватьсяip- Arrange
сговоритьсяpf
1 yes yes yes (поp)советоватьip Advise
1 yes yes yes соглашатьсяip- Consent, agree
согласитьсяpf
1 yes yes yes (поpf)требоватьip Ask for, demand,
request, require
1 yes yes yes уговариватьip- Persuade, induce,
уговоритьpf talk into
1 yes yes yes умолятьip-умолитьpf Move by entreaties
1 yes yes yes уполномочиватьip- Authorize
уполномочитьpf
1 yes yes yes упрашиватьip- Beg, entreat
упроситьpf
1 yes yes yes (поpf)хвастатьсяip Boast, brag
2 yes no yes наниматьсяip- Get a job
нанятьсяpf
3 yes yes no (поpf)боятьсяip 2 Fear
3 yes yes no забыватьip-забытьpf Forget
3 yes yes no замедлятьip- Be long in, delay
замедлитьpf
3 yes yes no медлитьip Be slow in
3 yes yes no мечтатьip Dream of
3 yes yes no (поpf)мешатьip 2 Prevent, hinder,
impede
3 yes yes no (поpf)надеятьсяip Hope for
3 yes yes no опаздыватьip- Be late
опоздатьpf
3 yes yes no опасатьсяip2 Be afraid of, beware
3 yes yes no отчаиватьсяip- Despair of
отчаятьсяpf
3 yes yes no покушатьсяip- Attempt

покуситьсяpf
3 yes yes no помогатьip-помочьpf Help, assist

ry # test test test
3 yes yes no привыкатьip- Get used to, get in
привыкнутьpf the habit of
3 yes yes no приниматьсяip- Start to
принятьсяpf
3 yes yes no рассчитыватьip Calculate on
3 yes yes no (не)(поpf)скупиться Hesitate
3 yes yes no соблазнятьip- Tempt, seduce
соблазнитьpf someone
3 yes yes no соблазнятьсяip- Tempt, seduce
соблазнитьсяpf
3 yes yes no сомневатьсяip Doubt, question
3 yes yes no стремитьсяip Strive
3 yes yes no (поpf)торопитьip Urge
3 yes yes no (поpf)торопитьсяip Hurry, rush
3 yes yes no уставатьip-устатьpf Get tired
3 yes yes no учитьip Teach
3 yes yes no учитьсяip Learn
4 yes no no боятьсяip 1 Be afraid
4 yes no no (поpf)брезговатьip- Disdain, hate, have
(поpf)брезгатьpf an aversion to
4 yes no no (поpf)гнушатьсяip Abhor, disdain
4 yes no no (поpf)желатьip 1 Desire, want, wish
for oneself
4 yes no no затруднятьсяip- Find difficult
затруднитьсяpf
4 yes no no избегатьip Avoid
4 yes no no любитьip Like, love
4 yes no no (поpf)мешатьip 1 Interfere with,
obstruct
4 yes no no мочьip 2 Can
4 yes no no ненавидетьip Hate
4 yes no no опасатьсяip1 Fear
4 yes no no остерегатьсяip- Beware of, be
остеречьсяpf careful
4 yes no no предоставлятьip- Leave, let
предоставитьpf
4 yes no no предпочитатьip- Prefer
препочестьpf
248 Dagmar Divjak

ry # test test test
4 yes no no рисковатьip-рискнутьpf Take the risk of
1
4 yes no no (поp)стеснятьсяip Hesitate
4 yes no no (поpf)стыдитьсяip Be ashamed
4 yes no no уметьip Know how to
5 no yes yes доверятьip-доверитьpf Entrust
5 no yes yes завещатьip&pf Leave, devise
5 no yes yes заставлятьip- Force
заставитьpf
6 no no yes вызыватьсяip- Volunteer, offer
вызватьсяpf
6 no no yes обязыватьсяip- Bind, pledge
обязатьсяpf oneself, undertake
6 no no yes (поpf)проситьсяip Ask, request
8 no no no бросатьip-броситьpf Give up
8 no no no изловчатьсяip- Contrive, manage
изловчитьсяpf
8 no no no кончатьip-кончитьpf End, finish,
complete
8 no no no ленитьсяip Be too lazy
8 no no no мочьip 1 Know how
8 no no no мочьip-смочьpf Be able
8 no no no намереватьсяip Intend
8 no no no начинатьip-начатьpf Begin, start
8 no no no переставатьip- Stop, cease
перестатьpf
8 no no no порыватьсяip Try
8 no no no предполагатьip Intend to
8 no no no прекращатьip- Stop, cease
прекратитьpf
8 no no no (поpf)пробоватьip Try
8 no no no продолжатьip Continue, proceed
8 no no no (поpf)пытатьсяip Try
8 no no no собиратьсяip [- Intend, be about
собратьсяpf]
8 no no no статьp Start
8 no no no (поpf)старатьсяip Try
8 no no no суметьpf Manage

ry # test test test
8 no no no успеватьip-успетьpf Find time, manage
8 no no no хотетьip Want, wish, desire
Notes
*
This research was carried out with the financial support of the Science Foun-
dation – Flanders (Belgium). I would like to thank Agata Kochańska for her
careful comments on passages relating to Cognitive Grammar theory and the
two reviewers for a variety of useful suggestions. The usual disclaimers apply.
1. Following Pustejovsky (1991) I will use “event” as a cover term for what is
expressed by one verb, be it in Smith’s terms (1997) a “non-dynamic state” or
a “dynamic event” such as an activity, an accomplishment, an achievement or
a semelfactive event.
2. Russian has 293 verbs that combine with an infinitive, disregarding imper-
sonal verbs and verbs of motion. I have culled all verbs for which combination
with an infinitive is attested in one general, explanatory dictionary (in its two
most recent editions) and four specific verb dictionaries, i.e. Ožegov and Šve-
dova (1995 and 1999), Daum and Schenk (1992), Benson and Benson (1995),
Denisov and Morkovkin (1978) and Apresjan and Pall (1982). Table 2 in the
appendix contains the approximately 100 productively used verbs that are sty-
listically neutral; if one stem combines with several prefixes (e.g. učit’ that
takes vy-, na-, ob-, pod-), only the base verb (i.e. učit’ ‘learn’ has been in-
cluded.
3. Although the absence of a specific pattern in a sufficiently large sample of the
language can provide support for the infelicity of that pattern (Stefanowitsch
2006), research into “alternations” is generally not corpus-based. Typically,
corpora are used to provide information about which constructions are attested
for a particular verb, whereas native speaker intuitions are used to decide
whether a particular construction alternates with another construction (com-
pare Levin 1993). As a consequence, the few available approaches to auto-
matic extraction of alternations are concerned with case studies conducted on
the basis of English language data (Schulte im Walde 2007).
4. The experiment was conducted over a period of three months in the form of a
weekly interview during which each native speaker was presented with ap-
proximately 25 verbs. Several measures were taken to minimize the negative
side-effects of this set-up. Among other things, native speakers were asked
both to judge ready-made sentences and to form sentences using particular
constructional devices; these sentences were on a later occasion presented to
the participant who had constructed them as well as to other participants.
250 Dagmar Divjak
5. In the elicitation experiment, infinitives were replaced by the pro-form

(s)delat’ ‘do’, and the results might therefore be restricted to combinations
with infinitives that relate to this action primitive.
6. In this case, the trigger questions were mixed with other, non-related questions
about aspects of Russian syntax and semantics.
7. The English translations are word-for-word translations of the Russian original
and glosses are provided for the phenomena that are in focus only; formal in-
formation irrelevant for the purposes of this paper is omitted and for the same
reason morphemic alignment of source and goal sentence has not been imple-
mented throughout. To improve readability, Cyrillic has been transliterated in
the running text, but has been maintained in the examples.
8. The phenomenon of referring to a process as if it were a thing occupies a cen-
tral position within the Pronominal Approach (PA) from which it was bor-
rowed (for a brief introduction and key references see Smessaert et al. 2005).
Within the PA this procedure is referred to as the “proportionality test”. This
test checks whether a lexical element can be replaced by a so-called pro-
nominal or pro-verbal paradigm that at the very least contains an assertive and
a non-assertive pro-form. Nouns are typically replaced by the pro-noun čto/ėto
‘what/that’, that stands “for any noun”, whereas verbs are replaced by the pro-
verb or pro-predicate čto/ėto (s)delat’ ‘do what/that’, that stands “for any
verb”. For the verbs considered in this paper the possibility of using an asser-
tive pro-noun ėto ‘that’ with or without (s)delat’ ‘do’ to abstract over a noun
and/or verb entails the possibility of using a non-assertive pro-noun čto ‘what’
with or without (s)delat’ ‘do’ to do the same. For this reason, this difference
will not be elaborated upon and will be referred to as “referential inclusion” or
“inclusion of do something under something”.
9. It is also possible that the finite verb combines with an infinitive only, that can
(e.g. in the case of rešit’ ‘decide’ or cannot (e.g. in the case of namerevat’sja
‘intend’ be referred to by means of a pro-noun, i.e. something. These cases
will not be considered here, but see Divjak (2004).
10. In other words, for planirovat’ ‘plan’ it does not matter what is being planned,
a thing or a process; the concept expressed by planirovat’ ‘plan’ is such that it
applies to both things and processes. As one of the reviewers pointed out, for
the complement it does of course make a difference whether it is conceptual-
ized as thing or as a process, but which form (nominal or verbal) is preferred
in which context remains beyond the scope of this paper.
11. Langacker (1990: 269–270) contends that each modal verb “evokes the con-
ception of an associated activity. One does not simply want, know or have a
physical capacity in the abstract – rather, one wants, knows, or has the capac-
ity to do something. Thus each verb makes schematic reference to another
process, which serves as a landmark and as the e-site for a relational comple-
ment”. Compare also Kemmer and Verhagen (1994) who define analytic cau-
satives as “necessarily invoking another action”; the thing-test visualizes this

claim.
12. Cristofaro (2003: 263) concludes: “some relation types involve easier concep-
tualization of the dependent State of Affairs as a thing”. According to Cristo-
faro, the reason for “easier conceptualization” lies in the nature of a thing, i.e.
a summarily scanned, non-relational entity that is characterized by stability
and persistence over time. Therefore, she claims, relation types pertaining to
the possible occurrence of “States of Affairs”, such as Modals, or to their
phasal development, such as Phasals, are basically incompatible with concep-
tualization of the dependent State of Affairs as a thing.
13. In Russian, that-clauses come in at least three forms, as a čto+indicative
clause, as a čtoby+infinitive clause and as a čtoby+past tense clause. A discus-
sion of these differences is beyond the scope of the present paper, however,
and the reader is referred to Divjak (2004).
14. Although Langacker (1991: 439–442) considers that, to, -ing and zero as
possible complementation markers for English, he stresses that these four dif-
ferent complementizers each “take a clause some distance along the path lead-
ing from a processual to a nominal profile”. Here, I consider the Russian coun-
terpart of that complementizers that arguably triggers a further distancing
effect (Langacker 1991: 447) and imposes a nominal profile to the highest ex-
tent: the demonstrative origin of English that underlines its function as nomi-
nalizer (cf. Langacker 1991: 440 fn. 21). Russian čto(by) may well relate to
the interrogative pronoun čto, augmented wth the particle by.
15. Wierzbicka (1988: 154, 156–158) observes that in Russian, the scope of the
infinitive construction is wider than in Polish (and French). Russian “uses the
bare infinitive for speech act verbs of strong expectation, such as prikazat’
‘order’ and zapretit’ ‘forbid’, but also for speech act verbs of weak expecta-
tion such as prosit’ ‘ask’ or umoljat’ ‘beg’ and even for negative verbs such as
otgovarivat’ ‘dissuade’. Similarly, there is no difference in this respect be-
tween mental verbs of strong expectation, such as rešit’ ‘decide’ and mental
verbs of weak expectation such as mečtat’ ‘wishfully dream’”.
16. Interestingly, native speakers remarked that it is easier to add conflicting tem-
poral modifiers if the two verbs are constructed as main clause and that-
complement clause; this supports my argument for the that-complement clause
as (likewise) signaling distance.
17. There are very few exceptions to this rule. For two verbs listed as “imperfec-
tive only” in dictionaries, high style molit’ ‘beg, plead’ and obsolete
uveščevat’ ‘admonish, exhort’, native speakers do accept conflicting temporal
modifiers to the finite verb and the infinitive. A similar effect occurs with
verbs that express coercion, e.g. prinevolit’ ‘force, make’ and vy-, pri-, po-
nudit’ ‘force, make’. Bogdanov (in Chrakovskij et al. 1983: 144) provides an
example with two different time adverbials added to the (admittedly, semanti-
cally related) finite verb zastavit’ ‘compel, force’ followed by an infinitive, i.e.
252 Dagmar Divjak
Vse utro mat’ zastavljala syna s’’ezdit’ večerom k babuške ‘all morning
mother was forcing her son to go by grandmother in the evening’. Important
here might be the fact that the finite verb has imperfective marking, i.e. has not
been taken to its natural ending.
18. My use of the term “bounded” does not imply reference to telicity.
19. Constructional data can be used to delineate semantically coherent categories
(cf. Croft 1999: 69–74; Divjak 2004, Divjak 2006, Levin 1993). If we look at
elements categorized as “1” in the first column of Table 2 in the Appendix, we
see that taking into account the details of the argument structure (i.e. preposi-
tion and case) yields subgroups that share meaning at a coarse-grained level.
For example, in category four we find verbs that express a positive or negative
attitude towards an event, e.g. bojat’sja ‘be afraid’, želat’1 ‘desire for oneself’,
izbegat’ ‘avoid’, opasat’sja1 ‘fear’, ostereč’sja ‘beware of’, stesnjat’sja ‘hesi-
tate’, stydit’sja ‘be ashamed’. For category 8 verbs, the constructional possi-
bilities of which are limited to combinations with an infinitive, there are no
conspicuous formal characteristics that can be used as the basis for further
subcategorization. Therefore, the verbs were sub-classified with the help of
data on their linear distribution, i.e. their mutual combinatorial possibilities in
verb triples. This discussion, however, exceeds the scope of this paper and the
reader is referred to Divjak (2004).
20. In Russian, construction kernels are sometimes termed predikator ‘predicator’.
Since this term is problematic in English I use the shorthand “P-verb”.
21. Adamec (1968) was the first to describe this class of verbs and termed them
modifikatory ‘modificators’. This term is fine in Russian and contrasts nicely
with predikatory ‘predicators’. Again, it is problematic in English and there-
fore a shorthand, “M-verb”, is used. Givón (1973: 894) uses the term M-verbs
(but attributes it in turn to Karttunen) to refer to “verbs requiring a sentential
complement and, in addition, requiring an equi-subject condition between the
main and complement verb”. I will restrict the term M-verb to verbs that, in
combination with an infinitive, form co-referential [VFIN VINF] constructions in
which the infinitive does not fill up one of the finite verb’s argument structure
slots, the infinitive cannot be extended to form a full-fledged complement; in
addition, clause and both verbs cannot take conflicting temporal modifiers.
22. One of the verbs that belongs to this category is ubedit’ ‘persuade’; interest-
ingly, Dal’ (1996: 575 vol. 2) does not mention the possibility of using ubedit’
‘persuade’ with an infinitive. A century ago, ubedit’ could only mean ‘force to
believe (in) something on the basis of facts, arguments, evidence’. In other
words, the verbs in this category might be loosing influence over the infinitive
they combine with, but they might also be in the process of acquiring such in-
fluence. A diachronic perspective could clarify this situation.
References
Adamec, P.
1968. K voprosu o modifikacijax (modal’nyx transformacijax) so znače-
niem neobxodimosti i vozmožnosti [On modifications (modal trans-
formations) expressing necessity and possibility]. Československa
Rusistika 13 (2): 88–94. (k VI. Mezinárodnímu sjezdu slavistů).
Apresjan, Jurij Derenikovič, and Erna Pall
1982 Russkij glagol – vengerskij glagol: upravlenie i sočetaemost’ [The
Russian Verb – the Hungarian Verb: Government and Collocation].
Budapest: Tankën’vkiado.
Atkins, B.T. Sue, and Beth Levin
1995 Building on a corpus: a linguistic and lexicographical look at some
near-synonyms. International Journal of Lexicography 8 (2): 85–
114.
Barentsen, Adriaan
s.d. The Amsterdam Corpus of Russian. Universiteit Amsterdam.
Benson, Morten, and Eveline Benson
1995 Russko-anglijskij slovar’ glagol’nyx slovosočetanij [Russian-English
Verbal Collocation Dictionary]. Moskva: Meždunarodnaja Škola
Perevodčikov.
Broccias, Christiano, and Willem B. Hollmann
(forthc) Do we need summary and sequential scanning in (Cognitive) Gram-
mar? Cogitive Linguistics 18 (4).
Chrakovskij, Vladimir S. et al.
1983 Kategorii glagola i struktura predloženija: konstrukcii s predikat-
nymi aktantami [Verbal categories and sentence structure: construc-
tions with predicative actants]. Leningrad: [s.l.].
Cristofaro, Sonia
2003 Subordination. (Oxford Studies in Typology and Linguistic Theory.)
Oxford: Oxford University Press.
Croft, William
1999 Some contributions of typology to cognitive linguistics, and vice
versa. In Cognitive Linguistics: Foundations, Scope, and Methodol-
ogy, Theo Janssen and Gisela Redeker (eds.), 61–93. (Cognitive Lin-
guistics Research 15.) Berlin/New York: Mouton de Gruyter.
2001 Radical Construction Grammar: Syntactic Theory in Typological
Perspective. (Oxford Linguistics.) Oxford: Oxford University Press.
Dal’, Vladimir I.
1996 Reprint. Tolkovyj slovar’ živogo velikorusskogo jazyka [Explanatory
Dictionary of the Great Living Russian Language]. Sankt-Peterburg:
Diamant.
254 Dagmar Divjak
Daum, Edmund, and Werner Schenk

1992 Die russischen Verben. Leipzig: Langenscheidt.
Denisov, P.N., and B.B. Morkovkin
1978 Učebnyj slovar' sočetaemosti slov russkogo jazyka [A Learner’s
Dictionary of Russian Word Collocations]. Moskva: Russkij Jazyk.
Divjak, Dagmar
2004 Degrees of Verb Integration. Conceptualizing and Categorizing
Events in Russian. Ph.D. diss., Dept. of Oriental and Slavic Studies,
K.U.Leuven (Belgium).
2006 Ways of Intending: Delineating and Structuring Near-Synonyms. In
Corpora in Cognitive Linguistics. Vol. 2: The Syntax-Lexis Interface,
St. Th. Gries and A. Stefanowitsch (eds.), 19–56. Berlin/New York:
Mouton De Gruyter.
2007 Witnessing a system collapse. The expression of complex events in
Russian vs Polish. Paper presented at BASEES 2007, Cambridge
(UK), March 31–April 2, 2007.
Dixon, Robert Malcolm Ward
1996 Complement clauses and complementation strategies. In Grammar
and Meaning: essays in honour of Sir John Lyons, Frank Robert
Palmer (ed.), 175–220. Cambridge: Cambridge University Press.
Eynde, Karel van den
1995 Methodological Reflections on Descriptive Linguistics. Knud To-
geby's Principles and the Pronominal Approach. In Studies in
Valency. Vol. I, Lene Schøsler and Mary Talbot, 111–130. (RASK
supplement 1.) Odense: Odense University Press.
Givón, Talmy
1973 The Time-Axis Phenomenon. Language 49 (4): 890–925.
2001 Syntax: an Introduction. Vols 1 and 2. Amsterdam/Philadelphia:
John Benjamins.
Goldberg, Adele
1995 Constructions: a Construction Grammar Approach to Argument
Structure. Chicago/London: Chicago University Press.
Hopper, Paul J., and Elizabeth C. Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
Janda, Laura A.
2004 A metaphor in search of a source domain: The categories of Slavic
aspect. Cognitive Linguistics 15 (4): 471–527.
Kemmer, Suzanne, and Arie Verhagen
1994 The grammar of causatives and the conceptual structure of events.
Cognitive Linguistics 5 (2): 115–156.
Lakoff, George, and Marc Johnson
1980 Metaphors We Live By. Chicago: University of Chicago Press.
1987a Nouns and verbs. Language 63 (1): 53–94.
1987b Foundations of Cognitive Grammar: Theoretical Prerequisites.
Stanford: Stanford University Press.
1991 Foundations of Cognitive Grammar: Descriptive Application. Stan-
ford: Stanford University Press.
Levin, Beth
1993 English Verb Classes and Alternations: a Preliminary Investigation.
Chicago: University of Chicago Press.
Ožegov, S.I., and N.Ju. Švedova
1995 Tolkovyj slovar’ russkogo jazyka [Explanatory Dictionary of Rus-
sian]. Moskva: Az’’.
1999 Tolkovyj slovar’ russkogo jazyka [Explanatory Dictionary of Rus-
sian]. Moskva: Azbukovnik.
Pustejovsky, James
1991 The syntax of event structure. In Lexical and Conceptual Semantics,
Beth Levin and Steven Pinker (eds.), 47–81. Cambridge MA/Oxford
UK: Blackwell Publishers.
Quine, Willard Van Orman
1964 From a Logical Point of View. Nine Logico-Philosophical Essays,
1–19. Cambridge MA: Harvard University Press.
RNC
2004 Russian National Corpus (www.ruscorpora.ru)
Schulte im Walde, Sabine
(2007) The induction of verb frames and verb classes from corpora. In Cor-
pus Linguistics. An International Handbook, Anke Lüdling, Merja
Kytö and Anthony Mc Enery (eds.). Berlin: Mouton de Gruyter.
Smessaert, Hans, Bert Cornillie, Dagmar Divjak, and Karel van den Eynde
2005 Degrees of clause integration. From endotactic to exotactic subordi-
nation in Dutch. Linguistics. An Interdisciplinary Journal of the
Language Sciences 43: 471–530.
Smith, Carlotta
1997 The Parameter of Aspect. Second Edition. (Studies in Linguistics and
Philosophy 43.) Dordrecht: Kluwer Academic Publishers.
Stefanowitsch, Anatol
2006 Negative evidence and the raw frequency fallacy. Corpus Linguistics
and Linguistic Theory 2 (1): 61–77.
Talmy, Leonard
2000 Toward a Cognitive Semantics. Vol. 1. Concept Structuring Systems.
(Language, Speech and Communication.) Cambridge MA: MIT
Press.
256 Dagmar Divjak
Wierzbicka, Anna
1975 Why ‘Kill’ Does Not Mean ‘Cause To Die’: The Semantics of Action
Sentences. Foundations of Language 13: 491–528.
1988 The Semantics of Grammar. (Studies in Language Companion Series
18.) Amsterdam/Philadelphia: John Benjamins.
The ‘impersonal’ impersonal construction in Polish.
A Cognitive Grammar analysis
Anna Słoń
Abstract
The aim of the paper is to examine one of impersonal constructions in Pol-

ish, namely the 3rd SG NEUTR construction. The construction is special
among other Polish impersonal constructions in that the instigator it defo-
cuses is non-human and inanimate. In this sense it is truly ‘impersonal’. It is
shown that the construction is used when the instigator is particularly diffuse
and difficult to identify. Various subconstructions of the 3rd SG NEUTR con-
struction differ in the construal which they impose on the event and in
whether the defocused instigator can be identified in corresponding ‘per-
sonal’ constructions (i.e. those which can have a subject). The ‘personal’
variants of one of the subtypes have either cognate or highly predictable
subjects; these subjects, however, cannot be treated as real instigators. Next
type’s ‘personal’ variants specify an instrument rather than an instigator. In
the third type, when a subject is at all available in a corresponding ‘per-
sonal’ construction, it is not an instigator but an active zone within the ex-
periencer. The analysis is conducted within the framework of Cognitive
Grammar (Langacker 1987, 1990, 1991, 1999).
Keywords: impersonal constructions, instigator defocusing, construal, ac-

tive zone, setting-subject constructions
1. Introduction
For decades Polish impersonal constructions have been the subject of lin-
guistic analyses, both theoretically-neutral, descriptive in character (see
Doros 1975; Grzegorczykowa 1996; Bogusławski 1984; Wolińska 1978)
and more generatively-oriented (e.g. Jackiewicz 1992; Kubiszyn-Mędrala
1994; Węgiel 1994). The descriptive analyses yielded many an insight into
the nature and the semantics of Polish impersonals; they failed, however, to
capture the more general characteristics of impersonals, to see impersonals
258 Anna Słoń
in a wider perspective of constructions fulfilling similar functions or to

account for the specifics of a given construction in terms of differences in
conceptualisation. Generative studies, on the other hand, did not contribute
much to the understanding of the nature of impersonal constructions, be-
cause they ignored what is most interesting in Polish impersonals, namely
their semantics: subtle semantic differences among different impersonal
constructions, the ways in which various grammatical markers contribute
to the meaning of a particular construction, or how the conceptualisation of
an event influences the choice of a particular coding.
Ignoring the semantic details of particular impersonal constructions in
Polish makes it impossible to account for their number and variety; an
overview of the most important types of impersonal constructions is given
in (1):1
[The -NO/-TO construction]

(1) a. Przesyłkę dostarczono dzisiaj rano.
parcel.FEM.ACC deliver.-NO/-TO today morning
‘The parcel has been delivered this morning.’
[The 3rd SG NEUTR + się construction]

b. Tak się to pisze.
2
so MM this.NEUTR.ACC write.3SG.PRES
‘One should spell it like this./ It should be spelt like this.’
[The 3rd PL VIR construction]

c. Dali mi podwyżkę!
give.3PL.VIR.PAST.PERF me.DAT rise.FEM.ACC
‘They gave me a rise!’
[The 1st PL VIR constr.]

d. Przez ulicę przechodzimy na
through/across street.FEM.ACC cross.1PL.PRES on
zielonym świetle.
[green light].NEUTR.LOC
‘We cross the street on green lights.’
The ‘impersonal’ impersonal construction in Polish 259
[The 2nd SG constr.]

e. Nigdy nie wiesz czego się
never NEG-PART know.2SG.PRES what.GEN MM
po nim spodziewać.
after him.LOC expect.INF
‘You never know what to expect of him.’
[The 3rd SG NEUTR construction]

f. Zatrzęsło mną z
shake.3SG.NEUTR.PERF.PAST me.INSTR from
oburzenia.
indignation.NEUTR.GEN
‘I was shaken with indignation.’
The problem is made even more complex due to the variation within a
particular construction. A given construction may be further analysed as a
category of subconstructions resulting from the merger of the construction
with a particular type of verbal predicates, the interplay of the impersonal
construction with some other construction, or the interaction with (widely
understood) context. Consider, for instance, example (1b). Normally it is
used as a generic statement describing the accepted way of behaving or
doing something. It is possible, however, to provide a context (e.g. a
teacher showing to a pupil how to produce a particular letter shape) in
which the sentence would mean ‘This is the way how it is being written’.
If, instead of the present tense, the future tense is used, the sentence Tak się
to napisze (so MM this.ACC write.3SG.FUT) would lose its generic meaning
and be roughly equivalent to ‘This is the way how it’s going to be written’
(with a complete lack of commitment on the part of the speaker). Similarly
non-committal is the statement Tak mi się to napisało (so me.DAT MM
this.ACC write.3SG.NEUTR.PAST.PERF)‘I just happened to write it like this’,
where the dative participant is seen as experiencing the situation rather
than as being responsible for it.
The present paper analyses one particular Polish impersonal construc-
tion, namely the 3rd SG NEUTR construction and its subtypes from a Cogni-
tive Grammar point of view (Langacker 1987, 1990, 1991, 1999). Cogni-
tive Grammar (henceforth CG) can handle what has been neglected in
previous approaches and provide a fuller, more adequate analysis of Polish
impersonals, taking into account their form, their meaning and the interre-
lation between the two. In general, CG, which treats meaning as conceptu-
alisation and claims that every morpheme and the sequence thereof is
260 Anna Słoń
meaningful, is particularly suited to the analysis of Slavic languages with

their rich inflectional morphology and (relatively) free word order. The
structure of the paper is as follows: in Section 2, I provide a brief explana-
tion of why the lack of a subject nominal cannot be always taken as an
indication of impersonality. Section 3 analyses different subcontructions of
the 3rd SG NEUTR construction in terms of the contrasting imagery that they
evoke, motivating why they all use the 3rd singular neuter form of the verb.
I also examine the possibility of coding the agent, or more generally the
instigator,3 of the event in a related ‘personal’ construction. In Section 4 I
show the affinity of the 3rd SG NEUTR construction to other impersonal con-
structions employing the 3rd singular neuter form of the verb. I offer con-
clusions in Section 5.
2. A note on Polish grammar: the absence of subject nominals
Unlike in English, French or German, subjects are frequently left unelabo-

rated in Polish. This is not exclusively or predominantly a phenomenon of
colloquial language, as is the case in English, but a characteristic of Polish
grammar – a characteristic connected with the information structure of
discourse. Whether it is a full nominal, a pronoun or null phonetic element
that elaborates the trajector in a clause depends on two main factors: the
accessibility of the subject’s referent at a given point in discourse (cf. Ariel
2000), and its attention-worthiness4 at this particular point in discourse.
The higher the accessibility of the subject’s referent, the less phonetic ma-
terial is necessary, so full nominals tend to introduce new participants into
discourse, while lack of elaboration on the phonological pole is frequently
anaphoric, as shown in (2a). Pronouns, on the other hand, tend to occur for
the sake of contrast (when more than one participant is prominent at a
given point in discourse or when the subject’s referent, despite its topical-
ity, is particularly attention-worthy), like in (2b).5 When the distinction
offered by pronouns is not sufficient to make an unambiguous reference,
full nominals can be used for contrast (and/or emphasis), as in (2c):
(2) a. Maria Kowalska? Niewiele o niej

[Maria Kowalska].NOM little about her.LOC
wiem. Ø/*Ona/*MK pracuje
know.1SG.PRES Ø/she.NOM/[MK].NOM work.3SG.PRES
w banku, Ø/*ona/*MK niezbyt
in bank.MASC.LOC Ø/she.NOM/[MK].NOM not-much
udziela się towarzysko Ø/*ona/*MK

has-an-active-social-life.3SG.PRES Ø/she.NOM/[MK].NOM
ma kota i Ø/*ona/*MK
have.3SG.PRES cat.MASC.ACC and Ø/she.NOM/[MK].NOM
mieszka gdzieś niedaleko centrum.
live.3SG.PRES somewhere near centre.NEUTR.GEN
‘Maria Kowalska? I don’t know much about her. She
works at a bank, she’s not very sociable, has a cat and lives
somewhere near the centre.’
b. Widziałeś tę parę na
saw.2SG.MASC.PAST.IMPERF [this couple].FEM.ACC on
ławce? Ona/*Ø kiedyś u nas
bench.FEM.LOC she.NOM/Ø once at us.GEN
pracowała, a on/*Ø to
work.3SG.FEM.PAST and he.NOM/Ø this.NEUTR.NOM
były poseł.
has-been.SG.MASC.NOM MP.MASC.NOM
‘Have you seen the couple on the bench? She used to work
with us and he’s an ex-MP.’
c. Widzisz te dwie laski tam

see.2SG.PRES [these two girls].PL.NON-VIR.ACC there
przy barze? Ta w czerwonej sukience
at bar.MASC.LOC this.FEM.NOM in [red dress].FEM.LOC
to Monika, a ta w
this.NEUTR.NOM Monika.NOM and this.FEM.NOM in
dżinsach to Aśka.
jeans.PL.NON-VIR.LOC this.NEUTR.NOM Aśka.NOM
Monika/*Ona/*Ø jest na prawie, a
Monika.NOM/she.NOM/Ø is on law.NEUTR.LOC and
Aśka/*Ona/*Ø właśnie skończyła
Aśka.NOM/she.NOM/Ø just finish.3SG.FEM.PAST
germanistykę.
German-studies.FEM.ACC
‘Do you see the two girls over there at the bar? The one in
a red dress is Monika, and the one in jeans is Aśka.
Monika’s studying law, and Aśka’s just graduated in Ger-
man studies.’
262 Anna Słoń
(3) a. ‘Pracuje w banku.’

work.3SG.PRES in bank.MASC.LOC
‘Czekaj, czekaj – kto pracuje w
[wait wait].2SG.IMPERAT who.NOM work.3SG.PRES in
banku?’ ‘No ona!/ No Kowalska!’
bank.MASC.LOC EXCL she.NOM/ EXCL Kowalska.NOM
‘Works at a bank.’ ‘Wait a moment – who works at a
bank?’ ‘Oh, she does!/Kowalska, of course!’
b. ‘Ciągle pada.’ ‘Czekaj, czekaj –

still fall.3SG.PRES [wait wait].2SG.IMPERAT
co pada?’ ‘No deszcz!’
what.NOM fall.3SG.PRES EXCL rain.NOM
Lit. ‘Still falling.’ ‘Wait a moment – what’s falling?’ ‘Oh,
rain, of course!’
c. ‘Przepraszam, poniosło
apologise.1SG.PRES carry-away.3SG.NEUTR.PAST.PERF
mnie.’ ‘Co niby cię
me.ACC what.NOM like you.ACC
poniosło?’ ‘???’
carry-away.3SG.NEUTR.PAST.PERF ‘???’
‘Sorry, I got carried away.’ ‘You got carried away by
what?’ ‘???’
When the trajector6 is not elaborated due to its referent’s high accessi-
bility and/or low attention-worthiness, the referent’s identity is nonetheless
recoverable from the context. This is illustrated in (3a), where one of the
interlocutors uses an unelaborated trajector assuming that its referent’s
identity is known to the other interlocutor. If it turns out that this is not the
case, the full nominal can be provided. A partially similar case is exempli-
fied in (3b). Here, however, the choice of referents is limited contextually,
and the expression itself is highly conventionalized. In contrast, (3c) illus-
trates a situation where the trajector cannot be elaborated because the iden-
tity of the trajector’s referent is difficult or even impossible to pin down. It
is this last type of trajector non-elaboration, and partly the trajector non-
elaboration accompanying the weather verbs, like the one in (3b), that oc-
curs in impersonal constructions.7
In sum, the absence of the subject nominal in Polish is not sufficient to
make a sentence impersonal. In impersonal constructions, like those in (1)
or (3c), there is no grammatical possibility of using a subject nominal. In

other words, the trajector in impersonal constructions is left obligatorily
unelaborated. The trajector’s referent is thus non-identified or potentially
non-identified (for various reasons, such as the diffuseness of the trajec-
tor’s referent, unclear reference, generic or obvious reference, a wish to
conceal its reference). That is contrasted with cases where the subject is
absent but can be grammatically supplied – like in (3a) – and the trajector’s
referent is identified or potentially identified (hence the anaphoric function
of such subject nominal omissions). It is also important to note that the
frequent omission of subject nominal in the anaphoric function is made
possible by the verbal morphology: the categories of number, person and
gender8 are marked on the verb. In contrast, verbs in a particular imper-
sonal construction have a specified form – they do not inflect for number,
person or gender. Frequently this specified form, as in the construction
under discussion, is the 3rd person singular neuter. The specified verb form
is a constitutive element of an impersonal construction and changing it
leads to the loss of the impersonal interpretation. Compare for instance the
impersonal (1d), which describes a desired way of (anyone’s) behaviour to
‘personal’ Przez ulicę przechodzą na zielonym świetle (the verb is present
3rd person plural), which states a fact about some more or less specified
group of people, or to ‘personal’ Przez ulicę przechodziliśmy na zielonym
świetle (the verb is past imperfective 1st person plural virile), which de-
scribes a past habit of some (again more or less specified) group of people
including the speaker.
3. The analysis
In Polish impersonals constitute a large and varied group of constructions.

They are distinguished within a wider category of instigator defocusing
constructions9 on the basis of the particular instigator defocusing mecha-
nism they employ: impersonals defocus the instigator by its obligatory non-
elaboration. Most impersonals leave unelaborated the instigator-agent
which is human. There is, however, one construction – the 3rd SG NEUTR
construction – in which the instigator is inanimate or at least so diffuse that
the issue of animacy is obliterated.10
Consider the following sentences exemplifying the 3rd SG NEUTR con-
struction:11
264 Anna Słoń
(4) a. Już świtało.

already dawn.3SG.NEUTR.PAST IMPERF
‘It was already dawning.’
b. Górę zasnuło
mountain.FEM.ACC veil.3SG.NEUTR.PAST.PERF
mgłą.
fog.FEM.INSTR
‘The mountain was veiled with/ in fog.’
c. Swędziało go.
itch.3SG.NEUTR.PAST.IMPERF him.ACC
‘He felt itching.’
d. Było zimno.
be.3SG.NEUTR.PAST cold.ADV
‘It was cold.’
In all of the above sentences the instigator’s referent is diffuse and/or

difficult to identify. In what way does this motivate the use of the 3rd per-
son singular neuter form of the verb? The 3rd person signals distinctness
from the speaker and hearer – the event unfolds independently from them
and they have no control over it. The singular is either used for single enti-
ties or for multiplex entities where individuation is not at issue, when they
are treated as a mass. The neuter gender can suggest non-animate nature by
virtue of its opposition to masculine and feminine (in their prototypical
value associated with animacy). It is pointed out by Doros (1975) that the
singular neuter may have some connection with the setting. More recently,
Tabakowska (2004) investigated other phenomena involving the singular
neuter category and pointed out that the 3rd person singular neuter verb
form signals the minimum individuation of the agent. Thus, if the agent is
minimally individuated, possibly multiplex, then its location/ setting may
stand for it metonymically.12
While the general motivation for the use of the 3rd SG NEUTR construc-
tion (whose morpho-syntactic exponents are the 3rd person singular neuter
form of the verb and the non-elaboration of the trajector) is the diffuseness
of the instigator, sentences in (4) differ as to why the instigator is defo-
cused. Sometimes, when a different construction is used to code the same
event, the instigator can be elaborated. In most cases, though, this ‘addi-
tional’ nominal does not elaborate an instigator but some other important
aspect of the situation depicted in the clause. The identity of this ‘addi-
tional’ nominal’s referent constitutes, therefore, another factor that distin-

guishes the subconstructions. The remaining part of the article is devoted
to examining the (sometimes very subtle) distinctions in the imagery dis-
played by the subtypes of the 3rd SG NEUTR construction.
3.1. What is it that does the dawning?
Sentence (4a) exemplifies the type of the 3rd SG NEUTR construction in

which the instigator is defocused because mentioning it explicitly would
not greatly contribute semantically to the overall meaning. Wolińska
(1978: 94) distinguishes three types of possible subjects occurring in the
corresponding personal construction: (a) tautological with the verb – the
nominal elaborating the agent would be a cognate subject, as in (5a); (b)
pleonastic, i.e. predictable from the verb, as in (5b), and (c) indefinite –
pronoun coś ‘something’. Note, however, that if the verb świtać ‘dawn’ is
used in an extra-meteorological context, as illustrated in (5c), the subject
will normally be elaborated since it is no longer predictable.13
(5) a. ?Świt świtał.

dawn.MASC NOM dawn.3SG.MASC.PAST.IMPERF
‘The dawn was dawning.’
b. Dzień świtał.14
day.MASC.NOM dawn.3SG.MASC.PAST.IMPERF
‘The day was dawning.’
c. Zaświtał jej
dawn.3SG.MASC.PAST.PERF her.DAT
nowy świetny pomysł.
[new excellent idea].MASC.NOM
‘A new excellent idea has dawned on her.’
The use of an impersonal construction to code a ‘meteorological’ situa-

tion is well motivated, because we naturally focus on the process or state
itself (and possibly on how it affects us) rather than on what or who is the
ultimate cause of this process or event. Instigator defocusing is also fa-
voured in a context of this kind because even when we recognise the exis-
tence of the cause, it is normally diffuse and/or difficult to pin down. The
contrast between the construals imposed by the impersonal and the ‘per-
sonal’ constructions is illustrated by the examples in (6a-b):
266 Anna Słoń
(6) a. Już świtało.

already dawn.3SG.NEUTR.PAST.IMPERF
‘It was already dawning.’
b. ?Świt już świtał.
dawn.MASC.NOM already dawn.3SG.MASC.PAST.IMPERF
‘The dawn was already dawning.’
c. It was already dawning.
In (6a) the only profiled element of the event is the process alone, while
in (6b), the profile encompasses not only the process but also quite a non-
prototypical instigator which is at the same time part of the setting and a
reification of the process. The English example in (6c) shows another con-
ceptual strategy of handling a diffuse, setting-like instigator, and is in-
cluded here for the sake of comparison with the Polish examples. English
it, providing an obligatory subject in weather expressions is treated by
Langacker (1991) as profiling a default setting. The imagery evoked by the
three sentences in (6) is shown schematically in Figure 1:
a Inst
setting b Inst
setting
tr
c Inst
setting
tr
Figure 1. a. The impersonal construction in Polish; b. A cognate-subject construc-

tion; c. The impersonal it construction in English (based on Langacker
1991: 366)
To conclude this section, let us point out that the availability of the
cognate-subject construction is limited by morphological considerations
and is thus language specific. The dawn is dawning example has as its Pol-
ish counterpart the sentence ?Świt świta, but Rain is raining or Snow is
snowing do not have such counterparts because Polish uses the verb padać
‘fall’, which is morphologically unrelated to Polish nouns for ‘rain’ deszcz
or ‘snow’ śnieg. The only available Polish version is Pada deszcz/ śnieg
‘The rain/ snow is falling’.
3.2. An instrument with no wielder
Sentence (4b), repeated below as (7a) illustrates another subtype of the 3rd
15
SG NEUTR construction. Here the instigator is likewise defocused, but the
instrument is elaborated, which opens up a possibility of treating it as an
instigator, as in sentence (7b). When an ‘additional’ participant is intro-
duced, it will be viewed as the wielder of the instrument,16 as shown in
(7c):
(7) a. Górę zasnuło

mountain.FEM.ACC veil.3SG.NEUTR.PAST.PERF
mgłą.
fog.FEM.INSTR
‘The mountain was veiled with/ in fog.’
b. Mgła zasnuła górę.
fog.FEM.NOM veil.3SG.FEM.PAST.PERF mountain.FEM.ACC
‘The fog veiled a/the mountain.’
c. Wiatr zasnuł
wind.MASC.NOM veil.3SG.MASC.PAST.PERF
górę mgłą.
mountain.FEM.ACC fog.FEM.INSTR
‘The wind veiled the mountain with/ in fog.’
The three construals are presented in Figure 2. In sentence (7a) and Figure
2a, the only elaborated participants are the patient and the instrument with
the instigator left obligatorily unelaborated. The 3rd SG NEUTR form of the
verb implicitly agrees with a diffuse, unidentified force. It seems likely that
when the causal element is difficult to identify and by virtue of its diffuse-
ness ‘occupies’ a considerable portion of the setting, it may be metonymi-
cally equated with the setting itself.17 The shading of the instigator and the
268 Anna Słoń
setting in Figure 2a symbolises the diffuseness of the former, while the

dotted-line circle signifies its obligatory unelaboration.
a setting IS
Inst Instr Pat
tr lm2 lm1
b Inst IS
Instr Pat
tr lm
c IS
Inst Instr Pat
tr lm2 lm1
Figure 2. a. The 3rd sg neutr impersonal construction – the instigator remains obli-
gatorily unelaborated; b. The narrowing of the immediate scope of predi-
cation resulting in the conferring of the trajector–instigator status on the
instrument; c. The personal, instigator–subject construction.
Sentence (7b), schematically presented in Figure 2b, is an instance of

immediate scope narrowing whereby the instrument is construed as the
instigating factor. The problem of diffuseness and/or lack of an identifiable
instigator is altogether avoided by removing the original causal element
from the immediate scope of predication. Note, however, that in sentence
(7b) we are not dealing with instigator defocusing, because the narrowed
immediate scope does contain a level of causation (it contains an instigator,
albeit a non-prototypical one, i.e. an entity which is just as likely construed
as a causal force and as an instrument), nor is the narrowing of the imme-
diate scope of predication accompanied by any formal devices, such as się

in the middle construction.18
Finally, sentence (7c), illustrated in Figure 2c, provides an explicit in-
stigator – in this case its identity can be recovered on the basis of common
knowledge – an ICM (an idealised cognitive model, see Lakoff 1987) of
natural/ meteorological phenomena. If the interpretation of the instigator
coincides with the information provided by the ICM, an overt mentioning
of the instigator is not particularly informative and thus it may be dis-
preferred. It must be noted that impersonal constructions in general defocus
19
ACCESSIBLE instigators, thus allowing the speaker to avoid mentioning
the instigator if it is recoverable to a sufficient degree and non-salient in a
given stretch of discourse. Compare, however, sentence (7c) – where the
overt mention of the instigator does not provide information which could
not be inferred from the ICM – to another sentence: Czarodziej zasnuł górę
mgłą. wizard.MASC.NOM veil.3SG.MASC.PAST.PERF mountain.FEM.ACC
fog.FEM.INSTR ‘The wizard veiled the mountain with/ in fog.’ In the latter
case the omission of the agent would considerably affect the amount of
information conveyed, because the subject nominal provides a piece of
information that is not supplied by the ICM (as the subject’s referent is not
part of this ICM).20
3.3. An instigator or an active zone?
The defocusing of a diffuse, unidentifiable instigator is even more con-

spicuous in (4c), repeated as (8a) below. In the ‘personal’ variant (8c), the
subject nominal does not elaborate the instigator but a location-patient
(which is part of the experiencer). Even the use of the indefinite pronoun
coś ‘something’ in (8b) does not facilitate an instigator interpretation. Coś
elaborates an unspecified region of the experiencer’s body. Therefore in
both the impersonal and the ‘personal’ variants, the instigator remains la-
tent; it is naturally defocused because of its perceptual (and possibly also
conceptual) non-salience.
270 Anna Słoń
(8) a. Swędziało go.

itch.3SG.NEUTR.PAST.IMPERF him.ACC
‘He felt itchy.’
b. Coś go swędziało.
something.NOM him.ACC itch.3SG.NEUTR.PAST.IMPERF
‘Something was itching him.’
c. Swędziała go ręka.
itch.3SG.FEM.PAST.IMPERF him.ACC hand.FEM.NOM
‘His hand was itching.’
In this particular example, the defocused instigator is normally concep-

tualised as part of the experiencer participant, but in many cases the dis-
tinction instigator-as-part-or-not-part-of-experiencer is blurred. Consider
the following examples:
(9) a. Kusiło/ Korciło go by jej

tempt/itch.3SG.NEUTR.PAST.IMPERF him.ACC to her.DAT
o tym powiedzieć.
about it.LOC tell.PERF.INF
‘He felt tempted/ was itching to tell her about it.’
b. Coś go kusiło/ korciło
something.NOM him.ACC tempt/itch.3SG.NEUTR.PAST.IMPERF
by jej o tym powiedzieć.
to her.DAT about it.LOC tell.PERF.INF
‘Something tempted/ was tempting him to tell her about it.’
c. Licho go kusiło/ ?*korciło
devil.NEUTR.NOM him.ACC tempt/itch.3SG.NEUTR.PAST.IMPERF
‘The devil tempted/ was tempting him to tell her about it.’
(10) a. Świerzbiło go by jej

itch.3SG.NEUTR.PAST.IMPERF him.ACC to her.DAT
o tym powiedzieć.
about it.LOC tell.PERF.INF
‘He felt an urge/ was itching to tell her about it.’
b. *Coś go świerzbiło
something.NOM him.ACC itch.3SG.NEUTR.PAST.IMPERF
c. Język go świerzbił
tongue.MASC.NOM him.ACC itch.3SG.NEUTR.PAST.IMPERF
‘He felt an urge/ was itching to tell her about it.’
(11) a. Świerzbiło ją by mu
itch.3SG.NEUTR.PAST.IMPERF her.ACC to him.DAT
przyłożyć.
beat.INF
‘She was itching/ felt an itch to box his ears.’
b. *Coś ją świerzbiło
something.NOM her.ACC itch.3SG.NEUTR.PAST.IMPERF
by mu przyłożyć.
to him.DAT beat.INF
c. Ręka ją świerzbiła
hand.FEM.NOM her.ACC itch.3SG.FEM.PAST.IMPERF
by mu przyłożyć.
to him.DAT beat.INF
‘Her hand was itching to box his ears.’
In example (9) with the verb kusić ‘tempt’, the temptation is neither
seen as coming entirely from outside, nor entirely from inside the experi-
encer participant; thus the situation can be conceptualised both ways. Sen-
tence (9a) suggests a more internal source of temptation, sentence (9c),
with its specified instigator a more external one, while (9b) is ambiguous
between the two, although it is more likely to imply an external source.
The verb korcić ‘tempt, itch’, does not allow the external interpretation
with a specified instigator – see (9c), but it does allow alluding to some
external force, as in (9b). The verb świerzbić resembles swędzieć (in their
physiological sense they are near synonyms, meaning ‘itch’), but in exam-
ples (10)–(11) the former is used metaphorically to render compulsion. The
source of the compulsion cannot be identified linguistically: sentences
(10b) and (11b) with coś are odd, whereas sentences (10c) and (11c) are
272 Anna Słoń
natural, though they do not specify the instigator, but like (8c), feature an
active zone within the experiencer participant. Note the acceptability of
(8b) with the indefinite subject: here, however, the meaning is not that
some unspecified force was making someone feel an itch, but that someone
felt an itching in some unspecified place in his body. With (10b) and (11b)
such an interpretation is impossible due to the idiomaticity of expressions
in (10c) and (11c).
Figure 3 illustrates the contrasting construals for the impersonal vari-
ants in (8a), (9a), (10a) and (11a), versus the active zone personal variants
in (8c), (10c) and (11c):
a Exp b
Exp
AZ
tr
lm lm
Figure 3. a. An impersonal construction with a diffuse instigator; b. The ‘personal’

variant – the subject nominal elaborates an active zone, not the instigator.
In Figure 3a, the experiencer constitutes the overall location for (or par-
ticipant affected by) the event whose instigator is not specified. Because it
is sometimes unclear whether the instigating force originates inside or out-
side the experiencer, the part of the arrow that reaches beyond the experi-
encer is drawn with a dashed line. In Figure 3b the trajector is elaborated,
but it does not specify the instigator. Instead it specifies an active zone (cf.
Langacker 1990, 1991) within the overall experiencer–location, i.e. the
active zone which is directly affected by the event.
A related subgroup of the 3rd SG NEUTR construction is exemplified in
sentence (4d), repeated below as (12a). In this subconstruction there is a
linking verb, most typically być ‘be’ or (z)robić się ‘grow, get, become’,
which is complemented with an adverb describing the state of a setting.21
Examples of such adverbs are: gorąco ‘hot’, ciepło ‘warm’, chłodno ‘cool’,
zimno ‘cold’, mroźno ‘freezing’, etc.; duszno ‘stuffy’, parno ‘muggy’,
wilgotno ‘humid’; ciemno ‘dark’, jasno ‘bright’; deszczowo ‘rainy’, bur-
zowo ‘stormy’, pochmurno ‘cloudy’; ładnie ‘nice’, brzydko ‘ugly’, ponuro
‘gloomy’. The construction is undoubtedly impersonal, because the trajec-
tor is unelaborated and cannot be elaborated by a nominal. It is nonetheless

possible to elaborate the setting,22 as in (12b):
(12) a. Było zimno.

be.3SG.NEUTR.PAST cold ADV
‘It was cold.’
b. Na dworze/ W pokoju
on outside.MASC.LOC / in room.MASC.LOC
było zimno.
be.3SG.NEUTR.PAST cold ADV
‘It was cold outside/ in the room.’
The setting, however, is not the only event element that can be elabo-
rated in this construction. Instead of referring to the state of the setting, one
can refer to the state experienced by a sentient participant, as shown in
(13a). The construal with an experiencer comes close to that triggered by
(8), and like in (8) it is possible to specify the active zone, as is done in
(13b):
(13) a. Było mi zimno.

be.3SG.NEUTR.PAST me.DAT cold ADV
‘I felt/was cold.’
b. Było mi zimno w ręce.
be.3SG.NEUTR.PAST me.DAT cold ADV in hands.NON-VIR.ACC
‘My hands felt/were cold.’
Another similarity between examples (12)–(13) and the previous exam-

ples from this section is that the sensation experienced by the participant
may have its source outside this participant (someone feels cold because it
is cold outside), as well as inside (someone feels cold because s/he has a
fever, low blood-pressure, etc.). In Figure 4 I suggest schematic representa-
tions for the sentences in (12) and (13).
In Figure 4a the situation is conceptualised as a pure state. It is possible
to elaborate the setting where the state takes place, as illustrated in Figure
4b, the experiencer of this state, as shown in Figure 4c, or even the affected
active zone within the experiencer, as in Figure 4d. However, the active
zone will not have the status of the trajector as in Figure 3b. The dotted
segment of the arrow shows that the state may originate both within the
experiencer and outside him/her.
274 Anna Słoń
a b
c Exp d Exp
AZ
lm lm
Figure 4. a. An impersonal construal of the state itself; b. An impersonal construal

of the state of some setting; c. An impersonal construal of the state affect-
ing the experiencer participant; d. An impersonal construal of the state af-
fecting an active zone within the experiencer participant.
4. A note on related impersonal constructions

No construction is an island and the 3rd SG NEUTR construction is no excep-
tion. The 3rd person singular verb form occurs also in the 3rd SG NEUTR +
się construction exemplified in (1b),23 the setting-subject-related imper-
sonal construction and in the impersonal locative construction.24 In note 12
I have already referred to the setting-subject construction, giving the fol-
lowing examples:
(14) a. Ogród roił się

garden.MASC.NOM swarm.3SG.MASC.PAST.IMPERF MM
od pszczół.
from bee.PL.NON-VIR.GEN
‘The garden was swarming with bees.’
b. Suknia mieniła się
dress.FEM.NOM glitter.3SG.FEM.PAST.IMPERF MM
cekinami.
sequin.PL.NON-VIR.INSTR
‘The dress was glittering with sequins.’
In those sentences the trajector status is conferred on the set-

ting/location although the actual instigators of the activity are bees and
sequins respectively. While English uses freely the same verb in the set-
ting-subject variant, as in the translations of (14a–b), and the instigator-
subject variant, exemplified in (15) and the translation of (16), Polish does
so relatively rarely. Thus the instigator-subject variant of (14b) is fully
acceptable – as shown in (16), whereas the instigator-setting variant of
(14a) would require a different verb – as in (17):25
(15) Bees were swarming in the garden.

(16) Cekiny mieniły się
sequin.PL.NON-VIR.NOM glitter.3PL.NON-VIR.PAST.IMPERF MM
na sukni.
on dress.FEM.LOC
‘Sequins were glittering on the dress.’
(17) Pszczoły krzątały się

bee.PL.NON-VIR.NOM hustle.3PL.NON-VIR.PAST.IMPERF MM
po ogrodzie.
on/across garden.MASC.LOC
‘Bees were swarming in the garden.’
In Polish, however, there is yet another construal available: the instiga-

tor is defocused, like in the setting-subject construction, but the set-
ting/location is not construed as a participant; it retains the status of set-
ting/location in its most typical coding, namely the prepositional phrase.26
This is illustrated in (18):
(18) a. W ogrodzie roiło

in garden.MASC.LOC swarm.3SG.NEUTR.PAST.IMPERF
się od pszczół.
MM from bee.PL.NON-VIR.GEN
‘The garden was swarming with bees.’
‘In the garden there were bees swarming.’
Lit.: ‘In the garden it/there was swarming from bees.’
276 Anna Słoń
b. Na sukni mieniło się

on dress.FEM.LOC glitter.3SG.NEUTR.PAST IMPERF MM
od cekinów.27
from sequin.PL.NON-VIR.GEN
‘The dress was glittering with sequins.’
Lit.: ‘On the dress it/there was glittering from sequins.’
The sentences in (18) are impersonal. The verb does not show agree-
ment with any element in the sentence; it has the form of 3rd person singu-
lar neuter. The construal offered by those sentences is that of focusing on
the process alone, with the actual instigators of the process and the set-
ting/location made less prominent. The three construals, i.e. those associ-
ated with the instigator-subject variant, setting-subject variant and the im-
personal variant, are shown in Figure 5:
a tr tr setting b setting
tr tr tr
tr tr tr
tr tr tr
tr tr tr lm
lm tr
c setting
lm
Figure 5. a. The non-setting-subject (i.e., instigator-subject) variant: the multiplex

instigators initiate an event in some setting; b. The setting-subject con-
struction: the multiplex instigator as an instrument or the stimulus for the
event – the instigator is defocused as a consequence of conferring trajec-
tor status on the setting; c. The impersonal setting-subject related con-
struction – only the process is in focus.
Example (19) illustrates a case that is partly similar to the alternation of

the setting-subject variant and the impersonal variant, as depicted in (14a–
b) and (18a–b), and the alternation referred to in note 22:
(19) a. Cały pokój pachniał

[whole room].MASC.NOM smell.3SG.MASC.PAST.IMPERF
kawą.
coffee.FEM.INSTR
‘The whole room smelt of coffee.’
b. W całym pokoju
in [whole room].MASC.LOC
pachniało kawą.
smell.3SG.NEUTR.PAST.IMPERF coffee.FEM.INSTR
‘The whole room smelt of coffee’
Lit.: ‘In the whole room it smelt of coffee.’
The situation in (19) differs form that in (14a–b) and (18a–b) because
kawa ‘coffee’ is not an instigator in the sense that bees or sequins are. No-
tice that the presence of coffee is not necessary at all; it is the smell of
coffee that is crucial. This is revealed by the awkwardness of the sentence
in (20): there coffee is treated as an instigator, so the contribution of the
locative phrase would be to suggest that the instigator had been distributed
all over it. In other words it would be coffee as a substance distributed all
over the room not its smell itself. Such an interpretation is evidently feasi-
ble but much less likely.
(20) ?Kawa pachniała w

coffee.FEM.NOM smell.3SG.FEM.PAST.IMPERF in
całym pokoju.
[whole room].MASC.LOC
?‘The coffee smelt in the whole room.’
Therefore (19b) resembles (12b) – the focus is on the state in a setting.

Because the setting is not a prominent element in such constructions, it can
be omitted, just like in (12a): Pachniało kawą ‘It smelt of coffee’ focuses
on the state alone and the location of that state can be more or less pre-
cisely inferred form the context (and/or general knowledge). In the imper-
sonal variant of the setting-subject construction, as shown in (18), the
omission of the prepositional phrase elaborating the setting/location is
278 Anna Słoń
more problematic: the setting/location is more prominent there than in

(19b) or (12b) because in (18) it is defined by the spatial extension of the
multiplex instigator. The omission would then have an anaphoric function.
Finally, the 3rd person singular neuter form of the verb as a marker of
impersonality features in the impersonal locative construction as exempli-
fied in (21c–d):
(21) a. Maria była dzisiaj w pracy.

Mary.FEM.NOM be.3SG.FEM.PAST today in work.FEM.LOC
‘Mary was/ has been at work today.’
b. Maria nie była dzisiaj w
Mary.FEM.NOM NEG-PART be.3SG.FEM.PAST today in
pracy.
work.FEM.LOC
‘Mary wasn’t/ hasn’t been at work today.’
c. Marii nie było dzisiaj w
Mary.FEM.GEN NEG-PART be.3SG.NEUTR.PAST today in
pracy.
work.FEM.LOC
‘Mary wasn’t/ hasn’t been at work today.’
Lit.: ‘(There) wasn’t Mary today at work.’
d. Marii było wszędzie pełno.
Mary.FEM.GEN be.3SG.NEUTR.PAST everywhere full.ADV
‘There was a lot of Mary everywhere.’
Lit.: ‘There was everywhere full of Mary.’
To explain the motivation for the use of the 3rd person singular neuter
for of the verb in (21c–d) as compared to the verb forms showing agree-
ment in (21a–b), we need to look at the contexts in which the two negated
constructions are used. Sentence (21b) – the personal negated locative con-
struction is used when the absence is viewed from the perspective of the
participant, while (21c) – the impersonal negated locative construction –
when the absence is viewed from the perspective of the location. Thus
Maria nie była dzisiaj w pracy bo ma urlop ‘Mary.NOM hasn’t been at
work today because she has a day off’ is natural when spoken at home by a
member of Mary’s family, for instance, whereas Marii nie było dzisiaj w
pracy bo ma urlop ‘Mary.GEN hasn’t been at work today because she has a
day off’ is natural when spoken at work by one of her colleagues. Similarly
the affirmative locatives: presence may be viewed from the perspective of

the participant, as in (21a), or from the perspective of the location, as in
(21d). The latter variant is in the situation at hand clearly marked both
emotionally and stylistically because it has to overcome the natural ten-
dency for single human participants to constitute a very good focus of at-
tention. The construction becomes much more natural though with multiple
participants (and in this sense has some affinity to the setting-subject re-
lated impersonal construction illustrated in [18]):
(22) Wszędzie było pełno ludzi.

everywhere be.3SGNEUTR.PAST full.ADV people.VIR.GEN
‘There were lots of people everywhere.’
Notice also that the preferred word order is the locative expression first.
Therefore the fact that the situation is viewed from the perspective of the
location motivates the use of the 3rd person singular neuter verb form.
The constructions analysed in this section show that the 3rd person sin-
gular neuter form of the verb is used when there is no participant elabo-
rated by a nominal in the nominative case and when the setting/location (if
elaborated) is elaborated by a prepositional phrase (or such forms as tutaj
‘here’, wszędzie ‘everywhere’, etc.). Thus the 3rd person singular neuter
form of the verb is evidently associated with the non-reified set-
ting/location of the event/state.
5. Conclusion
The ‘impersonal’ impersonal construction (i.e. the 3rd SG NEUTR construc-

tion) displays a great degree of variation. Unlike in the ‘personal’ imper-
sonals (more numerously represented in Polish), where the defocused in-
stigator is human, here the nature of the instigator is much more enigmatic.
The potential instigator is frequently diffuse and/or difficult to identify at
all, which makes it to some extent indistinguishable from the set-
ting/location. This is manifested formally by the use of the 3rd person sin-
gular neuter form of the verb. That this particular verb form is often em-
ployed to evoke the (non-reified) setting is further confirmed by the
analysis of related impersonal constructions, namely the setting-subject-
related impersonal construction and the impersonal locative construction.
The difficulty in the conceptualisation of the defocused instigator in the
3rd SG NEUTR construction is also demonstrated by the fact that the elabo-
280 Anna Słoń
rated trajector in the corresponding non-impersonal variants (where such

exist) may code the instigator, the instrument, the setting or an active zone
within the experiencer participant. Such a variety of construals available is
thus a clear indication of the unstable conceptual status of the defocused
instigator.
Notes
1. The data in the paper are in most cases made-up examples, based, however,
on real-life usage.
2. MM stands for “middle marker”. I follow here Kemmer, for whom “[a] middle
marker can be provisionally defined as a language specific morphosyntactic
marker that appears in the expression of some cluster of distinct situation
types (…) that are hypothesised to be semantically related to one another and
to fall within the semantic category of middle voice” (1993: 15). The Polish
middle marker się appears in the reflexive, reciprocal and related situation
types. By extension, it is also a marker of impersonality, as in (1b) (see Słoń
2001, 2003).
3. The term “instigator” used here refers to a causal factor that is not necessarily
human. For this reason the term is preferred over the term “agent”, especially
because the 3rd SG NEUTR construction is typically used to defocus non-human
instigators.
4. Attention-worthiness is related to newness of a particular item of information,
but it is not equivalent to it. While new information is clearly attention-
worthy, attention-worthiness may also refer to presenting an already men-
tioned item of information in a ‘new light’, so to say. In this interpretation,
topics are not attention-worthy as they constitute the background that contains
old information, necessary to keep the text coherent.
5. Pronouns can also be used for emphasis, which can be understood as an in-
stance of contrast with an element unexpressed linguistically.
6. Trajector in CG is defined as “the (primary) figure within a profiled relation”
(Langacker 1991: 555), while subject is “a nominal that elaborates the trajec-
tor of the process profiled at the clausal level of organisation. Its profile is
thus the primary clausal figure” (Langacker 1991: 554). It is necessary not to
confuse those two terms because the absence of subject does not imply the ab-
sence of trajector.
7. It must be noted that there is no sharp boundary between cases where the
identity of the trajector’s referent is fully recoverable despite the trajector’s
non-elaboration and cases where the identity of the trajector’s referent is so
diffuse that the trajector cannot really be elaborated at all. These two kinds of
situations constitute merely the endpoints of a continuum with a wide range of
intermediate cases. Thus, the greater the degree of irrecoverability of the iden-
tity of the trajector’s referent and/or the degree of conventionalisation of sub-
ject omission, the greater the likelihood of classifying the construction/ occur-
rence as impersonal. The picture gets more complex when context is taken
into account, because what is an instance of an impersonal construction in one
context, may be treated as an instance of zero anaphora in another.
8. Gender distinctions in Polish occur in the past tense and in the periphrastic
future tense with the so-called l-forms of content verbs. They do not occur in
the present tense, in the simple future tense and in the periphrastic future tense
with content verbs in the infinitive. This is illustrated by the sentences below.
Past tense singular:
(i) Jan (na)pisał list.
John.MASC.NOM write.3SG.MASC.PAST.(PERF)/IMPERF letter.MASC.ACC
‘John wrote a letter.’
(ii) Maria (na)pisała list.

Mary.FEM.NOM write.3SG.FEM.PAST.(PERF)/IMPERF letter.MASC.ACC
‘Mary wrote a letter.’
(iii) Dziecko (na)pisało list.

child.NEUTR.NOM write.3SG.NEUTR.PAST.(PERF)/IMPERF letter.MASC.ACC
‘The child wrote a letter.’
Present tense singular:
(iv) Jan/ Maria/ Dziecko pisze list.
John/ Mary/ child write.3SG.PRES letter.MASC.ACC
‘John/ Mary/ The child is writing a letter.’
Future tense singular:
(v) Jan/ Maria/ Dziecko napisze list.
John/ Mary/ child write.3SG.FUT letter.MASC.ACC
‘John/ Mary/ The child will write a letter.’
Periphrastic future tense singular (with the infinitive):
(vi) Jan/ Maria/ Dziecko będzie pisać list.
John/ Mary/ child be.3SG.FUT write.PERF.INF letter.MASC.ACC
‘John/ Mary/ The child will be writing a letter.’
Periphrastic future tense singular (with l-forms, which in form are like past
imperfective forms of the verb):
282 Anna Słoń
(vii) Jan będzie pisał list.

John.MASC.NOM be.3SG.FUT write.L-FORM.MASC letter.MASC.ACC
‘John will be writing a letter.’
(viii) Maria będzie pisała list.

Mary.FEM.NOM be.3SG.FUT write.L-FORM.FEM letter.MASC.ACC
‘Mary will be writing a letter.’
(ix) Dziecko będzie pisało list.

child.NEUTR.NOM be.3SG.FUT write.L-FORM.NEUTR letter.MASC.ACC
‘The child will be writing a letter.’
Similarly in the plural; there is, however, only the virile vs. non-virile gender
distinction.
9. Other instigator defocusing constructions include passives, middles and set-
ting-subject constructions (see Słoń 2002a, 2003).
10. This is why I call this construction ‘truly impersonal’ or ‘impersonal imper-
sonal’. In one sense, I use ‘impersonal’ to refer to a group of constructions in
which the trajector is obligatorily unelaborated (the ‘traditional’ sense; others
term such constructions as ‘person-less’, ‘nominative-less’ or ‘subject-less’,
where the last name is particularly infelicitous, as was explained in Section 2).
In the other sense, I use ‘impersonal’ to contrast with ‘personal’ meaning
‘human’. Therefore ‘impersonal impersonal’ refers to a construction in which
the trajector is obligatorily unelaborated and the trajector’s referent (however
vague its identity) is clearly non-human, or humanness is not at issue at all.
11. In most of the examples analysed here alternative word orders are possible.
Their contribution is either stylistic (e.g. rhythmic) or connected with the in-
formation structure of the sentence (whose discussion would require placing
these examples in context). As these considerations are less directly relevant
to the present analysis, they have been largely ignored in this paper. The ex-
amples are in the past tense because the gender distinctions are available then
(compare note 8).
12. This kind of metonymic relation is also present in setting subject construc-
tions, even though here the actual instigator is elaborated. In sentences (i) or
(ii) it is the bees and sequins respectively that are actual instigators, but be-
cause they are small and multiplex, the trajector status is conferred on their
location rather than on themselves.
(i) Ogród roił się od pszczół.

garden.MASC.NOM swarm.3SG.MASC.PAST.IMPERF MM from bee.PL.NON-
VIR.GEN
‘The garden was swarming with bees’
(ii) Suknia mieniła się cenkinami

dress.FEM.NOM glitter.3SG.FEM.PAST.IMPERF MM sequin.PL.NON-
VIR.INSTR
‘The dress was glittering with sequins’
There is an impersonal construction related to the setting-subject construction
illustrated above. It is briefly discussed in Section 4.
13. It is only in this metaphoric extra-meteorological context that the pronoun coś
‘something’ can be used. Coś is referential and Coś świtało must not be
treated as equivalent to English It was dawning (see below), but to Something
was dawning (on someone), meaning that someone was beginning to under-
stand something or getting some idea.
14. Notice that the cognate and pleonastic subjects are in this case masculine and
the verb agrees with them, so the verb is no longer neuter gender. Sentence (i)
is thus impossible:
(i) *Świt/ Dzień świtało.

dawn/day.MASC.NOM dawn.3SG.NEUTR.PAST.IMPERF
15. Doros (1975: 69) comments that this subconstruction is used in dialects or
literature rather than in everyday language and notes its lower frequency as
compared to Russian. In both languages the subconstruction is used with ref-
erence to forces of nature.
16. The identity of the instrument’s wielder can be predicted form the type of
process coded by the verb and the type of the instrument’s referent.
17. In this sense the 3rd person singular neuter marking of the verb may be consid-
ered to agree with the setting as Smith (1994) suggests for Russian imperson-
als.
18. Probably the original function of Polish się is a reflexive voice marker. In this
use it signals the co-reference of the direct object’s and subject’s referents, as
in (i):
(i) Marysia się uczesała.

Marysia.NOM herself/MM comb.3SG.FEM.PAST.PERF
‘Marysia has combed her hair.’
Here, się functions as a de-transitivising device. However, się appears also as
a marker of reciprocality, impersonality and middle voice (see Słoń 1999,
2001, 2003). As a middle voice marker, się is a de-causativising element.
Consider the following example:
(ii) Sukienka się pogniotła.

dress.FEM.NOM itself/MM crease.3SG.FEM.PAST.PERF
‘The dress got creased’
284 Anna Słoń
and compare it with the (implicitly) causative (i.e. transitive) version:
(iii) Marysia pogniotła (sobie) sukienkę.

Marysia.NOM crease.3SG.FEM.PAST.PERF (herself.DAT) dress.FEM.ACC
‘Marysia creased her dress.’
The function of the middle się is to narrow down the immediate scope of
predication, thereby excluding the causal element.
19. Although accessibility of the instigators that are defocused in the 3rd SG NEUTR
construction may seem problematic (it is true that they tend to be diffuse and
difficult to pin down), the defocused instigators of atmospheric impersonals
can have default reference (even though they are not very typical instigators
themselves).
20. Clearly it is possible to construct an ICM in which it would be perfectly natu-
ral to hold wizards responsible for meteorological conditions. However, this
usually requires explicitly establishing another mental space (see Fauconnier
1994) – a fairy tale mental space, in the case at hand.
21. Adverbs are particularly suitable to describe the state in/ of the setting because
they evolve out of the singular neuter form of corresponding adjectives.
22. According to Langacker (1999: 53–61), the prepositional phrase elaborating
the setting can be treated as the ‘search domain’ and can thus be potentially
construed as a thing. Apart from its being a non-prototypical nominal, it ap-
pears in a different kind of structure than a typical nominal would. Compare
(i) and (ii):
(i) W pokoju było zimno.
in room.MASC.LOC be.3SG.NEUTR.PAST cold .ADV
‘It was cold in the room’
(ii) Pokój był zimny.

room.MASC.NOM be.3SG.MASC.PAST cold.ADJ.SG.MASC.NOM
‘The room was cold.’
In sentence (xvi), the verb appears in the 3rd person singular neuter and is fol-
lowed by an adverb, while in (xvii), the verb agrees with the subject nominal,
as does the adjective which follows the verb. Moreover, w pokoju, is concep-
tualised as a setting and pokój as a participant (a patient).
23. For the details of this ‘personal’ impersonal construction see Słoń (2001,
2003).
24. Both constructions are analysed in Słoń (2003); the impersonal negated loca-
tive construction also in Słoń (2002b).
25. It is possible to use the verb roić się in the instigator-subject construction;
then, hovever, it no longer means ‘swarm’ but ‘leave the mother hive’.
26. Compare, however, note 22.
27. Note the oddness of (i) with sequins in the instrumental case:
(i) ?Na sukni mieniło się cekinami.

on dress.FEM.LOCglitter.3SG.NEUTR.PAST.IMPERF MM sequin.PL.NON-
VIR.INSTR
I cannot offer yet a satisfactory explanation why it is so.
References
Ariel, Mira
2000 The Development of Person Agreement Markers: From Pronoun to
Higher Accessibility Markers. In Usage-Based Models of Language,
Michael Barlow and Suzanne Kemmer (eds.), 197–260. Stanford:
CSLI Publications.
Bogusławski, Andrzej
1984 Polskie nieidentyfikacyjne wyrażenia osobowo-referencjalne [Polish
non-identifying personal-referential expressions]. Polonica 10: 49–
71.
Doros, Aleksander
1975 Werbalne konstrukcje bezosobowe w języku rosyjskim i polskim na
tle innych języków słowiańskich [Verbal impersonal constructions in
Russian and Polish as compared to other Slavic languages]. Prace
Komisji Słowianoznawstwa nr 32. Ossolineum.
Fauconnier, Gilles
1994 Mental Spaces: Aspects of Meaning Construction in Natural Lan-
guage. Cambridge: Cambridge University Press.
Grzegorczykowa, Renata
1996 Reguły deagentywizacji w polszczyźnie [Rules of deagentivisation in
Polish]. In Jazyk a jeho užívání, sborník k jubileu O. Uličneho [Lan-
guage and its use. Festschrift to O. Uličny], 58–65. Praha.
Jackiewicz, Alina
1992 Morphological Exponence of Impersonality with Special Reference
to Polish -no, -to Constructions and Their English and German
Equivalents. Linguistica Silesiana 14: 131–138.
Kemmer, Suzanne
1993 The Middle Voice. Amsterdam/Philadelphia: John Benjamins.
Kubiszyn-Mędrala, Zofia
1994 Problematyka konstrukcji bezmianownikowych w gramatyce genera-
tywno-transformacyjnej (na wybranych przykładach) [Problems of
nominativeless constructions in generative-transformative grammar
(on selected examples)]. Polonica 16: 163–179.
286 Anna Słoń
Lakoff, George
1987 Women, Fire, and Dangerous Things: What Categories Reveal about
the Mind. Chicago: University of Chicago Press.
1987 Foundations of Cognitive Grammar. Vol. 1: Theoretical Prerequi-
sites. Stanford: Stanford University Press.
1990 Concept, Image, and Symbol: The Cognitive Basis of Grammar
(Cognitive Linguistics Research 1). Berlin: Mouton de Gruyter.
1991 Foundations of Cognitive Grammar. Vol. 2: Descriptive Application.
Stanford: Stanford University Press.
1999 Grammar and Conceptualization. (Cognitive Linguistics Research
14.) Berlin/New York: Mouton de Gruyter.
Słoń, Anna
1999 A radial category of Polish się-constructions. Differences in con-
strual. In PASE Papers in Language Studies. Proceedings of the sev-
enth annual conference of the Polish Association for the Study of
English, Szczyrk, May 1998, Janusz Arabski (ed.), 151–166.
Katowice: Para.
2001 Konstrukcje nieosobowe a informacja pragmatyczna [Impersonal
construction and pragmatic information]. In Językoznawstwo
kognitywne II: Zjawiska pragmatyczne [Cognitive linguistics II:
pragmatic phenomena], Wojciech Kubiński and Danuta Stanulewicz
(eds.), 248–268. Gdańsk: Gdańsk University Press.
2002a Instigator Defocusing in Polish and English. A Cognitive Grammar
Analysis. In Proceedings of the 10th Annual Conference of the Pol-
ish Association for the Study of English. PASE Papers in Linguistics,
Translation and TEFL Methodology, Elżbieta Mańczak-Wohlfeld
(ed.), 81–93. Cracow: Jagiellonian University Press.
2002b Shifts in the relative prominence of participants, locations and set-
tings: Two negative constructions in Polish. In PASE Papers in Lan-
guage Studies: Proceedings of the Ninth Annual Conference of the
Polish Association for the Study of English. Gdańsk, 26–28 April
2000, Danuta Stanulewicz (ed.), 249–258. Gdańsk: Gdańsk Univer-
sity Press.
2003 Impersonal Constructions in English and Polish. A Cognitive
Grammar Approach. Ph.D. diss., Humanities Department, Maria Cu-
rie-Skłodowska University, Lublin.
Smith, Michael B.
1994 Agreement and Iconicity in Russian Impersonal Constructions. Cog-
nitive Linguistics 5 (1): 5–56.
2004 On Grammatical (Cross)Categorisation: Nominal Elements in Polish
Verbs – paper presented at Perspectives on Slavistics conference,
Leuven, 17th–19th September 2004.
Węgiel, Maria
1994 Czy w gramatyce transformacyjnej potrzebna jest reguła S Æ VP?
(Kwestia tzw. Czasowników zeroargumentowych) [Is there a need
for the S Æ VP rule in transformative grammar? (The problem of the
so-called zero-argument verbs)]. Polonica 16: 155–162.
Wolińska, Olga
1978 Konstrukcje bezmianownikowe we współczesnej polszczyźnie
[Nominativeless constructions in Modern Polish]. Katowice: Univer-
sity of Silesia Press.
ȱȱ
ȱǱȱ¢ȱȱ
A Frame Semantic account of morphosemantic
change: the case of Old Czech věřící
Mirjam Fried
Abstract
In this usage-based case study, I address the complex relationship between

form and meaning from a diachronic perspective, by examining the history
of the Old Czech ‘long’ present active participle věřící ‘(the one) believing’
in relation to the polysemous verb věřiti from which it is derived as part of
the inflectional paradigm. The verb was used in four major senses (BELIEF,
FAITH, TRUST, CREDIT). This general semantic structure is also attested
in the participle but the latter differs from the verb in its distribution across
the root meanings. I conclude that in explaining the relative survival rates of
individual uses it comes down to working out a relative equilibrium between
polysemy and isomorphism. The development of věřící is analyzed along
several dimensions: semantic shifts; syntactic behavior; pragmatic effects;
textual distribution; and chronological order. The result is a complex picture
of the factors that collectively motivate the layering and gradual crystalliza-
tion of the syntactic functions (reference vs. modification) and categorial
status of the form. The analysis is carried out within the framework of Con-
struction Grammar and Frame Semantics, since the interdependence
between lexical meaning and mosphosyntactic patterning and the subtle
changes in the mappings over time call for an approach that allows us to
represent meaning in terms of particular conventionalized patterns of under-
standing, in which lexical meaning, grammatical structure, and communica-
tive function form an integrated whole.
Keywords: morphosemantic change; categoriality; participles; Frame Se-

mantics; Construction Grammar
1. Introduction*
The Slavic field has always maintained strong interest in the study of lan-
guage change, and Slavic linguists have devoted substantial energy to de-
292 Mirjam Fried
veloping functionally oriented analyses of various historical phenomena,

traditionally focusing mostly on historical reconstruction. In recent dec-
ades, the interests of theoretically inclined linguists have shifted more to-
ward issues of formally oriented synchronic description and analysis, often
implying that historical phenomena cannot be of much use for a true under-
standing of grammatical structure, especially in light of the unavailability
of native speakers to prove or disprove a given theoretical proposal.
However, the emergence of grammaticalization theory (Traugott 1988,
1992; Heine et al. 1991; Hopper and Traugott 1993; Bybee et al. 1994;
Fischer et al. 2000; Haspelmath 1998, 1999; Bisang et al. 2004), prototype-
based historical semantics (Geeraerts 1983, 1992, 1997; Luraghi 2003), as
well as the recent focus on the value of old texts as valid sources of lin-
guistic data (Herring et al. 2000; Traugott and Dasher 2002) all have
(re)established the relevance of diachronic studies to our understanding of
the cognitive and communicative underpinnings of linguistic structure. It is
time, then, to align the traditional interests of Slavic linguists in diachronic
studies with the advances of various cognitively based theories, so that
both sides can enrich each other’s research, leading to deeper understand-
ing both of specific historical data and of the general cognitive and com-
municative processes that underlie language variation and change. The
present paper is intended as a step in that direction, introducing a particular
semantic theory of knowledge and understanding (known as Frame Seman-
tics) into the study of morphosemantic change as attested in a specific do-
main of Slavic morphology.
In particular, I examine the history of the Old Czech (OCz) word forms
věřící/věřúcí/věřijící, which in many ways illustrate the history of the so-
called ‘long’ present active participles in Czech. I will refer to these forms
as participial adjectives (PA), which is meant to reflect the categorial mis-
match between their internal and external morphology, i.e. an adjectival
CASE/NUMBER/GENDER suffix attached to a verbal (NT-participial)
stem, here labeled ppl;1 this structure is illustrated in (1), on the example of
věřiecí (later > věřící). The PA label is thus not intended as an indication
of the form’s functional status, which is much more complicated.
(1) [[V-root – theme – NT ]ppl – C/N/G]PA

[[ věř – ie – c ]ppl – í ]PA ‘(the one) believing’
The -NT- stem (i.e. present active participle) is derived from the present
(as opposed to past/infinitival) stem by adding the suffix -c, while the pre-
A Frame Semantic account of morphosemantic change 293
sent stem is formed by adding a thematic vowel to the root. The OCz ver-
bal system consisted of six conjugation classes and the thematic vowel was
-ú everywhere except in the i-stems, such as věřiti, which were marked
with -ie, as shown in (1). In addition, there was a highly productive sub-
class (the type kup-uj- ‘buy’) that represents an innovation in that it is
based on explicit aspectual morphology, deriving imperfective stems from
inherently perfective roots by adding -uj before the present stem suffix.
These patterns are summarized in the left portion of Table 1. The last two
columns of Table 1 illustrate cases in which the PAs deviate from the regu-
lar morphological template; following the terminology in Marvan et al.
(1963), I refer to those forms as pseudo-PAs. Their irregularity consists in
containing stem extensions (-ú, -ujú/-ijí) that belong to different morpho-
logical classes and as we can see, the verb věřiti involves both types of
irregular formations: the -ú suffix in the thematic vowel slot (deriving
věřúc-í instead of the expected věřiec-í) as well as the sequence -uj-ú- (de-
riving věřijící). It is also worth pointing out that pseudo-PAs often (though
not always) had a distinctly shifted meaning from the regular PAs, some-
times involving metaphorical shifts but by no means limited to those (cf.
Marvan et al. 1963; Fried 2003, 2005b).2
Table 1. Morphological patterns of OCz PAs and pseudo-PAs
Gloss Present stem -NT- CNG Pseudo-PAs

‘grow-ing’ rost –ú –c –… narost'-ujúc-
‘pass-ing’ min –ú –c –… min-ujúc-
‘pour-ing’ lej –ú –c –…
‘believe-ing’ věř – ie –c –… věř-úc- věř-ujúc-
‘call-ing’ volaj –ú –c –… (-ajúc > -úc )
‘buy-ing’ kup-uj –ú –c –…
Finally, with respect to věřící/věřúcí/věřijící in particular, a brief com-

ment is in order concerning certain orthographic practices in the manu-
scripts. I will continue using the Modern Czech (MCz) spelling věřící,
unless a given token is actually spelled differently in the manuscript itself.
However, using the modern spelling indiscriminately may not always be
historically accurate since it obliterates any differentiation between the
historically regular form věřiecí (> věřící) and the pseudo-form věřúcí (>
věřící). The problem is that the line differentiating between věřiecí and
věřúcí is a blurry one, since the distinct phonetic realizations -ie- and -ú-
became, at different chronological points, neutralized through regular
294 Mirjam Fried
sound change (first ú > í and later ie > í), the result of which was, in both
cases, spelled i and y in different texts. Consequently, we cannot be always
sure what the orthography in a given token represents etymologically (-ie-
vs. -ú-) after a certain point on the timeline.3 In contrast, the form věřijící is
a straightforward reflex of the u > i fronting (< věřujúcí).
The PAs offer valuable material that speaks to a number of questions of
general interest: the developmental potential of the hybrid form itself,
given the inherent tension between its internally verbal but externally
nominal character; the role of the communicative context in the PA’s inter-
pretation and function; the relative (in)stability in maintaining particular
form-meaning associations; and the role of collocations and fixed expres-
sions in semantic change. This case study focuses mainly on the first two
questions, by examining in detail the semantic relationship between the
polysemous verb věřiti and its corresponding PAs. However, the PAs’ se-
mantic development is also inextricably linked with their place and func-
tion in larger grammatical patterns and correlates with their textual distri-
bution as well. Only a systematic analysis of these interconnected
dimensions can lead to a comprehensive picture of all the factors that col-
lectively motivate the gradual shifts in the forms’ meaning and categorial
status, starting with a richly polysemous structure that spans several se-
mantic domains and three functional domains (referential, attributive,
predicative), and leads to the outcome we know from MCz: věřící as an
actor noun (‘practicing Christian’) and marginally as an adjective with the
same meaning. The analysis will demonstrate that the observed changes
can be best described in terms of Hopper and Thompson’s 1984 idea of the
promotion and demotion of particular aspects of meaning; in the case of
PAs these aspects have to do with the relative weight of their verbal and
nominal properties.
The nature of the data and the analytic goals – namely, capturing the
speakers’ knowledge of the meaning of a complex and functionally some-
what fluid morphological category – support the view that linguistic cate-
gories are best understood as functional prototypes (e.g. Daneš 1966; Hop-
per and Thompson 1985; Croft 1991, 2001) and call for an approach that
allows us (i) to represent meaning in terms of particular conventionalized
patterns of understanding, in which lexical meaning, syntactic function,
and communicative function form an integrated whole, and (ii) to structure
these patterns in a polysemy network that captures their mutual relatedness
and motivations for shifts. I will, therefore, appeal to the conceptual and
representational apparatus of the grammatical model known as Construc-
tion Grammar (esp. Fillmore 1989; Croft 2001; Fried and Östman 2004)
and its own theory of meaning, Frame Semantics (Fillmore 1982; Fillmore
and Atkins 1992; Atkins 1994; Atkins et al. 2003; Fillmore et al. 2003;
Fried and Östman 2003; Fried 2004, 2005a).4
In Frame Semantics, linguistically relevant semantic information is or-
ganized and structured in “interpretive frames” (Fillmore 1982: 122),
which represent the complete background scene associated with a given
linguistic expression: the scene’s participants, settings, and any other
unique semantic features (collectively referred to as “frame elements”) that
are necessary for speakers’ native understanding of what the lexical item
means and how it can be used in context. In the case of predicates, i.e. ar-
gument-taking lexemes, the frame also carries information about the con-
ventional expression of the syntactically relevant participants as they mani-
fest themselves in the syntactic organization of sentences. A single
linguistic expression may be (and often is) associated with multiple frames
and, conversely, a single frame may be shared by multiple expressions;
each such expression, then, represents a particular conceptualization of
certain parts of the larger background scene. Both aspects of frame sharing
are among the factors that have shaped the diachronic path of the PA
věřící.
In addition, the PA provides a good test case for one of the basic tenets
of Construction Grammar, which holds that the meaning or function of a
grammatical pattern is not just a sum of its parts, derivable compositionally
from the properties of its constituents. The PA has a particular meaning at
the word-level, as a symbolic sign that contributes semantic content to a
larger syntactic pattern it occurs in; at the same time, it is internally com-
plex, consisting of a sequence of inflectional morphemes attached to a
lexical root. This arrangement necessarily raises the question of how the
two dimensions are related: Is the overall meaning of the PA predictable
from its morphosemantic structure, or is it non-compositional, and in what
ways? The present analysis will show that the PA is indeed best treated as
a morphological construction as it is understood in Construction Grammar:
a conventionalized association between a complex form and its meaning.
Taking a diachronic perspective enhances our insight into how such a con-
ventionalization may arise.
The analysis is based on an extensive corpus of authentic data excerpted
manually from OCz texts that stretch across more than 300 years and pro-
vide a representative sample of genres (historical, biblical, administrative,
expository, poetry, popular entertainment, correspondence, drama, etc.).
My corpus contains 121 tokens of the PAs derived from věřiti and they
come from over 50 different texts. I excerpted about 74 different texts in
296 Mirjam Fried
their entirety, which has yielded more than 55% of the PA tokens in the
corpus. The remaining 45% come from about 120 additional texts (about
one quarter of them biblical) and were collected more or less at random
from the OCz archive at ÚJČ in Prague. The corpus contains more than
1200 tokens of PAs, which represent over 240 different verb roots, and
about one third of those roots show both a regular form and at least one
pseudo-PA. All three PA variants derived from the verb věřiti are amply
attested throughout the OCz period; this is not surprising, given the mean-
ings the verb expresses: as we shall see, they all have to do with salient
events of the speakers’ daily life. This fact together with the PAs’ rela-
tively high frequency across very diverse manuscripts is evidence that they
were well entrenched in the vernacular.
The paper is organized as follows. Section 2 summarizes the semantic
structure of the verb věřiti, establishing a basis for analyzing the meanings
of the PAs. Section 3 presents the details of the PA uses both as adnominal
modifiers and as syntactic nouns, incorporating them into the network of
frames associated with the verb root. Section 4 provides the overall dia-
chronic structure of the PA, including its irregular forms. Section 5 briefly
concludes the paper.
2. The semantics of the verb věřiti
The verb věřiti was used in several senses, which I label as BELIEF, FAITH,
CREDIT, and TRUST. They all can be described as having to do with the con-
cept of belief (accepting something as truthful or reliable) but they repre-
sent four distinct conceptualizations, which are manifested by different
complementation patterns and are motivated by the contexts in which they
are used.
The conceptually simplest scenario (BELIEF) is illustrated in (2). It pre-
supposes a person who believes (Believer) and a person who receives the
Believer’s trust (Recipient of Trust) with respect to some Content. In OCz,
the conventional way of encoding this semantic structure was to mark the
Believer in the nominative case, the Recipient in the dative, and the Con-
tent, which may or may not be explicitly mentioned, in various preposi-
tional phrases, often a locative – ‘to believe somebody (in/about some mat-
ter)’.
(2) a. Proč mi v tom nechceš věřiti?

why 1SG.DAT in that.LOC.SG NEG.want.PRES.2SG believe.INF5
[BawArn 79a; 1380; entertainment] 6
‘(It would harm my honor if I spoke lies.) Why do you not want
to believe me in that [i.e., in what I am saying]?’ 7
b. druh druhu málo věří

comrade.NOM.SG.M comrade.DAT.SG.M little believe.PRES.3SG
[AlxB 338; 1350; epic poetry]
‘comrade [but] little believes another comrade’
A distinctly different pattern is presented in (3). It involves a two-

participant scenario, where the second participant is in the accusative:
plain accusative in the oldest layer (3a) and with the preposition v ‘in’ (in
the directional, as opposed to static, sense) throughout OCz (3b) and also
in MCz.
(3) a. kteřít’ buoh věřieše

which.NOM.PL.M.PP god.ACC.SG believe.PAST.3SG
[Vít 40b; 1380; religious prose and poetry]
‘[those] who believed in God (got rid of all [their] pain
right away)’
b. {počet’ s ní sladce mluviti}

byt’ se modlám klaněla i
COND.PP RF idol.DAT.PL.F bow.PPL.SG.F and
v ně věřila
in 3PL.ACC believe.PPL.SG.F
[LyrVil 157,70 (Doroto, panno čistá), late 1300s; spiritual poetry]
‘{he tried to sweet-talk her into} bowing to idols and believing in
them’
Conceptually, the FAITH sense represents a fairly straightforward exten-

sion of BELIEF: its content-of-thought reading is cast strictly in the narrow
religious context, where the Believer is understood as a holder of the belief
in a deity(-like entity). Since this kind of belief is by definition absolute,
that is, the Content of belief is beyond any confusion or choice out of mul-
tiple possibilities, it also makes pragmatic sense that this participant is
never expressed. In some sense, the Recipient of Trust and the Content of
298 Mirjam Fried
belief are fused into a single entity, which is conventionally marked by the
(v +) Acc phrase. FAITH thus expresses a more restricted instance of the
concept of believing.
The examples in (4) illustrate the CREDIT sense, where the concept of be-
lieving somebody’s word is set in the context of a commercial transaction,
whereby one party (seller) provides a desired commodity to another
(buyer), in exchange for a payment. But the scene expressed by věřiti in the
commercial sense is more complex than that since the payment presup-
posed by a commercial transaction is necessarily missing here.
(4) a. {za kup’u ... } jemu samému věřil zač

3SG.M.DAT self.DAT.SG.M credit.PPL.SG.M for
kolivěk dlužen
whatever in.debt
[Rožmb 118; 1300–1350; oldest known record of OCz legal procedures]
‘hei gave him(self)j on credit whatever hej owed {for a purchase
or food}’
b. {jal se předivně kupčiti, bera draze na úvěrky...;}

což kto jedno chtěl věřiti
what.ACC who.NOM one want.PPL.SG.M credit.INF
[Budyš 30b; 1420; social satire]
‘{he [=king] started very strange trading, taking on expensive
credit…;} whatever who[ever] wanted to give on credit, (he
took it all)’
The seller (cast as a Creditor) is asked to simply believe that a payment

is forthcoming (under conditions the two parties agree upon), which leads
to the interpretation ‘give/sell something to somebody on the belief that the
buyer is honest about paying later’. The buyer is thus cast as a Debtor and
the belief has to do not so much with the content of someone’s thoughts
but, rather, with an expectation of an action at a later point. The comple-
mentation pattern associated with the CREDIT sense is the same as with
verbs of giving, reflecting a conceptual overlap between these two domains
as well: the Commodity is a direct object (as a given thing) and the new
owner (Debtor) is marked by the dative, as a recipient.
These three senses of věřiti are attested in the very oldest texts but
TRUST seems to have emerged somewhat later, as an additional extension of
the BELIEF sense and possibly motivated by the existence of the CREDIT
sense as well. The TRUST sense is exemplified in (5) and we note that it still
has three arguments (nominative-, dative-, and genitive-marked).
(5) a. já jich milosti toho úplně

1SG.NOM Their Excellency.DAT that.GEN.SG.N entirely
věřím
trust.PRES.1SG
[ArchČ 9,316; 1461; a will]
‘I completely trust Their Excellency with that [arrangement,
i.e., overseeing the protection of my widow]’
b. {aby o tom s ním bylo mluveno}

že mu i mého hrdla
that 3SG.M.DAT also my.GEN.SG.N throat.GEN.SG.N
nevěříte8
NEG.trust.PRES.2PL
[Výbor II/2; 1449; family correspondence]
‘{that it be discussed with himi [= my husband]} that you don’t
trust himi as regards my own life either’
Here the Recipient of Trust plays the role of a caretaker who is en-
trusted with a valued object and the Believer is thereby cast in the role of a
Dependent: somebody who depends on the trustworthiness of the care-
taker. In this respect, the TRUST sense shares the semantic shift that is pre-
sent in the CREDIT sense in highlighting the belief as expectation for an
action, rather than relating simply to the content of somebody’s thoughts.
In the oldest examples, this difference in meaning between BELIEF and
TRUST is consistently signaled by different case marking (Nom-Dat-Loc vs.
Nom-Dat-Gen), and the subsequent development only strengthens the se-
mantic contrast. In the BELIEF sense, the formal alternatives for expressing
the Content role are gradually expanded to a content clause introduced by
že ‘that’ (6a), whereas the corresponding role in the TRUST sense (Valued
Entity) is attested in the form of a purpose clause introduced by aby ‘in
order to/so that’ (6b).
300 Mirjam Fried
(6) a. já také věři že páni i také

1SG.NOM also believe.PRES.1SG that lord.NOM.PL and also
rytieři {… u dvora dobrú mají}
knight.NOM.PL
[Podk 59; around 1400; social satire]
‘I do believe that lords and knights {… live well at the court}’
b. a vám věřím, aby se tak k

and 2PL.DAT trust.PRES.1SG lest RF like.this toward
tomu měli aby se tak stalo
that.DAT.SG.N have.PPL.PL lest RF thus happen.PPL.SG.N
[ArchČ 8,30; 1471; pers. corresp.]
(a complaint about property damage caused by soldiers)
‘and I trust you [to see to it] that they set out to [do] it [=re-
pair] like this, that it will be done’
The chronological distribution of the four senses is summarized in Fig-

ure 1. The numbers across the top of the diagram indicate years on the
timeline, marking particular chronological layers.9 The bullets indicate the
first occurrence of a given meaning in the corpus and the usages that have
survived into the modern language are indicated by the arrows extending
into the MCz column; those lines that end at the 1500 boundary indicate
usages that are not present in MCz but the diagram says nothing about
when exactly (after 1500) those senses faded out.
1350 1380 1410 1450 1500 ... MCz

v LOC
BELIEF
že-S
CREDIT
FAITH
GEN
TRUST
aby-S
Figure 1. The chronological distribution of the OCz verb věřiti
The overall semantic structure, sketched in Figure 2, constitutes a

polysemy network of frames that are related through the general notion of
believing something to be true. That concept is expressed most directly in

the BELIEF sense, which can thus be considered the prototypical meaning
(indicated by the thick-line rectangle), while the remaining senses repre-
sent different conceptualizations of that frame, motivated by different
communicative contexts. These relationships are captured by the arrows
from the prototype toward the extensions; the dashed line between the
TRUST and CREDIT senses indicates the possibility that the emergence of the
TRUST sense might have been further facilitated or reinforced by its seman-
tic similarity to the CREDIT meaning and is not just an independent exten-
sion of the prototypical sense. Following the practice in Frame Semantic
representations (Fillmore et al. 2003), the frame names will always be
shown in a different font (BELIEF, TRUST, etc.) in order to indicate that these
labels represent abstract concepts, not particular linguistic instantiations;
they should not be understood as corresponding to the meaning of actual
words (or word forms) of the OCz lexicon. The reason for this practice is
to keep it in focus that the individual linguistic forms (whether finite forms
or PAs) all share a particular conceptual space associated with the root.
For purposes of clearer exposition, the diagram uses a stripped-down
version of the formalism that would be used in full Frame Semantic repre-
sentations. The information that is crucial to the issues discussed in this
paper includes the lexical category (cat) of the lexeme (lxm), in this case a
verb (cat v); the number and type of syntactically relevant frame partici-
pants (frame elements, FEs); and, through the valence (val) attribute, the
conventional associations between those participants and their morpho-
logical marking in an active sentence. Notice that the names of the partici-
pants change from frame to frame, reflecting the semantic modulations
inherent in each sense; however, the numerical indexing at the FE level
(e.g. #1 Believer, #1 Dependent, #1 Creditor, etc.) can be taken as a typo-
graphical shortcut for indicating that the co-indexed entities are members
of the same conceptual category. The gray rounded rectangles indicate
frames that are not inherently associated with the notion of believing but
overlap with specific uses of the verb věřiti, marking the relationship be-
tween věřiti and lexemes that belong to other lexical domains but share
with věřiti certain features, features which may be both semantic and for-
mal (e.g. specific case marking patterns for their complements or interpre-
tation of the event participants).
302 Mirjam Fried
(ex. 3)
cat v lxm věřiti
cat v lxm věřiti
Frame TRUST
(belief as an expectation) Frame FAITH
FEs: #1 Dependent (content of thought)
#2 Caretaker FEs: #1 Believer
#3 Valued-Entity #2 Deity
val {#1 [Nom], #2 [Dat], #3 [Gen] } val {#1 [Nom], #2 [Acc(v) ] }
(ex. 5, 6b)
[aby-S] cat v lxm věřiti
Frame BELIEF (content of thought)
FEs: #1 Believer
#2 Recipient-of-Trust
(ex. 4) #3 Content
cat v lxm věřiti val {#1 [Nom], #2 [Dat], (#3 [Locv]) }
Frame CREDIT (ex. 2, 6a)

[že-S]
(belief as an expectation)
FEs: #1 Creditor Frame COMMERCIAL
#2 Debtor TRANSACTION
#3 Commodity
FEs:
val {#1 [Nom], #2 [Dat], #3 [Acc] } Bueyr, Seller, Goods,
Cost, Tender...
Frame [ TRANSFER ]
FEs: #1 Giver
#2 Gift
#3 Recipient
Figure 2. The lexico-grammatical structure of the OCz verb věřiti
In the next section, we will examine how the verb meaning manifests it-
self in the PA form, focusing on the following questions: Which of the
meanings occur? Are there any correlations between the verb meanings and
the categorial status of the PA? To what extent can we detect semantic
shifts vis-à-vis a compositional interpretation of the PAs? And, finally,
does the regular vs. irregular morphology of the participial stem play a role
in any semantic distinctions?
3. PA functions
The PA occurs in all three syntactic environments we find with other OCz
PAs: in an adnominal position (7); as a syntactic noun, in (8) shown as a
dative-marked argument with a possessive modifier; and in a copular struc-
ture with být ‘be’(9). In the space of this paper, I can only address the
nominal and adnominal patterns, which, however, represent the vast major-
ity of attested uses (111 tokens out of the total of 121).
(7) lid v buoh věřící spasen

people.NOM.SG.M in god.ACC.SG believe.PA.NOM. SG save.PASS.SG.M
byl
be.PPL.SG.M
[PasMuzA 488a; mid 1300s; legend]
‘the people that believed in God were saved’
(8) Ducha Svatého dává v nich svým

Spirit.ACC Holy.ACC gives.PRES.3SG in 3PL.LOC his.DAT.PL
věřícím
believe.PA.DAT.PL
[ŠtítVyš 84b2; 1396; religious tract]
‘in them [=gifts], [Christ] gives the Holy Spirit to his believers’
(9) jsúce ... poslušní zjevení božieho

be.PART.PL obedient.NOM.PL.M revelation.GEN.SG divine.GEN.SG
a věřící slovu jeho
and believe.PA.NOM.PL word.DAT.SG his
[AktaBratr I 107b, 1470; expository prose]
‘while following the divine revelation and believing his word’
As already noted, the PAs combine properties that predispose them to-
ward both verbal and non-verbal behavior. Their verbal root creates the
potential for expressing verbal valence; the -NT- morphology explicitly
marks voice of the participial stem (encoding its inherently active orienta-
tion, cf. Haspelmath 1996) and relative tense, expressing an eventuality
that is temporally delimited by the main predicate (in the sense of contem-
poraneousness; cf. also Schultze-Berndt and Himmelmann 2004).10 At the
same time, the nominal inflection (CNG) makes the PAs externally non-
verbal entities: functionally they could be actor nouns (capitalizing on the
304 Mirjam Fried
potential of the PA as a ‘definite participle’, i.e. ‘the one V-ing’), and mor-
phologically they are unambiguously adjectives. By strengthening or sup-
pressing any of these a priori available features (nominal, adjectival, ver-
bal), a given PA can potentially develop any of the corresponding
functions: reference, modification, or predication. However, each function
displays a distinct preference with respect to a particular root meaning;
Table 2, as a preview of the subsequent discussion, summarizes the relative
frequencies. The number in parentheses refers to the purpose meaning,
shown in (12) below, which may or may not be an instance of PA, and the
copular uses (bottom row) are added for the sake of completeness only;
they require further commentary that is beyond the scope of this paper. It
must be stressed, though, that these numbers can only be used for their
orientational value. Since we are dealing with an incompletely attested
language statistical references are only useful as indicators of general ten-
dencies, not in any absolute terms.
Table 2. Relative frequencies of věřící in the corpus vis-à-vis the verb senses
BELIEF FAITH CREDIT TRUST
Adnominal PAs 6 (+ 26?) 13 -- --
PAs as syntactic nouns -- 46 20 --
PAs in (copular) predicates 6 4 -- --
The modification function in general subsumes several semantic devel-

opments, most of which are attested with the PAs derived from věřiti as
well. For easier orientation, Table 3 provides an overview of the meanings
found with věřící and the labels I will use for them.
Table 3. Meanings found with adnominal věřící in the corpus

Labels Definitions Examples
habitual ‘prone to believing’ (7)
passive/modal ‘believable, credible’ (11)
purpose ‘for (the purpose of) believing’ (12)
classificatory ‘faithful, devoted’ (15, 16)
As we shall see, they all have to do with the degree of manipulation of

the internal morphosemantic structure of the PA form and with moving
away from the compositional meaning of the PA as a semantically trans-
parent inflectional form: ‘[one] who V-s [at the time of the main event]’.
3.1. Adnominal PA uses
The expected structure and the most transparent meaning of the adnominal
PA would be a combination of a noun expressing the ‘agent’ of the PA
(Believer) and the PA as an adnominal modifier, as shown in (7), here re-
peated as (10); for easier identification of the syntactic structure, the PA
with its complements, if present, will always be enclosed in brackets <>:
(10) lid <v buoh věřící > ‘(the) people believing in God’
The adnominal PAs occur only in the BELIEF and FAITH senses. In the BE-
LIEF meaning, the PA is always bare and appears to be older. Interestingly,
even the earliest example, given in (11), shows a clearly shifted meaning:
the modified entity is not the Believer, as would be expected, but the Con-
tent, resulting in a distinctly passive/modal reading: ‘believable’, rather
than ‘believing’; structurally, the PA in (11) is a resultative secondary
predicate. Note also that this token must be classified as a pseudo-PA, and
not just by virtue of its age (the change ie > í had not taken place yet): the
manuscript spelling (-úcí) sometimes indicates the irregularity directly, as
in (11).
(11) Upravenie tvá < věřúcí >

testimony.NOM.PL.N your.NOM.PL.N believe.PA.NOM.PL
učiněna jsú velmi
make.PASS.PL.N AUX.3PL very
[ŽaltWittb 92,5; early 1300s; biblical] (Lat. credibilia) ‘credible’;
later variant: věři-dlná ‘believable’ in ŽaltKap)
‘your testimony is made very credible’
This usage is significant for two reasons. First, it follows the pattern
found with many pseudo-PAs, namely, a striking semantic shift; in this
case the meaning reflects a reconfiguration of the semantic participants
contributed by the verbal stem. In particular, the reconfiguration can be
attributed to the loss of the verbal category of voice, whereby the PA be-
comes fully dissociated from its active orientation signaled by the -NT-
morphology: the noun modified by the PA no longer bears the expected
agent relation to the event expressed by the root. All this could be taken as
evidence that the sequence -ú-c-í found in this pseudo-PA and imported
from another conjugation class had become a (partially?) frozen unit with-
out a transparent internal structure (Gebauer 1958: 89–92) and was at-
306 Mirjam Fried
tached to the root věř- as a single (derivational) suffix -úcí (we will have to
qualify this conclusion somewhat in section 4).11
The second point of interest is the kind of nouns this pseudo-PA collo-
cates with: upravenie ‘testimony’ in (11), svědek ‘witness’, etc. The gen-
eral context of testifying is common to all the other adnominal examples of
věřící in the BELIEF sense, whether used in the modal meaning ‘believable’
or the purpose meaning to be discussed below. It suggests that, unlike the
verb věřiti, which occurred in all kinds of contexts in which believing the
truthfulness of something was at issue, the adnominal PA was used more
narrowly and in a more specialized sense, largely restricted to situations
which require or presuppose the act of witnessing as a precondition for the
belief.
The PAs in the BELIEF sense appear to be fairly marginal. The modal
meaning (3 tokens in my corpus) disappears by the mid- to late 1400s, but
despite the small number of examples in the corpus, we cannot dismiss
these pseudo-PAs simply as scribal errors or a quirk of an individual
author, since the modal (and other non-active) interpretations are amply
documented by PAs formed from other verbs (Fried 2005b, forthcoming-b)
and the form věřící thus fits a more general pattern. What is perhaps more
surprising is the fact that the active meaning, in the habitual sense of
‘(prone to) believing’, seems to be equally rare, attested from about 1400
on (3 tokens). It is completely marginal in MCz as well (no tokens in the
800,000-word spoken Czech corpus, PMK, and only 20 in the
100,000,000-word corpus of written Czech, SYN2000), although it is of
course possible to form the PA; its formation remains morphologically
fully productive, as expected for an inflectional form. I will return to its
low incidence in actual texts at the end of the paper.
The corpus also contains a relatively large number of examples (26 to-
kens) involving one particular collocation: with the noun list ‘letter’, such
as in (12); for reasons to be addressed in a moment, I present this PA in the
actual OCz spelling.
(12) a. Tuto se vypisuje list ten

here RF write.PRES.3SG letter.NOM.SG.M that.NOM.SG.M
< wierzicij >
believe.PA.NOM.SG
[TovHád 55a; 1467; political alegorical satire]
‘Here is the letter of authorization/authentication’
b. piš napřed ty, komuž de

write.IMP first 2SG.NOM who.DAT.SG.PP go.PRES.3SG
kredencie neb < wierziczi > list
verification.NOM.SG.F or believe.PA.NOM.SG letter.NOM.SG.M
[ProkArs 176; late 1400s; rules for letter-writing]
‘you first [have to] write to whom[ever] the verification
document, or letter of authentication, is being sent’
This collocation appears to be used as a technical term that generally

fits in with the context of testifying. It does not mean ‘a letter that be-
lieves’ nor simply ‘believable letter’ but, rather, denotes a particular kind
of letter, defined by its purpose: it is to testify to the believabil-
ity/trustworthiness of a messenger or a document carried by the messen-
ger. The meaning is best glossed as ‘verifying/authenticating’ and the noun
list ‘letter’ is interpreted as a ‘witness’ to the trustworthiness of a messen-
ger who is only a proxy for the real Recipient of Trust. This is illustrated
nicely in an example that actually gives an explication of this meaning,
involving the verb věřiti itself:
(13) {Posieláme k vám ... Otíka z Račína..., jemuž o některé věci

s vámi mluviti poručili jsme, což nynie v tomto poselstvie
jménem naším s vámi mluviti bude, že}
jemu všeho jako nám samým úplně ...
3SG.M.DAT all.GEN.SG as 1PL.DAT self.DAT.PL entirely
věřiti budete
believe.INF be.FUT.2PL
[ArchČ 16, 12; 1491; pers. corresp.]
‘{We’re sending to you Otík from R., who we’ve ordered to talk
with you about a certain issue, and whatever he will be discuss-
ing with you, in our name, as part of this mission,} you will
believe him in everything the way you would [believe] us.’
It is not clear, though, that the forms in (12) are necessarily PAs. Since
the spelling in the manuscripts does not indicate length, we can only guess
whether the form in question represents a PA (long í, etymologically either
< ú or < ie), or a specific deverbal adjective dedicated to expressing pur-
pose and derived from the infitival stem + cí (in which case the vowel pre-
ceding -c- would be short), as shown in (14b):
308 Mirjam Fried
(14) a. [věř-í-] c-íPA ‘believing’

b. [věř-i-] cí ‘for believing’
Both analyses are plausible. What speaks in favor of (14b) is the pur-
pose meaning. On the other hand, the analytic uncertainty is peculiar to the
i-stems only since, without marking the length orthographically, we cannot
easily tell the difference between the present and infinitival stems (věří- vs.
věři-, respectively). With other stems, however, the distinction is unambi-
guous (cf. laj-ú-c-íPA vs. lá-cí from lá-ti ‘to reprimand’) and there is plenty
of evidence that the morphologically indisputable PAs were quite com-
monly used in the purpose reading as well (cf. Michálek 1963; Fried
2005b).
We can leave this question concerning věřící/věřicí open since nothing
in the overall account of the PA development hinges on resolving it. Let us
simply note that it is possible that the BELIEF sense of věřiti might have
been extended in the adnominally used PA also in the purpose direction, as
part of a generally available strategy (and one attested with other PAs) in
supressing the inherently active orientation of the PA.
Examples of the FAITH sense, shown in (10) above (lid v buoh věřící
‘people believing in God’), are first attested in the late 1300s and as is
apparent from Table 2, they are not very frequent either, especially in
comparison to the corresponding syntactic noun, to be discussed in section
3.2. The noun modified by the PA is typically lidi/lid ‘people’ or člověk
‘person’, and the generic meaning of these nouns also renders the colloca-
tion semantically equivalent to the syntactic nouns. In contrast to adnomi-
nal PAs in the BELIEF sense, the FAITH PAs sometimes express the non-
subject complement (Deity), especially in the oldest examples, as in (10).
This complement, however, does not provide a highly informative contri-
bution: the referent is always God/Christ, expressed either directly (buoh
‘god’, jezukristus ‘Christ’) or by a reflexive pronoun. This pattern suggests
pragmatic ‘emptying’ of the complement, whose presence thus serves
merely as a placeholder in the valence of the stem, not as an expression
identifying a pragmatically unpredictable, novel participant in the reported
event. As a result, the non-verbal potential of the PA is strengthened, de-
spite the presence of the root’s valence, and these PAs generally cast Chris-
tian faith as a property that characterizes a particular person or group of
people (those who can be classified as ‘having Christian faith’). Similarly,
lack of this property helps identify its opposite, as illustrated in (15).
(15) {pomsti svatej Kateřiny...}

nad < nevěřúcím > pohanstvem
over NEG.believe.PA.INS.SG pagandom.INS.SG
[LegKat 57a; late 1300s; spiritual poetry]
‘{avenge St. Catherine} against the pagans, who are without faith’
In many cases we see a clear shift towards a fully atemporal, classifica-

tory reading, extending the PA in the direction of ‘faithful, devoted’ as an
inherent quality. This reading is semantically supported especially by the
occasional collocation with křest’an ‘Christian’ (16), which otherwise
would seem redundant or even contradictory.12 Note also the syntax: the
bare PA in (16) is in the slot of an NP-internal modifier.
(16) naučenie všem < věřícím > křest’anuom

lesson.NOM.SG.N all.DAT.PL believe.PA.DAT.PL Christian.DAT.PL
budoucím
be.FUT.PA.DAT.PL
[AktaBratr I 196b; 1469/70; expository prose]
‘[this is] a lesson for all faithful Christians in the future’
In a further contrast to the BELIEF PAs, the FAITH examples are poten-
tially interesting with respect to the relative order of the PA and the noun it
modifies: while the tokens with complements, as in (10), show the linear
arrangement frequently found with the PAs in general ([NP – [comple-
ments – PA]]), the bare tokens, including the uses in the oldest layer (be-
fore 1380), tend to show the order in which the PA precedes the noun it
modifies, such as in (15)–(16). This can be taken as another indicator of a
gradual shift toward using the PA in a modifying function, for which the
Mod-N order was slowly establishing itself across all types of modifiers as
the only grammatical option (a more detailed discussion of the word order
and its role in the PA diachronic development can be found in Fried 2003
and forthcoming-b). Granted, there are also examples of the N-PA order,
but they do not seem to extend beyond mid-15th century and they all come
from biblical or religious texts, in which the N-PA order remained as a
genre-specific feature long after the original functional distinction between
contrastive and non-contrastive modifier became neutralized (cf. Kurz
1958). We can thus conclude that the syntactic patterning correlates with
the functional and semantic development. All these findings are arranged
chronologically in Figure 3. The lines with the bullets indicate the overall
presence of the PA in a given verb sense, the thin lines mark the presence
310 Mirjam Fried
of specific syntactic features and distinct meanings, and the dashed lines
indicate relatively low frequency in the corpus.
1350 1380 1410 1450 1500 ... MCz
BELIEF
meaning (habitual)
'credible'
'authenticating' ?
FAITH
Mod-Head order
(habitual)
meaning
'devoted'
Figure 3. Chronological structure of PAs in adnominal patterns
The issue, then, is to capture these developments in a way that reflects

both the shifts in the morphosemantic structure (i.e., the relative promi-
nence of the verbal and nominal features) and the attendant effects on the
external syntactic and semantic status of the PAs. When we relate the PA
formation to the network of the verbal frames in Figure 2, we arrive at the
representation in Figure 4, which incorporates not just the lexical seman-
tics of the root, but also the fact that the PA is a morphologically complex
pattern that may display a varying degree of mismatch between its internal
and external morphosemantic properties. Each PA is thus represented as a
construction (the outer boxes) in the sense of Construction Grammar.
In the regular PAs, the inside boxes represent the meaning of the verb
root (still part of the polysemy network, as indicated by the links to the
prototype) as well as the contribution of the participial stem (indicated by
the NT-ppl subscript, as a shorthand for spelling out the full morphological
structure, the details of which are not crucial to the concerns of this paper).
However, the pseudo-PA status of the passive/modal meaning within the
BELIEF sense is marked explicitly in the internal structure: the suffix -úcí is
a separate unit that attaches to the root. What is added (in bold) at the top
of each outer box represents the contribution of the PA itself and its overall
meaning: the crystallization of the external category (cat a), the PA’s
meaning (abbreviated as a label in the sem attribute), and the changes, if
any, we observe in expressing the root’s valence. We also note that this
particular pseudo-PA is attested only as a predicative rather than an attribu-

tive adjective (indicated by cat apred), in contrast to the regular PAs.
cat a pred. lform věřící (ex. 7,15)

sem ['credible'] cat a lform věřící
sem ['believing in Christ']
cat vroot lxm věřiti lform -úcí
val { ( #2 [Acc (v) ] ) }
Frame BELIEF
(content of thought) cat vNT-ppl lxm věřiti
FEs: #1 Believer Frame FAITH
#2 Recip. of Trust (content of thought)
#3 Content FEs: #1 Believer
(ex. 11) #2 Christ
val {#1 [Nom], #2 [Acc(v) ] }
Frame BELIEF
(content of thought)
FEs ... cat a lform věřící
sem ['devoted/faithful']
( ? collocation věřící list ) cat vNT-ppl lxm věřiti
'authenticating letter' (ex. 12)
Frame FAITH
cat a lform věřící (content of thought)
FEs: #1 Believer
val { #2 [Dat], (#3 [Loc]) }
#2 Deity
cat vNT-ppl lxm věřiti
val {#1 [Nom], #2 [Acc(v) ] }
Frame BELIEF
(content of thought) (ex. 16)
FEs: #1 Believer
#2 Recip. of Trust
#3 Content
val {#1 [Nom], #2 [Dat], (#3 [Loc]) }
Figure 4. Frame Semantic representation of věřící in adnominal patterns
The diagram also captures, at least roughly, the difference in the de-
grees to which the verbal component contributed by the root may fade in a
given PA class (indicated by the gray color). For example, the morphologi-
312 Mirjam Fried
cally regular BELIEF sense PA (lower left) remains a fully inflectional form
whose meaning corresponds directly to its morphemic structure, i.e., is
semantically fully compositional and hence, is not specified at the word
level. In contrast, the pseudo-PA (upper left) obliterates most of the inter-
nal verbal features; the only pieces of information that remain are the
meaning of the root (in the frame attribute) and reference to the participant
that is the referent of the NP modified by the PA (FE #3 Content). The
FAITH PA illustrates the gradual shift that is more typical of PAs in general:
in the early stage it retains the categorial tension between the NT-stem and
the external morphology, and its meaning and function are still more or
less predictable from the morphological structure, although the class of
referents for the Deity FE is fixed to ‘Christ’ (this restriction is indicated
by bold-face in the Frame specification), in a departure from the way the
finite verb is used (recall example (3a)). In subsequent development,
though, the interpretation of this PA extends further into a fully atemporal
classificatory meaning, suppressing any verbal/participial potential of the
form (in gray).
3.2. Syntactic nouns
The syntactic nouns start appearing in my corpus by mid-14th century, and

occur in a variety of texts and genres. Morphologically, the oldest tokens
are all pseudo-PAs of the form věřúcí, but later on we find also several
tokens of věřijící. Although they are attested in their bare form approxi-
mately as often as they have their non-subject complements present (ex-
pressed in the same form as would appear with a finite verb), their function
is that of an actor noun: ‘the one(s) who believe(s)’. And just as we saw
with the adnominal PAs, the syntactic nouns also cover only some parts of
the semantic spectrum of the verb root; in particular, they are attested only
in the CREDIT and FAITH verb senses.
Not surprisingly, the FAITH usage is found in all kinds of texts. It is also
attested as early as in the 1350 layer and as we saw in Table 2, it is by far
the most frequent one in the corpus. The oldest occurrences tend to main-
tain the verbal valence, expressing the second argument in the v + Acc
form; typical examples are given in (17). Note also that this pattern com-
pletely parallels the adnominal counterparts in the relative non-
informativeness of the complement: the second argument is occasionally
buoh ‘god’ or Kristus ‘Christ’ but the vast majority of cases (75% of the
tokens with complements) have only a pronoun, either personal (17a) or

reflexive (17b), always referring to God/Christ.
(17) a. {jsi}… všech v tě věřících

{you.are}… all.GEN.PL in 2SG.ACC believe.PA.GEN.PL
útěcha
solace.NOM.SG.F
[Legkat 65a; late 1300s; spiritual poetry]
‘{you are} … the solace of all those believing in you’
b. jenž všudy pomáhá v sě

who.NOM.SG.M everywhere help.PRES.3SG in self.ACC
věřícím
believe.PA.DAT.PL
[OtcB 196b; late 1300s; legend]
‘[Christ] whoi in all places helps those believing in himi’
The low semantic contribution of the complement downplays the PA’s

verbal potential and is consistent with the overall function of this use: al-
most all of the examples can be described as identifying a referent (‘a be-
liever [in God/Christ’]). This functional status is also reinforced syn-
tactically: about half of the examples contain the quantifier všichni ‘all’
and the PA thus invites the interpretation of being the head of an NP.
Nonetheless, the presence of the second argument, however unin-
formative, gives these PAs the flavor of a restrictive relative clause (‘those
who believe in God/Christ’) and the verbal potential still persists to some
degree. This is so not only because of the syntactically present complemen-
tation. Some contexts strongly suggest a ‘predicate’ reading, i.e. one in
which the PA resembles, functionally, a true participle, expressing an ad-
verbial clause, regardless of complementation. For example, the bare PA in
(18) can be just as easily interpreted as referring to an entity as expressing
a condition under which the event expressed by the main predicate holds.
(18) {Apolonius…nauči jej nasledovati přiebytka věčného a

trpělivě hledati božieho milosrdie...řka jemu}
že sě muož všecko dokonati
that RF can.PRES.3SG everything.NOM.SG.N complete.INF
314 Mirjam Fried
věříciemu
believe.PA.DAT.PL
[OtcB 19a; late 1300s; legend]
‘{Apolonius…taught him to aim for the eternal dwelling and
patiently seek God’s mercy…, telling him} that everything is
possible if one believes/for a believer’
This means that the syntactically nominal PA still could mark relative
tense and, crucially, only a particular context can determine which inter-
pretation is more accurate or is more likely intended by the speaker. It must
be noted that this is not a quirk of the form věřící, but follows the general
behavior of OCz PAs, both nominal and adnominal ones (Fried 2003 and
forthcoming-b).
The potential for ambiguity with respect to the relative strength of the
verbal character persists through the first part of the 15th century, but
overall, we observe a steady shift toward semantic nouns referring to enti-
ties that are characterized by certain properties, removed from any tempo-
ral or event-based interpretation that would be suggested by the participial
stem. ‘Having faith’ is not understood as an eventuality ascribed to an en-
tity (the restrictive relative clause interpretation) but as a classification of
an individual, independently of the event expressed by the main predicate.
This shift correlates with a gradual change in syntactic properties through-
out the 15th century. There is a noticeable erosion of the complement
structure: the attested occurrences are mostly bare (28 tokens out of 46), as
in (8), often accompanied by a universal quantifier (in 10 cases). If a com-
plement is present, it remains just as communicatively redundant as in the
oldest texts.
The low information value of the complement is nicely illustrated in
one particular text (VýklŠal, a religious tract from early 15th century, in-
terpreting the Song of Songs as a simile for proper Christian faith), which
consistently alternates between a bare PA věřící ‘believer’ and the phrase
věřící v Krista ‘believer in Christ’. The use of the latter suggests a fixed
collocation that simply stresses the Christian nature of the religious context
explicitly, but otherwise is not intended as a contributor of a novel event
participant. In fact, by mid-15th century we see the crystallization of the
PA’s meaning into denoting ‘a Christian’, i.e. as a label for a particular
social and religious group identity, or a class of people, rather than the
more general ‘one who believes in God/Christ’. Put differently, the
compositional meaning provided by the morphosemantic structure evolves
into a more specialized meaning that is not directly predictable from the
PA’s morphology. This semantic development is also evidenced by the fact

that this PA appears as a synonym of křest’an ‘Christian’ (e.g. HusBlud
293).
It is also during this period that the nominal syntax of věřící becomes
gradually more and more apparent. The shift involves an expanded inven-
tory of modifiers, such as an occasional possessive adjective, shown in (8)
above (cf. svým věřícím ‘to his own believers/faithful’ in (8) vs. the verbal
form v sě věřícím ‘to believers in him[self]’ in (17b), both reflexive forms
referring to Christ), as well as appearing in positions of adnominal geni-
tives (cierkev věřícíchGEN.PL ‘the church of believers’, VýklŠal 112). This
general trend culminates in the last half-century of the OCz period, when
the PA consistently displays distinct NP syntax (for example, virtually all
tokens now appear in quantified phrases), regardless of complementation.
Its meaning is fully established as ‘a Christian’, and certain administrative
as well as pastoral texts reflect an additional, metonymic shift, toward the
interpretation ‘a church member’, illustrated in (19). These kinds of uses
suggest highlighting group membership as a salient feature, not just the
individual identity based on a particular kind of belief.
(19) {žádajíc...spasení duše své a svejch starších a předchůdcích..., }

i jiných všech mrtvých
also other.GEN.PL.M all.GEN.PL.M dead.GEN.PL.M
věřících {voltář ke cti svaté Dorotě...založil a nadal}
believe.PA.GEN.PL
[ArchČ 9, 243; 1400; admin. record of dedicating a new altar]
‘{desiring…the salvation of his own soul and [the souls] of his
elders and predecessors…} and also of all other deceased mem-
bers [of his] congregation {he founded and endowed an altar to
honor Saint Dorota}’
The PA used in the CREDIT sense is interestingly different in several re-

spects. For one thing, it emerges somewhat later: we start seeing examples
only in the early 1400s. For another, it only shows up in administrative
texts, suggesting the status of a ‘technical’ term, rather than being part of
the common vernacular. Finally, it displays very distinct syntactic prefer-
ences. It is always used in its bare form, thus strongly inviting a referential
interpretation, denoting a person that occupies a particular role in a com-
mercial transaction, ‘creditor’. This is confirmed by the synonyms that
sometimes appear in the same text, especially in the late 1400s, where the
316 Mirjam Fried
PA is used interchangeably with the morphologically true actor noun

věřitel, as shown in (20).
(20) {a tak vždy máme vydávati na každý rok věřitelóm našim…,

dokudž bychom svrchupsaných zlatých jistinných i s úrokem}
našim věřícím svrchupsaným
our.DAT.PL believe.PA.DAT.PL above.written.DAT.PL
nedali {a nezaplatili úplně}
NEG.give.PPL.PL
[ArchČ 9, 294; 1451; a loan contract]
‘{and so we must pay each year to our creditors...until we might
give and pay in its entirety the above-mentioned principal and
interest} to our above-mentioned creditorsPA’
Moreover, this PA is always modified by a possessive pronoun (naši ‘ours’

in the above example), which marks the identity of the Debtor participant
and is a clear sign of the nominal character of this PA use as an actor noun.
To summarize, over the two centuries we witness a steady development
toward strengthening a referential function and the attendant semantic
shifts of věřící in the FAITH sense, from the reading ‘the one who believes in
a deity/Christ’ with a somewhat fluid functional status, to a fully referen-
tial reading ‘believer/Christian’, with a further, though in OCz still mar-
ginal, extension to ‘church member’. These developments are arranged
chronologically in Figure 5.
While the complementation pattern is often preserved with this PA and
always takes the form it would have in a finite verbal expression, the com-
plement is invariably low in informativeness. Moreover, the external syn-
tax of the PA is that of a noun and this status is consistently strengthened
by gradually adding other overt features of an NP (especially quantifica-
tion). It also seems that this grammatical pattern had been fairly robust and
well established by the time the sense ‘creditor’ emerges; the latter does
not show any traces of being possibly an inflectional form.
1350 1380 1410 1450 1500 ... MCz
FAITH sense
compl. (v ACC)
modifiers (Quant.)
'believer'
meaning
'Christian'
'church member'
CREDIT sense
compl. - none
modifiers (Possess.)
meaning 'creditor'
Figure 5. Chronological development of PAs as syntactic nouns
The place of the syntactic nouns in the semantic network is summarized

in Figure 6, using the same (simplified) formalism as was used for the ad-
nominal PAs.
The case of the CREDIT sense is straightforward: the verbal origin is re-
flected in the root meaning, but its valence is completely suppressed and
any verbal potential of the NT-morphology is neutralized. The FAITH PA, in
contrast, involves a somewhat surprising twist, in that it starts out as a
pseudo-PA. As already noted, the pseudo-forms are often associated with
shifted meanings that completely obliterate the inflectional structure of the
participial stem; we thus do not expect the present tense or active voice
marking as part of their semantic contribution, and the adnominal PAs
confirm this expectation. Yet, the early usage of the FAITH PA in the noun
slots patterns after the relatively transparent readings of regular PAs which
still retain much of their hybrid structure (recall examples in (17) or (18)).
For the purposes of Figure 6, I represent the irregular formation in the
same abbreviated way used in Figure 4 but indicate the potentially ‘visible’
internal morphology by including reference to tense; working out the for-
mal details of this variation will have to be left for future research. The
arrows again indicate the direction of semantic shifts; the plain lines sim-
ply remind us that some other frames, inherently independent of the BELIEF
network, share some features with a given sense in the network.
318 Mirjam Fried
(ex. 20)
COMMERCIAL
TRANSACTION cat n lform věřící
sem ['creditor']
TRANSFER lxm věřiti
cat vNT-ppl
Frame CREDIT
(belief as an expectation)
Frame BELIEF FEs: #1 Creditor
(content of thought)
#2 Debtor
FEs ... #3 Valued-Entity
val {#1 [Nom], #2 [Dat], #3 [Acc] }
(exs. 8, 17, 18)
cat n lform věřící
sem ['Christian']
val { ( #2 [Acc (v) ] ) }
cat vroot lxm věřiti lform -úcí
Frame FAITH tense ? cat n lform věřící

(content of thought) sem ['church member']
FEs: #1 Believer
#2 Christ (ex. 19)
val {#1 [Nom], #2 [Acc(v) ] }
Figure 6. Frame representation of noun semantics of OCz věřící
4. Overall diachronic structure of věřící
In order to understand the conventional interpretations of the OCz PAs,

together with the functional and categorial status they gradually acquired,
we must consider the range of their meanings in relation to the semantic
structure of the verb root from which they are derived. Without situating
the PAs in that semantic network, we could at best provide a disjointed list
of uses without any apparent motivation. Yet, we can see that the PAs form
coherent subnetworks within the semantic space associated with the verb.
Moreover, it is evident that the same morphological form (inflectional at
that) may take a number of paths that not only highlight different clusters
of functional properties, but also occupy different parts of the relevant

semantic space. And crucially, the paths are not necessarily predictable
from anything in the morphosemantic structure per se: for example, there is
no reason to expect that a particular sense should develop into a referential
vs. modification vs. predicative function. The only way we can establish
shifts within a semantic network is to study the details of how its (poten-
tial) members are used in context and what factors contribute to the sharp-
ening of an inherently available but vaguely delimited and initially context-
dependent function.
With respect to the semantic structure itself, we can make several gen-
eral observations regarding the distribution of the verbal senses across the
PAs and the relative age and longevity of each PA class, as well as assess
the significance of the pseudo-PAs in the overall development. Let us start
with a chronological summary of all the semantic and functional facts,
shown in Figure 7. For each sense of věřiti we can compare its attestations
as a verb (V, in gray) and in the corresponding PA form(s), either in the
attributive (A) or referential (N) functions. The dashed lines again indicate
relatively low incidence and the PAs are entered in their actual forms:
věřúcí (both attested and reconstructed) until late 1300s, věřící after that
point, and věřijící as appropriate.
We can start by observing that the senses CREDIT and TRUST did not last
beyond the OCz period, and their disappearance can perhaps be motivated
by appealing to their relative distance from the prototype. Both of these
senses represent a fairly significant semantic shift in that they cast the con-
cept of belief as an expectation, rather than the content of thought (shared
by the prototype and the FAITH sense). Especially the TRUST sense appears to
have been very marginal to begin with and the apparent lack of correspond-
ing PAs need not be, therefore, surprising. In contrast, both the verb and
the PA in the CREDIT sense are attested in abundance, but we also need to
keep in mind that they were only used as part of a particular technical vo-
cabulary. This PA always had a very specific meaning and syntactic func-
tion, not suggesting any transition from a productive inflectional form to a
more specialized usage. Not that such a transition should predict anything
about the survival rate of a given form, but its absence underscores the
‘outside’ status of the usage. The TRUST sense became gradually the domain
of a different verb derived from the same root (dů-věř-ovat ‘trust’) and the
CREDIT verb and its PA also gave way to competing expressions that were
in existence at the same time and were dedicated to these two meanings in
a morphosemantically transparent way: dávat na úvěr ‘give on credit’ and
věřitel ‘creditor’.
320 Mirjam Fried
1300 1350 1380 1410 1450 1500 ... MCz
V
BELIEF
'credible' (habitual)
A
věřúcí 'authenticating'
věřící list
V
CREDIT
'creditor'
N
věřící
V
FAITH
'believer' 'Christian'
N
'church member'
věřúcí 'Christian'
věřijící
'believing/Christian'
A
věřúcí 'devoted'
věřijící
TRUST V
Figure 7. Semantic and chronological overview of věřící/věřúcí/věřijící
The low incidence of the BELIEF PA in its active reading (labeled ‘habit-
ual’) can be perhaps motivated on pragmatic grounds. The CNG suffix pre-
disposes the PA toward developing readings in which the meaning of the
stem is interpreted as marking salient attributes of an entity or identifying
referents with such attributes, but this may be incompatible with what it
means to believe that something is true. We tend to perceive belief as a
fleeting state of affairs, not something that characterizes a person in a
noteworthy way (for that, Czech has a dedicated adjective, důvěřivý ‘trust-
ing/gullible’, which marks the habitualness and salience of the attribute by
the suffix -iv).
It is also significant that the PAs do not follow the same developmental
path for every sense. This fact is independent of the verb věřiti or any of its
senses, since the same unpredictable preferences for functional develop-
ment are commonly attested with other PAs. The diachronic patterning
only reveals something about this particular morphological form: while the
form itself carries the inherent potential to develop any (and all) of the
three syntactic functions (referential, attributive, predicative), we cannot
make a prediction about what direction of change we should expect. For
now I have to leave open the question of whether the diachronic tendencies
are truly unpredictable for every verb or whether there might be some very
general factors that predispose certain verbs or verb classes to a particular
functional outcome. But even if verb semantics (or other factors) turn out
to be relevant in this respect, such a result will not invalidate the observa-
tion that this is not a one-size-fits-all issue. The only generalization avail-
able at this point is the following: the external properties, not surprisingly,
have a stronger claim to dominance than the internal ones, hence a much
higher proportion of shifts toward reference and modification, as compared
to only marginal persistence of verbal behavior. But the competition be-
tween the formal (adjectival) and semantic (referential) potential of the
external morphology seems much less clear-cut. In the case of věřící, the
modificational uses cover a greater semantic territory within the semantic
network, but the referential ones turned out to be the most robust.
The oldest attestations are all within the prototypical (BELIEF) sense, but
they are all morphologically irregular – věřúcí (there is no example of a
clearly regular form indicated by the spelling věřiecí, even though the true
participle is attested quite commonly in the regular form věřiec ‘[while]
believing’). These pseudo-forms show a dramatic shift in meaning (‘be-
lievable’) in the adnominal usage, thus supporting the general hypothesis
that the irregular forms tend to be associated with some irregularity in
meaning or function. The corresponding active participial usage, ‘(the one)
believing’, does not appear until the point when the difference between the
regular and irregular formation is phonetically fully neutralized, as re-
flected in the spelling, at the end of the 14th century. From all this, we
could infer a diachronic development in which a pseudo-form starts out as
an independent form and is gradually pulled into the network of regular
PAs, potentially adjusting to their regular internal structure. However, as
noted in section 3.2, the same pseudo-form used in the nominal slots seems
to function the same way as the regular PAs, suggesting that the sequence
-ú-c-í must have retained some degree of transparency all along, at least
when used with this particular verb root and in this particular syntactic
function. It will require additional research to establish firmly that the dia-
chronic path of the pseudo-PAs vis-à-vis their regular counterparts sug-
gested above indeed is the correct one; and if it is, it also raises interesting
questions for the grammaticalization theory, minimally with respect to the
possibility that a frozen form may disintegrate back into its component
322 Mirjam Fried
parts. I cannot resolve these general issues here; I am simply proposing a

hypothesis for further testing. We may at least note, though, that it is fully
in the nature of things that language change does not proceed in clean-cut,
discrete stages and that overlaps in function and attraction across form-
meaning clusters are to be expected (cf. Hopper 1991; Traugott forthcom-
ing).
A separate pseudo-PA is attested in the FAITH sense. The form věřijící
did not stay in the language very long and it is possible that it was just a
one-time flourish, perhaps an attempt to reinforce the FAITH-related usage
as distinct from everything else. Notice also that this appears at the time
when the ie > í change had been under way, thus further obscuring any
distinction between the regular and irregular morphology; in contrast, the
sequence -ijí-c-í was unambiguously irregular for this morphological verb
type.
Finally, Figure 7 brings out the fact that the shifts do not progress in a
phase-by-phase pattern, one function at a time. Rather, multiple functions
coexist in overlapping layers across several generations of speakers, creat-
ing alternating stretches of rich polysemies and their subsequent ‘pruning’.
This is not surprising, especially when we consider that OCz was, after all,
primarily a spoken language (we can only study it through written records,
of course, but the vast majority of these texts were distributed mainly in
oral form) and we can, therefore, expect a high degree of variation at any
given point in time. In motivating and explaining the relative survival rates
of individual uses, some of which may be dedicated to very specific con-
texts or genres (e.g. administrative), while others are spread across the
spectrum, it ultimately comes down to working out a relative (and con-
stantly shifting) equilibrium between polysemy and isomorphism as two
competing and equally strong cognitive strategies (cf. Geeraerts’ 1985:
143–145 illuminating analysis of this tension and its effects in language
change).
5. Conclusions
This case study is concerned with the general developmental tendencies

that can be identified in a hybrid category such as the PA, a form that
straddles the boundary between inflection and derivation. The analysis of
the OCz PAs věřící/věřúcí/věřijící leads to the conclusion that it is not just
an issue of charting the ways in which the semantic space is divided among
different uses in different contexts, but also an issue of isolating the
specific properties (semantic, morphological, syntactic, and contextual)

that signal the erosion of the compositional morphosemantic structure of
the PA and identify the point(s) at which the erosion becomes fixed as a
full categorial and semantic shift. I hope to have shown that a functionally
oriented, frame-based constructional approach is particularly helpful in
capturing the categorially hybrid nature of the PA in that it allows us to
address the dynamic relationship between its internal and external proper-
ties, the interdependence of lexical meaning and morphosyntactic
patterning, and the subtle changes in the mappings over time. The Frame
Semantic approach to meaning provides additional tools for capturing the
semantic structure as a lexico-grammatical network of related or partially
overlapping meanings and for tracing the patterns of conventionalization in
the sense of ‘item x in environment y has function z’, not simply ‘item x
has function z’.
The paper touches on several issues of broad interest as well, suggesting
directions for further examination of these forms: (i) the role of isomor-
phism vs. polysemy in language change, (ii) the complex nature of layering
(Hopper 1991), (iii) the factors involved in categorial shifts known as
transpositional morphology, and (iv) adequate representation of such shifts
both diachronically and synchronically.
Notes
* I am greatly indebted to my colleagues in the Staročeské oddělení ÚJČ for

giving me access to their data archives and extensive library of Old Czech
texts; without their generous help, this paper would not have been possible. I
particularly wish to thank Milada Homolková and Věra Chládková for very
helpful discussions of my data. And I am very grateful to two anonymous re-
viewers for their useful practical comments, general encouragement, and in-
spiring questions for further work.
1. I use the -NT- label as the traditional way of classifying the I-E present active
participle, based on its reflex in Latin.
2. While the pseudo-PAs are no rarity in OCz texts and suggest a fairly robust,
semantically motivated type of word-formation (rather than being random
mistakes), they gradually disappeared from the language and survive only in
several frozen tokens in Modern Czech. Their exact status and their place in
the morphosemantic networks of OCz PAs still remain an open question
awaiting further research. The present paper is but the first attempt at mapping
out (some of) the relationships.
324 Mirjam Fried
3. I assume the commonly accepted cut-off point at the end of 14th century (e.g.
Vintr 1992): written i/y represents í < ú before that date, but is potentially am-
biguous after that.
4. As is well known, Frame Semantics and the version of Construction Grammar
presupposed in this paper incorporate the notion of prototype as a crucial or-
ganizational principle in representing speakers’ linguistic knowledge. Parts of
the analysis will thus be directly compatible with prototype-based research in
lexical semantics, especially as laid out in Geeraerts’ ground-breaking work in
cognitively oriented historical semantics. However, the present paper is not so
much concerned with the (re)organization of the conceptual structure that un-
derlies the meaning of lexical items (Geeraerts’ explicitly stated focus). I am
more concerned with changes in the lexico-grammatical organization associ-
ated with categorial changes in a particular inflectional word-form, for which
the apparatus of Construction Grammar, as a cognitively based model of lin-
guistic structure, is particularly suitable.
5. Glosses: NOM ‘nominative’, GEN ‘genitive’, DAT ‘dative’, ACC ‘accusa-
tive’, LOC ‘locative’, INS ‘instrumental’, M/F/N ‘masculine/feminine/ neu-
ter’, SG/PL’singular/plural’, NEG ‘negative’, PRES ‘present’, PST ‘past’,
FUT ‘future’, PPL ‘past participle’, PASS ‘passive participle’, IMP ‘impera-
tive’, INF ‘infinitive’, AUX ‘auxiliary’, RF ‘reflexive’, COND ‘conditional’,
PP ‘pragmatic particle’.
6. A note on presenting the examples: when additional context is helpful for
clearer understanding, it will be enclosed in curly brackets {} and left without
interlinear glossing, as in (3b). If the context in the original is too elaborate, I
will add an explanatory summary only in the English translation, enclosed in
parentheses (), as in (2a).
7. The cited texts and each example’s exact location in a text are identified by
the abbreviations and citing conventions established by the Old Czech Dic-
tionary (Staročeský slovník 1968).
8. The genitive in (5b) could be analyzed simply as a genitive of negation, but it
does not call into question the genitive as part of the conventional comple-
mentation pattern of this sense of věřiti. Notice that (5a) also contains the
genitive, even though there is no negation in the sentence.
9. While many of the manuscripts in the corpus can be dated quite precisely,
many others can only be placed within an estimated time period (a decade or
more). For the present purposes, it is sufficient to follow the practice estab-
lished by the Old Czech Dictionary (Staročeský slovník 1968), which recog-
nizes six chronological layers, each identified by reference to its upper bound-
ary ([up until] 1300, [up until] 1350, and so on).
10. In this respect it may also be worth noting that the copular use, sporadic but
persistent throughout the OCz period, as often as not functions as a predicate,
expressing simply a present-tense active event, as in (9) above. Moreover,
there are occasional examples of the PA used as a verb, e.g. in an absolute

construction.
11. Nevertheless, it would be an overgeneralization to conclude that the suffixes
forming the pseudo-PAs simply correlate with non-active meaning. Not all
pseudo-forms have the passive reading (including the actor-noun usage ad-
dressed in section 3.2), while a number of regularly formed PAs also show the
same shift in voice; cf. žádaj-ú-c-í ‘desired/desirable’ (< lit. ‘desiring’), as
discussed in Fried (forthcoming-b).
12. Support for this reasoning also comes from the possible antonyms. It is hard
to imagine what a ‘non-believing Christian’, in contrast to the locutio in (16),
would mean in the more literally participial (i.e., potentially temporally
grounded) reading, instead of the attributive reading of an atemporal quality
(‘faithful’).
References
Atkins, B.T.S.
1994 Analyzing the verbs of seeing: A frame semantics approach to corpus
lexicography. In Berkeley Linguistic Society 20: 42–56.
Atkins, Sue, Charles J. Fillmore, and Christopher R. Johnson
2003 Lexicographic relevance: Selecting information from corpus evi-
dence. International Journal of Lexicography 16: 251–280.
Bisang, Walter, Nikolaus P. Himmelmann, and Björn Wiemer (eds.)
2004 What Makes Grammaticalization? A Look from Its Fringes and Its
Components. Berlin: Mouton de Gruyter.
Bybee, Joan, Revere Perkins, and William Pagliuca
1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Lan-
guages of the World. Chicago: Chicago University Press.
Croft, William
1991 Syntactic Categories and Grammatical Relations: The Cognitive
Organization of Information. Chicago: Chicago University Press.
2001 Radical Construction Grammar. Oxford: Oxford University Press.
Daneš, František
1966 The relation of centre and periphery as a language universal. Travaux
linguistiques de Prague 2: 9–21.
1982 Frame Semantics. In Linguistics in the Morning Calm, The Linguis-
tic Society of Korea (ed.), 111–137. Seoul: Hanshin Publishing Com-
pany.
326 Mirjam Fried
1989 Grammatical Construction theory and the familiar dichotomies. In

Language Processing in Social Context, Rainer Dietrich and Carl F.
Graumann (eds.), 17–38. Amsterdam: North-Holland/Elsevier.
Fillmore, Charles J., and B.T.S. Atkins
1992 Toward a frame-based lexicon: The semantics of RISK and its
neighbors. In Frames, Fields, and Contrasts: New Essays in Seman-
tic and Lexical Organization, Adrienne Lehrer and Eva Feder Kittay
(eds.), 75–102. Lawrence Erlbaum Associates.
Fillmore, Charles J., Christopher R. Johnson, and Miriam R. L. Petruck
2003 Background to FrameNet. International Journal of Lexicography 16:
235–250.
Fillmore, Charles J., Miriam R.L. Petruck, Josef Ruppenhofer, and Abby Wright
2003 Framenet in action: The case of attaching. International Journal of
Lexicography 16: 297–332.
Fischer, Olga, Anette Rosenbach, and Dietrich Stein (eds.)
2000 Pathways of Change: Grammaticalization in English. Amster-
dam/Philadelphia: Benjamins.
Fried, Mirjam
2003 Dimensions of syntactic change: Evidence from the long -nt- partici-
ple in Old Czech texts. In American Contributions to the 13th Inter-
national Congress of Slavists in Ljubljana, Robert A. Maguire and
Alan Timberlake (eds.), 79–92. Bloomington, IN: Slavica Publishers.
2004 Predicate semantics and event construal in Czech case marking. In
Construction Grammar in a Cross-Language Perspective, Mirjam
Fried and Jan-Ola Östman (eds.), 87–120. Amsterdam/Philadelphia:
Benjamins.
2005a A frame-based approach to case alternations: The swarm-class verbs
in Czech. Cognitive Linguistics 16 (3): 475–512.
2005b The stability of meaning-form associations across time. In Verba et
historia, Petr Nejedlý and Miloslava Vajdlová (eds.), 77–86. Praha:
Ústav pro jazyk český AVČR.
Forthc. (a) Grammaticalization and lexicalization effects in participial morphol-
ogy: A Construction Grammar approach to language change. In
Grammaticalization and Grammar, Hubert Cuyckens, Kristin
Davidse, and Jean-Christophe Verstraete (eds.). Amster-
dam/Philadelphia: Benjamins.
Forthc. (b) Constructions and constructs: mapping a shift between predica-
tion and attribution. In Constructions and Language Change, Alex-
ander Bergs and Gabriele Diewald (eds.). Berlin: Mouton de Gruyter.
Fried, Mirjam, and Jan-Ola Östman

2003 The explicit and the implicit in the Suasion Frame. In Proceedings of
CIL 17. Eva Hajičová, Anna Kotěšovcová, and Jiří Mírovský (eds.),
1–22, CD-ROM. Prague: Matfyzpress.
2004 Construction Grammar: A thumbnail sketch. In Construction Gram-
mar in a Cross-Language Perspective, Mirjam Fried and Jan-Ola
Östman (eds.), 11–86. Amsterdam/Philadelphia: Benjamins.
Gebauer, Jan
1958 Historická mluvnice jazyka českého III [Historical grammar of the
Czech language]. Praha: F. Tempský.
Geeraerts, Dirk
1983 Reclassifying semantic change. Quaderni di Semantica 4: 217–240.
1985 Cognitive restrictions on the structure of semantic change. In
Historical Semantics: Historical Word-Formation, Jacek Fisiak
(ed.), 127–153. Mouton Publishers.
1992 Prototypicality effect in diachronic semantics: A round-up. In
Diachrony within Synchrony: Language History and Cognition,
Günter Kellermann and Michael D. Morrissey (eds.), 183–204.
Frankfurt/New York: Peter Lang.
1997 Diachronic prototype semantics: A contribution to historical
lexicology. Oxford: Clarendon.
Haspelmath, Martin
1996 Word-class-changing inflection and morphological theory. In
Yearbook of Morphology 1995, Geert Booij and Jaap van Marle
(eds.), 43–66. Dordrecht: Kluwer Academic.
1998 Does grammaticalization need reanalysis? Studies in Language 22:
315–351.
1999 Why is grammaticalization irreversible? Linguistics 37: 1043–1068.
Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer
1991 A Conceptual Framework. Chicago: Chicago University Press.
Herring, Susan C., Pieter van Reenen, and Lene Schøsler (eds.)
2000 Textual Parameters in Older Languages. Amsterdam/Philadelphia:
Benjamins.
Hopper, Paul
1991 On some principles of grammaticalization. In Approaches to
Grammaticalization I, Elizabeth C. Traugott and Bernd Heine (eds.),
17–35. Amsterdam/Philadelphia: Benjamins.
Hopper, Paul, and Sandra A. Thompson
1984 The discourse basis for lexical categories in universal grammar.
Language 60: 703–752.
1985 The iconicity of the universal categories ‘noun’ and ‘verb’. In
Iconicity in Syntax, John Haiman (ed.), 151–183. (Typological
Studies in Language 6.) Amsterdam/Philadelphia: Benjamins.
328 Mirjam Fried
Hopper, Paul, and Elizabeth C. Traugott

1993 Grammaticalization. Cambridge: Cambridge University Press.
Kurz, Josef
1958 K otázce doby vzniku slovanských adjektiv složených a jejich
původního významu [On dating the rise of Slavic long adjectives and
their original meaning]. In Studie ze slovanské jazykovědy, sborník
k 70. narozeninám akademika Fr. Trávníčka, 211–219. Praha: SPN.
Luraghi, Silvia
2003 On the Meaning of Prepositions and Cases: The Expression of Se-
mantic Roles in Ancient Greek. Amsterdam/Philadelphia: Benjamins.
Marvan, Josef, Emanuel Michálek, and Igor Němec
1963 Základní procesy v lexikálním vývoji českého jazyka [Basic patterns
of change in the lexical development of Czech]. Čs. Přednášky pro
V. mezinárodní sjezd slavistů v Sofii, 105–113. Praha: Nakladatelství
Československé akademie věd.
Michálek, Emanuel
1963 Staročeská adjektiva typu lácí, lajúcí [Old Czech adjectives of the
type lácí, lajúcí]. Listy filologické 86: 127–130.
Schultze-Berndt, Eva, and Nikolaus P. Himmelmann
2004 Depictive secondary predicates in crosslinguistic perspective.
Linguistic Typology 8 (1): 59–131.
Staročeský slovník
1968 [Old Czech dictionary]. Praha: Academia.
1988 Pragmatic strengthening and grammaticalization. In Berkeley Lin-
guistic Society 14: 406–416.
1992 From less to more situated in language: The unidirectionality of se-
mantic change. In Meaning and Grammar: Cross-Linguistic Per-
spectives, Michel Kefer and Johan van der Auwera (eds.), 497–517.
Berlin: Mouton de Gruyter.
Forthc. The grammaticalization of NP of NP constructions. In Construc-
tions and Language Change, Alexander Bergs and Gabriele Diewald
(eds.). Berlin: Mouton de Gruyter.
Traugott, Elizabeth C., and Richard B. Dasher
2002 Regularity in Semantic Change. Cambridge: Cambridge University
Press.
A prototype account of the development of
delimitative po- in Russian
Stephen M. Dickey
Abstract*
This paper applies principles of prototype semantics to explain the devel-

opment of the Russian prefix po- from a primarily resultative prefix to a de-
limitative prefix, which occurred from the eleventh to the nineteenth cen-
tury. It is argued that this development reflects a shift in its semantic
prototype from PATH/SURFACE-CONTACT to INGRESSIVE-PARTIAL TRAJEC-
TORY. Old Russian had a small class of po- delimitatives that could express
the relatively short duration of a situation (RELATIVE DELIMITATION) or its
duration for an entire specified interval (ABSOLUTE DURATION); only RELA-
TIVE DELIMITATION survives in modern Russian. Spatial po- with determi-
nate motion verbs underwent a similar development: Old Russian poiti ‘go’
could express either some initial portion of a motion event (INGRESSIVE-
PARTIAL TRAJECTORY) or the complete motion event (FULL TRAJECTORY);
only INGRESSIVE-PARTIAL TRAJECTORY survives in modern Russian. It is ar-
gued that the parallel semantic developments of these two senses of po- are
no coincidence, and that the development of modern Russian delimitatives
followed the development of po- as a perfectivizing prefix for determinate
motion verbs. The similarity of the new meaning of determinate-motion po-,
INGRESSIVE-PARTIAL TRAJECTORY, to the delimitative meaning of RELATIVE
DELIMITATION was a semantic point of contact that produced a new atelic
cluster of meanings in the network of old resultative po-. Atelic INGRESSIVE-
PARTIAL TRAJECTORY became its prototype and formed the semantic basis of
the new po- delimitatives. The salience of po- in poiti and other determinate
motion verbs facilitated the semantic productivity of INGRESSIVE-PARTIAL
TRAJECTORY and thus the rise of delimitative verbs in modern Russian.
Keywords: Russian, prefixation, delimitative verbs, motion verbs, proto-

type semantics, diachrony
330 Stephen M. Dickey
1. Preliminaries
In line with its assumption that all linguistic units are equally symbolic and
its rejection of a qualitative distinction between lexical and grammatical
units, cognitive linguistics has approached the analysis of the semantic
categories expressed by affixes and grammatical endings in essentially the
same way that it approaches the meanings of lexical units. Lehrer (2003)
argues that derivational affixes exhibit in principle the same kind of
polysemy that lexical elements do, even if to a lesser extent, given the fun-
damentally abstract nature of the meanings derivational affixes express.
Nikiforidou (1991) documents a structured polysemy for the genitive case
in Greek (and also cross-linguistically), which appears to reflect its seman-
tic development as a category.
Though these studies do not really consider issues of prototype theory,
the cognitive approach to meaning predicts that the categories expressed by
derivational and grammatical morphemes will exhibit prototype effects and
structure in a fashion comparable to lexical units. Tabakowska (1999) ar-
gues that the meanings of the Polish prefix po- form a structured network
containing extensions from three related prototypes, and Tabakowska
(2003) analyzes za- in a similar manner. Tabakowska’s analyses are con-
cerned only with synchrony, but the network structure based on extensions
from prototypes that she hypothesizes for Polish prefixes should have con-
sequences for analyses of diachronic development as well. Two recent
studies of prototype theory in its diachronic application are relevant in this
regard. Kemmer (1992) discusses the implications of (cross-linguistic)
prototypicality for the development of middle markers, and Geeraerts
(1997) discusses in detail the prototype effects (in a broad sense of the
term) affecting the semantic development of lexical units. A cognitive ap-
proach predicts that the semantic development of Slavic perfectivizing
prefixes (as derivational affixes) will display prototype effects similar to
those outlined by Kemmer and Geeraerts.
This paper applies principles of prototype theory to a diachronic analy-
sis of the meanings expressed by the Russian verbal prefix po-. Russian, as
a Slavic language, disposes of an aspect system consisting largely of pairs
of derivationally related imperfective (impf) and perfective (pf) verbs (for
a brief description of Slavic derivational aspect systems, see Dickey 2000:
7–12). Many such pairs are created by adding a perfectivizing prefix to a
simplex impf verb (e.g., stroit' (impf) – postroit' (pf) ‘build’), and among
the Russian perfectivizing prefixes po- has generally been recognized as
the most productive (cf., e.g., Čertkova 1996: 123–124 and Tixonov 1998:
A prototype account of the development of delimitative po- in Russian 331
36). Therefore, we are justified in assuming that the development of the

prefix po- was of significance for the development and grammaticalization
of the impf : pf aspect opposition in Russian (and some other Slavic lan-
guages). Though recent work on the grammaticalization of Russian aspect,
primarily Bermel (1997) and Nørgård-Sørensen (1997), has greatly en-
hanced our understanding of the spread of the aspect opposition in Russian,
it has not analyzed the active role of individual prefixes in creating the
system. Moreover, some studies focusing on the development of prefixa-
tion in the history of Slavic aspect, e.g., Strekalova (1968), inexplicably
fail to treat the development of po- in any detail. It is primarily for these
reasons that the present analysis is offered. However, another important
motivation is to show that the application of principles of cognitive linguis-
tics in an analysis of linguistic units and categories as typologically un-
usual as those involved in Slavic aspectual systems (on the typological
oddity of Slavic aspect, see Dahl 1985: 85) proves useful and allows us to
make sense of facts which have previously defied explanation.
The primary focus of this analysis is the class of DELIMITATIVE verbs,
i.e., perfective verbs prefixed in po- that express the indefinite (usually
relatively brief) duration of an action, e.g., posidet' ‘sit for a while’, poi-
grat' ‘play for a while’. These verbs, while perfective, differ markedly
from “ordinary” Russian perfective verbs in that they are atelic (cf., e.g.,
Flier 1985, Mehlig 1994, Sémon 1986, Zaliznjak and Šmelev 2000, Janda
2007): whereas napisat' means ‘write something to completion’, and the
parameters of the object being written impose an inherent limit (telos) on
the writing event beyond which it cannot continue, popisat' means only ‘be
engaged in writing for some indefinite period of time’ (no telos/inherent
endpoint of the writing event is profiled).1 (For a thorough discussion of
the aspectual differences between telic perfectives and delimitatives, the
reader is referred to Mehlig 2006.) The main function of delimitative verbs
is the coding of activity situations as pf in (narrative) sequences of events,
as illustrated in (1):
(1) Dva šara vošli v luzu, tretij pokrutilsja i zamer

Two balls entered into pocket, third PO-spun and stopped
na kraju
on edge
[Sloj 1: 21]
‘Two balls went into the pocket; the third spun a little and stopped
on the edge.’
If we recall that, as Hopper (1979) points out, one of the main functions
of pf verbs cross-linguistically is to sequence events in narratives, then it
becomes clear that by expressing a perfective construal of atelic activities,
delimitatives in fact perform a crucial systemic function in Russian – the
extension of the aspect opposition to a whole class of atelic verbs, i.e.,
Vendler’s activities.2 However, despite this crucial function, delimitative
verbs have not traditionally been considered pf “partner” verbs of their
source verbs on a par with ordinary telic pf verbs. Thus, while napisat'
‘write to completion’ is considered to be the pf “partner” verb of pisat'
‘write’, delimitative popisat' ‘write for a while’ has not traditionally been
considered to be paired with impf pisat' ‘write’ when the latter is construed
as an open-ended activity. Rather, delimitative popisat' ‘write for a while’
is considered to be an unpaired pf procedural verb, i.e., a special pf verb
which profiles some specific temporal configuration of a situation, but
which does not express the canonical, completed version of the situation.
And yet it is clear that delimitatives in po- are more important to the
system of Russian aspect than most other kinds of procedural verbs. For
this reason, recent treatments have recognized that delimitatives have a
special status among the types of Russian procedurals. Mehlig (2006) ob-
serves that po- delimitatives are the type of pf procedural that is most pro-
ductively used to perfectivize atelic impf verbs. Some treatments have even
allowed that they are “pf partner verbs” of their source verbs when the
action expressed by the source verb is construed as atelic (cf. in this regard
Čertkova 1996, Petruxina 2000: 187, Dickey and Hutcheson 2003, and
Dickey 2006). Given the current importance of po- delimitatives for the
Russian aspectual system, solving the puzzle of the development of delimi-
tative po- takes on a particular significance in a historical analysis of the
Russian aspectual system.
The view taken here is that the meanings of the prefix po- form (and
have formed) a family resemblance category (cf. Taylor 1995 and Geer-
aerts 1997), and that the semantic development of po- evident from histori-
cal sources indicates that there has been a shift in its prototypical center. In
particular, I argue that the category expressed by po- has shifted from a
cluster of meanings centered around a prototype of resultativity (based on
its spatial PATH/SURFACE-CONTACT meaning – see section 3) to a proto-
typical cluster centered around indefinite temporal delimitation. While this
initial point may be argued fairly easily on a descriptive level, the details of
the shift in the network prototype(s) are more complex than it would ap-
pear at first glance. This paper discusses the shift to delimitativity as the
semantic prototype of po- and offers a perhaps unexpected hypothesis on

the origin of this development.
2. Data and basic analysis
Let us take as a point of departure the statistical data on the semantic de-
velopment of po- given by Dmitrieva (1991: 71), presented here in figures
1 and 2. Dmitrieva, adopting Isačenko’s (1960: 209–309) classification,
classifies po- verbs into the following types: (1) resultatives, e.g., postroit'
‘build’, (2) delimitatives, e.g., postojat' ‘stand for a while’, (3) ingressives,
e.g., poletet' ‘fly/take to flight’, (4) distributives, e.g., pobrosat' ‘throw [all
of]’, (5) attenuatives, e.g., poostyt' ‘cool off somewhat’, (6) intermittent-
attenuative verbs, e.g., pobrasyvat' ‘throw from time to time/on and off’.
The resultatives, delimitatives, ingressives, distributives and attenuatives
(types 1–5) are all pf verbs; the intermittent-attenuative verbs (type 6) are
imperfectives, as they also contain the imperfectivizing -yva- suffix. As the
intermittent-attenuative verbs gain their meaning from the combination of
attenuative (or delimitative) po- with the impf suffix -yva-, they do not
represent a distinct sense of po- and are not considered in this analysis
(though I have left them in figures 1 and 2 to give an idea of their produc-
tivity). This leaves us with five distinct senses of po-: resultativity, delimi-
tativity, ingressivity (restricted primarily to determinate motion verbs3),
distributivity (either object- or subject-oriented), and attenuativity. Before
discussing the elements of each sense relevant to the analysis, let us first
examine figures 1 and 2 (taken from Dmitrieva 1991: 71) to get a basic
picture of the semantic development of po- since the eleventh century.
Dmitrieva’s data are based on the largest Old Russian (ORus) and Mod-
ern Russian (MRus)4 dictionaries available at the time of her research, the
MSDJa and SSRLJa (respectively). Though the SSRLJa contains many
more entries than the MSDJa, we may consider a comparison of the respec-
tive percentages of the different senses of po- in each dictionary to be a
fairly accurate reflection of the changes in the semasiological structure of
the prefix since the eleventh century. A comparison of Dmitrieva’s data
reveals that po- has undergone a significant shift in the relative frequencies
of its senses from ORus to MRus: in ORus, 73.5% of all po- verbs were
resultatives, whereas in MRus the resultatives are only 26% of all po-
verbs. On the other hand, in ORus delimitatives comprised only 3.8%,
compared to 31.8% in MRus. Regarding the statistics for delimitative verbs
in MRus, it must be emphasized that, as Isačenko (1962: 391–392) points
out, delimitative po- “is so productive that even the most comprehensive
dictionaries register only a small fraction of the delimitatives that actually
occur.” Thus, the 31.8% percentage for po- delimitatives in MRus should
be considered fairly conservative.
Intermittent-
Attenuative: 8.9%
Attenuative: 0.5%
Distributive: 9.1%
Ingressive: 4.9%
Delimitative: 3.8%
Resultative: 73.5%
Figure 1. Types of verbs prefixed with PO- in Old Russian (Dmitrieva 1991: 71)
Intermittent-
Attenuative: 9.2%
Resultative: 26%
Attenuative: 11.9%
Distributive: 19%
Ingressive: 2.1% Delimitative:

31.8%
Figure 2. Types of verbs prefixed with PO- in Modern Russian (Dmitrieva 1991:
71)
A comparison of figures 1 and 2 reveals that from the eleventh to the

nineteenth centuries there was a 64% decrease in the relative productivity
of resultative po-, whereas the productivity of delimitative po- increased by
a whopping 836% (again, the latter is a conservative estimate). It must also
be stressed that the delimitatives constitute the single largest group of po-
verbs in the MRus data. The other senses of po- changed less dramatically
(or at least with fewer consequences for the overall picture), and are in any
case not of primary importance to this analysis: distributives increased by
almost 110%, from 9.1% to 19%; attenuatives increased by 2,280%, from
0.5% to 11.9%; ingressives decreased from 4.9% to 2.1%. Thus, the overall
picture is that po- has shifted from a predominantly resultative prefix to a
delimitative prefix.5
This shift does not appear to have been a gradual one that spanned the
entire historical period of Russian (i.e., from the eleventh century onward).
Dmitrieva (2000: 32) observes that the 3.8% figure for po- delimitatives
characterizes the situation through the seventeenth century, adding that in
the eighteenth century the figure jumps to 18% and reaches the much
higher figure of 31.8% within another century. Dmitrieva’s data confirm
the findings by Sigalov (1975), who concludes that delimitative po- was
only marginally productive before the seventeenth century, when it began a
period of increasing productivity that continues into the present.6 Sigalov
(1975) also recognized the shift in terms of relative productivity from re-
sultative po- to delimitative po-. Thus, the question has been how to ex-
plain the sudden and dramatic semantic shift. The following analysis sug-
gests an answer to this question by treating the individual meanings of po-
in terms of prototypically clustered categories.
3. The semantics of po- in Old Russian
In order to examine the semantic categories expressed by po- in the history

of Russian, we should first briefly consider its original spatial senses dat-
ing back to Common Slavic. Němec (1954) hypothesizes that po- originally
had three etymologically distinct spatial meanings – ablativity, locativity
and allativity, which correspond to SOURCE, PATH, and GOAL (respectively)
in Shull’s (2003) system. ORus po- probably originally expressed all three
meanings. However, at a relatively early time SOURCE must have already
been fairly weak, and the verbs that originally expressed this meaning be-
gan to profile the inception of the event in time, e.g., poiti ‘go, set out’
(there were also a few ingressive verbs in po- with no original spatial
meaning, e.g., pozveniti ‘begin to ring’). This left two primary spatial tra-
jector-landmark configurations: the GOAL configuration (i.e., movement
toward a landmark; cf., e.g., ORus postignuti ‘reach’ and povyknuti ‘get
used to’ – note that in the latter the motion is metaphorical), and, more
importantly, the PATH configuration (i.e., movement or location along a
path or surface; cf., e.g., ORus poplavati ‘roam/wander [an area]’ and po-
voditi ‘lead about’). The PATH configuration was clearly dominant in
ORus, and was closely associated with mere contact with some surface, cf.
Shull’s (2003: 43, 160–161) remarks on the redundancy between these two
notions (as well as that of motion through space); by historical times, PATH
had given rise to a productive SURFACE-CONTACT meaning that was very
resultative in nature, cf., e.g., ORus posmoliti ‘cover with resin’. It should
probably be assumed that in the original situation PATH and GOAL were
distinct local prototypes in the network of po-. However, the
PATH/SURFACE-CONTACT meaning was a very telic one, i.e., it tended to
profile the complete affectedness of the surface in question, so that pop-
lavati meant ‘roam all over [an area]’, and posmoliti meant ‘cover [com-
pletely] with resin’. Thus, PATH/SURFACE-CONTACT could produce resulta-
tive verbs by metaphor on a par with the GOAL configuration. Inasmuch as
GOAL was still a distinct meaning of ORus po-, it was already weakening,
which is evidenced by the parallel demise of the GOAL preposition in his-
torical times: po in the spatial meaning of ‘up to’ was replaced by do, and
po in the goal-oriented meaning of ‘for/to get’ was replaced by za. Note
that allative verbs such as postignuti ‘reach’ easily allow an alternative
construal as goal-oriented PATH situations, inasmuch as reaching necessar-
ily involves motion along some path during the approach (it should also be
pointed out that MRus has also replaced the prefix po- in the spatial GOAL
meaning, cf., e.g., dostič' ‘reach)’. I therefore assume that the prototype of
po- in ORus was the PATH/SURFACE-CONTACT meaning, which by meta-
phorical extension had produced a large class of ordinary resultative verbs,
e.g., ORus postaviti ‘erect [on a surface]’, as well as a smaller number of
specifically distributive verbs, e.g., ORus povoziti ‘transport [all of]’. The
po- distributives appear to have originally been a direct metaphorical ex-
tension of the SURFACE-CONTACT meaning: the meaning of covering a full
space is transferred to the quantitative domain whereby the objects (or
subjects) are conceptualized as points along a surface which are covered or
encompassed by the predicate in question (cf. Dmitrieva 2000).7 Though
the central, prototypical meaning of po- was PATH/SURFACE-CONTACT, this
meaning was compatible with and linked to the other two meanings, GOAL
and (the remnants of) SOURCE, by a shared abstract schema (in terms of
Langacker 1987) of the TRANSITION FROM AN INITIAL STATE (S1) TO A RE-

SULTANT STATE (S2), which is in fact shared by all SOURCE, PATH and
GOAL prefixes, cf. Shull (2003: 184).
By historical times, the PATH/SURFACE-CONTACT sense of po- had pro-
duced another metaphorical extension, the delimitative sense, e.g., ORus
posěděti ‘sit for some time’: the metaphorical transfer of PATH/SURFACE-
CONTACT to the one-dimensional domain of time produced the meaning of
a predicate continuing for a length of time (cf. again Dmitrieva 2000, who
suggests basically the same meaning transfer but without reference to
metaphor). As Sigalov (1975) observes, such ORus delimitative verbs
comprised a very small class (consisting primarily of some basic stative
activity predicates8 such as poležati ‘lie for some time’, posěděti ‘sit for
some time’, pos"pati ‘sleep for some time’, etc.), and were quite old, being
in fact inherited from Common Slavic. Sigalov makes the important point
that these older delimitatives expressed two distinct senses of delimitativ-
ity: they could express both a relatively long duration, i.e., the continuation
of the predicate for some entire specified period of time, as shown in (2a),
(2b) and (2c), or a relatively short and usually indefinite duration (the latter
of which is familiar from MRus usage), as shown in (2d) and (2e).
(2) a. Po svjatom" kreščenii požive blaženyj knjaz' Volodimer"

After holy baptism PO-lived blessed prince Volodimer
28 lět".9
28 years
[ORus; 11th cent.; BLDR 1: 368]
‘After his holy baptism the blessed prince Vladimir lived 28
years.’
b. […] i poživ"ši že ej v" dobrě

[…] and having PO-lived emph. part. her.DAT in good
ispovědanii lěta mnoga, s mir"m" us"pe.
confession years many, with peace fell asleep
‘[…] and having lived many years in pious confession, she passed
away in peace.’
c. I posědě Kyevě na otni stolě 13 lět" i v

And PO-sat Kiev.LOC on father’s throne 13 years and in
lětě 6633 ot načala miru prestavisę.

year 6633 from beginning of world, passed away.
[ORus; 12th cent.; MSDJa 2: 1282]
‘And he sat in Kiev on his father’s throne for 13 years and in the
year 6633 after the beginning of the world he passed away.’
d. […] malo vrěmę poživ" povelěnyem' c(s)rkym" pride

[…] short time having PO-lived decree.INST royal came
v kostęntin" grad".
to Constantine’s city.
[ORus; 11th cent.; SDJa 6: 566]
‘[…] having lived a short while according to the royal decree, he
came to Constantinople.’
e. I posěděv" Davyd" malo, reče: “Gdě est'

And having PO-sat David little.ADV, said: “Where is
brat"?”
brother?”
‘And David sat for a little, and said: “Where is my brother?”
Examples such as (2a), (2b) and (2c) are much different from (2d) and
(2e), in that the extension of the predicate over the relatively long duration
in each example represents, as Sigalov (1975: 152) suggests, a resultative
sense of these po- delimitatives. Though the point is hard to argue conclu-
sively, in my view examples such as (2a) and (2c) represent telic predi-
cates, in the same way that walk a mile is a telic accomplishment predicate
as opposed to the atelic activity walk. Accordingly, (2a) and (2c) contain
telic predicates in the sense that the end of the specified intervals of time
(28 years and 13 years) are the endpoints of the respective predicates. In
any case, the predicates in (2a), (2b) and (2c) have the resultative feel men-
tioned by Sigalov (loc. cit.), due to the fact that in each case the duration of
the predicate reaches the maximum possible in the context, i.e., based on
the narrator’s knowledge of the world the predicates could not have con-
tinued longer than the specified period of time. It is for this reason that
ORus požiti ‘live for some time’ is used so often to sum up the duration of
rulers’ or saints’ lives preceding mention of their death, as in (2a) and (2b),
so that a major meaning of ORus požiti is something akin to ‘live the rest
of one’s days’ (this is probably why Sreznevskij’s MSDJa 2: 1082 defines
it first as MRus prožit' ‘live for a specified, long period of time’ and only
subsequently as požit' ‘live for some time’). Note that posědě 13 lět" ‘sat
for 13 years’ in (2c) is the exact same kind of usage, summing up the re-
mainder of the prince’s life before his death. Such usage is in clear contrast
with (2d) and (2e), where the delimitation of the predicate is clearly arbi-
trary, i.e., the predicates in (2d) and (2e) could have easily continued
longer than ‘just a little while’. It is because of this quality of arbitrarily
short delimitation that examples such as (2d) and (2e) have a much less
resultative feel (again, per Sigalov loc. cit.), and accordingly can hardly be
considered telic.10
The analysis that follows relies on the distinction between these two
senses of ORus delimitatives. Let us call the resultative/telic delimitativity
occurring in cases of a relatively long period of time ABSOLUTE DURATION,
and the delimitativity in cases of a relatively short and often indefinite pe-
riod of time RELATIVE DELIMITATION. ABSOLUTE DURATION usually in-
volves a precisely specified interval of time, such as ‘28 years’ in example
(2a), and, depending on the predicate, the interval may be longer than the
situation typically lasts; in any case, ABSOLUTE DURATION represents either
a relatively very long period of time and/or the maximum duration for
which the predicate could continue in the context. RELATIVE DELIMITA-
TION involves a relatively short interval of time for the predicate in ques-
tion, or one that is left indefinite, and perhaps even unspecified in which
case it is by default an interval of time typical for an episode of the situa-
tion in question.11 Again, note that delimitatives expressing ABSOLUTE
DURATION as in (2a) in fact functioned as resultative (telic) verbs (cf. Siga-
lov loc. cit.); indeed, (2a) occurs at the end of a narrative and sums up the
remainder of Vladimir’s life before his death. Not only was Sigalov (1975)
the first to distinguish these two distinct senses of ORus delimitative po-,
but he also recognized that, whereas ORus delimitatives could express both
ABSOLUTE DURATION and RELATIVE DELIMITATION, MRus delimitatives
have lost their ability to express ABSOLUTE DURATION, as shown by the
MRus translations of (2a), (2b) and (2c):
(3) a. Posle svjatogo kreščenija prožil blažennyj knjaz' Vladimir

After holy baptism PRO-lived blessed prince Vladimir
dvadcat' vosem' let.
twenty eight years
[MRus; = (2a); BLDR 1: 325]
‘After his holy baptism the blessed prince Vladimir lived 28
years.’
b. […] i, proživ mnogo let v iskrennem

[…] and, having PRO-lived many years in sincere
pokajanii, mirno skončalas'.
repentance, peacefully passed away
[MRus; = (2b); BLDR 1: 369]
‘[…] and, having lived many years in sincere repentance, passed
away peacefully.’
c. I v Kieve on prosidel na otcovskom prestole 13 let,

And in Kiev he PRO-sat on father’s throne 13 years,
i v 6633-om godu s načala sveta skončalsja.
and in 6633rd year from beginning of the world passed away
[MRus; = (2c)]
‘And in Kiev he sat out 13 years on his father’s throne, and in the
6633rd year since the beginning of the world passed away.’
Since the seventeenth century, Russian has developed an entire separate

class of PERDURATIVE verbs prefixed in pro- solely to express ABSOLUTE
DURATION; these perdurative verbs in pro- completely took over this func-
tion from old delimitative po- from around the seventeenth century (cf.
Sigalov 1977). This is why in MRus translations of ORus literature the
ORus po- delimitatives expressing absolute duration are regularly rendered
with perduratives in pro-, as shown in the examples in (3).12 The loss of the
ability of Russian delimitatives to express ABSOLUTE DURATION is a pecu-
liar development and is of crucial importance for this analysis, as will be
shown below.
Let us now turn to ORus PATH/SURFACE-CONTACT po- in combination
with determinate motion verbs. When PATH/SURFACE-CONTACT po- was
affixed to such verbs, e.g., ORus iti ‘go’, the original results were verbs
with a prominent ablative sense, e.g., ORus poiti ‘go [away], go forth’. I
consider it unnecessary to assume a distinct SOURCE meaning of po- in
ORus, for the following reasons. As Radden (1988: 382–383, cited in Is-
raeli 2002) has observed, the notion go (ORus iti) inherently focuses on the
spatial SOURCE of a trajectory. Moreover, since all motion occurs along
some path/in contact with some surface (cf. Shull 2003: 160–161), the
prefixation of determinate motion verbs (which by default profile move-
ment away from some source) with PATH/SURFACE-CONTACT po- did not
typically create verbs profiling the complete traversal of some path. Rather,
such verbs more often expressed a vaguely ablative sense that appears to be
almost indistinguishable from the expression of PATH/SURFACE-CONTACT,

as shown in (4).13
(4) […] proide v" vust'e dněpr'skoe, i ottole

[…] passed into mouth Dnepr.ADJ and from there
poide po Dněpru gorě.
PO-went along Dnepr upstream
[12th cent.; BLDR 1: 66]
‘[…] he passed into the mouth of the Dnepr, and from there he
went along the Dnepr upstream.’
Example (4) clearly profiles motion along a particular path (notably

with the PATH preposition po). Yet in the narrative it does not profile a
complete traversal of the subject’s path; its sense is more equivalent to ‘set
out/went off’. At some relatively early point in time, and it is not clear
when, the originally spatially ablative (i.e., SOURCE) sense of po- with mo-
tion verbs began to “bleach out” due again to the redundancy of the
PATH/SURFACE-CONTACT meaning of po- with iti ‘go’, so that ORus poiti
‘go forth’ began develop into an INGRESSIVE verb profiling the initial
phases of the motion in time, i.e., ORus poiti ‘go [initial phases]/start go-
ing’, cf. the intuition of native speakers of MRus that pojti ‘go’ profiles the
initial phases temporally, i.e., the inception of the motion as opposed to
spatial ablativity (which is expressed by otojti and ujti, each ‘go away’
with different nuances). Note that while the above example almost cer-
tainly profiles spatial ablativity, it also readily allows the ingressive inter-
pretation, which demonstrates how easily the metonymic development
from ablativity to ingressivity could take place.
In the case of ablative/ingressive poiti, the redundancy of
PATH/SURFACE-CONTACT po- with iti ‘go’ in ORus results in a paradigm
case of the semantic overlap between a prefix and its base verb (known in
the Slavic aspectological literature as SUBSUMPTION, cf. Nübler 1990 and
the references cited there), thus not only paving the way for the original
spatial meaning of the prefix to bleach out, but also producing an aspect
pair of lexically identical verbs iti (impf) – poiti (pf), in which po- is a
“lexically empty” marker of perfectivity in the prefixed correlate poiti (cf.
Mayo’s 1984: 30–32 view that in seventeenth-century Russian iti – poiti
was an aspect pair).
However, while I think it is correct to consider ORus iti – poiti ‘go’ an
aspect pair from around the sixteenth or seventeenth century, it is clear that
poiti (as well as MRus pojti) is not a prototypical telic pf verb (i.e., it does
not profile the attainment of an inherent limiting endpoint, in contrast to

MRus napisat' ‘write to completion’, etc.).14 So what pf meaning did po-
express in pf poiti? Following Shull’s (2003: 153) analysis of MRus pojti, I
assume that post-sixteenth century ORus poiti acquired a meaning closely
related to (yet distinct from) ingressivity, namely the initiation of a new
trajectory and traversal of some amount of that trajectory greater than
zero, the schema of which is shown in figure 3.
TR
TRy LM (background TRy)

TR = trajector, TRy = trajectory, LM = landmark
Figure 3. Shull’s (2003: 153) schema for Russian PO- with motion verbs
According to Shull’s synchronic schema for Russian po-, the landmark

of the schema is the full trajectory denoted by the base verb in the context,
as opposed to an external (physical) landmark, i.e., the trajectory that po-
expresses is the traversal of a portion of a known trajectory in space (which
itself is the landmark). This unusual configuration of the landmark and
trajectory has theoretical and descriptive motivations that need not concern
us here (for details, cf. Shull 2003: 147–180), but it is useful, for instance,
to point out that the shared knowledge of the full trajectory in a context
explains why someone can say Ivan uže pošel ‘Ivan has already gone’ only
if both interlocutors know Ivan’s intended final destination (this is in con-
trast to Ivan ušel ‘Ivan has gone away’, which requires no such knowledge
of the full trajectory, cf. Shull 2003: 154). It is also important to note that
Shull considers the above schema to be a spatio-temporal concept that is
applicable to both domains, and not primarily to space. According to
Shull’s formulation for MRus po-, while the trajectory profiled by po- is
always greater than zero, it is specified only vaguely and may be equal to
or less than the background trajectory. In order to assert a complete tra-
versal of the background trajectory explicitly, MRus employs projti ‘go
[from point A to point B]’.
With regard to this last point, it must be emphasized that the situation in
ORus differed from MRus to some extent. While the prefixation of deter-
minate motion verbs with PATH/SURFACE-CONTACT po- usually produced

an ablative/ingressive sense neutral for the notion go, in ORus poiti occa-
sionally functioned to express the complete transversal of some trajectory.
This is shown in the following ORus example, whose MRus equivalent
must employ projti:
(5) a. A xodjat na goru den' po odnomu čelověku:

And walk.3PL on mountain day by one man:
doroga tesna, a dvema poiti nelzě.
road narrow, and two.DAT PO-go impossible
‘And they walk a day up the mountain one by one: the road is nar-
row and two cannot pass.’
b. I puti na tu goru den', xodjat po odnomu

And trip.GEN onto that mountain day, walk.3PL by one
čeloveku: doroga uzka, dvoim projti nel'zja.
man: road narrow, two.DAT PRO-go impossible
[MRus; = (5a); BLDR 7: 355]
‘And it is a day’s trip up the mountain, they walk one by one: the
road is narrow and two cannot pass.’
In (5a) the full relevant trajectory is passing by another man along the
path; notice that an ablative/ingressive interpretation is completely inap-
propriate here. Perhaps a more illustrative example, albeit with another
lexical verb, poplyti ‘swim’, is given with its MRus translation in (6):
(6) a. Osen'ju, poslě Semenja dni, poplyl muž' moj

Autumn.INST, after Semen’s day, PO-swam husband my
Maksim" s" ust' Vitima rěki na ust' Olekmu.
Maksim from mouth Vitim.GEN river to mouth Olekma.ACC
[ORus; 17th cent.; SRJa XI–XVII 17: 89]
‘In the autumn, after Semen’s day, my husband Maksim swam
from the mouth of the Vitim River to the mouth of the Olekma.’
b. Osen'ju, posle Semenova dnja proplyl muž' moj

Autumn.INST, after Semen’s day PRO-swam husband my
Maksim iz ust'ja reki Vitim v ust'e reki

Maksim from mouth river.GEN Vitim to mouth river.GEN
Olekma.
Olekma
[MRus; = (6a)]
‘In the autumn, after Semen’s day, my husband Maksim swam
through from the mouth of the Vitim River to the mouth of the
Olekma.’
For ease of reference, Shull’s hypothesized schema for MRus po- with
motion verbs shall be termed INGRESSIVE-PARTIAL TRAJECTORY, and the
meaning expressed by MRus pro- with motion verbs shall be termed FULL
TRAJECTORY.
The situation regarding PATH/SURFACE-CONTACT po- with determinate
motion verbs thus shows an interesting parallel with the older delimita-
tives, which should not be overlooked. Just as ORus stative delimitatives in
po- could express two kinds of extension in time, ABSOLUTE DURATION
and RELATIVE DELIMITATION, ORus determinate motion verbs in po- could
express both FULL TRAJECTORY and INGRESSIVE-PARTIAL TRAJECTORY.
The only noteworthy difference lies in the fact that with delimitatives the
sense of ABSOLUTE DURATION was just as frequent as RELATIVE DELIMITA-
TION, whereas poiti most often expressed INGRESSIVE-PARTIAL TRAJEC-
TORY (which is expected, given that the default profile of go is the source
of motion), and only occasionally expressed FULL TRAJECTORY. The se-
mantic parallel between ORus determinate motion verbs in po- and delimi-
tatives in po- is summarized in table 1.
Let us sum up what we have established so far: ORus po- represented a
network originally consisting of a salient PATH/SURFACE-CONTACT mean-
ing, which was the central prototype of the prefix prior to the latter half of
the seventeenth century. The PATH/SURFACE-CONTACT sense of po- had as
metaphorical extensions salient meanings of resultativity (e.g., postaviti
‘build’) and distributivity (e.g., pometati ‘throw [all of]’). Another meta-
phorical extension evident in a small group of stative activity verbs was the
delimitative meaning (e.g., posěděti ‘sit for some time’); delimitativity had
two variants, ABSOLUTE DURATION and RELATIVE DELIMITATION. Lastly,
due to their particular spatial default meaning, determinate motion verbs
prefixed with PATH/SURFACE-CONTACT po- had an ingressive meaning, i.e.,
INGRESSIVE-PARTIAL TRAJECTORY, which can nevertheless be considered
an extension of the central PATH/SURFACE-CONTACT meaning; such verbs
could also express FULL TRAJECTORY.
Table 1. The semantic parallel between ORus determinate motion verbs and delimi-
tatives in PO-
PO- LIMITED MAXIMAL CONCEPTUAL
VERB TYPE TRAJECTORY TRAJECTORY DOMAIN
Determinate Ingressive- Full Trajectory Motion in Space
Motion Partial Trajectory
Verbs
Delimitatives Relative Absolute Duration Duration in Time
Delimitation
4. From resultative po- to delimitative po-: the shift in the prototype
In considering the issue of how the predominant meanings of resultativity

(as well as distributivity) gave way to a hyperproductive delimitative sense,
Sigalov (1975) and Dmitrieva (2000) emphasize the fact that most ORus
verbs in po- originally had resultative meanings, and subsequently lost
these meanings as they acquired the new delimitative meaning. For exam-
ple, ORus povoevati meant ‘conquer, seize militarily’, and then lost this
meaning around the eighteenth century, when it began to be attested in the
meaning of ‘soldier, wage war [for some indefinite period of time]’, and
ORus popisati meant ‘write to completion’, whereas in MRus popisat'
means ‘write for some time’. While this fact is clearly essential to solving
the puzzle, neither Sigalov nor Dmitrieva offers a satisfactory explanation
of how in fact po- developed from a primarily resultative prefix to a pri-
marily delimitative one. Since the resultative sense of po- preceded its
delimitative sense in many verbs, Sigalov (1975: 169) makes the muddled
suggestion that the “[temporally] delimitative circumstance” concomitant
to the construal of an action as complete and having produced a result was
“absorbed” by stative verbs prefixed with po- (e.g., požiti ‘live for some
time’; cf. examples 2a and 2b) and then “became” the meaning of the pre-
fix. In other words, Sigalov assumes that the alternation of ORus delimita-
tive verbs between a telic function (the expression of ABSOLUTE DURA-
TION) and an atelic function (the expression of RELATIVE DELIMITATION)
allowed the originally telic prefix po- to be reanalyzed as an atelic prefix
expressing primarily RELATIVE DELIMITATION. Dmitrieva simply asserts
that the SURFACE-CONTACT meaning allowed verbs in po- to begin to pro-
file not the entire trajectory to its end, but some subset of it.
I consider Sigalov’s and Dmitrieva’s explanations of the rise of the new

delimitatives expressing RELATIVE DELIMITATION to be unsatisfactory in
three important (and interrelated) respects. First, everything else being
equal I consider it very doubtful that in Russian, a language in which
telic/resultative verbal prefixation has thrived (to say the least), verbs con-
taining a resultative/telic prefix would simply lose their meaning of resulta-
tivity. I do not deny the possibility of reanalysis in principle; however, no
other Russian prefix has lost its telic meanings in this way. Second, the
suggestion that a prefixed verb simply transforms its aspectual meaning
from resultativity to delimitativity is problematic in that it lacks any kind
of cognitive, communicative or systemic motivation whatsoever. Third,
inasmuch as Sigalov’s hypothesis relies on the occasional meaning of
RELATIVE DELIMITATION of ORus stative delimitatives, it is highly ques-
tionable whether a few stative delimitatives would have the kind of linguis-
tic salience necessary to serve as tokens in an aspectual reanalysis whereby
a whole class of verbs such as povoevati ‘conquer’ loses its resultative
sense in favor of delimitativity.
Let us now consider the semantic change described above from the per-
spective of prototype semantics to see if we can make any sense of it.
Geeraerts (1997) argues convincingly that the polysemy inherent in proto-
typical categories allows for a straightforward account of semantic change:
prototypical categories change naturally by expanding, i.e., by gradually
being used to refer to referents that show lower and lower degrees of re-
semblance to the prototype (cf. Geeraerts’ case of Dutch legging ‘tights’).
At some point, a new meaning of the category may acquire such salience
and linguistic frequency that it either becomes a local prototype in the net-
work or also eventually supplants the original prototype as the category
center. Such cases of prototype shift are discussed by Kemmer (1992) re-
garding reflexive morphemes: a common category extension of reflexive
morphemes is their gradual spread to middle verbs, whereupon they either
retain or lose their original reflexive meaning. In the latter case, middle
voice becomes the prototype of the old reflexive marker. In accordance
with Geeraerts’ and Kemmer’s findings, I think it is advisable to seek the
change giving rise to the new delimitatives in an expansion of the category
expressed by po- to incorporate new meanings.
The network of ORus po- was organized around the PATH/SURFACE-
CONTACT meanings as its prototypical center. A small class of delimita-
tives existed from the outset. So on the face of it there appears to be no
categorial expansion whereby new concepts came to be expressed by po-.
Though prototype theory allows us to make sense of the development of
delimitatives on some basic descriptive level, we are left without any ap-
parent motivation for the shift in the prototype. This problem is com-
pounded by the fact that the new prototype, delimitativity, also inexplica-
bly loses the ability to express one of its senses, ABSOLUTE DURATION,
leaving MRus delimitatives able only to express RELATIVE DELIMITATION.
It is important to point out that this is an unexpected situation: as Dickey
and Hutcheson (2003) observe, in some other Slavic languages, e.g.,
Czech, the existing delimitatives have retained their ability to express AB-
SOLUTE DURATION, which is in clear contrast with MRus, as shown in (7a)
and (7b), taken from Petruxina (2000: 164).
(7) a. Celý večer poseděli, až zvonili na půlnoční. [Czech]

Entire evening PO-sat.3PL, until rang.3PL for midnight mass
‘They sat the entire evening, until they rang the midnight mass.’
b. Oni prosideli/*posideli celyj večer, poka ne zazvonil

They PRO-sat/*PO-sat entire evening, until rang
kolokol, opoveščaja o polunočnoj messe.
bell, announcing about midnight mass
[MRus; = (7a)]
‘They sat through the entire evening, until the bell rang, announc-
ing the midnight mass.’
Thus, the core problem is really this: how and why did the ORus delimi-
tatives lose the ability to express ABSOLUTE DURATION around the seven-
teenth century?
But for this problem, we can easily analyze the spread of the delimita-
tive meaning of po- as a case of semantic change from a subset, as outlined
by Geeraerts (1997: 68–79). For example, English meat originally meant
‘food’, with the meaning of ‘meat’ as a subset; gradually the meaning
‘meat’ develops into the sole meaning of the word. Similarly, Dutch winkel
has come to mean ‘shop’ in the following way. Originally winkel meant
‘corner’, which includes ‘street corner’ as a subset. From the subset ‘street
corner’ a metonymical extension produces the meaning ‘building located
on a street corner’, which in turn includes ‘shop located on a street corner’
as a subset. The older meaning of ‘corner’ continues to exist for some time
alongside the new meaning ‘shop located on a street corner’. A process of
semantic generalization from this second subset produces the simple mean-
ing ‘shop’. We may consider the delimitative meaning of ORus po- to be
ABSOLUTE DURATION, which includes RELATIVE DELIMITATION as a subset,
which then develops into an independent meaning. But while in the case of
English meat the salience of meat as a prized source of nourishment plau-
sibly motivates the change, and in the case of Dutch winkel the change is
motivated by the communicative need for clarity in matters of commerce
coupled with the salience of shops in our cultural awareness, it is almost
impossible to imagine an analogous motivation in the abstract domain of
time that would give ORus posěděti ‘sit [RELATIVE DELIMITATION]’ the
upper hand against posěděti ‘sit [ABSOLUTE DURATION]’ to the point where
the latter disappears (along with a whole class of resultative verbs in po-)
as the former becomes a salient token in the development of a productive
new derivational pattern.
This is where ORus poiti ‘go’ with its default meaning of INGRESSIVE-
PARTIAL TRAJECTORY becomes relevant. In various languages of the world,
the motion verb go has undergone various metaphorical extensions, the
most common of which involve the nearly ubiquitous TIME IS SPACE meta-
phor, which leads to its grammaticalization as a future auxiliary, cf., e.g.,
the French aller-future and the English gonna-future (cf. Hopper and
Traugott 2003: 23, 87–90), as well as the Hausa zaa-future (cf. Abdoulaye
2001; zaa is an ingressive motion verb). I suggest that ORus poiti (along
with the other determinate motion verbs prefixed in po-, e.g., poěxati
‘ride’, poplyti ‘swim’, etc.), with its default meaning of INGRESSIVE-
PARTIAL TRAJECTORY, served as the catalyst for the development of RELA-
TIVE DELIMITATION as the prototype of po-.
The INGRESSIVE-PARTIAL TRAJECTORY expressed by ORus poiti bears a
striking resemblance to the RELATIVE DELIMITATION meaning expressed by
delimitative verbs. For example, both the INGRESSIVE-PARTIAL TRAJEC-
TORY expressed by determinate motion verbs in po- and the RELATIVE DE-
LIMITATION expressed by ORus delimitatives in po- (as well as their MRus
counterparts) profile some indefinite quantity of an action that is less than
a canonical, completed event. In other words, the aspectual meaning ex-
pressed by ORus poiti and that expressed by a verb such as ORus pobesě-
dovati ‘discuss’ (in its RELATIVE DELIMITATION meaning) are parallel: in
the case of poiti the profile is less than the complete trajectory of the path,
and in the case of pobesědovati the profile is some amount of discussion
less than a complete discussion that produces a result. As suggested in
figure 4, the difference is primarily one of cognitive domain, i.e. determi-
nate motion in space versus communicative interaction in time.
When and how would ORus poiti come to serve as a token for the de-
velopment of a new class of delimitatives? In order to answer this question,
we must examine poiti as a perfective verb. Recall that the aspectual mean-
ing of poiti, i.e. its status as a perfective of iti ‘go’, was the result of the
redundancy of the PATH/SURFACE-CONTACT meaning of po- in combination
with iti ‘go’, which had a default focus on the source (inception) of some
trajectory (which always occurs on some path). It appears that ORus iti –
poiti became linked as a kind of aspectual pair relatively late in time.
Though the process appears to have been completed by the seventeenth
century (cf. Mayo 1984: 32), in earlier centuries ORus unprefixed iti ‘go’
regularly occurred in narrative sequences of events – the canonical context
for the pf in the modern language. A representative example is (8а), cf. the
MRus translation in (8b), which requires the pf pojti:
(8) a. Ol'ga s" synom svoim" Svjatoslavom" sobra voi

Ol'ga with son her Svjatoslav collected soldiers
mnogi i xrabry, i ide na Derv'sku zemlju.
many and brave, and ø-went to Derevlian land
‘Together with her son Svjatoslav Ol'ga gathered many brave war-
riors and went to the Derevlian land.’
b. Ol'ga s synom Svjatoslavom sobrala mnogo xrabryx

Ol'ga with son Svjatoslav collected many brave
voinov i pošla na Derevskuju zemlju.
soldiers and PO-went to Derevlian land
[MRus; = (8b); BLDR 1: 107]
‘Together with her son Svjatoslav Ol'ga gathered many brave war-
riors and went to the Derevlian land.’
Strekalova (1968: 49) observes on the basis of similar data that Polish
iść ‘go’ was not integrated into the Polish aspectual system as an impf verb
until around the sixteenth century (which involved it becoming paired with
pf poiść as its default pf). The establishment of ORus iti ‘go’ as an impf
verb must have occurred at approximately the same time, and likewise
depended on the development of poiti as its pf correlate (partner verb).
Accordingly, it is most likely that the ORus aspect pair iti – poiti became
established around the sixteenth century, and no earlier than the fifteenth
century. Thus, poiti became a pf correlate of iti not long before the advent
of the new delimitatives. If we assume that ORus pf poiti was a catalyst for
the rise of the new delimitatives in the seventeenth century, the chronology
makes sense: poiti as a pf correlate of iti was available to serve as model
for the creation of a new class of atelic verbs, the delimitatives, which be-
gan in the seventeenth century; however, poiti did not exist in this capacity
much earlier, which makes it unnecessary to explain why delimitatives did
not begin their productivity earlier.
The hypothesis advocated here is that innovative pf poiti, in which po-
expressed an emerging atelic meaning of INGRESSIVE-PARTIAL TRAJEC-
TORY in its capacity as a perfectivizing prefix, played a crucial role in the
development of the new delimitatives. This hypothesis accords well with
the mechanisms of prototypical change outlined above, in that a salient
token of a peripheral meaning in a category network can only facilitate the
development or strengthening of that meaning into a local prototype or
even the central meaning of the category. Moreover, poiti, as a form of the
highly frequent default motion verb ‘go’, is about the only verb in the lin-
guistic system of Russian that would have enough salience to contribute
significantly to such a reorganization of the semantic nature of po- in this
way. The relatively late emergence of poiti as the pf correlate to iti in-
volved a new kind of meaning expressed by po-, INGRESSIVE-PARTIAL TRA-
JECTORY, which as such would be incorporated into its network in one way
or another. I suggest that the new INGRESSIVE-PARTIAL TRAJECTORY mean-
ing assumed a relatively salient position in the network of po- given its
association with the highly salient motion verb poiti. The close resem-
blance between INGRESSIVE-PARTIAL TRAJECTORY and RELATIVE DELIMI-
TATION was the point of contact in the evolving network of po- that re-
sulted in the development of the new delimitatives: INGRESSIVE-PARTIAL
TRAJECTORY and RELATIVE DELIMITATION were reanalyzed together as the
core of a nascent cluster of atelic meanings of the prefix.15 With the addi-
tional conceptual and systemic salience of poiti, it seems much more plau-
sible to analyze the development of RELATIVE DELIMITATION as a case of
Geeraerts’ semantic change from a subset. Further semantic developments
shifted the prototypical center of the prefix away from the old telic cluster
of PATH/SURFACE-CONTACT and resultative meanings to the new atelic
cluster centered around INGRESSIVE-PARTIAL TRAJECTORY. The shift pro-
ceeded through two processes: (1) the swift expansion of the predicate
types producing new delimitatives, and (2) the realignment of some older
types of po- verbs with the new prototypical center.
Sigalov (1975) documents the spread of delimitative po- at the end of
the ORus period from the small class of stative delimitatives inherited from
Common Slavic. According to Sigalov (ibid., 171), the delimitative mean-
ing spread initially to po- derivatives of indeterminate motion verbs, e.g.,
ORus poběgati ‘run for a while’, poxoditi ‘walk for a while’, and “verbs of
psychological activity”, e.g., ORus pomolitisja ‘pray for while’ and
poveselitisja ‘be cheerful for a while’, in the sixteenth and especially sev-
enteenth centuries. Afterwards, delimitatives were derived from derivatives
of verbs of speech (e.g., pogovoriti ‘talk for a while’), verbs of sound (e.g.,
poguděti ‘drone for a while’) and verbs of physical activity (e.g., pokopati
‘dig for a while’), in the seventeenth and eighteenth centuries. Dickey
(2005: 45–47) briefly analyzes the spread of delimitativity among the vari-
ous lexical classes of predicates as a process of retextualization (cf. Nich-
ols and Timberlake 1991), based on the salience of the token poiti (as well
as the other determinate motion verbs in po-) in the development of the
new class of delimitatives, though without explaining the precise relation-
ship between the two predicate types – and as a consequence, without pro-
viding a plausible semantic motivation for the development. The prototype
analysis offered here provides the plausible motivation for the change: the
coalescence of the two atelic meanings expressed by po-, the RELATIVE
DELIMITATION of the old delimitatives and the newer INGRESSIVE-PARTIAL
TRAJECTORY of poiti, combined with the salience of poiti ‘go’ as a token of
the new category of atelic perfectivity created a situation which was ripe
for the spread of the derivational model of atelic po- perfectives (i.e., de-
limitatives) to various predicate types by analogy. Once iti – poiti was es-
tablished as an aspectual pair in Russian, it is reasonable to assume that
verbs of lesser salience, i.e. activity verbs of various kinds, would begin
systematically to derive atelic pf partner verbs by analogy (or metaphorical
extension). As suggested at the beginning of this paper, it is my view that
the development of the highly productive class of atelic delimitatives con-
tributed to the grammaticalization of Russian aspect in a major way by
extending the impf : pf aspect opposition to a large class of atelic activity
predicates.16
The assumption that poiti played a role in the rise of the new po- de-
limitatives in Russian allows us to make sense of the fact that the indeter-
minate motion verbs (e.g., ORus xoditi ‘walk’, ězditi ‘ride’, etc.) produced
delimitatives in po- relatively early. It is in fact unclear to what extent in
Late Common Slavic the so-called indeterminate motion verbs really com-
prised a distinct class with specific aspectual properties, as opposed to a
loose set of relatively stative manner of motion verbs (i.e., verbs expressing
a manner of motion, e.g., walking, as an activity without focusing on any
concomitant change of position, hence the term “stative”). Vaillant (1939:
294) expresses doubts on this issue, observing that (Common) Slavic pairs
of verbs such as nesti and nositi ‘carry’ are “just as independent of one
another as [French] aller [‘go’] and marcher [‘walk’]”. Thus, it is possible
(and in my view most accurate) to view indeterminate verbs in older stages
of Slavic as stative manner of motion verbs.17 Accordingly, we may suggest

that in ORus manner of motion verbs such as xoditi ‘walk’ were originally
primarily simply stative and thus peripheral (non-prototypical) motion
verbs (gaining their status as narrowly “indeterminate” correlates of deter-
minate motion verbs relatively late). If the motion verb poiti played a sig-
nificant role in the development of the new delimitative verbs, it makes
sense that stative manner verbs of motion would produce delimitatives
relatively early, as these atelic predicates were peripheral members of the
class of motion verbs (of which determinate motion verbs in po- were
clearly prototypical members).
The hypothesis that poiti ‘go’ was a salient token in a retextualization
of po- verbs as delimitatives that occurred early in non-prototypical motion
verbs (i.e., manner of motion or “indeterminate” motion verbs) receives
circumstantial evidence from the facts regarding the development of the
prefix po- with determinate motion verbs in Czech. Some Czech impf
verbs, primarily determinate motion verbs, form a synthetic imperfective
future-tense form with the prefix po-:18 jít ‘go’ – půjde ‘s/he will go’, jet
‘go [by vehicle]’ – pojede ‘s/he will drive’, letět ‘fly’ – poletí ‘s/he will
fly’, etc. (for details, see Kopečný 1962: 46–50). According to a historical
analysis by Bondarko (1961), these impf po- future-tense forms developed
directly from the ablative sense of Old Czech po-, so in fact we are dealing
with the same formant in basically the same meaning as in our discussion
of ORus. What is interesting about the Czech impf po- future is that, ac-
cording to Kopečný (1962: 48–49), in the colloquial language it has spread
from the narrow class of determinate motion verbs inherited from Common
Slavic (cf. the above examples) not only to newer loaned motion verbs, cf.,
e.g., fárat ‘transport’ – pofárá ‘s/he will transport’ (← Ger. fahren) but
also to various other verbs that do not express motion in its prototypical
sense, cf., e.g., růst ‘grow’ – poroste ‘it will grow’, kvést ‘bloom’ – pok-
vete ‘it will bloom’, mlít ‘grind’ (cf., e.g., mlýn pomele ‘the mill will
grind’) and lepit ‘stick’ (cf., e.g., lepidlo polepí ‘the glue will stick’), or at
all, e.g., kvákat ‘squawk’ – (cf., e.g., vrány snad zase pokvákají ‘the crows
will probably start squawking again’). Kopečný (ibid., 49) notes that the
verbs that form po- futures “are not a closed class,” and suggests that the
spread of the impf po- future to non-motion verbs has been the result of
analogy. It ought to be clear that a catalyst for this analogical spread of the
Czech impf po- future from determinate motion verbs to other verbs has
been the salience/frequency of the impf future of jít ‘go’, i.e., půjde, in the
Czech language. The importance of the Czech impf po- future for the
analysis here is that it is a case of the prefix po- in combination with a de-
terminate verb of motion serving as a salient token in the spread of a mean-

ing of the prefix (in the Czech case, futurity) to other verbs. Thus, the
emergence of Russian delimitative po- and the spread of the Czech impf
future in po- appear to be parallel in the sense that they each involve the
spread of a semantic category (INGRESSIVE-PARTIAL TRAJECTORY and futu-
rity, respectively) from a narrow class of determinate motion verbs to more
peripheral members of the class of motion verbs and (in the case of Rus-
sian delimitative po-) well beyond. This makes sense, given that the origin
of each phenomenon is the Slavic prefix po- in its spatial meanings
(PATH/SURFACE-CONTACT and SOURCE, respectively, which, as pointed out
in section 3, were not necessarily distinct in the relevant period of time).
In addition to the rapid growth of delimitatives as a class at the end of
the ORus and the beginning of the MRus periods, the changes in some of
the other types of po- verbs documented by Dmitrieva (2000) can be han-
dled easily by the prototype account suggested here. First, the sharp rise in
attenuative verbs, e.g., ORus ponakazati ‘punish mildly’, is easily ex-
plained as the transfer of the new category prototype of INGRESSIVE-
PARTIAL TRAJECTORY from the spatio-temporal domain to domains rele-
vant to individual predicates (often intensity of result). Whereas the few
attenuatives existing in ORus are to be considered metaphorical extensions
of PATH/SURFACE-CONTACT (or also a semantic subset of its resultative
meaning in the same way that RELATIVE DELIMITATION was simply a sub-
category of the ABSOLUTE DURATION expressed by the old delimitative
verbs), attenuatives were reanalyzed as tokens of the new INGRESSIVE-
PARTIAL TRAJECTORY meaning of po-. Thus, combining the INGRESSIVE-
PARTIAL TRAJECTORY of po- with MRus pf ostyt' ‘cool off’ yields attenu-
ative poostyt' ‘cool off somewhat’. We may explain the fact that attenu-
ative verbs have not become as productive in Russian as in Czech by as-
suming that INGRESSIVE-PARTIAL TRAJECTORY has increasingly become a
temporal category in Russian (which accords with what we know about the
development of Russian aspect on the whole, as well as with Shull’s 2003:
153 original synchronic characterization of MRus INGRESSIVE-PARTIAL
TRAJECTORY as a spatio-temporal concept, rather than a strictly spatial
one). The class of MRus degree achievements in po-, e.g., poxudet' ‘lose
weight’, can also be incorporated in the network quite easily, in the sense
that their use to profile any degree of change at all (i.e., less than some
maximum change) as a pf change of state S1 → S2 can be identified as a
clear case of INGRESSIVE-PARTIAL TRAJECTORY.
The twofold increase in distributive verbs in po-, e.g., ORus povoziti
‘transport [all of]’, can be accounted for as well. ORus distributive po-, as
a metaphorical extension of the old telic PATH/SURFACE-CONTACT cate-

gory, ought to become a peripheral type in a network clustered around an
atelic meaning such as INGRESSIVE-PARTIAL TRAJECTORY. While this is
basically the case, the particular development of Russian distributive po-
may be interpreted as evidence that it has nevertheless been integrated
successfully into the new network. Distributive po- has not become par-
ticularly productive in the creation of distributive verbs from simplex
stems in the standard language (cf., e.g., MRus po-brosat' ‘throw [all of]’),
as this function has been taken over by pere- (pere-brosat' cf. footnote 5
above); however, po- has retained its function of deriving distributive verbs
from stems already containing a prefix, cf., e.g., po-vy-brasyvat' ‘throw out
[all of]’. Assuming that po- now has an atelic prototype, distributive po-
occurring with verbs that already contain a prefix arguably does not present
a problem, as the telicity ordinarily concomitant with resultative pf verbs is
contained in the first prefix. If vy- ‘out’ in po-vy-brasyvat' ‘throw out’ pro-
files the change S1 → S2 in a completed event of ‘throwing out’, then we
may say that po- simply distributes that predicate over an explicitly quanti-
fied set of arguments (e.g., vse knigi ‘all the books’) without necessarily
expressing any telicity itself. In other words, the (abstract) extension pro-
filed by INGRESSIVE-PARTIAL TRAJECTORY is applied to a universally quan-
tified set of objects, thus including all of them as affected by the predicate
in question. Inasmuch as this solution seems far-fetched, the prototype
model predicts that in a given network there will be cases that are more and
less representative of a category, and we can define distributive po- as a
peripheral member of the category, conceptually relatively far removed
from INGRESSIVE-PARTIAL TRAJECTORY. This must be done in any case
with the considerable number of remnant resultatives (e.g., postroit'
‘build’) as well as the distributives derived from simplex verbs (e.g., po-
brosat' ‘throw [all of]’) surviving in MRus.19 Note that many resultatives
expressing the complete covering of a surface such as ORus posmoliti
‘cover with resin’ have not survived and were replaced with verbs contain-
ing other prefixes, e.g., MRus zasmolit' ‘cover with resin’, which is a pow-
erful piece of evidence that po- did move away from PATH/SURFACE-
CONTACT as its prototype.
Two more important pieces of circumstantial evidence for the hypothe-
sis advocated here deserve attention. The first is a general correlation be-
tween the existence of determinate motion verbs in po- and the productive
derivation of po- delimitatives in the individual Slavic languages. All the
Slavic languages that have developed productive delimitatives in po- (Rus-
sian, Belarusian, Ukrainian, Polish, Bulgarian and Macedonian) also have
(ingressive) pf determinate motion verbs prefixed in po-. Bulgarian has lost

poiti ‘go’, but Middle Bulgarian did have it (cf. Lilov 1964: 110–111; note
that Bulgarian still has other ingressive motion verbs in po-, e.g., ponesja
‘carry’, pobjagna ‘run’). Note that Croatian/Serbian poći does not disturb
the picture, because it is clearly spatially ablative and not a pf partner of ići
‘go’; rather, it is the pf partner verb of impf polaziti ‘depart’. So there is a
clear correlation between the development of delimitatives and the exis-
tence of determinate motion verbs prefixed in po- in a given Slavic lan-
guage. I know of no attempts to explain this correlation; the hypothesis
advocated here allows for a straightforward and principled explanation of
it. Note also that there is a narrower correlation between the existence of
determinate motion verbs in po- and the derivation of po- delimitatives
from indeterminate motion verbs in a given Slavic language: the languages
with determinate motion verbs in po- also have delimitatives derived from
indeterminate motion verbs (e.g., Rus poxodit' ‘walk for a while’). Lan-
guages lacking such determinate motion verbs in po- do not have such de-
limitatives: either they lack indeterminate motion verbs prefixed in po-
altogether, as in Croatian/Serbian, or such formations have retained the
older SURFACE-CONTACT meaning, as in Czech, cf., e.g., pochodit ‘walk
around’, pobíhat ‘run around’, etc. This fact is in my view a strong piece of
evidence for the hypothesis outlined above that the new INGRESSIVE-
PARTIAL TRAJECTORY meaning of po- in Russian determinate motion verbs
spread initially to non-prototypical motion verbs (i.e., manner of motion
verbs such as ORus xoditi ‘walk’).
The other piece of circumstantial evidence involves the “semantic
bleaching” of po- in Russian. Unlike all other MRus perfectivizing pre-
fixes, po- has completely lost its original spatial meanings, cf. Voloxina
and Popova (1997: 37–39), Camus (1998: 101) and Tixonov (1998: 36).
This situation is very odd: although many other Russian prefixes “bleach
out” semantically in combination with various individual verb stems as a
result of subsumption, e.g., na- ‘onto’ in napisat' ‘write [to completion]’,
none has thoroughly lost its spatial meaning(s) as a result. Judging from the
development of the preposition po, which lost its GOAL meaning in histori-
cal times (see section 3), one reason for this development must be the rela-
tive instability of a semantic network including both GOAL and
PATH/SURFACE-CONTACT; there have always been several GOAL preposi-
tions in Russian, so it makes sense that this meaning would not naturally be
retained by a preposition that often expressed a meaning as distinct from
GOAL as PATH/SURFACE-CONTACT. The situation was the same for the pre-
fix. But why did po- lose its spatial PATH/SURFACE-CONTACT meaning, and
why would the prefix lose this meaning when the corresponding preposi-
tion po has clearly maintained PATH/SURFACE-CONTACT as its spatial
meaning? This puzzling development is most easily explained by the re-
dundancy of a path prefix with a verb of motion (see section 3), as all mo-
tion occurs redundantly on some path. The salience of go as the highly
frequent default motion verb lent a corresponding salience to the new non-
spatial meaning of the prefix, INGRESSIVE-PARTIAL TRAJECTORY, which
arose by the subsumption of the spatial meaning of the prefix in the pre-
fixed motion verb. The subsequent reanalysis of INGRESSIVE-PARTIAL TRA-
JECTORY and RELATIVE DELIMITATION as a new central meaning of the
prefix laid the foundation for the complete loss of its original spatial mean-
ing.20 It should be stressed that the spatial meaning of the corresponding
preposition po would not be subject to subsumption by a verbal stem, and
so would be unlikely to be affected by developments in the verbal system.
The ability of the hypothesis presented here to explain this difference be-
tween the semantic properties of the prefix po- and the preposition po in
MRus, which to my knowledge has never been properly considered, let
alone satisfactorily explained, is an important point in its favor. (Note that
Tabakowska 1999 does examine the differences between the preposition po
and the prefix po- in Polish, but in my view her analysis does not carry
over to Russian.)
Finally, one last descriptive issue must be addressed: if RELATIVE DE-
LIMITATION as the meaning of the productive new class of Russian delimi-
tatives resulted in large part due to its conceptual proximity to INGRESSIVE-
PARTIAL TRAJECTORY, why is it that MRus pojti and other determinate
motion verbs in po- cannot combine with a facultative adverbial time
phrase, unlike MRus delimitative verbs, cf., e.g., *pojti neskol'ko minut ‘go
for a few minutes’ vs. posidet' neskol'ko minut ‘sit for a few minutes’?
First, it is worth pointing out that just because ORus poiti was the source of
the meaning of INGRESSIVE-PARTIAL TRAJECTORY that produced the new
delimitatives does not mean that this verb itself was (is) a member of that
specific class of verbs. Further, the contrasting combinability with adver-
bial time phrases is the expected consequence of the natures of the respec-
tive predicate types. Despite its particular source-oriented trajectory-
landmark profile, MRus pojti ‘go’ as a determinate verb of motion is never-
theless most often used in a goal-oriented sense, which is demonstrated by
the fact that pojti combines freely with GOAL prepositional phrases as in
pojti v magazin ‘go to the store’ (cf. in this respect Shull’s 2003 views on
the effect of the goal-oriented nature of language on Slavic prefixation).
The notion of linear progression towards a goal is so central to the meaning
of determinate motion verbs that it cannot be abstracted away from them

(which is surely connected with the fact that Russian has developed an
entire separate class of indeterminate motion verbs to express motion con-
strued as not proceeding towards a goal). So the event-completion implied
by pf pojti, as the default pf determinate motion verb meaning ‘go’, must
necessarily profile the traversal of some portion of the trajectory in space;
simultaneously marking time as the domain relative to which the trajectory
traversal is assessed conflicts, or at least is not naturally compatible with
the inherent spatial base of the profile of the pf aspect of pojti (again, as
the default pf motion verb). This restriction is not peculiar to Russian, cf.
English I read for two hours (= ‘spent two hours reading’) versus ?I went
for two hours (= ‘spent two hours going’). Circumstantial evidence for the
correctness of this view is the fact that profiling motion for a given span of
time is indeed possible, but must be expressed by a delimitative indetermi-
nate motion verb, e.g., poxodit' ‘walk for a while’. Thus, the inability of
determinate motion verbs prefixed in po- to profile the duration of the mo-
tion follows from the nature of perfectivity as evaluated for determinate
motion predicates in Russian. In contrast, delimitatives derived from activ-
ity predicates (e.g., posidet' ‘sit for a while’) involve no spatial goal-
orientation whatsoever, and therefore no conflict arises if the totality pro-
filed by the pf aspect is assessed in the domain of time. In the case of ordi-
narily telic predicates such as pisat' ‘write’, given the goal-oriented nature
of language, the default pf conceptualization is one of completion, i.e.,
napisat' ‘write something to completion’; however, such predicates may
alternatively be construed as atelic activities, in which case the totality
profiled by the pf aspect may be assessed relative to the domain of time, as
in the case of activity predicates (cf. Mehlig 1994, 2006 and Dickey 2006).
It should be pointed out that despite the general incompatibility of de-
terminate motion verbs in po- with adverbs of duration, Sémon (1986: 614)
discusses examples in which MRus pojti does co-occur with indicators of
delimitativity:
(9) a. Ob"javljaju: v tečenie tridcati minut po reke

Announce.1SG: in duration thirty.GEN minutes along river
Ljubimovke potečet šampanskoe.
Ljubimovka will PO-flow champagne
[MRus; Sémon 1986: 614]
‘I [hereby] announce: for thirty minutes champagne will flow in
the Ljubimovka river.’
b. Daže dremučij ded v konce neoxotno šel, za

Even dozing old man at last unwillingly went, behind
samoj poslednej telegoj - pojdet, pojdet i nedoverčivo
last wagon - PO-goes, PO-goes and suspiciously
ostanavljivaetsja.
stops
[MRus; Sémon 1986: 614]
‘Even the dozing old man went unwillingly at the back, behind the
last wagon - he goes, goes and then suspiciously stops.’
While such examples are fairly marginal, especially (9a), they neverthe-
less indicate a certain potential to express the duration of a motion event
inherent in determinate motion verbs in po-. This seems particularly true of
(9b), in which pojdet ‘PO-goes’ is reduplicated to produce a meaning akin
to ‘keeps going’ or ‘goes for a while’. Such reduplication is not very diffi-
cult to find, even in the past tense, cf. the following examples:
(10) a. Tak kak-to polučilos', čto ja ušel vpered i tak

So somehow happened, that I went off ahead and so
pošel-pošel-pošel, i dal'še to
PO-went-PO-went-PO-went, and farther sometimes
rasširjaetsja, to sužaetsja poloska, kamni krupnye,
widens, sometimes narrows strip, rocks large,
no v principe idetsja.
but in principle goes.REFL
[MRus]
‘So it happened somehow that I went out ahead and so kept going,
and farther on the strip alternately widens and narrows, with large
rocks, but it’s possible to pass through in principle.’
b. Noč'ju on netoroplivo pošel, pošel, pošel

Night.INST he unhurriedly PO-went-PO-went-PO-went
i okazalsja okolo vxoda v zoopark, gde razbegajutsja
and ended up near entrance to zoo, where run off
vo vse storony dorožki, i ostanovilsja v zadumčivosti,
in all directions paths, and stopped in pensiveness,
rešaja kuda idti dal'še, prjamo, napravo ili nalevo.

deciding where to go farther, straight, right or left
[MRus]
‘At night he went and kept going unhurriedly and ended up near
the entrance to the zoo, where paths go off in all directions, and
stopped, pensive, trying to decide which way to go on, straight
ahead, to the right or to the left.’
Such examples are interesting because they resemble the reduplication

of delimitatives, which produces a meaning of duration akin to English
‘keep Xing’:
(11) a. Poxodil-poxodil čelovek vokrug izbuški.

PO-walked-PO-walked man around little hut
[MRus]
‘The man kept walking around the little hut.’
b. Potom podumaet, podumaet i tože čto-to napišet.

Then PO-thinks, PO-thinks and also something writes
[MRus]
‘Then he thinks and thinks and also writes something down.’
Again, though examples (9) and (10) are admittedly marginal, such us-
age of determinate motion verbs in po- nevertheless does occur, and the
usage in (10) produces the same kind of durative meaning as reduplicated
po- delimitatives. It is important to point out that ordinary telic pf verbs are
to my knowledge unattested in such usage. Thus, despite the fact that de-
terminate motion verbs in po- are unable to profile the duration of the ac-
tion directly, they nevertheless do display a certain resemblance to po-
delimitatives when reduplicated. Sémon (1986: 614) attributes the usage in
(9) to the innovative productivity of delimitative po-, but it is much more
likely that this usage is in fact a relic of older patterns of usage of determi-
nate motion verbs in po-.
With regard to this last point, it is very important to understand that de-
terminate motion verbs prefixed with po- were not always unable to com-
bine directly with an adverbial time phrase. Aitzetmüller (1991: 172) ob-
serves that determinate motion verbs in po- sometimes had delimitative
meaning in Old Church Slavic, and gives the following example:
(12) […] ov"gda bo s', ov"gda on" vožd'

[…] sometimes for.CONJ this one, sometimes that one leader
byvajet" ti naročityj' god" poved" drugujemu
is.HAB and certain time having PO-led another.DAT
prědast", da t" paky vedet" […]
hands over, so he then leads […]
[Old Church Slavic; Aitzetmüller 1991: 172]
‘[…] for sometimes this one, sometimes that one is the leader and
having led for a certain time and hands over [power] to another,
so that then he leads […]’
ORus poiti is also attested with such time phrases as late as the fifteenth
century, as shown in (13):
(13) A is Čjuvilja pošli esmja do Pali 8 dni,

And from Chaul PO-went are.1PL.AUX to Pali eight days,
do indějskyja gory.
to Indian mountain
‘And from Chaul we spent eight days going to Pali, to the Indian
mountains.’
As far as I am aware, no such usage has been attested since the fifteenth
century, i.e., more or less a century before poiti began to function as the pf
correlate of iti ‘go’ (in any case, such usage would have disappeared once
indeterminate motion verbs in po- took on a delimitative meaning).21 In
view of this fact, I consider such usage to be a case of po- expressing AB-
SOLUTE DURATION, i.e., as a metaphorical transfer of FULL TRAJECTORY,
which was in turn derived from the old PATH/SURFACE-CONTACT meaning
of the prefix. Thus, such older usage does not affect the hypothesis advo-
cated here. In support of this view, it should be pointed out that attestations
of poiti expressing ABSOLUTE DURATION before the sixteenth century also
accord with the documented fact that during the sixteenth and seventeenth
centuries pro- (‘through’) took over the expression of PATH/SURFACE-
CONTACT and thus FULL TRAJECTORY from po-. In this respect, pro-, as the
new PATH/SURFACE-CONTACT prefix, played an important role in the shift
to INGRESSIVE-PARTIAL TRAJECTORY as the prototypical meaning of po-.
One may only speculate as to why ORus would develop a new expres-
sion of PATH/SURFACE-CONTACT. It is likely that the new atelic INGRES-
SIVE-PARTIAL TRAJECTORY created an unacceptable or unstable polysemy
in that, as an atelic concept, it could not be incorporated easily into the

telic schema S1 → S2 that united the old spatial meaning(s), and thus ne-
cessitated a new expression for PATH/SURFACE-CONTACT. It is also quite
possible that it was not so much a matter of unacceptable polysemy as it
was a matter of clarity: as po- developed new meanings, its older meaning
of PATH/SURFACE-CONTACT became less readily accessible and was taken
over by another existing SURFACE-CONTACT prefix, pro-, in order to pro-
vide this basic spatial notion the clear expression it requires (and thereby
simplifying the network of po- as well). In any case, it was only when pro-
had taken over PATH/SURFACE-CONTACT from po-, and thus the expression
of FULL TRAJECTORY (cf. examples 5 and 6 above), that pro- became pro-
ductive as a perdurative prefix, i.e., as the prefix expressing ABSOLUTE
DURATION (cf. examples 2a, 2b, 2c, 3 and 7 above).
5. Concluding remarks
In terms of prototype semantics, we may say that Russian po- began as a

category clustered around a telic prototype of PATH/SURFACE-CONTACT;
the expansion of the category involved its spread to concepts bearing lesser
and lesser resemblance to the prototype, notably the atelic concept of IN-
GRESSIVE-PARTIAL TRAJECTORY. As the semantic network of po- began to
spread and develop a local prototype from meanings that were beyond
those that were clearly derived from PATH/SURFACE-CONTACT, another
prefix (pro-) took over the old telic prototype of po-. Considered in this
way, the development of po- resembles the development of middles in
Turkish and Hungarian hypothesized by Kemmer (1992: 159–160), accord-
ing to which “an old reflexive marker developed middle uses, then at some
point would be replaced in reflexive function by an unrelated form.” In
both cases, a the meaning of a linguistic unit expands to include salient
meanings beyond its prototype, to be subsequently replaced in the old pro-
totypical meaning by another form. An important difference between the
development of Russian po- and the universal tendency for reflexives to
develop into middles discussed by Kemmer, or the paths of development of
the function words BE and HAVE discussed by Koch (1999), is that the
paths of development of Russian (and Slavic) perfectivizing prefixes, as
simultaneously lexical and grammatical units, will never fit neatly into the
universal developmental tendencies documented in such studies. Rather,
the paths of their semantic development will frequently turn out to be more
idiosyncratic.
This is why the otherwise highly valuable findings and analyses pre-
sented by Bybee, Perkins and Pagliuca (1994) will probably never be of
more than secondary use to scholars attempting to figure out how Slavic
aspect developed into a grammaticalized category. It is due to the lexical
nature of the prefixes involved that the semantic pathways of the develop-
ment of Slavic perfectivizing prefixes as prototypical categories bear more
resemblance to the lexical cases discussed by Geeraerts (1997) than to the
demonstrated paths of purely grammatical units. (The considerably “lexi-
cal” nature of the grammatical category of Slavic aspect presents no diffi-
culties for a cognitive analysis, which assumes that there is a continuum
between lexical and grammatical units and moreover that grammatical and
lexical units can and often do share the same semantic organization.)
Though the importance of universal, recurrent mechanisms of change (e.g.,
metaphor and metonymy, etc.) remains beyond doubt, the watershed se-
mantic developments in the history of the Slavic perfectivizing prefixes in
fact appear to have been of an accidental nature: Dickey (2005) concludes
that the highly productive Slavic resultative prefix s-/z- was primarily the
accidental result of a sound change, and the analysis presented here argues
that the development of Russian perfectivizing po- was in large part a con-
sequence of the singular combination of the prefix with a particularly sali-
ent predicate type (determinate motion verbs such as iti ‘go’). In this latter
case as well, the development of Slavic perfectivizing prefixes conforms to
Nichols’ and Timberlake’s (1991: 129) suggestion that processes of gram-
maticalization are “less straightforward and obvious than is usually as-
sumed.”
It is worth pointing out that the account of the development of the prefix
po- given here comports with recent views on the timing of the develop-
ment of Russian aspect. Bermel (1997) demonstrates that Russian aspect
was not grammaticalized early, but rather developed more or less continu-
ously throughout the Old Russian period and into the Modern Russian pe-
riod. Nørgård-Sørensen (1997) concludes that Russian aspect must have
emerged as a grammaticalized category in Russian in the seventeenth cen-
tury. The evidence of a significant shift in the semantic network of po-
beginning in the sixteenth century, as well as the establishment of delimita-
tives as a productive class in the seventeenth and eighteenth centuries, fits
in well with their conclusions, especially those of Nørgård-Sørensen on the
importance of the seventeenth century in the development of Russian as-
pect. Given the coincidence in timing, I suggest that the development of
po- was another important change that occurred in the Russian language at
this time, and that the significance of the shift in the meaning of po- (as
manifested in the rise of delimitative verbs) for the Russian aspect system
is more or less comparable to that of the rise of imperfective suffixation.
As Hopper and Traugott (2003: 71) observe, causal explanations are
impossible in investigations into grammaticalization, and it makes sense
only to speak in terms of motivations. Similarly, Nichols and Timberlake
(1991: 131) completely dispense with any pretense (!), conceding that “the
task of investigating historical semantics of grammatical morphemes is in
principle impossible” and hoping only “to sketch what seems a plausible
path of development.” In a case as messy as Russian po-, these are surely
the only reasonable approaches. The one thing that does seem to be clear
about po- is that it has been characterized by a frustrating degree of
polysemy. For this reason, a prototype approach suggests itself naturally,
given its demonstrated suitability to handle complex cases of polysemy.
Though the analysis presented above cannot claim to have solved the entire
puzzle (and the chronology must remain at the level of century-scale ap-
proximations), at the very least it has the advantage of utilizing independ-
ently motivated principles of prototype theory as a tool for organizing our
knowledge of the semantic development of Russian po- in order to hy-
pothesize plausible paths of development. Moreover, it again demonstrates
the value of prototype theory and its ability to handle the semantic devel-
opment of Slavic perfectivizing prefixes, which straddle the distinction
between lexicon and grammar and each of which must be understood fully
in their diachronic development if we are to solve the problem of the
grammaticalization of Slavic aspect.
Notes
* This article was supported in part by a New Faculty General Research Fund
Grant awarded by the University of Kansas. This support is acknowledged
with gratitude. I would like to sincerely thank Alina Israeli and Irina Six for
their frequent insights as well as some Modern Russian translations of Old
Russian sentences, and Erin Moulton for providing some of the Old Russian
data contained here. I am also grateful for the comments of two anonymous
reviewers that helped improve this article. Any errors or inaccuracies are natu-
rally mine alone.
1. There is no universal consensus on the telic vs. atelic status of po- delimita-
tives. However, it should be pointed out that the majority of Slavic aspectolo-
gists consider them to be atelic. Bulygina (1982: 60) makes the crucial point
that if one can describe a situation as zanimalsja (impf) ‘occupied oneself
with/did’, then no matter how short the duration of the activity, it may be de-
scribed perfectively with pozanimalsja (pf) ‘PO-occupied oneself with/PO-did’;
Kučera (1984) makes the same point. This fact proves beyond doubt the at-
elicity of po- delimitatives in Russian. Mehlig (e.g., 1994, 2006) uses the term
netransformativnyj, and Janda (2006) labels atelic verbs uncompletable. Both
these terms are equivalent to atelic for purposes of this analysis.
2. Note that not all Slavic languages have developed delimitatives as a produc-
tive class of verbs in order to extend the aspect opposition to activity predi-
cates; for details, see Dickey and Hutcheson (2003).
3. Russian has divided motion verbs into two classes, DETERMINATE (e.g., idti
‘go’) profiling a one-way, goal-oriented trajectory, and INDETERMINATE (e.g.,
xodit' ‘walk/go’) profiling motion without an obviously linear trajectory (ha-
bitual, aimless, etc.). The exact nature of the latter class in older stages of Rus-
sian is in fact less than clear (see section 4).
4. No single periodization of the Russian language is universally accepted, but
we may generally divide its history into the following periods: Old East Slavic
(from the eleventh to the fourteenth century), Middle Russian (from the late
fourteenth to the seventeenth century), and Modern Russian (from the eight-
eenth century to date). Spanning the end of the Middle Russian period and the
first century or so of the Modern Russian period was an important transitional
period that saw the formation of the Russian national language (from the late
seventeenth to the early nineteenth century). As the distinction that is primarily
relevant for this discussion is that between pre-modern Russian and modern
Russian after the transitional period (i.e., from the beginning of the nineteenth
century), these two periods are hereinafter labeled Old Russian and Modern
Russian respectively.
5. There are many interesting issues involved with this change that are not im-
mediately relevant for the purpose at hand. One that deserves a brief comment
is the healthy increase in distributive verbs. While it is true that the percentage
of po- distributives increased considerably, distributive po- has not become
nearly as productive as delimitative po-: pere- became the primary distributive
prefix in MRus (cf., e.g., perebrosat' ‘throw [all of one after another]’), largely
taking over this function from po- except in the derivation of distributives
from verbs already containing a prefix, e.g., po-vy-brasyvat' ‘throw out [all
of]’ (from vy-brasyvat' ‘throw out’). The replacement of distributive po- by
distributive pere- should in fact be viewed as part of the overall process that
led to po- being a primarily delimitative suffix.
6. Regarding Dmitrieva’s and Sigalov’s claims, one reviewer wonders whether
po- delimitatives are absent in ORus texts because they are characteristic of
colloquial style. I believe that it is better to take the texts at face value in this
case, as po- delimitatives do occur in ORus texts, but only for a small set of
stative activity predicates, e.g., poležati ‘lie for a while’, posěděti ‘sit for a
while’ and postojati ‘stand for a while’ – the very same predicates for which
po- delimitatives exist in those Slavic languages, such as Czech and

Serbian/Croatian, in which delimitative po- never became productive; cf. the
discussion by Dickey and Hutcheson (2003).
7. This mapping is difficult to prove, but is hardly controversial given the similar
extension of another Slavic prefix expressing ‘application to a surface’,
namely na- (cf., e.g., MRus nakraxmalit' ‘apply starch to’) to express quantity
(e.g., MRus nakupit' ‘buy a lot of’).
8. The term “stative activity predicate” may at first glance seem self-
contradictory and contrary to Vendler’s distinction between states and activi-
ties, but agentive activities range from very dynamic (e.g., push, run, work) to
very stative (e.g., sit, lie, rest). Note that Mehlig (2006) makes the distinction
between stativnye netransformativy (stative activities) and dinamičeskie ne-
transformativy (dynamic activities), neither of which are states (essentivy in
Mehlig’s terms).
9. The transcription of ORus examples follows the orthography of the published
source; different editions clean up the orthography to different extents. How-
ever, orthography is irrelevant to the issues discussed here.
10. An anonymous reviewer objects that MRus delimitatives occur with adverbials
indicating specific periods time, which must render them as telic as (2a, c), cf.
the following: V marinade krevetke dostatočno posidet' čas, potom ee nado
obsušit' na salfetke ‘It is sufficient for the shrimp to sit [PO-sit] in the mari-
nade for an hour, then it should be dried on a napkin’; Poètomu esli rebenka
kormili v tečenie 40 minut načinaja s 10 časov utra, posle čego on pospal 1
čas 20 minut i načal est' snova, interval sostavljaet 2 časa ‘Therefore, if the
child was fed for a period of 40 minutes beginning at 10 a.m., after which he
slept for 1 hour and 20 minutes and began to feed again, the [resulting] pe-
riod is 2 hours’. But the point is this: in each of the cited cases, the predicate
in question (‘sitting in a marinade’, ‘sleeping’) is presented as easily able to
continue for a longer period of time. This fact accounts for the phrase dosta-
točno ‘it is sufficient’ in the first example – the shrimp could easily be left in
the marinade all day and thus an hour is not a maximum but a minimum, as
well as the conditional esli ‘if’ in the second example – the sleeping time of in-
fants after breast feeding varies greatly even with a single infant, and there is
no intention here to present one hour and twenty minutes as some maximum
duration beyond which an infant could not sleep. Thus, these predicates are
clearly atelic. If the speaker wished to present the durations involved in these
examples as maximum, the corresponding perduratives in pro- would be used,
i.e., prosidet' ‘sit [for a specified, long period of time]’ and prospat' ‘sleep [for
a specified, long period of time]’ (see below for a discussion of such verbs in
MRus).
11. In MRus, po- delimitatives in fact occur most typically in texts with no tempo-
ral adverbial at all.
12. Note that Veyrenc (1968) characterizes the distinction between MRus po- and
pro- as that between “extension relative” and “extension absolue du procès”,
which I take to be an independent confirmation of the relevance of RELATIVE
DELIMITATION and ABSOLUTE DURATION for these prefixes in (the history of)
Russian.
13. I assume that this situation was partially the result of the incorporation of an
etymologically distinct SOURCE meaning into the network of the prototypical
PATH meaning, but this issue is not immediately relevant for the present dis-
cussion.
14. Note the stipulation prototypical. It is hard to find detailed commentary on the
actionality (and perfectivity) of MRus pojti. As an ingressive verb, it is ordi-
narily assumed to reach its telos “immediately, when the movement starts,” as
one anonymous reviewer phrases it. However, Zaliznjak and Šmelev (2000:
109) observe that “the ingressive meaning of such verbs is attenuated,” as
phrases such as pojti v kino ‘go to the movies’ or poexat' v Pariž ‘drive to
Paris’ do not mean ‘begin to go’ but refer to the actions “as such”. Zaliznjak
and Šmelev attribute this effect to a kind of metonymy whereby the beginning
of the action represents the entire action. While this could be the case,
Zaliznjak’s and Šmelev’s view comports with Shull’s (2003) hypothesis that
spatial po- indicates the initiation of the trajectory plus some unspecified
amount of the background trajectory: if the very beginning of a trajectory can
metonymically represent an entire trajectory, then the beginning of a trajectory
plus some additional non-zero amount of its traversal must certainly also be
able to represent the entire trajectory metonymically. Moreover, there are
several reasons to believe that spatial po- with determinate motion verbs
expresses more than merely the beginning of the action. First, as Shull (2003:
149–150) points out, pojti need not refer to the initiation of motion, but can
refer to the initiation of a new trajectory within a larger motion event, cf., e.g.,
her example Sobaka podbežala k kovru, pobežala po kovru i ubežala dal'še
‘The dog ran up to the carpet, ran [PO-ran] on the carpet and ran on.’ In this
respect, as Shull (ibid., 155–157) observes, spatial po- differs markedly from
the main MRus inceptive prefix za-, as the latter must express the transition
from the lack of an activity to the activity, e.g., zabegat' ‘start running
[transition from the state of not running to the state of running]’. Second, when
pojti is reduplicated it expresses duration in a fashion resembling true
delimitatives in po- (see examples 10–11), which is very hard to explain if
spatial po- does not express something more than the absolute beginning of the
action; note that other inceptive verbs such as zabegat' ‘begin to run’ and
momentary verbs such as skočit' ‘jump’ are not reduplicated to express
duration.
15. Note that Aitzetmüller (1991: 171) has also suggested a general connection
between the spatial and temporal meanings of po- in the evolution of the
Slavic aspectual system, cf. his example in (12) below. I am grateful to Tho-
mas Daiber for making me aware of Aitzetmüller’s view.
16. The development of this new class of pf verbs was also a contributing factor in
a significant change in the semantics of the pf aspect in Russian, which unfor-
tunately cannot be considered here; for discussion, see Dickey and Hutcheson
(2003) and Dickey (2005).
17. This view is based on numerous considerations that cannot be taken up here.
Briefly, we may say that, as Stern (2002) and Greenberg and Dickey (2006)
observe, in OCS and ORus the putative “indeterminate” motion verbs did in
fact occur in determinate contexts (e.g., i sam" si nosę kr'st' izide v" nari-
caemoe kranievo město, eže glagolet" sę evreisky gol"gafa ‘and himself bear-
ing his cross he went out to the place called the place of the skull, which in
Hebrew is called Golgotha’). Such usage is much easier to explain if such
verbs were manner of motion verbs, as opposed to strictly “indeterminate”. It
is also worth noting that Veyrenc (1966) suggests a similar situation in MRus,
arguing that “indeterminate” motion verbs such as xodit' ‘walk, go’ represent a
kind of activity verb (verbes du fonction) as opposed to genuine motion verbs
(verbes du déplacement).
18. These forms resemble conjugated forms of pf determinate motion verbs pre-
fixed in po- in the East Slavic languages (e.g., Rus pf pojdu ‘I will go’ < pojti
‘go’), but are future tense forms of the Cz unprefixed impf verbs. Note that,
with the exception of jít ‘go’, there are synonymous compound future tense
forms for these verbs (e.g., budu jet alongside pojedu ‘I will drive/be driv-
ing’); cf. Kopečný (1962: 47) on the “absolute” synonymy of the Czech impf
compound and po- future tense forms. It should be pointed out that though
such po- futures are undoubtedly impf, they do not necessarily emphasize an
ongoing process on a par with English progressive forms (e.g., ‘s/he will be
going’); rather, they are the default future tenses of Czech impf motion verbs,
and thus have been translated with English simple-tense forms.
19. Shull (2003: 160–164) argues that her schema for po- given in figure 3 also
accounts for resultatives, in that po- contributes merely the notion that some
amount of the trajectory denoted by the verb is completed, and the construal
expressed by the source verb itself determines the telicity or atelicity of the
po- derivative (i.e., stroit' ‘build’ is a telic notion, whereas sidet' ‘sit’ is atelic).
I think that this is a viable approach, but cannot consider the issue in the nec-
essary detail here. Suffice it to say that it seems possible to integrate the sur-
viving resultative verbs in the hypothesized network for MRus po- as marginal
cases, but cases that are nonetheless incorporated in the network in a princi-
pled way.
20. In this respect, it should be pointed out that po- has lost its spatial meaning in
Ukrainian (cf. Šerex 1951: 292) and Bulgarian (cf. Ivanova 1966: 124), as
well as in Polish (cf. Śmiech 1986: 18). Thus, it appears that in the individual
Slavic languages there is a three-way correlation between the prefixation of
determinate motion verbs with po-, the productivity of delimitative po-, and
the lack of a spatial meaning for po-. If this is true, it cannot be a mere coinci-
dence.
21. Another remnant of this older situation seems to be Bulgarian poida ‘go for a
short time’, given in the RSBKE 2: 599. The verb is tagged ‘rare’, and judged
to be dialectal by Bulgarian informants.
References
Abdoulaye, Mahamane L.
2001 The grammaticalization of Hausa zaa ‘be going’ to future. Journal of
African Languages and Linguistics 22 (1): 1–32.
Aitzetmüller, Rudolf
1991 Altbulgarische Grammatik als Einführung in die slavische
Sprachwissenschaft. Freiburg: U. W. Weiher.
Bermel, Neil
1997 Context and the Lexicon in the Development of Russian Aspect.
Berkeley: University of California Press.
Bondarko, A. V.
1961 K voprosu o glagolax dviženija v češskom jazyke (formy tipa
ponesu, poletím) [On the question of verbs of motion in Czech
(forms of the type ponesu, poletím)]. Slavia 30: 527–547.
Bulygina, T. V.
1982 K postroeniju tipologii predikatov v russkom jazyke [Toward con-
structing a typology of predicates in Russian]. In Semantičeskie tipy
predikatov [Semantic Types of Predicates], O. N. Seliverstova (ed.),
7–85. Moscow: Nauka.
Bybee, Joan, Revere Perkins, and William Pagliuca
1994 The Evolution of Grammar. Tense, Aspect and Modality in the Lan-
guages of the World. Chicago: University of Chicago Press.
Camus, Remi
1998 Quelques considérations sur le préverbe po- en russe contemporain.
Revue des Études Slaves LXX (1): 101–112.
Čertkova, Marina
1996 Grammatičeskaja kategorija vida v sovremennom russkom jazyke
[The Grammatical Category of Aspect in Contemporary Russian].
Moscow: Izdatel'stvo Moskovskogo universiteta.
Dahl, Östen
1985 Tense and Aspect Systems. Oxford: Blackwell.
Dickey, Stephen M.
2000 Parameters of Slavic Aspect: A Cognitive Approach. Stanford: CSLI.
2005 S-/Z- and the grammaticalization of Slavic aspect. Slovene Linguistic

Studies 5: 3–55.
2006 Aspectual pairs, goal orientation and PO- delimitatives in Russian. In
Glossos 7, Laura Janda (ed.), (http://seelrc.org/glossos/issues/7/).
Dickey, Stephen M., and Julie Hutcheson
2003 Delimitative verbs in Russian, Czech and Slavic. In American Con-
tributions to the Thirteenth International Congress of Slavists,
Robert A. Maguire and Alan Timberlake (eds.), 23–36. Blooming-
ton, IN: Slavica.
Dmitrieva, Ol'ga
1991 Formirovanie semantičeskoj struktury russkogo glagol'nogo prefiksa
po- [The formation of the structure of the Russian verbal prefix po-].
In Aktivnye processy v jazyke i reči [Active Processes in Language
and Discourse], 68–74. Saratov: Izdatel'stvo Saratovskogo univer-
siteta.
2000 Formirovanie sistemy russkix delimitativnyx glagolov [The forma-
tion of the system of Russian delimitative verbs]. I n Predloženie i
slovo: paradigmatičeskij, tekstovyj i kommunikativnyj aspekty
[Clause and Word: Paradigmatic, Textual and Communicative As-
pects], T. V. Kočetkova ( ed.), 28–33. Saratov: Izdatel'stvo Sara-
tovskogo pedagogičeskogo instituta.
Flier, Michael
1985 The scope of prefixal delimitation in Russian. In The Scope of Slavic
Aspect, Michael Flier and Alan Timberlake (eds.), 41–58. Columbus,
OH: Slavica.
Geeraerts, Dirk
1997 Diachronic Prototype Semantics. Oxford: Clarendon Press.
Greenberg, Marc L., and Stephen M. Dickey
2006 Slavic *Jazditi ‘to ride’ and its implications for the category of (in-)
determinacy. In Jezikovna predanost. Akademiku prof. dr. Jožetu
Toporišiču ob 80-letnici [Linguistic Dedication. To Prof. Dr. Jože
Toporišič on his 80th Birthday], Marko Jesenšek and Zinka Zorko
(eds.), 153–158. Maribor: Slavistično društvo.
Hopper, Paul J.
1979 Aspect and foregrounding in discourse. In Discourse and syntax, T.
Givon (ed.), 213–241. (Syntax and semantics 12.) New York: Aca-
demic Press.
Hopper, Paul J., and Elizabeth Closs Traugott
2003 Grammaticalization (Second Edition). Cambridge: Cambridge Uni-
versity Press.
Isačenko, A. V.
1960 Grammatičeskij stroj russkogo jazyka v sopostavlenii s slovackim.
Čast' vtoraja: morfologija [The Grammatical Structure of Russian
Contrasted with Slovak. Part Two: Morphology]. Bratislava: Iz-

datel'stvo akademii nauk.
Israeli, Alina
2002 Russian verbs of motion: focus, deixis and viewpoint. Cahiers
Chronos 10: 97–118.
Ivanova, Kalina
1966 Desemantizacija na glagolnite predstavki v sâvremennija bâlgarski
knižoven ezik [The Desemanticization of Verbal Prefixes in Contem-
porary Literary Bulgarian]. Sophia: Bâlgarska akademija na naukite.
Janda, Laura
2007 Aspectual clusters of Russian verbs. Studies in Language 31 (3):
607–648.
Kemmer, Suzanne.
1992 Grammatical prototypes and competing motivations in a theory of
linguistic change. In Explanation in Historical Linguistics, Garry W.
Davis and Gregory K. Iverson (eds.), 145–166. Amsterdam: John
Benjamins.
Koch, Peter
1999 Cognitive aspects of semantic change and polysemy: The semantic
space HAVE/BE. In Historical Semantics and Cognition, Andreas
Blank and Peter Koch (eds.), 279–305. Berlin: Mouton de Gruyter.
Kopečný, František
1962 Slovesný vid v češtině [Verbal Aspect in Czech]. Prague: Nak-
ladatelství Československé akademie věd.
Langacker, Ronald
1987 Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford Uni-
versity Press.
Lehrer, Adrienne
2003 Polysemy in derivational affixes. In Polysemy: Flexible Patterns of
Meaning in Mind and Language, Brigitte Nerlich, Zazie Todd, Vi-
mala Herman, and David D. Clarke (eds.), 217–232. Berlin: Mouton
de Gruyter.
Lilov, Metodi
1964 Semantični razvoj na glagolnata predstavka po- v bâlgarskija ezik
[The semantic development of the verbal prefix po- in Bulgarian]. Iz-
vestija na Instituta za bâlgarski ezik 10: 65–157.
Mayo, Peter J.
1984 The Morphology of Aspect in Seventeenth-Century Russian.
Columbus, OH: Slavica.
Mehlig, Hans Robert

1994 Gomogennost' i geterogennost' v prostranstve i vremeni [Homogene-
ity and heterogeneity in space and time]. Revue des Études Slaves 64
(3): 585–606.
2006 Glagol'nyj vid i vtoričnaja gomogenizacija oboznačaemoj situacii
posredstvom kvantifikacii: K upotrebleniju delimitativnogo sposoba
dejstvija v russkom jazyke [Verbal aspect and the secondary homog-
enization of a denoted situation by means of quantification: On the
Use of Delimitative Procedurals in Russian]. In Glagol'nyj vid i lek-
sikografija. Semantika i struktura slavjanskogo vida [Verbal Aspect,
Lexicography and the Structure of Slavic Aspect] IV, V. Lehmann
(ed.), 235–276. Munich: Otto Sagner.
Němec, Igor
1954 O slovanské předponě po- slovesné [On the Slavic verbal prefix po-].
Slavia 23: 1–22.
Nørgård-Sørensen, Jens
1997 Tense, aspect and verbal derivation in the language of the Nov-
gorod birch bark letters. Russian Linguistics 21: 1–21.
Nichols, Johanna, and Alan Timberlake
1991 Grammaticalization as retextualization. I n Approaches to Gram-
maticalization. Vol. 1. Focus on Theoretical and Methodological
Issues, Elizabeth Closs Traugott and Bernd Heine, ( eds.), 129–146.
Amsterdam: John Benjamins.
Nikiforidou, Kiki
1991 The meanings of the genitive: A case study in semantic structure and
semantic change. Cognitive Linguistics 2 (2): 149–205.
Nübler, Norbert
1990 Zum Begriff der ‘Subsumptionspräfixe’ in der Aspektforschung.
Anzeiger für slavische Philologie XX: 123–134.
Petruxina, Elena
2000 Aspektual'nye kategorii glagola v russkom jazyke v sopostavlenii s
češskim, slovackim, pol'skim i bolgarskim jazykami [Aspectual
Categories of the Verb in Russian contrasted with Czech, Slovak,
Polish and Bulgarian]. Moscow: Izdatel'stvo Moskovskogo univer-
siteta.
Radden, Günter
1988 The concept of motion. In Understanding the Lexicon: Meaning,
Sense and World Knowledge in Lexical Semantics, W. Hüllen and R
Schulze (eds.), 380–394. Tübingen: Max Niemeyer.
Sémon, Jean-Paul
1986 Postojat' ou la perfectivité de congruence. Definition et valeurs tex-
tuelles. Revue des Études Slaves 58 (4): 609–635.
Shull, Sarah
2003 The Experience of Space: The Privileged Role of Spatial Prefixa-
tion in Czech and Russian. Munich: Otto Sagner.
Sigalov, Pavel S.
1975 Istorija russkix ograničitel´nyx glagolov [The history of Russian
delimitative verbs]. Trudy po russkoj i slavjanskoj filologii: Serija
lingvističeskaja [Studies in Russian and Slavic Philology] 24: 141–
181.
1977 Istorija russkix perdurativnyx glagolov [The history of Russian per-
durative verbs]. Trudy po russkoj i slavjanskoj filologii: Serija ling-
vističeskaja [Studies in Russian and Slavic Philology] 23: 101–118.
Śmiech, Witold
1986 Derywacja prefiksalna czasowników polskich [The Prefixal
Derivation of Polish Verbs]. Wrocław: Zakład narodowy im.
Ossolińskich.
Stern, Ariann Nicole
2002 The Verbs of Motion in Old Russian Texts: A Comparative Gram-
matical Analysis of a Nascent Verb Class. Ph.D. diss., University of
California at Los Angeles.
Strekalova, Zoja Nikolaevna
1968 Iz istorii pol'skogo glagol'nogo vida [From the History of Polish
Verbal Aspect]. Moscow: Nauka.
Šerex, Jurij
1951 Narys sučasnoji ukrajins'skoji literaturnoji movy [A Sketch of Con-
temporary Literary Ukrainian]. Munich: Naukove tovarystvo im.
Ševčenka.
1999 ‘Pobłądzić po malowniczych zaułkach Starego Miasta’: Semantyka
polskiego przyimka po i przedrostka po- [‘To roam around the pic-
turesque lanes of Stare Miasto’: The semantics of the preposition po
and the prefix po-]. In Collectanea Linguistica in Honorem Casimiri
Polański, Maria Brzezina and Halina Kurek (eds.), 269–278.
Kraków: Księgarnia Akademicka.
2003 Space and time in Polish: The preposition za and the verbal prefix
za-. In Motivation in Language. Studies in Honor of Günter Radden,
Hubert Cuyckens, Thomas Berg, René Dirven, and Klaus-Uwe Pan-
ther (eds.), 153–177. Amsterdam: John Benjamins.
Taylor, John R.
1995 Linguistic Categorization. Prototypes in Linguistic Theory. Oxford:
Clarendon Press.
Tixonov, Aleksandr
1998 Russkij glagol: Problemy teorii i leksikografirovanija [The Russian
Verb: Theoretical and Lexicographical Problems]. Moscow: Akade-
mia.
Veyrenc, Jacques
1966 Russe idti et xodit': mouvement de déplacement et mouvement de
fonction. Bulletin de la Société de Linguistique de Paris 61 (1): 21–
41.
1968 Expansion syntaxique et classement des préverbes. Slavica 8: 245–
249.
Voloxina, G. A., and Z. D. Popova
1997 Kategorija glagol'nogo vida v svete semantičeskogo ustrojstva glago-
l'nyx pristavok [The category of aspect in light of the semantic orga-
nization of verbal prefixes]. In Trudy aspektologičeskogo seminara
Filologičeskogo fakul'teta MGU im. M. V. Lomonosova [Studies of
the Aspectological Seminar of the Philological Faculty of Moscow
State University]. Vol. 3, M. Ju. Čertkova (ed.), 34–41. Moscow: Iz-
datel'stvo Moskovskogo universiteta.
Zaliznjak, Anna A., and Aleksej D. Šmelev
2000 Vvedenie v russkuju aspektologiju [An Introduction to Russian As-
pectology]. Moscow: Jazyki russkoj kul'tury.
Sources
BLDR 1 = D. S. Lixačev, L. A. Dmitrieva, A. A. Alekseevna and N. V. Ponyrko,

eds. (1997) Biblioteka literatury drevnej Rusi. Tom 1 [Library of
the Literature of Old Russia. Vol. 1]. St. Petersburg: Nauka.
BLDR 7 = D. S. Lixačev, L. A. Dmitrieva, A. A. Alekseevna and N. V. Ponyrko,
eds. (2000) Biblioteka literatury drevnej Rusi. Tom 7 [Library of
the Literature of Old Russia. Vol. 7]. St. Petersburg: Nauka.
MSDJa = Sreznevskij, I. I. (1958) Materialy dlja slovarja drevnerusskogo jazyka
po pis'mennym pamjatnikam [Materials for a Dictionary of Old Rus-
sian based on Written Texts]. Moscow: Gosudarstvennoe izdatel'stvo
inostrannyx i nacional'nyx slovarej.
PLDR XVII.2 = Lev Aleksandrovič Dmitriev and Dmitrij Sergeevič Lixačev, eds.
(1989) Pamjatniki literatury drevnej Rusi. XVII vek. Kniga vtoraja
[Texts of the Literature of Old Russia. XVIIth Century. Book Two].
Moscow: Xudožestvennaja literatura.
RSBKE 2 = Romanski, Stojan, ed. (1957) Rečnik na sâvremenija bâlgarski
knižoven ezik. Tom 2 [Dictionary of Contemporary Literary Bulgar-
ian. Vol. 2]. Sophia: Bâlgarskata akademija na naukite.
Sloj 1 = Strogal'ščikov, Viktor. (2003) Sloj 1 [Layer 1]. Moscow: Palmira.

SDJa 6 = I. S. Uluxanov, ed. (2000) Slovar' drevnerusskogo jazyka (XI–XIV vv.).
Tom 6 [Dictionary of Old Russian (XI–XIV cent.). Vol. 6]. Moscow:
Azbukovnik
SRJa XI–XVII 17 = S. G. Barxudarov, ed. (1991) Slovar' russkogo jazyka XI~XVII
vv. Vypusk 17 [Dictionary of the Russian Language of the XI–XVII
Centuries. Issue 17]. Moscow: Nauka.
SSRLJa = Akademija Nauk SSSR. (1950–65) Slovar' sovremennogo russkogo
literaturnogo jazyka [Dictionary of Contemporary Literary Russian].
Moscow: Izdatel'stvo Akademii nauk SSSR.
The rise of an epistemic pragmatic marker in
Balkan Slavic: an exploratory study of nešto
Eleni Bužarovska
Abstract
The article focuses on the semantic change of the indefinite pronoun nešto
into an epistemic mitigation modal in Macedonian, as part of a wider Balkan
Slavic context. The presence of the secondary nešto in other South Slavic
languages suggests that the development of the pronominal nešto into a
mitigation marker represents a motivated change based on some universal
conceptual mechanisms. The author explores the possible paths of this
change and suggests that strengthening of invited inferences and subjectifi-
cation were the two cognitive mechanisms that played a major role in this
metonymically-based process. Due to contextual inferences, the schematic
referential nešto was subject to change in colloquial speech. The result of
this semasiological change was an epistemic approximative modal nešto that
functions as a modal of epistemic stance to encode the speaker’s slight de-
crease of the assertoric force of the proposition. The theoretical framework
of the analysis of the semantic change of nešto is based on the Taylor’s
model of schema-instances relations (Taylor 2002) and Traugott and
Dasher’s (2002) model of invited inferencing theory of semantic change
(IITSC). The former is used to account for the development of the polyse-
mous, secondary nešto, whereas the latter explains the rise of several epis-
temic meanings from the secondary nešto in specific contexts. On this view,
the development of the secondary nešto represents a case of coded inferen-
tial change that resulted in the creation of two pragmatic utterance-type
meanings of the secondary nešto: approximative quantification and epis-
temic modal. Via metonymic extension the utterance-type meanings have
further developed into nine interrelated utterance-token, context-dependent
meanings.
Keywords: Balkan Slavic, mitigation marker, inferencing, subjectification
376 Eleni Bužarovska
1. Introduction
This article investigates the change of the Balkan Slavic indefinite pronoun
nešto into an epistemic pragmatic marker. In colloquial usage nešto has
undergone a major semantic shift into a non-assertoric epistemic particle
that marks approximation and mitigation as the following Serbian (1), Bul-
garian (2) and Macedonian (3) examples illustrate:1
(1) Nešto mi je sumnjiva.

something 1SG.DAT.CL is suspicious.SG.F
‘She behaves in a somewhat suspicious manner.’
(2) Mislja če nešto ne săm

think.1SG.PRES that something not be.1SG.PRES
zapisal verno mobilnija ti telefon.
written.SG.M correctly mobile 2SG.DAT.CL telephone
‘I think I might not have written down your mobile phone correctly.’
(3) Nešto glavata me boli.

something head-the 1SG.ACC.CL hurt.3SG.PRES
‘I sort of have a headache.’
The comparison between examples (1), (2) and (3) and the correspond-
ing minimal pairs without nešto (1’), (2’), and (3’) shows that speakers,
with the help of nešto, “soften” their assertions. Thus, the mitigation sense
of nešto in examples (1), (2), and (3) is rendered in English by three differ-
ent lexemes: ‘somewhat’, ‘might’ and ‘sort of’.
(1’) Sumnjiva mi je.

suspicious.SG.F 1SG.DAT.CL is
‘She behaves in a suspicious manner.’
(2’) Mislja če ne săm zapisal

think.1SG.PRES that not be.1SG.PRES written.SG.M
verno mobilnija ti telefon.
correctly mobile 2SG.DAT.CL telephone
‘I think I have not written down your mobile phone number cor-
rectly.’
The rise of an epistemic pragmatic marker in Balkan Slavic 377
(3’) Glavata me boli.

head-the 1SG.DAT.CL hurt.3SG.PRES
‘I have a headache.’
The paper traces the development of nešto in Macedonian, a Balkan

South Slavic language. Despite the frequent distribution of the epistemic
nešto in colloquial Macedonian and other Balkan Slavic languages, its
semantic change has not received attention in literature except for a brief
comment by Koneski (1987). Citing a late nineteenth century example
(1987: 340), he observes that nešto, traditionally defined as an indefinite
pronoun, can lose its independent pronominal meaning and assume the
function of a “strengthener”:
(4) Što nešto ste se uplašile

what something be.2PL.PRES REFL frightened.PART.PL
i ste se stutkale?
and be.2PL.PRES REFL huddled.PART.PL
‘Why did you seem to be frightened and you huddled like that?’
However, in (4) the function of nešto cannot be defined as that of a

“strengthener”, but rather of a pragmatic “softener” of the assertion. An
analysis of data collected from various texts on the web and from oral
communication shows that nešto has developed a series of secondary
meanings in colloquial Macedonian that can be subsumed under two sub-
types: approximate quantification and assertoric mitigation.2
The main goal of this paper is to establish the pathways of change and
the polysemy pattern of nešto by exploring the functions of the secondary
nešto. The following questions are addressed in the discussion: (a) How
was nešto coerced into performing the function of a modal operator? and
(b) How are the secondary senses related?
By providing possible answers to these questions, I will uncover the
mechanisms and pathways of change, thereby demonstrating that the tra-
jectory traced by nešto illustrates a highly regular type of pragmatically
driven semantic change. The theoretical framework of the analysis of the
semantic-syntactic derivation of nešto is primarily based on the invited
inferencing theory of semantic change (IITSC) model developed in
Traugott and Dasher (2002) and Taylor’s model of schema-instances rela-
tions (Taylor 2002).3 The IITSC model accounts for the mechanisms that
are involved in the internal semantic change of nešto: strengthening of
invited inferences and subjectification (Traugott 1989).4
The strengthening of pragmatic inferences causes the semantic change

of the coded meanings into new polysemous coded meanings. The authors
argue that semantic change is motivated by the conversational principle of
relevance or R(elevance)-heuristic (see Horn 1984) “say no more than you
must, and mean more thereby”. They point out that “[t]he R-heuristic leads
to change because it evokes utterance meanings beyond what is said; in
other words it involves pragmatic strengthening.” (Traugott and Dasher
2002: 19).
Traugott and Dasher (2002: 49) underline the central role of speakers as
the prime initiators of change and add that “[t]he main tendency in seman-
tic change from a semasiological perspective is toward greater subjectivity
or grounding in SP/W’s [speaker/writer’s] attitude and perspective”.
As a complementary model to invited inferencing theory of semantic
change, Taylor’s model of schema and instances relation (Taylor 2002:
125) is used to explain the development of the secondary nešto as well as
the types of conceptual relations between the extended meanings.5 The
cognitive “link” between the original and the derived sense can be repre-
sented as an “inverted” triangle, with a superordinate schema at its bottom
(see the top of Figure 2). The schema abstracts what is common in the in-
stances characterized by additional specifications. Taylor claims that “[a]n
instance inherits the specifications of the schema, but fleshes out the
schema in more detail” (2002: 124). In the case of nešto, all extended
meanings share a common concept of an indeterminate entity, so they may
be considered different instances of the same schema.
The paper is organised in the following manner: the following section
examines the semantic structure of the pronominal nešto. Section 3 and 4
present the classification of the two main secondary meanings based on
distributional criteria of nešto in different contexts. Section 5 provides a
network of interrelated secondary meanings and section 6 explains the
pathways of the semantic change of nešto.
2. The primary function of nešto
The pronouns nekoj ‘someone’ and nešto ‘something’ are the central expo-
nents of the category of reference in standard Macedonian. Out of two
existing universal ontological categories, humans and objects, nešto desig-
nates objects. In contrast to nekoj, nešto exhibits a higher degree of ab-
stractness due to the lack of human reference. Nekoj ‘someone’ is marked
for both gender and number; its function in an NP is to carry information
about a human referent that the speaker is unable to identify. The indefinite
pronoun nešto, marked for neuter gender, denotes things and belongs to the
category of NPs.6 The NPs that contain the pronominal nešto carry infor-
mation that the topic of conversation is unidentified both to the speaker and
to the other participants of the speech situation. On the syntactic plane,
nešto functions as a phrasal head:
(5) Jas gledam nešto.

I see.1SG.PRES something
‘I see something.’
The reference of nešto is not limited only to things, but also extends to
any indeterminate entity such as events and objects of thoughts. The pro-
nominal nešto pertains to a physical object in (5), to an event in (6) and to
a proposition in (7).
(6) Se sluči nešto.

REFL happen.3SG.AOR something
‘Something happened.’
(7) Toj znae nešto.

he know.3SG.PRES something
‘He knows something.’
In examples (5), (6), and (7), the pronominal nešto invokes the notion of
‘an indeterminate entity’ as its underspecified lexical meaning makes it
schematic both for things and events.7 The full interpretation of nešto de-
pends on the context and knowledge of the speech event circumstances.
Because the pronominal nešto lacks specific meaning and has an extremely
wide scope of reference, in certain contexts it can create specific pragmatic
inferences of the speaker’s slight doubt in the truth of the assertion. In such
contexts, the speaker’s avoidance to provide any additional characteriza-
tion of the thing/event s/he is talking about (coded by nešto) might imply
that s/he epistemically evaluates the event. For instance, when the speaker
utters sentences like I see something, He knows something or Something
happened the hearer infers that the speaker may imply “I think so, but I do
not claim it”. Namely, the “failure” of the pronominal nešto to express the
specifications of the profiled event produces pragmatic inferences of a
weak epistemic force.8 Presumably the speaker’s ignorance or unwilling-
ness to elaborate the content of nešto creates invited inferences of the

speaker’s slight indeterminacy about the truth of the proposition.
It was argued above that the schematic referential meaning of nešto
makes possible the rise of pragmatic inferential meanings in certain con-
texts. The inference of indeterminacy that the pronominal nešto creates is
semanticized as a new coded meaning of the speaker’s non-assertion. Thus,
the inferences of the primary nešto have become a conventionalized mean-
ing in the secondary nešto, thereby producing polysemy of the same lex-
eme. The new, secondary nešto functions as a modal of epistemic stance
with the help of which the speaker slightly decreases the assertoric force of
the proposition:
(8) Nešto ne mi e dobro.

something not 1SG.DAT.CL is well
‘I don’t feel quite well.’
It is reasonable to assume that the reasons for this semasiological

change lie in the highly schematic semantic structure of nešto. As a nomi-
nal concept with a maximum degree of conceptual schematicity, nešto
becomes a viable candidate for semantic change. The semantically poor
nominal concept under favorable pragmatic conditions changes into a rela-
tional concept.9 Figure 1 below sums up the main features of the polyse-
mous nešto.
primary nešto secondary nešto

nominal concept relational concept
indeterminate object/event > indeterminacy, uncertainty of the
speaker
Figure 1. Semantic features of nešto
According to Croft and Cruse (2004: 20), the dictionary view of linguis-
tic meaning generally describes only the concept’s profile. Due to the cen-
tral position of profile in determining a word’s meaning, any shift in pro-
file has truth-conditional consequences.
In the case of secondary nešto the profile of indeterminate entity has
been shifted to a different semantic frame, i.e. of events in the mental
world of the speaker. In other words, nešto underwent semantic change due
to profile shifting from denoting an entity (in the real world) that the
speaker conceives as indeterminate to the very relation of indeterminacy
between a conceived entity and the speaker. The result of this semantic
change was the rise of two secondary polysemous nešto, syntactically and
semantically different. The first nešto signifies a small approximative
degree, whereas the meaning of the second nešto may be described as epis-
temic approximation because it expresses slight uncertainty of the speaker
with respect to the epistemic status of the proposition. Subsequently, these
two schemas in different contexts created a network of several related
meanings that will be analyzed in the following sections.
3. The rise of approximative nešto
3.1. Contextual extensions
The analysis of the collected examples has helped identify four quantifica-
tion functions of nešto. They occur in different contexts and all four are
characterized by the inferential meaning of ‘smallness’ or ‘paucity’ and
approximative quantification. The reason for the contextual extensions of
nešto lies in its simple semantic matrix. Langacker claims that “[t]he ma-
trix for most predicates is complex, requiring specifications in numerous
domains both basic and abstract” (1987: 163). Since nešto does not have
specifications in any of the domains, it is prone to semantic shift in con-
texts where the profiled entity lacks quantitative specification. In addition,
the contextually dependent quantification meaning of nešto is enriched
with a “paucal” approximation component.
Below follows the classification of the approximative nešto according
to its function in different phrases. The quantifier nešto has developed four
contextual meanings depending on its syntagmatic co-occurrence. As an
indefinite quantifier of the head in a comparative AdjP, nešto designates an
approximative small degree of difference in a shared property; with NPs,
nešto denotes an approximative small quantity of an object or a substance;
with temporal PPs, nešto approximates the time specification; and with
VPs, it quantifies a small temporal scope over an event.
3.2. Comparative quantifier (N1)10
Nešto functions as an approximative small degree quantifier in a compara-

tive AdjP/AdvP to denote a slightly higher or lower degree of some shared
property. The comparative construction evokes comparison between two
entities that possess the same property in different degrees. The approxima-
tive nešto codes a small indeterminate difference between the degrees of
the shared property in the two entities.
(9) Novata kola mi e nešto

new-the car 1SG.DAT.CL be.3SG.PRES something
pogolema od starata.
bigger than old-the
‘My new car is somewhat bigger than the old one.’
(10) Ovoj mesec rabotev nešto pomalku.

this month work.3SG.IMPERF something less
‘This month I worked somewhat less.’
3.2.1. Nominal quantifier (N2)
Nešto serves as an approximative quantifier of an NP to mark a small quan-

tity of some object or substance.
(11) Imaš nešto pari da mi

have.2SG.PRES something money SUBJ 1SG.DAT.CL
pozajmiš za taksi?
lend.2SG.PRES for taxi
‘Do you happen to have a little money to lend me for a taxi?’
(12) Verojatno ќe se najde nešto

Probably will REFL find.3SG.PRES.PF something
hrana vo frižiderot.
food in fridge-the
‘There will probably be some food in the fridge.’
In (11) the request without nešto is more direct, while the omission of
nešto in (12) conveys the meaning that the speaker’s expectations of find-
ing food in the fridge are slightly higher.
3.2.2. Approximative temporal quantifier (N3)
Nešto modifies a temporal specification realized as a PP. Before the prepo-

sition po ‘after’, nešto conveys the meaning of a small span of time after
the exact hour or part of the day indicated by the PP.
(13) Se vrati nešto po polnoќ,

REFL return.3SG.AOR something after midnight
ušte ne bevme legnati.
till not be.1PL.IMPERF lain.PART.PL
‘S/he returned a little after midnight, we had not gone to bed yet.’
(14) Menuvačnicata beše ograbena nešto

exchange-office-the.SG.F be.1PL.IMPERF robbed.SG.F something
posle eden časot popladne.
after one hour afternoon
‘The exchange office was robbed just after one o’clock in the
afternoon.’
3.2.3. Approximative “event” quantifier (N4)
In a VP, nešto functions as an approximative quantifier that quantifies over

the second participant of a process coded as the implicit direct object of the
verb (15), or over the temporal scope of a process (16). In both examples
nešto signifies more than indeterminate quantity; in addition, it conveys
the meaning of small quantity. The semantic shift proceeds from the source
meaning of approximation which, depending on the context, implies either
a small indeterminate quantity of a participant involved in the event (15),
or a small indeterminate temporal quantity of an event (16).
(15) Nešto isprav no imam ušte

something wash.1SG.AOR but have.1SG.PRES more
dve mašini za perenje.
two loads for washing
‘I did some washing but I have two more loads to wash.’
(16) Se čuvstvuvam podobro, otspav nešto.

REFL feel.1SG.PRES better sleep.1SG.AOR something
‘I am feeling better, I slept a little.’
The syntactic position of nešto depends on its semantic interpretation.

In (17) the pronominal nešto occupies the position of the direct object and
occurs in non-contrastive context (without but-sentence).
(17) Isprav nešto.

wash.1SG.AOR something
‘I washed something.’
However, in a contrastive context, due to the rise of small quantity in-

ference, nešto becomes ambiguous between the primary and secondary
meanings (18). This new bridging context paves the way for the reanalysis
of nešto into a small degree quantifier.
(18) Isprav nešto no imam ušte

wash.1SG.AOR something but have.1SG.PRES more
dve mašini za perenje.
two loads for washing
‘I washed some things but I have two more loads to wash.’
The disambiguation of nešto in (18) can occur via syntactic change: in

sentence-initial position nešto takes on the role of a small degree quantifier
(19). Thus functional change causes syntactic change – nešto undergoes
category conversion into an adverbial modifier similar to malku ‘a little’.
Having gained broader scope, nešto gains greater syntactic mobility: it
moves to the initial position (see Lehmann 1995) both with transitive (15)
and intransitive verbs (19).
(19) Otspav nešto > Nešto otspav.

sleep.1SG.AOR something something sleep.1SG.AOR
‘I slept a little.’
3.2.4. Inferences
In order to explain the mechanism of creation of the inference of (small)

quantification it is useful to recall Langacker’s claim about schematic
landmarks: “Every relational predication is attributed a landmark of some
kind as part of its profile, regardless of whether this landmark receives
overt expression through syntagmatic combination with other symbolic
units” (Langacker 1987: 163). Similarly, in the above examples nešto func-
tions as an underspecified landmark because it does not require specifica-

tions in basic and abstract domains. The lack of quantitative specification
of the scope of the event or of its unelaborated landmark (DO) gives rise to
inferences of small quantity. The meaning of ‘smallness’ results from the
semanticization of the inference ‘small thing’ inherited from the schema of
the primary nešto. The “paucal” approximation component has likely de-
veloped from the inferences of small object that the pronominal nešto cre-
ates in specific contexts. Prototypically, nešto designates a thing: a com-
pact relatively small object, as opposed to some diffuse, all-encompassing
entity. When the speaker says Gledam nešto vo…‘I see something in…’
s/he uses the pronoun nešto to refer to an unidentified but relatively small
object in a bigger “frame”. The hearer typically infers that the speaker in-
tends the prototypical meaning (if there are no indications to the contrary).
This inference may have given rise to the meaning of small quantity of the
approximative nešto.11
The second inference of approximation associated with the quantifier
nešto arises due to the strengthening of pragmatic implicatures of the pro-
nominal nešto in specific contexts (such as the above). To account for the
rise of the approximation meaning, I invoke Topolińska’s observation re-
lated to the semantics of indefinite pronouns (1974: 160). She notes that
nešto ‘something’ and nekoj ‘someone’ have acquired the ability to denote
approximation, as there exists a correlation between definiteness and the
semantic category ‘number’. Namely, the indefinite pronouns nešto and
nekoj have the ability to carry the meaning of approximation. Nekoj occurs
with cardinal numbers as in nekoi pet ženi ‘some five women’, but never
with approximative quantifiers malku ‘a little’ or mnogu ‘a lot’. I assume
that the profile of nekoj is elaborated by a precise quantitative specification
formalized as a cardinal number. When the speaker is not sure of the exact
number of the referents coded by the cardinal number, s/he may use the
plural of nekoj ‘someone’ to denote lack of certainty. In this specific con-
text, the collocation nekoi ‘some, several’ and a cardinal number creates
the new meaning of approximation of the expression.
Bearing in mind the conceptual closeness of nekoj ‘someone’ and nešto
‘something’, it is reasonable to assume that the latter contains the same
approximation semantic component that is activated only in specific con-
texts where the speaker’s communicative goal is to denote approximative
quantification over some participant or specification of an event. The most
suitable means appears to be nešto ‘something’, because nekoj ‘someone’
displays more semantic and syntactic restrictions than nešto. This sugges-
tion is supported by the fact that the sentences with nešto are conceptually
related to corresponding sentences containing NPs with ‘nekoj ‘someone’.

In the examples below, nešto quantifies over direct object NPs or time
span NPs which are explicitly specified in sentences with nekoj, but are left
implicit in sentences with nešto. The contextual implications of indetermi-
nate small quantity that NPs with nekoj create are similar to those created
by the small degree quantifier nešto. Therefore nešto in such contexts
functions as a colloquial counterpart of nekoj-NPs, but with stronger
“paucal” implications.
(20) Nešto isprav ≈ isprav nekoi raboti

something wash.1SG.AOR wash.1SG.AOR some.PL thing.PL
‘I washed a few things.’ ≈ ‘I washed some things.’
(21) Nešto po polnoќ ≈ nekoe vreme po polnoќ

something after midnight some.SG.N time.SG.N after midnight
‘A little after midnight’ ≈ ‘Some time after midnight’
(22) Nešto hrana ≈ nekoi vidovi hrana

something food some.PL kind.PL food
‘Some food’ ≈ ‘Some kinds of food’
4. The rise of epistemic nešto
The secondary nešto comprises a series of modal meanings that denote the
speaker’s decreased assertoric force. These meanings arise when the
speaker makes motivated semantic extensions from the conventionalized
secondary meaning in specific contexts. They should be viewed as contex-
tual variants because they all express the speaker’s “weakened” epistemic
force but at the same time they include “particulars of the speech situation
that are not linguistically coded” (Langacker 1987: 157).12 They differ with
respect to the salience of inferential meanings that the epistemic nešto cre-
ates in different contexts due to these “particulars”. Given the similarity of
these meanings, it is important to explore their hierarchical organization
and the mechanisms that triggered the change of one meaning into another.
The classification of the compiled examples produced five senses of the
epistemic nešto. These senses were established on the basis of semantic-
pragmatic criteria and the type of contexts in which the epistemic nešto
occurs.
4.1. “Circumstantial” epistemic nešto (N5)
The sentence initial nešto expresses the speaker’s insufficient knowledge

about the cause, reason or temporal characteristic of the event s/he is talk-
ing about. This “circumstantial ignorance” may be ascribed to the
speaker’s lack of knowledge or unwillingness to talk about the specifics of
the event. By using nešto, the speaker unobtrusively, tactfully protects
her/himself from further inquiries. In (23) the speaker does not want to talk
about the details of the agreement, in (24) s/he does not know why the
participant is summoned, and in (25) the speaker does not know or does not
want to reveal the reasons for the participants’ falling out.
(23) Dali Ana ќe se vrati do 6? Nešto se

Q Anna will REFL return.3SG till 6? something REFL
dogovaravme da izlezeme zaedno večerva.
agree.1PL.IMPERF SUBJ go-out.1PL.PRES.PF together tonight
‘Will Anna come back by 6? I think we had plans to go out tonight.’
(24) Kade e Petar? Vikni go da dojde.

Where is Peter? call.2SG.IMP him SUBJ come.3SG.PRES.PF
Nešto majka mu go vika.
something mother 3SG.DAT.CL 3SG.ACC.CL call.3SG.PRES
‘Where is Peter? Tell him to come. His mother wants him for
something.’
(25) a. Zašto ne sedat zaedno?

why not sit.3PL.PRES together
‘Why aren’t they sitting together?’
b. Nešto se zamerija pa ne se
something REFL anger.1PL.AOR and not REFL
družat.
socialize
‘They seem to be angry with each other, so they don’t socialize.’
The epistemic meaning has resulted from the metonymical extension of

the meaning of indeterminacy in the semantic structure of nešto due to the
interplay of two factors: invited inferences of causal or “circumstantial”
indeterminacy created in specific contexts and the underspecified indeter-
minate meaning of nešto. The contexts in the above examples create the so
called “circumstantial” inference because they lack specifications in the

causal or other domains of the communicated event. The newly-arisen in-
ference coded by nešto signals the speaker’s lack of knowledge or unwill-
ingness to talk about the causal specification of the event. Thus in (26)
nešto invites the inferences that the speaker does not know why his/her car
broke down or does not want to talk about it at that moment. Wanting to
imply more than s/he actually says the speaker uses nešto to replace a
causal proposition that s/he finds irrelevant for communication. As a result,
the causal or “circumstantial” inferences of nešto trigger its reanalysis into
an adverbial and its subsequent movement into the preverbal position in
declarative (26) and conditional sentences (27). It should be pointed out
that there are no semantic and syntactic restrictions in the distribution of
the “circumstantial” nešto – it occurs with verbal predicates in all tenses.
(26) Ќe zadocnam denes. Kolata nešto

will be-late.1SG.PRES today car-the something
mi se rasipa.
1SG.DAT.CL REFL break-down.3SG.AOR
‘I will be late today. It looks like my car broke down.’
(27) Ako nešto ti zatrebam, javi

if something 2SG.DAT.CL need.1SG.PRES.PF call.2SG.IMP
mi se.
1SG.DAT.CL REFL
‘If you happen to need me, call me.’
However, in interrogative sentences expressing a supposition, nešto fa-

vors the postverbal position. The interaction of sentence type and word
order may derive from the “circumstantial” inference that nešto (N5) in-
vites. As the speaker does not explicitly mention the cause of the event in a
subsequent adverbial clause, nešto takes postverbal or preverbal position.
Examples like (28) suggest a close relation between the non-fact modality
and final nešto in interrogative-negative sentences.
(28) Da ne se nerviraš nešto?

SUBJ not REFL upset.2SG.PRES something?
‘Are you upset for some reason?’
4.2. Mitigator of a modifier (N6)
As a wide scope modifier, the initial nešto mitigates the quantitative or

manner specification of an event. In (29) nešto decreases the assertoric
force of the temporal adverbial rano ‘early’, while in (30) and (31) it miti-
gates the quantifiers mnogu ‘much’ and malku ‘little’, respectively. The
inferences that the adverbials create are those of indeterminate specifica-
tion of the event coded by the predicate. The quantifiers mnogu ‘much’ and
malku ‘little’ are inherently indeterminate; similarly, adverbials like rano
‘early’ express a subjective evaluation of the event. By using nešto before
these adverbials, the speaker conveys two messages: that s/he slightly miti-
gates the “excessive” value expressed by the adverbial and that this as-
sessment reflects his/her personal view (coded by ‘seem’ in English).
(29) Nešto rano se vrati.

something early REFL return.3SG.AOR
‘He seems to have returned somewhat early.’
(30) Nešto mnogu čini.

something much cost.3SG.PRES
‘It seems to cost a lot.’
(31) Nešto malku jadeš.

something little eat. 2SG.PRES
‘You don’t seem to eat much.’
The target meaning of mitigation again results from the semantic exten-
sion of the source meaning: approximative quantification and insufficient
circumstantial knowledge. The speaker implies that s/he does not know
why the event characterized by some property possesses this property to
the extent (or degree) “quantified” by the adverbial. In addition to the “cir-
cumstantial” inference “I don’t know why p is true”, the invited inference
‘it seems to me’ arises in this context. The English translation of nešto with
the verb seem is also indicative of the “appearance” inference. The blend-
ing of these two inferences into “I don’t know why p is true but it seems to
me that p is true” produces the mitigation sense of nešto.13
With second person subjects, nešto serves to soften the disapproval of
the manner in which the addressee performs the activity. In such contexts
nešto often co-occurs with the dative short pronoun mi ‘to me’, the so
called dativus sympatheticus marker, to express the speaker’s emotional

involvement and weaken the critical remark.
(32) Nešto slabo mi učiš.

something badly 1SG.DAT.CL study.2SG.PRES
‘You don’t seem to study enough.’
4.2.1. Attitudinal epistemic (N7)
The sentence initial nešto conveys the speaker’s decreased assertoric force
about situations involving his/her physical or mental state.14 The inferential
meaning ‘it seems to me’ arises in propositions that refer to the mental or
emotional world of the speaker. In contexts with first person subjects nešto
acquires the “appearance” inference ‘it seems to me’. The new inference,
in turn, fuses with the source meaning ‘insufficient causal or circumstantial
knowledge’ inherited from the schema of the epistemic “circumstantial”
nešto. Like in the previous type, this contextual meaning also consists of
two inferences: “circumstantial” and subjective appearance. However, it
differs from the mitigating nešto in two respects: the quantification infer-
ence is missing and the “circumstantial” inference is stronger than that of
“appearance”. Due to the strengthening of the ‘it seems to me’ inference,
nešto conveys the meaning ‘for some reason I think that p is true, but I
don’t assert that p’. Therefore the non-assertoric epistemic function of
nešto is similar to the function of a pragmatic assertoric mitigator, but it
has not fully become one.
(33) Denes nešto ne mi se jade, cel

today something not 1SG.DAT.CL REFL eat. 3SG.PRES all
den samo kafinja pijam.
day only coffees drink.1SG.PRES
‘Somehow I don’t feel like eating today, I’ve been drinking coffee
all day.’
(34) Sakam da sedneme nekade. Nešto

want.1SG.PRES SUBJ sit.1Pl.PRES somewhere. something
sum umorna.
be.1SG.PRES tired.SG.F
‘I would like to sit somewhere. I feel a bit tired.’
(35) Veќe pomina polnoќ a mene nešto

already pass.3SG.AOR midnight but me.DAT something
ne mi se spie.
not 1SG.DAT.CL REFL sleep.3SG.PRES
‘It’s already past midnight but I don’t feel like sleeping.’
4.2.2. Assertoric mitigator (N8)
In contexts with non-first person subjects, the non-assertoric modal nešto

undergoes the highest degree of subjectification in order to express the
speaker’s more subjective assessment of an event in which s/he does not
participate. The meaning conveyed by nešto becomes increasingly based in
the speaker’s attitude towards the proposition (cf. Traugott 1989: 35) as it
indexes the relation between the proposition and the extra linguistic world
interpreted through the “eyes” of the speaker. This is marked by the wide-
spread use of dativus sympatheticus pronoun mi illustrated in examples
(36–38). The analysis of the corpus examples with mitigator nešto shows
that there is a correlation between the occurrence of nešto and the dative
mi. The latter, being a marker of subjectification, codes the speaker’s em-
pathy.15
(36) Nešto si mi bleda.

something be.2SG.PRES 1SG.DAT.CL pale.SG.F
‘You seem somewhat pale to me.’
(37) Nešto si mi neraspoložen.

something be.2SG.PRES 1SG.DAT.CL bad-tempered.SG.M
‘You look somewhat indisposed to me.’
(38) Nešto si mi vesela denes.

something be.2SG.PRES 1SG.DAT.CL cheerful.SG.F today
‘You look happy today.’
This highly subjective, mitigating nešto has derived from the attitudinal
epistemic nešto in the new context involving events in the real world via
the strengthening of the “appearance” (‘it seems to me’) inference. The
difference from the attitudinal nešto lies in the reverse salience of the two
inferences that constitute its meaning: “appearance” and “circumstantial”
indeterminacy. The latter has faded in this context at the expense of the ‘it
seems to me’ inference. In (39) nešto signals that the proposition that refers
to the hearer or another participant of a speech situation expresses the
speaker’s viewpoint and hence does not have to be true. In (40) this mean-
ing is enriched by the inference ‘he worries me’ coded by the dative mi.
(39) Ana nešto čudno se odnesuva vo

Anna something strangely REFL behave.3SG.PRES in
posledno vreme.
last time
‘Anna has been acting somewhat strangely lately.’
(40) Nešto mi e somnitelen.

something 1SG.DAT.CL be.3SG.PRES suspicious.SG.M
‘He looks suspicious.’
There seems to exist a correlation between the presence of the dative mi

and the assertoric mitigator (N8). The corpus examples with nešto (N8)
show that this co-occurrence is characteristic of present tense contexts that
involve second person subjects. This means that the utterances in question
function as speech acts that describe the addressee’s state or “looks”
through the speaker’s subjective assessment, while the dative mi reinforces
his/her subjectivity.16
The semantic extension of nešto to mitigation function occurs only in
contexts with second or third person subjects and propositions in the pre-
sent time frame. This semantic constraint supports Bybee et al.’s view
(1991: 30) that changes occur only in highly specific contexts and that they
reflect how the inferential meaning arises out of language use.
4.2.3. Marker of narrative hedging (N9)
In narrative discourse, nešto signals the speaker’s lack of commitment to

the truth of a proposition that refers to some past event. The speaker cannot
vouch for the factivity of this event because s/he is uncertain about the
exact circumstances.
(41) Vo emisijata učestvuvaše edna interesna

in program paricipate.3SG.IMPERF one interesting
žena. Nešto Germanka beše ili Šveġanka ...

woman something German was or Swedish
‘An interesting woman took part in the program. She may have been
German or Swedish...’
In specific types of discourse, such as narration of dreams, nešto is used

to express uncertainty about details or the course of events.
(42) Sonuvav kako otidovme vo nekoi planini. Pa

dream.1SG.IMPERF how go.1PL.AOR in some mountains. and
nešto se kačuvavme na vrvovi, pa
something REFL climb.1PL.IMPERF on tops and
nešto vleguvavme vo pešteri…
something enter.1PL.IMPERF in caves
‘I dreamt that we went to some mountains. It seemed that we had
climbed some peaks, then entered caves...’
An analysis of these examples collected from oral communication

shows that nešto has developed a series of secondary modal meanings in
specific contexts. These contexts require elaborations in five different do-
mains: cause or other circumstances of an event, degree of a property of an
event, epistemic assessment of the physical/emotional state of the speaker,
assessment of the physical/ emotional state of the hearer or another partici-
pant, and epistemic assessment of past events. In the majority of examples,
the sentence initial nešto functions as an epistemic modal marker isofunc-
tional with a superordinate predicate ‘it seems’. Nešto slightly decreases
the assertoric force of the proposition by ascribing “paucal” meaning to
propositions that the speaker quantitatively or epistemiologically assesses
in his/her mental subjective world. It signals to the addressee that the asser-
tion introduced by nešto is ranked somewhat lower in terms of its veracity
than an assertion without it, so it may not correspond exactly to the situa-
tion in the real world. The fact that nešto expresses a small degree of doubt
in the truth of the proposition but does not determine how small it is, sug-
gests that nešto has scalar properties.17 The scale invoked by nešto renders
the speaker’s own perspective and therefore nešto functions as a “weak”
epistemic adverb in the sense that it slightly reduces the assertoric force of
a proposition and therefore can be defined as a pragmatic marker of mitiga-
tion and hedging, respectively. This classification is in line with Fraser’s
definitions of pragmatic and discourse markers. According to Fraser
(1996), pragmatic markers carry additional, representational meaning that
signals both the message force and the content. They belong to aspects of
meaning subjectively construed as they linguistically encode “clues which
signal the speaker’s potential communicative intentions.” (Fraser 1996:
168). The aspect of meaning is subjectively construed in all five epistemic
senses of nešto, but to a different degree. Thus, all contextual instances of
nešto except the “circumstantial” one carry the inferential meaning ‘it
seems to me’, which is more strongly felt in the last two due to the higher
degree of subjectification.
5. Semantic change
The previous sections have investigated the secondary meanings of nešto

without discussing the reasons and the direction of its semantic develop-
ment. To account for the pathways of the semantic change of nešto it is
necessary to explain the reasons for the rise of the secondary nešto and the
development of its modal meanings
5.1. From primary nešto to secondary nešto
It was assumed earlier that the rise of the epistemic nešto was motivated by
the semantics of the pronominal nešto and R-heuristic principle. According
to the containment theory (Willet 1998), the semantic structure of the
source element contains the meaning of the target element. In the process
of the derivation of some lexical unit the basic meaning is preserved in the
semantics of the target element. In modal nešto the dominant meaning of
indeterminacy and the inference of “small” quantification blends with the
meaning of subjective impression. The rise of inferential meaning of small
quantification was already explained in section 2.5. By using nešto the
speaker signals that the small degree of uncertainty in the truth of the
proposition comes from his/her own impression. While the pronominal
nešto is characterized by referential indeterminacy in the real world, the
epistemic nešto expresses “subjective” indeterminacy which involves the
mental world of the speaker. Accordingly, this semantic change triggers the
conversion of nešto from an indefinite pronoun to an epistemic degree
adverbial.
Although the semantic change of nešto involves mapping between enti-
ties in two different domains, and therefore is an example of how a mem-
ber of one category is represented in terms of another, there are grounds to
consider it an instance of metonymy rather than metaphor.18 Metonymy

involves a conceptual mapping between entities within a single domain.
Both mechanisms are conceptual in nature and both lead to transfer of
meaning (discrete in metaphor and gradual in metonymy) so it is difficult
to determine which mechanism of change is at work in the shift of nešto. In
explanation of the semantic change of nešto I will accept Traugott’s (1988)
position on a metonymically-based change. She maintains that inferencing,
being an instance of metonymy, involves “explicit coding of relevance and
informativeness that earlier was only covertly implied” (Traugott 1988:
413), i. e. strengthening of invited inferences.
There are two more theoretical points that support the hypothesis about
the metonymic change of nešto. Firstly, as Kövasces and Radden (1998)
argue, metonymy provides mental access to another conceptual entity
within the same domain, but, on the other hand, ontological metonymies
such as CONCEPT FOR THING/EVENT metonymy (1998: 42) cut through dis-
tinct ontological realms. Presumably, the development of the secondary
nešto belongs to the ontological CONCEPT FOR THING/EVENT metonymy
because the concept of indeterminacy stands metonymically for both an
indeterminate thing and indeterminate event.
Secondly, Thornburg and Panther (1997) have shown that indirect
speech acts are cases of metonymy and that the use of metonymic, rather
than literal language is motivated by considerations pertaining to politeness
and face-saving. The use of the assertoric mitigator nešto also indicates the
speaker’s considerations for the addressee’s reactions to the assertoric
force of certain claims and therefore can be classified as a face-saving
strategy.
5.2. Modal nešto
It was suggested in the previous section that the metonymic change of

nešto was triggered by the strengthening of the indeterminacy inferences
into the meaning of the speaker’s weak epistemic commitment. The second
cognitive mechanism that played a major role in the rise of the epistemic
nešto was subjectification (see Traugott 1989, 2002). Traugott and Dasher
(2002: 30) define subjectification as a metonymically-based process that
belongs to semasiological change because it brings about meaning changes
in specific lexical items. Examples of subjectification include the devel-
opment of a lower degree of truth-conditional or referential meaning; in the
modal nešto the poor referential meaning has disappeared entirely. On the
other hand, subjectification involves internalization of external experience.

It is grounded in discourse because it expresses the speaker’s attitude to-
wards the discourse content and to the addressee (cf. Traugott and Dasher
2002: 95). This is evident in the case of nešto, which has become a suitable
marker to code the speaker’s personal stance or impression. In most mitiga-
tor examples (N8) in the corpus, nešto co-occurs with the dative clitic mi
‘to me’ to signal the speaker’s emotional involvement in the event and
familiarity with the addressee.19
(43) Nešto si mi konfuzen denovive.

something be.2SG.PRES 1SG.DAT.CL confused.SG.M days-the
‘You look somewhat confused these days.’
(44) Što ti e? Nešto si mi

what 2SG.DAT.CL is? something be.2SG.PRES 1SG.DAT.CL
bleda.
pale
‘What’s wrong with you? You look a bit pale.’
Section 2 showed that the secondary nešto consists of four quantifica-

tion meanings and five epistemic meanings; these have developed from
nešto’s semantic component of indeterminacy blended with quantification.
The existence of a common conceptual schema of indeterminacy represents
a semantic base for the creation of extended meanings. The contextual
senses of nešto are considered different instances of the schema that have
inherited its specifications (cf. Taylor 2002: 124) and in addition have
more specific properties of their own. These specifications are further de-
veloped into specific properties by invited inferences and subjectification,
the two mechanisms of change involved in the internal semantic change of
nešto. The crucial role in the activation of these mechanisms belongs to
context. Firstly, in specific contexts the communicative implicatures of
nešto increase at the expense of the cognitive meaning, due to the low in-
formational content of nešto. Thus the meaning of nešto becomes increas-
ingly context-dependent. Secondly, nešto most frequently occurs in
modally marked contexts and with lexemes that carry subjective perspec-
tives of the participants of the speech situation. The speaker-marked con-
text enhances the subjectification process responsible for the rise of several
epistemic meanings.
If the creation of the secondary nešto represents a case of coded infer-
ential change resulting in polysemy, the network of modal meanings should
be viewed as contextual variants of nešto. They have not been semanticized

as new polysemous lexemes as their meaning heavily depends on the con-
text. These modal interrelated extensions of the secondary nešto represent
pragmatic utterance-token meanings. Therefore, the extensions of the sec-
ondary nešto do not qualify as instances of polysemy; rather they represent
a cluster of extended meanings grouped around the non-assertoric epis-
temic modal nešto. In other words, the conversational implicatures of these
meanings have not become generalized (cf. Levinson 2000). As a result,
the source meaning, which is retained in each subsequent target meaning,
fuses with new inferred meanings (see tables 1 and 2). In the process of
semantic derivation these meanings acquire different salience depending
on the context. On the other hand, the two senses of the secondary nešto:
approximative quantification and epistemic modal should be viewed as
pragmatic utterance-type meanings with generalized conversational impli-
catures.
The above discussion has shown that secondary nešto functions as a
marker of epistemic modality: it codes that the speaker does not want to
fully commit himself or herself epistemically to a proposition that is inde-
terminate in certain respects. Functional shift accompanies the semantic
one: from a quantifier of a phrasal head, nešto has become a modal modi-
fier of the whole sentence. Taking into consideration that a new modal was
born, this pragmatically-driven, semantic-syntactic language change can be
described as grammaticalization rather than lexicalization.20 However,
since both the source and the target nešto belong to functional, closed-item
classes, the change of nešto may represent a special type of grammaticali-
zation.
5.3. Frequency
An important factor instrumental to grammaticalization is frequency. Two

interrelated questions should now be addressed: (i) What is the role of
frequency in the change of nešto? and (ii) Should the change of nešto be
defined as grammaticalization if the frequency criterion is applied?
According to Bybee (2003: 602), frequency is an active force instigat-
ing changes that occur in grammaticalization. Typically, high-token fre-
quency words are subject to change. The pronominal nešto has a very high
token frequency, higher than the modal nešto, a fact that can be confirmed
in any Google search. However, pronouns are more resistant to change
(e.g., they keep the case system) in comparison to full nouns (Bybee and
Thompson 2000: 68). So in spite of its pronominal status, the primary

nešto most probably became eligible for change because of its morpho-
logical unmarkedness and its indeterminacy semantics of broad referential
scope, rather than its high frequency.
The frequency criterion can be applied in defining the semantic change
of nešto as grammaticalization or just as a semantic change. The outcome
of grammaticalization is characterized by increasing frequency after the
change. As Fischer and Rosenbach (2000: 27) note, “[i]f a lexical, open-
class item turns into a functional, closed-class item, it is quite obvious that
it should be used more frequently.” Hence, the semantic shift of nešto to-
wards more subjective meaning of a mitigator should lead to more frequent
use. However, the target nešto is certainly less frequent than its pronominal
source. I assume that the decreased frequency of the grammaticalization
outcome derives from the different functional status of primary and secon-
dary nešto in different cognitive domains. During “grammaticalization” the
pronominal nešto which belongs to the content domain was reanalyzed into
secondary nešto which functions in other two domains: epistemic and
speech act (cf. Sweetser 1991). In other words, the outcome of this kind of
grammaticalization was a new lexeme, a pragmatic marker that spread into
epistemic and speech act domains (see example 11: 146) but with limited
functional application. Namely, it can be used only in certain contexts – a
fact that accounts for its lower frequency rate.
6. The semantic network of nešto
6.1. Map model of secondary nešto
All the secondary meanings of nešto are graphically represented in Figure

2. The upper diagram shows the creation of the secondary nešto by instan-
tiation of the general concept – the schema – by way of a schema-instances
triangle. The instances of nešto are linked by the relation of similarity in-
dicated by the broken line; the solid line indicates the relationship of in-
stantiation.
The lower map represents the hierarchical relations among the nine in-
terrelated senses of the secondary nešto. In this iconic semantic map model
(see Van der Auwera and Plungian 1998, Haspelmath 2003), where the
multidimensional conceptual space is reduced to two dimensions, the ar-
rows between the senses of the secondary nešto indicate derivational paths
while the spatial distance between the senses indicates cognitive distance.
primary meaning (Np) secondary meaning (Ns)
Np (prototype) Ns (extension)
indeterminate indeterminate
(small) thing relation
N schema
indeterminacy
approximation
Nq (quantifier) Nm (modal)
N1 compar.
quantifier N3 temporal N4 event N5 epistemic of
quantifier quantifier indeterminate cause
N7 attitudinal N6 mitigator
N2 nominal epistemic of modifier
quantifier
N9 hedge N8 mitigator
Figure 2. The semantic network of nešto
6.2. Inferential meaning in approximative nešto
Figure 2 can be further elaborated in two tables that show the change of
meaning of approximative and epistemic senses of nešto, a change trig-
gered by the presence of context-dependent invited inferences.
The approximative nešto results from inferences of approximation
and/or small degree/quantity that arise in particular contexts: with com-
parative AdjP/AdvP, nešto becomes an approximative small degree quanti-
fier (N1), with NPs, nešto serves as an approximative nominal quantifier

(N2), with temporal PPs, nešto functions as an approximative temporal
modifier (N3), and with VPs, it quantifies a small scope (temporal or other)
over predications (N4).
Table 1. The development of approximative nešto

(N1) (N2) (N3) (N4)
context: compAdjP/AdvP NP temporal PP VP
inference: small degree small quantity approximation small scope
6.3. Inferential meaning in epistemic nešto
Epistemic nešto has developed five context-dependent senses on the basis

of the causal/circumstantial indeterminacy inference contained in “circum-
stantial” epistemic nešto (N5). With certain adverbials, when the “circum-
stantial” inference is enriched with an “appearance” inference (‘it seems to
me’) nešto becomes a mitigator of a modifier (N6). Attitudinal epistemic
nešto (N7) is produced through the interplay of “circumstantial” indeter-
minacy inference and “appearance” inference (abbreviated: circ. ind. and
appear., respectively) in first-person contexts, while assertoric mitigator
nešto (N8) results from strong “appearance” inference and weaker “cir-
cumstantial” indeterminacy inference in non-first person contexts. The
final nešto (N9) functions as a hedge when the “circumstantial” indetermi-
nacy inference in narrative contexts develops into the meaning of speaker’s
uncertainty (uncert.).
Table 2. The development of epistemic nešto

(N5) (N6) (N7) (N8) (N9)
context: any adverbs 1st pers. other pers. narrative
inference: circ. ind. circ. ind. circ. ind. appear. circl ind.
- +appear. +appear. +circ ind. +uncert.
7. Conclusion
In the light of the previous discussion, the following conclusions can be

drawn:
Firstly, the rise of the secondary nešto represents a case of coded infer-
ential change that resulted in the creation of nine interrelated meanings.
Each secondary meaning, quantitative or modal, has developed from the

previous one via metonymic extension.
Secondly, the network of modal meanings should be viewed as contex-
tual variants of nešto, and not as new polysemous lexemes. The crucial role
in the development of the modal series belongs to subjectification. This
cognitive mechanism turns nešto into a marker that encodes the speaker’s
slightly weaker assertoric force and hence has a subjective epistemic char-
acter. As it is speaker-oriented and serves the purpose of intimacy or po-
liteness, it belongs to an “interpersonal” mode (Brinton 1996: 38).21
Thirdly, the analysis of all secondary senses of nešto shows how mean-
ings acquire different salience depending on the context. The importance
of context in the semantic change of nešto offers support for Traugott and
Dasher’s view (2002: 14) that pragmatics is the chief driving force in the
process of regular semantic change.
Fourthly, certain cross-linguistic evidence such as the presence of the
secondary nešto in other South Slavic languages, as well as the similar use
of Russian (čto-to) and Polish (coś) strongly suggests that the development
of this grammatically unmarked, highly abstract member of a reference
category into a modal mitigation marker represents a motivated change that
has to do with conceptual mechanisms of some universality.
Fifthly, the above discussion on the change of nešto represents a practi-
cal application of Taylor’s and Traugott’s models. The former accounts for
conceptual links between different meanings and the development of the
secondary epistemic meaning. The latter is used to explain the mechanisms
of this pragmatically-driven semantic change with a special emphasis on
the rise of several epistemic meanings from the secondary nešto.
Notes
1. The following abbreviations were used in the glossed examples:

ACC – Accusative, AOR – aorist, CL – clitic, DAT – dative, F – feminine,
IMP – imperative, IMPERF – imperfective tense, INF – infinitive, M – mascu-
line, N – neuter, PART – participle, PL – plural, PRES – present (imperfec-
tive) tense, PRES.PF – present (perfective) tense, Q – interrogative marker,
QT – question tag, REFL – reflexive marker, SG – singular, SUBJ – subjunc-
tive marker
2. The colloquial character of the secondary nešto influences its distribution to
the types of discourse characteristic of oral communication. The secondary
nešto is a feature of spoken language, seldom found in written language (with
the rare exception of example 4). The absence of written historic evidence has
restricted our treatment of nešto to synchronic analysis. The corpus of sen-
tences with the secondary nešto consists of 300 examples collected from vari-
ous chat rooms on the internet. It also includes examples taken from conversa-
tions or interviews that the author has conducted with students and friends.
3. Traugott and Dasher’s (2002: 16–49) IITSC model is based on semanticiza-
tion of pragmatic implicatures of a lexeme in new specific contexts. On this
view, semantic changes arise from an usage-based process in which speakers
exploit invited inferences of a lexeme and “re-analyze” context-dependent
implicatures to the point where the lexeme acquires a new meaning. The se-
mantic change from the source meanings to new coded meanings goes through
two intermediate stages (a) utterance-token meanings (or invited inferences,
IINs) and (b) utterance-type meanings (generalized invited inferences, GIINs).
Invited inferences that arise in context become generalized invited inferences
(akin to generalized conversational implicatures) after they undergo strength-
ening in specific contexts. The GIINs represent the preferred, pragmatically
polysemous meanings that have not been crystallized into new coded mean-
ings. Pragmatic strengthening is achieved via associations and metonymy
rather than analogy and metaphor.
4. The term ‘invited inferences’, borrowed from Geis and Zwicky (1971), is used
here to denote deductive processes of thought – inferences – that arise in spe-
cific contexts. They are “invited” in the sense that they are suggested by the
context. Traugott and Dasher (2002: 5) point out that they do not restrict the
term “invited inferences” to generalized implicatures. It has a broader inter-
pretation in their IITSC model as “[i]t is meant to elide the complexities of
communication in which the speaker/writer evokes implicatures and invites
the addressee/reader to infer them.”
5. Taylor’s model is based on Langacker’s network model of knowledge struc-
tures (1987: 162) according to which a knowledge system can be described as
a network of nodes and arcs. The nodes correspond to conceived entities,
while the arcs correspond to the conceived relationships between the entities.
6. Topolińska distinguishes pronominal and nominal nešto; the latter has nomi-
nal categorial properties and can be pluralized: Mnogu nešta go grizea nego-
voto srce ‘A lot of things tormented his heart’ (1974: 168). The process oppo-
site to the conversion of a pronoun to a nominal category is noted by Heine
and Kuteva (2001: 196). They underline the close relation between indefinite
pronouns and nouns denoting ‘thing’ cross-linguistically and argue that “[t]his
grammaticalization appears to be a more general process whereby generic
nouns give rise to pronominal categories.”
7. Van Kemenede (1999: 1001) finds that “[t]hose elements that have under-
specified lexical meaning are more prone to grammaticalization than others.”
8. The term ‘epistemic force’ is used here to denote the degree of belief with
which one asserts the content of an assertive speech act (cf. Sweetser 1991:
84). For indirect-quote constructions Givón (1990: 531) uses the term ‘epis-
temic commitment’.
9. Relations are conceptually dependent because they cannot be conceptualized-
without also conceptualizing the entities they interconnect. Langacker (1987:
215) maintains that a nominal predication profiles a thing, i.e. a region in
some domain, whereas a relational predication puts interconnections in pro-
file, rather than presupposing them as part of the base.
10. The symbols (N1), (N2) etc stand for the nine secondary meanings of nešto
(N), represented in Figure 2.
11. I would like to thank the editors for this insightful suggestion.
12. According to Langacker (1987: 157) the contextual meaning of an expression
represents: “[t]he richly detailed conceptualization that constitutes our full
understanding of the expression and includes all relevant aspects of the con-
ceived situation.”
13. The view that inferences, as part of the inherent meaning of a lexeme, can be
coded into a new meaning was expressed by Bolinger in the early 1970s: “It is
probably a fact, that in the course of time, inferences become references”
(1971: 522). Traugott and Dasher (2002: 5) by using the term ‘invited infer-
ences’ emphasize the active role of the speaker in rhetorical strategizing of the
communicative act. Therefore, in the treatment of nešto the term ‘invited in-
ferences’ or ‘inferences’ is preferred to ‘context-induced inferences’.
14. The Russian indefinite pronoun čto-to ‘something’ may also assume the func-
tion of an attitudinal epistemic especially in negative assertions to “soften” the
force of personal feelings or attitudes:
(i) Mne čto-to ne xočetsya idti domoj.

to-me something not feel.3SG.PRES.REFL go.INF home
‘I somehow don’t feel like going home.’
15. Examples (36–38) without the dative mi ‘to me’ are characterized by the pres-
ence of two inferences: speaker’s impression ‘it seems to me’ and mode of
knowledge ‘I don’t know why’. The clitic mi enriches the implicational mean-
ing of examples (36–38) with the “empathy” inference: ‘I feel sorry for you’
(36), ‘that worries me’ (37), and ‘I feel happy for you’ (38).
16. In Macedonian and Serbian corpus examples, the dative mi typically co-
occurs with statal predicates such as upset, familiar, suspicious, tired, pale,
confused, etc. In Macedonian example (i) mi conveys the speaker’s compas-
sion or involvement; mi can reinforce the irony of the whole utterance as illus-
trated in Serbian example (ii). It is worth noting that there were no sentences
with “appearance” nešto and mi among Bulgarian examples.
(i) Nešto si mi nervozna denes. Nervozna si, neli?

something are to-me upset.SG.F today. upset.SG.F are QT
‘You look somewhat upset today. You are, aren’t you?’
(ii) Nešto si mi poznat. Da nismo

something are to-me familiar.M.SG SUBJ not-be.1PL.PRES
zajedno ginuli za otađbinu?
together die.Pl. PART for fatherland
‘You seem familiar to me. Haven’t we fought for the fatherland to-
gether?’
17. Schwenter and Traugott (2000: 16) note that any lexeme recruited into an
epistemic modal domain acquires scalar values. On how non-scalar expres-
sions acquire scalar values, see König’s (1991) treatment of the particle even.
18. Heine et al. (1991) consider metaphorical extension as the main mechanism of
semantic change. Similarly, Sweetser (1991: 19) underlines the role of meta-
phor in the creation of polysemy patterns based on metaphorical intra-domain
connections. Arguing that “[w]e cannot escape using the general concept of
multiple domain-structures in our analysis of linguistic meaning” (1991: 147),
she distinguishes three different domains represented by sentences: real world
content, epistemic and speech acts. The latter two domains are structured in
terms of the basic content domain (1991: 146). However, Traugott and Dasher
argue (2002) that metaphorisation often is an outcome of metonymical
change. They note that “[n]either conceptual metaphorisation nor conceptual
metonymisation in principle exclude each other: easily comprehended meta-
phors are consistent with typical associations; both exploit pragmatic mean-
ing; both enrich meaning” (Traugott and Dasher 2002: 29). Similarly, the
metaphorisation of the pronominal nešto into an epistemic modal may have
metonymical origins in inferences arising in particular contexts: the link
between the notion of referential indeterminacy and the notion of epistemic
uncertainty may become metaphorical through the process of decontex-
tualization and conventionalization.
19. The use of dative in highly subjective contexts is not accidental. Manoliu
(2001: 304) claims that “[t]he dative case carries a special pragmatic connota-
tion, since it points to the most salient constituent after the Agent.”
20. The change of nešto qualifies as grammaticalization rather than lexicalization.
Both processes are historical changes that result in the production of new
forms: grammaticalization in functional and lexicalization in lexical forms
(see Brinton and Traugott 2005: 96–97). Lexicalization usually involves fu-
sion of a compound or a syntactic construction into a new lexical, contentful
form. The inputs of lexicalization are semantically highly specified items, a
criterion that excludes nešto from being considered an eligible candidate for
lexicalization change.
21. According to Brinton (1996: 38), “[t]he interpersonal mode is the expression
of the speaker’s attitudes, evaluations, judgments, expectations and demands,
as well as of the nature of the social exchange, the role of the speaker and the
role assigned to the hearer.”
References
Bolinger, Dwight
1971 Semantic overloading: a study of the verb ‘remind’. Language 47:
522–547.
Brinton, Laurel J.
1996 Pragmatic Markers in English. Berlin/New York: Mouton de
Greyter.
Brinton, Laurel J., and Elizabeth C. Traugott
2005 Lexicalization and Language Change. Cambridge: Cambridge Uni-
versity Press.
Bybee, Joan, William Pagliuca, and Revere D. Perkins
1991 Back to Future. In Approaches to Grammaticalization. Vol. 2,
Traugott and Heine (eds.), 17–58. Amsterdam: John Benjamins.
Bybee, Joan, and Sandra A. Thompson
2000 Three Frequency Effects in Syntax. Berkeley Linguistic Society 23:
65–85.
Bybee, Joan
2003 Mechanisms of Change in Grammaticalization: The Role of Fre-
quency. In The Handbook of Historical Linguistics, B. D. Joseph and
J. Janda (eds.), 602–623. Oxford: Blackwell.
Croft, William, and D. Alan Cruse
2004 Cognitive Linguistics. Cambridge: Cambridge University Press.
Fraser, Bruce
1996 Pragmatic Markers. Pragmatics 6: 167–190.
Fischer, Olga, and Anette Rosenbach
2000 Introduction. In Pathways of Change. Grammaticalization in Eng-
lish, Olga Fischer, Anette Rosenbach, and Dieter Stein (eds.). Am-
sterdam: John Benjamins, 1–37.
Geis, Michael L., and Arnold M. Zwicky
1971 On invited inferences. Linguistic Inquiry 2: 561–566.
Givón, Talmy
1990 Syntax. A Functional-Typological Introduction. Vol. 2. Amsterdam:
John Benjamins.
Haspelmath, Martin
2003 The geometry of grammatical meaning: semantic maps and cross-
linguistic comparison. In The New Psychology of Language. Vol. 2,
M. Tomasello (ed.), 211–242. Mahwah, NY: Lawrence Erlbaum As-

sociates.
Heine, Bernd, Urlike Claudi, and Friederike Hünnemeyer
1991 Grammaticalization: A Conceptual Framework. Chicago: University
of Chicago Press.
Heine, Bernd, and Tania Kuteva
2001 World Lexicon of Grammaticalization. Cambridge: Cambridge Uni-
versity Press.
Horn, Laurence
1984 A new taxonomy for pragmatic inference: Q-based and R-based
implicatures. In Meaning, Form and Use in Context, D. Schiffrin
(ed.), 11–42. Washington, DC: Georgetown University Press.
Koneski, Blaže
1987 Gramatika na makedonskiot jazik [Grammar of Macedonian Lan-
guage]. Skopje: Kultura.
König, Ekkerhard
1991 The Meaning of Focus Particles: a Comparison Perspective. Lon-
don: Routledge.
Kövesces, Zoltan, and Günter Radden
1998 Metonymy: Developing a Cognitive Linguistic View. Cognitive Lin-
guistics 9: 37–77.
1987 Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford Uni-
versity Press.
Lehmann, Christian
1995 Thoughts on Grammaticalization. München: Lincom Europa.
Levinson, Stephen C.
2000 Presumptive Meaning: The Theory of Generalized Conversational
Implicature. Cambridge, MA: MIT Press.
Manoliu, Maria M.
2001 Conventional Implicature and Language Change. In Historical Lin-
guistics 2001. Selected Papers from the 15th International Confer-
ence on Historical Linguistics, B. J. Blake and K. Burridge (eds.),
303–320. Amsterdam: John Benjamins.
Schwenter, Scott, and Elizabeth C. Traugott
2000 Invoking Scalarity: the development of in fact. Journal of Historic
Pragmatics 1: 7–25.
Sweetser, Eve E.
1991 From Etymology to Pragmatics. Cambridge: Cambridge University
Press.
Taylor, John R.
2002 Cognitive Grammar. Oxford: Oxford University Press.
Thornburg, Linda, and Klaus-Uwe Panther

1997 Speech act metonymies. In Discourse and Perspectives in Cognitive
Linguistics, W. Liebert, G. Redeker, and L. R. Waugh (eds.), 205–
219.
Topolińska, Zuzanna
1974 Gramatika na imenskata fraza vo makedonskiot literaturen jazik
[The Grammar of the Noun Phrase in Macedonian Literary Lan-
guage]. Skopje: MANU.
1988 Pragmatic strengthening and grammaticalization. Berkeley Linguistic
Society 14: 406–416.
1989 On the rise of the epistemic meaning in English: An example of sub-
jectification in semantic change. Language 65: 31–55.
Traugott, Elizabeth C., and R. B. Dasher
2002 Regularity in Semantic Change. Cambridge: Cambridge University
Press.
Van der Auwera, Johan, and Vladimir A. Plungian
1998 Modality’s Semantic Map. Linguistic Typology 2: 79–94.
Van Kemenade, Ans
1999 Functional categories, morphosyntactic change, grammaticalization.
Linguistics 37: 997–1010.
Willet, Thomas
1998 A cross-linguistic survey of the grammaticalization of evidentiality.
Studies in Language 12 (1): 51–97.
ȱȱ
ȱǱȱ¢ȱȱȱ
ȱ
Iconicity and linear ordering of constituents within
Polish NPs
Elżbieta Tabakowska
What is uppermost in mind comes up first
(Dwight Bolinger)
Abstract
The paper deals with the ordering of multiple (mainly double) adjectival
modifiers within Polish nominal phrases. Linear ordering of constituents
within those structures, generally considered haphazard and erratic, is shown
to be motivated. Two basic assumptions are made: (i) although in the dia-
chronic development of language the motivation can be bleached and the
particular structure sanctioned by convention, linear arrangement of ele-
ments within an expression is always motivated, and (ii) the motivation is
basically iconic. The structure of NPs is shown to depend on the traditional
division of adjectives into two categories: the characterising (attributive)
and the specifying (restrictive). The borderline between the two, however, is
fuzzy, with the allotment of an adjective to one or the other category de-
pending on communicative needs, which are often discourse-sensitive. In the
speaker’s choice of a construal, iconic principles conspire (or are overrid-
den) for optimum effect.
Keywords: categories, characterising adjectives, phrasal integrity, prox-

imity, salience, scope of modification, sequentiality, specifying adjectives,
thematic-rhematic structure
1. Introduction
Irrespective of actual language data selected for description and of the

theoretical framework chosen for analysis, rules offered by contemporary
linguistics generally fall into one or the other of two basic categories. On
the one hand, there are objectivist descriptions of conventional aspects of
linguistic structure that produce systems of conventional (or symbolic)
rules; on the other, subjectivist analyses are made which lead to the formu-
412 Elżbieta Tabakowska
lation of principles (rather than rules sensu stricto) that are basically ex-
pressive (or iconic) in nature. The latter are considered to be more univer-
sal because of the nature of mechanisms that motivate them (cf. Haiman
1985, ch. 6). In spite of its search for language universals, highly formal-
ised contemporary mainstream linguistics, firmly rooted in Saussurean and
American structuralism, has been traditionally focusing on rules of the first
type. This general approach applies also to what constitutes the topic of the
present paper, that is, the order of constituents within the nominal phrase.
The area of interest is fairly narrow: the following discussion will focus on
the ordering of multiple (mainly double) adjectival modifiers within Polish
nominal phrases. It could indeed be hard to justify this limitation, unless it
can be claimed that it might have some more general implications. And this
is precisely the claim that the present paper makes: the case, limited as it is,
clearly demonstrates that there are linguistic phenomena which – seen from
the traditional objectivist point of view – seem haphazard and erratic, or at
best “extremely troublesome” (cf. Łuczyński 1993), but when they are
analysed in a different, wider perspective, obvious regularities can be re-
vealed. In the case under discussion, these regularities stem from iconic
motivation.
In recent Polish works that deal with word order phenomena in general,
and with linear ordering of modifiers within the nominal phrase in particu-
lar, two main trends may be observed: analyses focus either on syntax,
aiming at formulating grammatical sequencing rules, or on the rhematic-
thematic (topic-comment) sentence structure (for a discussion of these two
general directions, see Zakrzewska 2001: 661). Within the first group, one
of the most comprehensive descriptions was offered by Topolińska (1984:
381 ff.), whose basic assumption is that sequencing of elements within the
NP reflects the “derivative history” of the phrase (cf. also Gębka-Wolak
2000: 139). However, the source of direct inspiration for the present paper
was a recent study of the problem, contained in chapter 3 of the monograph
written by a Polish linguist of generativist persuasion, Dorota Szumska (cf.
2006: 166 – 176).
Like Topolińska, Szumska gives a detailed and insightful description of
what things are like, without, however, answering the question why they
are as they are. Numerous works by Polish authors – older and more recent
alike (cf. e.g. Śliwiński 1990, Łuczyński 1993, Gębka-Wolak 2000) –
explicitly postulate the need for integrating the syntactic, the semantic and
the pragmatic perspectives as a necessary prerequisite to any adequate de-
scription of word order phenomena. Moreover, careful reading of these
works reveals that many authors do in fact refer, albeit only implicite, to
Iconicity and linear ordering of constituents within Polish NPs 413
what is defined as principles of iconicity. There are also some who are
quite explicit: in Zakrzewska’s discussion of functional explications of
linear ordering within Polish grammatical structures, iconicity appears as
one of “the four most important principles that govern word order in Pol-
ish” (2001: 668; translation – E.T.). As will be demonstrated by the follow-
ing discussion, the other three factors can be indeed classified as various
manifestations of iconicity.
The increasing bulk of evidence from linguistics, psychology and cogni-
tive science has been showing that iconicity, defined as “form miming
meaning”, or the perceived similarity between the shape of a sign and the
object that this sign stands for, is one of the basic mechanisms that underlie
cognitive processes taking place in the human mind and that it manifests
itself in linguistic expressions as a reflection of those processes in human
language. Iconic motivation is both universal and ubiquitous in natural
languages, and thus it is only natural to expect that its workings will be-
come apparent also in word order phenomena.
Iconic motivation of linguistic structure – and of linear arrangement of
constituents in particular – has recently attracted the attention of linguists
of the cognitive persuasion. In general, discussions that deal more or less
explicitly with iconic motivation as a mechanism governing word order
focus on the word order of main sentence constituents in SVO vs. SOV
languages (cf. e.g. Hopper and Traugott 2003, Kubiński 1999). As far as
our present area of interest is concerned, the noun-adjective ordering was
described for Old English by a Dutch historical linguist Olga Fischer
(2000), and for Modern English, briefly but insightfully, by the German
cognitive linguist Guenter Radden (1991). In the same cognitivist vein, the
ordering of modifiers within Modern Polish NPs was dealt with by Taba-
kowska (2001a); for Modern Italian, a similar study (though not openly
“cognitive” in the theoretical approach) was carried by Magajewska
(2006).
In spite of numerous attempts made within various theoretical frame-
works, principles governing the ordering of multiple (double) adjectival
modifiers within Polish NPs have not been systematically described, and
the question remains basically unresolved. The present paper is yet another
attempt at providing an answer. Underlying the analysis are two fundamen-
tal assumptions: (i) linear arrangement of elements within an expression is
always motivated (although in the diachronic development of language the
motivation can be bleached and the particular structure sanctioned by con-
vention), and (ii) the motivation is basically iconic. Before we proceed to
the actual analysis one important proviso should be made: “[the] act of
interpretation must always proceed from meaning to form and not the other
way round” (Nänny and Fischer 1997: 2): motivation that underlies the
form of an expression can only be searched for after the meaning of the
expression had been established.
2. Discussion
It goes without saying that every pragmatically-oriented study should be

based on a large corpus of authentic language data. The present paper does
not fulfill this condition, since the data (unless stated otherwise, taken from
popular Polish fiction) will only be used as illustration of a more general
problem. We shall analyse a couple of authentic examples of structures
containing sequences of attributive adjectives used as pre-positioned noun
modifiers, leaving aside predicative structures and constructions with ad-
jective modifiers in postposition. Thus the discussion will be limited to
such modifiers that can naturally occur in a sequence preceding the noun;
they will be analysed with two questions in view: how they are ordered and
why the particular order was chosen.
2.1. Formal motivation
Formal factors governing linear sequencing of modifiers are thus of inter-

est here only in so far as they might be found to interfere with semantic and
pragmatic considerations (see 2.4. below). Thus what should certainly be
mentioned in this context is the general principle known as language in-
dependent preferred order of constituents, whereby phonologically shorter
(or “light”) constituents precede elements that are longer (or “heavy”) (for
a discussion, see Zakrzewska 2001: 666 – 667). The length and the level of
complexity of modifying structures (or their relative “lightness” or “heavi-
ness”) is also found to influence the positioning of modifiers in Polish
NPs: shorter phrases tend to occur in preposition, and longer ones – in
postposition (Gębka-Wolak 2000: 134), irrespective of their semantic func-
tion (for a discussion, see Tabakowska 2001a).
2.2. Semantic motivation
The generally accepted semantic rule governing the linear ordering of ad-
jectival modifiers within Polish NPs, whereby characterising (attributive)
modifiers occur in preposition and specifying (restrictive) modifiers in
postposition, is believed to reflect mental processes of categorisation. The
assumption is that categories are arranged in a hierarchical system, with the
logical relation of inclusion holding between categories of higher and
lower levels. This model, going back to the Aristotelian system of categori-
sation, is fundamental for generative semantics. Applied to the traditional
classification of Polish adjectives, it also underlies the basic dichotomous
division of adjectives into those that “characterise” objects to which they
refer and those which perform the function of “specification”. The former
are used to describe ad hoc categories, which are created by the human
mind to express occasional and pragmatically conditioned relations be-
tween objects and their attributes; the latter serve the purpose of referring
to “objective” categories: such that correspond to well-established relations
perceived in the surrounding reality. In other words, while the latter corre-
spond to the Aristotelian differentiae specificae, the former reflect “free
choice” on the part of the speaker (for a discussion, see Tabakowska
2001a: 579). Hence the claim – made also by linguists describing lan-
guages other than Polish – that while the “specifying” adjectives have “lit-
eral, proper meaning, relevant for the contents of the noun”, the “character-
ising” ones “have meanings that are less precise, general, stylistically
marked, often figurative” (Magajewska 2006: 9).
Although the difference, traditionally described as “syntactico-
semantic”, is found to condition syntactic rules of pre- and postposition (in
such languages which allow both positions; but consider also the descrip-
tion of Old English NP in Fischer 2000), the same claim is made, mutatis
mutandis, in reference to the ordering of multiple elements within phrases
exemplifying any of the two positions: in either pre- or postposition, in
“natural” (unmarked) word order “characterising“ adjectives are found to
precede the “specifying” ones.
On the other hand, in nominal phrases containing multiple “specifying”
modifiers, the ordering reflects the rule (which is not language-specific)
pertaining to the scope of modification – the closer an attribute is to the
noun, the narrower becomes the scope of modification:
(1) (…) ten taki mały owalny koralik

(…) that such little oval-shaped bead
‘You know, that little oval-shaped bead.’ (Topolińska 1984: 376)
However, what becomes problematic in confrontation with actual lan-

guage data is finding the criteria that govern the dichotomy itself: the same
lexical items can serve as either “characterising” or “specifying”, which
makes some linguists believe that “the internal ordering within segments
depends entirely upon communicative intentions of the author of the text,
which becomes reflected in the linear way by the derivative history of the
phrase” (Topolińska 1984: 383, emphasis added). Or, in other words,
pragmatic considerations override the “objective” (or objectivist) semantic
factors. The reason for this, however, is no longer a matter of controversy:
like in biology, categorisation in natural language comes closer to Eleanor
Rosch’s prototype theory rather than to Carolus Linnaeus’s system of sci-
entific classification. Moreover, since categories are created rather than
discovered, their number, ontological status and membership are not set
once for ever: “specification” can be degraded to mere “characterisation”,
and “characterisation” elevated to “specification”, with “communicative
intentions” of speakers being the decisive factor.
This explains why there exist potential combinations which are (by de-
fault) considered unacceptable on cognitive grounds, such as
(2) a. dobra wysoka dziewczyna

good tall girl
‘a good tall girl’
which “strike one as incongruous on purpose” (example and assessment

from Topolińska 1981: 167). Like Topolińska, we shall claim that it would
be difficult to find a situation in which the first adjective might be used as
an attribute characterising in a pragmatically relevant way the category of
‘tall girls’. Concatenation of the two attributes seems equally unwarranted
(? ‘good and tall’). On the other hand, as noted by Szumska,
(2) b. dobry wysoki koszykarz

good tall basketball-player
‘a good tall basketball player’
is acceptable, as the formation of the category ‘tall basketball player’ is

justified on pragmatic grounds (cf. Szumska 2006: 173). Pragmatic motiva-
tion proves to be decisive.
2.3. Pragmatic (iconic) motivation
As has been said, iconicity is, more or less explicitly, taken to be one of the
“most fundamental principles governing word order” (Zakrzewska 2001:
63). In her discussion on word order phenomena in modern Polish (which
includes a section on the ordering of adjectives within the NP),
Zakrzewska (2001: 664 – 665) enumerates three types of iconicity: sali-
ence, scope of modification and phrasal integrity. Other factors – logical
consequence, increasing complexity and rhematic-thematic structure – are
ascribed to mechanisms other than iconicity (Zakrzewska 2001: 668). The
remainder of the present paper aims at justifying the claim that linear ar-
rangement of adjectival modifiers within Polish NPs is basically pragmatic,
with iconicity principles underlying all word order varieties. It is those
principles (rather than rules) of iconicity (which will be classified as dia-
grammatic rather than imagic; for a discussion of this typological distinc-
tion, see Tabakowska 2001b) that govern “communicative intentions” of
speakers, and the same principles – petrified as “generalized intentions” in
the diachronic development of language – ultimately give rise to grammati-
cal conventions.
2.3.1. Principle of proximity
The principle of proximity is a “modernized version of Behagel’s first law;

it states that things that are conceptualized as being close conceptually tend
to be put close together in linguistic expressions” (Tabakowska 2001b: 8).
In other words, psychological (or cognitive) distance becomes reflected in
the distance between constitutive elements within linguistic structures.
With the noun constituting the core of an NP, the distance is that between
the noun and its modifiers. Thus in
(3) długa brązowa sztruksowa sofa

long brown corduroy sofa
‘a long brown corduroy sofa’
the most inherent property of the object (sofa ‘sofa’) is the material it is
made of, and thus the adjective takes the position closest to the noun. The
colour (brązowa ‘brown’), also inherently connected with the object, fills
the next closest slot, while the adjective referring to size – and thus less
objective and prone to reflect the speaker’s individual relative assessment –
is situated in the position most distant from the noun. None of the three
adjectives is a good candidate to grasp a pragmatically justified stable dif-
ferentia specifica: all are “characterising” rather than “specifying”. In con-
sequence, any change in the ordering of the three adjectives yields a con-
struction which would be judged unacceptable. Now consider
(4) różowa domowa spódnica

pink house skirt
‘a pink house (indoor) skirt’
In (4), the property ‘house’ or ‘indoor’ (domowa) is used to define a

category of clothes (domowa spódnica ‘house skirt’) which have some
established properties (e.g. are made of soft material, have a loosely fitting
cut, etc.) that make it suitable to be worn at home. The cluster of specific
properties subsumed under the adjective domowa makes it – pragmatically
– more useful for grasping the relevant differentia specifica than the colour
adjective (różowa). A skirt comfortably worn at home must be soft, lose,
etc, but it can be of any colour. As such, the category establishing adjective
would be naturally felt to be conceptually closer to the noun.
The same principle accounts for gradual emergence in natural language
of terms which come to function as labels for categories that cognitive
linguistics defines as basic level categories: their members share a large
number of features and differ most from members of other related basic
level categories. Potential candidates for such terms (e.g. within a proper
pragmatic context at some stage of language development, the present sub-
ordinate level expressions domowa [spódnica] ‘house skirt’ and wiec-
zorowa [spódnica] ‘evening skirt’ might well develop into basic level
terms) are habitually combined with their nouns (or, in terms of the iconic
principle of proximity the conceptual distance between them becomes re-
duced) and gradually turn into well-entrenched collocations, which may
then become transformed into compounds or blends, with resultant pho-
netic and phonological changes (cf. e.g. the English rasp berry giving rise
to the modern raspberry, or the Polish żywy płot (lit. living fence) turning
into żywopłot ‘hedge’). Thus Zakrzewska’s notion of phrasal integrity in-
deed reveals its iconic motivation. Integrated phrases, however, may only
arise as the result of convention, or common consensus as to the way in
which a given fragment of the world should be perceived and conceptual-
ized.
It would thus not be unjustified to hypothesise that the change in status
of an adjective from characterising to specifying in languages like Polish,
where compounds and blends are not readily created, would result in shift-
ing the adjective from pre- to postposition, thus turning the structure into a
(potential) basic level term. In this way prepositioned slots would become
free for potential “characterising” modification. Vacillation between the
two possibilities would then mark either a transitory stage in diachronic
development or an incidental “one-shot” pragmatic motivation:
(5) a. (...) wszystkie miały mocny, sceniczny makijaż

(…) all had strong stage makeup
‘(…) they were all wearing strong stage makeup’
b. zestaw do makijażu scenicznego

kit for makeup stage
‘a stage makeup kit’
In (5a) the adjective sceniczny ‘stage’ performs the characterising func-

tion, while in (5b) it specifies a subcategory of the category ‘makeup’.
What merits additional comment is perhaps the use (or lack of use) of
the comma in expressions like (5a). One of “the most troublesome issues”
in the description of Polish punctuation (Łuczyński 1993: 334), the ques-
tion concerning the principle that governs the use of punctuation in such
structures has not yet been resolved in a satisfactory way. Some authors
classify this kind of comma as “facultative”, considering it a “mere reflec-
tion of the author’s or the editor’s mannerism” (Topolińska 1984: 81),
others, while acknowledging its role in imposing upon the reader “the most
natural reading” (Szumska 2006: 175), fail to provide clear criteria of such
“naturalness”. The comma – a written counterpart of an intonation break –
prevents readings that assume phrasal integrity, or – in terms of iconicity
principles – marks the existence of a cognitive distance between notional
constitutive elements within an overall conceptualization. Conversely, its
absence suggests the reduction of this distance. In (5a) the speaker charac-
terises ‘their’ makeup as ‘strong’ (an autonomous subjective assessment),
which brings about an association with ‘stage makeup’. The makeup is thus
‘strong’ and ‘stage-like’, while in the comma-free variant of (5a),
(5) c. (...) wszystkie miały mocny sceniczny makijaż

(…) all had strong stage makeup
‘(…) they were all wearing strong stage makeup’
the makeup is strong (even) for the stage makeup as a specific category,
and iconicity imposes a default reading, with characterising modifiers pre-
ceding specifying ones.
Unlike in mental images, in linguistic renderings of images linear struc-
ture of language imposes ordering of individual elements of descriptions;
users of language try to override this limitation, an example being lexical
blends or portmanteau words. Whenever two or more modifiers are treated
“on a par”, that is, should indeed be expressed simultaneously as reflecting
simultaneity of perception, less innovative constructions allow for alternate
orderings:
(6) a. mała cudowna knajpka

small wonderful restaurant.DIMIN
‘a small wonderful restaurant’
b. cudowna mała knajpka

wonderful little restaurant.DIMIN
‘a wonderful little restaurant’
But as shown by the English equivalents, even then the variants are not
“free”, as one or the other property of ‘the restaurant’ still “comes first”.
And sequentiality itself – apart from the resulting notion of distance – has
its own important role to play.
2.3.2. Principle of sequentiality
The principle of sequentiality refers to what has a long story as the concept
of ordo naturalis; primarily, it is manifested as a diagram that reflects tem-
poral or spatial succession, but – by metonymic and metaphoric extensions
– can also signify many abstract notions, such as continuity or change. It is
well known, for instance, that the default reading of a narrative imposes
upon the reader a chronology of events corresponding to the order in which
they had been described.
Adjectives express atemporal relations, therefore their linear ordering
within NPs reflects what cognitive grammar defines as a variety of abstract
motion: a mode of viewing a scene through sequential scanning, i.e. by

perceiving its particular aspects in temporal succession. This “perceptual
linearity” (term from Bolinger 1952) accounts for the ordering of modifiers
in structures like
(7) wysoka, piękna, ciemnowłosa kobieta

tall beautiful darkhaired woman
‘a tall, beautiful, dark-haired woman’
where the linear ordering of adjectives reflects the actual “zooming” order
in which the attributes were (or can be) perceived in reality, and the com-
mas correspond to pauses, marking consecutive stages of the conceptual-
izer's zooming abstract motion. The same effect can be observed in Szum-
ska's example
(8) Turyści spacerują po wąskich, brukowanych

turists walk along narrow stone-paved
uliczkach. (Szumska 2006: 184)
streets.DIMIN
‘Tourists walk along narrow, stone-paved little streets.’
where the streets are perceived first (i.e. from a larger distance) as ‘nar-
row’ and only later (at a shorter distance) as ‘stone-paved’.
On a higher level of abstraction, linear sequentiality can be taken to re-
flect the cause-result relationship, with the spatiotemporal distance iconi-
cally (metaphorically) extended to stand for causal relationship, the natural
experiential sequence being “the cause before the result”. Such can indeed
be the explanation of Szumska’s further examples. In
(9) małe, ciasne mieszkanie (Szumska 2006: 190)

small cramped flat
‘a small (and therefore) cramped flat’
the iconic interpretation is encouraged by experience: the lack of room is

prototypically caused by the small size of the flat. On the other hand, in
(10) a. wygodny, obszerny dom

comfortable spacious house
‘a comfortable (and) spacious house’ (ibid.)
the cause-result relationship is not relevant enough (being comfortable

does not a necessarily imply being spacious, and being spacious does not
guarantee being comfortable) to prevent alternate ordering and conjunctive
interpretation:
(10) b. obszerny, wygodny dom

spacious comfortable house
‘a spacious (and) comfortable house’
In other words, although the attributes are semantically and pragmati-

cally related, none of them is perceived as being the cause of the other. In
expressions like (9) and (10) the comma – far from being a mere “manner-
ism” – signifies a conceptual pause in the conceptualization (cf. above,
2.3.4).
2.3.3. Principle of quantity
“The last principle of iconicity, that of quantity, accounts for the tendency
to maintain the relation of direct proportion between the amount of form
and the amount of meaning: more form tends to carry more meaning while
less form is usually associated with less meaning” (Tabakowska 2001b: 8).
This type of iconicity is found on all levels of linguistic structure; as far as
word order phenomena are concerned, it finds a reflection in the well-
known phenomenon of the thematic-rhematic (or topic-comment) structur-
ing of expressions, whereby the new – or most relevant – information is
placed where it attracts most attention. In Polish syntax, there are two such
positions, marked by the increased amount of stress and phonological ma-
terial: at the end and at the beginning of expressions. Iconicity will natu-
rally favour the first of these possibilities, mainly due to the working of the
simple mechanism that makes human beings put “first things first”; what
speakers usually consider most important is what is closest to them, as
assessed from their particular point of view, or “what is at the moment
uppermost in the speaker’s mind tends to be the first expressed” (Hopper
and Traugott 2003: 61). A quarter of a century earlier Dwight Bolinger said
much the same thing (cf. the motto of this paper).
Like in other instances of verbal communication, speakers may simply
agree, by convention, as to what does, or should become uppermost in
their minds and therefore first in the relevant expression in a situation of a
given type, and gradually a linguistic rule emerges that reflects the conven-
tionalization. An illustrative example is provided by the formula conven-

tionally placed in Polish obituaries:
(11) zmarł po długiej i ciężkiej chorobie

died.MASC.SG after long and serious illness
‘[he] died after prolonged illness’
In (11) both adjectives are characterising, and no obvious reason – se-

mantic, pragmatic, phonological, etc. – can be seen why the conventional-
ized fixed order should have been chosen.
2.4. Overriding the principles
2.4.1. Structural symmetry
As is the case with flouting the maxims of conversation rather than just
following them, it is overriding the principles that carries extra informative
value. A possible exception is, probably, modification of iconicity-
governed linear ordering resulting from purely phonetic, or phonological,
considerations: because of their phonetic properties, some sequences just
“sound better” than others. It would be, however, fairly difficult to find
instances where the ordering would be absolutely free from other prag-
matic considerations, and, anyway, the result can never be identical as far
as the meaning is concerned.
In (12) the parallelism of structure suggest sameness of interpretation,
while in fact the last two adjectives are both characterising, allowing for
alternative ordering (biała aksamitna tunika vs. aksamitna biała tunika; cf.
also example (4) above). Yet it is only the “knowledge of the world” that
excludes the specifying interpretation.
(12) miała na sobie różową domową spódnicę

had.FEM.SG on her pink house skirt
pod białą aksamitną tuniką
under white velvet tunic
‘she was wearing a pink house skirt under a white velvet tunic’
2.4.2. Discourse-specific pragmatic considerations
Some of the pragmatic factors that condition the speaker’s choice of an

adjective for category description are discussed at length in Tabakowska
(2001a). Let us recall at this place that the speakers’ choices result from
using different strategies chosen in different contexts, in order to achieve
best results in differing communicative situations. Thus, to quote the ex-
ample from Tabakowska, the supermarket display will offer to its patrons
sery żółte (lit. cheeses yellow ‘hard cheeses’), while the same product,
served by the same patron to his family for dinner, will be referred to as
żółty ser (2001a: 585). While the shop management needs a category (for
pragmatic reasons, e.g. better spatial orientation and clearer spatial organi-
zation of sectors), the consumer is satisfied with mere characterisation.
And in both cases iconicity is at work. In Tabakowska (2001a) the focus is
on the opposition between pre- and postpositioned adjectives, but the prin-
ciple as such holds good for the issue presently under discussion: the as-
cription of the specifying function to an adjective is ultimately a matter of
pragmatic requirements of particular discourse. As is the decision of not
allotting this function to any of the modifiers.
When the iconic principles of proximity and/or sequentiality become
overridden, it is the principle of salience that wins. Compare (10a) and
(10b), repeated below for convenience:
(10) a. wygodny, obszerny dom

comfortable spacious house
‘a comfortable (and) spacious house’
b. obszerny, wygodny dom

spacious comfortable house
‘a spacious (and) comfortable house’
The choice of ordering depends on what the speaker happens to con-

sider more salient, or “what comes first to the speaker’s mind”. Consider
(13) W szarym, idiotycznym Peerelu

in grey idiotic Polish People’s Republic
poczułem się pewnie
(I)felt myself sure
‘In the grey, idiotic Polish People's Republic I felt secure.’
The reversal of the canonical ordering (evaluation preceding inherent

property – colour) elevates the adjective ‘idiotic’ to the function of cate-
gory-forming, objective, specifying modifier, thus giving more impact to
(13). Thus what would probably be traditionally called a stylistic (free)
variant reveals an important aspect of meaning.
Ultimately, the chief purpose of using language is maintaining easy cod-
ing and decoding of verbal messages (cf. Zakrzewska 2001: 667) in order
to fulfill particular communicative purposes. Here the principle “first
things first”, translated as the definition of discoursive salience, proves
decisive – for linear ordering in the same measure as for other aspects of
language structure. Examples abound. For instance, in a recent written
advertisment of a furniture shop we find
(14) tanie piękne meble

cheap beautiful furniture
‘beautiful cheap furniture’
where the adjective ‘cheap’ is shifted to the first position (against the prin-
ciple of perceptual linearity) as the most salient from the point of view of
the dealer, trying to attract the prospective buyer. Similarly, in
(15) plastikowa, czerwona lampa

plastic red lamp
‘a plastic red lamp’
as opposed to the canonical order exemplified by
(16) białe perkalowe zasłony

white calico courtains
‘white calico curtains’
the reversal of the conventional order signifies that what the observer no-
ticed fist was the cheap material the lamp was made of, the epitome of the
cheap quality of the entire interior of which the lamp is an element (cf. also
(3) above). Such indeed is the function of (15) in its wider context (the
narrative in Anna Krzemieniecka’s novel Dziewczyna z Buenos). One
could say that in (15) the conventional iconicity principle is overridden by
“one-shot” discourse iconicity, which subdues conventional (intersubjec-
tive) perceptual sequentiality to the subjective sequentiality of an individ-
ual act of perception and conceptualization.
Linear ordering is of course closely connected with semantico-

pragmatic congruity: in any order
(17) a. ciasne, przytulne mieszkanie

cramped cosy flat
‘a cramped, cosy flat’
b. przytulne, ciasne mieszkanie

cosy cramped flat
‘a cosy, cramped flat’ (Szumska 2006: 192, cf. also (9) above)
will strike the hearer/reader as incongruous, unless an explanation is found

that makes the sequence semantically and pragmatically justified (e.g.
‘cramped but cosy nevertheless’, cf. Szumska 2006: 192).
2.4.3. Non-specifying postposition
Overriding the iconic principles of proximity and sequentiality by the prin-

ciple of salience can mean a “conflict of iconicities”, which can only be
avoided by combating default (canonical) interpretations. Thus in
(18) a. szerokie, czerwone, drewniane wsuwki

broad red wooden hairpins
‘broad and red wooden hairpins’
b. czysty, zielony wiejski pejzaż

clear green country landscape
‘a clear, green, countryside landscape’
c. mała, błękitna, wiejska chata

small blue country hut
‘a small blue country hut’
salience of notionally distant characterising attributes can be unambigu-

ously signalled by shifting them to postposition, with intonation (punctua-
tion) preventing specifying interpretation, and the paratactic conjunction
equalizing the status of the two adjectives Thus:
(19)' a. drewniane wsuwki, szerokie i czerwone

wooden hairpins broad and red
‘wooden hairpins, broad and red’
b. wiejski pejzaż, czysty i zielony

country landscape clear and green
‘countryside landscape, clear and green’
c. wiejska chata, mała i błękitna

country hut small and blue
‘a country hut, small and blue’
Alternative variants of (19) differ in the degree of acceptability:
(19)'' a. wsuwki, drewniane, szerokie i czerwone

hairpins wooden broad and red
‘hairpins, wooden, broad and red’
b. ? pejzaż, wiejski, czysty i zielony

landscape country clear and green
* ‘a landscape, countryside, clear and green’
c. ?? chata, wiejska, mała i błękitna

hut country small and blue
* ‘a hut countrylike, small and blue’
A possible explanation is that while ‘country hut’ functions as a basic

level term, ‘countryside landscape’ and (even less so) ‘wooden hairpin’ do
not. This obviously begs the question of how exactly the status of basic
level categories can be established. Any attempt at discussing this issue
would mean going far beyond the scope of this paper. Here, it must suffice
to say that pragmatic considerations are decisive, with frequency of occur-
rence of particular adjective-noun combination showing emergent basic
level meanings of fixed collocations.
2.4.4. Conflicting principles at work
In actual discourse, individual principles may conspire, but the communi-

cative effect is often achieved also as the result of conflict. Like in other
areas, pragmatic force of an expression seems to be in reverse proportion to

the degree of its formal conventionalization (for a similar claim, see
Zakrzewska 2001: 669). Consider
(20) Jej oczy duże, ciemne, wymyte spokojnym snem

her eyes large dark washed calm.INSTR sleep.INSTR
‘Her eyes – large, dark, washed clean by a peaceful night's sleep’
(20), quoted after Pawica (2006: 135), comes from a literary text (A.
Szczypiorski’s Początek) and its professional, published English transla-
tion reveals the translator’s interpretation of the meaning: the possesive
adjective her preposes the noun (as it does in Polish) and reflects the lack
of (notional) distance between the possessor and the possessed. Large, in
closest postposition, but not specifying (with the dash marking the non-
specificity in the English version), comes as the most salient element of
(subjective) perception. Dark (which canonically should come closer to the
noun, i.e. precede large), reflects the iconic perceptual “zoom”. Finally,
washed clean... is the narrator’s interpretation rather than actual percep-
tion, and as such it follows the adjectives opening the “perceptual linear-
ity” chain. The ultimate effect is what is traditionally – and imprecisely –
called “literary style”.
3. Conclusion
In linear ordering of (prepositioned multiple) modifiers iconicity principles

conspire to produce overall communicative effects. The principles, as dif-
ferentiated in theoretical descriptions of iconicity, are difficult to separate
in a practical analysis, even though an attempt to do so must often be made
for the purposes of description. Apparent free variants are always (iconi-
cally) motivated, although the working of the mechanism is not wholly
predictable. The discussion above only covers a small fraction of word
order phenomena, but in view of the nature of the mechanisms generaliza-
tion seems justified. A particular grammatical word order is a stylistic vari-
ant (or, in cognitive linguistics terms, an alternative scene construal); some
such variants become conventionalized because the speakers find them to
be communicatively (functionally) effective: they become more frequent
than others and in consequence are more readily expected by the receiver,
ultimately accounting for default readings of messages. Departures occur
following the speakers’ individual choices, giving rise to (possible) future

systematic innovations.
References
Bolinger, Dwight
1952 Linear modification. Language 67: 1117–1144.
1977 Meaning and Form. London: Longman.
Fischer, Olga
2000 Position of the adjective from an iconic perspective. In The Moti-
vated Sign. Iconicity in Language and Literature 2, Olga Fischer and
Max Nänny (eds.), 249–276. Amsterdam/Philadelphia: John Benja-
mins.
Gębka-Wolak, Małgorzata
2000 Związki linearne między składnikami grupy nominalnej we
współczesnym języku polskim [Linear relations involving the compo-
nents of the nominal group in current Polish]. Toruń: Wydawnictwo
Uniwersytetu Mikołaja Kopernika.
Haiman, John
1985 Natural Syntax. Cambridge: Cambridge University Press.
Hopper, Paul J., and Elizabeth Closs Traugott
2003 Reprint. Grammaticalization. Cambridge: Cambridge University
Press, 1993.
Kubiński, Wojciech
1999 Word Order in English and Polish. On the Statement of Lineariza-
tion Patterns in Cognitive Grammar. Gdańsk: Wydawnictwo
Uniwersytetu Gdańskiego.
Łuczyński, Edward
1993 O interpunkcji w wyrażeniach typu samotny biały żagiel i
niespokojne, wzburzone morze [On punctuation in expressions
analogous to samotny biały żagiel and niespokojne, wzburzone
morze]. Język polski [The Polish language] LXXIII (4–5): 334–339.
Magajewska, Mirosława
2006 La posizione e la funzione dell'aggettivo qualificativo in Italiano.
Summary of an unpublished Ph.D. diss., Jagiellonian University,
Kraków, Poland.
Nänny, Max, and Olga Fischer
1997 Iconicity as a creative force in language and literature. Leaflet an-
nouncing the symposium “Iconicity in Language and Literature”,
University of Zűrich, 20–22 March 1997.
Pawica, Michał
2006 The English equivalents of the Polish instrumental case. Ph.D. diss.,
Université Laval, Québec.
Radden, Guenter
1991 The Cognitive Approach to Natural Language. Series A. Paper no.
300. Duisburg: L.A.U.D.
Szumska, Dorota
2006 Przymiotnik jako przyłączone wyrażenie predykatywne [Adjective as
an attached predicative expression]. Kraków: Universitas.
Śliwiński, Władysław
1990 Łączliwość składniowo-semantyczna przymiotników z rzeczownikami
we współczesnym języku polskim [Syntactic and semantic adjective-
noun combinability in current Polish]. Kraków: Jagiellonian Univer-
sity Press.
2001a O szyku wyrazów w obrębie wyrażeń argumentowych przedmi-
otowych [On word order in object argument expressions]. Prace
filologiczne [Philological works] XLVI: 661–671. Warsaw: Institute
of the Polish Language, Warsaw University.
2001b Iconicity. In Handbook of Pragmatics, 1–17. Amster-
dam/Philadelphia, John Benjamins.
Topolińska, Zuzanna
1981 Remarks on the Slavic Noun Phrase. Wrocław: Ossolineum.
1984 Składnia grupy imiennej [The syntax of the nominal group]. In
Gramatyka współczesnego języka polskiego. Składnia [The grammar
of current Polish. Syntax], Zuzanna Topolińska, (ed.), 301–389.
Warszawa: Państwowe Wydawnictwo Naukowe.
Zakrzewska, Ewa D.
2001 Funkcjonalna eksplikacja szyku wyrazów w polszczyźnie [A
functional explanation of word order in Polish]. Prace filologiczne
[Philological works] XLVI: 661–671. Warsaw: Institute of the Polish
Language, Warsaw University.
Discourse-aspectual markers in Czech sound
symbolic expressions: Towards a systematic analysis
of sound symbolism
Masako U. Fidler
Abstract*
This article deals with variation among sound symbolic expressions (SSEs)
(uninflected, largely onomatopoeic expressions) in Czech: repetition,
lengthening, and suffix-like formations. By showing consistent correlation
between these three types of variation and functions related to aspectuality
and discourse, I demonstrate that SSEs do not merely imitate sounds and
motions, but accentuate specific aspects of events, set the locus of perspec-
tive, make implicit contrast between possible worlds or texts, foreground
events, or sum up a stretch of text. The major goal of this article is to ex-
plore the possible relationship between SSEs and grammar. Contrary to de
Saussure’s view that SSEs are “not an organic part of the language,” I argue
that there are properties that bridge SSEs and more conventionalized part of
the language. I show that the discourse-aspectual functions of variants
among SSEs can help explain some processes of word derivation and varia-
tion in language, e.g. cases of SSE-based aspectual derivation and nominali-
zation, as well as vowel length fluctuations in Colloquial Czech. Some of
the suffix-like formations among SSEs parallel quantification and diminu-
tion expressed by grammatical morphemes. By showing systematic func-
tions of components of SSEs, I argue that a SSE can be seen as a composite
structure that consists of the basic form and of a discourse-aspectual marker,
which parallels the model presented in Langacker (2003:67) for
[FLING+ER]; the basic SSE describes the properties of sound and/or mo-
tion, while the discourse-aspectual marker specifies functions related to as-
pectuality and discourse.
Keywords: Sound symbolism, onomatopoeia, Czech, aspect
432 Masako U. Fidler
1. Introduction
Sound symbolic expressions (henceforth SSEs) are commonly used lin-

guistic devices in various languages, but they by no means constitute a
homogeneous group, as shown in Hinton et al. (1994: 10), whose observa-
tions on SSEs are summarized below. SSEs can be:
(a) sounds that reflect human psychological states and physical symptoms
(e.g. sounds indicating displeasure, exclamation of surprise, cough, hic-
cups);
(b) representations of sounds produced by animate beings (e.g. chirping of a

bird, barking of a dog, hissing of a snake);
(c) representations of sounds involving and/or produced directly by inani-

mate entities (e.g. a fist hitting a door, rain falling on a surface, the splashing
sound produced by an object falling into water);
(d) representations of the perceptual characteristics of how an action is car-

ried out (e.g. speaking unclearly (e.g. mumbling), moving with one’s body
dragging heavily, piercing something with a sharp object);
(e) representations of perceptual characteristics of objects (e.g. size, shape).
The list is not intended to be exhaustive. In fact, it does not explicitly

include interjections that function as speech acts, such as the Czech pši
[pʃi], which is a SSE used to drive away small domestic animals. The list
also shows that “sound symbolism” is not necessarily a perfect label for all
the forms concerned, since the term actually includes expressions that rep-
resent not strictly sound-related phenomena, e.g. motion, size and shape.
There may also be different degrees of abstractness in representing a sound
within a group. For example, the Czech cha [xa] representing jeering of
contemptuous laughter, and n [n] expressing a variety of contradicting
emotions (e. g. agreement and disagreement, understanding and suspicion)
may belong to (a) in the list above; in such cases there is no difference
between the SSE and what the speaker actually says. However, the same
forms may be used when the speaker represents someone else’s utterance.
Such cases involve a higher degree of abstraction; the speaker may be
highlighting only selected aspects of the original speaker’s utterance (e.g.
Czech sound symbolic expressions 433
emotions), which may in reality consist of a whole sentence or a longer

stretch of text rather than a single SSE.
Finally, implicit in this list is the existence of at least two approaches to
sound symbolism. One of them is exemplified by Jespersen (1933) and
Jakobson and Waugh (1987). Their approach aims at finding consistent
connections between meaning and sound within a language, e.g. vowel
height and size-shape imagery. This is the type of sound symbolism re-
search widely known to Slavists. The other approach, common in sound
symbolism research of non-European languages, deals specifically with
expressions that primarily mimic sounds and motions; these expressions
seem to constitute a distinct category within the language because of their
phonological and morphological uniqueness. This second approach is taken
in this article for Czech. Thus, by SSEs I mean primarily those uninflected
forms representing sounds and motions. The term “onomatopoeia” loosely
covers this group of expressions, although for some of them sound-
mimicking is not the primary function (e.g. šup [=up] represents a quick
movement, and chmat [xmat] capturing of an object by hand).1
Why were uninflected SSEs in Slavic languages not studied as exten-
sively as in other languages? This situation is probably based on a belief
that onomatopoeic forms are not quite a full-fledged part of language. De
Saussure (1959: 69) states that iconic or onomatopoeic words are “not an
organic part of language”. Kořínek (1934: 173) says that onomatopoeic
forms are associated with the speaker’s “intimate and naïve contact with
nature,” and that interjections “figure prominently in the speech of a psy-
chologically primitive speaker”. Implicit in these statements are impres-
sions that (a) SSEs have little in common with the rest of the language; that
(b) SSEs are naïve and perhaps childish attempts at imitating natural
sounds.
Are SSEs truly distinct from the rest of the language? If they are not an
“organic part of the language,” why is it possible for them to function
within a sentence the way that is illustrated below?
(1) Kubát vzal cep a BÁC, BÁC! Zloděj se skácel. (Werich 1960: 61)
‘Kubát took the flail and [baɕɕts baɕɕts]!2 The thief was knocked
down.’
(2) Čert ... dělal hercum-percum a brlbrlbrl .... (Kubátová 1994: 62)
‘The devil ... did (made) [hertsum pertsum a brlbrlbrl] ...’
(3) ... colt štěk BENG! BENG! (Werich 1960: 142)

‘... the colt barked [beŋk]! [beŋk]!’
The SSEs above can function as a predicate as in (1), as a direct object

(2), and as a quotation (3).
It is true that all these forms are uninflected. The apparent lack of struc-
ture may be what separates SSEs from the more conventionalized part of
language. But recent studies of SSEs find quite the opposite. SSE are found
to be “grammatically explicit” in Pastaza Quechua (Nuckolls 1996: 11).
Rhodes (1994) observes that formation of some English SSEs can be seen
as a special type of morphological process resembling derivation. Such
studies suggest the possibility that a SSE might consist of some type of a
composite structure. Such a structure may not be exactly identical to Lan-
gacker’s model for conventionalized word forms (2003: 67), but it is rea-
sonable to investigate whether there are parallel structures for SSEs.
Besides the potential structural similarities between SSEs and gram-
matical forms, existing literature suggests that SSEs may have systematic
discourse functions. For example, Clark and Gerrig (1990) mention that
onomatopoeia, along with quotations, is a type of “demonstration,” a dis-
course process by which the speaker reports not so much the content of the
utterance as a selected form of delivery. The notion of demonstration is
illustrated by the following examples in Clark and Gerrig (1990: 764):
(4) a. She said that she wanted to buy an ant.

b. She said, “I want to buy an ant.”
The indirect quotation in (4a) describes what she said: wanting to buy
an ant. The direct quotation in (4b) depicts how she said it. Its focus is
more on the acoustic aspects of how this message was delivered than on
the content of the message. SSEs are similar to direct quotation in this re-
spect, as shown below:
(5) Vážený pane, bla bla bla… videokač, bla bla bla… proto vás zveme…
(Disney 1991: 3/6)
‘Sir, [bla bla bla]… Videoduck, [bla bla bla]… therefore we invite
you...’
The SSEs in (5) above are used to illustrate the somewhat cumbersome
manner of speaking rather than to convey the entire utterance.
Grice’s theory of conversational implicatures suggests further potential

discourse functions of SSEs. The Gricean model (Grice 1975: 45) is based
on the cooperative principle “make your contribution such as is required, at
the stage at which it occurs, by the accepted purpose or direction of the talk
exchange in which you are engaged.” This principle contains four maxims:
the maxim of quality (try to make your contribution one that is true); the
maxim of quantity (make your contribution informative as it is required);
the maxim of relevance (make your contribution relevant); and the maxim
of manner (be perspicuous). Many utterances may appear to be in violation
of these maxims. Grice argues, however, that such utterances can be inter-
preted as tacitly conveying something other than what is being said (i.e.
delivering conversational implicatures).
As SSEs are not precise reproduction of existing sounds, and as they
tend to be reduplicated or repeated, they might seem to be in violation of
the maxim of quality and manner (“SSEs are not precise”), of the maxi-
mum of quantity (“SSEs are redundant”), and with the maximum of rele-
vance (“there is no point in using SSEs”). When we consider Grice’s coop-
erative principle, however, we can anticipate that the major function of
SSEs may not be imitative, but discourse-related. In fact, Schiffrin (1987:
73–127) demonstrates that interjections such as oh and well in English
have important discourse functions.
In sum, existing literature suggests that SSEs may have structure and
that they are likely to have discourse functions. In order to investigate these
two issues, I will examine certain types of variation among SSEs –
additions or modifications of the basic shape of a SSE: lengthening, repeti-
tion, and suffix-like formations. These features occur not only in Czech
SSEs, but also in SSEs cross-linguistically. As mentioned above, a process
resembling word formation is found in English SSEs (Rhodes 1994: 299–
290). Repetition and lengthening are considered to interact with aspect in
Pastaza Quechua and Japanese (Nuckolls 1996: 62–78; Hamano 1998:65–
67, 71–73). Presented below are samples of repetition ((6a)), lengthening
of a segment ((6b)), and suffix-like formations ((6c)).
(6) a. chi [xi]: chi-chi, br [br]: brbr

b. bac [bats]: bác [baɕts], br [br]: brr [brɕ]
c. dup [dup]: dup-ity [dupiti], ťap [cap]: ťap-ity [capiti]
buch [bux]: buch-y [buxi], buc [buts]: buci
I will call these types of variation “discourse-aspectuality3 markers.”

Discourse-aspectuality markers are added to the basic form of a SSE. They
accentuate selected aspect(s) of a sound, a movement, a situation, or spec-

ify the role of the SSE in a text. A discourse-aspectuality marker may, for
example, draw more attention to the internal properties of an event than to
its relation to other events. When this occurs, the point of perspective is
placed internal to the event, and the event is presented as if it were gradu-
ally unfolding in front of the addressee. This mode of presentation is simi-
lar to a film shot in which an event progresses in slow motion with accom-
panying sound effects.4
On the other hand, a discourse-aspectual marker may also draw more at-
tention to the relationship between one event and other events. In such
cases the point of perspective is external to the event, and there is focus on
the logical or sequential order of events. This is similar to viewing a se-
quence of fast-motion actions in a film. Discourse-aspectual markers can
also signal the role of the SSE within an episode: e.g. marking the SSE as
representing the climax of an episode, or as summarizing the entire unit of
text.
Examination of discourse-aspectuality markers will be useful in three
respects. First, the notion of discourse-aspectuality markers will allow us to
see sound symbolic expressions holistically, without having to eliminate
samples excessively from the database. Second, analyzing SSEs as consist-
ing of a basic form and a discourse-aspectuality marker will suggest that
(a) SSEs are not chaotic, and that (b) SSEs can exhibit a composite struc-
ture similar to those seen in more conventionalized parts of a language.5
Third, discourse-aspectual functions of sound properties in SSEs can ac-
count for the directions of grammaticalization and de-sound-symbolization
of SSEs (i.e. development of less sound symbolic meanings); they can also
help explain the occurrences of certain sound properties in conventional-
ized parts of language that are not directly based on SSEs in variation and
derivational processes in grammar.
Samples of SSEs were collected from several sources: the 1989 diction-
ary of Standard Czech (Havránek: 1989), data extracted from the Czech
National Corpus (Syn2000 and Syn2005), and manually collected data
from texts including children’s literature, folktales, and comic strips.6
2. Discourse-aspectuality marker: repetition
The first discourse-aspectuality marker to examine is repetition. By repeti-

tion I mean here repetition of a SSE (e.g. ťuk ťuk).7 A repeated SSE (multi-
plex variant) may represent more than one occurrence of a sound, as in

(7b) as opposed to (7a).
(7) a. zase ho vedli pod šibenici,

‘they took him to the gallows again,’
zase kat šáh po vazu,
‘the executioner put his hand on the neck,’
soudce Ryba neřek nic,
‘judge Ryba said nothing,’
a křup! a Franta se houpal… na špagátě. (Syn2000)
‘and [kȋup]! And Franta was dangling again,…on the
rope.’
b. Bože, to je krása, ... to je zvláštní, jak sníh vypadá

‘God, it’s beautiful, ... it’s strange, how snow looks’
starodávně a venkovsky. Křup, křup, pan Rybka si hledal
‘ancient and rustic. [kȋup kȋup], Mr. Rybka looked for’
nepošlapanou cestičku, jen pro tu radost, že to tak křupe;
(Syn2000)
‘an untrodden path, just for the joy [of hearing] that it crunches
like that;’
In (7a) the simplex SSE is used instead of a verb. It illustrates the act of
hanging (by reporting the cracking of the neck), which constitutes one of
the sequentially ordered events. Example (7b) is different from (7a) in that
it represents more than one occurrence of the cracking sound. The differ-
ence, however, is not merely related to sound imitation. Unlike (7a), the
SSE in (7b) presents how the snow crunches as the character walks. The
point of perspective is internal to the event, giving the impression as
though the reader were seeing the snow and hearing the sounds through the
character’s eyes.
Events viewed from the internal vs. external perspective can render sen-
sations other than sound and motion as in (8a) and (8b):
(8) a. V neděli jsme zabalili, hráli chvíli kostky a po obědě

‘On Sunday we packed up, played dice for a while and after
lunch’
(houbový gulášek - mňam) šli na vlak. (Syn2000)
‘(the mushroom goulash – [mȁam]) we went to the train.’
b. Ve Francii jsem koupila Hugovi veletáfli, vážila aspoň dvě kila,

‘In France I bought Hugo a chocolate bar, it weighed at least 2kg’
a v ní oříšky jako třešně, ty chrupaly, mňam, mňam. (Syn2000)
‘and it had walnuts like cherries, they crunched, [mȁam
mȁam].’
(8a) presents one of the points in the story (packing up, playing dice,
having lunch where the speakers had a good mushroom gulash, and getting
on a train). The SSE notes the good quality of the meal and the speed with
which the food was eaten up (because it was good). In this regard, the sim-
plex variant can be viewed as part of the sequence of events. It is very
close to sentences such as: byl velmi dobrý and/or hned jsme ho snědli ‘it
[the goulash] was very good, we ate it up right away.’
The repeated mňam, mňam in (8b) is part of a text in which the speaker
recalls the pleasure of savoring the chocolates. Co-occurring with the im-
perfective verb chrupaly ‘they crunched,’ the multiplex variant stretches or
zooms in on the scene of savoring the chocolate, as if the speaker were
reliving the experience.
Functional approximations of (7a, 8a) and of (7b, 8b) are given in Fig-
ures 1 and 2 respectively. Here, t is the temporal axis, E0 refers to the event
associated with the SSE (E-1 and E+1 refer to events preceding and follow-
ing E0), and the thick lines represent the profiled parts of the event. The
line representing the event in Figure 2 is profiled in dots to show its indi-
vidual components.
E-1 E0 E+1
t
Figure 1. Functional approximation of (7a) and (8a): Simplex variants
E0
t
Figure 2. Functional approximation of (7b) and (8b): Multiplex variants
Figure 1 places the perspective external to E0; this results in profiling of

the sequential order of events rather than E0. There is more interest in the
relationship, most likely temporal and/or causal, between E0 and other
events. The internal contents of the E0 as represented by the SSE are less
accentuated.8 Figure 2 shows that the perspective in (7b) and (8b) is inter-
nal to E0. Consequently, more attention is paid to the internal properties of
the event (including quantity). The relationship between E0 and other
events is not immediately relevant.
This contrast essentially overlaps with Chung and Timberlake’s (1985:
223–224) description of perfective and imperfective verbal aspect.9 Just
like the imperfective aspect, the multiplex variant presents the event as
open on the propositional level (because the event is viewed from inside
the event); just like the perfective aspect, the simplex variant presents the
event as closed (because the event is viewed from outside the event). It is
natural that repetition in SSEs is linked to the Czech aspectual system.
Types of repetition can be found only in imperfective verbs, such as blábo-
lit ‘talk nonsense’ (cf. bla bla) and chechtat ‘laugh loudly’ (cf. che che).
Simplex SSEs combine with the semelfactive suffix -nou- to form perfec-
tive verbs: bacnout ‘hit (once)’ mňamknout ‘eat up fast.’ Forms exhibiting
some type of repetition do not easily combine with the suffix -nou-:
*blábolnout, ?chechtnout.
Repetition, however, is not merely connected to aspectuality, as shown
in (9a) and (9b):
(9) a. Řítí se k zemi hlavou dolů. ... Země

‘He is crashing head on towards the ground. …The ground’
se kvapem blíží. Bum! Hlasitý náraz. Leží na
‘quickly approaches. [bum]! A loud crash. He is lying on’
chodníku s polámanými údy. (Syn2000)
‘the sidewalk with broken limbs.’
b. Bum. Bum. Bum. Bum. To je zdola,

‘[bum bum bum bum] This is from downstairs,’
z montérského pokoje. Bum.Bum. Bum. Bum. Bum.
‘from the builder’s room. [bum bum bum bum bum].’
Bydlí tam ten hoch... Borek Trojan. (Syn2000)
‘That fellow... Borek Trojan lives there.’
The simplex variant in (9a) represents the most dramatic event within
the stretch of text. The event of the fall is not only part of a sequence of
events, it is likewise the most foregrounded part of the storyline. The mul-
tiplex variant in (9b) can be considered a discourse correlate of (7b, 8b). It
represents a sound that resonates simultaneously as the character talks. The
point of perspective in this example is internal to the entire speech situa-
tion.
Figures 3 and 4 are functional approximations of (9a) and (9b). T0 is the
current segment of text. T-1 is the text segment that precedes T0 and T+1 the
text segment that follows it. The arrow in Figure 3 represents text progres-
sion. In Figure 3 the text segment T0 is viewed as more important and pro-
filed in relation to the other text segments that belong to the same episode,
T-1 and T+1. The perspective point is outside T0. In Figure 4, the perspective
point is internal to the entire stretch of text, creating the impression that the
reader is inside the speech event, hearing the utterance and the sound si-
multaneously. The entire T0 is profiled. The SSE is represented as a dotted
line as a part of this text; it is dotted to reflect focus on details (including
quantity) of the sound.
text progression
T-1 T0 T+1
Figure 3. Functional approximation of (9a): A simplex variant
T0
Figure 4. Functional approximation of (9b): A multiplex variant
The discourse function of multiplex variants is often outside the dialog

bubbles and handwritten in cartoons as in (10):
(10) “Ten muž musí být šílený.” “Ani ne šílený, jako spíš...”
Uch juch juch uch
(www.idnes.cz/IdsKosile/MIMORADNE/KOMIKS/komiks.asp?x
=include/komiks/_moravicus/2001/0519/0519_1.htm, accessed
5/1/07)
‘[in the left dialog bubble] “This man must be crazy!” [in the right
dialog bubble] “Not really crazy, but rather…” [in the background,
hand-drawn] [ux jux jux ux]10’
The SSEs represent sounds that are continuously present throughout the
conversation between the two men.
In as much as simplex forms are associated with closed events, there is
a tendency for them to take the focus away from the internal properties of
the sound or motion (i.e. how something sounds or moves). Consequently,
they are more likely to develop non-sound symbolic meanings, in particu-
lar, to become verbs reporting what happened (rather than how something
happened), reporting cause-effect relationship between events. This is ex-
emplified by the SSE paf. Paf is a SSE that represents a gun shot when it
occurs in multiplex forms.
(11) Napodoboval zvuk střelby: Paf, paf, paf! (Syn2000)

‘He imitated the sound of shooting a gun: [paf paf paf]!’
In simplex forms paf departs from this sound-related meaning:
(12) … Taková akční bomba!!! Hrál jsem i tu hru a byl jsem z toho paf!!
(www.kfilmu.net/uzivatele.php?akce=vicenazoru&koho=3230,
accessed 9/7/06)‘
‘Such a cool action [movie] [lit. such an action bomb]!!! I played the
game, too, and was floored by it!!’
Paf in this example is used as a predicate that refers to an event of being

surprised. This non-sound symbolic meaning implicitly refers to a cause
(there must be a situation causing the surprise); clearly, a logical-sequential
relationship between events is built into the meaning.
We have seen how the contrast between simplex and multiplex variants
is consistently associated with aspectuality and discourse. The multiplex
and simplex variants interact with the locus of perspective, either on the
level of events (thereby overlapping in function with verbal aspect), and/or
on the level of texts. The discourse-aspectual functions of these forms are
consistent with the directions in which they are integrated into more con-
ventionalized parts of language: multiplex variants are integrated into im-
perfective verbs; simplex variants combine more easily with the semelfac-
tive suffix -nou-. Simplex variants are also more likely to develop into non-
sound symbolic predicates which implicitly refer to a cause-effect relation-

ship.
3. Discourse-aspectuality marker: lengthening
Within a SSE, a vowel or a consonant can be optionally lengthened or

shortened, e.g. a banging sound can be represented by either the long bác
[baɕts] or the short bac [bats], and a buzzing sound bz [bz] or bzz [bzɕ].
Added length may iconically reflect a prolonged sound. In example (13a)
[p=] represents a short abrupt sound, while example (13b) with lengthened
[=ɕ] represents a sustained sound:
(13) a. “Promiň! Promiň! Kolikrát ti to musím říct?”

‘“Sorry! Sorry! How many times do I have to tell you?”’
“Pš,” řekla Elma a přiložila prst ke rtům. (Syn2000)
‘“[p∫],” said Elma and put her finger to her lips.’
b. “Chrisi... já…” “Pššš...všechno bude fajn.” Objal mě.

(Syn2000)
‘“Chris, I…” “[p∫ɕɕ] everything will be OK.” He hugged me.’
However, the difference between (13a) and (13b) is not limited to sound
iconicity, but also concerns aspectuality. The short variant is used to tell
the addressee to stop talking immediately; the long variant is used to urge
the addressee to calm down without a sense of urgency. In other words, the
former is associated with an immediate and abrupt change of state, the
latter with a process. The main goal of the speaker in (13b) is to achieve a
state where the addressee is calm, but this change is seen as gradual transi-
tion from an upset state to calmness.
The contrast between long and short variants can be roughly repre-
sented by Figures 5 and 6.
S ~S
t
Figure 5. Functional approximation of (13a): A short variant
t
Figure 6. Functional approximation of (13b): A long variant
S above represents a state. Figure 5 marks the transition (marked as a

profiled circle) from one state S to a different state ~S. ~S is profiled in
contrast to S. Figure 6 does not implicitly contrast the situation S with any
other states. It focuses on the progressively changing nature (heterogeneity)
of S (in this case the state of being upset, less upset, and calm).
This contrast can be extended to the discourse level as in the examples
below:
(14) a. Vítr vehnal zase do místnosti spršky deště.

‘The wind drove showers of rain into the room again.’
Obloha byla tmavězelená. Padala sem voda.
‘The sky was dark-green. Water was falling here.’
Br, otřáslo se děvče. (Syn2000)
‘[br], the girl gave a shudder.’
b. Jéžiš, vono hřmí! Už aby tady byli, nesnáším to.

‘Jesus, it thunders! If only they were here, I can’t stand it.’
Brrrr, to je ale zima, vůbec si nemůžu zahřát ty nohy. (Syn2000)
‘[brɕɕ], it is really cold, I can’t warm up my feet.’
The short variant of the motion-symbolic expression br in (14a) repre-

sents an immediate reaction. The character suddenly feels an unpleasant
sensation. There is a change in her perceptive knowledge from her preced-

ing text (a text of the type “It isn’t/may not be too bad here”) to the current
text (a text of the type “It is unpleasant”). The SSE marks this point of
change. The long variant brrrr in (14b) is not associated with a contrast
between texts. The SSE summarizes the speaker’s emotional evaluation of
the entire situation: that it is cold, and that the character cannot warm up
her feet; the SSE, as it were, “says it all.” The linearity of text progression
in (14b) is downplayed; instead, the SSE functions as the general headline,
under which individual items are listed. The difference between these two
examples can be roughly represented as in Figures 7 and 8 below.
In Figure 7 the SSE represents the turning point between the speaker’s
previous text (T) and the speaker’s new text (~T). The newly discovered
knowledge ~T – the unpleasantness of the environment – is profiled. In
Figure 8 the SSE represents all the properties described in the text. The
SSE is therefore decomposable and has heterogeneous internal properties,
i.e. text segments;11 the SSE is similar to the headline supported by other
text segments. This relationship between the SSE and the text segments is
represented by arrows directed towards the SSE.
T ~T
text progression
Figure 7. Functional approximation of (14a): A short variant
SSE
T1, T2, ………… Tn
Figure 8. Functional approximation of (14b): A long variant
When examining the discourse-aspectual functions of added length, it is

necessary to consider the relative length, as shown in the examples below:
(15) a. “Né, né, né” volá maminka, “mé, mé, mé” piští děti. (Knoblo
chová 2000: 4)
‘“No, no, no,” the mother calls, [meɕɕ meɕɕ meɕɕ],” squeak the
kids.’
b. “Já jsem plesnivý kozel Černá Brada a všechny vás potrkám!

‘“I am the old he-goat the Black Beard and I will butt you all!!”
Mé, mé, meeé!” Hrozivě při tom kroutil očima (Knoblochová
2000: 7)
‘He rolled his eyes menacingly [meɕɕ meɕɕ meɕɕɕɕ]”’
ɕɕ
In (15a) the long vowel [meɕ] represents the neutral bleating sound of
little goats. In comparison to this SSE, the SSE [meɕɕ] in (15b) is a form
with added length. Thus (15a) should be considered the short variant and
(15b) the long variant for this particular SSE. The discourse functions of
these examples are consistent with the short-long contrast discussed above.
(15a) marks a text boundary: the text of the mother goat and the text of the
kids who reject the mother’s text. In (15b) the SSE sums up the major emo-
tion (a threat), which is elaborated by the other parts of the utterance.
When discussing the contrast between long and short vowels, it is worth
considering the two different directions in which the semantics of SSEs
prask and prásk seem to develop. According to the 1989 dictionary both
prask and prásk render a cracking, bursting or tearing sound; the long
prásk represents a more resonating and louder sound than prask. The ex-
amples elsewhere are more or less consistent with the 1989 dictionary
definitions, as illustrated in (16a) and (16b) below.
(16) a. “Tumáte!” “Jau!” PRASK (Disney 1991: 1/78)

‘”Take that!’ ‘Ouch!’ [prask]’ [an old lady hits the character
on the head with an umbrella.]
b. PRÁSK “Co máte s tím motorem?”

‘[praɕɕsk] “What do you have with the engine?”
“Vlastní vynález: alkohol se střelným prachem.” (Disney1991:
2/56)
‘“My own invention: alcohol with gun powder.”’
In (16a) an umbrella suddenly hits someone’s head, while in (16b) the

engine makes a loud explosive sound. Clearly, (16b) represents a louder
sound than (16a). However, the difference between the two is also dis-
course-related. Example (16a) contrasts two texts: the character rescues a

cat out of a tree, but instead of being rewarded for his act (the anticipated
text), he gets hit by the owner who thought he was tormenting the animal
(the modified text). In (16b) the contents of the bubbles explain the unusu-
ally loud noise produced by the engine; the SSE serves as the center-piece
of the entire communication.
The distribution of prask and prásk as SSEs is different, however; the
former is far less frequent than the latter. In fact, Syn2000 lists only the
homonymous past tense l-participle verb form as in the examples below.12
(17) a. ... prask mu vřed. (Syn2000)

‘His ulcer burst.’
b. Na nový pugét prask poslední zbytek mých...úspor, (Syn2000)

‘By [buying] the new bouquet of flowers the last remainder
of my…savings popped.’
c. ... jsem o pololetí prask z latiny (Syn2000)

‘... I flunked Latin in the middle of the school year.’
The examples above all refer to a change of state on the lexical level. In
(17a) the exterior of the ulcer breaks open, allowing its contents to flow
out. In (17b) the money bursts out of the place where it was kept. In (17c)
the speaker receives a failing grade, thereby being demoted out of a group
of students where he had belonged. None of these examples represents an
audible sound.13
In contrast, prásk occurs more frequently than prask as an indeclinable
SSE (9 for prask as opposed to 114 in Syn2000).14 L-participles, which are
homonymous with prásk, nonetheless do exist, but they have different
properties from those homonymous with prask:
(18) a. Milošek Straka prásk vlastního tátu, že poslouchá Hlas

Ameriky. (Syn2000)
‘Milošek Straka informed on [his] own dad that he was
listening to the Voice of America.’
b. .... pak prásk dveřmi a šel domů. (Syn 2000)

‘… then he slammed the door and went home.’
c. Námořník popadne pistoli a prásk ho – jak mouchu.(Syn2000)

‘The sailor grabbed the gun and shot him – like a fly.’
Admittedly, example (18a) is similar to (17a), (17b) and (17c). It can be

seen as a metaphorical extension of a sound representation; informing on a
person is to figuratively tear apart an individual (as in an explosion) – with
serious consequences – by exposing his secrets. However, the other mean-
ings in (18bc) retain association to sound. Example (18b) reports a loud
noise when slamming the door. Example (18c) involves a shooting.
In contrast to verbs that retain sound associations, prásk as an inde-
clinable SSE can occur in a less sound-related meaning as in the following
example:
(19) Nedovedou si představit, jakou jsem měl radost! Sedmnáct let jsem ji
neviděl!
‘They can’t imagine how glad I was! I hadn’t seen her in 17 years!’
Ale pak zas: Prásk! Maďarsko! A mamince ten pas nedali.
(Syn2000)
‘But then again: a big surprise! Hungary! And they didn’t give mom
the passport.’
(20) Měla jsem za ten dlouhý život chvíle, kdy jsem si připadala
‘During this long life I had moments when I felt’
jako ta nejšťastnější a prásk kozu do vozu,
‘like the happiest [person], and a big surprise [lit. [praɕɕsk] the goat
into the cart],
za tři neděle bylo všechno jinak. (Syn2000)
‘after three weeks everything was different.’
In the examples above, prásk does not represent a loud physical sound,
but reports a state: ‘All of a sudden there was a big surprise like an explo-
sion.’ Likewise, the second example prásk, which is accompanied by a
complement-like component in the accusative, compares an unhappy pe-
riod to the sound of slamming a goat into a cart. In each sample prásk
represents the existence of a set of interrelated states and situations that
constitute a crisis or a misfortune.
Even less sound-symbolic meaning of prásk can be found in a syntactic
function that resembles a nominal:
(21)“Není to špatné,” řekl uznale.

‘“Not bad,” said [the coach] appreciatively.’
“Ovšem techniku nemáš žádnou. Nevadí.
‘“Of course you don’t have any technique. No problem.’
Hlavně když máš prásk v ruce. A ten ti neschází.” (Syn2000)
‘The main thing is that you have power in [your] hand. And that you
have plenty.”’
The form can be considered a direct object of the verb mít ‘to have’.
Here, prásk is a group of properties that are more important than technique
for the athlete.15
Examples (17–21) show that the discourse-aspectual properties of prask
and prásk are at the base of how they develop less sound symbolic mean-
ings. The designation of transition is present both in the SSE prask and in
the verb in which prask is the root. The summary function is present in
both in the SSE prásk and in the SSEs that depart from sound representa-
tion. The discourse-aspectual functions of these SSEs parallel Langacker’s
two cognitive processes of summary and sequential scanning (1987: 247–
248).
Sequential scanning involves the processing of the component states in
a sequence. As this type of cognitive processing is said to be prototypical
of the meaning a verb, we can anticipate that prask is predominantly incor-
porated as part of a verb. On the other hand, summary scanning designates
all facets of the complex scene as simultaneously available and it is proto-
typical of the meaning of a noun (Langacker 1987: 199–200, 247). This
explains the noun-like behavior and the stative-predicate-like behavior in
(19–20) with prásk.
Sporadic occurrences of vowel length and shortening in the spoken Pra-
gue Czech is also consistent with the correlation between the two types of
cognitive processing and short-long vowel contrast. In spoken Prague
Czech, which is becoming the interdialect for Czech (Kučera 1961: 16–
20), variation in vowel length seems to be correlated with parts of speech.
According to the examples in Townsend (1990: 40–43), the most wide-
spread shortening is in verb forms e.g. in půjčit ‘lend’ > pučit and půjdu ‘I
will go’ > pudu. In contrast, vowel lengthening is much more frequent in
nominal stems and nominal-based adverbs than in verbs: dveře ‘door’ >
dvéře, nahoru/nahoře ‘up(stairs)’ (originally a preposition na + horu/hoře
> nahóru/nahóře.
Lengthening occurs also in familial words16 such as bratr ‘brother’ >
brácha, sestra ‘sister’ > ségra, tatínek ‘father’ > táta. The referential terms
such as ségra, brácha and táta suggest the speaker’s personal relationship
to the referents. For example, ségra is not merely anyone’s sister, but the
speaker’s sister or the sister of someone whom the speaker knows well.
The term can therefore be viewed as emphasizing the large amount of the
speaker’s knowledge of the referent.17
The discourse-aspectual functions of long and short variants of SSEs
are correlated with two types of cognitive processing. This correlation pro-
vides a potential clue to motivating the lengthening and shortening in more
conventionalized parts of the language.
4. Discourse-aspectuality marker: suffix-like formations
Discourse-aspectuality markers can resemble suffixation as well. The term

“suffix-like formations” is used because of the limited productivity of these
formations.
One of the most frequent suffixation-like processes involves addition of
the component –y between two identical SSEs. SSEs with –y compactly
represent multiple sounds without extensively repeating the same SSE. The
song below describes a rainy day when frogs are having a big party.
(22) U tří louží stará žába kváká,

‘At the three puddles an old frog is quacking’
déšť do blatouchů si dělá “kápy-káp”,
‘the rain [falling] into marsh-flowers goes “[kaɕɕpikaɕɕp]”,’18
kváká na žáby a na mejdan je láká.(Syn2000)
‘she quacks at the frogs and lures them to the party.’
A simple repetition káp, káp would render rain drops falling one by one
with significant intervals; it would also help the reader imagine the number
of drops, as though s/he were witnessing them (Figure 2). The form with -y
represents indefinite multiple drops in a more compact form; -y thus essen-
tially marks plurality. Furthermore, the rain drops in this context constitute
the necessary condition for the merry-making to take place; the frogs
would not quack so loudly and evoke the image of a big party without the
rain.
The following example represents the sound and the motion of letter-
writing played out in a film:
(23) [Lidé na filmovém plátně] ba mohou i dopis napsat, ovšem jen

‘[the people in the film screen] can even write a letter, of course only’
v rekordní rychlosti, škrty škrt, a už to musí být. (Syn2000)
‘at a record speed, [∫krti∫krt], and it must be done.’
The context points to the unusual speed at which letter-writing must be

accomplished in a film. The actors in a film must do it quickly solely for
the purpose of letting the audience see the completion of the letter. The
sound symbolic expression škrty škrt reflects not only the fast completion
of an action consisting of multiple movements, but presents the action as a
necessity for the purpose of making the film scene. The repeated sound
symbolic expression škrt škrt would reflect the physical sound of a pen or
any sharp pointy object scratching against a surface with a clear interval
between the strokes.
In short, SSEs with –y are less imitative than repeated variants. While
the form with -y implies a succession of sounds, each individual and sepa-
rable sound unit is not its major focus. Not only do SSEs with –y evoke
plurality, but they also represent the sound as an indispensable condition
for some situation (e.g. the rain drops for the party to take place; the fast
movement of the pen for creating the film scene). In other words, there is
an implicit contrast such as the one roughly represented in Figure 9. Be-
low, w represents a world in which situation x holds. ~w represents a world
in which x does not hold. The dotted line is the sound or motion repre-
sented by the SSE with –y. The dots represent plurality of the sounds.
The condition represented by SSE is indispensable for situation x to
hold. The world in which this relationship holds is w and is profiled. This
world is implicitly contrasted with a different world ~w in which x does
not hold because the condition represented by the SSE is absent.
w
x
~w
~x
Figure 9. Functional approximation of (23) with –y

SSEs may be accompanied by yet another suffix-like component –ity.

This form can imply the smallness of the source of the sound-motion as
well as the smallness of the sound-motion itself. The near-minimal pair
below illustrates the size difference of the source of sound and the sound
itself:
(24) a. A už bylo za tři minuty šest.... Když vtom

‘And it was already three minutes to six…When suddenly’
cupy dupy, dupy cupy, blíží se k nim jelínek ... (Syn2000)
‘[tsupi dupi dupi tsupi], a stag approaches them ...’
b. Běhá to okolo chalupy a dělá to cupity dupity. (Syn2000)
‘It runs around the cottage and does [tsupiti dupiti].’
Cupy dupy in (24a) represents sounds produced by the hooves of a

small deer. The sound source in (24b) is even smaller and the sound itself
softer because it refers to rain drops. The SSE with –ity functions like a
diminutive form of the suffix-like formation with –y. The form with –ity, in
addition, has a discourse function. Example (24b) is a children’s riddle; the
reader is expected to guess that “it” refers to rain. By using cupity dupity,
the speaker evokes a scene where some unknown tiny creature is running
around the cottage, but simultaneously anticipates that in the context of a
riddle the reader would interpret it further as a scene of rain drops falling.
The focus of the text is not on the sound itself, but on the type of object
being indirectly described.
Forms with -ity can also be used to produce irony as a diminutive form
does (Dressler and Barbaresi 1994):
(25) Může za to průvodce! To je porád: Nalevo vidíte,

‘It’s the tour-guide’s fault! It’s always: On the left you see,’
napravo vidíte, nalevo vidíte... houpity, houpity, žbluňk!
(www.techdesk.cz/diskuse/03/09/43.html, accessed 1/16/06)
‘on the right you see, on the left you see… [houpiti houpiti],
[ȡbluȁk]!’
The text is a reader’s comment on the news of an accident where tour-

ists drowned after their boat capsized. The SSE is not used to describe
small individuals bouncing in the boat nor the smallness of motion, but it is
used for an ironic effect. The purpose of the SSE is not so much to render
vividly the way the tourists bounced and were thrown into the water as to
report the carelessness that led to the tragedy. This use of -ity underscores
the inappropriate treatment of the situation as if it were minute and insig-

nificant, when it was actually very serious (it concerned the life and death
of the passengers).
The discourse function of –ity can be roughly represented in Figure 10
below. In this figure x represents an entity or a situation, w a world, and T a
text. The dotted line represents the sound or motion represented by the
SSE. In text T1 the SSE is an indispensable component of the entity or
situation x1 (cf. x1 does not hold in the world ~w1 in which the component
represented by the SSE is absent). This text evokes T2 (the arrow from T1
to T2). In T2 the SSE is an indispensable property of x2. Both w1 and w2 are
profiled.
w1 w2
x1 2 x2
T1 1 T
~w ~x1 ~w2 ~x2
Æ
Figure 10. Functional approximation of (24b) and (25) with –ity
More concretely, in (24b) the sound represented by the SSE is essential

for the scene where a small creature is running around the cottage (x1);
without the SSE this scene would not hold. This text (T1) is expected to
evoke another text T2 in which SSE is essential to the scene x2 in which
rain drops are falling around the cottage. In (25) the attitude represented by
the SSE is essential for situation x1 to hold (a light-hearted carefree attitude
as an essential part of a non-serious situation) in text T1. This text (T1)
evokes another text T2 in which the attitude represented by the SSE is an
essential part of another situation x2 (a light-hearted carefree attitude as the
cause of the tragic situation).
This section does not present an exhaustive view of suffix-like forma-
tions in SSEs, but it shows that these forms have consistent discourse-
aspectuality functions. Addition of semi-suffixes can report plurality, dimi-
nution, and can evoke a contrast between possible worlds and texts. As
seen in Figures 9 and 10, some of these suffix-like components involve
complex discourse operations that are far from reflecting a “naïve contact
with nature.”
5. Conclusions
Added repetition, length and suffix-like formations are not as chaotic as

they appear; they take part in accentuating specific aspects of events, in
setting the locus of perspective, in contrasting possible worlds and texts,
and in specifying the role of the SSE in a text. Contrary to de Saussure’s
view that SSEs are “not an organic part of the language,” uninflected ono-
matopoeic expressions and the more conventionalized part of the language
share properties. Systematicity in discourse-aspectuality markers points to
the possibility that SSEs may be seen as composite structures, which con-
sist of the basic form of a SSE and a discourse-aspectual marker. This
structure may not be exactly identical to the model in Langacker (2003: 67)
of the type [FLING+ER], but seems to parallel it. For example, the SSE
form bum bum bum can be roughly analyzed as [BUM + REPETITIONDA-
marker], prásk as [PRASK + LENGTHDA-marker], and dupydup as [DUP + YDA-
marker]. The basic SSE represents properties of sound and/or motion, while
the discourse-aspectual marker specifies its functions related to aspectual-
ity and discourse. As seen in (5), a SSE is a “demonstration.” They do not
merely imitate sounds or motions, but are selective in conveying their
properties; it is therefore possible to claim that the relationship between the
SSE basic form and the referent is symbolic. It is also conceivable that
some SSEs may not be broken down into a composite structure; this also
parallels Langacker’s model which allows varying degrees of analyzability
for grammatical forms. Construction of a more detailed model for SSEs is
a goal for future research.
Shared properties between SSEs and grammar can help explain some
processes of word derivation and variation in language. Specifically, dis-
course-aspectual functions of repetition can be used to explain why some
SSEs develop into imperfective verbs, while others combine with the se-
melfactive suffix -nou- to form perfective verbs. Discourse functions of
added length motivate the development of non-sound-related meanings in
nominal forms; I also suggested that these functions can provide at least a
partial explanation for vowel length variation in Common Czech.
Some of the suffix-like components in SSEs are associated with quanti-
fication and diminution, functions similar to those of grammatical mor-
phemes.19
Notes
* I want to express my gratitude to Laura Janda, Tore Nesset, the editors of this
volume as well as the two anonymous referees for their feedback. Obviously
all remaining errors and inconsistencies are entirely my own.
1. For examples of SSEs that represent motion, see e.g. Hamano 1998:2.
2. Sound symbolic expressions in Czech will not be translated, but will be repre-
sented in IPA. Since this study focuses not on grammatical categories, but on
discourse-aspectual functions, examples are not glossed, but are translated
side-by-side with the original.
3. In this article the term “aspectual” is an adjective connected with aspectuality.
Aspectuality is a semantic property, which overlaps but is not identical to ver-
bal aspect, which is a grammatical category.
4. Nuckolls (1996: 78) also notes similarities between aspectual functions of
SSEs on the one hand and cinematique techniques and gesture on the other.
This suggests that visual representation is highly relevant for the description of
SSEs, hence the use of diagrams in this study.
5. My approach belongs to a group of studies that use the general concept of
multiple domain structure in analyzing linguistic meaning (discussed in Sweet-
ser 1990: 147). The current study, however, differs from the existing literature
in dealing with those phenomena which have thus far been treated inconsis-
tently in lexicography and grammar and whose meanings are said to be situ-
ated in the phonological space rather than in the semantic space (Langacker
1987: 79–80). This article also finds that apparently random sound properties
in SSEs are connected to some of the very basic cognitive operations such as
placement of perspective and scanning processes.
6. The observations in this article are based on those onomatopoeic expressions
for which variation in length, repetition, and suffix-like formations are attested
in a database extracted from the 1989 8-volume dictionary. 406 indeclinable
expressions that are associated with sound and/or motion were identified. The
criteria used for selection were (1) indeclinability (often labelled as interjec-
tions) and (2) the definition of the lexicon that indicates connection to sound
and/or motion (in the presence of key words such as označuje/vyjadřuje zvuk,
pohyb ‘signifies/ expresses sound, motion’). More contexts for these variants
were sought in other sources (Syn2000, Syn2005, internet, literature, and com-
ics) and analyzed in detail in near-minimal pairs. In this article I use near-
minimal pairs rather than quantitative data for discussion for two reasons.
First, onomatopoeia is one area that is challenging to automatic tagging, which
is otherwise an indispensable tool for linguists. For example, bac in bac! bac!
napodobil ránu ... ‘[bac]! [bac]! he imitated a shooting sound ...’ is tagged as
a verb bacit in Syn2000, although it is more probably an indeclinable SSE. An
identical set of grammatical attributes is given not only to SSEs but to other
difficult-to-tag words such as abbreviations; e.g. the attributes for the ono-
matopoeic br in Br… moje hlava!! ‘[br]… my head!’ and the chemical sym-
bol Br in Význam některých prvků La, Ce, As, Br, Pb, Sb pro organismy ‘The
significance of some elements La, Ce, As, Br, Pb, Sb for organisms’
(Syn2005). Both samples are labeled as nouns. Second, quantifying the occur-
rences of SSEs is problematic. The National Corpus automatically counts each
instance of an SSE as one token even when it is repeated in the same sentence.
Thus in a single context where the same onomatopoeic expression occurs three
times in a string (e.g. bum, bum, bum), the Corpus enters them as three tokens.
7. Although partial repetition is also possible (e.g. pif paf), it is not included in
the discussion here.
8. This statement is made in relation to the multiplex variant.
9. Examples in Chung and Timberlake (1985) concern Russian verbal aspect, but
this difference between the imperfective and perfective aspect is sufficiently
fundamental to be applied to Czech verbal aspect as well.
10. Juch juch reports the speaker’s state of euphoria.
11. This summarizing function is actually present in (7b) as well. The SSE is the
most important speech act and the other utterance of the speaker (everything
will be fine) corroborates this speech act.
12. I use only the masculine 3rd person singular forms as examples to show that
the form occurs only as a homonymous l-participle verb form rather than as an
indeclinable SSE. The presence of this form, of course, assumes that there are
other l-participle forms of the verb prasknout such as prasknul, praskla,
prasknula, which contain the root prask.
13. All the 9 occurrences of prask in Syn2000 are masculine singular l-participle
forms, but are automatically tagged as “X” (unknown).
14. The actual number of tokens for prásk is 99 after eliminating those that count
each occurrence of prásk in a single sentence. All the tokens are automatically
tagged as interjections, although some of them show different properties as
shown in the examples here.
15. Prásk (and not Prask) is also the name of a TV talk show where the host
interviews celebrities for recent gossip; this use can be considered as prásk
representing a set of properties and situations (i.e.. a collection of up-to-date
gossip).
16. Implied solidarity in dimunitives is compatible with the notion of decomposa-
bility in that it imposes a set of interrelated properties shared between the
speaker and the referent (and/or the addressee) Here, too, lengthening is pos-
sible (oheň ‘fire’ > dim. ohýnek, květ ‘flower’ > dim. kvítek, sud ‘barrel’ >
dim. soudek). Conditions for lengthening, however, may be influenced by
other factors, which are beyond the scope of this article. I thank an anonymous
referee for this comment.
17. Length in conventional words has similar functions, as in the following exam-
ple: (a) Vždy když jsem vylezl, křičel tatíí, tatíí. Pozná mě i po hlase.
(Syn2005) ‘Whenever I crawled out, I shouted daddyyy, daddyy. He would
recognive me even by the voice.’ (b) “... prosím, prosím, prosím, prosííím...”
To poslední prosím se vzneslo vysoko. Proniklo stropem a usadilo se na větvi
magnolie. (Syn2005) ‘“... please please please pleeease…” This last please
soared high. It penetrated the ceiling and settled on a branch of the magnolia
tree.’ In (a), not only the word tati ‘daddy’ but also the way it is delivered fa-
cilitates easy identification of the speaker. Similarly, (b) reports a distinct way
in which the word prosím resonates and travels. In each instance length im-
plicitly refers to a set of distinct properties of way in which the word was de-
livered.
18. This example was taken from a song. The influence of rhyming on the choice
of the suffix formation in this case, however, seems minimal, as káp, only the
second part of the SSE, rhymes with čáp: ...déšť do blatouchů si dělá “kápy-
káp”,… /že jim ulít' čáp.
19. In fact, the forms –y and –ity contain phonological elements that resemble with
the most prevalent plural suffix –y in nominal declension, as pointed out by an
anonymous referee.
References
Chung, Sandra, and Alan Timberlake

1985 Tense, aspect, and mood. In Language Typology and Syntactic De-
scription. Vol. III: Grammatical Categories and the Lexicon, Timo-
thy Shopen (ed.), 202–58. Cambridge: Cambridge University Press.
Clark, Herbert H., and Richard J. Gerrig
1990 Quotations as demonstrations. Language 66 (4): 746–805.
Disney, Walt
1991 Walt Disneys’ Donald Duck. Bratislava: Egmont.
Dressler, Wolfgang U., and Merlini Lavinia Barbaresi
1994 Morphopragmatics: Diminutives and Intensifiers in Italian, German,
and Other Languages. Berlin/New York: Mouton de Gruyter.
Grice, Paul H.
1975 Logic and conversation. Syntax and semantics 3: 41–58.
Hamano, Shoko
1998 The Sound-symbolic System of Japanese. Tokyo: CSLI.
Havránek, Bohumil (ed.)
1989 Slovník spisovného jazyka českého [Dictionary of Czech Literary
Language]. Prague: Academia.
Hinton, Leanne, Johanna Nichols, and John J. Ohala

1994 Introduction: sound-symbolic processes. In Sound symbolism,
Leanne Hinton, Johanna Nichols, and John J. Ohala (eds.), 1–12.
Cambridge: Cambridge University Press.
Jakobson, Roman, and Linda R. Waugh
1987 The Sound Shape of Language. Berlin/New York/Amsterdam: Mou-
ton de Gruyter.
Jespersen, Otto
1933 Symbolic value of the vowel i. In Linguistica: Selected Papers in
English, French and German, 283–303. Copenhagen: Levin and
Munksgaard, 283–303.
Knoblochová, Hana Phillips
2000 Chytrá koza Líza [Smart Goat Líza]. Prague: Kvarta.
Kořínek, Jozef Miroslav
1934 Studie z oblasti onomatopoje [A Study of Onomatopoeia]. Prague:
Charles University Press.
Kučera, Henry
1961 The Phonology of Czech. The Hague: Mouton de Gruyter.
Kubátová, Marie
1994 Pohádky o Krakonošovi. [Fairy Tales about Krakonoš]. Prague:
Leon.
Langacker, Ronald, W.
1987 Foundations of Cognitive Grammar, Vol. 1: Theoretical Prerequi-
sites. Stanford, CA: Stanford University Press.
2003 Constructions in cognitive grammar. English Linguistics 20 (1): 41–
83.
Nuckolls, Janis B.
1996 Sounds Like Life: Sound-symbolic Grammar, Performance, and
Cognition in Pastaza Quechua. New York: Oxford University Press.
Rhodes, Richard
1994 Aural images. In Sound symbolism, Leanne Hinton, Johanna Nichols,
and John J. Ohala (eds.), 276–292. Cambridge: Cambridge Univer-
sity Press.
Saussure, Ferdinand, de
1959 Course in General Linguistics. Ed. by Charles Bally, Albert Seche-
haye, and Albert Reidlinger. Translated by Wade Baskin. New York:
Philosophical Library.
Schiffrin, Deborah
1987 Discourse Markers. New York/New Rocelle/Melbourne/Sydney:
Sweetser, Eve
1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects
of Semantic Structure. New York/Port Chester/Melbourne/Sydney:
Ústav Českého národního korpusu FF UK
2000, 2005
Český národní korpus [Czech National Corpus] – SYN2000,
SYN2005. Prague. Accessible at http://ucnk.ff.cuni.cz.
Townsend, Charles Edward
1990 A Description of Spoken Prague Czech. Columbus, OH: Slavica.
Werich, Jan
1960 Finfárum [Finfarum, fairy stories]. Prague: Československý spisova-
tel.
Subject index
absolute duration, 329, 339, 340, 156, 157, 158, 159, 160, 161,
344, 345, 347, 348, 353, 360, 162, 164, 165, 166, 167, 168,
361, 366 169, 171, 172, 173, 174, 175,
action 176, 177, 178, 184, 187, 188,
– completable action, 84, 105 189, 192, 194, 195, 196, 197,
– non-completable action, 84, 93, 201, 202, 211, 224, 236, 251,
105 252, 263, 281, 293, 330, 331,
– non-singularizable action, 93 332, 333, 349, 351, 352, 353,
– singularizable action, 93 355, 363, 367, 438, 439, 441,
active zone, 257, 269, 272, 273, 274, 453, 455
280 – canonical imperfective, 125,
actual/actuality, 120, 126, 130, 131, 140
132, 133, 134, 136, 137, 138, – non-canonical imperfective,
140, 142, 143, 146, 159, 161, 125, 127, 133, 134, 139,
163, 164, 175, 177, 178, 179 141
adjective/adjectival – perfective, 1, 6, 12, 13, 83, 84,
– adjectival predicate, 21, 23, 51 85, 86, 87, 88, 89, 90, 91, 92,
– attributive adjective, 311, 414 93, 94, 95, 97, 98, 102, 105,
aspect, 1, 5, 6, 10, 11, 13, 83, 84, 85, 106, 111, 112, 113, 114, 115,
86, 87, 88, 89, 90, 91, 92, 93, 116, 117, 118, 122, 125, 126,
105, 116, 117, 118, 119, 122, 127, 128, 130, 133, 134, 135,
123, 124, 143, 145, 150, 152, 136, 139, 141, 142, 145, 149,
184, 185, 187, 191, 192, 193, 150, 151, 152, 153, 156, 157,
197, 205, 208, 211, 234, 236, 158, 159, 160, 161, 162, 164,
244, 330, 331, 332, 351, 353, 165, 166, 167, 168, 169, 171,
357, 362, 363, 367, 435, 439, 172, 173, 174, 175, 176, 177,
441, 454, 455 178, 184, 185, 187, 188, 189,
– aspect markers, 12, 121 191, 194, 195, 196, 199, 201,
– aspect opposition, 351, 364 202, 203, 204, 205, 208, 209,
– aspect pair, 341, 349 211, 224, 227, 236, 293, 330,
– cluster model, 83, 92, 95, 96 331, 332, 348, 349, 439, 453,
– discourse-aspectual marker, 14, 455
431, 436, 453 – perfectivizing prefix, 14, 85,
– imperfective, 1, 6, 12, 13, 83, 106, 118, 329, 330, 350, 355,
84, 85, 86, 88, 89, 90, 91, 92, 361, 362, 363
93, 94, 95, 96, 97, 98, 99, 101, assertoric mitigation, 377
102, 104, 105, 106, 111, 112,
113, 114, 115, 116, 117, 118, binding, 233, 242
122, 123, 124, 125, 126, 127, – binding scale, 13, 221, 222
130, 131, 134, 136, 140, 145, – degree of binding, 244
146, 149, 150, 151, 152, 153,
460 Subject index
case, 1, 2, 5, 11, 12, 23, 40, 144, conjunction

252, 292 – conditional conjunction, 181,
– accusative, 6, 75, 297, 447 182, 183, 186, 189, 190, 191,
– case marking, 237, 299, 301 193, 207, 208, 211, 212
– case system, 397 – temporal conjunction, 183, 186,
– dative, 12, 55, 57, 58, 59, 60, 188, 189, 190, 191, 197, 199,
61, 62, 63, 64, 65, 66, 67, 68, 211
69, 70, 72, 73, 74, 75, 76, 144, construal, 6, 9, 10, 55, 60, 64, 74,
259, 296, 298, 299, 303, 389, 76, 84, 93, 94, 119, 124, 134,
391, 392, 396, 403, 404 136, 143, 151, 172, 174, 175,
– dativus sympatheticus, 55, 57, 176, 232, 240, 257, 265, 267,
58, 390, 391 272, 273, 274, 275, 276, 280,
– genitive, 57, 299, 315, 324, 330 332, 336, 345, 367, 411, 429
– instrumental, 12, 21, 22, 23, 26, construction
27, 28, 29, 33, 34, 36, 37, 38, – [VFIN VINF] construction, 221,
40, 41, 42, 43, 45, 46, 47, 48, 222, 228, 229, 237, 238, 239,
49, 50, 51, 285 241, 242, 252
– nominative, 6, 12, 21, 23, 25, – conditional construction, 181,
26, 27, 28, 30, 34, 37, 38, 41, 197, 198, 199, 200, 207, 209,
42, 43, 45, 46, 47, 48, 50, 51, 210
279, 296, 299 – constructional compositionality,
categoriality 181, 182, 186, 210
– categorial, 291, 292, 294, 302, – coordinate construction, 13,
312, 318, 323, 324, 346, 402 181, 201, 202, 205, 206, 209
cognitive grammar, 2, 12, 116, 118, – impersonal construction, 5, 13,
119, 143, 151, 152, 176, 225, 257, 258, 259, 260, 262, 263,
249, 257, 259, 420 265, 269, 272, 274, 279, 281,
collocation, 294, 306, 307, 308, 309, 283, 284
314, 385, 418, 427 – possessive construction. See
complementation, 221, 229, 231, possession
233, 238, 242, 244, 313, 315 – predictive construction, 183,
– complementation marker, 251 184, 187, 188, 190, 191, 193,
– complementation pattern, 296, 194, 195, 196, 199, 204, 207,
298, 316 208, 209
– that-complementation, 232, – setting-subject construction,
233, 236, 238, 239, 240, 241 274, 275, 277, 282, 283
conditionality, 13, 181, 182, 184, – temporal construction, 187,
185, 191, 198, 199, 204, 205, 197, 203
207, 209, 211 construction grammar, 12, 59, 181,
– conditional clause, 138 291, 294, 295, 310, 324
– conditionals, 182, 183, 184,
185, 190, 196, 197, 198, 200, epistemic, 9, 11, 12, 13, 14, 111,
201, 202, 207, 209, 210, 212 116, 119, 120, 121, 127, 129,
– truth-conditional, 10, 380, 395 137, 139, 140, 141, 142, 145,
Subject index 461
146, 149, 151, 152, 153, 154, force

155, 158, 159, 160, 161, 162, – epistemic force, 170, 171, 172,
163, 164, 165, 166, 167, 168, 379, 386, 403
169, 170, 171, 172, 173, 174, – force-dynamic, 178
175, 176, 178, 182, 183, 190, frame semantics, 12, 291, 292, 295,
196, 199, 212, 375, 376, 377, 324
381, 386, 387, 390, 391, 393, frequency, 11, 89, 99, 100, 102, 103,
394, 395, 397, 398, 399, 400, 106, 113, 124, 283, 296, 310,
401, 403, 404 346, 352, 397, 398, 427
– basic epistemic model, 119, infrequency, 31
120, 121
– elaborated epistemic model, grammaticalization, 11, 222, 241,
119, 145 244, 331, 348, 351, 362, 363,
– epistemic background, 138 397, 398, 402, 404, 436
– epistemic distance, 189 – grammaticalization theory, 12,
– epistemic immediacy, 13, 121, 56, 292, 321
138 grounding, 87, 116, 119, 121, 129,
– epistemic import, 151 162, 164, 168, 178, 378
– epistemic meaning, 13, 149,
151, 165, 166, 182, 375, 387, habitual, 21, 113, 132, 133, 136,
396, 401 141, 157, 160, 177, 306, 320, 364
– epistemic stance, 5, 114, 115,
139, 181, 183, 184, 185, 186, iconicity/iconic
191, 192, 197, 198, 199, 204, – principle of proximity, 417, 418
207, 208, 210, 211, 212, 375, – principle of quantity, 422
380 – principle of salience, 424, 426
– epistemic value, 13, 111, 114, – principle of sequentiality, 420
138, 142, 149, 151, 152, 160,
161, 165, 167, 172, 174, 175, metaphor, 1, 2, 3, 8, 11, 57, 84, 93,
176 94, 103, 105, 236, 336, 348, 362,
event 395, 402, 404
– argument structure, 221, 224, – metaphorical, 2, 84, 90, 92,
228, 229, 231, 233, 234, 235, 236, 293, 336, 337, 344, 348,
237, 239, 240, 241, 242, 243, 351, 353, 354, 360, 404, 447
252 metonymy, 2, 7, 362, 366, 395, 402
– complex event, 5, 11, 221, 233, – metonymic, 282, 315, 341, 375,
240, 242, 243 395, 401, 420
– event integration, 221, 222, – metonymical, 2, 136, 347, 387,
233, 235, 242 404
– multiple event, 239, 242 mood
– temporal event structure, 221, – conditional mood, 186, 187,
222, 223, 240, 242, 243 188, 189, 190, 191, 192, 193,
208
– indicative mood, 191
462 Subject index
motion subjectivity, 9, 111, 130, 134, 167,

– abstract motion, 421 378, 392
motivation – subjectification, 11, 14, 111,
– iconic motivation, 14, 412, 413, 116, 129, 144, 375, 377, 391,
419 394, 395, 396, 401
– metaphorical motivation, 84, 92 – subjective, 9, 51, 150, 151, 152,
167, 168, 173, 175, 176, 389,
participle 390, 391, 392, 393, 394, 396,
– participial adjective, 292, 293, 398, 401, 404, 419, 426, 428
294, 295, 296, 301, 302, 303, – subjectively construed, 129,
304, 305, 306, 307, 308, 309, 140, 151, 169, 394
310, 311, 312, 313, 314, 315,
316, 317, 318, 319, 320, 321, tense
322, 323, 325 – future tense, 31, 90, 137, 145,
– pseudo PA, 293, 296, 305, 306, 187, 194, 202, 259, 281, 367
310, 311, 312, 317, 319, 321, – historical present, 115, 142
322, 323, 325 – non-past tense, 13, 151, 152,
polysemy, 3, 8, 9, 11, 14, 291, 294, 153, 158, 162, 167, 172, 175,
300, 310, 322, 323, 330, 346, 176
360, 361, 363, 377, 380, 396, – past tense, 31, 90, 142, 151,
397, 404 156, 166, 168, 172, 183, 184,
– polysemous, 13, 56, 291, 294, 185, 187, 193, 196, 251, 281,
375, 378, 380, 381, 397, 401, 282, 446
402 – present tense, 13, 111, 112,
prototype semantics, 12, 14, 329, 113, 114, 116, 117, 118, 119,
346, 361 120, 121, 122, 124, 125, 126,
127, 128, 129, 130, 131, 132,
reality 134, 137, 138, 139, 140, 141,
– basic reality, 164, 168 142, 143, 144, 145, 146, 153,
– elaborated reality, 169 155, 159, 177, 183, 185, 187,
rhematic-thematic structure, 412, 193, 194, 195, 197, 203, 210,
417, 422 259, 281, 317, 392
semantic change, 14, 129, 291, 294, usage-based, 10, 11, 291, 402
346, 347, 350, 375, 377, 378,
380, 381, 394, 395, 396, 398, verb
401, 402, 404 – bi-aspectual, 12, 13, 83, 84, 85,
semantic shift, 144, 228, 291, 299, 87, 88, 89, 90, 91, 92, 94, 95,
302, 305, 316, 317, 319, 323, 96, 97, 98, 99, 100, 101, 102,
335, 376, 381, 383, 398 103, 104, 105, 106
sound symbolism, 11, 431, 432, 433 – delimitative verb, 329, 331,
– sound symbolic expression, 5, 332, 333, 337, 345, 348, 352,
14, 431, 432, 436, 450, 454 353, 356, 357, 363
Subject index 463
– delimitative prefixation, 14, – virtual event, 132, 134, 139,

94, 98, 329, 332, 333, 334, 142
335, 339, 340, 345, 350, – virtual plane, 132, 177
353, 359, 364, 365, 368 – virtual reality, 130
– resultative verb, 336, 348, 367 – virtual world, 131, 132, 134,
– resultative verbal prefixation, 136, 138, 140, 142, 143, 145
346 vowel length, 431, 448, 453
– verb integration, 13, 237, 238, – lenghtening, 431, 435, 442,
239 448, 449, 455
virtual/virtuality, 11, 111, 114, 120, – shortening, 448, 449
132, 133, 135, 136, 137, 138,
140, 141, 142, 143, 146, 152, word order, 5, 11, 117, 260, 279,
162, 164, 175, 176, 177, 178 282, 309, 388, 412, 413, 415,
417, 422, 428

Cognitive Paths Into The Slavic Domain (Eds. D.divjak&a.kochańska)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cognitive Paths Into The Slavic Domain (Eds. D.divjak&a.kochańska)

Uploaded by

Copyright:

Available Formats

Cognitive Paths into the

Library of Congress Cataloging-in-Publication Data

Cognitive paths into the Slavic domain / edited by Dagmar Divjak,

Bibliographic information published by the Deutsche Nationalbibliothek

Dagmar Divjak and Agata Kochańska

Part one. The nominal system: the meaning of case

Why double marking in the Macedonian dativus sympatheticus? ....... 55

Perfectives, imperfectives and the Croatian present tense .................. 111

Conflicting epistemic meanings of the Polish aspectual variants in

Conjunctions, verb forms, and epistemic stance in Polish and

Part three. The sentential system: non-archetypal event

The ‘impersonal’ impersonal construction in Polish. A Cognitive

Part four. Changing language: category shifting

A prototype account of the development of delimitative po- in

The rise of an epistemic pragmatic marker in Balkan Slavic:

Part five. Motivating language: iconicity in language

Discourse-aspectual markers in Czech sound symbolic expressions:

Subject index ………………………………………………………... 459

Dagmar Divjak, Laura A. Janda and Agata Kochańska

1. The cognitive paradigm and Slavic linguistic research

ues to constitute an important source of insight and inspiration for cogni-

Moscow Semantic School propound that language structures on all levels

linguistic perspective could be considered as merely recasting old ideas,

2. Slavic languages: an ideal laboratory for a cognitive linguist

Slavic languages have multiple attractions in store for a cognitive linguist,

of the counters) there are at least a dozen Slavic languages, spoken by

pected combinations such as imperfective aorists and perfective imperfects.

3. The theoretical framework of cognitive linguistics

3.1. The prototype approach to categorization

Categorization is not a matter to be taken lightly. There is nothing more

Categorization, in other words, matters to the linguist in at least two ways,

via another – typically more concrete – domain of experience – cf. e.g.,

3.2. A conceptual and imagistic approach to meaning

Meaning in the cognitivist framework is no longer defined in terms of out-

3.3. A usage-based approach to language

The third theoretical assumption we would like to highlight here concerns

4. Why cognitive linguists should read this volume

occurring language structures are “hints” helping language users construct

of a borrowed verb plays in determining whether it will be recognized as a

semantics to explain the development of the Russian prefix po- from a

1. This is, of course, an oversimplification of the situation. Where cognitive

guists therefore claim that a cognitive approach to language “should rely on

1988. A View of Linguistic Semantics. In Topics in Cognitive Linguistics,

This article reexamines the nominative/instrumental variation of adjectival

Keywords: Russian language, adjectival predicates, instrumental, limita-

The nominative/instrumental variation of adjectival predicates, including

(1) Včera on byl veselyj.

(2) V detstve on byl veselym.

(3) a. Togda, šest' let nazad ètot mal'čik byl

b. Kogda on byl ešče molodym i

c. Odnako, v tu poru, o kotoroj idet reč', ona

Recently, Gasparov (1996: 227), comparing the nominative long form

(4) a. Da, ded byl staryj i xlopot s

b. No ded byl starym, i azbuku Morze

This article will examine different types of narrative time, comparison,

2. Types of narrative time and comparison

(5) a. Znanie rodnoj istorii bylo očen' vysokim v

b. v te gody, kogda my ešče žili v Staroj Russe i

(6) Iz soroka čelovek, zakončivšix kogda-to 7 “B”, do sedyx

(7) a. Malina byla sadovaja, očen' krupnaja …

b. *Malina byla sadovoj, očen' krupnoj …

In (7b) the permanent nature of sadovyj ‘garden variety’, similar to kir-

(8) Dom byl kirpičnyj i, vidimo, očen' drevnij.