Igor Mel'čuk

Ten Studies in Dependency Syntax

Trends in Linguistics
Studies and Monographs

Editor
Chiara Gianollo
Daniël Van Olmen

Editorial Board
Walter Bisang
Tine Breban
Volker Gast
Hans Henrich Hock
Karen Lahousse
Natalia Levshina
Caterina Mauri
Heiko Narrog
Salvador Pons
Niina Ning Zhang
Amir Zeldes

Editor responsible for this volume

Chiara Gianollo

Volume 347
Igor Mel'čuk
Ten Studies in
Dependency Syntax
ISBN 978-3-11-069470-3
e-ISBN (PDF) 978-3-11-069476-5
e-ISBN (EPUB) 978-3-11-069481-9

Library of Congress Control Number: 2020941531

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data are available on the Internet at http://dnb.dnb.de.

© 2021 Walter de Gruyter GmbH, Berlin/Boston

Typesetting: raumfisch.de/sign, Berlin
Printing and binding: CPI books GmbH, Leck

www.degruyter.com
Acknowledgments
 This book would not exist if there were not several extraordinary people who
3
have helped me at different stages of my research.

The first to be named is, of course, Lidija Iordanskaja, my Reader in Residence;

she read and reread every piece in the volume many times, from the first sketch
to the last proofs.
The second to be named is David Beck, without whose advice, suggestions
and corrections—over many years!—I could never have reached the results that
I present here. He has been a severe critic and an efficient editor of all my texts.
Then come three colleagues and friends, with whom all my texts were dis-
cussed and who hunted down many mistakes and ironed out countless inconsis-
tencies: Jasmina Milićević, Alain Polguère and Elena Savvina.
And finally, Margarita Alonso Ramos, Jurij Apresjan, Igor Boguslavskij,
­Alexan­der Grosu, Leonid Iomdin, Sylvain Kahane, François Louis, Sébastien
Marengo, Simon Mille, Nikolaj Pertsov, Rafaël Poiret, Leo Wanner and Anton
Zimmerling went through different chapters of the volume, help­ing me with
their judicious remarks to significantly improve the presentation. Thanks to the
several constructive suggestions by the Anonymous Reviewer (of the Publisher)
I was able to better organize the volume.
Chiara Gianollo, Frank Junghanns, Barbara Karlson and Birgit Sievert—the
de Gruyter team that worked on this book—pulled off a miracle commensurable
with the proverbial transformation of water into wine: they turned my uncouth
manuscript into this beautiful volume.
Very special thanks are also due to the people who have offered me their con-
sultations and advice concerning various languages:

Acehnese: Mark Durie

Amele: John Roberts
Basque: Georges Rebuschi
Chinese (Mandarin): Haitao Liu, Jian-Yun Nie, Ruochen Niu, Rafaël Poiret
English: David Beck
Georgian: Zurab Baratashvili
Hindi: Omkar Koul
Korean: Chai-Song Hong, Mi-Hiyun Kim, Seong-Heon Lee,
­Geun-Seok Lim, John B. Whitman
Tagalog: Jean-Michel Fortis
Tongan: Yoko Otsuka

https://doi.org/10.1515/9783110694765-201
vi Acknowledgments

I did my best to take into account my colleagues’ proposals, and I thank them
from the depth of my heart. It goes without saying that I am alone responsible for
the final product.
Contents

Acknowledgements  v

Symbols, abbreviations and writing conventions  ix

Introduction  1

Part I: A Brief Overview of the Meaning-Text Model

1 Meaning-Text linguistic model  7

Part II: Surface-Syntactic Relations

2 A general inventory of surface-syntactic relations
in the world’s languages  31

3 Syntactic subject: syntactic relations, once again  117

4 “Multiple subjects” and “multiple direct objects” in Korean  179

5 Genitive adnominal dependents in Russian:

surface-syntactic relations in the N→Ngen phrase  205

Part III: Some Hard Nuts in Syntax Cracked by Dependency

Description
6 Relative clause: a typology  235

7 e sli …, to … ‘if …, then …’:

syntactic description of binary conjunctions in Russian  275

8 The East/Southeast Asian answer to the European passive  289

9 Pronominal idioms with a blasphemous noun in Russian

and syntactically similar expressions  309
viii Contents

Part IV: Word Order – Linearizing Dependency Structures

10 Word order in Russian  335

11 Linear ordering of genitive adnominal dependents

cosubordinated to a noun in Russian  369

References  387

Index of definitions  403

Index of notions and terms, supplied with a glossary  406

Index of languages  441

Index of semantic and lexical units  443

Symbols, abbreviations and writing conventions
All the relevant notions and formalisms cannot be explained here; the reader is
kindly invited to consult the Glossary (pp. 406ff ) and the monograph Mel’čuk
2012–2015.
All interlinear glosses in the examples are literal; two English words that cor-
respond to one foreign form are united by a dot: Rus. boli ‘of.pain’.

Symbols
|C condition part of a linguistic rule
L a particular language
L a particular lexical unit
«L» a particular fictitious lexeme (in the deep-syntactic structure)
L(‘X’) a particular lexical unit L expressing the meaning ‘X’
L(x1, x2, …, xn) (x1, x2, …, xn) is the syntactics of lexical unit L, x1, x2, …, xn being syntactic
features
˹L1 … Ln˺ a particular idiom L1 … Ln
L1–sem→L2 L2 directly depends on L1 semantically
L1–synt→L2 L2 directly depends on L1 syntactically
L1–⋯→L2 L2 indirectly depends on L1
L1 L2 L1 and L2 are co-referential (= L1 and L2 have the same referent)
NB important but tangential (= logically not necessary) information
r a particular surface-syntactic dependency relation
R Rheme (communicative value)
RDSynt deep-syntactic Rheme (communicative value)
RSem semantic Rheme (communicative value)
‘σ’ a particular semanteme
‘σ’ a particular semanteme that is communicatively dominant within
the semanteme configuration it belongs to
‘σ̃ ’ a particular configuration of semantemes
T Theme (communicative value)
TDSynt deep-syntactic Theme (communicative value)
TSem semantic Theme (communicative value)
«they» the expression «they» represents the indefinite-personal pronoun,
such as they (in the sentence In Yorkshire they say “eh” whenever
they don’t understand something), Fr. on, Ger. man
{xi} a set of elements xi
«x» x, a feature of syntactics of a linguistic sign

https://doi.org/10.1515/9783110694765-202
x Symbols, abbreviations and writing conventions

⟨x, y, …, z⟩ an ordered set of elements x, y, …, z

X a linguistic expression
*X an ungrammatical linguistic expression
?
X an incorrect or dubious linguistic expression
#
X  a pragmatically deficient or semantically anomalous linguistic
expression
(X) an optional X
[X] 1) government pattern
2) additional explanation
X ⟨Y⟩ Y, a variant of X
X|Y Y, conditions of use of X
X ⇔ Y  correspondence between linguistic entities X and Y of two adjacent
represent­ation levels ⟨= ‘X corresponds to Y and vice versa’⟩
X the context X of a linguistic rule
X+Y Y follows X immediately
X+…+Y Y follows X with a possible gap
⟦‘X’⟧ a presupposed semantic component ‘X’
|, ||, ||| pauses of increasing duration and importance
1, 2, 3 pronominal/verbal person 1, 2, 3
I, II, ..., VI DSynt-actants I, II, ..., VI
Ø zero sign (= linguistic sign whose signifier is empty)
⊕ operation of linguistic union
3 directly relevant important information
☛ explanations concerning conventions and notations

Abbreviations
-A actant
A ⟨= ADJ⟩ adjective (part of speech)
ACC accusative (grammeme of nominal/adjectival case)
ACT active (grammeme of verbal voice)
ADV adverb (part of speech)
AgCo agentive complement (a clause element)
AOR aorist (grammeme of verbal tense)
APPEND the APPENDITIVE deep-syntactic relation
ART article
ATTR the ATTRIBUTIVE deep-syntactic relation
CLAUS clausative (part of speech)
colloq. colloquial (stylistic label)
 Symbols, abbreviations and writing conventions xi

COORD the COORDINATIVE deep-syntactic relation

CONJ conjunction (part of speech)
CONV converb (grammeme of verbal finiteness: deverbal adverb)
D (syntactic) dependent
D- deep (sublevel of a linguistic representation)
DAT dative (grammeme of nominal/adjectival case)
DECL declarative (grammeme of verbal mood)
DEF definite (grammeme of nominal determination)
DET determiner (syntactic class of lexemes)
DirO direct object (a clause element)
DMorphR deep-morphological representation
DPhonR deep-phonic (= phonological) representation
DSyntA deep-syntactic actant
DSynt-AnaphS deep-syntactic anaphoric structure
DSynt-CommS deep-syntactic communicative structure
DSynt-ProsS deep-syntactic prosodic structure
DSyntR deep-syntactic representation
DSyntRel deep-syntactic relation
DSyntS deep-syntactic structure
ECD Explanatory Combinatorial Dictionary
fEm feminine (grammeme of adjectival/verbal gender)
fem feminine (gender; a value of the syntactic feature GENDER of a
noun)
FUT future (grammeme of verbal tense)
G (syntactic) governor
GER gerund (grammeme of verbal finiteness)
GP Government Pattern
HON honorific (grammeme of the category of politeness)
iff if and only if
impers impersonal (value of a syntactic feature of a pronoun)
IMPF imperfective (grammeme of verbal aspect)
IND indicative (grammeme of verbal mood)
INF infinitive (grammeme of verbal finiteness)
IndirO indirect object (a clause element)
INDEF indefinite (grammeme of nominal determination)
intrans intransitive (value of the syntactic feature TRANSITIVITY of a verb)
LDOCE Longman Dictionary of Contemporary English
LF lexical function
LU lexical unit
lit. literal
xii Symbols, abbreviations and writing conventions

masc masculine (grammeme of adjectival/verbal gender)

masc masculine (gender; value of the syntactic feature GENDER of a noun)
MTM Meaning-Text model
MTT Meaning-Text theory
MV Main Verb
N noun (part of speech)
NEU neuter (grammeme of adjectival/verbal gender)
neu neuter (gender; value of the syntactic feature GENDER of a noun)
neut. neutral (stylistic label)
NOM nominative (grammeme of nominal/adjectival case)
NUM cardinal numeral (part of speech)
OBL oblique (grammeme of nominal case [Eng. me, him, her, …])
OblO oblique (≈ prepositional) object (a clause element)
PART participle (grammeme of verbal finiteness)
PASS passive (grammeme of verbal voice)
PAST past (grammeme of verbal tense)
PERF perfective (grammeme of verbal aspect)
PL plural (grammeme of nominal/adjectival/verbal number)
PREP preposition (part of speech)
PRES present (grammeme of verbal tense)
pron pronominal (value of a syntactic feature)
-R representation (linguistic)
RefS referential structure
RhetS rhetorical structure
S- surface (sublevel of a linguistic representation)
-S structure
SAE Standard Average European (language)
Sem- semantic
SemA semantic actant
Sem-CommS semantic-communicative structure
SemR semantic representation
SemS semantic structure
SG singular (grammeme of nominal/adjectival/verbal number)
SMorphR surface-morphological representation
SSyntA surface-syntactic actant
SSyntR surface-syntactic representation
SSyntRel surface-syntactic relation
SSyntS surface-syntactic structure
SUBJ subjective (grammeme of nominal case)
Synt- syntactic
 Symbols, abbreviations and writing conventions xiii

SyntR syntactic representation

SyntSubj syntactic subject (a clause element)
trans transitive (value of the syntactic feature TRANSITIVITY of a verb)
V verb (part of speech)
VFIN finite verb
vulg. vulgar (stylistic label)

Writing conventions
Linguistic examples are in italics.

Textual glosses are in roman and between ‘semantic quotes’.

Lexical units are in small capitals.

Grammemes ⟨= inflectional values⟩ are in UPPER CASE: PAST, PL(URAL), etc.

Derivatemes are in helvetica italics caps: ‘one who [L-s]’ (read+er from readL,
teach+er from teachL).

The names of lexical functions are in Courier New: S0, Magn, Oper1, etc.

At their first mention (and sporadically where it is deemed useful), technical

terms are in Helvetica in the main text: antonymy, dependency, semanteme, etc.
Introduction Introduction

This volume, which you, my dear reader, are (I hope) about to start perusing,
presents a num­ber of case studies in dependency syntax carried out within the
Meaning-Text approach. The expression “dependency syntax” refers to a type of
linguistic description in which the syntactic structure of a sentence is described
in terms of syntactic dependencies—hierarchical binary rela­tions between lexical
units of the sentence. In point of fact, there is no other, *non-dependency, syntax.
The syntactic structure of a sentence cannot be described in any other way—
for instance, in terms of constituents, or phrases, as the overwhelming major-
ity of linguists have striven to do over the last half-century. This impossibility is
strictly logical: phrases in the sentence under production are themselves a way to
express its syntactic structure, and therefore, they cannot simultaneously be part
of what they express. Indeed, phrase structure in syntax will someday be seen
in the history of science in a similar light to Ptolemy’s epicycles in astronomy,
phlogiston and the luminiferous ether in physics, or the Scientific Socialism of
the Soviet era in social sciences. However, owing to near uncontested reign of
the phrase structure perspective in linguistics over the last several decades, it is
necessary to emphasize from the start the strictly dependential nature of syntax
as it is understood in this book, a point that I will return to again and again.
It is impossible to enter here into the intricacies of the appearance and devel-
opment of dependency representations in syntax. Suffice it to mention the trail-
blazing book Tesnière 1959 and the first dependency descriptions mentioned in
Mel’čuk 2014a; see also Mel’čuk 1988. For main references on dependency syntax,
see Chapter 2 below.
Right now, I need to present the overall organization of this volume.
The subsequent discussion will be conducted in terms of what is known as
a Meaning-Text model [MTM] of natural language. An MTM is a logical device
(= a system of rules) that is intended to represent the functional nature of a lan-
guage—namely, the transition from a chunk of meaning ‘σ’ to the text or several
synonymous texts Ti(‘σ’) that express ‘σ’; indeed, the formula ‘σ’⇔Ti(‘σ’) could be
taken to be the trademark of the whole Meaning-Text business: ‘σ’ is the meaning,
Ti(‘σ’) are the corresponding texts, and ⇔ is the Language. A brief sketch of the
linguistic Meaning-Text model is offered in Chapter 1.
As far as syntax is concerned, the Meaning-Text approach presupposes three
levels of syn­tactic description for a sentence: 1) the deep-syntactic representa-
tion, 2) surface-syntactic repre­sentation, and 3) the deep-morphological repre-
sentation (Chapter 1, Subsection 1.2.2, pp. 11ff ):

{DSyntRs}⇐deep syntax⇒{SSyntRs}⇐surface syntax⇒{DMorphRs}

https://doi.org/10.1515/9783110694765-001
2 Introduction

This volume considers only the transition between the surface-syntactic repre-
sentation of the sentence to be produced and its deep-morphological representa-
tion. The linguistic rules that ensure said transition constitute the surface-syntac-
tic module of the Meaning-Text model; thus, I will speak exclusively of surface
syntax.
The full characterization of surface syntax requires three elements:

–– the specification of the surface-syntactic representation [SSyntR];

–– the specification of the deep-morphological representation [DMorphR];
–– the specification of the MTM surface-syntactic module, which maps SSyntRs
onto DMorphRs.

A complete presentation of general surface syntax would need several volumes;

as a result, my goals have to be limited to just two aspects of syntactic descrip-
tion: surface-syntactic rela­tions and word order rules, which are part of the
MTM’s surface-syntactic [SSynt-]module. In conformity with this limitation, the
book is divided in four thematic parts.

Part I – A brief overview of the Meaning-Text model

This part contains only one chapter—Chapter 1, which exposes the theoretical
framework for what follows.

Part II – Surface-syntactic relations

Surface-syntactic relations [SSyntRels] form the sur­ face-syntactic structure
[SSyntS] of a sentence, and the SSyntS is the central component of an SSyntR.
Strange as it may seem, as of today, no language has a full universally agreed-
upon in­ventory of SSyntRels. Therefore, I feel justified in dedicating Chapters 2
through 5 to the various problems of creating such an inventory. More specifi-
cally, these chapters present:

Chapter 2 – a tentative set of SSyntRels found in various languages and grouped

according to their syntactic properties: actantial, modifying, attributive, auxil-
iary, etc. This chapter consti­tutes the foundation on which rest all other subse-
quent deliberations.

Chapter 3 – linguistically universal formal definitions of syntactic subject and

direct object, the subjectival and direct-objectival SSyntRels being the basic
SSyntRels.

Chapter 4 – a discussion of so-called “multiple subjects and objects” in Korean.

 Introduction 3

Chapter 5 – an analysis of the syntactic organization of genitive adnominal

dependents in Rus­sian; in other words, this chapter proposes several SSyntRels
for the description of such dependents.

Part III – Some Hard Nuts in Syntax Cracked by Dependency Description

Here four very different syntactic phenomena are analyzed which constitute a
challenge for a coherent lin­guistic description; this is done in four chapters:

Chapter 6 is dedicated to a typology of relative clauses.

Chapter 7 deals with so-called binary conjunctions (e.g., if …, then…); the surface-
syntactic description of sentences containing these conjunctions is proposed.

Chapter 8 shows that what is sometimes considered а passive construction in

Mandarin Chinese is in fact a verb meaning ‘[to] undergo’ that introduces a full-
fledged clause; there is no passive in Chinese. As a basis for the definition of
passive construction, some requirements on a good linguistic definition are for-
mulated.

Chapter 9 offers а syntactic description of Russian phrasemes that are indefinite

pronouns (e.g., čërt ego znaet ‘devil it/him knows’). This chapter also includes
a universal typology of phrasemes.

Part IV – Word order: linearizing dependency structures

Word order is studied in lin­guistics much better that surface-syntactic relations,
although not enough has been done from the angle of dependency structure;
Chapters 10 and 11 deal with formal word order rules exactly in a strictly depen-
dency-oriented perspective. More specifically:

Chapter 10 presents a sketch of a general method for linearizing a dependency

syntactic structure, introducing the input and output formalisms for lineariza-
tion rules and detailing major types of these rules. The proposed methodology is
developed on the basis of Russian, a language famous for its extremely flexible,
but not arbitrary, word order.

Chapter 11 analyzes a particular case of SSyntS linearization—the arrangements

of the genitive nouns cosubordinated “in parallel” to a noun in Russian; this
arrangement is compared with the ordering of Russian adjectives also cosubor-
dinated to the same noun.

Given the character of the present volume, which includes abundance of formal-
isms and technical terms, I have also provided a Glossary, where all the terms
4 Introduction

used in the text are defined and (succinctly) explained; see pp. 406ff. On top of
this, the volume includes:

–– An index of definitions proposed therein, so that the reader can easily

compare them.
–– An index of languages mentioned and discussed.
–– An index of semantemes and lexical units treated in this volume one way or
another.

And so with that,

WISHIND, PRES

subject dir-object

indir-object

Journeysg
I Yousg

determinative obl-object

modificative
e xclamative

THROUGH
A Pleasant
prepositional

volumesg

determinative

this
Part I: A Brief Overview of the Meaning-Text Model
1 Meaning-Text linguistic model
1.1 Functional models in sciences
1.2 Meaning-Text theory and functional models of natural language: Meaning-Text models
1.2.1 The Meaning-Text theory’s three postulates
1.2.2 Linguistic representations in a Meaning-Text model
1.2.3 Linguistic rules in a Meaning-Text model
1.2.3.1 Semantic Meaning-Text rules
1.2.3.2 Deep-syntactic Meaning-Text rules
1.2.3.3 Surface-syntactic Meaning-Text rules
1.2.3.4 Morphological Meaning-Text rules
1.2.4 Modeling two important linguistic phenomena: paradigmatic and syntagmatic
lexical choices
1.2.4.1 Paradigmatic lexical choices
1.2.4.2 Syntagmatic lexical choices
1.2.4.3 Correlations between the meaning and the collocates of a lexical unit
1.3 The value of functional models in linguistics

The sciences do not try to explain, they hardly even try to interpret, they mainly make
models. By a model is meant a mathematical con­struct which, with the addition of certain
verbal interpretations, des­cribes observed phenomena. The justification of such a math-
ematical construct is solely and precisely that it is expected to work—that is, correctly to
describe phenomena from a reasonably wide area. Furthermore, it must satisfy certain
esthetic criteria—that is, in rela­tion to how much it describes, it must be rather simple.
John von Neumann, in L. Leary (ed.), The Unity of Knowledge (1955: 158)

All discussions of syntactic phenomena in this volume are conducted strictly

within the formal framework of the Meaning-Text approach. In order to help
the reader navigate the technical complexities of the following proposals, a few
words have to be said about the Meaning-Text approach in general and about the
Meaning-Text linguistic model in particular: this approach and this model essen-
tially underly all the syntactic studies presented below.

1.1 Functional models in sciences

The Meaning-Text approach is based on the idea of functional modeling, so it is
natural to start with the notion of functional model.
How do we know what happened a few seconds after the Big Bang? Nobody
was there, and no observation device could exist at the moment of the creation of
the Universe. However, we know a lot, and this is made possible by abstract cos-
mological models—systems of equations, which, based on known physical laws

https://doi.org/10.1515/9783110694765-002
8 1 Meaning-Text linguistic model

and general logic, relate to one another various facts observable ≈ 14 billion years
after the Big Bang. From these models, scientists can reach conclusions about the
state of the Universe at different stages of its existence.
This is but one example among many that could be cited. In countless cases,
a researcher who is in no position to directly observe the internal structure of an
object or a phenomenon has recourse to a model. Exaggerating a bit, any hard
science is mainly the construction of models. This has been well known at least
since Galileo. “There is no scientist who does not reason in terms of models—even
if he does not admit this to others or to himself” (Auger 1965: 4).
Linguistics, which has natural language as its object, is in the same position
with respect to language as cosmology with respect to the Big Bang. Language, an
extremely complex system of rules, is encoded in the brains of its speakers and
thus it is inaccessible to direct observation: linguists cannot open the skulls or
penetrate the brain with electrodes at their will. The only solution is the recourse
to models. And this trend—constructing formal models of (fragments of) lan-
guages—has already launched.
On the one hand, N. Chomsky’s Generative-Transformational Grammar, since
1950s till the beginning of the 21st century, has solidly implanted the idea of mod-
eling in linguistics. Thus, as early as in Nagel et al. (eds.) 1962 we find several
articles that discuss the topic of linguistic models.
On the other hand, intensive work in different branches of computational
linguistics has heavily contributed to this trend. Today we can take it for granted
that modeling is de facto fully accepted in linguistics. However, it remains to be
established what types of linguistic models are the most promising and to make
the notion of model more specific—so that it can be accepted de jure. In fact, the
term model itself is ambiguous; in order to eliminate confusion, a rigorous defini-
tion must be proposed.
Let there be an entity E (an object or a system of objects); E functions in the
sense that it receives observable inputs and produces for them corresponding
observable outputs. The researcher is interested in the functioning of E rather
than in its internal structure (which is in any case not observable). To describe E,
he constructs a functional model M of E, that is, M(E).

Definition 1.1 – functional model

A system of symbolic expressions M(E) created by the researcher to describe
the function­ing of E is a functional model of E if and only if it associates with
the given inputs the same outputs as E does.
 1.2 Meaning-Text theory and functional models of natural language 9

The model M is functional in two senses: 1) it seeks to represent the function-

ing of E, and 2) it does so by using mathematical functions, so that M is actually
a very complex function in the mathematical sense, mapping inputs to outputs.
NB Our term functional [model], having the indicated meaning, should not be confused with
functional in functional theories of grammar, which take the functionality of language and its
elements to be the key to understanding linguistic processes and structures.

1.2 Meaning-Text theory and functional models of natural

language: Meaning-Text models

Meaning-Text theory [MTT] is characterized here in three steps: the postulates

that underlie it (1.2.1), the linguistic representations it uses (1.2.2), and some basic
types of MTT rules (1.2.3).

1.2.1 The Meaning-Text theory’s three postulates

More than 60 years ago, work started on the development of a functional model
of Natural Language, the Meaning-Text model [MTM]. The project was begun in
Moscow in the 1960s by the present author, together with several colleagues,
principally—A. Zholkovsky and Ju. Apresjan (see, e.g., Žolkovskij & Mel’čuk 1967,
Mel’čuk 1974, 1988, 2012–2015, 2016). The linguistic theory underlying MTMs is
known as Meaning-Text theory; it is based on the following three postulates.

Postulate 1 – Answer to the question “What is natural language?”

A natural language is a system of rules that are stored in the brains of spea-
kers and describe the correspondence between a denumerable (= infinite,
but countable) set of linguistic meanings and a denumerable set of lingu-
istic texts.

Linguistic meanings (in the technical sense of the term) appear as formal symbolic
objects called semantic representations [SemRs], and texts—as phonetic representa-
tions [PhonRs]. Postul­ate 1 can then be expressed in symbolic form as (1):

(1) {SemRi} ⇐language⇒ {PhonRj} | i ≠ j, 0 < i, j ≤ ∞

Logically, the Meaning-Text correspondence (i.e., “⇐language⇒”) is bidirec-

tional and represents equivalence; yet in linguistics it should be studied and
described in the Meaning-to-Text direction: natural language is mainly about
speaking, not understanding. Linguistic synthes­is, or text production, is much
10 1 Meaning-Text linguistic model

more important for linguistics than analysis, or text understanding. This is so

because text understanding inevitably involves, to a greater or lesser extent, the
under­standing of the subject matter, i.e., the understanding of the extralinguistic
reality, which muddles the picture. Going from a given meaning to all the texts
that can express it liberates the researcher from huge “non-linguistic” difficulties
leaving him face-to-face with the language in its purest state. To this it can be
added that any text, before it can be analyzed, has to be synthesized: text synthe-
sis is obviously primary to analysis.
The meaning-to-text orientation of linguistic research and description gives
absolute priority to the study of synonymy, in particular—of linguistic paraphrase
(Milićević 2007).

Postulate 2 – Answer to the question “What is a description of a language?”

The Meaning-Text correspondence in (1) is described by a logical device, or
system of formal rules, which constitutes a functional model of language:
an MTM.

An MTM takes meanings, or SemRs, as its inputs, and produces texts, or PhonRs,
as its out­puts—in the same way that native speakers do. It is in this sense that an
MTM is a mathematical function: f(SemR) = {PhonRj}. Applied to a SemR, it pro-
duces the set of all (nearly) synonymous PhonRs that correspond to it. (An MTM
can also be used in the inverse direction: taking texts as inputs and extracting
meanings from them. In this chapter, however, only the Meaning-to-Text direc-
tion is considered.)
The Meaning-Text correspondence is many-to-many: one SemR can corre-
spond to an astro­nomical number of PhonRs (several million; there is incredibly
rich synonymy: see, e.g., Mel’čuk 2012–2015: vol. 1, 65–67, 155), and one PhonR
can express many SemRs (ambiguity). Because of this, Postulate 3 is needed.

Postulate 3 – Answer to the question “How should an MTM be structured?”

To successfully describe the complicated Meaning-Text correspondence,
two intermediate levels of linguistic representation are needed in an MTM:
Synt(actic)R, corresponding to sentences, and Morph(ological)R, corres-
ponding to wordforms.

As a result, an MTM has the following general architecture:

(2) {SemRi} ⇔ {SyntRk} ⇔ {MorphRl} ⇔ {PhonRi}

semantics syntax morphology +
phonology
 1.2 Meaning-Text theory and functional models of natural language 11

The boldfaced words are the names of MTM’s major components, or modules.
Linguistic representations of all levels, except for the semantic level, are each
subdivided into deep, or meaning-geared, and surface, or form-geared, sublevels.
Including the final—phonic—level, this gives us a total of seven representations.
NB To avoid unnecessary complications, in this book only graphic representations of actual sen-
tences will be used.

1.2.2 Linguistic representations in a Meaning-Text model

To make clearer the basic ideas underlying an MTM, examples of the basic, or
main, struc­tures of the linguistic representations of all levels will be supplied
(and in the next subsection, a few rules relating them). Due to lack of space, many
approximate descriptions will be used and many explanations foregone.

Semantic structure
The semantic structure [SemS] is one of the four components of a SemR—its basic
structure. Formally, it is a network whose nodes are labeled with semantemes
(meanings of disambiguated lexical units [LUs] of the language in question) and
arcs are labeled with numbers used to distin­guish the arguments of a predicate.
The other three components of a SemR—the peripheral structures, namely,
the semantic-com­municative structure [Sem-CommS], the rhetorical structure
[RhetS] and the referential structure [RefS]—are not shown.

The SemS in Figure 1.1 (next page) can be verbalized by a huge number of sen-
tences, of which only three are shown in (3):

(3) Three of the sentences that can be obtained from the SemS in Figure 1.1
a. 
Abu-Khalaf has been permitted by Damascus to step up the flow of terrorists­
into Iraq to 30 a month.
b. 
The government of Syria let Abu-Khalaf increase the number of terrorists
slipping into Iraq up to 30 per month.
c. 
Abu-Khalaf has the permission of the Syrian government to raise the number
of terrorists going to Iraq to 30 each month.

For the reader’s conevenience, all illustrative representations are given for sen-
tence (3a).
12 1 Meaning-Text linguistic model

‘before’
‘permit’ 1 2 ‘now’
1 2 ‘cause2’ ‘increase1’
‘government’
2 1 2
1

‘Syria’ ‘Abu-Khalaf’ ‘set and’

‘become’
‘rate’ 1
2 2
1 ‘during’ ‘each’
‘go’
‘Iraq’ 2 1 1 2 1
‘terrorists’ ‘30’ ‘month’

☛ 1. The underscoring of a semanteme (here, ‘permit’) indicates its communicative-dominant­

character (Mel’čuk 2001: 29ff ). In other words, the whole Sem-network above is about
‘permitting’.
2. The semanteme ‘setand’ represents a logical conjunction of elements, in our case—of the
semantemes ‘increase1’ and ‘become’. Since all elements of a conjunction are semanti-
cally “equal,” the branches leading from ‘setand’ to its Sem-dependents are not labeled.
The semanteme ‘increase1’ represents the meaning of the intransitive verb increase1:
‘become more intense’.

Figure 1.1 A semantic structure

Deep-syntactic structure
The deep-syntactic structure [DSyntS] of a sentence is an unordered labeled depen-
dency tree (the physical disposition of its nodes on the page has no logical rel-
evance).

–– Its nodes represent only the (semantically) full lexical units [LUs] that compose
the sentence, including ordinary (i.e. non-auxiliary) lexemes, idioms, ficti-
tious lexemes, lexical functions (see below, 1.3.3), and such complex LUs
as compound numerals (such as THREE MILLION TWO HUNDRED THOUSAND
THIRTY-FOUR= 3 200 034) and compound proper names (such as ROBERT
MALCOLM WARD DIXON). Each node of DSyntS is labeled with the name of one
of these LUs. Structural (auxiliary, or grammatical) LUs are not present in a
DSyntS. In a language that has inflectional morphology, each LU in a DSyntS
is supplied with appropriate deep, or semantic, grammemes (= grammemes
which have their source in the semantic representation, i.e. which carry
meaning; such as, for instance, the number and definiteness of nouns and
the voice, mood, aspect and tense of verbs). Surface grammemes, imposed
by government and agreement (such as the case for nouns, the person and
number for verbs, the gender, number and case for adjectives), are not shown.
–– Its branches represent the DSynt-relations [DSyntRels] that link the LUs in
the sentence and are labeled with the names of 13 universal DSyntRels; see
Chapter 2, 2.2.1, pp. 32ff.
 1.2 Meaning-Text theory and functional models of natural language 13

The DSyntS of sentence (3a) is presented in Figure 1.2.

PERMIT PASS, IND, PERF , PRES

I II III
CausPredPlus

ABU-KALAFSG DAMASCUS SG II ATTR

FLOWDEF, SG TO
II
I II

THIRTY
TERRORIST NON -DEF , PL I RAQ SG ATTR

«PER »
II

MONTH INDEF, SG

Figure 1.2 The deep-syntactic structure of sentence (3a)

The other three components of a DSyntR—the peripheral structures, namely, the

deep-syntactic-communicative structure [DSynt-CommS], the deep-syntactic-
anaphorical structure [DSynt-AnaphS] and the deep-syntactic-prosodic structure
[DSynt-ProsS]—are not shown.

Surface-syntactic structure
The surface-syntactic structure [SSyntS] of a sentence is also—just like the DSyntS—
an unordered labeled dependency tree; however, its composition and, conse-
quently, labeling are different from that of a DSyntS.

– Its nodes represent all actual lexemes of the sentence, including all pro-
nouns and structural (= grammatical) words, and are labeled with their
names. Each lexeme, just as in the DSynt-structure, is supplied with appro-
priate deep grammemes. (All the surface grammemes are introduced in the
DMorph-structure at the following step—by SSynt-rules.)
– Its branches represent the SSyntRels that link the lexemes and are labeled
with the names of language-specific SSynt-relations. Their number seems to
be about 60–70 per language. (See Chapter 2, Section 2.5 for a general inven-
tory of SSyntRels in the languages of the world.)
14 1 Meaning-Text linguistic model

HAVEIND , PRES

subject perf-analytical

BEPPART
ABU-KHALAFSG passive-analytical

PERMITPPART
agentive dir-objectival

BY TO
prepositional prepositional

DAMASCUSSG STEP(V)

verb-mark-analytical circumstantial
dir-obj
UP FLOWSG TO

subj-adnom
deter obl-obj prepositional

OF INTO THIRTY
THE
prepositional prepositional «per»-appositive

MONTHSG
TERRORISTPL IRAQ SG
determinative

A
Figure 1.3 The surface-syntactic structure of sentence (3a)

Deep-morphological structure
The deep-morphological structure [DMorphS] of a sentence is a string of all its
lexemes supplied with all the values of their inflectional categories, i.e., with all
their grammemes.

Abu-KhalafSG | haveIND, PRES, 3, SG bePPART permitPPART by DamascusSG ||

to stepINF up |
the flowSG of terroristPL into IraqSG |
to thirty a monthSG |||

☛ The symbols “|,” “||” and “|||” stand for pauses of different duration; these pauses are ele-
ments of the deep-prosodic structure of the sentence.

Figure 1.4 The deep-morphological structure of sentence (3a)

Surface-morphological structure
The surface-morphological structure [SMorphS] of a sentence is a string of mor-
phemes that are fed to the SMorph-module of an MTM, to produce a string of
morphs supplied with necessary prosodies.
 1.2 Meaning-Text theory and functional models of natural language 15

{Abu-Khalaf}+{sg} |
{have}+{ind.pres}+{3.sg} {be}+{ppart} {permit}+{ppart} {by} {Damascus}+{sg} ||
{to} {step}+{inf} {up} |
{the} {flow}+{sg} {of} {terrorist}+{pl} {into} {Iraq}+{sg} |
{to} {thirty} {a} {month}+{sg} |||

Figure 1.5 The surface-morphological structure of sentence (3a)

The phonemic and the phonetic structures of sentence (3a) are not shown, since
they are irrelevant to the topic of the present monograph.

1.2.3 Linguistic rules in a Meaning-Text model

Linguistic representations of adjacent levels are linked by corresponding lin-

guistic rules, which map a linguistic representation of the level n onto the repre-
sentation of the level n+1 (and vice versa). To simplify the exposition, only rules
dealing with the basic structures of representa­tions are considered.

Three major classes of linguistic rules will be illustrated:

– semantic: {SemSs} ⇔ {DSyntSs}
– syntactic: {DSyntSs} ⇔ {SSyntSs} ⇔ {DMorphSs}
– morphological: {DMorphSs} ⇔ {SMorphSs} ⇔ {DPhonSs}
The rules presented below are used in transitions between the structures in
Figures 1.1 through 1.5.

1.2.3.1 S
 emantic Meaning-Text rules
The semantic Meaning-Text rules come in three major types:
– Lexicalization rules map configurations of semantemes (in the starting
semantic structure) on deep lexical units, which label the nodes of the deep-
syntactic structure of the sentence to be synthesized.
– Morphologization rules map configurations of semantemes on deep gram-
memes, which are subscripted to the deep lexical units.
– Arborization rules map semantic relations between semantemes on the deep-
syntactic relations.
It is impossible to give examples of every type of Sem-rule here; I will limit myself
to just three Lexicalization rules.
☛ 1. The expression of the form “L(‘σ’)” stands for lexical unit L that expresses the meaning ‘σ’.
2. Shading marks the context of the rule—that is, elements that are not impacted by the
rule, but whose presence in the input structure is necessary for the rule to apply.
16 1 Meaning-Text linguistic model

Individual lexemic Sem-Rule

‘rate’ FLOW (N)

1 I II
‘go 1 ’
2 1
‘Y’ ‘X’ L(‘X’) L(‘Y’)

Figure 1.6 The Sem-rule for the lexeme flow(N) (‘go1’ ≈ ‘move towards’)

The meaning ‘rate of Xs that go1 to Y’ can be expressed by the lexeme flow(N) [of
Xs (in)to Y] (daily flow of visitors to the museum). This is, roughly speaking, a part
of the lexicographic entry for the noun flow(N) ‘movement of X’s …’—namely, its
definition.

Individual lexical-functional Sem-Rule

‘cause 2’ CausPredPlus

2
‘become’ II
1 2

‘Y’ ‘more’ L(‘Y’)

☛ The semanteme ‘cause2’ represents voluntary, teleological causation, as in John killed the
wolf [‘X is the causer of Y’]; ‘cause1’ stands for non-voluntary, spontaneous causation, as in
The bullet killed the wolf [‘X is the cause of Y’].

Figure 1.7 The Sem-rule for the lexical function CausPredPlus (see 1.2.4.2, p. 22)

The meaning ‘cause2 Y to become more [than Y was]’ can give rise to the lexical
function [LF] CausPredPlus, whose value is specified for its second argument,
i.e., its DSynt-actant II, in the dictionary: CausPredPlus(flow(N)) = step up the
flow; CausPredPlus(gap(N)) = widen the gap; CausPredPlus(ties(N)) = solidify
the ties; etc.

Individual metonymic Sem-rule

‘government’

2 L(‘capital of Y’)

‘Y’

Figure 1.8 The Sem-rule for the metonymy: ‘government of country Y’ ~ ‘capital of country Y’
 1.2 Meaning-Text theory and functional models of natural language 17

The meaning ‘the government of country Y’ can be expressed by the name of the
capital of Y (‘government of Russia’ ⇔ Moscow, ‘government of USA’ ⇔ Wash-
ington, etc., as in strong ties between Moscow and Riyadh).
For syntactic rules, it is necessary to present DSynt- and SSynt-rules sepa-
rately; I will give three rules of each type.

1.2.3.2 Deep-syntactic Meaning-Text rules

DSyntS ⇔ SSyntS rules (deep syntax)
Establishing the value of an LF (step up [the flow])
CausPredPlus STEP
verb-marker-analytical
II II
UP

FLOW (N) FLOW (N)

Figure 1.9 The DSynt-rule for the value of LF CausPredPlus

The elements of the value of an LF f applied to L are taken from L’s lexical entry,
in this case L being flow(N).
NB For the SSynt-relations whose names appear in the rules, see Chapter 2.

Expressing the agent of a passive verb (permitted by Damascus)

L 1(V) PASSIVE

L 1(V) PASSIVE
passive-agentive

II BY
prepositional
L2

L2

Figure 1.10 The DSynt-rule for the agentive complement

Expressing the Genitivus Subjectivus complement of a noun (flow of terrorists)

L1(N)

L1(N)
subjectal-adnominal

I
OF
prepositional
L2(N)

L2(N)

Figure 1.11 The DSynt-rule for the subjectal adnominal complement

18 1 Meaning-Text linguistic model

1.2.3.3 Surface-syntactic Meaning-Text rules

SSyntS ⇔ DMorphS rules (surface syntax)
☛ 1. The symbol “+” specifies linear order of lexemes, and “…”, a possible gap between two
lexemes.
2. The expression “II[N]” in the syntactics of a verb indicates that its DSynt-actant II can be
expressed by a prepositionless noun—that is, that this verb is transitive.

Constructing the direct-object phrase (step up the flow; Have you seen (just) him?)

L1(V, II[N])
Condition
L1(V, II[N]) L2(N)
direct-objectival If L2(N) = L (pron),
+…+
then L2(N)OBL and no gap between L1 and L2
except for verbal adjuncts
L2(N)

☛ “OBL” stands for ‘oblique case of pronouns’ (me, you, him/her, us, them, whom).

Figure 1.12 The SSynt-rule for a direct object

Constructing the determinative phrase (the new computer, our president, these
and other girls)

L1(N, non-pron)
Condition
L1(N, non-pron) L2(N)
determinative +…+ If L2(determ) = THIS/THAT
then AGREE ADJ(N) (L2(determ), L1(N))
L2(determ)

☛  GREE ADJ(N) (Lʹ, L) is an agreement operator (set of rules) that ensures the agreement in
A
number between the demonstrative pronominal adjectives and the governing noun (this
tree vs. these trees).

Figure 1.13 The SSynt-rule for a determiner

Constructing the perfective-analytical phrase (Having, as everybody knows,

already written to Father, …)

HAVE
HAVE L(V)PPART
perfect-analytical
+…+

L(V)PPART

Figure 1.14 The SSynt-rule for a perfect form

 1.2 Meaning-Text theory and functional models of natural language 19

1.2.3.4 Morphological Meaning-Text rules

Two DMorph- and two SMorph-rules are given below.

DMorphS ⇔ SMorphS (Deep Morphology)

PL ⇔ {PL} (nominal plural suffixes)
PAST ⇔ A PAST
I⇒æ
(past tense apophony of the sing ~ sang type)

SMorphS ⇔ morph string (Surface Morphology)

{PL} ⇔ /z/ | /C[–sibilant], [+voiced] /__
⇔ /s/ not
| /C
with an N(-ǝn) /__
[–sibilant], [–voiced]

⇔ /ɪz/ | /C[+sibilant] /__

⇔ /ǝn | with an N(-ǝn)
A PAST
I⇒æ
(/drɪ́nk/) ⇔ /drǽnk/

There is no need to enter into the details of Meaning-Text morphology since the
interested reader has Mel’čuk 1992–2000 and 2006a at his disposal.
To conclude this section, a general architecture of an MTM is presented in
Figure 1.15.
Note that conceptics (= the module describing the correspondence between
a Conceptual Representation of reality and the SemR of the utterance) and
­phonetics (= the module responsible for the correspondence between the SPhonR
and actual articulated sound) remain outside of an MTM. They belong to a more
general model of linguistic behavior—a Reality-Speech model.

SPhonRs = Texts
phonemics
DPhonRs
surface morphology
SMorphRs
deep morphology
synthesis

analysis

DMorphRs
surface syntax
SSyntRs
deep syntax
DSyntRs
semantics
SemRs = Meanings

Figure 1.15 General structure of a Meaning-Text linguistic model

20 1 Meaning-Text linguistic model

1.2.4 M
 odeling two important linguistic phenomena: paradigmatic and
syntagmatic lexical choices

As we have known since F. de Saussure and R. Jakobson, any linguistic activity

is carried out by the Speaker along two axes: paradigmatic, where the selection of
linguistic units happens, and syntagmatic, where the units selected are combined
to produce utterances. To highlight certain advantages of MTMs, one particular
aspect of this activity can be considered: lexical choices.

1.2.4.1 Paradigmatic lexical choices

A choice of a lexical unit [LU] L by the Speaker is called paradigmatic if L is selected
from the lexical stock of the language more or less independently of its eventual
contextual neighbors. Two types of paradigmatic lexical choices are known: free
choices, where L is selected only for its meaning (and maybe some other spe-
cific properties, syntactic and/or morphological); and restricted choices, where
L is selected instead of another LU L′ as function of L′; in the text, L replaces L′.
Free choices are modeled by means of semantic decompositions, and restricted
choices, by means of semantic derivations.

Semantic decomposition
The sentences (4a) and (4b) are synonymous:

(4) a. John is sure that Mary is in town. ≡

b. John does not doubt that Mary is in town.

This means that they are mutually substitutable in text salva significatione (‘with
complete preservation of the meaning’). Both sentences are not factive and there-
fore both can be continued by … but this is not true.
Now, what type of information must a Speaker have in his brain about the
lexemes sure and doubt(V) in order to be able to manipulate them as he actually
does? We do not know for sure, but we can propose a plausible model.
Following A. Zholkovsky, A. Bogusławski and A. Wierzbicka, MTT proposes
that the mean­ings of these lexemes (like all lexical meanings, i.e., all semantemes)
consist of simpler meanings—in other words, that meanings are decomposable.
NB These simpler meanings are decomposable into even simpler meanings and so forth, until
semantic primitives, or meaning atoms, are reached (see Mel’čuk 2012–2015: vol. 2, 287).
 1.2 Meaning-Text theory and functional models of natural language 21

Consider the dataset (5):

(5) a. I think that Alan has come, but I am not sure of this.
b. I am sure that Alan has come, #but I do not think this.
c. I think that Alan has come, #but I doubt this.
d. i. I am sure that Alan has come. ≡
ii. I don’t doubt that Alan has come.
e. i. I am not sure that Alan has come. ≅
ii. I doubt that Alan has come.
☛ The symbol “ # ” indicates a pragmatically unacceptable continuation.

To enable a logical device to construct the sentences in (5) and to establish their
acceptability and synonymy, it is sufficient to represent the meanings under con-
sideration as follows:

(6) a. X is sure that P: ‘⟦Thinking that P takes place,⟧

X is not prepared to admit that P does not take place’.
b. X doubts that P: ‘⟦Not thinking that P takes place,⟧
X is prepared to admit that P does not take place’.

☛ 1. Special brackets ⟦…⟧ indicate a presupposition—the part of the meaning that does not
undergo negation or questioning in case of negation or questioning of the whole meaning.
2. ‘X thinks that P’ = ‘⟦Having the statement “P takes place” in X’s mind,⟧ X believes that
this statement is true’.

With the definitions in (6), one obtains for sentences in (5) the following semantic
decompo­sitions:

(7) a. ‘I think that A. has come, but⟦, thinking that A. has come,⟧ I am prepared
to admit that A. hasn’t come’.
b. ‘⟦Thinking that A. has come,⟧ I am not prepared to admit that A. hasn’t
come, #but I do not think that A. has come’ (a contradiction is marked
in boldface).
c. ‘I think that A. has come, #but⟦, not thinking that A. has come,⟧ I am
prepared to admit that A. hasn’t come’ (a contradiction is marked in bold-
face).
d. i. ‘⟦Thinking that A. has come,⟧ I am not prepared to admit that A. hasn’t
come’. =
ii. ‘⟦Thinking that A. has come,⟧ I am not prepared to admit that A. hasn’t
come’.
22 1 Meaning-Text linguistic model

e. i. ‘⟦Thinking that A. has come,⟧ I am prepared to admit that A. hasn’t

come’. ≈
ii. ‘⟦Not thinking that A. has come,⟧ I am prepared to admit that A. hasn’t
come’.
NB The case of (7d) vs. (7e) requires an additional explanation. (7e) represents a negation
of (7d); if (7d-i) and (7d-ii) are equivalent, how can it be that their negations are not equiva-
lent? The answer is the idio­matic character of the negation in (5d): I don’t doubt is not a
simple, “free” negation of I doubt, but an antonym meaning ‘I am sure’. (Technically, this is
manifested by the impact of the negation on the presupposition of ‘[to] doubt’, which
should not have happened under a free negation; see Iordanskaja 1986. If, in (5d), instead
of I don’t doubt we have a free negation It is not true that I doubt, sentences (5d-i) and ­(­5d-ii)
cease to be semantically equivalent.) For (7e), the contrast of the two presuppositions
creates the effect of a lesser certainty of the Speaker in (7e-ii) with respect to (7e-i).

Thus, the proposed decompositions allow for a formal and coherent description
of the data in (5). Semantic decompositions constitute an important descriptive
tool in MTT.

Semantic Derivation
Suppose the Speaker wants to talk about the person who is at the wheel of a car;
he needs the lexeme driver. But a person “driving” a locomotive is called an
engineer, the one “driving” a plane, a pilot, and a ship’s “driver” is a helms-
man. Similarly, a person who manages (= “drives”) a farm is a farmer, and the
one managing a shop is a shopkeeper. On the other hand, the client of a restau-
rant is a client or patron, that of a hospital, a patient, of a prison, a prisoner
or inmate, of a theater, a spectator, of a school, a student, etc. We see that
there are regular semantic links of the type car ~ driver, locomotive ~ engi-
neer, plane ~ pilot, ship ~ helmsman: they remind one of derivation, except
that in this case there can be no formal similarity between the members of the
pair. For the purposes of formal modeling of text produc­tion such lexical rela-
tions should be explicitly specified. MTT proposes to do that by means of lexical
functions [LFs], which are introduced in the next subsection; more precisely, these
are paradigmatic LFs.

1.2.4.2 Syntagmatic lexical choices

A choice of an LU L by the Speaker is called syntagmatic if L is selected as a func-
tion of one of its contextual neighbors; such a choice reflects what is known as
restricted lexical cooccurrence. Thus, English says MAKE a mistake but DO a favor,
TAKE an action, but BE ENGAGED in an activity, HOLD influence over N, but GIVE
N orders, etc. Similarly, we have a GREAT achievement, [a] DRASTIC action, [a]
 1.2 Meaning-Text theory and functional models of natural language 23

FIRM believer, SOLID grounds, [a] CATEGORICAL protest, quick AS LIGHTNING,

[to] rain HARD, etc.
Restricted lexical cooccurrence is a serious problem for any lexicographic
description. MTT proposes to solve it based on the following hypothesis:

In the majority of cases, restricted (i.e., synchronically arbitrary) lexical

cooccurrence manifests itself in the expression of a limited number of very
abstract, nearly-grammatical meanings.

Thus, in the first series of examples above, such a meaning is ≈ ‘do’, and in the
second, ≈ ‘very/intense’.
The crucial fact here is that such a meaning corresponds to a function f (in
the mathematical sense): the lexical unit L to which this meaning is applied is
the argument of this function and the set of appropriate collocates is its value.
Formally, we have
f(L) = {L1, L2, ..., Ln}.

To describe the lexically restricted expressions in the above examples, two such
functions are proposed; these are simple standard lexical functions (see immedi-
ately below):

Support verb Oper1

Oper1(mistake) = make [ART ~] Oper1(favor) = do [ART ~]
Oper1(rage) = be [in ART ~] Oper1(action) = take [ART ~]
Oper1(activity) = be engaged [in ART ~] Oper1(belief) = hold [ART ~]
Intensifier Magn
Magn(achievement) = great Magn(action) = drastic
Magn(believer) = firm Magn(dispute(V)) = hotly
Magn(quick) = as lightning Magn(sell(V, intrans)) = ˹like hot cakes˺

The phrases described by these LFs are nothing other than collocations: the collo-
cation’s base, selected independently by the Speaker for its meaning, is the argu-
ment of the LF, and the collocate, selected as a function of the base, is one of the
elements of its value.
Three properties of standard LFs prove to be especially important:
– Standard LFs are not numerous: there are about 60.
– Standard LFs are equally convenient for the description of both paradigmatic
and syntagm­atic restricted lexical choices. In other words, they allow for a
homogeneous and systematic des­cription of semantic derivations as well as
collocations.
24 1 Meaning-Text linguistic model

– Standard LFs are cross-linguistically valid, that is, they are language-univer-
sal in the sense that they are more or less sufficient for the description of
restricted lexical choices in all lan­guages. (It is certain that some languages
lack some of the proposed standard LFs, but it is not very probable for a new
language under consideration to require a new standard LF.)

Several examples of LFs will make the notion of LF clearer. (For more on LFs, see
Mel’čuk 1974: 78–109, 1982, 2007, 2012–2015: vol. 3, Ch. 14, and Wanner (ed.) 1996.)

Paradigmatic LFs
Action/property noun S0
S0(accept) = acceptance S0(intrude) = intrusion
S0(capable) = capacity S0(angry) = anger

Patient noun S2
S2(award) = recipient S2(shoot) = target
S2(sell) = merchandise S2(talk [to N]) = addressee
Active possibility adjective Able1
Able1(harm) = harmful Able1(rebellion) = restive
Able1(fight) = bellicose Able1(afraid) = cowardly

Syntagmatic LFs
Positive evaluation adjective Bon
Bon(contribution) = valuable Bon(service) = quality
Bon(idea) = good, promising Bon(weather) = fine, lovely, nice, …
Support verbs Oper, Func and Labor
Oper1(apology) = offer [ART ~] Oper2(apology) = receive [ART ~]
Func1(support) = comes [from NX] Func2(support) = goes [to NY]
Labor123(inheritance) = leave [NY as ~ to NZ]
Labor321(inheritance) = receive [NY as ~ from NX]
Realization verbs Real, Fact and Labreal
Real1(duty) = discharge [NX’s ~] Real2(treatment) = respond [to ART ~]
Fact0(film) = is playing, is in the theaters
Fact1(river) = empties [into NX] Fact2(bomb) = falls [on NY]
Labreal12(artillery) = hit [NY with ~]
Labreal21(invitation) = take up [NX on A(poss)(NX) ~]
Locative/temporal preposition Locin
Locin(list) = on [ART ~] Locin(end) = at [ART ~] Locin(program) = on [ART ~]
Locin(holiday) = on [~] Locin(socialism) = under [~] Locin(past) = in [ART ~]
 1.2 Meaning-Text theory and functional models of natural language 25

It is impossible to touch here on several interesting aspects of LFs: standard vs.

non-standard LFs, simple vs. complex LFs, configurations of LFs, fused elements
of the value of LFs, etc.

1.2.4.3 Correlations between the meaning and the collocates of a lexical unit
The proposed description of lexical meanings and restricted lexical cooccurrence
leads to sharpening of lexicographic definitions. Take, for instance, the noun
applause. In Longman Dictionary of Contemporary English, it is defined as ‘the
sound of many people hitting their hands together and shouting, to show that
they have enjoyed something’. This definition would be OK, if it weren’t for the LFs
Magn/AntiMagn of the noun applause: deafening, rapturous, … vs. thin, scat-
tered, etc. These adjectives indicate that the applause is gradable: the strength
and frequency of hitting hands together is (roughly) proportional to the approval/
enjoyment by the applauder. Therefore, the definition of applause (and that of
the verb applaud) must be corrected:

X applauds Y : ‘X claps hands to express X’s approval and/or enjoyment of Y,

the strength and frequency of clapping being proportional to X’s
approval/enjoyment’.

In this way, the lexicographic definition of a headword L and L’s collocations are
buttressing each other (see Iordanskaja & Polguère 2005 and Iordanskaja 2007).
Such a link is vital for the description of language, a system ‘où tout se tient’ [F.
de Saussure].

The Explanatory Combinatorial Dictionary

All types of information about an LU L necessary to ensure the correct use of L
in any con­text are stored in a dictionary of a special type: the Explanatory Com-
binatorial Dictionary [ECD]. It is explanatory since each L receives in it a semantic
decomposition (L’s “explanation”); it is combi­natorial because it specifies each
L’s syntactic and restricted lexical cooccurrence. In the framework of MTT, such
a lexicon plays a central role: L’s lexical entry contains all the semantic data con-
cerning L and all the combinatorial data that are used by MTM grammatical rules.
In this sense, MTT is lexically based.
Since the ECD has been described in detail in numerous publications (Mel’čuk
1974: 110–140, Mel’čuk et al. 1984–1999, Mel’čuk & Zholkovsky 1984[2016], Mel’čuk
& Polguère 2007, Mel’čuk 2012–2015: vol. 2, Ch. 11), suffice it here to state its defin-
ing properties:
26 1 Meaning-Text linguistic model

General properties of the ECD

–– The ECD is a theoretical lexicon elaborated within the framework of a full-
fledged linguistic theory, namely MTT.
–– The ECD is a formalized lexicon, written in terms of several predefined lexi-
cographic metalanguages.
–– The ECD is exhaustive at the level of each entry.

Specific properties of the ECD

–– The ECD is an active dictionary, supplying all the data in the direction Mean-
ing-to-Text.
–– The ECD is a semantically based dictionary: the definition (= semantic repre-
sentation) of the headword L underlies and determines L’s entry.
–– The ECD fully and systematically covers L’s restricted lexical cooccurrence—
in terms of LFs.
–– The ECD treats lexemes and idioms in the same way: all of them constitute
headwords of the corresponding entries.
–– Each ECD’s entry describes one (monosemous) LU; LUs related to each other
by polysemy are united within a superentry, called a vocable.

1.3 The value of functional models in linguistics

An MTM of a natural language is speculative by its very nature: analyzing speak-
ers’ be­havior, the linguist observes the associations between (understood) mean-
ings and (perceived) texts and makes inferences as to the underlying representa-
tions and formal rules that relate them. The question naturally arises: How can
we validate the model being proposed? At the present time, at least two experi-
mental techniques are available:

– Computerization of linguistic MTMs and the use of the resulting systems in

all branches of Natural Language Processing: machine translation, text gen-
eration, question-answering, etc.
– Psycholinguistic experimentation, which could shed precious light on the
psychological reality of the basic oppositions and descriptive formal objects
put forth in MTT. Thus, psycholinguistic experiments may contribute to our
knowledge of whether (or to what extent) it is correct to insist, as MTT does,
on the following five oppositions:
1) Linguistic synthesis (i.e. Meaning ⇒ Text) vs. linguistic analysis (i.e. Text
⇒ Meaning).
2) Static linguistic knowledge vs. dynamic procedural knowledge.
 1.3 The value of functional models in linguistics 27

(The information of the first type is specified in linguistic rules, which

constitute the MTM itself; the infor­mation of the second type is embodied
in algorithmic rules that manage the application of linguistic rules.)
3) Linguistic representations of various levels vs. modules (i.e. rules) of the
MTM that relate them.
4) The semantic representation, which targets meaning of sentences, ignor-
ing their “physical” organization, vs. the syntactic-morphological-pho-
nic representations, which target the structure of sentences, ignoring
their meaning.
5) The lexicon, where all data concerning individual LUs are stored, vs.
grammar, which pre­sent the information about classes of LUs and gram-
matical signs (affixes, apophonies, morphological conversions and mean-
ingful syntactic constructions).

It seems crucial to know whether the actual behavior of speakers is based on

these opposi­tions. We need to know more (much more!) on the psychological and
neurological differences between speakers’ encoding and decoding of texts, on
the way the strictly linguistic data are stored in the brain in contrast to procedural
knowledge, etc.
At the same time, the psychological correlates of MTT’s descriptive formal
objects are no less interesting:

–– The semantic structure: Is meaning really represented in the brain by

networks similar (isomorphic?) to those of MTT? Do speakers use semantic
decompositions? If so, exactly how? Is it true that the production of a sen-
tence begins with the shallowest and quite approximate semantic structure
available (as MTT has it)?
–– The syntactic structure: Is a sentence really represented in the brain by a
structure similar (isomorphic?) to a dependency tree of the MTT type?
–– The lexicon: Is the storage of lexical information in the brain similar (iso-
morphic?) to what is presupposed by the ECD of MTT? What are neurologi-
cal differences between encoding the meaning of sentences vs. encoding the
meaning of LUs in the lexicon? What are the mechanisms allowing a speaker
to apply the descriptions of LUs in his brain to the starting SemS in order to
produce the DSyntS of the future sentence? How is the interaction between
the starting SemS and the sentence under production carried out?
NB What is said about psycholinguistics here is no more than a wish aimed at the science of
tomorrow (or even of the day after tomorrow). The present level of research in psycholinguis-
tics seems insufficient for what is needed to validate/invalidate the Meaning-Text approach.
28 1 Meaning-Text linguistic model

The studies into the acquisition of language by children and adult learners, into
aphasic disorders, into diachronic developments, etc. also could contribute their
share to the acceptance /rejection of a given functional model.
To sum up: Functional models in linguistics, including MTMs, do not lack
ways and means of validation. These models are of high practical utility in at least
three technological and social domains: natural language processing; teaching
and learning languages; manufacturing reference books, such as dictionaries,
pedagogical grammars, and manuals.
The formal character of MTMs and their orientation (‘How is such-and-such a
thought expressed in such-and-such a language?’) are especially valuable in this
connection.
The theoretical impact of MTMs appears even more important. Scientific
progress until today has been basically addressing the problems of the physi-
cal universe: matter and energy. Since Homo sapiens started speaking, we have
developed new means of transportation (including spacecraft), enhanced our
physical strength manifold (remember the H bomb!), improved our organs of per-
ception (electronic microscopes and radio telescopes), widened our communica-
tion abilities (electronic media, the Internet), etc. We have penetrated the atom
and the depths of the Universe; we know a lot about the origins of our world
and the structure of our genes. But we have as yet made no comparable headway
in the mastery of information (in the scientific sense)—this evasive “substance,”
which is so central to life in general and to the life of humans in particular. We
do not know enough about the workings of our brain, while the enhancement of
the brain remains task number one for today’s science. Facing the challenges of
the 21st century, the humanity badly needs good models of human thinking and
reasoning (and, why not, of human emotions). This seems to be well understood
by the international scientific community, and the majority of scientists would
probably be in agreement with such a program.
However, strangely enough, people tend to forget—or disregard?—this vital fact:
 he only reliable key to human thinking, in all its complexity, is natural
T
language.
Without a profound understanding of how language is functioning in our psyche,
there will be no good understanding of information processing by the human
brain. That is why functional models of language, and MTMs in particular, now-
adays have acquired quite a special signi­ficance. Linguistics must take a place
of honor among the “hard” sciences, and functional models, which embody the
typical scientific approach to complex phenomena, will make their contribution. 1

A heavily reworked version of Mel’čuk 2009b.

Part II: Surface-Syntactic Relations
2 A
 general inventory of surface-syntactic
relations in the world’s languages
2.1 Introductory remarks
2.2 The deep-syntactic structure
2.2.1 Deep-syntactic relations
2.2.2 Fictitious lexemes
2.3 The notion of surface-syntactic relation
2.4 Conditions and criteria for establishing surface-syntactic relations in language L
2.5 An inventory of surface-syntactic relations found in the world’s languages
Appendix A Alphabetical index of surface-syntactic relations
Appendix B Index of passive surface-syntactic valences of word classes

2.1 Introductory remarks

This chapter—as this monograph in general—stands upon the following three

assumptions concerning the syntactic structure [SyntS] of a sentence:

1. The SyntS is described in terms of syntactic dependencies.

2. The SyntS is described on two levels of representation: deep-syntactic
structure [DSyntS] and surface-syntactic structure [SSyntS].
3. The SyntS uses a set of labeled relations—deep-syntactic relations [DSynt­
Rels] and sur­face-syntactic relations [SSyntRels].

Chapter 2 is dedicated to SSyntRels. For more information on linguistic depen-

dency in general, see Mel’čuk 1988 and 2009a; on SSyntRels in particular, see
Iordanskaja & Mel’čuk 2009a.
Let it be emphasized lest there be a misunderstanding: the proposed inven-
tory of SSyntRels is not an a priori universal construction; it is “universal” only
in the most trivial sense—it represents a set-theoretical union of the lists of
­SSyntRels established empirically for several languages. That is why it is called a
general inventory (rather than a universal inventory, as is the inventory of deep-
syntactic relations, see Section 2.2). It is not claimed that all of the SSyntRels
listed in Section 2.5 are encountered in any language. Rather, the set of these
SSyntRels is considered sufficient for the description of SSyntSs in the languages
that have been studied.
Tentative lists of SSyntRels for particular languages—Russian, English,
French, German, Spanish, Arabic, etc.—have been published over the last 60 years
(see the references in Iordan­skaja & Mel’čuk 2009a: 153). The general inventory of

https://doi.org/10.1515/9783110694765-003
32 2 A general inventory of surface-syntactic relations in the world’s languages

SSyntRels proposed below is based on a list of Russian SSyntRels (Mel’čuk 1974:

221–235, 2012c: 135–144; Iomdin 2010c), a list of English SSyntRels (Mel’čuk &
Pertsov 1987, Mel’čuk 2012–2015: vol. 3, 444–453 [3.4.11] and 2016: 184–194), a list
of German SSyntRels (Zangenfeind 2012), and two partial lists of French SSynt­
Rels (Iordanskaja & Mel’čuk 2009a, Poiret & Liu 2019); these lists have ­profited
from many previous works, duly indicated in the above titles. The resulting set
of SSyntRels underwent numerous corrections and additions, embodying the
experience acquired during the last decade. Therefore, any mismatch between
the present list and previous publications must be resolved in favor of the former.
NB There is also a universal inventory of SSyntRels based on syntactic dependency tree banks
for over 70 languages: “Stanford Universal Dependencies [UD]” (de Marneffe et al. 2014, de
Marneffe & Manning 2008/2015), https://universaldependencies.org. Unfortunately, it is impos-
sible to compare the SSyntRel inventory proposed below with Stanford Universal Dependencies:
the theoretical framework and the methodology are so different that a comparison would require
a serious special study. See Gerdes et al. 2018, which proposes a modification of Stanford UD,
making it closer to a linguistically valid inventory of SSyntRels.

The rest of this chapter is organized as follows: Section 2.2 offers a cursory char-
acterization of the DSynt-structure; Section 2.3 contains a few informal remarks
on the notion of SSyntRel; Section 2.4 describes the criteria for establishing the
inventory of SSyntRels in a given language; finally, Section 2.5 presents the list of
SSyntRels known to me today.

2.2 The deep-syntactic structure

To help the reader evaluate the suggested SSyntRels, it is necessary to present
the set of DSyntRels and that of fictitious lexemes appearing in DSynt-structures
of sentences: the fact is that each SSyntRel in the inventory of Section 2.5 is sup-
plied with the information of its corres­pondence to DSyntRels and/or fictitious
lexemes. (On DSyntS, see Chapter 1, Subsection 1.2.2, p. 12.)

2.2.1 Deep-syntactic relations

In sharp contrast with the SSyntRels, the DSyntRels are linguistically universal
in the following sense of the term universal:

Used together with fictitious lexemes, the proposed DSyntRels are suffi-
cient to describe the DSyntSs of any language.
 2.2 The deep-syntactic structure 33

There are thirteen DSyntRels, which have been established by theoretical reason-
ing (based, of course, on linguistic data).

•All syntactic constructions known in the world’s languages are divided into
two major families: coordinative vs. subordinative constructions.
The coordinative constructions are described by two DSyntRels:
COORD(inative) and PSEUDO-COORD.
NB In several previous publications, the PSEUDO-COORD DSyntRel (and the corresponding
SSyntRel) was called QUASI-COORDINATIVE. In order to improve the terminology, it was
decided to use the prefix quasi-[X] for an element that is not an X, but—under appropriate
conditions—can be treated as an X, i.e., quasi-Xs can be confounded with genuine Xs (for
instance, quasi-elementary [sign], quasi-gram­meme, quasi-morph). An element that is not an X
and can never be confounded with Xs, but resembles X to a sufficient degree will be called
pseudo-[X]. This modification concerns also such names of SSyntRels as *quasi-­subjectival
⇒ pseudo-subjectival, etc.

• The subordinative constructions are subdivided, in their turn, into weak-sub-

ordinative vs. strong-subordinative. The weak-subordinative constructions are
described by the DSyntRels ADDRESS(ative) and APPEND(itive).
• The strong-subordinative constructions fall into two subsets: the modifying
DSyntRels vs. the actantial DSyntRels; this division corresponds to the two
main syntactic phenomena: modification (or characterization) vs. complemen-
tation.
– T he modifying constructions are described by the DSyntRels ATTR(ibutive)
and ATTRdescr(iptive).
– The actantial constructions are described by the DSyntRels I, II, ..., VI, and
IIdir(ect).sp(eech).
NB A deep-syntactic actant [DSyntA] of a lexeme L is, as a rule, a syntactic dependent of L
that expresses one of L’s semantic actants—that is, a DSynt-actant of L is controlled by L’s
active valence. For instance, in the sentence Because of this, John reminded Mary about the
exam in my presence the lexemes JOHN, MARY and EXAM are DSyntAs of REMIND, since they
are imposed by the meaning of the verb (‘X reminds Y of Z’); however, the expressions
because of this and in my presence are not the verb’s actants, but freely added circumstan-
tials.

A lexical unit may have up to six DSynt-actants, which gives six actantial DSyn-
tRels. An additional DSyntRel is introduced for Direct Speech, which functions as
an object of a commun­ication verb:

[Mickey] shouted: –IIdir.sp→“Come [over right away!”].

Let me sum up. The 13 DSyntRels used in the Meaning-Text approach are as follows:
34 2 A general inventory of surface-syntactic relations in the world’s languages

Two coordinative DSyntRels

– COORD, or the coordinative DSyntRel, which represents normal coordina-
tion, either with­out a conjunction or with one:
Mary–COORD→John–COORD→or Ann ⇔ Mary, John or Ann
[Mary] comePAST,–[John]–COORD→leavePAST ⇔ Mary came, John left.
[Mary] comePAST,–COORD→and [John leavePAST] ⇔ Mary came, and John left.
– PSEUDO-COORD, or the PSEUDO-COORDINATIVE DSyntRel, which represents
syntactic construc­tions of elaboration, where, for instance, a prepositional
phrase follows—normally without a conjunction—another such phrase:
in–[New York]–PSEUDO-COORD→on–[Manhattan]–PSEUDO-COORD→at
[John’s] ⇔ [I stayed] in New York, on Manhattan, at John’s.

Eleven subordinative DSyntRels

– APPEND, or the APPENDITIVE DSyntRel, which subordinates such “extra-
structural” elements as parentheticals, interjections and prolepses to the
syntactic head of the clause:
sorry←APPEND–[I]–beIND, PRES, NON-PERF, NON-PROGR [busy] ⇔ Sorry, I am busy.
ok←APPEND–[I]–beIND, PRES, NON-PERF, NON-PROGR–APPEND→fortunately [here]
⇔ OK, fortunately I am here.
– ADDRESS, or the ADDRESSATIVE DSyntRel, which subordinates direct address
expressions to the syntactic head of the clause:
John←ADDRESS–[where]–beIND, PRES, NON-PERF, NON-PROGR [you?]
⇔ John, where are you?
ok–ADDRESS→Captain ⇔ OK, Captain.

In contrast to the APPEND DSyntRel, the ADDRESS DSyntRel is non-repeatable.

Two modifying DSyntRels

– ATTR, or the ATTRIBUTIVE DSyntRel, which describes all types of restrictive
modifier construc­tions (minus descriptive ones, see the next DSyntRel):
red←ATTR–flag ⇔ red flag
man–ATTR→of [great courage] ⇔ man of [great courage]
very←ATTR–interesting ⇔ very interesting
drive(V)–ATTR→fast ⇔ [John was] driving [very] fast
NB The name ATTR is in fact an abbreviation for ATTRrestr, that is, it is used for all restrictive
modifiers. A restrictive modifier restricts, or constraints, the denotation of the lexical
expression modified: thus, the LU [a] book denotes any book, but the phrase [a]
French←ATTRrestr–book denotes only a book printed in French. Since most modifiers are
restrictive—it is the default case—it is possible to omit the subscript restr for simplicity’s sake.
 2.2 The deep-syntactic structure 35

– ATTRdescr, or the descriptive-attributive DSyntRel, used for descriptive

modifiers, which do not restrict the denotation of the expression modified,
but simply qualify it:

Mary,–ATTRdescr→tired [and hungry, left] ⇔ Mary, tired [and hungry, left.]

Seven actantial DSyntRels

– I, as in
John←I–read ⇔ John is reading.
my/John←I–trip ⇔ my trip, John’s trip
translation–I→by [John] ⇔ translation [of this novel] by John

– II, as in
book←II–read ⇔ [John is] reading [a] book.
John←II–expulsion ⇔ John’s expulsion
for–II→John ⇔ for John

– III, as in
[book←II–]send–III→John ⇔ [Mary] sends [a] bookII to JohnIII. /… sends
JohnIII [a] bookII.
– IV–VI, as in
[hundred dollarPL←II–]lend–IV→month ⇔ [Would you] lend meIII
$$100II for a monthIV?
[Istanbul←III–]mission–IV→month ⇔ [a] mission to IstanbulIII for
–V→study a monthIV to studyV [Turkish]
– IIdir.sp, as in
whisper–IIdir.sp→comeIMPER ⇔ [John] whispered: “Come [back!”]

The full inventory of DSyntRels is given in Table 2.1, starting with the strongest
subordin­ative dependencies and going towards the weakest coordinative links.

Table 2.1 Universal inventory of deep-syntactic relations

subordinative DSyntRels COORDINATIVE DSyntRels

STRONG DSyntRels WEAK DSyntRels
Actantial DSyntRels Attributive DSyntRels
I II III IV V VI IIdir.speech ATTRrestr ATTRdescr address append pseudo-coord COORD
1 2 3 4 5 6 7 8 9 10 11 12 13
36 2 A general inventory of surface-syntactic relations in the world’s languages

2.2.2 Fictitious lexemes

Along with DSyntRels, the DSynt-structure uses fictitious lexemes, which carry
the lexical-type meanings expressed by meaningful SSynt-constructions (Mel’čuk
2012–2015: vol. 2, 37–42, 2018c). In other words, a fictitious lexeme represents, in
the DSyntS, a meaning expressed on the surface by a meaningful syntactic con-
struction, that is, a type of two-lexeme phrase where the surface-syntactic rela-
tion carries, in addition to the information about the syntactic link itself, also
a lexical-like chunk of meaning. Such constructions are quite idiosyncratic and
language-specific. In the DSyntS they have to be represented by artificial lexical
units—in order to avoid multiplying the set of DSyntRels. Let me give a couple of
examples.

Example 1: indirect object of Beneficiary

(1) John has painted me a beautiful landscape.

The wordform me is, of course, a surface-syntactic indirect object. But at the

DSynt-level it does not correspond to a DSynt-actant of the verb paint; semanti-
cally, it expresses the person for whose benefit the painting was created. To rep-
resent this construction in the DSyntS, if we do not want to introduce another
DSyntRel, we have to use a fictitious lexeme: «for».

☛ 1. A fictitious lexeme is indicated by angular quotes: « ».

2. On the form of linguistic rules, see Chapter 1, Subsection 1.2.3, pp. 15ff.

deep-syntactic structure surface-syntactic structure

L(V)
L(V)
ATTR

«FOR» ⇔ indirect-objectival

II
L(N)
L(N)

Figure 2.1 Deep-syntactic rule for the fictitious lexeme «for», that is, for producing the English
construction of the type “paint me [something]”
 2.2 The deep-syntactic structure 37

Example 2: “derision” schm-noun

(2) Politics, schmolitics! | Theory, schmeory. | Books, schmooks. | Baby, schmaby.

In English, the so-called schm-reduplication of a noun N expresses the Speaker’s

derision and skepticism about (the referent of) N; the N, schm-N phrase means
≈ ‘I dismiss N as being ludicrous and worthless’. (schm- is the name of the cor-
responding derivational means, which adds to the derived N the prefix /šm/- and
deletes the initial prevocalic cluster, if any.)

deep-syntactic structure surface-syntactic structure

L(N) L(N)

ATTR ⇔ reduplicative

«DERISION» SCHM-L(N)

Figure 2.2 Deep-syntactic rule for the fictitious lexeme «derision», that is, for producing the
English construction “N, schm-N”

Note that because of variegated lexical-type meanings attached to the operation

of lexeme reduplication in the world’s languages, a special reduplicative SSyntRel
is needed (or even several different reduplicative SSyntRels, see Section 2.5, No.
113, p. 108).
Here is an approximate inventory of fictitious lexemes. (It is by no means
complete, since it can be established only empirically, by collecting fictitious
lexemes found in different lan­guages.)

Table 2.2 Some fictitious lexemes in the world’s languages

«affect» «condition» «if» «represent»

«after» «dat_eth» «ifirr» «result_in»
«although» «derision» «include» «say»
«and_then» «distance» «instrument» «should»
«as_if» «during» «manner» «term»
«be» «every» «maybe» «title» (professor
«be_able» «for» (buy her a dress) «more» Drouin)
«be_from» «from» (one from these) «movedir» «very»
«be_located» «goal» «name» «while»
«belong» «have» «per» «will_be»
«cause(n)» «have_to» «price» «with»

NB A fictitious lexeme is not a semantic unit; it is only a conventional name for a (possibly, quite
complex) configuration of semantemes.
38 2 A general inventory of surface-syntactic relations in the world’s languages

2.3 The notion of surface-syntactic relation

A SSyntRel r (in the phrase L1–r→L2) is a linguistic sign of language L of the syn-
tactic level—a triplet ⟨signified ; signifier ; syntactics⟩ that describes a family of
similar syntactic constructions of L.

– The signified of this sign is the name r itself; r is characterized by a bundle of

the deep-syntactic relations to which it corresponds (= which it can express).
– Its signifier is a bundle of formal properties of the phrase L1–r→L2—namely,
word order, prosodic and morphological means applied to L1 and L2 in order
to express r in the sentence.
– Its syntactics is a bundle of combinatorial properties—that is, all indications
concerning cooccurrence of r with other SSyntRels in an SSynt-structure.

The inventory of SSyntRels of language L, just like the inventory of its gram-
memes or its pho­nemes, is established based on special criteria of three types:
Criteria A – C, given in Section 2.4.
Each SSyntRel r is described by the corresponding surface-syntactic rule;
SSynt-rules have the following form:

L1
L1m 1 L2m 2

r ⇔ + … + C

L2

The left-hand side of a SSynt-rule is a two-node subtree featuring the SSyntRel

linking lexemes L1 and L2; its right-hand side is the corresponding linear string
of the same lexemes supplied with syntactic grammemes, if any (m1 and m2); for
simplicity’s sake, I don’t consider prosodies, which mark particular SSyntRels in
some tonal languages; C stands for conditions under which this correspondence
holds.
SSyntRels are (by definition) language-specific, just as grammatical cases
or phonemes. Thus, the subjectival SSyntRel in Russian is, of course, not iden-
tical to the subjectival SSyntRel in Basque or Lezgian, neither semantically nor
formally—in the same way as the Russian nominative case is not identical with
the nominative in Kurdish or Dargwa, and as the Russian phoneme /t/ is by no
means identical with the phoneme /t/ in English, Mandarin Chinese or Hawaiian.
These facts, however, do not prevent linguists from compiling general inventories
of nominal cases and those of phonemes, based on the structural resemblance of
 2.3 The notion of surface-syntactic relation 39

the corresponding entities. With all possible distinctions between them, nomina-
tives of different languages are the cases of nomination; and /t/ is every­where a
voiceless dental plosive consonant. Quite similarly, the (syntactic) subject—the
depend­ent element of the subjectival SSyntRel—is the most privileged among the
surface-syntactic actants of a finite verb in any language, while the direct object—
the dependent element of the direct-objectival SSyntRel—is the second most
privileged among the actants of a transitive verb (see Chapter 3, Subsection 3.8);
etc. The idea of a general list of SSyntRels known today suggests itself, and the
present chapter picks up the challenge.

Sufficiency of the set of SSyntRels proposed. The SSyntRels on the list have
been esta­blished based on the following premise:

The surface-syntactic structure—in combination with the SSynt-communi-

cative and SSynt-anaphoric structures—must be sufficient for determining
word order, prosodiza­tion and morphologization of the sentence under syn-
thesis.

To put it differently, the SSyntS-to-DMorphS transition should not require infor-

mation from the preceding levels of representation (the DSyntR or the SemR):
whatever information is needed must be contained in the SSyntRels used—plus, of
course, the SSynt-communicative and SSynt-anaphoric structures as well as all the
necessary lexicographic information about the properties of lexical units involved.
NB This does by no means imply that our present list claims exhaustiveness. It is, of course, not
sufficient for the description of SSyntSs of any language: to compile such a list for about 7000
languages of the world is an enormous empirical task. But the intention behind this list is exactly
being exhaustive. In practical terms, this requires that criticisms leveled at the list should aim at
indicating syntactic cons­tructions not covered by it. 

Necessity of each SSyntRel on the list. The inventory of SSyntRels proposed

below (in Section 2.5) has not been systematically checked for minimality—that
is, for the necessity of each SSyntRel included; it is the SSyntRel sufficiency that
is pursued in the first place. At the same time, I strive for the maximal clarity, so
that I do not collapse two SSyntRels even if this seemed formally possible, but
would make the system less transparent. Therefore, some SSyntRels listed below
may turn out redundant.
40 2 A general inventory of surface-syntactic relations in the world’s languages

2.4 Conditions and criteria for establishing surface-syntactic

relations in language L

Each r must meet the following two general conditions (Iordanskaja & Mel’čuk
2009a):

Condition 1. The syntactic constructions described by the SSyntRel r resemble each

other to a sufficient degree—that is, they share enough linguistic properties. The
properties underlying SSyntRels are not linguistically universal in the strict sense
of the term: generally speaking, they may be relevant in language L under consid-
eration and irrelevant or non-existent in another lan­guage. Moreover, the proper-
ties relevant for one lexical class (say, for verbs) may be useless or inapplicable to
a different class. (For more on relevant linguistic properties, see Chapter 3, 3.2.3,
pp. 122ff.)

Condition 2. The SSyntRel r satisfies the formal requirements of Criteria A–C, given
below.
There are three groups of criteria for SSyntRels (Mel’čuk 2009a: 27–33, 2012–
2015: vol. 3, 411ff ):

– Criterion A  RESENCE of a SSyntRel between lexemes L1 and L2 in an

P
utterance.
– Criteria B1–B3 ORIENTATION of the SSyntRel between lexemes L1 and L2 in
an utterance.
– Criteria C1–C3 TYPE of the SSyntRel between lexemes L1 and L2 in an
utterance.

Criterion A – PRESENCE of a surface-syntactic relation between lexemes L1 and L2

For there to be a SSyntRel r between lexemes L1 and L2 in a given utterance U,
Criterion A requires two things:

– The configuration L1–r–L2 must be expressible by a prosodic unit, that is, by

a phrase of language L—not necessarily in the utterance U itself, but at least
in L in general.
– The linear position of one of the lexemes L1 and L2 in the utterance U must be
determined by the other.
2.4 Conditions and criteria for establishing surface-syntactic relations in language L 41

Prosodic unity and linear arrangement in the configuration L1–r –L2

In a given utterance U of L, the lexemes L1 and L2 can have a direct Synt-
dependency link, that is, they can form a configuration L1–r–L2, if and only
if both of Conditions 1 and 2 are simultaneously satisfied:
Condition 1
(a) General case
L1–r–L2 can be implemented by a phrase of language L, such as N—V,
V—N, ADJ—N, PREP—N, ADV—ADJ, NUM—N, etc.
(b) Special case
L1–r–L2 alone cannot be implemented by a phrase of L, but taken
together with a convenient configuration of lexemes from the set {Li}
appearing in the same utterance it can, such that the following three
configurations are implementable by phrases of L:
1) L1—{Li-1} L2—{Li-2}, 2) L1—{Li-1} and 3) L2—{Li-2}.
Condition 2
The linear position of one of the lexemes L1 and L2 in the phrase L1–r–L2 is
specified with respect to the other.

Examples

The Case (b) of Condition 1 covers two types of expressions:

(i) L1—L2—L3(N), as in oneL1—ofL2—themL3. Here, *one of is not a phrase, while the
utterances of them and one of them are phrases; consequently, the configura-
tion one—of is allowed for.

(ii) L1—{Li-1} L2(CONJ)—{Li-2}, as in

It becameL1—{obvious}{Li-1} thatL2—{he wasn’t there}{Li-2}.

Here, *became that is not a phrase, while became obvious and that he wasn’t there
are phrases, with became and that as their syntactic heads (see Criteria B immedi-
ately below); therefore, the configuration became—that is accepted as legitimate.

Criteria B – ORIENTATION of the SSyntRel between lexemes L1 and L2

In each configuration of lexemes L1–r–L2 in the utterance U, one of them syntac-
tically dominates the other, i.e., is its SSynt-governor, or the SSynt-head of the
phrase L1–r–L2. Inform­ally speaking, the SSynt-head of a phrase is the lexeme
that determines—at least, to a greater ex­tent than the other lexeme (its SSynt-
dependent)—different properties of the phrase according to Criteria B1–B3.
Let there be a phrase L1–r–L2.
42 2 A general inventory of surface-syntactic relations in the world’s languages

Criterion B1 – The passive SSynt-valence of a phrase

In the phrase L1–r–L2, the lexeme L1 is the Synt-governor of L2 if the follow-

ing condi­tion is satisfied:
The passive SSynt-valence of the L1–r–L2 phrase is determined by the
passive SSynt-valence of L1 to a greater extent than by that of L2; then
we have L1–r→L2.

To put it differently, the passive SSynt-valence of the phrase L1–r→L2 is rather that
of L1 than that of L2; the SSynt-head of a phrase determines more than any other of
its elements all the external syntactic links of the phrase.

Criterion B2 – The morphological links between the elements of a phrase and its external context
(in a language that has inflectional morphology)

If the phrase L1–r–L2 in which the passive SSynt-valence of its components

does not allow one to establish the SSynt-head is such that
– L1 controls the inflection of lexemes external to the phrase
– or L1’s inflection is controlled by such lexemes,
then L1 is the SSynt-head of the phrase: L1–r→L2.

The morphological “contact point” of the phrase L1–r→L2 is rather L1 than L2; it
is the SSynt-head of a phrase that, as a rule, interacts morphologically with its
context.

Criterion B3 – The semantic content of a phrase

If the phrase L1–r–L2 in which neither the passive SSynt-valence nor the
morphology allows one to establish the SSynt-governor means ‘a kind/an
instance of L1’ rather than ‘a kind/an instance of L2’, then L1 is the SSynt-
head of the phrase: L1–r→L2.

The denotation of the phrase L1–r→L2 is determined rather by the denotation of

its SSynt-head.
For Criteria B to be satisfied, at least one of the Criteria B1 > B2 > B3 must be
satisfied, such that other Criteria B higher in the hierarchy are not operational.
2.4 Conditions and criteria for establishing surface-syntactic relations in language L 43

Criteria C1–C3 – TYPE of the syntactic dependency between two lexemes

Criterion C1: presence of semantic contrast (Minimal pair test)
Notation: wi(L) is a wordform of lexeme L.

One and the same hypothetical SSyntRel r should not describe two phrases
w1(L1)–r→w2(L2) and w3(L1)–r→w4(L2)
if Conditions 1 and 2 are simultaneously satisfied:

Condition 1
These phrases contrast semantically, the contrast being manifested either
in the form of the phrases themselves or in the syntactic behavior properties
of their members.

Condition 2
If these phrases differ in their form, they differ only by some syntactic
means of expression—by word order of their elements, syntactic prosody or
syntactic gram­memes.

If Criterion C1 is satisfied—that is, if Conditions 1 and 2 are both satisfied, r should

be split into two different SSyntRels, r1 and r2, r1 ≠ r2.
For example, the Russian phrases žena–synt→drugaGEN ‘wife of.friend’ and
žena–synt→drugNOM [žena-drug] ‘wife who is a friend’ should be described by
two different SSyntRels, since these phrases semantically contrast and formally
differ only by the case of DRUG: the geni­tive case in the first phrase and the same
case as that of ŽENA in the second (žena–obj-ad­nom→druga, Section 2.5, No. 49;
žena[-]–qualif-appos→drug, ibidem, No. 78).
NB Criterion C1 is formulated here with an addition, previously absent. Namely, Condition 1 now
foresees the possible physical manifestation of the semantic contrast not only in the form of the
phrases under analysis, but also “… in the syntactic behavior properties of their members.” Syn-
tactic behavior of an LU includes its combinability with other LUs and word order with respect to
other phrases). This is an important amendment, which makes Criterion C1 more sensitive.1

1 Here is an example. In Russian, two genitive phrases dependent on the same N are mutually
ordered according to their SSyntRel:
obj-adnom
qual-adnom
(i) portret neobyčajnoj formy našego otca lit. ‘portrait of.unusual form of.our father’
vs. *portret našego otca neobyčajnoj formy
Two genitive phrases have identical form, but manifest different syntactic behavior of their mem-
bers—their different linear positions.
44 2 A general inventory of surface-syntactic relations in the world’s languages

Criterion C2 – syntactic substitutability (Substitution test)

An SSyntRel r must have a prototypical dependent that is allowable with

any governor.

For example, the Russian phrases xoču–synt→vypitʹ [kofe] ‘I.want drink [coffee]’
and mogu–synt→vypitʹ [kofe] ‘I.can drink [coffee]’ should be described by two
different SSyntRels—direct-objectival and infinitive-objectival:

1) The direct-objectival SSyntRel has a prototypical dependent, possible with any

governor: NACC; some—but not all—governors accept also VINF and čto/čtoby-
clause:

xoču–dir-obj→kofeACC ‘I.want coffee’ ~

xoču–dir-obj→vypitʹ [kofe] ‘I.want to.drink [coffee]’ ~
xoču–dir-obj→čtoby [on pil kofe] ‘I.want that [he drink coffee]’

2) The infinitive-objectival SSyntRel also has a prototypical dependent, possible

with any gover­nor: VINF; no governor accepts NACC:

mogu–inf-obj→vypitʹ [kofe] ‘I.can to.drink [coffee]’ ~

*mogu–inf-obj→kofeACC ‘I.can coffee’

If these SSyntRels are not distinguished, the “unified” SSyntRel will have no pro-
totypical dependent.

Criterion C3 – no limited repeatability (Cooccurrence test)

An SSyntRel r must be either unlimitedly repeatable or non-repeatable—

that is, it can­not be limitedly repeatable.

The phrases write–synt→after lunch, write–synt→in the next room, write–

synt→out of frust­rastion, etc. can all be described by the same SSyntRel: circum-
stantial, since the number of such dependents appearing simultaneously with
the same governor is theoretically unlimited. On the contrary, the phrases [They]
returned–synt→all and [They] returned–synt→drunk require two different SSyn-
tRels (floating-copredicative and subject-copredicative), since otherwise the depen-
dent will be repeatable exactly twice (They returned all really drunk).
 2.5 An inventory of surface-syntactic relations found in the world’s languages 45

2.5 A
 n inventory of surface-syntactic relations found in the
world’s languages

And now, to the surface-syntactic relations themselves. But before a general

inventory of SSyntRels in various languages is presented, let me illustrate the
transition between a DSynt-structure and an SSynt-structure in order to demon-
strate the correspondences between DSynt-relations and fictitious lexemes, on
the one hand, and the SSynt-relations, on the other hand. This is done for sen-
tence (3):

(3) Mary washed the floor clean.

WASHIND, PAST, NON-PERF, NON-PROGR WASHIND, PAST

object-result-
I II ATTR subjectival copredicative
«RESULT_IN» dir-objectival
⇔ MARY CLEAN
MARY FLOORSG, DEF
II
FLOORSG

CLEAN determinative

THE
Figure 2.3 Deep-syntactic and surface-syntactic structures of sentence (3)

The SSyntRels in this inventory are not supplied with full systematic explanations
and justi­fications; I limit myself to a minimum of examples and cursory remarks.
It goes without saying that the present proposal is still a sketch that needs to
be improved and sharpened in many respects.
For better surveyability, the SSyntRels described here are grouped as follows:

• First, SSyntRels are divided into subordinative and coordinative relations.

• Second, the subordinative SSyntRels are subdivided into two sets:
– Clausal SSyntRels, which typically hold between the heads of phrases
within a clause. These SSyntRels either link the verb to its actants/circum-
stantials, or are similar to such SSyntRels; certain clausal SSyntRels can
also hold between the lexemes within a phrase (for example, passive-agen-
tive-completive and comparative-objectival SSyntRels).
– Phrasal SSyntRels, which hold only between the lexemes within a phrase,
never between heads of phrases within a clause (for example, determina-
tive and modificative SSyntRels).
46 2 A general inventory of surface-syntactic relations in the world’s languages

• Third, inside of each subdivision of subordinative SSyntRels, the line is drawn

between valence-controlled SSyntRels, which necessarily embody complemen-
tation (e.g., compare→N1 to N2), and non-valence-controlled SSyntRels, which
can be either modifying (e.g., shining←mo­ment) or ancillary (= involving
structural, or grammatical, LUs; e.g., have→written).

An SSyntRel G–r→D is said be valence-controlled if and only if r is explicitly

mentioned in the government pattern of its syntactic governor G.
NB The terms modification and modifying are used here in a very broad sense—to refer
to any syntactically subordinated element that does not represent complementation and is
not an ancillary (structural, or grammatical) element.

The name of an SSyntRel is an adjective derived from the name of its dependent
member: subject ~ subjectival, direct object ~ direct-objectival, etc. The systematic
effort to have “self-ex­planatory,” logically derived names for SSyntRels some-
times results in names that are too long and cumbersome; such is, for instance,
the direct-object-comparative-conjunctional-completive SSyntRel, which subordi-
nates the complement of a comparative conjunction that is semantically corre-
lated with the direct object of the clause (see below, No. 102). For practical use,
such names can, of course, be abbreviated at will (in this case, it could be dir-obj-
compar-conj).
In the examples, the SSynt-dependent of the SSyntRel under examination
is boldfaced, and words not participating in the construction illustrated are
included in brackets.
First, a synopsis of the inventory of SSynt-relations possible in various lan-
guages.

I Subordinative surface-syntactic relations: 1–114

I.1 Clause-level SSyntRels: 1–46
I.1.1 Valence-controlled [= actantial, quasi-actantial and actantial-like]
­SSyntRels: 1–26
I.1.2 Non-valence-controlled SSyntRels: 27–46
I.1.2.1 Actantial SSyntRels: 27–28
I.1.2.2 Copredicative SSyntRels: 29–32
I.1.2.3 Circumstantial SSyntRels: 33–38
I.1.2.4 Extra-structural SSyntRels: 39–46
I.2 Phrase-level SSyntRels: 47–114
I.2.1 Any type of phrase SSyntRels, non-valence-controlled: 47
I.2.2 Nominal phrase SSyntRels: 48–83
 2.5 An inventory of surface-syntactic relations found in the world’s languages 47

I.2.2.1 Valence-controlled: 48–59

I.2.2.2 Valence-controlled and non-valence-controlled: 60–64
I.2.2.3 Non-valence-controlled: 65–83
I.2.3 Adpositional phrase SSyntRels, valence-controlled: 84–85
I.2.4 Verbal phrase (= analytical form) SSyntRels, non-valence-con-
trolled: 86–96
I.2.5 Conjunctional phrase SSyntRels, valence-controlled: 97–107
I.2.6 Word-like phrase SSyntRels, non-valence-controlled: 108–114
II Coordinative surface-syntactic relations: 115–122

Each SSyntRel r is supplied with the following three types of data:

– r’s correspondence with DSyntRels and/or with a fictitious lexeme. If there
is no such correspondence (that is, r is introduced into the SSynt-structure
by DSynt-to-SSynt-structure rules), r is not required to have a prototypical
dependent.
– r’s standard (= most frequent, normal) governor G. Other, more specific gov-
ernors may be indicated in the examples.
– r’s prototypical dependent D.

I Subordinative surface-syntactic relations: 1–114

I.1 Clause-level (= clausal) SSyntRels: 1–46
I.1.1 Valence-controlled SSyntRels: 1–26

Actantial, quasi-actantial and actantial-like SSyntRels (≈ subjects and objects)

It is worth reminding the reader that a SSynt-actant of a lexeme is not necessarily
valence-controlled by it (Mel’čuk 2012–2015: vol. 3, 94–99); for instance, the fol-
lowing types of SSynt-actant are non-valence-controlled:
– Beneficiary, as in [He] did–[it]–oblique-objectival→ for you
– Price, as in [He] did–[it]–oblique-objectival→ for two hundred dollars
– Raised Possessor, as in Fr. [On ne] lui ←indirect-objectival–trouve [pas de défauts]
lit. ‘People don’t to.him find of defects’
– Dativus Ethicus, as in Fr. Goûtez–dat-ethic→ moi [ça] ! lit. ‘Taste on.me this!’

In the DSynt-structure, such actants are introduced by means of fictitious lexemes

(in the above examples, «for», «price», and «dat_ethic»).

1. Subjectival SSyntRel. It expresses the DSyntRel I; the prototypical G is a

finite verb VFIN, the prototypical D is a prepositionless N.
48 2 A general inventory of surface-syntactic relations in the world’s languages

The subject is the most privileged dependent element of the clause in the lan-
guage under consideration (Chapter 3, Section 3.3, Defini­tion 3.1, p. 135).
[As the] reader←subj–will [see ...] | I←subj–am [fine.] | It←subj–was [dawning.]
That←subj–[John left]–amazed [us.] | It←subj–amazed [us that John left.]
To←subj–[mention this point]–is [important.]
It←subj–is [important to mention this point.]
NB In the sentences of the type of It amazed us that John left and It is important to mention this
point the boldfaced phrases are described as pseudo-subjects, see below, No. 6, p. 51.

To←subj–[read]–is [to empower,] to←subj–[empower]–is [to write.]

Carrying←subj–[out attacks]–became [increasingly difficult.]
Enough←subj–has [been said on this topic.]
dir-obj
[Which←way] to←subj–[choose]–must [be decided later.]
[The] easiest←subj–[of these solutions]–turned [out to be the last one.]

— The G = VFIN can be a zero wordform of one of the Russian lexemes bytʹ ‘be’:

Russian

Ivan←subj–ØBYTʹ [star] lit. ‘Ivan old’. | Ivan←subj–ØBYTʹ [v Londone] lit. ‘Ivan in London’.
Ja←subj–[tebe ne]–ØBYTʹ [mama!] lit. ‘I to.you not mom!’ = ‘I am not your mom!’
[Vot] ØBYTʹ–[tebe moja]–subj→ruka lit. ‘Here to.you my hand’. = ‘Take my hand’.

The G can be different from a VFIN; the following two cases are found.

— The G is an invariable interjection that expresses an instantaneous act and

functions as the Main Verb (one of so-called verbal interjections):

Russian (a colloquial construction)

Ivan←subj–bac [emu po morde] lit. ‘Ivan «bang» to.him on mug’. =
‘Ivan punched him in the face’.
Ivan←subj–xvatʹ [ego za šivorot] lit. ‘Ivan «grab» him by collar’. =
‘Ivan grabbed him by the collar’.
NB The interjection of this type is readily coordinated with a “normal” Main Verb: 
(i) Ivan podbežal k Petru i bac emu po morde
‘Ivan came.running to Petr and «bang» to.him on mug’.
(ii) Ivan bac Petru po morde i ubežal
‘Ivan «bang» to.Petr on mug and ran.away’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 49

— The G is an invariable adverb (particle?) that introduces Direct Speech:

German
“Die Reduktion der CO2-Emissionen war noch nicht so wichtig”,]
so–[der]–subj→Professor lit. ‘The reduction of CO2 emissions was still not so
important”, so [= ‘said’] the professor’.

The D can also be different from a prototypical prepositionless noun. Besides

quite common subjectival subordinate clauses, infinitives, adverbs, etc., illus-
trated above, the following cases should be mentioned:

— The D is a zero lexeme—a personal-indefinite pronoun with the meaning of

‘«they»’ (see Symbols, p. ix) or a “meteorological” pronoun:

Russian
[Tam] Ø(3,
people
pl)
←subj–rabotajut lit. ‘There «they» are.working’.
= ‘People are working there’.
sg, neu)←subj–bylo [temno]
[Na dvore] Ø(3,METEO lit. ‘Outside was dark’. ~
[Na dvore] Ø(3,METEO
sg, neu)←subj–ØBYTʹ
[temno] lit. ‘Outside dark’. = ‘Outside is dark’.

Spanish

(3, sg) ←subj–está [lloviendo] ‘It is raining’.

ØMETEO

On zero lexemes, see Mel’čuk 2006a: 469ff.

— The subject is a dummy (= semantically empty) pronoun introduced by

DSynt-to-SSynt-structure rules:
It←subj–is [great that you are here.] | It←subj–rained.

— The subject is a finite verb that is the head of a pseudo-relative clause (Chapter 6,
Sub­section 6.3.2.2):
dir-obj
[What he] has←subj–[written]–is [interesting.]

Note that in some languages, under specific conditions, the subject can be
doubled by a resumptive clitic:
50 2 A general inventory of surface-syntactic relations in the world’s languages

French
dir-obj
Le premier ministre←subj–considère-t-il que cette conversion n’est pas allée à son
terme ? lit. ‘The Prime Minister believes-he that this conversion did not run its
course?’
NB 1. The doubling of an actant by a resumptive pronoun—most often, a clitic—does not contra-
dict Crite­rion C3 (for establishing surface-syntactic relations), which forbids limited
repeatability of a particular type of dependent. The doubling clitic does not represent
another dependent: it is a syntax-imposed repetitive marking of the same dependent
(among other things, the actant and the doubling clitic are coreferential). Cf. the situation
with clitic doubling for objects, No. 14, p. 58.
2. A resumptive clitic cannot be described by a special SSyntRel since this clitic can appear
alone in its own right as a subject: Considère-t-il que cette conversion n’est pas allée à son
terme ?

2. 
Consecutive-subjectival SSyntRel. It expresses the DSyntRel I and the ficti-
tious lexeme «and_then»; the G is an invariable verbal form (a VIMPER, 2, SG
or a VINF, with or without preposition) and the prototypical D is a preposi-
tionless N.

— The G is an invariable imperative form in the 2nd person singular:

Russian (a colloquial construction)

Ivan←subj–vozʹmiIMPER, 2, SG [i skažiIMPER, 2, SG èto vslux] ‘Ivan up and said this aloud’.

— The G is an infinitive or a preposition that introduces an infinitive:

Russian (a colloquial construction)

[A] Ivan←subj–bežatʹINF, IMPF lit. ‘And Ivan to.run’. = ‘And Ivan took immediately to
his heels’.

French (a formal narrative construction; see Melis 2000)

[Et] moi←subj–de [répondreINF que je préfère les versions originales]

lit. ‘And me to answer that I prefer the original versions’.

Conditional-subjectival SSyntRel. It expresses the DSyntRel I and the ficti-

3. 
tious lexeme «if»; the G is a VFIN, and the prototypical D is a preposition-
less N.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 51

Russian (mostly colloquial construction)

Znaet–cond-subj→on, [čto ja ego ždu, – xorošo] lit. ‘Knows he that I him am.waiting,
is.good’. = ‘If he knows that I am waiting for him, this is good’.
Pridëšʹ–cond-subj→ty [vo-vremja – vsë budet v porjadke] lit. ‘Will.come you
on.time …’ = ‘If you come on time, everything will be in order’.
[O,] znal–[by]–cond-subj→ja[, čto tak byvaet!] lit. ‘ Oh, would.know I …’ =
‘If only I knew that it can be like this!’ [B. Pasternak].

NB In the last example the meaning of irrealis is rendered by the conditional-subjunctive form of
the verb (marked by the particle by), not by the SSyntRel itself.

4. 
Irrealis-subjectival SSyntRel. It expresses the DSyntRel I and the fictitious
lexeme «ifIRR» (≈ ‘if only’); the G is VIMPER, 2, SG, and the D is a preposition-
less N.

Russian

(4) Uznaj–irr-subj—→ja [ob ètom, vsë by bylo v porjadke]

learn-IMPER.2.SG I-NOM about this everything would be in order
‘Had I learned about this, everything would be in order’.

5. D
 ebitive-subjectival SSyntRel. It expresses the DSyntRel I and the fictitious
lexeme «have_to»; the G is a VIMPER, 2, SG, and the D is an N.

Russian (a colloquial construction)

(5) [A] ja←deb-subj–rabotaj!

and I-NOM work-IMPERF.IMPER.2.SG
‘And I have to work [when some other people don’t]!’

6. Pseudo-subjectival SSyntRel. It expresses the DSyntRel I or is introduced by

DSynt-rules; the G is a VFIN, and its prototypical D varies according to the
language (thus, in English it is a that-clause or to-infinitive).

[It←subj–]amazes–[us]–pseudo-subj→that [John left.]

[It←subj–]is–[vital]–pseudo-subj→to [keep accurate records.]

French

[Il←subj–est] venu–[trois]–pseudo-subj→voisins lit. ‘It has come three neighbors’.

[Il←subj–est] venu–[ton]–pseudo-subj→frère [et ses enfants]
lit. ‘It has come your brother and his kids’.
52 2 A general inventory of surface-syntactic relations in the world’s languages

— Pseudo-subjects also describe clefts:

[It←subj–]is–[novels that John]–pseudo-subj→prefers.

[It←subj–]was–[John who]–pseudo-subj→reacted [first.]
[It←subj–]was–[to John]–pseudo-subj→that [I spoke first.]
[It←subj–]was–[(to) John to whom I]–pseudo-subj→spoke [first.]

French

[C’←subj–]est–[moi qui]–pseudo-subj→ai [ouvert la fenêtre]

lit. ‘It is me who have opened the window’ [in French, the Main Verb of a relative
clause agrees not with the relative pronoun, as in English, but with its anteced-
ent, in this case—with moi].
NB In the preceding five examples, the lexemes novels, John, to [John] and moi depend on the
form of the verb ‘be’ by the copular-completive SSyntRel, No. 21, p. 63.

— Pseudo-subjects include a placeholder (= a dummy lexical element needed

to occupy a partic­ul­ar linear position; the Main Verb does not agree with a
placeholder, but agrees with the actual subject):

There←pseudo-subj–exists[–[a]–subj→condition for ...] vs.

There←pseudo-subj–exist[–subj→conditions for ...]
There←pseudo-subj–appear[–[to be better]–subj→rooms.]

German

Es←pseudo-subj–haben[–[einige interessante]–subj→Vorstellungen stattgefunden]

lit. ‘It have some interesting shows taken.place’. = ‘Some interesting shows have
taken place’.

7. 
Direct-objectival SSyntRel. It expresses the DSyntRel II; in most cases, the
G is a V(trans), and the prototypical D is an N.

The direct object [DirO] is the second most privileged clause element (see Chapter 3,
Section 3.8, Definition 3.2, p. 174).

[He wanted to] see–dir-obj→John. | [He] knew–dir-obj→this.

[He was] given–[the]–dir-obj→permission [to carry out his plans.]
[He] knew–dir-obj→that [Mary was in town.]
[He] knew–[Mary]–dir-obj→was [in town.]
[He] knew–[why Mary]–dir-obj→was [in town.]
[He] knew–[what method]–dir-obj→to [adopt.]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 53

[Which] way←dir-obj–[to]–choose [must be decided later.]

make–[possible]–dir-obj→neutralizing [the consequences]
make–dir-obj→it [possible to neutralize the consequences]
NB Cf. make–[it]–object-attributive-completive→possible [to neutralize the consequences], No. 24.

want–dir-obj→to [know]; [I] need–dir-obj→to [know–[what]–dir-obj→to [expect.]]

[This piece of history, the negotiators have] chosen–dir-obj→to [ignore.]

— The DirO can be a finite verb—the head of a pseudo-relative (= “free relative”)

clause:

[I] saw–[what John]–dir-obj→had [written.]

— The DirO can be a bare infinitive (= infinitive without preposition):

French

[Il] aime–dir-obj→nager ‘He likes to.swim’.

[Le député] dit–dir-obj→avoir [rencontré le patriarche orthodoxe Ignatius]
lit. ‘The congressman says have met the orthodox patriarch Ignatius’.

— The DirO can be a pronominal clitic:

French

[Nous] l’←dir-obj–avons [perdu] ‘We have lost it’.

Serbian

[Mama] ga←dir-obj–[je jedva]–naterala [da nosi jaknu]

lit. ‘Mom him has barely made that he.wears jacket’. =
‘Mom barely made him wear a jacket’.

— The DirO can be a demonstrative nominal pronoun modified by a relative

clause (see No. 73, p. 89):

Russian

[Ja] videl–dir-obj→to[, čto on napisal] lit. ‘I saw that what he had.written’.

French

[J’ai] vu–dir-obj→ce [qu’il avait écrit] lit. ‘I have seen that what he had written’.
54 2 A general inventory of surface-syntactic relations in the world’s languages

— The DirO can be a dummy pronoun introduced by DSynt-to-SSynt-structure

rules:

[I] made–dir-obj→it [clear that I am serious about tackling the problem.]

— In some languages, the DirO can be introduced by a preposition (under spe-

cific conditions, for instance, if the DirO is animate, definite, etc.):

Spanish

Vi–dir-obj→a [tu amigo] lit. ‘I.saw to your friend’.

Romanian

[L’am] văzut–dir-obj→pe [prietenul tău] lit. ‘Him I.have seen to friend.the your’.
NB In this sentence we see the doubling of the DirO by a clitic: l’[am]; see the remark on p. 58.

— The DirO can be a cognate object that expresses a Deep-Syntactic manner

circumstantial:

[He] died–[a terrible]–dir-obj→death ≈ ‘He died in a terrible way’.

NB Cf. the description of cognate objects in Russian and Arabic on p. 72.

— The G of a DirO can be an invariable clausative:

Russian

Doloj–dir-obj→carja! lit. ‘Down.with tsar!’

Von–dir-obj→eë [otsjuda!] lit. ‘Off.with her from.here!’

— A special case: an IndirO/OblO masquerading as a DirO (Mel’čuk 2012–2015:

vol. 3, 498).

[It is quite] like–indir-obj→John. | [be] worth–[a]–indir-obj→trip

German

WasACC←obl-obj–fragst [Du michACC?] lit. ‘What ask you me?’ =

‘About what are you asking me?’

8. 
Quasi-direct-objectival-1 SSyntRel. It expresses the DSyntRel II; the G is a
V(trans) of a particular semantic class, and the prototypical D is a preposi-
tionless N; see Iordanskaja & Mel’čuk 2009a: 190–192.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 55

This SSyntRel describes “deficient” DirOs, as those seen below:

[The ticket] cost–[300]–quasi-dir-obj-1→dollars.

[The table] smells–[the]–quasi-dir-obj-1→herring.

These DirOs are deficient in the sense that they do not have all the properties of
normal DirOs: for instance, they do not passivize.

 uasi-direct-objectival-2 SSyntRel. It expresses the DSyntRel III; the G is a

9. Q
light (= collocative) verb, and the D is a prepositionless N.

This SSyntRel describes, in particular, a V→N collocation that functions syn-

tactically as one transitive verb. In terms of lexical functions, the collocate V is
Labor12 of the collocation base noun N; see Mel’čuk 2012–2015: vol. 3, 93–94.

(6) a. Korean (see Chapter 4, Section 4.6.2.2, pp. 198–200)

dir-obj
John+i enehak +ɨl koŋpu+lɨl←quasi-dir-obj-2–hay+ss +ta
SUBJ linguistics ACC study(N) ACC do PAST DECL
NB 1. SUBJ is the subjective case, which marks the subject and the pseudo-conjuncts
of the subject (No. 122, p. 112), but cannot be used for nomination.
2. The collocation ‘do study(N)’ functions as a transitive verb ‘[to] study [some-
thing]’; ENEHAK ‘lin­guistics’ is a DirO.

b. Persian
Madär Ramin-ra bedar←quasi-dir-obj-2–kärd
mother DirO wakening(N) did
‘The mother woke Ramin’.
NB The collocation ‘do wakening(N)’ is used as a transitive verb ‘[to] wake up [someone]’.
See Chapter 9, Section 9.2.1.1.2, (13), p. 320.

10. Pseudo-direct-objectival SSyntRel. It expresses the DSyntRel II; the G is a

V(trans), and the prototypical D is a that-clause.

dir-obj
make—[it clear]–pseudo-dir-obj→that [we want to neutralize the consequences.]
dir-obj
[He] doubts—[it]–pseudo-dir-obj→that [we want to neutralize the consequences.]
[The rumor] has–[it]–pseudo-dir-obj→that [you are looking for a job.]
make–[it possible]–pseudo-dir-obj→to [neutralize the consequences]
[Girls] like–[it very much]–pseudo-dir-obj→when [you think of them.]
56 2 A general inventory of surface-syntactic relations in the world’s languages

11. Infinitival-objectival SSyntRel. It expresses the DSyntRel II; the G is a V(modal),

and the D is a VINF. (In English, it will be a bare infinitive, i.e. without to.)
can–inf-obj→read; should–inf-obj→read
NB But Fr. [Je] peux–dir-obj→nager ‘I can swim’. ~ [Je] le←dir-obj–peux lit. ‘I can it’.

12. Direct-speech-objectival SSyntRel. It expresses the DSyntRel IIdir.sp; the G

is a communication verb/noun, and the D is the head of a Direct Speech
expression.

[Churchill] declared–[in front of a large audience in Brooklyn: “I]–dir-sp-obj→am

[sure that the great struggle of the future would be between English-speaking
nations and communism.”]
NB Cf. He whispered,–“[Three]–dir-sp-obj→words!” vs. He whispered–[three]–dir-obj→words.
[Then Edmund Burke uttered his famous] sentence–[on the wrongs of Ireland: “No
country, I believe,]–dir-sp-obj→suffered [so much on account of religion.”]

Russian

[On tolʹko] ulybnulsja–⟨maxnul–[rukoj]⟩–[: “Ja vsë]–dir-sp-obj→sdelaju!”

lit. ‘He just smiled ⟨waved with.hand⟩: “I’ll do everything!” ’
NB Such verbs as ulybnutʹsja or maxnutʹ are intransitive and cannot have a DirO. They are not
even bona fide communication verbs, but they can be used as such to express the fictitious
lexeme «say». In Russian, Direct Speech can be introduced by a verb denoting its author’s
gesture (like ‘smile’), a brusque change of state (like ‘flare up’) or a brusque action (‘wave his
hand’); see Iordanskaja & Mel’čuk 1981.

13. Affected-objectival SSyntRel. It expresses the fictitious lexeme «affect»,

which represents the meaning of the Mandarin preposition bǍ , in the
construction
G–ATTR→«affect»–II→D
and is the result of Possessor Raising; the G is V(trans)FIN, and the D is a
bǍ→N phrase (Chapter 8, Section 8.4, p. 304).

Mandarin Chinese

(7) a. [Wǒ] bǎ←aff-obj–[John]–bǎng -le [liǎng zhī jiǎo]

I BǍ tie.up PERF two CLASS(ifier) foot
lit. ‘I as.for John tied.up two feet’. = ‘I tied up John’s feet’.

Here, jiǍ o ‘foot’ is a DirO; cf. No. 7, p. 52.

 2.5 An inventory of surface-syntactic relations found in the world’s languages 57

b. [Wǒ] bǎ←dir-obj–[wǒ-de qìchē]–mài-le

lit. ‘I BǍ my car sold’. = ‘As for my car, I sold it’.
NB The preposition bǍ introduces a special DirO that is a Focalized Given Theme and is
anteposed to the Main Verb. This construction is possible only in case of a transitive
verb that expresses an action upon the referent of the DirO.

For the affected object in Korean, see Chapter 4, 4.6.2.4, p. 201.

14. Indirect-objectival SSyntRel. It either expresses the DSyntRel II or III, or is

the result of a raising, or else expresses a fictitious lexeme «for»; the G is
a V(trans), and the prototypical D is a PREP→N phrase or an NDAT.

The IndirO is the third most privileged clause element (after the Subj and the DirO).
give–indir-obj→John ⟨him⟩ [some money]
[France] offers–indir-obj→Christians [asylum after Mosul threat.]
give–[some money]–indir-obj→to [John, who needs it]

French
[Marie] luiII←indir-obj–ressemble ‘Mary resembles–dir-obj→him/her’.

— In French, an IndirO often is a raised Possessor of the DirO; the same happens
in some other Romance and Slavic languages:

French
[Le bandit] lui←indir-obj–a [cassé le bras] lit. ‘The bandit to.him has broken the arm’.
[Va] te←indir-obj–laver [les mains] lit. ‘Go to.you wash the hands’.

Spanish
Le←indir-obj–han [robado la cartera] lit. ‘[«They»] to.him have stolen the wallet’.

Serbian
Proučavali–[smo]–indir-obj→mu [život] lit. ‘Having.studied we.are to.him life’. =
‘… his life’.

Russian
[Ona] porvala–indir-obj→mne [rubašku] lit. ‘She tore to.me shirt’. = ‘She tore my
shirt’.
[Ja] tebe←indir-obj–[ne]–ØBYTʹ [mama!] lit. ‘I to.you not mom!’ = ‘I am not your
mom!’
58 2 A general inventory of surface-syntactic relations in the world’s languages

— In many languages, the indir-objectival SSyntRel expresses the fictitious

lexeme «for»:

Mary cooked–indir-obj→them a good supper.

Russian
Uberi–indir-obj→mame komnatu! ‘Clean to.mom room!’ = ‘Clean mom’s room!’
oblique-obj
NB Cf. Postavʹ èto mame←possessive–[v]–komnatu! ‘Put this to.mom into room!’ = ‘Take this to
mom’s room!’ [*Postavʹ èto mame!]. See No. 60, p. 84.

In some languages, under specific conditions, the direct-objectival and indirect-

objectival SSynt­Rels are also repeatable just twice, like the subjectival SSyntRel,
pp. 49–50. More precisely, an object can or must (depending on the language) be
repeated (= “resumed”) by the corresponding clitic. As indicated above, this does
not contradict Criterion C3 of syntactic dependency, since the resumptive clitic is
not another DirO/IndirO, but an additional marker of the same object.

(8) Spanish
a. (i) La←dir-obj–veré–dir-obj→a [María]
lit. ‘Her I.will.see to Maria’.
(ii) Le←indir-obj–di–indir-obj→a [María el libro]
lit. ‘To.her I.gave to Maria the book’.
Bulgarian
b. (i) Knigata ja←dir-obj–četa veče cjal mesec
lit. ‘The.book it I.read already whole month’.
(ii) I←indir-obj–stana–[lošo]–indir-obj→na [Marija]
lit. ‘To.her became badly to Maria’. = ‘Maria became sick’.
NB A clitic can of course appear alone—without the noun to resume. This is the reason why
it cannot be treated as a different clause element.

In contrast to the DirO, the IndirO can have a noun as its G:

[John’s] answer–indir-obj→to [Mary]; [one of my first] gifts–indir-obj→to [her]

Now we come to oblique-objectival SSyntRels. Two (or more) oblique-objectival

SSyntRels are needed for two reasons:

1) In order to control the linear disposition of different OblOs in the absence of

communic­ative indications.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 59

2) For languages where different OblOs are marked by different grammatical

cases (rather than by prepositions), in order to determine the corresponding
case under synthesis.

All OblOs have the same or almost the same syntactic properties; they are distin-
guished according to their correspondence to DSynt-actants. The G of an OblO
can be, as in the case of an IndirO, also a noun.

15. Oblique-objectival-1 SSyntRel. It expresses one of the DSyntRels II–VI; the

G is a lexeme with a corresponding government pattern, and the proto-
typical D is a PREP→N phrase or an N in an oblique case; a completive
clause and a to-infinit­ive are also possible.

The OblO is the fourth (fifth, sixth, …) most privileged clause element.

help–[Mary]–obl-obj-1→in [her studies]; help–[Mary]–obl-obj-1→move [to London]

[McGuire] weighed–[in]–obl-obj-1→on [what is wrong with our school.]
[They] proposed–[Alan]–obl-obj-1→as [director.]
[They] held–[Alan]–obl-obj-1→for [a poet.]
[old] enough–obl-obj-1→for [this book]; [old] enough–obl-obj-1→to [understand]
held–obl-obj-1→in [contempt by colleagues]
translate–obl-obj-1→from [Hungarian into Greek]
translation–obl-obj-1→from [Hungarian into Greek]
sentenced–obl-obj-1→to [death for his crimes]
agreement–obl-obj-1→between [Stalin and Hitler]
[with no] objections–obl-obj-1→from [the Minister]
NB The synonymous phrases the Minister’s objections and objections by the Minister have
­different SSyntSs: the Minister’s←possessive–objections and objections–subj-adnom→by [the
Minister].

Down–obl-obj-1→with [the Mullahs!]

[It is ten] feet←obl-obj-1–high. | [fifteen-thousand]-foot←obl-obj-1–high [peak]
too–[tired]–obl-obj-1→to [go out]; too–[sweet]–obl-obj-1→to [eat]
[John was] clever(GP1)–obl-obj-1→to [leave.] ~ [To leave was] clever(GP2)–obl-obj-1→of [John.]
NB These sentences contain the adjective clever with two different government patterns [GPs].
[travel to several European cities,] such–obl-obj-1→as [London, Paris and Florence]

French

[Il] en←obl-obj-1–ressort [que les dépenses n’ont pas augmenté]

lit. ‘It from.this follows that the expenses have not risen’. =
‘This indicates that expenses …’.
60 2 A general inventory of surface-syntactic relations in the world’s languages

— Three interesting special cases of oblique object:

• The OblO of a superlative (most, least, best, highest, … ); this OblO indi-
cates the scope of the superlative (best in what set?):

[the] most–[expensive car]–obl-obj-1→in [France] ~

[this car, the] most–[expensive]–obl-obj-1→in [France]
NB Cf. [the] most–[expensive]–elective→of [French cars] (see No. 58, p. 83).
[the] best–[care]–obl-obj-1→possible; [the] highest–[energies]–obl-obj-1→available
[the] best–[care we]–obl-obj-1→can [provide]

• A completive clause/a to-infinitive:

[Mary has] such–[beautiful eyes]–obl-obj-1→that [she got a job as a make-up model.]

[Mary was] so–[tired]–obl-obj-1→that [she could not eat.]
[Mary was] too–[tired]–obl-obj-1→to [eat.]

Russian

[Marija] sliškom–[ustala,]–obl-obj-1→čtoby [poestʹ]

lit. ‘Mary too became.tired in.order.to eat’.

• A prepositionless noun
[Parents] named–[her]–obl-obj-1→Mary
Rus. [Roditeli] nazvali–[eë]–obl-obj-1→MarijaNOM/MariejINSTR [idem]

16.  Oblique-objectival-2 SSyntRel. It expresses one of the DSyntRels III–VI; the

G is a V with a corresponding government pattern, and the prototypical D
is a PREP→N phrase or an N in an oblique case.

translate–[from Hungarian]–obl-obj-2→into [Greek]

translation–[from Hungarian]–obl-obj-2→into [Greek]
sentenced–[to death]–obl-obj-2→for [his crimes]

Hungarian
obl-obj-2
(9) fordítás–obl-obj-1→magyar +ról orosz +ra
translation Hungarian DEL(ative) Russian SUBL(ative)
‘translation from.Hungarian into.Russian’

17.  Infinitive-oblique-objectival SSyntRel. It expresses the DSyntRels II or III

that correspond to Sem-actants 2 and 3; the G is a verb with a correspond-
ing govern­ment pattern, and the prototypical D is a (PREP→)VINF.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 61

[John] ordered–[his platoon]–inf-obl-obj→to [hide behind growth nearby.]

[John was] forced–inf-obl-obj→to [leave.]
[John] made–[his platoon]–inf-obl-obj→laugh.
[John] had–[better]–inf-obl-obj→leave [right away.]

French

[Je me] hâte–inf-obl-obj→de [partir]

lit. ‘I myself hurry to leave’. = ‘I am in a hurry to leave’.

18.  Infinitive-copredicative-objectival SSyntRel. It expresses the DSyntRel III

that corres­ponds to a “part” of Sem-actant 2 of the governing verb, the
other part being expressed by DSyntRel II; the G is a semantically biac-
tantial transitive verb with a corresponding government pattern, and the
prototypical D is a (PREP→)VINF.

This SSyntRel describes the Accusativus cum infinitivo construction, which has the
following semantic and deep-syntactic structures:2
‘see–2→waltz–1→people’ ⇔ people←II–see–III→waltz
See–[peopleII]–inf-copred-obj→waltzIII, see–[people]–inf-copred-obj→dance!
[Ximénez] observed–[the animals]–inf-copred-obj→to [cross waters more than 250 m
wide.]
[The test was] determined–[by the UN]–inf-copred-obj→to [be in violation of a UN
resolution.]
[I] like–[her]–inf-copred-obj→to [be slim.]
NB Cf. [I] like–[her]–obj-attr-obj→slim (No. 24, p. 65).

Latin

[Ceterum] censeo–[Carthaginem]–inf-copred-obj→esse [delendam]

‘Moreover, I.consider CarthagoACC to.be which.must.be.destroyed’ [Cato the Elder].

2 Accusativus cum infinitivo/participio

The Accusativus cum Infinitivo/Participio construction appears with semantically bi-actantial
verbs whose SemA 2 is a statement P: ‘X knows that P’ or ‘X says that P’. ‘P’ itself means ‘Y Z-s’,
so that in the construction under discussion, the lexeme L(Y) becomes a DirO (≈ “accusative”)
and L(Z) is implemented as an inf-copred-objectival SSyntRel (“infinitive”) or an obj-attr-objectival
SSyntRel (“participle”). The particularity of this construction is that, contrary to other actantial
infinitives/participles, its Ds do not directly correspond to SemAs of their G, but realize on the
surface each a “part” of its SemA 2.
62 2 A general inventory of surface-syntactic relations in the world’s languages

19.  Agentive-completive SSyntRel. It expresses the DSyntRel I; the G is a non-

finite verb form V (an infinitive or a gerund), and the prototypical D is
either a for→N phrase (in English) or a prepositionless N.

An agentive complement [AgCo] is a transform of a subject with a VFIN.

— The G is an infinitive:
obl-obj-1
[His thumb is too sore] for←agent-compl–[him]–to [play next week.]
obl-obj-1
[He asked] for←agent-compl–[the British]–to [stay longer.]
oblique-obj-1
NB Cf. [He] asked–[the]–dir-obj→British to [stay longer.]

Me←agent-compl–worry?

Spanish

[¿Qué estaba haciendo antes de] aparecer–[los]–agent-compl→problemas?

lit. ‘What was [s/he] doing before to appear the problems?’ = ‘… before the prob-
lems appeared?’
[Al] morir–[su]–agent-compl→madre [la niña quedó sóla]
lit. ‘At.the to.die [her] mother the girl remained alone’.

— The G is a so-called personal infinitive, which agrees with its agentive com-
plement [AgCo] in person and number:

Portuguese

(10) [O guarda fez sinal] para←agent-compl–[os motoristas]–par+ar +em

the guard made signal for the motorists stop INF 3.PL
‘The guard made the signal for the motorists to stop’.

— The G is a gerund (= a converb):

Spanish

[Nos casamos hace 50 años] estando–agent-compl→yo [sin trabajo]

lit. ‘We married 50 years ago being I without work’.

Turkish (c = /ǯ/)

Hasan←agent-compl–var+incaCONVERB [denize girelim] ≈ lit. ‘Hasan having.arrived,

to.sea let’s.go’. = ‘As soon as Hasan arrives, let’s go swimming’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 63

Kazakh

Kün←agent-compl–bat+ɨpCONVERB, [el orɨnġa otɨrdɨ] lit. ‘Sun having.set, people on.place

sat.down’. = ‘When the sun set, people sat.down on their seats’.

More generally, the construction CONVERB–agent-compl→L is found in many

Altaic and Uralic languages (Turkish, Mongolic, Tungusic, Finnic), in Nakh-
Daghestanian family, in Inuktitut (Eskimo), and elsewhere. Thus, an agentive
complement depending on a converb, is also typical of Korean (Chapter 4, Sub-
section 4.6.1.2, p. 195).
NB The construction John being sick, [we were unable to leave] is different. Here the noun John
is the head of the phrase John being sick, and the participle depends on it via the absolute-modifi-
cative SSyntRel, No. 62, p. 85.

20. Passive-agentive-completive SSyntRel. It expresses the DSyntRel II; the G is

a VPASS, and the prototypical D is a PREP→N phrase or an N in an oblique
case.

written–pass-agentive→by [McGuire]; baffled–pass-agentive→by [quantifiers]

[She was] sent–[a letter]–pass-agentive→by [McGuire.]

French

[Il a toujours été] aimé–pass-agentive→des [femmes]

lit. ‘He has always been loved of.the [= ‘by.the’] women’.

Russian

[On vsegda byl] ljubim–pass-agentive→ženščin+amiINSTR

lit. ‘He always was loved by.women’.

The SSyntRels 21–25 describe different copular complements, which appear with
verbs of a particular semantic type—namely, copular verbs.
NB A copular verb is a copula, i.e. be and become, or a verb whose signified includes the seman-
teme ‘be’ not in the dominant position; for instance, seem, appear [as], look [nice], etc.

21. Copular-completive SSyntRel. It expresses the DSyntRel II; the G is a copula

V—‘be’ or ‘become’, and the D is an ADJ, which semantically bears on the
subject and can agree with it.

NB If the G is a copula that means ‘be identical’ or ‘be an element of the class’, the prototypical
D is an N:
[It 〈This person⟩] was–cop-compl→John.
[John] is–cop-compl→engineer.
64 2 A general inventory of surface-syntactic relations in the world’s languages

be–cop-compl→easy; become–cop-compl→easy
be–[a]–cop-compl→teacher; become–[a]–cop-compl→teacher
[To read] is–cop-compl→to [empower.] | [He has the right to] be–cop-compl→it.

— The copula can be a zero wordform:

Russian

[Ivan] ØBYTʹ–cop-compl→bolen+ØMASC, SG ‘Ivan [is] ill’. ~

[Ivan] byl—cop-compl→bolen+ØMASC, SG ‘Ivan was ill’.

22. Copular-genitive-completive SSyntRel. It expresses the DSyntRel II; the G is

a copula V—‘be’ or ‘become’, and the D is an NGEN or a PREP→N phrase.

Russian

byl–[takogo že]–cop-gen-compl→tipa ‘[It] was of the same type’.

NB [Glavnoe prepjatstvie] ØBYTʹ–[ego solidnyj]–cop-compl→vozrastNOM
‘The.main obstacle [is] his advanced age’. vs.
[On uže] ØBYТʹ–[solidnogo]–cop-gen-compl→vozrastaGEN
‘He [is] already of an advanced age’.

23. Subject-attributive-completive SSyntRel. It expresses the DSyntRels I–III;

the G is a copular V, and the prototypical D is an ADJ, which semantically
bears on the subject and can agree with it.

[This task] seems–subj-attr-compl→easy. | [This task was] found–subj-attr-obj→easy.

[This task] seems–subj-attr-compl→to [be easy.]
[This task was] found–subj-attr-compl→to [be easy.]

French

[Le problème(masc)] semble–subj-attr-compl→intéressant+ØMASC

‘The problem seems interesting’. ~
[La tâche(fem)] semble–subj-attr-compl→intéressant+eFEM
‘The task seems interesting’.
[Il] s’appelle–subj-attr-compl→Jean
lit. ‘He calls himself Jean’. = ‘His name is Jean’.
Élu–subj-attr-compl→directeurIII, [Alain partit en Inde]
lit. ‘Elected director, Alain left for India’.
Élue–subj-attr-compl→directriceIII, [Helen partit en Inde]
lit. ‘Elected director, Helen left for India’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 65

24. Object-attributive-completive SSyntRel. It expresses the DSyntRel III; the G

is a V of a particular semantic class, and the prototypical D is an ADJ or
an as→ADJ phrase, which semantically bears on the DirO (and agrees
with it in languages having agreement).

consider–[him]–obj-attr-compl→happy ~
consider–[him]–obj-attr-compl→to [be happy]
consider–[him a]–obj-attr-compl→fool ~
consider–[him]–obj-attr-compl→to [be a fool]
believe–[him]–obj-attr-compl→to [be dumb] ~
believe–[him]–obj-attr-compl→dumb
?

[John] finds–[this task]–obj-attr-compl→easy.

make–[it]–obj-attr-compl→possible [to neutralize the consequences]
judge–[him]–obj-attr-compl→guilty; [They] want–[him]–obj-attr-compl→in [jail.]
identify–[this element]–obj-attr-compl→as [vital / a suffix]

French

[Je] trouve–[le problème]–obj-attr-compl→intéressant+Ø

‘I find the problem interesting’. ~
[Je] trouve–[la tâche]–obj-attr-compl→intéressant+e
‘I find the task interesting’.

— The object-attributive complement D can be a participle in an Accusativus

cum participio construction, see footnote 2, p. 61):

[I] heard–[them]–obj-attr-compl→stomping [out of the cabin.]

Latin

[Nemo] audiebat–[eumSG.ACC]–obj-attr-compl→querentemSG.ACC ⟨–[eosPL.ACC]–obj-attr-­

compl→queren­tesPL.ACC⟩ lit. ‘Nobody heard him complaining ⟨them complaining⟩’.

Ancient Greek

[Hē gunḕ] eporãi–[minMASC.SG.ACC]–obj-attr-compl→exióntaMASC.SG.ACC

‘The woman saw him going.out’.

Finnish

Pekka kuuli–[junanSG.GEN]–obj-attr-compl→saapuvanSG.GEN
lit. ‘Pekka heard train arriving’.
66 2 A general inventory of surface-syntactic relations in the world’s languages

— The object-attributive complement D can also be a subordinate clause intro-

duced by a relative pronoun (see Chapter 6, Subsection 6.2.2, p. 241, (4c)) :
Italian
obj-attr-compl
Ho visto→Chiara che usciva dal cinema
lit. ‘I.have seen Chiara who was.going out of the movie theater’. =
‘I saw Chiara go out 〈going out〉 of the movie theater’.
obj-attr-compl
L’ho vista che usciva dal cinema
lit. ‘Her I.have seen who was.going out of the movie theater’. =
‘I saw her go out 〈going out〉 of the movie theater’.

25. Predicate-attributive-completive SSyntRel. It expresses the DSyntRel II or III;

the G is a V having a syntactic feature «pred-attr», and the D is an ADJ,
which semantical­ly bears on the predicate itself.

smell–pred-attr-compl→good; feel–pred-attr-compl→miserable
playing–pred-attr-compl→small; [to] win–pred-attr-compl→big

26. Comparative-objectival SSyntRel. It expresses the DSyntRel II; the G is a

comparative word, and the prototypical D is a CONJ(compar) (which intro-
duces the object of comparison: the comparand N).

This SSyntRel describes all cases of comparison: ‘X is more/less L than Y’ and ‘X

is as L as Y’.
more–[important]–compar-obj→than [Peter]; older–compar-obj→than [Peter]
as–[important]–compar-obj→as [Peter]
[Beliefs are] so–[important]–compar-obj→as [to have people been killed for them.]
[John loves Mary] more–compar-obj→than [Peter.]
Russian
[byl] silʹnee–compar-obj→IvanaGEN
lit. ‘was stronger of.Ivan’ [= ‘was stronger than Ivan’] ~
[byl] silʹnee–compar-obj→čem [IvanNOM] ‘was stronger than Ivan’ ~
bolee–[silʹnyj]–compar-obj→čem [IvanNOM] lit. ‘more strong than Ivan’

I.1.2 Non-valence-controlled SSyntRels: 27–45

I.1.2.1 Actantial SSyntRels: 27–28

The following two SSyntRels represent the situation mentioned above, I.1.1, p. 47:
SSynt-actants that are not valence-controlled.

27. Dative-ethical-objectival SSyntRel. It expresses the fictitious lexeme «dat_

eth»; the G is a V, and the D is a PRON(pers)DAT—as a rule, of the 1SG or 2SG.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 67

German

Lieb’–dat-eth-obj→mir [nur keinen Hippy!]

lit. ‘Love to.me only no hippie!’ = ‘Don’t you love a hippie on me!’

French
[Marie] te←[m’]–dat-eth-obj–a [donné une de ces gifles !]
lit. ‘Mary to.you to.me has given one of those slaps.in.the.face!’ =
‘Mary gave me such a bloody slap in the face!’
Bulgarian
[Ex, da] ti←dat-eth-obj–pipna [az mitnica!]
lit. ‘Oh, that to.youSG I.seize I customs!’ =
‘Oh, if only I could become the master of the customs!’

— The Ethical Dative can be a clitic form of the masculine substitute pronoun of
3SG:

Bulgarian
[Ja] mu←dat-eth-obj–udariIMPER [edna rakija!]
lit. ‘It to.him hit one vodka!’ = ‘Have one vodka!’
NB Ja ‘sheFEM.SG.ACC’ is a resumptive clitic repeating the DirO rakija(fem) ‘vodka’.

— Two Ethical Datives are possible in the same clause, at least, in Romanian:

Romanian
dat-eth-obj
[Luând pe băiat de urechi] mi ţi←dat-eth-obj–[-l]–bătea
lit. ‘Grabbing to boy by ears, [he] me youSG him beat.up’.
These two cooccurring Ethical Datives violate Criterion C3 of syntactic depen-
dency; is it because of the pragmatically charged character of the Dependents?

28. Modal-objectival SSyntRel. It expresses the fictitious lexeme «should»,

«will_be» or «be_able»; the G is a VINF, and the D is an NDAT.
Russian
Mne←mod-obj–ostatʹsja? lit. ‘To.me to.stay?’ = ‘Should I stay?’ [«should»].
Emu←mod-obj–[by]–obratitʹsja [k vraču] lit. ‘To.him to.turn to doctor’. =
‘He should see a doctor’ [«should»].
[Nu,] bytʹ–mod-obj→skandalu! lit. ‘Well, to.be to.a.scandal!’ =
‘Well, there will be a scandal!’ [«will_be»].
Tebe←mod-obj–[ètogo bylo ne]–ponjatʹ lit. ‘To.you this was not to.understand’. =
‘You couldn’t understand this’ [«be_able»].
68 2 A general inventory of surface-syntactic relations in the world’s languages

I.1.2.2 Copredicative SSyntRels: 29–32

A copredicative complement is a non-valence-controlled non-actantial depen-

dent of a verb that semantically bears on an actant of this verb.

29. Subject-copredicative SSyntRel. It expresses the fictitious lexeme «be»:

L1←I–L(V)–ATTR→«be»–I→L1
–II→L2(ADJ)

The G is a V with the corresponding syntactic feature, and the prototypical D is an

ADJ, which semantically bears on this V’s subject (here, L1).

[John] returned–subj-copr→rich. |
[John] returned–subj-copr→in [a new uniform.]
[John] arrived–subj-copr→third. |
[Visitors] returned–[fervent]–subj-copr→admirers [of Mao.]
[They] parted–subj-copr→enemies.
[The fighting] continued–subj-copr→unabated.
[John] served–[Mary the salad]–subj-copr→undressed
[‘John was undressed’] (Wechsler 1995: 93).

Russian
[Ja] vstretil–[Mariju]–subj-copr→starikomINSTR ‘I [male] met Maria an.old.man’.

30. Object-copredicative SSyntRel. It expresses the fictitious lexeme «BE»:

L1←II–L(V)–ATTR→«BE»–I→L1
–II→L2(ADJ)

The G is a V with the corresponding syntactic feature, and the prototypical D is an

ADJ, which semantically bears on this V’s direct object (here, L1).

[They] sent–[John home]–obj-copr→rich. | [They] buried–[Mary]–obj-copr→alive.

[John] served–[Mary the salad]–obj-copr→undressed [‘the salad was undressed’]
(Wechsler 1995: 93).

Russian
[Ja] vstretil–[Mariju]–obj-copr→staruxojINSTR ‘I [male] met Maria an.old.woman’.
[Marija] vstretila–[menja]–obj-copr→starikomINSTR ‘Maria met me [male] an.old.
man’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 69

31. Object-resultative-copredicative SSyntRel. It expresses the fictitious lexeme

«result_in»:
L(V)–ATTR→«result_in»–II→L1(ADJ)

The G is a V with the corresponding syntactic feature, and the D is an ADJ, which
semantically bears on this V’s DirO.3 (See Goldberg & Jackendoff 2004.)

wash–[the floor]–obj-result-copr→clean
wash–obj-result-copr→clean [the inside of the cup]
hammer–[the box]–obj-result-copr→flat
beat–[the prisoner]–obj-result-copr→dead
push–[the door]–obj-result-copr→open

32. Floating-copredicative SSyntRel. It has no correspondent in the DSyntS,

but is intro­duced by DSynt-to-SSynt-structure rules; the G is a V, and the
D is a pronom­inal quantifier adjective such as all or each.

[Such sentences] all←float-copr–contain [a negative word.]

NB Cf. All←determ–[such]–sentences [contain a negative word.]
[John and Mary] both←float-copr–were [expelled.] ~
[John and Mary] were–[expelled]–float-copr→both.
[These phonemes] have–[two allophones]–float-copr→each.
NB Cf. Terrier dogs closely resemble–dir-obj→each other; ˹EACH OTHER˺ is an idiom (see No. 110,
p. 106)

A language that has a rich syntactic morphology may necessitate several floating-­
copredicative SSyntRels, as, for instance, Russian. In this language, a floating
quantifier can bear semantically on the SyntSubj, the DirO or the IndirO and
agrees with its “antecedent” in gender and case:

Russian

(10) a. [RabotnicySyntSubj] javljalisʹ–[v kontoru]–subject-float-copr→každ+aja v svoj denʹ

lit. ‘Female.workers were.coming to the.office each on her day’.
b. [RabotnicDirObj] posylali–[v kontoru]–dir-object-float-copr→každ+uju v svoj denʹ
lit. ‘Female.workers were.coming to the.office each on her day’.
c. [RabotnicamIndirObj velili] javitʹsja–[v kontoru]–indir-obj-float-copr→každ+oj
v svoj denʹ
lit. ‘To.female.workers «they» ordered to.come to the.office each on her day’.

3 There are cases of subject-resultative-copredicative SSyntRel: He froze stiff ⟨The pond froze
solid⟩. However, I do not know to what extent they are widespread.
70 2 A general inventory of surface-syntactic relations in the world’s languages

I.1.2.3 Circumstantial SSyntRels: 33–38

33. Circumstantial SSyntRel. It expresses the DSyntRel ATTR and, in many

cases, a fictitious lexeme that expresses the type of the circumstantial
(temporal, locative, instrumental, of purpose, etc.); the prototypical G is
a V (not necessarily a VFIN); the prototypical D is an ADV, a VING, a PREP or
a CONJ(subord).

Let it be emphasized that a circumstantial can (and often does) depend on a

noun, see the remark at the very beginning of I.2.2.3, p. 87.

walk–circum→fast; delve–circum→deeply
[John] works–circum→there ⟨in [this office]⟩.
[John] works–circum→abroad ⟨in [Germany]⟩.
[Don’t] waste–[time]–circum→playing [computer games!]
Having←circum–[rushed off, John]–forgot [his umbrella.]
[John] works–circum→with [several assistants.]
When←circum–[summer approaches,]–start [preparing your car.]
[Mary] received–[John]–circum→as [a queen.]
[Mary] received–[John]–circum→as [a king.]
[Mary] sang–circum→˹as if ˺ [she knew me.]
[Sometimes animals] act–circum→like [us.]
[He will] write–[next]–circum→Ø(prepos)
temp
[week.]
[He will] write–circum→tomorrow.
[A new store] opened–[three]–circum→miles–circum→West [from here.]
[John] kissed–[Mary three]–circum→times.
Had←circum–[John been here, he]–could [have helped us.]
˹Holidays←circum–[or no holidays˺, I]–have [to finish my paper.]
NB We see here a syntactic idiom ˹X or no X˺ ‘no matter whether there is X or no X’.

French

˹Une fois˺←circum–[son travail terminé, Jean]–devra [retourner à Nice]

lit. ‘Once his work being.over, Jean will.have to.return to Nice’.

[Jean] travaille–circum→Ø(prepos)
LOC
[Place de la Nation] ‘John works at Place de la Nation’.
NB The zero prepositionØ(LOC
prepos)
appears with the names of streets, squares, etc. (Mel’čuk 2018b).
 2.5 An inventory of surface-syntactic relations found in the world’s languages 71

[Jean] mange–circum→beaucoup [et voracement] ‘Jean eats a lot and greedily’.

NB But [Jean] mange–dir-obj→beaucoup [de fruits] ‘Jean eats a lot of fruit’. Cf.:
(i) On le fait manger beaucoup ‘They make him eat a lot’.
vs. (ii) On lui fait manger beaucoup de fruits ‘They make him eat a lot of fruit’.
In the French causative construction faire ‘make’ + V, the Causee must be a DirO (in the
accusative, if a clitic) if V has no DirO, and an IndirO (in the dative, if a clitic) if V has a DirO.
The example shows that the adverb beaucoup ‘a lot’ is not a DirO, but a Circumstantial,
while the noun phrase beaucoup de Ns is a genuine DirO.

— A circumstantial can constitute an absolute construction:

[John] went–[out, his]–circum→gun [in his left hand.]

With←circum–[her paper finished, Helen]–can [afford this trip.]
[The sellers] offered–[500 tons,]–circum→delivery [to be made in October.]

Latin (Ablativus absolutus)

[Mortuo] CaesareABL←circum–[bella civilia orta]–sunt

‘With Caesar dead, civil wars have started’.

— A circumstantial can be an N marked by a semantically full case (boxed in

the examples below) that encodes its role; in the DSynt-structure, the cor-
responding semanteme is expressed by a fictitious lexeme:

Russian, circumstantial of duration

[Ivan] rabotal–[celuju]–circum→nedelju ACC ‘Ivan worked the whole week’.

Russian, circumstantial of time

[Ivan] rabotal–circum→pjatogo GEN [aprelja] ‘Ivan worked on.the.fifth of.April’.

[Ivan] rabotal–[rannim]–circum→utrom INSTR ‘Ivan worked early morning’.

Russian, circumstantial of instrument

[Ivan] rešil–[zadaču obyčnym]–circum→metodom INSTR

‘Ivan solved the.problem by.ordinary method’.

Ancient Greek, circumstantial of relation

Athēnaῖos–[tó]–circum→génos ACC ‘Athenian with. respect.to [= ‘by’] birth’

Kámnō–[toùs]–circum→ophtalmoús ACC ‘I.suffer with.respect.to [= ‘from’] the eyes’.
72 2 A general inventory of surface-syntactic relations in the world’s languages

— The circumstantial can be a cognate object:

Russian

[On] umer–[užasnoj]–circum→smertʹju INSTR lit. ‘He died by.a.terrible death[cognate object]’.

NB In the DSyntS, a cognate object corresponds to a circumstantial of manner (‘[died] in a terrible

way’).

Arabic

Dafaʕa–[al+walad+aACC]–circum→dafʕаt+a ACC +nINDEF [kabīrаt+aACC+nINDEF]

lit. ‘S/He.pushed the.boy a.push big’. = ‘S/He pushed the boy hard’.

— The circumstantial can be a VINF of purpose (with or without to):

To←circum–[simplify the procedure, Dr. Copulatti]–has [recourse to the following

technique.]

Russian

[On] uexal–[v Kanadu]–circum→učitʹsjaINF ‘He went to Canada to.study’.

— The G of a circumstantial can be a noun:

[Experts share their] lessons–circum→from [the last year.]

[John wants to buy a] lodge–circum→at [the lake.]

34. Durative-circumstantial SSyntRel. It expresses the DSyntRel ATTR and the

fictitious lexeme «during»; the prototypical G is a V (not necessarily a
VFIN); the D is a noun that denotes a period of time.

[John] worked–[three]–durative-circum→days.
[John] worked–[the whole]–durative-circum→year.

35. Distance-circumstantial SSyntRel. It expresses the DSyntRel ATTR and the

fictitious lexeme «distance»; the G is a V (not necessarily a VFIN) whose
meaning includes the component ‘directed motion’; the D is a noun that
denotes a distance.

[John] walked–[three]–dist-circum→miles.
[John] walked–[four]–dist-circum→blocks [down High Street.]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 73

The SSyntRels 36–38 are of circumstantial type, but their governor is neces-
sarily a VFIN. Their triple distinction is parallel to the distinction between the three
adnominal SSyntRels:

 modifier-circumstantial ~ modificative
[Amazingly] successful,←mod-circum–[his solution]–became [generally accepted.] ~
[his] successful←modif–solution

 apposition-circumstantial ~ appositive
[An old] man,←appos-circum–[the officer]–told [us …] ~
[The] officer,–[an old]–appos→man, [told us …]

 attribute-circumstantial ~ attributive
Abroad,←attrib-circum–[an American]–is [always preoccupied.] ~
[An] American–attrib→abroad [is always preoccupied.]

36. Modifier-circumstantial SSyntRel. It expresses the DSyntRel ATTR; the G is

a VFIN, and the D is an ADJ/ADV.

[As always] elegant,←mod-circum–[John]–walked [away.]

[As always] elegantly,←mod-circum–[John]–walked [away.]

37. Apposition-circumstantial SSyntRel. It expresses the DSyntRel ATTR; the G

is а VFIN, and the D is an N.

[An old] man,←appos-circum–[John]–works [less.]

 ttribute-circumstantial SSyntRel. It expresses the DSyntRel ATTR; the G is

38. A
а VFIN, and the prototypical D is an ADV or a PREP.

Abroad,← attr-circum–[John]–works [less.]

Without←attr-circum–[his computer, John]–feels [lost.]

I.1.2.4 Extra-structural SSyntRels: 39–46

39. Parenthetical SSyntRel. It expresses the DSyntRel APPEND; the G is a VFIN,

and the D is the head of the parenthetical expression and has the corre-
sponding syntactic feature.

Oddly,←parenth–[John]–works [less.]
[John,] oddly,←parenth–works [less.]
74 2 A general inventory of surface-syntactic relations in the world’s languages

[John] works–[less,]–parenth→oddly.
[John,] naturally,←parenth–accepted [the offer.]
[John] accepted–[the offer,]–parenth→naturally.
NB Cf. [John] accepted–[the offer quite]–circum→naturally [‘in a natural manner’]. Here the
adverb naturally is subordinated to the Main Verb in the DSyntS by the ATTR DSyntRel.
As←parenth–[we have known for some time, John]–works [less.]
To←parenth–[give an example, I]–will [consider nominal suffixes.]
[It] was,–parenth→as [John said, a very hot day.]
In_general←parenth–[John]–is [happy.] | [John]–is,–parenth→in_general, [happy.]
NB The underscoring of a space (“_”) between two words means that they form in fact “one
word”—that is, in spite of its official spelling as two words, in_general is actually one wordform,
since its internal structure does not correspond to syntactic rules of English: there is no
*PREP→ADJ phrase.

40. Quasi-parenthetical SSyntRel. It expresses DSyntRel APPEND; the G is the

head of a Direct Speech expression, and the D is a VFIN, the head of a
Direct-Speech Introductor.

[“Alan] will–[visit us,” John]–quasi-parenth→shouted, [“next Friday.”]

[“I] am–[not going there!”,]–quasi-parenth→shouted [John.]
[“I] am–[not going to kill the project,” McGuire]–quasi-parenth→declared [in front
of cameras.]

Russian

[“Kak vy] smeete!”–quasi-parenth→vspyxnula [Elena] lit. ‘«How you dare!»—

flared.up Elena’.
NB Many verbs that appear as the head of a subordinated parenthetical Direct-Speech Introduc-
tor are impossible in the superordinated Direct-Speech Introductor:

French

« C’est un secret ! », élude Isabela Ono ‘«This is a secret! », eludes Isabela Ono’.
vs.
*Isabela Ono élude: « C’est un secret ! » ‘Isabella Ono eludes: «This is a secret! »’
(see Danlos et al. 2010).

41. Adjunctive SSyntRel. It expresses the DSyntRel APPEND; the G is a VFIN, and
the prototypical D is an interjection.

OK,←adjunct–[John]–will [go.] | Wow,←adjunct–isn’t [she stunning!]

 2.5 An inventory of surface-syntactic relations found in the world’s languages 75

42. Proleptive SSyntRel. It expresses the DSyntRel APPEND; the G is a VFIN,

and the D is an N that is, as a rule, the expression of a Focalized Theme or
Rheme. (For the proleptive SSyntRel in Russian, see a detailed analysis in
Sannikov 2010a.)

[This] Collins,←prolept–[we]–hate [him.] | [This] film,←prolept–[I]–find [it gorgeous.]

Prolepses are quite typical of French and of many South-East Asian languages.

French
prolept
[Ma] mère,←prolept–[mes amis, elle les]–adore
lit. ‘My mom, my friends, she them adores’.
prolept
Partir, [c’est] mourir [un peu] lit. ‘Leave, it is die a bit’.

Korean

(12) John+i←prolept–[kho +ka]—kil +ta

SUBJ nose SUBJ be.long DECL(arative)
‘[It is] John [whose] nose is long’.

43. A
 ddressative SSyntRel. It expresses DSyntRel ADDRESS; the G is a VFIN, and
the D is an N that is the designation of the entity to which the utterance is
addressed; quite often, it is a proper name.

Mary,←address–[where]–are [you?]

44. P
 resentative SSyntRel. It expresses the DSyntRel APPEND; the G is a VFIN,
and the D is the particle èto ≈ ‘this, it’.

Russian

Èto←present–[Vanya tam]–sobiraetsja lit. ‘This Vanya there is.packing’

[for instance, as an answer to the question What is happening? or What is this?].
NB Cf. Èto←restr–Vanya [tam sobiraetsja] lit. ‘This [is] Vanya [who] there is.packing’ (see No. 47).

Èto←present–stučit [doždʹ po kryše] lit. ‘This is.drumming rain on roof’.

45. Auxiliary-conjunctive SSyntRel. It expresses the DSyntRel APPEND; the G

is a VFIN, and the D is a coordinating conjunction that marks the begin-
ning of a sentence.

And←auxiliary-conjunct–[everyone]–burst [out laughing.]

But←auxiliary-conjunct–[everyone]–disagreed.
76 2 A general inventory of surface-syntactic relations in the world’s languages

Russian
A←auxiliary-conjunct–[zori zdesʹ]–ØBYTʹ [tixie] lit. ‘And dawns here [are] quiet’. =
‘The dawns here are quiet’ [a known 1972 Soviet film].
No←auxiliary-conjunct–[esli Ivan ušël, nado]–budet [ždatʹ ego vozvraščenija]
lit. ‘But if Ivan has.left, necessary will.be wait [for] his return’.

46. Auxiliary SSyntRel. It does not correspond to any DSynt-relation, but is

introduced by DSyntS-to-SSyntS rules in different cases specific to differ-
ent languages.

Here are two examples.

— In a language with subordinating binary conjunctions, when expanding the

DSynt-node of a binary con­junction, the auxiliary SSyntRel subordinates the
second component of the binary conjunction—a specific particle that marks
the beginning of the superordinate clause—to the MV of this clause.
circumstantial
[If John gets the job,] then←auxil–[he]–will [stay in town.]

Russian
circumstantial
[Tolʹko ja priotkryl dverʹ,] kak←auxil–poslyšalsja [tixij golos]
lit. ‘As.soon I slightly.opened door as was.heard low voice’.
For binary conjunctions, see Chapter 7, p. 275ff.

— In a language that has a lexemic nominalizer, i.e. an “auxiliary” noun that

is only used for converting a clause C into a nominal, the auxiliary SSyntRel
subordinates C to this noun (under­scored in the example):

Korean (see Chapter 6, p. 254, (17b))

(13) Nay+ka paŋ + ɨl č huŋsoha+nɨn←auxil–kes +ɨl towačwuseyyo

I SUBJ room ACC clean.up PARTICIPLE «thing» ACC help.IMPER
lit. ‘I room cleaning.up «thing» help!’ = ‘Help me clean up the room!’

I.2 Phrase-Level SSyntRels: 47–113

I.2.1 Any type of phrase SSyntRel, non-valence-controlled: 47

47. R
 estrictive SSyntRel. It expresses the DSyntRel ATTR; the G is any lexeme,
and the D is a particle.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 77

still←restr–taller; [is] still←restr–here; not←restr–here; not←restr–me

so←restr–rich; too←restr–tired; that←restr–far; boys–restr→only
not←restr–only←restr–me
[John] just←restr–arrived.

Russian

my–restr→že ≈ ‘but we’/‘as for us’ (že is a clitic particle that expresses contrast)

French

Ne←restr–[me]–quitte–restr→pas ! [J. Brel] lit. ‘No me leave not!’

[Je] ne←restr–dors–restr→pas ‘I am not sleeping’.
[Je] ne←restr–lisais [alors] que←restr–[des]–polars ‘I read then only whodunits’.

— In particular, the restrictive SSyntRel subordinates the first component of a

coordinative binary or repeated conjunction (see Chapter 7, Section 7.3, p. 280):
[I liked] both←restrictive–[the]–movie[–coord→and–[the]–conj-coord→play.]

Russian

[Ivan ljubit] i←restr–mjaso[,–coord→i–conj-coord→rybu,–coord→i–conj-coord→

ovošči] lit. ‘Ivan likes and meat, and fish, and vegetables’.
NB The linear position of a restrictive particle is controlled either by its syntactic feature
(«antepos» or «postpos») or by the communicative structure.

I.2.2 Noun phrase SSyntRels: 48–83

I.2.2.1 Valence-controlled SSyntRels: 48–59

For more details on SSyntRels Nos. 48–53, see Chapter 5, p. 205ff.

48. Subjectival-adnominal-completive SSyntRel. It expresses the DSyntRel I; the

G is an N, and the prototypical D is a phrase PREP(subj-adnom)→N or an N in
an oblique case (in English, the depending N is in the possessive form).

The subjectival adnominal complement is an adnominal transform of the subject;

in a language having case, it is marked on the surface, as a rule, by the genitive
and is known as Genitivus Subjectivus.

[a] translation–subj-adnom→by [McGuire]; objections–subj-adnom→by [the minister]

[a] translation–subj-adnom→of [McGuire: ‘McGuire translated the piece’]
McGuire’s←subj-adnom–translation [‘McGuire translated the piece’]
78 2 A general inventory of surface-syntactic relations in the world’s languages

arrival–subj-adnom→of [the President]

shooting–subj-adnom→of [the hunters: ‘the hunters shoot’]
NB 1. [a] translation–obj-adnom→of [McGuire: ‘Somebody translated a text by McGuire’] is also
pos­sible. However, *a translation of Verlaine of McGuire is ungrammatical; the correct
expres­sion is [a] translation of Verlaine by McGuire.
2. McGuire’s←possessive–translation; see No. 60, p. 83.
tons–subj-adnom→of [debris] (‘ton–1→debris’: a ton of debris is debris)
hundreds–subj-adnom→of [books] (‘hundred–1→books’)
walls–subj-adnom→of [the building] (‘walls–1→building’)
John’s←subj-adnom–heart (‘heart–1→John’)

Russian

priezd–subj-adnom→Ivana ‘arrival of.Ivan’

mečty–subj-adnom→Ivana ‘dreams of.Ivan’
kuča–subj-adnom→peska ‘pile of.sand’ (‘pile–1→sand’)
sistema–subj-adnom→padežej ‘system of.cases’ (‘system–1→cases’)

French

effet–subj-adnom→de [ses actions] ‘effect of his actions’

rien–subj-adnom→d’[intéressant] lit. ‘nothing of interesting’

— In some languages the subj-adnom SSyntRel describes the phrases of the form
“ADV(quantitative) + N”:
Rus. mnogo–subj-adnom→knigPL.GEN lit. ‘much of.books’
Fr. beaucoup–subj-adnom→de [livres] lit. ‘much of books’

49. Objectival-adnominal-completive SSyntRel. It expresses the DSyntRel II; the

G is an N, and the D is a PREP(attr)→N phrase or an N in an oblique case (in
English, the depending N is in the possessive form).

The objectival adnominal complement is, roughly, an adnominal transform of the

direct object; in a lan­guage having case, it is marked, as a rule, by the genitive
and is known as Genitivus Objectivus.
shooting–obj-adnom→of [the hunters: ‘the hunters are shot at’]
[John’s] description–obj-adnom→of [Alan] ~
*[Alan’s] description–obj-adnom→of [John] [in the sense of ‘John describes Alan’].
Alan’s←obj-adnom–description[–subj-adnom→by John]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 79

Spanish
poss-adnom
subj-adnom
el retrato–obj-adnom→de Enrique VIII de Holbein del barón Thyssen
‘the portrait of Henry VIII by Holbein owned by Baron Thyssen’

Russian

ubijstvo–obj-adnom→Ivana ‘murder of.Ivan’

priëm–obj-adnom→Ivana ‘reception of.Ivan’ [= ‘Ivan is being received’]
sposob–obj-adnom→razdelenija ‘technique of.separation’
(‘dobyča←1–sposob–2→razdelenie’ = ‘extraction←1–technique–2→separation’)

50. Q
 ualificative-adnominal-attributive SSyntRel. It expresses the DSyntRel
ATTR; the G is an N1, and the D is a PREP(attr)→N2 phrase or an N2-GEN, where
N2 is a predicative noun whose Sem-actant is the G [‘N2(N1)’].

imagesN1–qual-adnom→of [superhuman beautyN2] (‘beauty–1→images’)

[a] man–qual-adnom→of [courage]; dress–qual-adnom→of [a beautiful color]
[a] manN1–[the same]–qual-adnom→ageN2 (‘age–1→man’)

Russian

tranšejaN1–[bolʹšoj]–qual-adnom→glubinyN2 lit. ‘trench of.great depth’

The SSyntRels subj-adnom, obj-adnom, act-attr (No. 56, p. 82) and qual-adnom are
distinguished from the “simple” attributive SSyntRel (No. 63, p. 85) and among
themselves because of different placement of their Ds:

Russian
qual-adnom
kuča–[morskogo]–subj-adnom→peska [ogromnogo] razmera lit. ‘pile of.sea sand
of.huge size’
vs. ?kuča ogromnogo razmera morskogo peska

French
subj-adnom
pompe–obj-adnom→à [essence] du [camion] lit. ‘pump to gas of.the truck’
vs. *pompe du camion à essence
act-attr
moulin–obj-adnom→à [café] à [piles] lit. ‘grinder for coffee with batteries’
vs. *moulin à piles à café
80 2 A general inventory of surface-syntactic relations in the world’s languages

51. Possessive-adnominal-attributive SSyntRel. It expresses the DSyntRel ATTR

and the fictitious lexeme «belong»; the G is an N1, and the D is a
PREP(attr)→N2 phrase or an N2-GEN.

Russian
sad–poss-adnom→sosedaGEN ‘garden of.neighbor’
stadion–poss-adnom→universitetaGEN ‘stadium of.University’

French
jardin–poss-adnom→du voisin ‘garden of.the neighbor’
stade–poss-adnom→de l’Université ‘stadium of the University’
Cf. the possessive SSyntRel: No. 60, p. 83.

52. Characterizing-adnominal-attributive SSyntRel. It expresses the DSyntRel

ATTR; the G is an N1, and the D is a PREP(attr)→N2 phrase or an N2-GEN.

Russian

rukopisʹ–[šestnadcatogo]–charact-adnom→vekaGEN ‘manuscript of.sixteenth century’

stol–[krasnogo]–charact-adnom→derevaGEN ‘table of.red wood’ = ‘mahogany table’

53. Metaphorical-adnominal-attributive SSyntRel. It expresses the DSyntRel

ATTR and the fictitious lexeme «as_if»; the G is an N1, and the D a
PREP(attr)→N2 phrase or an N2-GEN, N1 being a metaphor of N2.

curtainN1–metaph-adnom→of [the nightN2]

[the bitter] bread–metaph-adnom→of [exile]

Russian

zerkaloN1–metaph-adnom→ozeraN2-GEN ‘mirror of.lake’

kolonny–metaph-adnom→sosenGEN ‘columns of.pines’

54. E valuative-adnominal-attributive SSyntRel. It expresses the DSyntRel ATTR

and the fictitious lexeme «be»; the G is an N1, and the D a PREP(attr)→N2
phrase or an N2‑GEN, N1 expressing an “evaluation” of N2. (See Foolen 2004
and Polguère 2014.)

[my] husbandN2–ATTR→«be»–II→idiotN1 ⇔ [my] idiotN1–eval-adnom→of [a husbandN2]

husbandN2←I–
[a] bitch–eval-adnom→of [a problem]; [a] beast–eval-adnom→of [a night]
[a] whale–eval-adnom→of [a project]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 81

French
[votre] fils–ATTR→«être»–II→ingénieur ⇔ [votre] ingénieur–eval-adnom→de
[fils] lit. ‘your engineer of son’
[ton] pharmacien–eval-adnom→de [mari] ‘your pharmacist of husband’
[ce] bijou–eval-adnom→du [lac] ‘this jewel of lake’
German
[dieser] Schuft–eval-adnom→von [einem Hausmeister] ‘this scoundrel of a super-
intendent’

55. Modifier-attributive SSyntRel. It expresses the DSyntRel ATTR and the ficti-
tious lexeme «be»; the G is an ADJ, and the D a PREP(attr)→N phrase, ADJ
expressing a modification of N.

This construction is rather exotic: contrary to the “normal” case, where a modify-
ing ADJ syntactically depends on the modified N, here the ADJ that semantically
subordinates the modified N, syntac­tically also subordinates it by means of a
preposition; that is, a modified noun is implemented in the SSyntS as an attri-
bute of its own semantic modifier. We find this construction in French, although
only with four adjectives (chouette ‘nice’, drôle ‘strange, funny’, putain ≈
‘bloody’, and vache ‘impressive’):4
French
(14) un drôle–modif-attrib→de garçon ‘a strange boy’
a-MASC.SG strange-MASC.SG of boy[MASC]-sg
une drôle de voiture ‘a strange car’
a-FEM.SG strange-FEM.SG of car[fem]-sg
ces drôles de garçons ‘these strange boys’
this-masc.pl strange-MASC.PL of boy[masc]-pl
ces drôles de voitures ‘these strange cars’
this-FEM.PL strange-FEM.PL of car[fem]-pl

4 Here are two more examples (for a detailed analysis of the construction in question, further
examples and a bibliography, see Gaatone 1988):
un vache de garçon ‘an impressive boy’
a-masc.sg impressive-masc.sg of boy[masc]-sg
une vache de voiture ‘an impressive car’
a-fem.sg impressive-fem.sg of car[fem]-sg
ce chouette de garçon ‘this nice boy’
this-masc.sg nice-masc.sg of boy[masc]-sg
cette chouette de voiture ‘this nice car’
this-fem.sg nice-fem.sg of car[fem]-sg
82 2 A general inventory of surface-syntactic relations in the world’s languages

However, in Sardinian, this construction is fairly productive: it is possible for all

prenominal adjectives.

Sardinian (Jones 1993: 76–79)

(15) unu bette–modif-attrib→ de pittsinnu ‘a big boy’

a-MASC.SG big-MASC.SG of boy[MASC]-SG
una ruja de makkina ‘a red car’
a-FEM.SG red-FEM.SG of car[fem]-sg
sa manna de ampulla ‘the big bottle’
the-MASC.SG big-MASC.SG of bottle[fem]-sg
cudda de makkina ‘that car’
that-FEM.SG of car[fem]-sg

56. A
 ctantial-attributive SSyntRel. It expresses the DSyntRel I, II, III, … whose
depend­ent does not correspond to the subject or the direct object; the G
is an N1, and the D is a PREP(attr)→N2 phrase or an N2-GEN.

French

moteur–act-attr→à [essence] lit. ‘engine to gas’ = ‘gas←compos–engine’

(see No. 61, p. 84)
gâteau–act-attr→au [chocolat] lit. ‘cake to.the chocolate’ = ‘chocolate←compos–cake’
farine–act-attr→de [maïs] lit. ‘flour from corn’

 ctantial-appositive SSyntRel. It expresses the DSyntRel II or III; the G is

57. A
an N, and the prototypical D is an N.

Russian

vesom–[odna]–act-appos→tonnaNOM lit. ‘[having] weight one ton’

[pri] vysote–[odin]–act-appos→metrNOM lit. ‘with height one meter’
vesom–act-appos→v [odnu tonnu] lit. ‘[having] weight in one ton’
[po] cene–[tri]–act-appos→rublja–act-appos→štukaNOM lit. ‘at price three rubles piece’

French

ticket–act-appos→restaurant lit. ‘ticket restaurant’ = ‘meal←compos–voucher’

(see No. 61)
espace–act-appos→enfants lit. ‘space children’
assurance–act-appos→vie lit. ‘insurance life’
début–act-appos→mai ‘beginning May’; mai–act-appos→2016 ‘May of 2016’
 2.5 An inventory of surface-syntactic relations found in the world’s languages 83

58. Elective SSyntRel. It expresses the DSyntRel II; the G is an ADJSUPERL or a

NUM; the D is a PREP→N phrase.

[the] poorest–elect→among [peasants]; [the] best–elect→of ⟨from⟩ [these boys]

[the] most–[intelligent]–elect→of ⟨from⟩ [these boys]
one–elect→of [them]; five–elect→of [these books]
NB 1. [the] poorest–elect→of [region’s peasants] vs. [poorest–[peasants]–obl-obj-1→in [the
region] ~ these peasants, the poorest–obl-obj-1→in the region
2. [the] best–elect→of [national announcers] vs. the best national←modif–announcer ~ the
best–[announcer]–obl-obj-1→of the nation ~ this announcer, the best–obl-obj-1→of the
nation

59. Sequential SSyntRel. It does not express a DSyntRel, but links the SSynt-
“reflexes” of DSynt-actants I–III of L; the G is an N, and the D is an N.

man–sequent→machine [interactionL];5 Paris–sequent→London [flightsL]

English–sequent→German [dictionaryL]; English–sequent→to [German translationL]

I.2.2.2 Valence-controlled and non-valence-controlled SSyntRels: 60–64

60. Possessive SSyntRel. It expresses DSyntRel I, II, III, ATTR or a fictitious

lexeme; the G is an N, and the D is an N in the form that depends on the
language.

Semantically, the possessives are quite variegated; syntactically, the possessive

SSyntRel represents simply the syntactic dependence of a “possessor” on the
“possessed” noun, and the numer­ous particularities of the possessive construc-
tion are specific to the corresponding language. Some illustrations are supplied
from three languages.

— English has a special possessive form of the noun:

John’s←poss–arrival; John’s←poss–execution; John’s←poss–book
[Last] year’s←poss–wishes [are this] year’s←poss–apologies.

5 Government pattern for the noun interaction (‘X’s interaction with Y’)
X⇔I Y ⇔ II
1. N’s 1. with N
2. of N
3. between NX and NY
4. –compos→NX–sequent→NY
84 2 A general inventory of surface-syntactic relations in the world’s languages

— In Russian we find “possessive” phrases either with the prepositions u ‘at’

and k ‘to’ or with NDAT:

[Stul stoit] u←poss–[Maši v]–komnate lit. ‘Chair is at Masha in room’. ≡

[Stul stoit v] komnate–poss→u [Maši] lit. ‘Chair is in room at Masha’. =
‘…in Masha’s room’.

[Otnesi stul] k←poss–[Maše v]–komnatu! lit. ‘Carry chair to Masha into room!’ ≡
[Otnesi stul v] komnatu–poss→k [Maše!] lit. ‘Carry chair into room to Masha!’ =
‘…into Masha’s room’.

[Otnesi stul] MašeDAT←poss–[v]–komnatu! lit. ‘Carry chair to.Masha into room!’ ≡

[Otnesi stul v] komnatu–poss→Maše! lit. ‘Carry chair into room to.Masha!’ =
‘…into Masha’s room’.
NB 1. The dative case in this construction expresses the fictitious lexeme «affect», that is, the
additional meaning ‘N is affected by the action in question’. This means that either we
have to consider the dative case to be semantically full in this construction (and mark it in
the SSyntS), or to introduce another SSyntRel. For the time being, I am unable to solve this
dilemma.
2. The choice between the prepositions u vs. k is determined by the meaning of the verb on
which depends the “possessed” noun: if this verb denotes static localization, then u; if it
denotes directed movement, then k or NDAT.
3. Cf.: (i) [Ja] otnës–[stul]–indir-obj→MašeDAT ([v komnatu])
‘I carried the.chair to.Masha (into the.room)’.
(ii)  [Stul] naxoditsja–obl-obj→u Maši
‘The.chair is at Masha’s’: u NGEN can mean ‘at N’s’.

— Bulgarian uses for the “Possessor” the clitic dative form of a personal pronoun:

Gradina+ta–poss→mi ⟨ti, mu, i⟩ [e tam] ‘Garden.the to.me ⟨to.youSG, to.him, to.her⟩

is there’. = ‘My ⟨your, his, her⟩ garden is there’.

As a clitic should, the possessive dative clitic is linearly positioned after the “pos-
sessed” N or after the first wordform of the phrase:

Prekrasna+ta mi gradina e tam ‘Beautiful.the to.me garden is there’.

61. Compositive SSyntRel. It expresses the DSyntRel I, II or a fictitious lexeme;

the prototypical G is an N, and the D is an N.

man←compos–[-machine]–interaction; car←compos–repair
noun←compos–phrase
modif
[secure] smartphone←compos–shipping←compos–box
 2.5 An inventory of surface-syntactic relations found in the world’s languages 85

NB A dependent in a compositive phrase [here, shipping] that is the governor of another com-
positive depend­ent [smartphone] can accept an adjectival modifier [secure]. This is one of the
facts preventing the treat­ment of compositive phrases in English as compound words, because
in this case an internal com­ponent (shipping) of a presumed compound noun (smartphone+
shipping+box) would have its own modifier outside the compound.

fax←compos–transmission←compos–network←compos–access
color←compos–blind; tone←compos–deaf; tax←compos–free
road←compos–test [a car]; guest←compos–conduct [an orchestra]

62. Absolute-modificative SSyntRel. It expresses DSyntRel I (with head-switch-

ing and possible omission of the copula verb) or one of the fictitious
lexeme «after», «while», «with»; the G is an N, and the prototypical D
is an ADJ, including participles.

[With the Central] Bank–abs-modif→refusing [to budge, there were no ruble buyers.]
[Without] me–abs-modif→asking [her, Mary offered me help.]
[John went out, his] anger–abs-modif→gone.
John–abs-modif→being [sick, we remained with him.]
[His first] attempt–[a]–abs-modif→failure, [John decided to try again.]
[He went out, (with) his] gun–abs-modif→in [his left hand.]

French

[Le] chat–abs-modif→étant [parti, les souris se sont mises à danser]

‘The cat having left, the mice started dancing’.

— Ablativus absolutus:
Latin

(16) a. Ciceron+e—abs-modif→viv +o [bellum civile Romae erat.]

Cicero SG.ABL alive MASC.SG.ABL war civil in.Rome was
‘With Cicero alive, there was a civil war in Rome’.
b. Oppid+is(neu)—abs-modif→incens +is [exercitus signa movit.]
town PL.ABL set.alight NEU.PL.ABL army standards moved
‘With the towns set on fire, the army marched away’.

63. A ttributive SSyntRel. It expresses the DSyntRel ATTR and a fictitious
lexeme: «be», «have», «designed_for», …; the G is an N, and the proto-
typical D is a PREP(attr)→N phrase or an NGEN.

[a young] man–attr→from [Nantucket]; Detroit–attr→after [dark]
[every] path–attr→on [the island]; life–attr→abroad
86 2 A general inventory of surface-syntactic relations in the world’s languages

learners–attr→with [different backgrounds]

years–attr→of [war]

Russian

rebënok–attr→s [bolʹšim životom] ‘child with big tummy’

čelovek–attr→bez [očkov] ‘person without glasses’
Ivan–attr→s [Mašej] ‘Ivan with Masha’

French

carnet–attr→d’[étudiant] ‘student gradebook’ vs.

carnet–subj-adnom→de [l’étudiant] ‘gradebook of the student’
tronc–attr→d’[arbre] ‘tree trunk’ vs.
tronc–subj-adnom→de [l’arbre] ‘trunk of the tree’
roue–attr→de [vélo] ‘bicycle wheel’ vs.
roue–subj-adnom→d’[un vélo] ‘wheel of a bicycle’
robe–attr→de [mariée] ‘wedding dress’ vs.
robe–subj-adnom→de [cette mariée] ‘dress of this bride’
NB un carnet–attr→d’étudiant périmé lit. ‘a gradebook of student expired’ ~
un carnet–[périmé]–subj-adnom→de l’étudiant lit. ‘a gradebook expired of the student’
*un carnet périmé d’étudiant lit. ‘a gradebook expired of student’ ~ *un carnet de l’étudiant
vs. 
périmé lit. ‘a gradebook of the student expired’ 

course–attr→à [obstacles] ‘obstacle race’

— The attribute can be an infinitive introduced by a preposition with the

meaning ‘intended for’:

books–attr→to [read]

French

[un] appartement–attr→à [louer] ‘an apartment to rent’

64. D
 escriptive-attributive SSyntRel. It expresses the DSyntRel ATTRdescr and a
fictitious lexeme, for instance, «be_from»; the G is an N, and the proto-
typical D is a PREP(loc)→N phrase or an N.

[Professor] Wanner,–descr-attr→from [Stuttgart, was also present.]

[Professor] Wanner,–descr-attr→Stuttgart, [was also present.]
NB Cf. [A] professor–attr→from [Stuttgart was also present.]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 87

The semantic contrast between restrictive and non-restrictive (= descriptive, or

qualifying) modifiers is well known: a restrictive modifier restricts a set of pos-
sible referents of the governor to a narrower subset (‘the dogs that are healthy’ is
a subset of ‘dogs’), while a descriptive modifier expresses an additional charac-
terization of the elements of the same set (‘these dogs, which are healthy’ is the
same set as ‘these dogs’).
It is worthwhile to indicate the following proportionality (Mel’čuk & Pertsov
1987: 152):

modif : descr-modif : modif-circum =

attrib : descr-attrib : attrib-circum =
appos : descr-appos : appos-circum

I.2.2.3 Non-valence-controlled SSyntRels: 65–113

Along with the elements subordinated to the noun by the SSyntRels Nos. 65–113,
a noun phrase can contain non-valence-controlled circumstantials of all types—
prepositional phrases and adverbs, such as tower on Fifth Avenue, their meeting
yesterday, etc. They are covered by the circumstantial SSyntRel, No. 33, p. 70.

65. Determinative SSyntRel. It is used in two cases: either it is introduced in

the SSyntS to subordinate an article (which implements a deep gram-
meme) to a G = N; or it expresses the DSyntRel ATTR, the G being also an
N, and the D a determiner other than an article.

a←determ–bed; those←determ–beds; my←determ–bed

66. Quantitative SSyntRel. It expresses the DSyntRel ATTR; the G is an N, and

the D is a NUM(quant).

three←quant–beds; [three←num-junct–]thousand←quant–people
NB Cf. thousands–attr→of–prepositional→people (here thousand is an N).

67. Approximate-quantitative SSyntRel. It expresses the fictitious lexeme

«maybe»; the G is an N, and the D is a NUM(quant).

Russian

knig–approx-quant→dvadcatʹ ‘maybe twenty books’ ~

dvadcatʹ←quant–knig ‘twenty books’
88 2 A general inventory of surface-syntactic relations in the world’s languages

knig–[na]–approx-quant→dvadcatʹ ‘maybe for twenty books’ ~

[na] dvadcatʹ←quant–knig ‘for twenty books’

68. Ordinal SSyntRel. It expresses the DSyntRel ATTR; the G is an N, and the
D is an ADJ(ordinal).

[the] third←ordin–rank; [on the hundred forty-]third←ordin–day

69. Approximate-ordinal SSyntRel. It expresses the DSyntRel ATTR and the fic-
titious lexeme «maybe»; the G is an N, and the D is an ADJ(ordinal).
Russian

denʹ–[na]–approx-ord→šestoj lit. ‘day on sixth’ = ‘maybe on the sixth day’ ~

[na] šestoj←ordin–denʹ ‘on the sixth day’

 odificative SSyntRel. It expresses the DSyntRel ATTR; the prototypical G

70. M
is an N, and the D is an ADJ.

The modificative SSyntRel covers the most typical and semantically neutral adjec-
tival modifi­cation. The linear position of the ADJ with respect to the N it modifies
is controlled by general syntactic rules of the language, the type of the ADJ (ante-
posed/postposed), the type of the N (e.g., “genuine” N vs. nominal pronoun), the
phraseological character of the ADJ, etc. However, in some cases, the position of the
ADJ expresses a meaning, thus creating a different SSyntRel, which semantically
contrasts with the modificative SSyntRel: the special-modificative SSyntRel (No. 71).

comfortable←modif–beds; visible←modif–stars; French←modif–production

nothing–modif→interesting; [a] house–modif→ablaze
secretary–modif→general, notary–modif→public, God–modif→Almighty,
knight–modif→errant (these examples represent phraseologized expressions,
namely collocations).
NB As is the case with the name of DSynt-relation ATTR (see Subsection 2.2.1, p. 34), which is a
convenient abbreviation of ATTRrestr, the name of SSyntRel modificative is an abbreviation for
restrictive-modific­ative; it is opposed to descriptive-modificative, No. 72 below.

French

dernier←modif–jour ‘last day’; message–modif→inattendu ‘unexpected message’

personne–modif→curieuse1 ‘curious/indiscreet person’
curieuse2←modif–personne ‘strange/bizarre person’
˹Jugement–modif→Dernier˺ ‘Last Judgment’
 2.5 An inventory of surface-syntactic relations found in the world’s languages 89

Russian

papa–modif→rimskij lit. ‘Pope Roman’

˹Mama–modif→ródnaja!˺ [interjection idiom] lit. ‘Mom natural!’ = ‘Goodness!’

— An ADJ can modify another ADJ:

burning←modif–hot, icy←modif–cold, dark←modif–green, light←modif–green

Russian

takoj←modif–milyj lit. ‘such nice’ = ‘so nice’

˹takoj←[že]˺–modif–milyj lit. ‘such že nice’ = ‘equally nice’
˹tot←[že]˺–modif–samyj lit. ‘that že same’ = ‘the same’

71. Special-modificative SSyntRel. It expresses DSyntRel ATTR plus a semantic

addition, which depends on the language and is expressed in the DSyntS
by a fictitious lexeme; the G is an N, and the D is an ADJ.

[All] stars–spec-modif→visible [are named after famous astronomers.] vs.

[All] visible←modif–stars [are named after famous astronomers.]

[Every] cent–spec-modif→available [was put into the project.] vs.

[Every] available←modif–cent [was put into the project.]

In English, special-modifying adjectives are postposed; they express “ephem-

eral,” temporary properties; in French, special-modifying adjectives (they are
anteposed) express subjective, emo­tional evaluation; in Russian, special-modi-
fying adjectives (postposed) express terminological, rather than qualifying, char-
acter of the expression (tigr sibirskij lit. ‘tiger Siberian’); etc.

 escriptive-modificative SSyntRel. It expresses the DSyntRel ATTRdescr; the

72. D
G is an N, and the D is an ADJ.

[these] beds,–descr-modif→comfortable [and not expensive], ...

[There she met] John,–descr-modif→tired [of loneliness.]

73. Relative SSyntRel. It expresses the DSyntRel ATTR; the prototypical G is an

N, and the D is a VFIN, the head of a relative clause (see Chapter 6, pp. 235ff ).
90 2 A general inventory of surface-syntactic relations in the world’s languages

The reasons to have the relative SSyntRel different from the general modificative
SSyntRel is the fact that the modificative SSyntRel is unlimitedly repeatable, while
the relative SSyntRel is non-repeat­able.
NB In fact, the name relative is an abbreviation for restrictive-relative, in the same way and for the
same reason as the name of the modificative SSyntRel is an abbreviation for restrictive-modificative.

[the] paper–[that I]–rel→read [yesterday]

[the] paper–[I]–rel→read [yesterday]
[the] girl–[who]–rel→came [first]
[the] country–[where I]–rel→could [live]
[the] country–[I]–rel→could [live–obl-obj→in]

— The G can be a nominal correlative pronoun:

Russian (the pronoun totII.1 ‘that.one’; for totII.2, see below, SSyntRel 112, pp. 107–108)
[Vernëmsja k] tomu–[, o čëm my]–rel→govorili
lit. ‘Let’s.return to that about what we.were.talking’.
[Pogovori s] temi–[, komu ty]–rel→posylal [pisʹmo]
lit. ‘Talk to those to.whom you have.sent the.letter’.
nastolʹko–[prošče, naskolʹko èto]–rel→bylo [vozmožno]
lit. ‘so simpler as.much.as it was possible’ = ‘simpler to the extent that it was possible’
~ [prošče] nastolʹko–[, naskolʹko èto]–rel→bylo [vozmožno]

— The G can be an ADJ:

Spanish
[¡Lo] hermosas–[que]–rel→son [esas chicas!]
lit. ‘The beautiful which are these girls!’ = ‘How beautiful are these girls!’

74. Descriptive-relative SSyntRel. It expresses the DSyntRel ATTRdescr; the G is

(proto­typically) an N, and the D is a VFIN.
NB The descriptive-relative clauses that depend not on a noun (= sentential relative clauses,
Quirk et al. 1991: 1118–1120) are considered by many as a special syntactic construction
different for the relative clause proper; this construction is called the explanatory clause. In
my previous publications (e.g., Mel’čuk 2016: 194) sentential relatives were treated as a
subtype of the coordinate clause (they were described as depending on the MV of the super-
ordinate clause by the explanatory-coordinative SSynt­Rel). However, sentential relat­ives are
covered by the definition of relative clause proposed in this volume (Chapter 6, Definition
6.2, p. 240), and I do not see reasons that would force me to change this definition.

[Тhis] paper–[, which I]–descr-rel→read [yesterday, seems interesting.]

John–[, who]–descr-rel→loves [Mary so much, should return.]
[Mary] gave–[me a smile, which]–descr-rel→was [nice.]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 91

Russian

[Ivana sčitali] lenivym,–[ čto mne]–descr-rel→kazalosʹ [ošibkoj]

lit. ‘Ivan was.considered lazy, which to.me seemed a.mistake’.
[Ivana sčitali] lenivym,–[kakovym on, odnako ne]–descr-rel→byl
lit. ‘Ivan was.considered lazy, which he, however, was not’.
[Ivanu poručili] gotovitʹ,–[čem on]–descr-rel→ljubil [zanimatʹsja]
lit. ‘Ivan was.entrusted with.cooking, which he liked to.do’.

75. WH-pronominal SSyntRel. It expresses the DSyntRel ATTR; the G is a

PRON(rel), and the D is a VFIN from a small lexical set.

[John disappeared God] knows←WH-pronominal–where.

[John does you] will←WH-pronominal–[never guess]–what.

Russian

[Ja budu rabotatʹ] gde–[vam]–WH-pronominal→budet [ugodno]

‘I will work where to.you [it] will.be pleasant’.

The WH-pronominal SSyntRel describes an open set of expressions equivalent to

indefinite pronouns, such as somewhere or whatever; see Chapter 6, 6.3.2.2,
p. 249, and Chapter 9, 9.2.1.1.1. p. 312.
NB Compound indefinite pronouns where the element depending on the L(pron, rel) is not a non-
phraseologiz­ed clause but a fixed lexical unit from a very limited set—such as Russian aby←kto
‘no.matter who’ or kto→-to ‘somebody’—are described by the intraphrasemic SSyntRel, No. 114,
p. 109.

76. Specifying-appositive SSyntRel. It expresses the DSyntRel ATTR and the

fictitious lexeme «be»: G–ATTR→«be»–II→D; the G is an N(prop), and the
prototypical D is an ADJ.

Peter–[the]–specif-appos→Great; Nicholas–specif-appos→II

Russian

Pëtr–spec-appos→Pervyj lit. ‘Peter First’

NB Cf. pervyj←ordin–Pëtr lit. ‘[the] first Peter’, because here the ordinal ADJ denotes one of
several Pëtrs who is the first in a series.
92 2 A general inventory of surface-syntactic relations in the world’s languages

77. Identity-appositive SSyntRel. It expresses the DSyntRel ATTR and the ficti-
tious lexeme «be»: G–ATTR→«be»–II→D; the G is an N, and the proto-
typical D is an N.

[the] term–ident-appos→“suffix”; [the Polish] word–ident-appos→CIASTKO ‘pastry’

78. Qualifying-appositive SSyntRel. It expresses the DSyntRel ATTR and the

fictitious lexeme «be»: G–ATTR→«be»–II→D; the G is an N(prop), and the
prototypical D is an N.

Russian

utës[-]–qual-appos→velikan lit. ‘rock giant’

devuška[-]–qual-appos→počtalʹon lit. ‘girl postman’
raketa[-]–qual-appos→nositelʹ lit. ‘rocket booster’
uragan[-]–qual-appos→ubijca lit. ‘hurricane killer’

79. Descriptive-appositive SSyntRel. It expresses the DSyntRel ATTRdescr and

the fictitious lexeme «be»: G–ATTRdescr→«be»–II→D; the G is an N, and
the D is an N.

[This] term–descr-appos→(“suffix”) [will be considered later.]

John,–[a professional–descr-appos→vet, [came over.]
[You forget about] me,–[your]–descr-appos→mother.
[The sales totaled] $10,000,–descr-appos→down [from June.]

 itle-appositive SSyntRel. It expresses DSyntRel ATTR and the fictitious

80. T
lexeme «title»:
G–ATTR→«title»–II→D; the G is an N, and the D is an N that denotes a title.

General←title-appos–Wanner vs.
Wanner,–[a]–descr-appos→general [in the Catalan army]
Mother←title-appos–Teresa vs.
Teresa,–[your]–descr-appos→mother
Father←title-appos–Patrick; Sir←title-appos–Nicholas
NB Cf. General←title-appos–Wanner,–[the]–descr-appos→commander [of 32nd Catalan division]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 93

81. Name-appositive SSyntRel. It expresses DSyntRel ATTR and the fictitious

lexeme «name»: G–ATTR→«name»–II→D; the G is an N, and the D is an
N(prop) or the phrase “of N(prop).”

[the] Gobi←name-appos–desert; Lake–name-appos→Erie

[the] Volga←name-appos–river; [the] river–name-appos→Thames
[the] Vancouver←name-appos–island; [the] island–name-appos→of [Madagascar]
NB The choice of the linear position for the proper name in cases such as the Volga river vs. the
river Thames or the Vancouver island vs. the island of Madagascar is done according to the syn-
tactic features of the proper name.
[the heavy] cruiser–name-appos→“Saratoga”; [the] USS–name-appos→Enterprise
[the] town–name-appos→of [Mount-Royal]
equation–name-appos→(23); Section–name-appos→B; [World] War–name-appos→II
NB 1. Nicholas–specif-appos→II (No. 76), since here “II” is not the name of Nicholas.
2. On the English constructions of the type the prefix un-, the painting “Seated Woman”, the
poet William Blake, etc., which are described by the SSyntRels Nos. 76, 77, 79 and 81, see
Jackendoff 1984.

82. “Per”-appositive SSyntRel. It expresses DSyntRel ATTR and the fictitious

lexeme «per»: G–ATTR→«per»–II→D; the G is a measure noun N1, and
the D is another measure noun N2 , with or without a preposition.

[100] kilometers–[an]–“per”-appos→hour; [100] dollars–[a]–“per”-appos→week

[100] kilometers–“per”-appos→per [hour]; [100] dollars–“per”-appos→per [week]

Russian

[100] kilometrov–“per”-appos→v [čas] lit. ‘100 km in hour’

[100] dollarov–[každuju]–“per”-appos→nedelju lit. ‘100 dollars each weekACC’

83. A
 dnominal-linking SSyntRel. It has no correspondent in the DSyntS, but
is intro­duced by DSynt-to-SSynt-structure rules. The G is an N, and the D
is a linker—a lexeme that depends on this N and is used to introduce N’s
postposed modifiers and attributes of various types; as a rule, a linker
agrees with its G (= the modified noun) in gender, number, case and defi-
niteness.
94 2 A general inventory of surface-syntactic relations in the world’s languages

(17) Albanian

— The linker introduces an ADJ:

sistem(masc)+Ø +Ø–adnom-link–→ i[—modif—→mirë]
system SG.NOM NON-DEF MASC.SG.NOM.NON-DEF good
‘a.system i good’ = ‘a good system’

sistem(masc)+e +Ø–adnom-link–→ të[—modif—→mirë]

system PL.NOM NON-DEF MASC.PL.NOM.NON-DEF good
‘systems të good’ = ‘good systems’

sistem(masc)+Ø +i—adnom-link–→ i[—modif—→mirë]

system SG.NOM DEF MASC.SG.NOM.DEF good
‘the.system i good’ = ‘the good system’

sistem(masc)+e +t—adnom-link–→e[—modif—→mirë]
system PL.NOM DEF MASC.SG.NOM.DEF good
‘the.systems e good’ = ‘the good systems’

— The linker introduces an NGEN:

sistem(masc)+Ø +Ø–adnom-link–→i[—subj-adnom—→edukim+i +t]
system SG.NOM NON-DEF MASC.SG.NOM.NON-DEF education SG.GEN DEF
‘a.system i of.the.education’

sistem(masc)+e +Ø–adnom-link–→të[—subj-adnom—→edukim+i +t]

system PL.NOM NON-DEF MASC.PL.NOM.NON-DEF education SG.GEN DEF
‘systems të of.the.education’

sistem(masc)+Ø +i–adnom-link–→i[—subj-adnom—→edukim+i +t]

system SG.NOM DEF MASC.SG.NOM.DEF education SG.GEN DEF
‘the.system i of.the.education’

sistem(masc)+e +t–adnom-link–→e[—subj-adnom—→edukim+i +t]

system PL.NOM DEF MASC.PL.NOM.DEF education SG.GEN DEF
‘the.systems e of.the.education’
 2.5 An inventory of surface-syntactic relations found in the world’s languages 95

I.2.3 Adpositional phrase SSyntRels, valence-controlled: 84–85

84. Prepositional-completive SSyntRel. It expresses the DSyntRel II; the G is a

PREP, and the prototypical D is an N.

in–prepos-compl→bed; without–[three hundred]–prepos-compl→dollars

to–prepos-compl→go to–prepos-compl→bed
Given–[this]–prepos-compl→postulate, [what are the values] for–[the]–prepos-
compl→velocity?
NB Here, GIVEN is a preposition.

— The D of the prepositional-completive SSyntRel can be a VINF:

[Do you ever do anything] besides–prepos-compl→offer [your apologies?]

French

sans–prepos-compl→parler ‘without to.speak’ = ‘without speaking’

— The D of the prepositional-completive SSyntRel can be a that-clause:

[The iota operator is different] in–prepos-compl→that [its interpretation depends

on the context.]

French

[Il faut battre le fer] pendant–prepos-compl→qu’[il est chaud]

lit. ‘You have to.strike the iron while that it is hot’.

Spanish

[El hecho] de–prepos-compl→que [“gordo” funciona como un nombre no afecta]

a–prepos-compl→si [es fraseologizado]
lit. ‘The fact of that gordo functions as a noun does.not affect to whether [it] is
phraseologized’.

85. Postpositional-completive SSyntRel. It expresses the DSyntRel II; the G is a

post­position, and the D is an N.

[ten] centuries←postpos-compl–ago; [a few] years←postpos-compl–back

[the whole] month←postpos-compl–through
[The motion passed, our] objection←postpos-compl–notwithstanding.
96 2 A general inventory of surface-syntactic relations in the world’s languages

Hungarian

(18) a. a szobá+n←postpos-compl–kívül ‘outside the room’

the room SUPERESS(ive) outside
b. anya +Ø←postpos-compl–szerint ‘according to Mother’
Mother NOM(inative) according.to

Hungarian does not have prepositions, only postpositions.

It is necessary to distinguish a prepositional and a postpositional SSyntRel, since

they can coexist in the same language. Thus, English has a postpositional SyntRel
for such postpositions as ago, back (three years and a half back) and notwith-
standing; they cannot be lumped together with prepositions, since their behav-
ior is too different.

I.2.4 V
 erbal phrase (= analytical formation) SSyntRels, non-valence­-
controlled: 86–94

Analytical SSyntRels are needed to describe two types of phrase:

– Inflectional analytical forms, such as has→written, are→writing, was→writing,

was→writ­ten, have→been→writing or more←intelligent, most←intelligent,
etc. These are genuine inflectional forms (= lexes) of the corresponding
lexemes.
– Derivational “analytical formations,” such as English phrasal verbs ˹give→up˺
‘abandon, surrender’, ˹do→in˺ ‘kill, destroy’, ˹put→off˺ ‘postpone’, etc.
or German verbs with separable prefixes, such as Ger. X ruft–[Y]→an ‘[X]
phones [Y]’ (anrufen ‘[to] phone’), X gibt–[Y]→auf ‘[X] gives Y up’ (aufge-
ben ‘[to] give up’), X teilt–[Y ZDAT]→mit ‘[X] communicates Y to Z’ (mit­teilen
‘[to] communicate’), etc. These are idioms (lexemic or morphemic) split into
two parts by particular syntactic rules.

As one sees, there is no sufficient parallelism between these two types of phrase:
in an analy­tical form one of its components expresses a grammeme (or a configu-
ration of grammemes), while the components of a derivational analytical forma-
tion have no meaning of their own.
Since at the DSynt-level, an inflectional form of a lexeme as well as an idiom
is always represented by one node, the analytical SSyntRels do not correspond to
any DSyntRel.
An analytical form consists minimally of a lexical part, or a full lexeme
(WRITE, INTELLIG­ENT), and an auxiliary part—that is, grammatical lexemes, which
 2.5 An inventory of surface-syntactic relations found in the world’s languages 97

either serve as the markers of the corresponding grammemes (for instance, have
expresses PERFECT; be expresses PROGRESSIVE or PASSIVE; more expresses COMPARA-
TIVE; most expresses SUPERLATIVE), or represent the separated derivateme marker.

There are two major types of analytical forms:

1) Either the full lexeme is the syntactic governor, while the auxiliary lexeme—
the gram­meme/deriva­teme marker—is a (mostly invariable) particle, syntac-
tically depending on it: more←intelligent or stand→up. Since the dependent
component in this type of construction is a grammatical marker, the corres-
poning SSyntRel can be generally called marker-analytical.
2) Or the auxiliary lexeme—the grammeme marker—is the syntactic governor of
the full lexeme; in all such cases known to me the auxiliary lexeme is the Main
Verb of the clause, while the lexical verb, which depends on it, is in one of
its non-finite forms: an infinitive, a participle, a converb, as in has→written,
was→writing, etc. The SSyntRels that describe these analytical forms can
be generally called lexical-analytical, since their dependent member is a full
lexeme (or a preposition/conjunction that introduces a full lexeme, see No. 93,
p. 101, the future-analytical SSyntRel in Spanish, Russian and Serbian).

Let us now consider the two families of analytical SSyntRels in more detail.

Marker-analytical SSyntRels. If in language L analytical markers are used only with

lexemes of one part of speech, this construction can be naturally described by
one SSyntRel, which will be simply marker-analytical. But if L uses such markers
with two or three parts of speech, for instance, verbs, nouns and adjectives, these
constructions do not have a prototypical dependent and different SSyntRels
are needed: verb-marker-analytical, noun-marker-analytical and adjective-marker-
analytical. The first one links the analytical tense-aspect-voice markers to verbs,
the second—the analytical number-case-definiteness markers to nouns, and the
third—the analytical degree markers to adjectives. Such a situation obtains in
Polynesian languages:

Maori

(19) a. kei.te←verb-mark-analyt–moe ‘be sleeping’

kua←verb-mark-analyt–moe ‘have slept’
i←verb-mark-analyt–moe ‘slept’

b. Kua moe te tamaiti ‘The child has slept’.

PERF sleep the child
98 2 A general inventory of surface-syntactic relations in the world’s languages

Kua whakareri te tamaiti i←noun-mark-analyt–[te]–rama

PERF prepare the child ACC the torch
‘The child has prepared the torch’.
Kua moe +a te tamaiti e←noun-mark-analyt–[te]–nanakia
PERF sleep PASS the child INSTR the monster
lit. ‘Has been.slept.with the child by the monster’. =
‘The monster has taken the child as wife’.
c. pai ~ pai–adj-mark-analyt→atu ‘better’ ~ pai––adj-mark-analyt→rawa ‘best’
good more most

Thus, in this family we can expect three SSyntRels: 86–88.

86. Verb-marker-analytical SSyntRel.

— The marker of the future tense:

Bulgarian

piša ~ šte←verb-analyt-mark–piša; pišeš ~ šte←verb-analyt-mark–pišeš

I.write will I.write youSG.write will youSG.write

— The marker of the conditional-subjunctive mood:

Russian

pisal–verb-analyt-mark→by ‘would/should write’

wrote

— The marker of the imperative mood:

Russian

Puskaj/Pustʹ←verb-analyt-mark–[on]–ujdët! lit. ‘That he goes!’ = ‘Let him go!’

Hawaiian

E←verb-mark-analyt–hele [‘oe i ke kula!]

lit. ‘Let go you to the school!’ = ‘Go to school!’
E←verb-mark-analyt–hele [kākou i ke kula!]
lit. ‘Let go we to the school!’ = ‘Let’s go to school!’
E←verb-mark-analyt–hele [ia i ke kula!]
lit. ‘Let go he to the school!’ = ‘Let him go to school!’
 2.5 An inventory of surface-syntactic relations found in the world’s languages 99

— The marker of the reflexive:

French

se←verb-mark-refl-dir-analyt–laver lit. ‘oneself wash’ = ‘to wash oneself’

oneself wash
[Tu] te←verb-mark-refl-dir-analyt–laves lit. ‘You yourself wash’ = ‘You wash yourself’.
youSG yourself wash
[Nous] nous←verb-mark-refl-dir-analyt–sommes [lavés] lit. ‘We ourselves have washed’.
we ourselves are washed
[Nous] nous←verb-mark-refl-indir-analyt–sommes [acheté une maison]
we ourselves are bought a house
‘We have bought ourselves a house’.
NB The verb-mark-refl-dir-analyt SSyntRel covers the reflexive as a DirO (‘I wash myself’) and the
verb-mark-refl-indir-analyt SSyntRel covers the reflexive as an IndirO (‘I wash the hands to.myself’).

— The marker of the “gérondif” (= of the converb):

French

en←verb-mark-analyt–lavant ≈ ‘[while] washing’

wash-PRES.PART(iciple)

This SSyntRel is also used to describe the dependence of a separable deriva-

tional/inflectional prefix of the German or Hungarian type, as well as that of
verbal adjuncts in English phrasal verbs:

— A separable prefix:

German

[Er] will [die Tür] aufmachen ‘He wants to.open the door’.
[auf- is a prefix that, added to the verb machen ‘make’, forms a morphemic idiom with the meaning
‘to open’; if a verb with such a prefix appears in an independent clause as the Main Verb, the prefix
is separated from the stem and put into the rightmost position].
vs.

[Er] macht–[die Tür]–verb-analyt-mark→auf ‘He opens [lit. ‘makes up’] the door’.
he makes the door up
100 2 A general inventory of surface-syntactic relations in the world’s languages

Hungarian

Elutazott [Párizsba] ‘[S/he] travelled to.Paris’.

[el- is a prefix of the perfective aspect; such a prefix is separated from the verb under negation
and postposed to the verb]
vs.

[Nem] utazott–verb-analyt-mark→el Párizsba ‘[S/he] did.not travel to.Paris’.

— An idiomatic verbal adjunct (see Jackendoff 2002[2010]):

˹put–verb-mark-analyt→up˺ [for the night]; ˹bring–verb-mark-analyt→down˺

NB Free, that is, non-idiomatic, verbal adjuncts are subordinated to the verb by the circumstan-
tial SSyntRel: climb–circum→up, run–circum→away, etc. Cf.: Up he climbed! vs. *Up he put me!

The particularities of the syntactic behavior of these elements—in the first place,
their linear positioning—can be taken care of thanks to the special indications in
their syntactics.

87. Noun-marker-analytical SSyntRel.

Tagalog

— The nominal plural marker MGA /máŋa/:

mga←noun-mark-analyt–aklat; mga←noun-mark-analyt–anak
PL book PL child

88. A
 djective-marker-analytical SSyntRel.

less←adj-mark-analyt–intelligent [than his brother]

as←adj-mark-analyt–intelligent [as his brother]
most←adj-mark-analyt–frequent
Rus. samyj←adj-mark-analyt–častyj ‘most frequent’

Lexical-analytical SSyntRels. In this family, the auxiliary verb—in “coopera-

tion” with the inflectional form of the lexical verb—can in principle express all
verbal grammemes: voice, aspect, tense, polarity, etc.

89. Passive-analytical SSyntRel. (See Chapter 10, 10.3.2.1, SSynt-rule I.A-7, p. 354.)

was–pass-analyt→written
 2.5 An inventory of surface-syntactic relations found in the world’s languages 101

90. Perfective-analytical SSyntRel.

has–perf-analyt→written

Serbian

sam–perf-analyt→pisao lit. ‘am having.written’ = ‘I wrote/I was writing’.

Swahili

(20) Ni +li +kuwa–perf-analyt→ni +me +soma

1.SG PAST be 1.SG PERF read
lit. ‘I.was I.have.read’. = ‘I had read’.

91. Progressive-analytical SSyntRel.

was–progr-analyt→writing

Swahili

(21) Ni +li +kuwa–progr-analyt→ni +ki +soma

1.SG PAST be 1.SG SIMULT read
lit. ‘I.was I.read’. = ‘I was reading’.

92. Preterit-analytical SSyntRel.

Catalan (ig = /č/, j = /ž/, v = /b/)

Vaig–pret-analyt→menjar lit. ‘I.go eat’. = ‘I ate’.

93. Future-analytical SSyntRel.

will–fut-analyt→write

Spanish

Van–fut-analyt→a [escribir] ‘They.are.going to write’.

Russian

[Ja] budu–fut-analyt→pisatʹ ‘I will write’.

Serbian

(i) [Ja] ću–fut-analyt→pisati ‘I will write’. =

(ii) [Ja] ću–fut-analyt→da[–subord-conj-compl→pišem]
lit. ‘I will that I.write’. = ‘I will write’.
(iii) Pisa←fut-analyt–ću (⇐ pisati ću) lit. ‘Write I.will’.
102 2 A general inventory of surface-syntactic relations in the world’s languages

94.  Negative-analytical SSyntRel.

[He] doesn’t–neg-analyt→understand.

Finnish, the verb ANNA- ‘give’

(22) anna+Ø +n ~ e +n—neg-analyt→anna+Ø

PRES 1.SG don’t 1.SG PRES
‘I.give’ ‘I.don’t give’.
anno+i +t ~ e +t—neg-analyt→anta+nut
PAST 2.SG don’t 2.SG PAST.PARTICIPLE
‘you.gave’ ‘youSG.didn’t give’
anta+isi +Ø ~ e +i—neg-analyt→anta+isi
IRREAL 3.SG don’t 3.SG IRREALIS
‘he.would.give’ ‘he.wouldn’t.give’

95. Interrogative-analytical SSyntRel.

Does–[he]–interrog-analyt→understand?

96. Assertive-analytical SSyntRel.

[He] does–interrog-analyt→understand.

I.2.5 Conjunctional-completive Phrase SSyntRels, Valence-controlled: 97–107

97.  Subordinative-conjunctional-completive SSyntRel. It expresses DSyntRel II;

the G is a CONJ(subord), and the prototypical D is a VFIN.

[I’ll never be the same] since–[John]–subord-conj-compl→came [into my life.]

For empty complementizers such as that, Fr. que ‘that’, Rus. čto ‘that’, etc.,
which do not appear in the DSyntS, the subord-conj-compl SSyntRel is postulated
by analogy:
[Suppose] that–[John]–subord-conj-compl→comes.

98. Subordinative-conjunctional-infinitival-completive SSyntRel. It expresses

DSyntRel II; the G is a CONJ(subord-inf), and the D is a to→VINF phrase.

˹so as˺–[not]–subord-conj-inf-compl→to [irritate Leo]

˹as if˺–subord-conj-inf-compl→to [show his support]
˹in order˺–subord-conj-inf-compl→to [avoid irritating Leo]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 103

The subordinating conjunctions CONJ(subord-inf) cannot introduce a completive that-­

clause.

99. Coordinative-conjunctional-completive SSyntRel. It expresses DSyntRel II;

the G is a CONJ(coord), and the prototypical D is a lexeme of the same part
of speech as the G of the CONJ(coord).

[Alan] and–coord-conj-compl→Helen; [Alan,] but–[not]–coord-conj-compl→Helen

[Do you have a place for us] or–[we]–coord-conj-compl→must [leave now?]

100. Comparative-conjunctional-completive SSyntRel. It expresses DSyntRel II;

the G is a CONJ(compar), and the prototypical D is an N.

than–compar-conj-compl→Helen

— The D can also be a VFIN, an ADV or a CONJ(subord):

[more] than–[Vanya]–compar-conj-compl→does
as–compar-conj-compl→always
[We are never as unhappy] as–compar-conj-compl→when [we lose love.]

In Russian, the morphological case of a nominal comparate (= what the com-

parand is being compared with) depends on the case of the comparand, while
there is no direct syntactic link between the two. As a result, Russian requires
five comparative SSyntRels. Each of these results from the ellipsis of a DSynt-
configuration where the semantic relations are explicitly reflected. Thus, for No.
101, we have ljubitʹ–ATTR→silʹnyjCOMPAR–II→ljubitʹ–I→Vanja, i.e. ‘love more
than Vanya loves’. The following five SSyntRels—Nos. 101–105—are needed in the
syntax of Russian as well as in that of other languages that have nominal cases:
other Slavic languages, German, Hungarian, Finnish, etc.

101. Subject-comparative-conjunctional-completive SSyntRel. It expresses DSynt­

Rel I; the G is a CONJ(compar), and the prototypical D is an N.

[PetjaNOM ljubit LenuACC silʹnee,] čem–subj-compar-conj-compl→VanjaNOM

‘Petya loves Lena more than Vanya does’.
[PetjaNOM ljubit LenuACC,] kak–subj-compar-conj-compl→VanjaNOM
‘Petya loves Lena like Vanya does’.
104 2 A general inventory of surface-syntactic relations in the world’s languages

102. Direct-object-comparative-conjunctional-completive SSyntRel. It expresses

DSyntRel II; the G is a CONJ(compar), and the prototypical D is an N.

[PetjaNOM ljubit LenuACC silʹnee,] čem–dir-obj-compar-conj-compl→VanjuACC

‘Petya loves Lena more than he loves Vanya’.
[PetjaNOM ljubit LenuACC,] kak–dir-obj-compar-conj-compl→VanjuACC
‘Petya loves Lena as he loves Vanya’.

103. Indirect-object-comparative-conjunctional-completive SSyntRel. It expresses

DSyntRel III; the G is a CONJ(compar), and the prototypical D is an N.

[LeneDAT dostalosʹ bolʹše,] neželi–indir-obj-compar-conj-compl→VaneDAT

lit. ‘To.Lena [it] got more than to.Vanya’. = ‘Lena was through more [mishaps] than
Vanya’.
[Ja tebeDAT verju,] kak–indir-obj-compar-conj-compl→VaneDAT
lit. ‘I believe to.you as I believe to.Vanya’.

104. Oblique-object-comparative-conjunctional-completive SSyntRel. It expresses

­DSyntRel III; the G is a CONJ(compar), and the prototypical D is an N.

[Vanja privjazan k Lene bolʹše,] neželi–obl-obj-compar-conj-compl→k [Maše]

‘Vanya is.attached to Lena more than [Vanya is.attached] to Masha’.

105. Circumstantial-comparative-conjunctional-completive SSyntRel. It expresses

­DSyntRel ATTR; the G is a CONJ(compar), and the prototypical D is an N.

This SSyntRel is not valence-controlled; it is placed in this subsection by analogy.

[Teperʹ oni živut lučše,] čem–circum-compar-conj-compl→v [Kazani]

‘Now they live better than [they lived] in Kazan’.

106. Absolute-conjunctional-completive SSyntRel. It expresses DSyntRel II; the

G is a CONJ(subord, abs), and the D is an N.

This SSyntRel subordinates a complete absolute construction—that is, a noun

plus an adject­ival/an adverbial—introduced by a subordinating conjunction.

French

˹Une fois˺–[le]–abs-conj-compl→bateau[–abs-modif→redressé, stabilisez-le]

‘Once the boat [is] straightened.up, stabilize it’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 105

107. 
Elliptic-absolute-conjunctional-completive SSyntRel. It expresses DSynt­
Rel­ II; the G is a CONJ(subord, ellipt-abs), and the prototypical D is an ADJ.

This SSyntRel subordinates an incomplete absolute construction—that is, an N,

an ADJ or an adverbial—introduced by a subordinating conjunction.

[Obama’s voting record] while–ellipt-abs-conj-compl→senator [made him the most

liberal person in Congress.]
If–[a]–ellipt-abs-conj-compl→pronoun[, the grammatical subject may ...]
[even] if–[too]–ellipt-abs-conj-compl→weak [to seize power]
[The baby,] if–ellipt-abs-conj-compl→young [enough to be easily controlled, need
not be regularly dressed.
while–ellipt-abs-conj-compl→in [bed]; once–ellipt-abs-conj-compl→here

I.2.6 Word-like phrase SSyntRels, non-valence-controlled: 108–114

108. Numeral-junctive SSyntRel. It has no correspondent in the DSyntS, where

a compound numeral is represented by one node. This SSyntRel is
introduced into the SSynt-structure by DSynt-to-SSynt-structure syn-
tactic rules; the G is a NUM/ ADJ(ordin), and the D is a NUM.

two←num-junct–hundred←num-junct–fifty←num-junct–three [= 253]
fifty←num-junct–third

— The lexeme and (and its semantic equivalents in other languages) in com-
pound numerals is not a CONJ(coord):

two←num-junct–hundred←num-junct–and←num-junct–three [= 203]
one←num-junct–hundred←num-junct–and←num-junct–third [= 103rd]

German

drei←num-junct–und←num-junct–vierzigster [Band]
lit. ‘three and fortieth volume [of a periodical]’ = ‘forty-third volume’
NB three–pseudo-coord→and–[five]–coord-conj→sixths ‘3 5/6’ ([one] SIXTH, as the names of all
fractions, is an N; see SSyntRel No. 122, p. 111).

109. Name-junctive-1/2/3 SSyntRels. They have no correspondents in the

DSyntS, where a compound human name is represented by one node,
but are introduced into the SSyntS by DSynt-to-SSynt-structure rules;
106 2 A general inventory of surface-syntactic relations in the world’s languages

the G is an N(prop, hum, first_name), and the D is an N(prop, hum, second_name). These
SSyntRels are different in different languages as function of the struc-
ture of the proper human names in the language.

Spanish (Vincze & Alonso Ramos 2011; in Spanish, a person has two family names: father’s and
mother’s family names)

Margarita–name-junct-1→Alonso–name-junct-2→Ramos
name-junct-1
José–name-junct-3→Luis Rodríguez–name-junct-2→Zapatero
José–name-junct-1→Rodríguez

The name-junctive-1 SSyntRel subordinates the first family name to the given name.
The name-junctive-2 SSyntRel subordinates the second family name to the first
family name.
The name-junctive-3 SSyntRel subordinates the second given name to the first
given name.

The person officially called Margarita Luisa Alonso Ramos can be referred to as
Margarita Alonso Ramos, Margarita Alonso, Margarita Luisa Alonso, Margarita,
Margarita Luisa, Alonso Ramos and simply Alonso. (The second family name
alone—in this case, Ramos—can be used only for an outstanding and well-known
person, preceded by a title: la presidente/la doctora Ramos.)

Russian
name-junct-1
Igorʹ–name-junct-2→Aleksandrovič Mel’čuk

The name-junctive-1 SSyntRel subordinates the family name to the given name.
The name-junctive-2 SSyntRel subordinates the patronymic [= a derivative of the
father’s given name] to the given name.

In Russian the person officially called Igorʹ Aleksandrovič Melʹčuk can also be
referred to as Igorʹ Melʹčuk, Igorʹ Aleksandrovič, Igorʹ and simply Melʹčuk.

110. B
 inary-junctive SSyntRel. It has no correspondent in the DSyntS, where a
“bipart­ite word” is represented by one node, but is introduced by
DSynt-to-SSynt-struc­ture rules.
NB The G is the element that cannot be omitted; if both elements are not omissible, then
the G is the element that receives the “external” morphological impact.

each–bin-junct→other [from–prepos→each–bin-junct→other]
 2.5 An inventory of surface-syntactic relations found in the world’s languages 107

French
[Cette vente fera époque parmi les marchands] autant–bin-junct→que[–coord-
conj→parmi les amateurs] lit. ‘This sale will.make history among the merchants
as.much as among the amateurs’.
The idiom ˹AUTANT QUE˺ ‘as well as’ is a “normal” coordinating conjunction. Con-
sider, however, the following sentence:
[Cette vente fera époque] autant←restr–parmi les marchands–coord→que–coord-
conj→parmi les amateurs lit. ‘This sale will.make history as.much among the
merchants as among the amateurs’.
This sentence contains another—binary coordinating—conjunction ˹AUTANT [X]
QUE [Y]˺, also an idiom, but with a different SSynt-structure (for details on binary
conjunctions, see Chapter 7, pp. 275ff ).

Russian
nikto ‘nobody’ ~ ni←bin-junct–[dlja]–kogo lit. ‘not for body’ = ‘for nobody’
odin←bin-junct–drugogo(masc) lit. ‘one other’, odin←bin-junct–[k]–drugomu(masc)
lit. ‘one to other’ ~ odna←bin-junct–druguju(fem) lit. ‘one other’, odna←bin-junct–
[k]–drugoj(fem) lit. ‘one to other’
drug←bin-junct–druga ‘each other’ ~ drug←bin-junct–[dlja]–druga lit. ‘each for
other’ = ‘for each other’

111. C
 olligative SSyntRel. It has no correspondent in the DSyntS, but is intro-
duced by DSynt-to-SSynt-structure rules; the G is a PARTPASS, and the D
is a stranded PREP.

[The patient is] operated–collig→upon.

[John was] done–[away]–collig→with.
NB Cf. [the] problem (which) we deal–obl-obj→with in Chapter 7; here, the complement of the
preposition WITH is the relative pronoun WHICH, which can be omitted on the surface.

112. C
 orrelative SSyntRel. It has no correspondent in the DSyntS, but is put
into SSynt-structure by DSynt-to-SSynt-structure rules; the G is the cor-
relative pro­noun Fr. ce, Rus. totII.2, etc., and the D is a semantically
empty complementizer that intro­duces a completive clause.6

6 т
 отI is a pronominal adjective meaning ‘that [passenger]’
(Te passažiry, kotorye sledujut do Moskvy, … ‘Those passenger who go to Moscow…).
тот II.1 is a pronominal correlative noun meaning ‘that one’
(Tot, kto rabotaet, živët neploxo ‘That.one who works lives well’).
108 2 A general inventory of surface-syntactic relations in the world’s languages

The correlative pronoun is used in order to “nominalize” a completive clause—

in particular, to fit it into a construction with a preposition or with a verb that
requires an oblique case.

French

[Ne vous attendez pas à] ce–correl→que [nous vous rappelions de sortir vos ordures]
lit. ‘Don’t expect this that we remind you to take.out your garbage’.

Russian

[dlja] togo,–correl→čtoby [Ivan spal] lit. ‘for this that Ivan sleep’ = ‘for Ivan to sleep’
[nedovolen] tem,–correl→čto [Ivan spal] lit. ‘not.happy with.this that Ivan was.
sleeping’
NB Let it be reminded that we see here two correlative pronouns totII, namely:
(i) totII.1 governs a pseudo-relative clause turning it into a relative:
Tot,–[kto]–rel→xočet, možet ujti
lit. ‘That who wants, can leave’ [SSyntRel No. 73, p. 89].
(ii) totII.2 governs a completive clause:
Na to,–correl→čto Ivan spal, možno ne obraščatʹ vnimanija
lit. ‘On this that Ivan was.sleeping, it.is possible not pay attention’.

German

darauf–[beharren,]–rel→daß [wir zu dieser Frage konsultiert werden]

lit. ‘on.this insist that we on this question consulted are’ =
‘insist that we (should) be consulted on this issue’

113. R
 eduplicative SSyntRel. It expresses a fictitious lexeme (depending on
the lan­guage) and subordinates Lʹ—a reduplicate of L—to L.

resolutionsL ,–redupl→schmesolutionsLʹ: the fictitious lexeme is «derision».

A particular language can require more that one reduplicative SSyntRel. Thus,
Russian needs several, because of the following contrasts:

Russian (on Russian constructions with repeated wordforms, see Sannikov 2010b)

blizko–redupl-1→blizko lit. ‘close, close’ = ‘close [insistently]’: Prud uže blizko,

blizko ‘The.pond [is] already close, close’.
vs.
blizko–redupl-2→blizko lit. ‘close-close’ = ‘very close’: Prud – blizko-blizko ‘The.
pond [is] quite close’. | Nebo bylo sinee-sinee ‘The.sky was very blue’.
 2.5 An inventory of surface-syntactic relations found in the world’s languages 109

gde–redupl-1→gde
lit. ‘where, where = ‘where [insistently]’: Gde že on, gde? ‘Where is he, where?’
vs.
gde–redupl-3→gde
lit. ‘where-where’ = ‘I am not sure about anywhere else’: Gde-gde, a u nas kofe
estʹ ‘I am not sure about anywhere else, but we do carry coffee’.
počitaj–redupl-1→počitaj
lit. ‘read, read’ = ‘Read! [insistently]’: Počitaj, počitaj, prošu tebja ‘Read, read, [I]
beg you’.
vs.
počitaj–redupl-4→počitaj
lit. ‘read-read’ = ‘Don’t you dare to read! [a threat]’: Počitaj-počitaj! ‘Try and read
on me!’
NB The four types of Russian reduplicative phrases carry different prosodies.

Hindi needs at least two reduplicative SSyntRels:

do–redupl-1→do [laṛke] lit. ‘two two boys’ = ‘two boys at a time’: the fictitious
lexeme is «distributive».
roz–redupl-2→roz lit. ‘day day’ = ‘every day’: the fictitious lexeme is «every».

114. I ntraphrasemic SSyntRel. It has no correspondent in the DSyntS, but is

introduced by DSynt-to-SSynt-structure rules; it describes idioms in
which each component, except the central one, follows or precedes its
governor immediately and is invariable; the linear position of the
depending component is marked on it as a syntactic feature.

˹kingdom–intraphras→come(postoposed)˺; ˹by–intraphras→far(postposed)˺
˹as–intraphras→yet(postposed)˺; ˹as–intraphras→if(postposed)˺
˹as–intraphras→of(postposed) –intraphras→yet(postoposed)˺
[for] ˹each–intraphras→other(postposed)˺
NB But Russian odin←bin-junct–[dlja]–drugogo lit. ‘one for the other’ (No. 110).

— A special case: compound indefinite pronouns

Russian (Chapter 9, 9.2.1.1, pp. 310ff)

malo(anteposed)←intraphras–kto lit. ‘few who’ = ‘few people’
malo(anteposed)←intraphras–gde lit. ‘few where’ = ‘in few places’
malo(anteposed)_li←intraphras–kto lit. ‘few whether who’ = ‘many people’
malo(anteposed) _li←intraphras–gde lit. ‘few whether where’ = ‘in many places’
110 2 A general inventory of surface-syntactic relations in the world’s languages

kto–intraphras→-to(postposed) ‘somebody’
gde–intraphras→-to(postposed) ‘somewhere’
KTO–intraphras→BY(postposed)–intraphras→TO(postposed)–intraphras→NI(postposed)
–intraphras→BYLO(postposed) ‘no matter who’
GDE–intraphras→BY(postposed)–intraphras→TO(postposed)–intraphras→NI(postposed)
–intraphras→BYLO(postposed) ‘no matter where’

Russian (syntactic idioms with repeated lexemes; see Sannikov 2010b: 200–208)
Zavtra,–intraphras→tak–intraphras→zavtra lit. ‘Tomorow, so tomorrow’. = ‘Let it
be tomorrow’. = ‘I don’t care whether this is tomorrow or not’.

II Coordinative surface-syntactic relations: 115–122

115. C
 oordinative SSyntRel. It expresses DSyntRel COORD; the G is a lexeme of
any part of speech, and the prototypical D is a lexeme of the same part
of speech as G.

John,–coord→Mary,–coord→Peter; fast,–coord→gently,–coord→skillfully
John–coord→and[–coord-conj→Mary]; fast,–coord→but [gently]
[for] John–coord→and [Mary] vs. for–[John]–coord→and[–coord-conj→for Mary]
[coordination of prepositions]
[John] was–[reading,]–coord→and[–[Mary patiently]–coord-conj→waited.]
three–coord→or [four times a year]

116. E lliptic-coordinative SSyntRel. It expresses DSyntRel COORD; the G is a

lexeme of any part of speech, and the prototypical D is also a lexeme of
any part of speech.

[He] works–[a lot,]–ellipt-coord→but [only at night.]

[He eats] vegetables,–[however, not]–ellipt-coord→boiled, [but fried.]

Now it is the turn of a special family of SSyntRels introduced to describe what

looks like coordination of different syntactic roles. Russian (and several other
languages, see Patejuk & Prze­piórkowski 2019) has a quirky coordinative con-
struction: if different actants and circum­stan­tials of a verb are all expressed by
interrogative, negative or some quantifier-like pronouns, they are con­joined in the
most standard way, although they must be in different grammatical cases: Rus.
Kto, kogo, komu i kak poslal? lit. ‘WhoNOM, whomACC, to.whomDAT and how sent?’;
Nikogo, nikto i ničem ne kormil lit. ‘NobodyACC, nobodyNOM and with.nothingINSTR
fed’; Ivan vsegda vsem vsë i pro vsex rasskazyvaet lit. ‘Ivan always to.everybody
everything and about everybody tells’. To properly represent the grammatical
 2.5 An inventory of surface-syntactic relations found in the world’s languages 111

cases of these actants, Russian needs five more coordina­tive SSyntRels: 117–121
(at the DSynt-level, this construction is described by actantial DSyntRels linking
the Main Verb to each actant and without the conjunction i ‘and’).

117. Subject-coordinative SSyntRel.

Nikogo,–subj-coord→nikto[–obl-obj-coord→i ničem ne kormil]

lit. ‘NobodyACC, nobodyNOM and with.nothingINSTR not fed’. = ‘Nobody fed anybody
with anything’.

118. Direct-object-coordinative SSyntRel.

Kto,–dir-obj-coord→kogo [i komu poslal?] lit. ‘Who, whom and to.whom sent?’

119. Indirect-object-coordinative SSyntRel.

Kto,–indir-obj-coord→komu [i kogo poslal?] lit. ‘Who, to.whom and whom sent?’

120. Oblique-object-coordinative SSyntRel.

Kto,–obl-obj-coord→čem [i kogo kormil?] lit. ‘Who, with.what and whom fed?’

121. Circumstantial-coordinative SSyntRel.

Russian
[Kto,–dir-obj-coord→kogo,–indir-obj-coord→]komu–circum-coord→i [kak poslal?]
lit. ‘Who, whom, to.whom and how sent?’

NB A
 ll these “strangely” coordinated SSynt-actants can correspond to DSynt-actants depending
on dif­ferent verbs; cf.:
kto←I–xotetʹ–II→poslatʹ–II→čto ⇔
–III→kto
Kto,–dir-obj-coord→čto–indir-obj-coord→i komu xočet poslatʹ?
lit. ‘Who what and to.whom wants to.send?’

This fact does not create any additional problem for the SSynt-representation of the
construction under consideration—it is taken care of by DSyntS-to-SSyntS rules.

122. Pseudo-coordinative SSyntRel. It expresses DSyntRel PSEUDO-COORD.

The pseudo-coordination resembles normal coordination in its formal aspect

only: the D of the pseudo-coordinative SSyntRel necessarily follows the G, has the
same form, and carries the enumeration prosody. But a coordinating conjunction
112 2 A general inventory of surface-syntactic relations in the world’s languages

in this construction is, generally speaking, impossible; and semantically, the

pseudo-coordinative D adds a more detailed characterization to its G.

in–[Siberia,]–pseudo-coord→on–[the Ob shore,]–pseudo-coord→close [to Novosi-

birsk]
[six] dollars,–[80]–pseudo-coord→cents
[six] dollars,]–pseudo-coord→and [80 cents]
tomorrow–pseudo-coord→night; Monday–[next]–pseudo-coord→week
from–[fifty]–pseudo-coord→to [seventy pounds]
[Responses ranged] from–[the indifferent,]–pseudo-coord→to–[the surly,]–pseudo-
coord→to [the downright obscene.]

out_of–[political limbo,]–pseudo-coord→towards [the bright lights of liberty]

Saturday–pseudo-coord→night,–pseudo-coord→at [a quarter to eleven]
Saturday,–pseudo-coord→at–[night,]–pseudo-coord→after–[dinner,]–pseudo-coord­
→at [a quarter to eleven]
[He had] everything:–pseudo-coord→family[, friends, good health.]
[Such are all voiced] consonants–[, in particular,]–pseudo-coord→/b/.

Pseudo-coordinative dependents, or pseudo-conjuncts, are prominent in Korean,

where they accompany the subject, the DirO, etc.; see Chapter 4, pp. 179ff. (SUBJ
stands for the subjective case, which marks the subject and the subject’s pseudo-
conjuncts.)

Korean

(23) a. Kay+hantey John+i–pseudo-coord→son +i mul+li +ess +ta

dog DAT SUBJ hand SUBJ bite PASS PAST DECLAR(ative)

lit. ‘By.dog John hand was.bitten’. = ‘John was bitten by the dog on the
hand’.
b. Kay+ka John+ɨl –pseudo-coord→son +ɨl mul+Ø +ess +ta
dog SUBJ ACC hand ACC bite ACT PAST DECLAR(ative)
lit. ‘Dog John hand bit’. = ‘The dog bit John on the hand’.

An important particular case of the pseudo-coordinative SSyntRel is its use to
describe the verb series (Haspelmath 2016):
 2.5 An inventory of surface-syntactic relations found in the world’s languages 113

(24) a. Ewe
Ku—[tsi]—pseudo-coord→klɔ́ [ŋkú.me]
2.SG.IMPER-scoop water 2.SG.IMPER-wash face
‘Scoop some water and wash your face’.
b. Paamese
Ma+kuri +ko—pseudo-coord→lo +va +haa
1.SG IMMED-take 2.SG 1.DU.INCL IMMED go
lit. ‘I.will.take.you I.and.you.will. go’. = ‘I will take you with me’.

There is an interesting particular case of the verb series: Russian so-called double
verbs (Vajs 2000), which can also be described by means of the pseudo-coordina-
tive SSyntRel:

Russian

[Ona] sidit–pseudo-coord→xoxočet lit. ‘[She] is.sitting is.laughing.uproariously’.

[Oni] xodjat–pseudo-coord→pobirajutsja lit. ‘[They] are.walking.around are.beg­
ging’.
Davaj–pseudo-coord→ešʹ! lit. ‘GiveIMPER.2.SG eatIMPER.2.SG!’ [reinforced incitement]

Appendix A
Alphabetical index of surface-syntactic relations

absolute-conjunctional-completive 106 attributive No. 63

absolute-modificative No. 62 auxiliary No. 46
actantial-appositive No. 57 auxiliary-conjunctive No. 45
actantial-attributive No. 56
binary-junctive No. 110
addressative No. 43
adjective-marker-analytical No. 88 characterizing-adnominal-
adjunctive No. 41 attributive No. 52

adnominal-linking No. 83 circumstantial No. 33

affected-objectival No. 13 circumstantial-comparative-

conjunctional-completive No. 105
agentive-completive No. 19
circumstantial-coordinative No. 121
apposition-circumstantial No. 37
colligative No. 111
approximate-ordinal No. 69
comparative-conjunctional-
approximate-quantitative No. 67 completive No. 100
assertive-analytical No. 96 comparative-objectival No. 26
attribute-circumstantial No. 38
compositive No. 61
114 2 A general inventory of surface-syntactic relations in the world’s languages

conditional-subjectival No. 3 metaphorical-adnominal-

consecutive-subjectival No. 2 attributive No. 53
modal-objectival No. 28
coordinative-conjunctional-
completive No. 99 modificative No. 70
coordinative No. 115 modifier-attributive No. 55
copular-completive No. 21 modifier-circumstantial No. 36
copular-genitive-completive No. 22 name-appositive No. 81
correlative No. 112 name-junctive-1/2/3 No. 109
dative-ethical-objectival No. 27 negative-analytical No. 94
debitive-subjectival No. 5 noun-marker-analytical No. 87
descriptive-appositive No. 79 numeral-junctive No. 108
descriptive-attributive No. 64 object-adnominal-completive No. 49
descriptive-modificative No. 72 object-attributive-completive No. 24
descriptive-relative No. 74 object-copredicative No. 30
determinative No. 65 object-resultative-copredicative No. 31
direct-object-comparative- oblique-object-comparative-
conjunctional-completive No. 102 conjunctional-completive No. 104
direct-object-coordinative No. 118 oblique-object-coordinative No. 120
direct-objectival No. 7 oblique-objectival-1 No. 15
direct-speech-objectival No. 12 oblique-objectival-2 No. 16
distance-circumstantial No. 35 ordinal No. 68
durative-circumstantial No. 34
parenthetical No. 39
elective No. 58 passive-agentive-completive No. 20
elliptic-absolute-conjunctional- passive-analytical No. 89
completive No. 107
“per”-appositive No. 82
elliptic-coordinative No. 116
perfective-analytical No. 90
evaluative-adnominal-attributive No. 54
possessive No. 60
floating-copredicative No. 32 possessive-adnominal-attributive No. 51
future-analytical No. 93 postpositional-completive No. 85
predicate-attributive-completive No. 25
identity-appositive No. 77
prepositional-completive No. 84
indirect-object-comparative-
conjunctional-completive No. 103 presentative No. 44
indirect-object-coordinative No. 119 preterit-analytical No. 92
indirect-objectival No. 14 progressive-analytical No. 91
infinitive-copredicative-objectival No. 18 proleptive No. 42
infinitive-objectival No. 11 pseudo-coordinative No. 122
infinitive-oblique-objectival No. 17 pseudo-direct-objectival No. 10
interrogative-analytical No. 95 pseudo-subjectival No. 6
intraphrasemic No. 114 qualificative-adnominal-
irrealis-subjectival No. 4 attributive No. 50
 2.5 An inventory of surface-syntactic relations found in the world’s languages 115

qualifying-appositive No. 78 subject-attributive-completive No. 23

quantitative No. 66 subject-comparative-
conjunctional-completive No. 101
quasi-direct-objectival-1 No. 8
subject-coordinative No. 117
quasi-direct-objectival-2 No. 9
subject-copredicative No. 29
quasi-parenthetical No. 40
subjectival No. 1
reduplicative No. 113 subordinative-conjunctional-
completive No. 97
relative No. 73
subordinative-conjunctional-
restrictive No. 47 infinitival-completive No. 98

sequential No. 59 title-appositive No. 80

special-modificative No. 71
verb-marker-analytical No. 86
specifying-appositive No. 76
subjectal-adnominal-completive No. 48 WH-pronominal No. 75

Appendix B

Index of passive surface-syntactic valences of word classes

(the numbers identify the SSyntRels of which a lexeme of the given class may be a Dependent)

This index is supposed to help the reader find the SSyntRel that represents a given
construction. For instance, what SSyntRels link the elements of the phrases could
resist and resist joining in the sentence Few writers could resist joining this society?
The wordform resist is here a (bare) VINF, so that we have to choose between SSynt­
Rels Nos. 7, 11, 17–18, 33, 84, 92, 93; only No. 11 (the infinitival-objectival SSyntRel)
is good. Similarly, joining is a VING, and the possible choices are SSyntRels Nos. 1,
7, 33, 62 and 91; No. 7 is good—the direct-objectival SSyntRel.

N  1–5, 7–9, 14–16, 19–24, 28–30, 33–35, 37, 42–43, 48–57, 59–64, 77–82,
84–85, 99–107, 113
N(measure) 34, 35
N(pron, pers) 27
N(proper) 43, 76, 81, 109
V
VFIN 1 [in a pseudo-relative], 7[in a pseudo-relative or an asyndetic comple-
tive], 12 [in a Direct Speech clause], 40 [in a Direct Speech clause],
73–75, 91, 97, 100
VINF 7, 11, 17–18, 33, 84, 92, 93
VING 1, 7, 33, 62, 91
VPART 89–91
116 2 A general inventory of surface-syntactic relations in the world’s languages

ADJ 21, 23–25, 29–31, 36, 62, 70–72, 76, 107

ADJ(determ) 65
ADJ(ordinal) 68, 69
ADJ(pron) 32, 65

LINKER 83

NUM(quant) 66, 67, 108

ADV 1, 33, 36, 38

ADV(interj) 41
ADV(adjunct) 86 [go→down]
there 6
Fr. beaucoup
‘much’ 48
Rus. mnogo
‘much’ 48

PREP  7 [in some languages], 14–20, 22, 29, 33, 38, 48–56, 63–64, 111
Chin. bǍ 13
by 48
for 19
to(inf) 1, 6, 7, 10, 98
Rus. u 60

CONJ
and 108
as 39
CONJ(coord) 45, 115–120
CONJ(subord) 33, 106
CONJ(compar) 26, 39
CONJ(complementizer) 1, 6, 7, 84, 93, 106
that5 6, 7, 10, 106

PARTICLE 46, 47, 86–88

Rus. èto1 47 [Èto Ivan razbil čašku! ‘It is Ivan who broke the cup!’]
Rus. èto2 
44 [Èto Ivan tam v mjač igraet ‘This is Ivan who plays ball there’.]
Rus. to2 46 [Esli X > Y, to uravnenie ne imeet rešenija ‘If X > Y, then the.
equation has no solution’.]

DIRECT SPEECH 1, 12

A preliminary version of Chapter 2 was published as Mel’čuk 2015–2016.

3 Syntactic subject: syntactic relations, once again
3.1 The problem stated
3.2 Conceptual preliminaries
3.2.1 *Grammatical relations ⇒ syntactic relations
3.2.2 Syntactic relations are syntactic dependency relations
3.2.3 Syntactic subject as the dependent member of the subjectival SSyntRel
3.2.4 Definitional vs. characterizing parameters of the syntactic subject
3.2.4.1 Introductory remarks
3.2.4.2 Definitional parameters of clause elements
3.2.4.3 Characterizing parameters of clause elements
3.2.5. Subjecthood properties
3.2.6. Syntactic subject and ergativity
3.3 Syntactic subject: an attempt at a definition
3.4 Establishing the syntactic subject in a language 
3.4.1 No agreement on the Main Verb
3.4.2 Monopersonal agreement of the Main Verb
3.4.3. Polypersonal agreement of the Main Verb
3.5 Syntactic subject problems related to impersonal constructions
3.6 A difficult case: the syntactic subject in Lushootseed
3.7 The syntactic subject: its syntactic role vs. its semantic and communicative roles
3.8 The direct object
3.9 Summing up
3.9.1 Defining surface-syntactic relations
3.9.2 Cross-linguistic universality of particular surface-syntactic relations

To the fond memory of Sasha Kibrik (26 Mar 1939 – 31 Oct 2012)
Linguistics owes him a lot; several languages that he helped to save from oblivion owe him
a lot; his students, many of whom are profes­sors now, owe him a lot; I, his friend, owe him
a lot. And these debts will never be repaid.

3.1 The problem stated

Syntactic subject (as well as syntactic object) has always been and still is a popular
topic in linguistics, especially in typology: it suffices to indicate such studies as
Keenan 1976, Foley & Van Valin 1977, Van Valin 1981, Kozinskij 1983, Lazard 1994,
Palmer 1994, Givón 1995, Kibrik 1997 and 2001, Testelec 2001: 317–359, Farrell
2005, Falk 2006, Bickel 2011, Siewierska & Bakker 2012, Zimmerling 2012, Hand-
schuh 2014, etc., as well as the collections Li (ed.) 1976, Cole & Sadock (eds.) 1977,
Yaguello (ed.) 1994, Burgess et al. (eds.) 1995, Aikhenvald et al. (eds.) 2001, Davies
& Dubinsky (eds.) 2001, Bhaskararao & Subbarao (eds.) 2004, and Suihkonen et
al. (eds.) 2012. The notions of syntactic subject and syntactic object are known in

https://doi.org/10.1515/9783110694765-004
118 3 Syntactic subject: syntactic relations, once again

modern linguistic literature as grammatical relations, and they continue to gen-

erate controversy, especially, the notion grammatical subject—a highly conten-
tious topic since the time the term subject appeared in the linguist’s toolbox.1 For
instance, some linguistic theories claim that all clauses of all languages must have
a grammatical subject, while other theories insist that there is no such category
consistent for all languages. As is typical of the science of language, the problem
resides in the absence of a rigorously defined notion of subject—that is, for the
n-th time, we are dealing with a notional/terminological problem. (Thus, in the
two-volume collection Bhaskararao & Subbarao (eds.) 2004 we find 28 papers on
subjects and subjecthood, but not one definition of syntactic subject or even an
attempt at a definition.)
The same things can be said about the direct object; however, this latter
notion is, in a sense, “derived” from that of subject, so in what follows I will focus
on the subject in order to say a few words about the direct object at the end of the
chapter (Section 3.8).

 The goal of this chapter is to propose rigorous definitions for both above notions,
3
the syntactic subject [SyntSubj] and the direct object [DirO], and discuss, in suf-
ficient detail, several complex cases involving the SyntSubj.

The problem of the definition of SyntSubj and DirO is complicated: it involves

the represen­tation of the (surface- and deep-) syntactic structure of sentences,
the actants, the diathesis and the voice, transitivity, ergativity, agreement and
government, zero lexemes, and many other things. As a consequence, I am forced
to limit myself to approximate and sketchy characterization of many relevant
notions. At the same time, I have to analyze data from languages that I do not
sufficiently know, so that factual mistakes are quite probable. This is, however,
not that dangerous: the main thrust of this chapter is not in communicating new
linguistic facts, but in showing the logical links between facts (independently of
whether they are correct or not) and corresponding abstract statements.

1 The term subject was introduced into scientific literature somewhere in the early Middle Ages,
seemingly by Boethius: “In Boethius (5th–6th c.) we find ‘subiectum’ and ‘praedicatum’, but
these are terms which belong to logic rather than to grammar” (Lepschy 1994: 278). In other
words, the term subject was originally meant to designate a logical notion. According to Lepschy,
its use for the syntactic role of a lexical expression is known only from 18th century; the term
object is 100 years younger.
 3.2 Conceptual preliminaries 119

3.2 Conceptual preliminaries

3.2.1 *Grammatical relations ⇒ syntactic relations

First things first: language has no such things as *grammatical relations; the rela-
tions that are under discussion are syntactic. Generally speaking, the relations
between lexical units in a sen­tence include relations of semantic, syntactic, and
morphological dependency; see, e.g., Mel’čuk 1988: 105–149 and 2012–2015: vol.
3, Ch. 18. (The relation of coreference is ignored here as being of a completely
different nature: coreference is not dependency, but equivalence.) Therefore, the
only term allowed in this book for what we are dealing with is syntactic relations.
The present discussion is based on the following three postulates:

1. In any language, an utterance is represented at the syntactic level by its syn-

tactic structure.
2. The syntactic structure is a dependency structure: only this type of structure
represents syn­tactic relations directly and explicitly.
3. There are two sublevels of syntactic structure: the deep-syntactic structure
and the surface-syntactic structure. SyntSubj and DirO belong to the surface-
syntactic structure; therefore, in what follows we speak only of surface-syn-
tactic relations [SSyntRels].

As soon as we agree on these postulates, no more discussion is possible as to

whether syntactic relations as such are universal: of course they are, and that, in
the strongest sense possible—namely, syntactic relations are necessarily present
in any multilexemic utterance of any language, and they always form a connected
structure, since all words of an utterance are syntactically linked. (Strange as
it may seem, you can find in the literature statements to the effect that “Gram­
matical Relations” are not cross-linguistically universal. Would their proponents
say that syntactic relations are not cross-linguistically universal? I did not think
so and frankly believed that just replacing “grammatical” by “syntactic” would
already move us in the right direction. However, I have discovered that some lin-
guists speak quite seriously about languages without syntactic rela­tions…2 I keep
silent, in conformity with L. Wittgenstein’s Major Proposition 7.3)

2 For instance: “Contrary to common assumptions, syntactic relations, especially those of sub-
ject and object, are not universal, but are only one of several possibilities of organizing relational
clause structure” (Kibrik 1997: 279). See also Gil 1994 (about a language “without syntax”—Riau
Indonesian) and Dryer 1997. It is clear that statements of this kind are due to the absence of a
rigorous notional apparatus.
3 “Wovon man nicht sprechen kann, darüber muß man schweigen” [‘Whereof one cannot speak,
thereof one must be silent’] (Wittgenstein 1922: 162).
120 3 Syntactic subject: syntactic relations, once again

Now, from what was just said it does not, of course, follow that any particular
SSyntRel—in our case, the subjectival (and also the direct-objectival) SSyntRel—is
universal; that is what has to be explored. (The subjectival SSyntRel seems to be
cross-linguistically universal, as I hope to show. But the direct-objectival SSyntRel
is not universal: thus, it is absent from ergative languages, see below, Sections 3.2,
p. 134 and 3.8, p. 174..)

3.2.2 Syntactic relations are syntactic dependency relations

Our discussion of SyntSubj is based on a dependency representation of the syn-

tactic structure of sentences. It is impossible to present here all the necessary
notions, and I will limit myself to three references (Mel’čuk 1988, 2004 and 2009a)
concerning the notion of actant, and another reference (Mel’čuk 2006a: 181–262)
for the notions of diathesis and (grammatical) voice; the latter two notions have
their rigorous definitions in Chapter 8.
A SSyntRel r represents a family of two-lexeme syntactic constructions—that
is, a set of syn­tactically similar two-lexeme phrases. Thus, the expression “L1–
r→L2” describes all phrases (of language L) that can be produced out of lexemes
L1 and L2, if L2 depends on L1 via SSyntRel r. To illustrate, the descriptions of two
SSyntRels will be presented: the prepositional-completive SSyntRels of English and
the postpositional-completive SSyntRel of Hungarian (see Chapter 2, Section 2.4,
Nos. 84 and 85, p. 95).

The prepositional SSyntRel of English

The prepositional-completive SSynt-relation subordinates a noun L2 to a preposi-

tion L1 (I ignore here all more complex cases in which the dependent element is
not a noun/a pronoun); cf.:

(1) forL1–prepos→JohnL2 (for John)

onL1–[this shaky]–prepos→benchL2 (on this shaky bench)
withoutL1–prepos→he L2-OBL (without him)

What does the name of this SSyntRel—prepositional–completive—stand for? It carries

three types of information:

• Linear position of L2 with respect to L1: 1) L2 follows L1; 2) only some types of
dependents of L2 are allowed to be placed between L2 and L1 (an exhaustive
specification of these depend­ents is needed, of course).
• Inflection of L1 (in particular, as a function of L2, i.e. agreement): none.
 3.2 Conceptual preliminaries 121

• Inflection of L2 (in particular, as a function of L1, i.e. government):

– none, if L2 is a noun;
– OBL(ique case), if L2 is a personal pronoun (with meOBL) or the interroga-
tive/relative pronoun WHO (with whomOBL).

The linear position of L2 with respect to L1 (plus the specification of lexical units
that can appear between L1 and L2) and the possible inflection (or the absence
thereof) of both L1 and L2 are, generally speaking, necessary definitional properties
of any SSyntRel (see Subsection 3.2.3).

The postpositional-completive SSyntRel of Hungarian

Hungarian has only postpositions, which play the same syntactic role as English
prepositions; a postposition L1 subordinates its noun L2 by the postpositional-com-
pletive SSyntRel, which is specified as follows:

• Linear position of L2 with respect to L1: L2 immediately precedes L1; only ele-
ments coordinat­ed with L2 are allowed between L2 and L1.
• Inflection of L1 (in particular, as a function of L2, i.e. agreement):
– none, if L2 is a noun different from a personal pronoun;
– if L2 is a personal pronoun, L1 agrees with L2 in the morphological category
of possession, that is, L1 receives the corresponding possessive form and L2
itself is deleted (e.g., the meaning ‘with you’ is expressed literally as ‘your.
with’).
• Inflection of L2 (in particular, as a function of L1, i.e. government): L2 receives
the case governed by L1 (boxed in (2)).

(2) Hungarian
az új ház +ØL2←postpos–mellettL1 ⇔ az új ház mellett
the new house NOM close.to ‘close to the new house’

énL2←postpos–mellettL1 ⇒mellettL1+em ⇔ mellettem

I close.to close.to 1.SG lit. ‘my-close.to’ = ‘close to me’

az új ház +onL2←postpos–kivülL1 ⇔ az új házon kivül

the new house SUPRESSIVE outside ‘outside the new house’

Examples (1) and (2) illustrate the descriptions of two of the least controversial
SSyntRels. In point of fact, these descriptions are rough informal presentations
of surface-syntactic rules ({SSyntSs} ⇔ {DMorphSs}) in a Meaning-Text model.
An interested reader can see some SSynt-rules of Russian and English in Mel’čuk
1974: 260–300 and Mel’čuk & Pertsov 1987: 178–470.
122 3 Syntactic subject: syntactic relations, once again

3.2.3 Syntactic subject as the dependent member of the subjectival SSyntRel

What is being specifically considered in this chapter is the syntactic subject—

SyntSubj, Rus. podležaščee, a purely syntactic entity, that is, a clause element—a
lexical expression that is logic­ally independent of semantic and communicative
roles it can play (i.e., it is not a logical subject and not the discourse topic or
something of the kind). Thus, talking about SyntSubjs actually means talking
about the subjectival SSyntRel, of which the SyntSubj is the dependent member.
Since the classic paper Keenan 1976, SyntSubj has been understood as a
cluster concept defined inductively. In Keenan’s view, the notion of SyntSubj
is based on 1) some intuitively clear cases in the simplest sentences possible—
canonical SyntSubjs, and 2) a list of cross-linguistically universal syntactically
relevant properties of clause elements (omissibility/non-omissibility, imposing
grammemes on other clause elements, undergoing grammeme imposition, a
particular grammat­ical case, obligatory definiteness, particular linear position,
participation in syntactic processes, etc.). Different SSynt-elements are compared
to canonical SyntSubjs according to these proper­ties; those SSynt-elements that
are similar enough to the canonical SyntSubjs are also recognized as SyntSubjs.
Keenan supplied a detailed checklist of syntactically relevant properties—some
30 plus of them. In this chapter, I am using this list, although modified; see, for
instance, Iordanskaja & Mel’čuk 2009a, where the syntactic properties of clause
elements are discussed for French.
In accordance with Keenan’s approach, I will define the SyntSubj as the
most privileged SSynt-actant of the Main Verb; in other words, the SyntSubj is
the most privileged clause element in the given language.

A clause element CE1 is more privileged than another clause element CE2 if
and only if CE1 has more Keenan’s properties than CE2; the most privileged
clause element in L has more Keenan’s properties than any other clause
element. (See Comment 2 after Definition 3.1, p. 135.)

However, the word property is ambiguous—between ‘parameter’ (which has

several values) and ‘a particular value of the parameter’. Thus, in our examples
with prepositions in 3.2.2, “linear position of L2 with respect to L1” is the name of
a parameter, while the indications “L2 follows L1, not neces­sarily immediately” or
“L2 immediately precedes L1” are ones of its possible values. To avoid confusion,
from now on, I will speak of parameters and of their values, using the term prop-
erty only for ‘a value of the corresponding parameter’.
 3.2 Conceptual preliminaries 123

3.2.4 Definitional vs. characterizing parameters of the syntactic subject

3.2.4.1 Introductory remarks
Keenan (1976: 324) divided syntactic subjecthood parameters on his checklist into
two major classes: coding parameters and behavioral parameters.
Coding parameters, which I am calling definitional, specify the way the given
SSyntRel is implemented (= encoded, expressed) in the sentence—roughly, the
linear placement and inflection of its members. If and only if these parameters
are satisfied to a sufficient degree—that is, if enough of these have the expected
positive values for the clause element being tested, this clause element is the
SyntSubj. The definitional parameters concern exclusively the governor and the
dependent of the SSyntRel r under consideration; they do not involve any third,
“outside” syntact­ic element. These parameters are six in number (see below), and
they are established by logical reasoning; therefore, they are language-universal,
that is, applicable to any language.
Behavioral, or characterizing, parameters, in sharp contrast to definitional
ones, obligatorily mention an element that is exterior with respect to the expres-
sion of the L1–r→L2 configuration. For instance, one of the most exploited char-
acterizing subjecthood parameters is the possibility of relativization (the well-
known Accessibility Hierarchy of Keenan and Comrie, see Chapter 6, 6.4.3, p. 258):
“if some clause element in L can be relativized, then the SyntSubj can”; this state-
ment is not about the SyntSubj, but about the relativization, for the possibility
of which the fact of being the SyntSubj, DirO, IndirO, etc. serves as a necessary
condition. The characterizing parameters are established empirically and, there-
fore, they are in principle not language-universal. However, some of these are so
widespread that they can be called quasi-universal.
Subjecthood behavioral properties are defined for SSynt-elements in surface-
syntactic struc­tures; therefore, these elements must themselves be defined prior
to and independently from their syntactic behavior. Therefore:

 Any SSynt-relation—in particular, the subjectival SSyntRel—must be defined

3
only by its definitional (= coding), and not by the characterizing (= behavioral),
properties (Iordan­skaja & Mel’čuk 2009a: 159–160).

Cf. the following relevant remark in Croft 1994: 30: “I wish to invert the usual pri-
ority in syntactic theory of behavioral over coding properties of subjects.”
The viewpoint proposed here can be illustrated with a simple comparison.
What is a woman? The only definitional property of a woman is her gender
physiology, allowing for childbirth. Nothing in the physical appearance, social
standing or behavior defines a woman as such; no matter whether she looks and
124 3 Syntactic subject: syntactic relations, once again

dresses like a man, whether she has full civic rights or none whatso­ever, she
remains a woman: her unique definitional “privilege” is the potential capacity
of bring­ing children into the world. The same can be said about SyntSubjs: a
SyntSubj’s definitional properties make it what it is, while its behavioral (= char-
acterizing) properties may vary from language to language without changing its
fundamental nature.
Once defined, the SyntSubj of language L must, of course, be characterized
by its syntactic behavior in larger formations: for instance, its ability to relativize,
to control deverbal adverbials and/or reflexives, to control deletions under coref-
erence, etc. This can throw an interesting light on it, as well on some other clause
elements—yet this behavior can by no means define the SyntSubj.
It seems that the root of disagreement between linguists with respect to the
identification of SyntSubjs lies in the adopted principle for defining them: either
solely by their definitional (= coding) properties or by their syntactic behav-
ior—that is, by their participation in syntactic pro­cesses, with or without coding
properties. In my approach, the choice is clear-cut: any clause element, and the
SyntSubj in particular, must be defined exclusively by its coding properties and
then characterized by its syntactic behavior.

3.2.4.2 Definitional parameters of clause elements

A clause element CE is the dependent member of a particular SSyntRel r: L–r→CE.
The values of the definitional parameters of r specify the properties of L and CE
necessary for the configuration L–r→CE to appear in the surface-syntactic struc-
ture of a clause and to be imple­mented in its deep-morphological structure. If we
exclude (for simplicity’s sake) the prosody from consideration, there are six such
parameters. Since this inventory is established logically, the six parameters are
cross-linguistically universal—in the sense that they are sufficient for defining all
SSyntRels in all languages. (Some of them, of course, prove irrelevant for some
languages—for instance, because of a very flexible word order or the absence of
inflection in a given language.)

Table 3.1 Definitional parameters of a surface-syntactic clause element CE (L–r→CE)

1. CE’s dependence on the MV as its actant.

2. CE’s omissibility from the syntactic structure of the clause.
3. CE’s linear position with respect to L.
4. L’s form:
a. L’s modification as a function of CE (≈ agreement);
b. L’s valence-changing (voice-like) modification affecting the CE.
5. CE’s form as a function of L (≈ government).
6. CE’s pronominalization.
 3.2 Conceptual preliminaries 125

NB 1. Although the above definitional parameters are valid for all SSynt-clause elements—that
is, for all SSyntRels, I will comment on them using as illustration the SyntSubj, since it is
the latter that constit­utes my actual target.
2. The names of the parameters are not formulated rigorously: they are meant to refer to sets
of properties that have to be sharpened for each specific language.
3. The list of definitional parameters in Table 3.1 features some modifications with respect to
the list given in Iordanskaja & Mel’čuk 2009a.

Comments

1. Parameter 1 reflects the assumption that a full-fledged clause contains one and
only one Main Verb, as well as one and only one SyntSubj (except for coordi-
nation). Thus, the clause element that expresses the Agent of a non-finite verb
form—an infinitive or a converb—is not considered to be a SyntSubj. Note that in
some languages the syntactic head of a clause can be a unit different from a finite
verb—for instance, an infinitive, so that, in point of fact, I am talking here about
the dependence on the head of the clause whatever it is (see on the SyntSubj in
Russian, Section 3.3 below, pp. 136ff ).

2. Parameter 2 is aimed at omissibility of a clause element from the syntac-

tic structure of the sentence, not simply from the sentence itself; an omissible
element must be omissible from the starting semantic structure and be absent
from the SSyntS. Thus, in a Pro-Drop language, the SyntSubj, if it is a pronoun,
can or must be absent from the sentence, while it (or its source) is still present in
the sentence’s syntactic structure; this absence is known as ellipsis. One can see
this, for instance, in Spanish: a sentence such as Desapareció detrás de la esquina
lit. ‘Disappeared behind the corner’ actually means ‘He/She/It disappeared…’,
where the SyntSubj él/ella ‘he/she/it’ is not present in the sentence itself, but
appears of course in the SemS and SyntS of the sentence. Let us consider an
example of ellipsis from Navajo.

(3) Navajo (Foley & Van Valin 1977: 300–301)

a. (i) ‘Ashkii ‘at‘ééd yi+ztał
lit. ‘Boy girl kicked’. = ‘The boy kicked the girl’.
and (ii) At‘ééd yi+ztał
lit. ‘He/She girl kicked’. = ‘He/She kicked the girl’.
vs. b. (i) ‘At‘ééd ‘ashkii bi+ztał
lit. ‘Girl boy was.kicked’. = ‘The girl was.kicked by the boy’.
and (ii) ‘At‘ééd bi+ztał
lit. ‘He/She girl was.kicked’. = ‘He/She was.kicked by the girl’.
126 3 Syntactic subject: syntactic relations, once again

Here none of the actants of the MV is omissible from the sentence’s Synt-struc-
ture: its physic­al absence from the sentence is due to its pronominalization with
the subsequent Pro-Dropping. However, in an English sentence such as The
bridge was destroyed the Synt-actant expressing the Agent is not present in the
Sem-structure nor in the Synt-structure: the sentence does not mean ‘… destroyed
by HIM/HER/THEM’. In other words, the Agent of an English passive verb need not
be recoverable from discourse (and so it is not amenable to pronominalization); it
need not be known or knowable to the Speaker.

NB A SyntSubj can be absent from the sentence while being present in its Synt-structure according
to two scenarios. 
– In a Pro-Drop language: the SyntSubj is pronominalized and then omitted from the sentence.
– In a language that has the inflectional category of predicativity (see Mel’čuk 1992–2000:
vol. 2, 221ff ), such as Altaic and Samodian languages, as well as Korean: the SyntSubj can
be “fused” with the verb ‘be’ into one wordform, so that it is impossible to distinguish the
SyntSubj and the Main Verb; cf.: 
(i) Korean
Pul +i +ta! lit. ‘Fire.is!’ = ‘There is fire!’
fire be DECLAR(ative)
Even if we accept that the SyntSubj is not explicitly present in sentence (i), it is, of course, present
in its Synt-structure: it is pul ‘fire’. (See also Chapter 4, 4.5.1, p. 191.)

3. Parameter 3 presupposes a preferred word order in a clause without any com-

municative effects. Thus, in English, in a simple, communicatively neutral clause
the SyntSubj precedes the MV (and all MV’s other actants); in Malagasy, the Synt­
Subj follows the MV (and all its other actants).

4. Parameter 4 covers two cases: 4a and 4b.

Parameter 4a concerns, roughly speaking, the agreement of the L1 with the CE in
question.
No rigorous definition of agreement can be given here (see, e.g., Mel’čuk
2006а: 58ff), but an intuitive understanding seems to be sufficient. It must be
emphasized that

“The lexeme L1 agrees with the lexeme L2” does not mean that L1 faithfully
copies some features of L2; this only means that some features of L2 control
the morphological form of L1.

Thus, the Russian MV agrees with the prepositional phrase po + NP ≈ ‘NP each
…’ in the role of SyntSubj by taking the grammemes SG, NEUTER: Prixodil+o
[NEU.3.SG] po pjat′ posetitelej v čas lit. ‘Came each five visitors in hour’. = ‘Each
hour five visitors came’. Similarly, in Arabic, the MV agrees with the SyntSubj,
although the rules of this agreement are by no means straightforward. Namely:
 3.2 Conceptual preliminaries 127

• If the SyntSubj denotes humans:

– If the MV precedes the SyntSubj, then, whatever the number of the
SyntSubj:
– if the SyntSubj denotes male humans, then the MV must be in the sin-
gular masculine;
– if the SyntSubj denotes female humans, then the MV must be in the
singular feminine.
– If the MV follows the SyntSubj, then it has the gender and the number of
the SyntSubj.
• If the SyntSubj does not denote humans:
The MV must be in the singular feminine, whatever the gender and the number
of the SyntSubj.

Speaking of MV agreement, two possible complications should be kept in mind.

First, we must make sure that in L the agreement of the MV is indeed con-
trolled by a syntact­ically determined unit rather than by a semantic or communi-
cative factor. Two examples:

– In Awa Pit (Kibrik 2003: 158–160), the MV agrees with the actant higher on
the person hierarchy (1 > 2 > 3), whatever its syntactic role:

(4) Awa Pit

a. (i) Na+Ø pjan+ni +s ‘I will hit youSG/youPL/him/them’.

I NOM hit FUT 1.SG
and (ii) Na+wa pjan+ni +s ‘YouSG/YouPL/He/They will hit me’.
I ACC hit FUT 1.SG

b. (i) Nu +Ø /Us+Ø pjan+ni +zi ‘YouSG/He will hit him/them’.

youSG NOM/he NOM hit FUT 2/3.SG
and (ii) Nu +wa /Us+a pjan+ni +zi ‘He/They will hit youSG/him’.
youSG ACC /he ACC hit FUT 2/3.SG

– In Kinyarwanda (Dalrymple & Nikolaeva 2011) the agreement of the MV is

determined by the communicative role of the controller rather than by its
syntactic function:

(5) Kinyarwanda (Roman numerals in the glosses indicate noun classes)

a. Umu+huûngu a+ra +som+a igi+tabo

I boy I PRES read IND VII book
‘The.boy is.reading the.book’.
128 3 Syntactic subject: syntactic relations, once again

vs. b. Igi+tabo ki +ra +som+a umu+huûngu

VII book VII PRES read IND I boy
‘It is the boy who is reading the book’.

In (5b), the MV agrees in noun class with its DirO, because the SyntSubj is Rhe-
matic and Focalized.
In a language with semantically or communicatively controlled agreement of
the MV, this agreement should not, of course, be considered among definitional
properties of the SyntSubj.
Second, the MV often agrees with a zero dummy SyntSubj, as, for instance, in
Russian sentence [Nadkus sdelan, i] pal′cem Ø(neu, dummy
ØBYTʹ smjatoNEU, SG lit. ‘[A.bite
3sg) MV

is.done, and] with.finger [it] is crumpled’ [M. Zoščenko]. When the MV has the
“unmarked/neutral/default form” (NEUTER, 3SG) in the absence of an overt Synt­
Subj, this means that there is a semantically empty zero lexeme SyntSubj Ø(neu, dummy
3sg)
,
which imposes this agreement (Mel’čuk 2006a: Ch. 9, especially p. 475). The failure
to have recourse to a zero SyntSubj leads to bizarre results, such as treating a
normal DirO as a “derived subject.” 4
Parameter 4a foresees not only well-known phenomenon of agreement of the
MV with the SyntSubj (in person, number, and gender/noun class), but also more
complex situations. For instance, in Dyirbal, only the SyntSubj can be the seman-
tic target of the frequentative verbal suffix ‑ḑay, which expresses a large quantity
of referent(s) of the SyntSubj:

4 Consider the example from Biblical Hebrew in Keenan 1976: 325, (26b):
(i) Bě-γorɔl yeḥoleq ’εθ hɔ-’ɔrεṣ
by lot divide-PASS.IND.PAST.3.SG.MASC DirO the land
lit. ‘By lot [it] was.divided the land’.
The phrase ‘the land’ is not a “derived” SyntSubj; it is a DirO, explicitly marked as such by the
corresponding preposition ’εθ. The SyntSubj here is a dummy (= empty) zero, corresponding to
the English IT; this is an impersonal construction.
Of course, a dummy zero SyntSubj must be postulated with caution:
– It should not be introduced if the MV does not show agreement at all (e.g., Lezgian).
– It should not be introduced for the only reason that the MV is in the least marked form, as,
e.g., in the Hindi sentence (ii-a), where the compound MV dekhā hai is MASC.3.SG:
(ii) a. Rītā+ne laṛkī +Ø+ko dekh+Ø +ā hai
Rita INSTR girl(fem) SG DAT see PERF.PART MASC.SG be-PRES.3.SG
‘Rita has seen the girl’.
Here the perfect participle DEKHĀ agrees with LAṚKĪ (recall that agreement does not necessarily
mean the identity of features); to see this, it suffices to replace LAṚKĪ with an inanimate noun that
appears in the nominative, and the MV reacts to this modification by reflecting the grammatical
gender of the DirO (ii-c):
 3.2 Conceptual preliminaries 129

(6) Dyirbal (Dixon 1972: 250)

a. Bayi+Ø yaŗa+Ø ɲinan +ḑa +ɲu ‘Many men sat down’.
the NOM man NOM sit.down FREQ PRES/PAST
b. (i) Balam+Ø miraɲ +Ø ba+ŋgul yaŗa+ŋgu
the NOM black.bean NOM the INSTR man INSTR
gundal +ḑa +n
get.collected FREQ PRES/PAST
‘Many black beans got collected by the man’.
vs. (ii) Bayi+Ø yaŗa+Ø gundal +ŋa +ḑa +ɲu
the NOM man NOM get.collected PASS FREQ PRES/PAST
ba +gum miraɲ +gu
the DAT black.bean DAT
‘Many men collected black beans’.5

b. Rītā+ne kamr +ā/e +Ø dekh+Ø +ā /+e hai /haĩ

Rita INSTR room(masc) SG/PL NOM see PERF.PART MASC.SG/PL be-PRES.3.SG/be-PRES.3.PL
‘Rita has seen the room/s’.
c. Rītā+ne čiṭṭhī +Ø/yã+Ø dekh+Ø +ī hai /haĩ
Rita INSTR letter(fem) SG/PL NOM see PERF.PART FEM.SG/PL be-PRES.3.SG/be-PRES.3.PL
‘Rita has seen the letter/s’.
5 The passive in Dyirbal. Dixon 1972: 65–67 speaks simply of the -ŋay form and the -ŋay con-
struction: 50 years ago, no theoretical tools were available to properly describe the phenomenon.
But here are his own examples (the SyntSubj, called “pivot” by Dixon, is boxed):
(i) a. Bayi bargan+Ø X⇔I baŋgul yaŗa+ŋguX⇔II ḑurga+Ø +n
the wallaby NOM the man INSTR spear ACT PRES/PAST
‘The man is spearing the wallaby’.
b. Bayi yaŗa+Ø X⇔I (baŋun bargan+duX⇔II) ḑurga+ŋa +ɲu
the man NOM the wallaby DAT spear PASS PRES/PAST
‘The man is spearing (at the wallaby)’.
This is an obvious diathesis modification marked on the verb—that is, a voice. True, termino-
logically it is not OK to call it “passive”—because of semantic connotations of the term passive,
since in this diathesis the verb acquires an “active” meaning; many linguists call it “antipassive.”
Formally, this voice marks the following diathetic modifica­tion:
X Y X Y
I II
⇒ II I
It also turns a transitive verb into an intransitive one: in (i-a), the tense suffix -n is that of transi-
tive verbs, while in (i‑b), its counterpart, -ɲu, is used only with intransitives.
This is, of course, a classic passive schema, not some “antipassive.” What is “anti-” here is
Dyir­bal itself: being ergative, it is “anti-nominative,” in that all its verbs are semantically oriented
in a way that is a mirror image of our verbs. ‘X spears Y’ corresponds in Dyirbal to ‘Y undergoes
spearing (by X)’. Since the term passive jars as applied to (i-b), the terms direct voice and converse
voice could be used; in nominative languages, direct vs. converse voices appear as active vs. pas-
sive. (For a review of “antipassive” constructions in various language types, see Cooreman 1994.)
130 3 Syntactic subject: syntactic relations, once again

To put it differently, in Dyirbal only the SyntSubj can have the multiplicity of its
referents to be reflected in the MV. (Dyirbal does not have inflectional nominal
number.)
Parameter 4b concerns different valence-changing (= actant-manipulating)
inflections of the verbal governor, which may affect the syntactic status of the CE
under analysis; such are voices, including reflexives, and voice-like inflectional
categories such as (in)transitivization. The stock example is the passive that turns
objects into SyntSubjs.

5. Parameter 5 covers, in addition to case government of the SyntSubj by the MV,

more “exotic” cases as well. For instance:
– In Ilocano, the noun in the role of SyntSubj can be only definite, so that to
express the mean­ing ‘A man can kiss the woman’, a passive must be used—in
order to demote ‘a man’ to the agent­ive complement [AgCo]:
(7) Ilocano (Schwartz 1976)
a. Maka +bisito ti lalaki iti babay ‘The man can kiss a woman’.
can-ACT kiss the man a woman
b. *Maka +bisito iti lalaki ti babay ‘A man can kiss the woman’.
can-ACT kiss a man the woman
c. Ma +bisito ti babay iti lalaki ‘The woman can be kissed
can-PASS kiss the woman a man by a man’.
The obligatory definiteness is a definitional property of the SyntSubj in Ilocano.6
– The syntactic role of the CE under consideration may be marked not only
by a grammatical case, but also by a structural lexeme. This happens, for
instance, with the direct object in Spanish, Romanian and Hebrew, where it is
introduced by a preposition (under particular conditions: the DirO is human,
the DirO is definite, etc.):
(8) a. Spanish (human DirO)
Saludamos a nuestros hermanos lit. ‘[We] are.greeting to our brethren’.
b. Romanian (human and definite DirO)
Îi salutăm pe frații noștri lit. ‘Them [we] are.greeting on our brethren’.
c. Hebrew (definite DirO)
Ani soger et ha-dalet lit. ‘I am.closing to the.door’.

6 The same state of affairs is observed in other Malayo-Polynesian languages, for instance, in
Malagasy. — It can be the case that the obligatory definiteness is typical not of the SyntSubj, but
of the syntactic-communicative Theme, while the Theme can be expressed in these languages
only by a SyntSubj. I have no necessary data to solve this problem, but, fortunately, this is not
relevant for my general topic here.
 3.2 Conceptual preliminaries 131

6. Parameter 6 requires considering the pronominalization of L, since pronouns

often behave differently from nouns (as far as their morphology and linear posi-
tion are concerned). See, for instance, the Tongan actant clitics, Subsection 3.4.1,
(21), p. 146.
Let it be emphasized the a clause element’s—and, in particular, the SyntSubj’s—
definitional parameters must be tested in the simplest clauses of L: declarative
and communicatively most neutral. The MV must be taken in its least marked
form: in the present tense of the indicative, in the imperfective (if L has aspects),
in the active (if L has voices), without negation, etc.
Now, I can switch to characterizing parameters.

3.2.4.3 Characterizing parameters of clause elements

Most of the standard characterizing (= behavioral) parameters of clause element
CE have been known for several decades (since Keenan 1976). Here are five of the
most frequently mentioned, which are typical of hosts of languages:

– Control of relativization (the capacity of the CE in a relative clause to be core-

ferential with the modified noun)
– Control of reflexivization (the capacity of the CE to impose reflexivization)
– Control of converbs, in particular, gerunds (the capacity of the CE to be core-
ferential with the Agent of a converb)
– Control of gapping under coordination (a.k.a. “Equi-NP Deletion”; the capac-
ity of the CE to require the deletion of a coreferential CE in a coordinated
clause)
– Control of floating quantifiers (the capacity of the CE to launch a floating
quantifier, as in They arrived all at the same time.)

In addition to this quasi-universal inventory of characterizing parameters, a lan-

guage may have its own specific characterizing parameters, which should be, of
course, also taken into account. For instance, in Malagasy, the interrogative par-
ticle ve, which marks a general question, can be linearly placed only before the
SyntSubj:

(9) Malagasy (ao = /o/, o = /u/)

N +anome vola an-dRabe ve ianao?
PAST give money to Rabe INTERR youSG
‘Did you give money to Rabe?’

In this language, the control of linear placement of sentence particles is a charac-

terizing parameter of CEs.
132 3 Syntactic subject: syntactic relations, once again

As one immediately sees, all characterizing parameters of CEs concern their

capacity to control (= to be a necessary condition for) the syntactic behavior of
a “third party,” that is, of another CE. This is quite an essential property, which
should be described in all the details, but not in connection with the definition of
the CE that interest us.
From now on, I will concentrate specifically on subjecthood properties.

3.2.5 Subjecthood properties

Definitional properties of clause elements—and, in particular, subjecthood prop-

erties
– are language-specific;
– can undergo seeming violations;
– accrue in different degrees to prototypical and to peripheral clause elements.

Subjecthood properties are language-specific. A checklist of definitional

parameters for SyntSubj is universal; a checklist of characterizing parameters
for SyntSubj is quasi-universal. However, lists of subjecthood definitional and
characterizing properties—that is, of concrete values of these parameters—are
language-specific; each such list has to be established specifically for language L.
Thus, L may have no agreement on the MV, it may lack case government of
the actants, and its word order may be too flexible to be relevant. Or to take the
linear position of the SyntSubj with respect to the MV: In English, one of the defi-
nitional subjecthood properties is to be linearly placed before the MV (and before
all its other actants, which follow the MV). In Ilocano, (7) above, a definitional
property of the SyntSubj is to occupy the closest position after the MV— before
all its other actants. And in Malagasy (see (9)), an important Synt-subjecthood
privilege is to be placed after the MV and after all its other actants. Similarly, the
control of the MV’s agree­ment is in itself an important subjecthood property in
English, Russian or French, because in these languages only one clause element
can control the personal-number agreement of the MV. But in Acehnese, where
both the SyntSubj and the DirO impose agreement on the MV and the only actant
of the MV can be either SyntSubj or the DirO (see 3.4.3, Subtype 3b, pp. 163–165),
the control of the MV’s agreement is not a SyntSubj definitional property. Here,
such a property is to impose on the MV the agreement by a prefix, which is obliga-
tory and cannot be linearly separated from the verb, while the agreement suffix is
not obligatory and can migrate from the verb to the last word of the verb phrase.
Non-omissibility is a subjecthood property in English, French, etc., because only
the SyntSubj is not omissible in these languages, but not in Tagalog, where any
 3.2 Conceptual preliminaries 133

actant of the MV is omissible, including the SyntSubj: May dumating lit. ‘There.is
having.arrived’. = ‘Someone or something has arrived’.
Therefore:

 The list of definitional and characterizing properties (= values of the corre-

3
sponding para­meters) of the SyntSubj in L is specific for L .

“Violation” of subjecthood properties. A definitional property of the SyntSubj

valid for L may seem to be violated on several—sometimes even frequent—occa-
sions. Yet if a given viola­tion is triggered by a clearly statable factor (that is, the
violation happens only and always in its presence), it is irrelevant. Therefore:

 The situation where a definitional property is not satisfied under precisely

3
stated condi­tions must be ignored—as if it were satisfied.

Thus, in Finnish, SyntSubj is defined as marked by the nominative; however, if

the referent of a non-pronominal SyntSubj is indefinite, the SyntSubj is in the
PART(itive):

(10) Finnish
Lapse+t leikk+i +vät ulkona ‘The children played outside’.
child PL.NOM play PAST 3.PL outside
vs. Laps+i +a leikk+i +Ø ulkona ‘(Some) children played outside’.
child PL PART play PAST 3.SG outside

This violation—i.e., the SyntSubj in the partitive instead of the nominative—can

be safely ignored, since it has an obvious semantic motivation (the SyntSubj’s
indefiniteness), unrelated to the syntactic role of the SyntSubj.
Prototypical (= canonical) vs. non-prototypical (= deviant) SyntSubjs. As
foreseen by Keenan (1976) and developed in Iordanskaja & Mel’čuk 2009a (espe-
cially, 159ff), the whole cluster of a CE’s definitional properties does not necessar-
ily apply to all types and varieties of this CE: it can prove valid only for a subset
of CEs—namely for prototypical, or canonical, CEs; other CE do not satisfy all
definitional properties, but are similar enough to prototypical CEs to be accepted
as such. For instance, in Russian the prototypical SyntSubj is a noun in the nomi-
native; however, using syntactic analogy, also an infinitive, a prepositional po-
phrase or a ČTO-clause are also considered SyntSubjs.
134 3 Syntactic subject: syntactic relations, once again

3.2.6 Syntactic subject and ergativity

Most cases of problematic SyntSubjs come from languages with ergativity. But
what exactly is ergativity? This term is too vague; it does not correspond to a
clearly defined notion. If we con­sider the adjective ergative, it is much easier to
make its meaning more precise: it is applicable to at least three different nouns—
language, construction and case—and has accordingly three dif­ferent interpreta-
tions (Mel’čuk 1988: 251).

– Ergative language is a language in which a bi-actantial verb semantically

corresponding to a transitive verb in a non-ergative language (for instance,
‘build’ or ‘kill’) has as the generic com­ponent of its meaning the semantic
expression ‘X undergoes a change, caused by an action of Y on X’ (in a non-
ergative language the corresponding meaning is converse: ‘Y, by an action on
X, causes that X changes’). As a result, a verb in an ergative language cannot,
generally speaking, have a DirO; since a transitive verb is a verb that allows a
DirO, an ergative language normally does not have “genuine” transitive verbs
(it can have such verbs exceptionally, if they are produc­ed by diathetic modi-
fications; see Note 16, p. 154). What is called a transitive verb in an ergative
language is simply a verb that semantically corresponds to a transitive verb
of a non-ergative language.
Ergative languages include, for instance, Dyirbal, Lezgian, Avar and Archi,
see below. This is what could be called deep, or semantic, ergativity. (The
current term is syntactically ergative lan­guages.)

– Ergative construction is a construction SyntSubj←subj–MV where the

SyntSubj is marked by a case other than the nominative, something like ‘By.
me am.reading a.book’. This construction is found, for instance, in Geor-
gian, Hindi, Chukchi, Inuktitut and Warlpiri; the presence of an ergative
construction characterizes surface, or syntactic, ergativity. (The current term
is morpholo­gically ergative languages.) Note that an ergative language in this
sense should not, as a general rule, have an ergative construction, although
logically it is not excluded.

– Ergative case is a case that exclusively marks either a certain type of SyntSubj—
namely, a “tran­sitive,” or “active,” SyntSubj. The ergative case is found, for
instance, in Lezgian (where it marks the agentive complement), Georgian and
Basque; two dead languages of Asia Minor, Urartean and Hurrian, also had
an ergative case. The ergative case does not imply the existence of an ergative
construction, and vice versa: the ergative case can be used outside the erga-
tive construction (when it does not mark the SSynt-subject), and an ergative
construction can exist without ergative case. (See the remarks on the necessity
 3.3 Syntactic subject: an attempt at a definition 135

of distin­guishing ergative construction and ergative case in Tchekhoff 1979:

28–29.) The ergative as a typical case of certain SyntSubjs in certain languages
is opposed to the nominative, which is also typical of certain SyntSubjs in
certain languages. The nominative, however, is defined not by its syntactic
functions, but by the fact that it is the case of nomination: the least marked
grammatical case of nouns, appearing, in the first place, when a noun is used
to designate an entity (Mel’čuk 1988: 208). For other grammatical cases that
mark SyntSubjs in some languages, see Subsection 3.4.1, p. 141.

3.3 Syntactic subject: an attempt at a definition

Here is a universal definition of SyntSubj, applicable to any language.

Definition 3.1 – syntactic subject

The syntactic subject in a clause of L is the most privileged SSynt-actant of

this clause’s Main Verb.

Comments

1. What exactly are syntactic privileges in L has to be indicated by a list of specific

SyntSubj privileges (= properties) elaborated for L. The syntactic privileges of the
SyntSubj in L are, as a general rule, the values of all or at least of some of the six
universal definitional parameters, indicated in Table 3.1 above.

2. Definition 3.1 entails the existence of SyntSubj in any L: it is presumed that a

language necessarily has the most privileged SSynt-actant. It is logically possible
for two actants to share the same syntactic privileges, but the SSynt-actants of
the same governor must be distinguishable one way or another, and one of them,
most probably, stands out.

3. Definition 3.1 does not entail the existence of SyntSubj in any clause of any L:
subjectless clauses are quite common. There are, first, various “degenerate
clauses” without a finite MV: What a beautiful day!, Ouch!, Never in my life, etc.;
and second, full-blown clauses with a finite MV, but without a SyntSubj—in lan-
guages that allow for such a state of affairs, such as Lezgian: for instance, Čhimida
lit. ‘Is.hot’. = ‘It is hot’ (the Lezgian verb features no agreement with the SyntSubj,
so that there is no justification for a zero dummy subject).

4. Definition 3.1 fully corresponds to the hierarchy of clause element types stated
in Keenan & Comrie 1977: SyntSubj > DirO > IndirO > Obl(ique)O. This hierarchy
136 3 Syntactic subject: syntactic relations, once again

was based on the diminish­ing accessibility of noun phrases for relativization;

later it was shown that it also covers many other syntactic operations.
Since Definition 3.1 does not mention any particular properties of any par-
ticular language, it makes the SyntSubj cross-linguistically universal. However,
in a different sense, the SyntSubj is at the same time language-specific in so far
as syntactic privileges are different in different languag­es: thus, in many Indo-
European languages the main privilege of a clausal element is to impose agree-
ment on the Main Verb, while in Malagasy it is to occupy the clause-final position.
The general notion of SyntSubj can be well illustrated with Russian data,
because in Russian it is straightforward; the same state of affairs is observed in
many other languages—Slavic and, more generally, Indo-European languages.
In Russian, the subjectival SSyntRel and, consequently, the SyntSubj—boxed
in the examples— is defined by the following properties. (For simplicity’s sake, I
consider only nominal—that is, prototypical, or canonical—SyntSubjs.)

Parameter 1. The SyntSubj L2 depends only on the head L1 of the clause, be it a finite
verb or any other element (an infinitive, a verbal interjection, a special VIMPER, 2, SG
form, etc.; the syntactic head of the clause—the Synt-predicate—is boldfaced):

(11) Russian
a. Ivan spit 〈spal〉
‘Ivan is.sleeping 〈was.sleeping〉’.
b. A Ivan – nu orat′INF i vyskočil iz komnaty
lit. ‘And Ivan—NU to.yell and ran.out of room’. =
‘And Ivan started.yelling and ran out of the room’.
c. Ivan bac Petru po morde i vyskočil iz komnaty
lit. ‘Ivan smack! to.Peter on [his] mug [= ‘smacked Peter’s mug’] and
ran.out of room’.
d. Pridi Ivan vo-vremja, vsë bylo by v porjadke
lit. ‘ComeIMPER, 2, SG Ivan [= Had Ivan come] on.time, everything would.have
been in order’.

Parameter 2. In Russian, the SyntSubj L2 is non-omissible from the Synt-struc-

ture of the clause whose head, i.e., MV, is a finite verb, since the form of this verb
is controlled by the SyntSubj (= the finite verb agrees with the SyntSubj). The sen-
tences in (12) include zero subjects —lexemes having empty signifiers and percep-
tible only due to their syntactics (a dummy Ø(neu,
dummy
3, sg)
, similar to Eng. IT and Fr. IL;
 3.3 Syntactic subject: an attempt at a definition 137

the indefinite personal Ø «PEOPLE»

(3, pl)
, similar to Fr. ON and Ger. MAN; and the imper-
sonal Ø «ELEMENTS»
(neu, 3, sg)
).7

(12) Russian
a. MenjaACC Ø(neu, 3, sg) tošn+it3, SG
lit. ‘[It] me nauseates’. ≈ ‘I feel nauseated’.
b. MneDAT Ø(neu, 3, sg) byl+oNEU, SG prijatno
lit. ‘[It] to.me was pleasant’. ≈ ‘I felt good’.
c. MneDAT Ø(neu, 3, sg) povezl+oNEU, SG
lit. ‘[It] to.me favored’. ≈ ‘I was lucky’.
d. MenjaACC Ø«PEOPLE»
(3, pl) xorošo prinjal+iPL
lit. ‘[«They»] me well received’. ≈ ‘I was well received’.
e. (i) MostACC Ø«PEOPLE»
(3, pl) snesl+iPL
lit. ‘[«They»] the.bridge demolished’.
(ii) MostACC Ø«ELEMENTS»
(neu, 3, sg) snesl+oNEU, SG
lit. ‘[«It»] the.bridge destroyed’ [e.g., a flood or a hurricane].

Parameter 3. In a declarative clause, the SyntSubj L2 normally linearly precedes

its governor L1, although in several cases L2 may follow L1 (as determined by a
number of particular factors, mainly communicative ones; a list of these is, of
course, necessary).

Parameter 4. From a morphological viewpoint, the Synt-head L1 of the Russian

clause agrees (in person, number and gender) with the SyntSubj L2 and with no
other actant (Property 4a) and can undergo passivization, which demotes the
SyntSubj and promotes the DirO to the SyntSubj (Property 4b).
NB This holds, of course, only if L1 is a finite verb capable of agreement: thus, in examples (11b–d),
the boldfaced L1 is invariant and cannot show agreement; in examples (11a–c), L1 cannot be passivized.

(13) a. Agreement
Ja pokupaj+u dom ‘I am.buying a.house’.
~ My pokupaj+em dom ‘We are.buying a.house’.
Ty pokupaj+ešʹ dom ‘YouSG are.buying a.house’.
~ Vy pokupaj+ete dom ‘YouPL are.buying a.house’.
On pokupuj+et dom ‘He is.buying a.house’.
~ Oni pokupaj+ut dom ‘They are.buying a.house’.

7 Recall that a linguistic zero sign is simply a meaningful absence; see Mel’čuk 2006a: 469–516.
138 3 Syntactic subject: syntactic relations, once again

b. Passivization
(i) IvanNOM-SyntSubj pokupaet domACC-DirO ‘Ivan is buying the house’.
~ DomNOM-SyntSubj pokupaetsja IvanomINSTR-AgCo ‘The house is being bought
by Ivan’.
(ii) IvanNOM-SyntSubj kupil domACC-DirO ‘Ivan bought the house’.
~ DomNOM-SyntSubj byl kuplen IvanomINSTR-AgCo ‘The house was bought by
Ivan’.

Parameter 5. In Russian, the SyntSubj L2 is marked by the nominative, except for

two situa­tions:
– if L2 is not a nominal and cannot have cases (e.g., the infinitive in (14a), the
finite verb as the head of a subordinate subject clause in (14b), and the con-
junction čto, also as the head of a subordinate subject clause in (14c));
– if an overriding factor intervenes—for example, if the case of L2 is controlled
by a numeral (14d), or if L1 is negated (14e–f).

(14) Russian
a. Idti bylo trudno ‘To.walk was difficult’.
b. Čego on xočet , bylo nejasno ‘What he wanted was unclear’.
c. Čto on bolen, bylo očevidno ‘That he [was] sick was obvious’.
d. IxGEN bylo pjatero ‘They were five’.
e. Pis′maPL.NOM ne prišli ‘The.letters did not arrive’. ~
PisemPL.GEN ne prišlo ‘No letters arrived’.
f. IvanSG.NOM ne byl na beregu ‘Ivan wasn’t on the beach’. ~
IvanaSG.GEN ne bylo na beregu ‘There was no Ivan on the beach’.

In any language that has grammatical cases the nominative is the case of
nomination; in other words, when in a language the Speaker names something by
a noun, this noun is in the nominative. The nominative is therefore a privileged
case, and it is generally expected that the SyntSubj be marked by the nominative.

Parameter 6. Pronominalization does not affect the Russian SyntSubj’s proper-

ties in any special way.
All the other Synt-subjecthood parameters on the Keenan’s checklist—charac-
terizing para­meters—concern not so much the syntactically defined clause ele-
ments, but some semantic or communicative entities; I will give three examples.
– The control of the coreferential Actor in a phrase of the form čtoby + VINF
‘in.order.to V’ belongs in Russian to the semantic Actor rather than to the
SyntSubj (pace Kozinskij 1983: 18–19); or, to put it differently, the possibility
of using such a phrase depends on the coreference not with the SyntSubj, but
with the semantic Actor—the person or body that fired employees in (15):
 3.3 Syntactic subject: an attempt at a definition 139

(15) Russian
Mnogie sotrudniki byli uvoleny 〈*lišilis′ raboty〉, čtoby sokratit′ štaty
‘Many employees were fired 〈*lost [their] jobs〉 in.order.to reduce [the] staff’.

The choice of the čtoby + VINF construction happens during the SemS ⇔ DSyntS
transition, and it is only natural that the conditions for this choice are semantic
(i.e., unrelated specifically to SyntSubj).

– In a similar vein, Nichols et al. 1980: 376–377 demonstrate that, on the one
hand, the control of deverbal adverbials in Russian, traditionally ascribed to
the SyntSubj, can depend on its Thematicity (= Topicality):

(16) Russian
a. The SyntSubj is thematic:
Pereexav v Moskvu, IvanTHEME ustroilsja na ètot post
‘Having.moved to Moscow, Ivan obtained this position’.
vs. b. The SyntSubj is the rhematic focus:

*Pereexav v Moskvu, na ètot post ustroilsja IvanRHEM.FOCUS
‘Having.moved to Moscow, it is Ivan who obtained this position’.

On the other hand, the authors aptly note (pp. 383–384) that the control of
deverbal adverbials by a dative IndirObj with psychological predicates (Uznav ob
ètom, mneDAT zaxotelos′ poznako­mit′sja s nim lit. ‘Having.learned this, to.me the.
desire.came to.meet him’) does not constitute an argument in favor of its sub-
jecthood. Its control capacity—to the extent that such sentences are accepted by
speakers—is explained by its semantic and communicative roles: it denotes the
human Experiencer and is thematic.

– The control of the coreference with the understood “subject” of an infinitive

is not an exclu­sive syntactic property of SyntSubj, either. For instance, in
(17), such control belongs to an obvious oblique object dlja Ivana ‘for Ivan’,
which is coreferential with the “subject” of the infinitive (it is Ivan who will
be going to London):

(17) Dlja Ivana važno poexat′ v London ‘For Ivan [it is] important to.go to
London’.

This property accrues to a semantic role (the Experiencer, in this sentence—Ivan,

for whom it is important) rather than to a syntactic entity. (For a detailed review
of characterizing, or behav­ioral, properties of the Russian SyntSubj, see Testelec
2001: 317–359.)
140 3 Syntactic subject: syntactic relations, once again

Thus, in Russian, the SyntSubj can be defined clearly and robustly since it
is specified by the positive values of all definitional parameters of SyntSubjs: it
depends only on MV (or, more generally, on the head of the clause); it is non-
omissible; in a declarative sentence, it precedes the MV (if communicative factors
do not require inversion, which constitutes an explicable “viola­tion”); it is the
only actant of the MV that controls the MV’s agreement; it is marked by the nomi-
native case; its role is targeted by the passive; and its pronominalization does not
affect its status in any way.8 However, the theoretical debate over SyntSubjs (and
DirOs) started not with Russian, but with other languages, where this notion is
not so straightforward. Therefore, I will discuss the notion of SyntSubj in some
languages considered problematic in this respect.

3.4 Establishing the syntactic subject in a language

The most “material,” easily observable properties of the SyntSubj is agreement

on the MV and the case marking of the SyntSubj itself; so let us begin with agree-
ment. Based on the agreement properties of the MV, three major different types of
languages must be examined: in the sentences of L, the MV either does not agree
with its actants at all (= no agreement on the MV): 3.4.1; the MV agrees with only
one actant (= monopersonal agreement on the MV): 3.4.2; or else the MV agrees
with more than one actant (= polypersonal agreement on the MV): 3.4.3.
Recall that agreement must be considered in the simplest clause possible,
with the least marked form of the MV (since in a more complex form of the MV the
agreement with the SyntSubj might be different).

8 Of course, Russian also has some problematic SyntSubjs, for instance:

– In ÈtoSG byliPL moi druz´jaPL ‘This were my friends’ the copula agrees not with the SyntSubj èto
‘this’, but with the nominal attribute of the copula.
– The sentence MneDAT xočetsja pokoj+aGEN ≈ ‘I want some peace’ = lit. ‘[It] wants.itself to.me
of.peace’ does not have an overt SyntSubj, but manifests a dummy zero SyntSubj; the same is
true for ImDAT èt+ogoGEN xvataet lit. ‘To.them of.this [it] suffices’. = ‘This is sufficient for them’.
– A number of verbs (usually with the prefix NA-) admit the SyntSubj in the genitive:
(i) Naexali sjuda vsjakieNOM
lit. ‘Came here anybodies’. = ‘God knows who came here en masse’.
~ Naexalo sjuda vsjakixGEN
[idem, but more colloquial and more depreciative with respect to the Actor].
On the SyntSubj in Russian bi-nominative sentences (of the type Director laboratorii – Ivan ≈ ‘It
is Ivan who is the lab director’), see Mel’čuk 2012a.
 3.4 Establishing the syntactic subject in a language 141

3.4.1 No agreement on the Main Verb

Language type 1. If in language L the MV does not agree with any of its actants,
then we have two situations: L either has nominal cases, or it does not.

Subtype 1a. In L the MV does not agree with its actants, but actants are case-
marked for their syntactic role.
The SyntSubj is the actant L that is marked, generally speaking, by one of
four cases:

– either 1) by the nominative (= the least marked case, that of nomination);

– or 2) by a special case called the subjective (= the case used, in a given lan-
guage, to mark all SyntSubjs, including the only actant of an intransitive
verb; the best known sub­jective is found in Japanese—the case in -ga);
3) by another special case, the absolutive (the case used to mark intransitive
SyntSubjs and DirOs; we find it, for instance, in Tongan); 4) the SyntSubj can
be in a different case, but only exceptionally—with some lexically marked
verbs and under special conditions.

A good example of Subtype 1a language is Lezgian.

Lezgian. The Lezgian verb does not agree with its actants (no person-number or
class inflec­tion of the verb); there is no voice and no voice-like categories. The
actants of a verb are distin­guished solely by case markings: the only actant of a
monoactantial MV is in the nominative, as in (18a), while with a biactantial MV
the actant that expresses the Agent is in the ergative case in -di, and the other one
that expresses the Patient is in the nominative, see (18c):

(18) Lezgian (Mel’čuk 1988: 207–249)

a. Gada+Ø/jar+Ø χta +na
boy SG/PL NOM return AOR
‘[The] boy/s returned’.
b. *χta +na
return AOR
‘There.was.returning’.
c. Buba+Ø+di gada+Ø/jar+Ø gatha+na
father SG ERG boy SG/PL NOM beat AOR
‘Father beat.up [the] boy/boys’. = lit. ‘By.father [the] boy/s got.a.beating’.
142 3 Syntactic subject: syntactic relations, once again

d. Gada+Ø/jar+Ø gatha+na
boy SG/PL NOM beat AOR
‘[The] boy/boys got.a.beating’.
e. *Buba+Ø+di gatha+na
father SG ERG beat AOR
lit. ‘By.father [somebody] got.a.beating’.
f. Buba+Ø+di čhukur+izva
father SG ERG run PRES
lit. ‘Father is running’. = ‘By.father there.is.running’.
g. Čhukur+izva
run PRES
‘There.is.running’.
h. Gišin +da
hungry PRES
‘There.is.hunger’.

The actant in the ergative is always omissible, as in (18c) vs. (18d), even if it is
the only actant explicitly present in the clause, as in (18f) vs. (18g). The actant in
the nominative is, on the contrary, not omissible, cf. (18b) and (18e). Crucially,
(18d) is an absolutely normal, context-independent, current type of sentence. If
both actants are present with a biactantial MV, the NNOM is positioned closer to
the MV. Now, some sentences, such as (18f–g), might give the impression that
the nominative actant is absent, yet it is not the case: the verb čhukur+un ‘[to]
run’ is, in point of fact, a contraction of the phrase čhukur av+un ‘running do’,
so that the noun čhukur, not used as such in Lezgian anymore, plays the role of
SyntSubj and it is in the nominative. Sentences of the type of (18f–g) can be pro-
duced only with such “contracted” verbs (which are rather numer­ous in Lezgian).
Genuine subjectless sentences are possible, but just with semantically specific—
for instance, meteorological—verbs: Meqʻida ‘[It] is.cold’, Mičʻida ‘[It] is.dark’,
etc. The corres­ponding Indo-European sentences have either an overt dummy
SyntSubj—Eng. it, Fr. il, Ger. es —or a zero lexeme SyntSubj Ø(3, sg), which imposes
the 3.SG or SG.NEU grammemes on the verb (and on the attributive adjective if
any): Sp. Hac+e3.SG frío lit. ‘[It] does cold’ or Rus. Byl+oSG.NEU xolod­n+oSG.NEU ‘[It] was
cold’. But Lezgian meteorological sentences have no zero dummy SyntSubj, since
the Lezgian verb knows no number-person agreement; (18h) is a really subject-
less sentence, of the only possible kind in Lezgian.
Without going into more details (see Mel’čuk 1988: 207ff for additional argu-
ments), I con­clude that the SyntSubj in Lezgian is the actant marked by the nomi-
native, because it has four out of possible six subjecthood properties (= syntactic
privileges):
 3.4 Establishing the syntactic subject in a language 143

Parameter 1. The exclusive dependence on the MV.

Parameter 2. Non-omissibility from the syntactic structure.
Parameter 3. The linear position immediately before the MV.
Parameter 5. The nominative marking.

 arameters 4 and 6 are irrelevant: Lezgian has no MV agreement and no voice-

P
like categories, and the pronominalization of the SyntSubj in no way affects it.
Thus, Lezgian does not have an ergative construction, because its SyntSubj
is always in the nominative; however, it does have an ergative case: the agentive
complement is necessarily in this case, which is not used for anything else. And
most importantly, Lezgian is an ergative language —the diathesis of its transitive
verbs is inverse with respect to that of the transitive verbs in Standard Average
European [SAE] languages (or, for that matter, in Hindi). The English verb ‘X beats
up Y’ corresponds in Lezgian to a verb meaning ‘Y gets a beating from X’; ‘X sees
Y’ is in Lezgian ‘Y is.visible to X’; etc.9
As languages with the SyntSubj marked by the subjective case, I will consider
Tagalog and Tongan.
Tagalog. My description of SyntSubj in this language is based on Kroeger
1993.10

(19) Tagalog (the marker of the oblique case ng is pronounced /naŋ/; voice markers
are pre­fixes, infixes and suffixes, sometimes combined within one wordform;
the past—more pre­cisely, the perfective—is expressed by the absence of a
reduplication of the radical)

9 M. Haspelmath’s detailed grammar of Lezgian proposes a different description of Lezgian

predicative con­struction (Haspelmath 1993: 287–298), in which the NERG is considered the Synt-
Subj and the ergativity of Lezgian as a language is rejected; see also Haspelmath 1991.
10 The Tagalog case expressed by the analytical marker ang is the subjective, but by no means
the nominative (as it is frequently named): it is not used for nomination. The nominative in Ta-
galog has a zero marker—as a well-behaved nominative should—and is used for nomination,
as well as in several syntactic constructions: as the marker of an address (Oy, lalakeNOM! ‘Hey,
man!’), of a nominal attribute (Aking apóNOM siSUBJ Ramon ‘Ramon [is] my grandson’), of the com-
plement of a measure noun (dakot na bigasNOM ‘handful [of] rice’, where na is a linker—marker
of syntactic dependency, similar to the idafa marker in Persian), of a nominal modifier (bahay
ng mambubukidNOM ‘peasant house’, etc., wherever the noun is not referential or indefinite (e.g.,
Siyá’y nagsalitáng parang batàNOM ‘He speaks like [a] child’ or May libroNOM sa mesa ‘There.is [a]
book on [the] table’). The nominative is also used for the SyntSubj—in an alternation with the
subjective (under specific conditions): Kumain na siSUBJ nanay lit. ‘Has.eaten already Mother’. ~
Kumain na nanayNOM; Napakatamad angSUBJ batang ito lit. ‘Very.lazy.is child this’ ~ Napakata-
mad batangNOM ito ‘This child is very lazy’; Napakatamad angSUBJ lahat ng anak niya lit. ‘Very.
lazy.are all children his’. ~ Napakatamad lahatNOM ng anak niya’; etc. (My most hearfelt thanks to
J.-M. Fortis for his consultations on Tagalog.)
144 3 Syntactic subject: syntactic relations, once again

a. D +um +ating ang aking apó ‘My grandson came’.

come…PERF.ACT… come SUBJ my grandson
b. May d +um +ating ‘There.is havе.come’. =
exist come… PERF.ACT…come ‘Somebody came’.
c. S +um +ulat ang aking apó ng liham sa titser
write…PERF.ACT… write SUBJ my grandson OBL letter DAT teacher
‘My grandson wrote the teacher the/a letter’.
d. S +in +ulat +Ø ng aking apó ang liham sa titser
write…PERF… write PASS-DIR OBL my grandson SUBJ letter DAT teacher
‘The letter was.written by my grandson to the teacher’.
e. S +in +ulat +an ng aking apó ng liham ang titser
write…PERF…write PASS-INDIR OBL my grandson OBL letter SUBJ teacher
lit. ‘The teacher was.written.to the/a letter by my grandson’.

Summing up, the SyntSubj’s privileges in Tagalog are four:

– Parameter 1. It depends only on the MV.
– Parameter 4a. It imposes optional plural agreement on the MV (Kroeger 1993:
24–25).
– Parameter 4b. It is buttressed by a rich system of passives—among them, a
direct and an indirect, or locative, passive, which are shown in (19d–e) and
which promote other clause elements to SyntSubj status.11
– Parameter 5. It is marked by the subjective case.

The Tagalog SyntSubj is omissible form the Sem- (and Synt-) structures of the
clause (Parameter 2), its linear position does not distinguish it from other actants
of the MV (Parameter 3), and the pronominalization of its SyntSubj does not affect
the latter (Parameter 6).

11 The Tagalog SyntSubj has several typical behavioral characteristics: only the SyntSubj can
launch a floating quantifier ‘all’, only it can relativize, only it controls coreferential deletability,
etc. (Kroeger 1993: 19–36). However, the history of ideas surrounding the subjecthood in Tagalog
is quite interesting; I learned it from Kroeger 1993: 19–20 and 2007, and I feel it is worth telling
here in a few words. The founders of Tagalog studies (beginning with L. Bloomfield in 1917) had
no problem with the Tagalog SyntSubj—they identified it exactly as it is done in this chapter. But
then in 1958 an eminent American specialist in Austronesian languages, Howard McKaughan,
for several false reasons changed the terminology and proposed to call the SyntSubj in Taga-
log and structurally similar languages “the Topic.” Unfortunately, the idea caught. Later McK-
aughan realized how wrong he had been and wrote in a 1973 paper: “Please, reader, forgive me for
confusing the issue by calling these subjects the ‘topic’ of the sentence” (a rare example of real
scientific honesty and sincerity). However, inexplicably, his incorrect proposal was accepted by
accla­mation and still persists, while his strong retraction was practically paid no attention at all…
 3.4 Establishing the syntactic subject in a language 145

Tongan, genetically related to Tagalog (both belong to the Malayo-Polynesian

branch of Austronesian family), is structurally rather different. As in Tagalog, the
Tongan verb has no num­ber-personal agreement, and the linear placement of
actants with respect to the MV does not give a reliable clue as to their syntac-
tic role, since it is relatively flexible. Again as does Tagalog, Tongan has cases,
also expressed analytically, among which I will indicate four: the nominative
(unmark­ed, i.e., having a zero marker Ø), the absolutive marked by ’a,12 the erga-
tive with the marker ’e and the dative with the marker ki. The SyntSubj is boxed.

(20) Tongan (Tchekhoff 1979, Otsuka 2000 and 2010)

a. ’Oku ’alu ’a Sione ‘John is leaving’.
PRES leave ABS John
b. ’Oku ’alu ‘He/She [mentioned in the preceding context]
PRES leave is leaving’.

Unlike Tagalog, where the SyntSubj is omissible, cf. (19b), in Tongan the SyntSubj
is not omissible: in (20b), it is present in the Synt-structure, but is elided from the
sentence (although not from its structure!) by a Pro-Drop rule.

c. ’Oku sio+Ø ’a Sione

PRES see NEUTR ABS John
‘John sees’. = ‘John is not blind’.
d. ’Oku sio+Ø ’a Sione ki Mele
PRES see NEUTR ABS John DAT Mary
‘John sees Mary’.
e. ’Oku sio+’i ’a Mele ’e Sione
PRES see TRANS ABS Mary ERG John
‘John stares at Mary’.
f. ’Oku sio+’i ’a Sione
PRES see TRANS ABS John
‘He/She [mentioned in the preceding text] stares at John’.
g. ’Oku sio+’i ’e Sione
PRES see TRANS ERG John
‘John stares at him/her [mentioned in the preceding text]’.

12 The name of the Tongan absolutive should not be confounded with the name “absolutive”
often given to the nominative case in languages with the ergative construction: the Tongan ab-
solutive is formally different from the nominative. Note, however, that this absolutive optionally
alternates with the nominative in full referential NPs:
(i) ‘Oku ‘alu ‘aABS e tamasi. ~ ‘Oku ‘alu ØNOM e tamasi lit. ‘Is leaving the boy’.
146 3 Syntactic subject: syntactic relations, once again

For the grammemes NEUTR(al) and TRANS(itivizer), see immediately below.

However, the case marking does not allow us to decide which of the two
actants of a transitive MV in (20e) is more privileged. They are both non-omis-
sible (cf. (20f–g)), and, in sharp contrast with Tagalog, Tongan has no voice-like
(= actant-shuffling) verbal alternations. Yet there are two phenomena that are
helpful: cliticization of the actants and transitivization of the MV with the suffix -’i
(20e–g).
Cliticization. Personal pronominal clitics (of the three numbers—singular,
dual and plural, of the three persons, and in addition exclusive vs. inclusive),
which are the only signs13 allowed between the tense marker and the MV, cor-
respond to the single actant of a V(intrans) and to the ergative-marked actant of a
V(trans); the clitics replacing the NABS and the NERG are homophonous:

(21) a. ’Oku ne /ou ’alu

PRES he-ABS /I-ABS leave
‘He is leaving’./‘I am leaving’.
b. ’Oku ne /ou sio+Ø
PRES he-ABS/I-ABS see NEUTR
‘He sees’./‘I see’.
c. ’Oku ne /ou sio+’i ’a Sioné
PRES he-ERG/I-ERG see TRANS ABS John
‘He stares at John’./‘I stare at John’.
d. ’Oku ne /ou sio+’i
PRES he-ERG /I-ERG see TRANS
‘He stares at him’./‘I stare at him’.
e. *’Oku ne /ou sio+’i ’e Sione
PRES he-ABS/I-ABS see TRANS ERG John
‘John stares at him [⇔ ne]/me [⇔ ou]’.

Examples in (21) show that clitics correspond either to the NABS with a V(intrans), as
in (21a), or to the NERG with a V(trans), as in (21b–d), but not to the NABS with a V(trans),
as in (21e); one can con­clude that an NABS with a V(intrans) and an NERG with a V(trans)
are SyntSubjs, as shown by the boxes in (21).
Transitivization. The suffix -’i attached to a semantically bi-actantial
V(intrans) turns it into a V(trans), without affecting its semantic valence; the verb V+’i
requires that its second semantic actant be explicitly expressed syntactically as a

13 With the exception of a handful of adverbs, such as ‘often’ and ‘again’.

 3.4 Establishing the syntactic subject in a language 147

DirO—i.e., as an NABS. (NEUTR(al) and TRANS(itivizer) are grammemes of the inflec-

tional category of transitivization, see Note 23, p. 161.) Cf. (21b–d) and (22b), which
also identify NERG as the SyntSubj:

(22) a. ’Oku ’uma+Ø ’a Sione mo Mele

PRES kiss NEUTRAL ABS John and Mary
lit. ‘John kisses with/at Mary’.
b. ’Oku ’uma+’i ’a Mele ’e Sione
PRES kiss TRANS ABS Mary ERG John
‘John kisses Mary’.

The three SyntSubj’s privileges in Tongan then are as follows:

– Parameter 1. It depends only on the MV.
– Parameter 4b. Its case is affected by transitivization.
– Parameter 6. It is the only clause element expressible by a preverbal pronomi-
nal clitic.

Omissibility (Parameter 2) does not distinguish the SyntSubj from the DirO, the
word order (Parameter 3) is quite flexible, and the SyntSubj case marking (Param-
eter 5) is not decisive.
To sum up: unlike Tagalog, Tongan does have an ergative case and an erga-
tive construction, but it is—like Tagalog—a non-ergative language; in this respect
it resembles Georgian, Basque and Hindi and contrasts with Lezgian and Archi.
However, the description of Tongan subjectival constructions proposed here
faces a problem: the absolutive case. The existence of this case—different from
the nominative—in other languages is an open question (as far as I know, other
Polynesian languages do not have it), and as such, it weakens my proposal: typo-
logical plausibility is required.

Subtype 1b. In L the MV does not agree with its actants and the actants are not
case-marked for their syntactic role; the MV is not inflected at all.14
In such a language, the SyntSubj is the actant L that occupies a special linear
position in the sentence. Vietnamese and Mandarin Chinese are good examples:
here, the SyntSubj immediately precedes the MV (as before, in the examples the
SyntSubj is boxed, and the MV boldfaced).

14 In a language without grammatical cases where the verb does not agree with its actants, but
has voices, the SyntSubj will be identifiable by the passive permutation (plus, of course, linear
position). Such languages are, for instance, Malagasy and Malay/Indonesian.
148 3 Syntactic subject: syntactic relations, once again

Vietnamese (Trương 1970)

(23) a. Tôi/Giáp đã về lit. ‘I/Giap PAST return’. =

‘I/Giap returned’.
b. Tôi/Giáp đã đọc quyển sách lit. ‘I/Giap PAST read book’. =
‘I/Giap read [the] book’.

Vietnamese has no voice, so the dependence on the MV and the preverbal linear
position are the only privileges of the SyntSubj here. (I do not know about addi-
tional definitional properties of the SyntSubj specific to Vietnamese.) However,
to prevent possible misunderstandings, let me indicate that the preverbal noun
in Vietnamese can also be a prolepsis that expresses the Theme of the sentence:

c. Giáp, nó đã đọc quyển sách

lit. ‘Giap, he PAST read book’. = ‘As for Giap, he read [the] book’.

A sentence of the type (23d) shows a DirO turned into a prolepsis (Trương 1970: 105):

d. Quyển sách, đã đọc

lit. ‘Book, PAST read’. = ‘The book, somebody read it’.

The SyntSubj is absent from the SyntS (and the SemS) of sentence (23d).
In Mandarin Chinese the preverbal noun is also necessarily either a SyntSubj,
or a prolepsis, which expresses the Theme of the sentence:

(24) Mandarin Chinese (Li & Thompson 1994: 234–242)

a. Zéi kāi -le mén le ‘Thieves opened the door’.
thief open PERF door CRS [= particle signaling a Currently Relevant State of affairs]
b. Mén kāi-le ‘The door opened’.
door open PERF/CRS
↗
c. Mén, || zéi kāi-le lit. ‘Door, thieves opened’. =
↗
‘The door, the thieves opened it’.
d. Mén, || kāi-le
door open PERF/CRS

In (24a–b) we see two different lexemes of the vocable kāi, just like the English
verb open: a transitive and an intransitive one (such verbs are known as labile).
(24c–d) show mén ‘door’ in the syntactic role of a prolepsis (it is marked by a
pause and a mounting contour); in (24c) the DirO of the verb kāi and in (24d),
both the SyntSubj and the DirO are not expressed on the surface.
 3.4 Establishing the syntactic subject in a language 149

(24d) shows that in Mandarin the SyntSubj is omissible; here is another

couple of examples:

e. – Zuò shénme? lit. ‘Do what?’ – Chī zhe lit. ≈ ‘Eating be’.

This exchange is possible in any circumstances with the question put to some-
body about himself or about any third party (‘What am/is/are I/he/you/they
doing?’ – ‘I/He/You/They is/am/ are eating’.).

f. Diū -le yí kuài biǎo lit. ‘[Somebody] lost a watch’. =

lose PERF one CLASS watch ‘A watch was lost’.

The same state of affairs is characteristic of many other languages that lack inflec-
tional morphology.

3.4.2 Monopersonal agreement of the Main Verb

Language type 2. If in L any MV, intransitive or transitive, agrees with only one
of its actants, then this actant is the SyntSubj.
This must be true at least for the basic (= least marked) forms of the MV, for
instance, the imperfective stem; with the perfective stem, the transitive MV may
agree with the DirO.
Enga. The simplest case of Type 2 language known to me is the New Guinea
language Enga, which, as far as the SyntSubj is concerned, presents a very clear
picture: its MV has strictly monopersonal agreement (in all forms), and this iden-
tifies the SyntSubj uniquely.

(25) Enga (Van Valin 1981: 367–371)

a. Nambá+Ø p+é +ó ‘I went’.
I NOM go PAST 1.SG
b. Namba+mé mená+Ø dóko p +í +ó ‘I hit [= killed] the pig’.
I ERG pig NOM the hit PAST 1.SG

Like Hindi and Georgian (see below), Enga has the ergative construction, but
without split—it is used in all tenses; unlike Hindi, but like Georgian, it has a
special ergative case. Enga is, of course, not an ergative language.
150 3 Syntactic subject: syntactic relations, once again

Hindi. The situation in Hindi is more complex. Here a perfective transitive MV

controls an ergative construction (with the SyntSubj in the instrumental; see (26c)):

(26) Hindi (ai = /ε/, aĩ = /ɛ̃ /)

a. Maĩ ā +Ø +Ø hũ ̄ ‘I [a man] have come’.
I-NOM [male] come PERF.PART MASC.SG be-PRES.1.SG
Maĩ ā +Ø +ī hũ ̄ ‘I [a woman] have come’.
I-NOM [female] come PERF.PART FEM.SG be-PRES.1.SG
Ve ā +Ø +e haĩ ‘They [men] have come’.
they-NOM [males] come PERF.PART MASC.PL be-PRES.3.PL
Ve ā +Ø +ī haĩ ‘They [women] have come’.
they-NOM [females] come PERF.PART FEM.PL be-PRES.3.PL

b. Maĩ čiṭṭhī +Ø /+yã likh+Ø rah +Ø +ā hũ ̄

I-NOM [male] letter(fem) SG.NOM/PL.NOM write CONV remain PERF.PART MASC.SG be-PRES.1.SG
lit. ‘I [a man] letter/s writing remained am’. = ‘I am writing a letter/letters’.
Ham čiṭṭhī +Ø /+yã likh+Ø rah +Ø +e haĩ
we-NOM [males] letter(fem) SG.NOM/PL.NOM write CONV remain PERF.PART MASC.PL be-PRES.1.PL
lit. ‘We [men] letter/s writing remained are’. = ‘We are writing a letter/letters’.

vs. Tū čiṭṭhī +Ø /+yã likh+Ø rah +Ø +ī

youSG-NOM [female] letter(fem) SG.NOM/PL.NOM write CONV remain PERF.PART FEM.SG
hai
be-PRES.2.SG
lit. ‘You [a woman] letter/s writing remained are’. = ‘You are writing a letter/
letters’.
Tum čiṭṭhī +Ø /+yã likh+Ø rah +Ø +ī
youPL-NOM [females] letter(fem) SG.NOM/PL.NOM write CONV remain PERF.PART FEM.PL
ho
be-PRES.2.PL
lit. ‘You [women] letter/s writing remained are’. = ‘You are writing a letter/
letters’.

c. Maĩ+ne čiṭṭhī +Ø /+yã likh +Ø +ī

I INSTR [male] letter(fem) SG.NOM/PL.NOM write PERF.PART FEM.SG/PL
hai /haĩ
be-PRES.3.SG/be-PRES.3.PL
lit. ‘By.me [a man] letter/s written is/are’. = ‘I [a man] have written a letter/
letters’.
 3.4 Establishing the syntactic subject in a language 151

vs. Tū +ne čiṭṭhī +Ø /+yã likh +Ø +ī

youSG INSTR [female] letter(fem) SG.NOM/PL.NOM write PERF.PART FEM.SG/PL
hai /haĩ
be-PRES.3.SG/be-PRES.3.PL
lit. ‘By.you [a woman] letter/s written is/are’. = ‘You [a woman] have written a
letter/letters’.

We can be sure, however, that in (26c) the noun čiṭṭhī(yã) ‘letter(s)’ is a DirO:
thanks to the passive, which—as shown in (26d)—promotes this noun to the
SyntSubj, demoting the former SyntSubj to an Ag(entive) Co(mplement), which is
dispreferred in Hindi (ǯā ‘go’ is the passive auxi­liary, here in the form of converb
≈ gerund; rahā ‘remain’ is the progressive auxiliary, which takes the converb of
the lexical verb):15

d. Čiṭṭhī +Ø likh +Ø +ī ǯā+Ø rah +Ø +ī

letter(fem) SG.NOM write PERF.PART FEM.SG go CONV remain PERF.PART FEM.SG
hai
be-PRES.3.SG
lit. ‘Letter written going remained is’. = ‘The letter is being written’.

and Čiṭṭhī +yã likh +Ø +ī ǯā+Ø rah +Ø +ī

letter(fem) PL.NOM write PERF.PART FEM.PL go CONV remain PERF.PART FEM.PL
haĩ
be-PRES.3.PL
lit. ‘Letters written going remained are’. = ‘The letters are being written’.

Hindi is thus a non-ergative language: its transitive verb admits a DirO, and the
meaning of a Hindi transitive verb typically has ‘[to] cause’ as the generic (=
central) component of its defini­tion. Hindi has no special ergative case, either,
but it does have an ergative construction—with a transitive MV in a perfective
form, where the SyntSubj in the instrumental. With an imperfective MV, Hindi
uses a nominative construction, and the verb agrees then with the SyntSubj. (In
other words, Hindi manifests what is known as split ergativity: the ergative con-
struction appears under special conditions—in this case, with perfective series
tense forms; elsewhere we have the nominative construction.) In a perfective
form, the MV agrees only with the DirO.

15 The AgCo is only used in Hindi either in administrative/legal register (with the postposition
dvara ‘through/by’) or in non-assertive sentences, which express the ability of the Agent to per-
form the action (in the ablative in -se); see Kachru 2006: 204–205.
152 3 Syntactic subject: syntactic relations, once again

Thus, the SyntSubj’s privileges in Hindi are the following five (pronominal-
ization being irrelevant):

– Parameter 1. It depends on the MV.

– Parameter 2. It is non-omissible.
– Parameter 3. It occupies the linear position before the MV (and other actants).
– Parameter 4:
– 4a. It controls the agreement of the MV (but only in an imperfective form).
– 4b. It is “passivizable” (that is, it is the target of promotion by the passive).
– Parameter 5. It is marked by the nominative case (again, only with an imper-
fective MV).

Archi. The things are substantially different in Archi, a Daghestanian lan-

guage. Just like Hindi, Archi has a monopersonal agreement—if the MV is in the
one of the least marked synthetic forms, see in (27). But the single actant of the
MV that controls its noun-class agreement is—in sharp contrast to Hindi—always
in the nominative; it is not omissible and its syntactic position is targeted by an
actant-manipulating voice-like transformation (as before, this actant is boxed in
the examples; it is the SyntSubj, as will be shown).

(27) Archi (Kibrik 1977, 2003: 332–368; Roman numbers stand for noun classes)
a. Buwa +Ø+Ø da+qʻa ‘Mother came’.
mother(II) SG NOM II come-PERF
b. Dija +Ø+mu buwa +Ø+Ø χir a+r+u
father(I) SG INSTR mother(II) SG NOM behind do.II.do-PERF
lit. ‘By.father mother behind did’ [˹behind do˺ is an idiom meaning ‘bring
with oneself’]. = ‘Father brought mother with him’.
c. Dija +Ø+mu dos +Ø+Ø χir a+w+u [⇒ aw]
father(I) SG INSTR friend(I) SG NOM behind do.I.do-PERF
‘Father brought a friend with him’.
d. Dija +Ø+mu dos +til+Ø χir a+b+u
father(I) SG INSTR friend(I) PL NOM behind do.III.do-PERF
‘Father brought friends with him’.
All plural Archi nouns belong to the noun class III; the verb as ‘do’ shows class III agree-
ment with the plural dostil ‘friends’.
e. Dija +Ø+n buwa +Ø+ɬ̄u anχ +Ø+Ø a +Ø+u [⇒ aw]
father(I) SG GEN mother(II) SG COMIT fight(N, IV) SG NOM do.IV.do-PERF
lit. ‘Father’s with.mother fight was.done’. = ‘Father fought with mother’.
 3.4 Establishing the syntactic subject in a language 153

f. (i) Balah +Ø+Ø dita +b+u b +erχin

trouble(III) SG NOM soon.III.soon III forget-IMPERF
‘Trouble gets forgotten quickly’.
(ii) Arša horōk ej +b +u iškul +Ø+Ø da+b +l +u
Archi-INESS long.ago very.III.very school(III) SG NOM open.III.open PERF
‘A school opened in Archi very long time ago’.
(iii) D+ez un malgan
II I-DAT youSG(II)-NOM be.dear
lit. ‘To.me you [sg, female] are.dear’. = ‘I love you’.

The SyntSubj in Archi has six privileges:

– Parameter 1. It depends only on the MV.

– Parameter 2. The SyntSubj is non-omissible, while all other actants of the MV
can be absent from the Synt- and Sem-structure of the sentence; this is true
even for causative sentences:

(28) a. (i) Zari nokɬʻ +Ø +Ø ekʻ +Ø+u

I-INSTR room(IV) SG NOM sweep.IV.sweep-PERF
lit. ‘By.me room underwent.sweeping’. = ‘I swept [the] room’.
and (ii) Nokɬʻ +Ø +Ø
ekʻ +Ø+u
room(IV) SG NOM sweep.IV.sweep-PERF
lit. ‘Room underwent.sweeping’.
vs. b. (i) Dija +Ø +mu zari nokɬʻ +Ø +Ø ekʻ +Ø+a +s
father(I) SG INSTR I-INSTR room(IV) SG NOM sweep.IV.sweep INF
a+Ø+w
do.IV.do-PERF
 lit. ‘By.Father by.me room to.undergo.sweeping underwent.causa-
tion’. = ‘Father made me sweep the room’.
and (ii) Zari
nokɬʻ +Ø +Ø ekʻ +Ø+a +s a+Ø+w
I-INSTR room(IV) SG NOM sweep.IV.sweep INF do.IV.do-PERF
lit. ‘By.me room to.undergo.sweeping underwent.causation’. =
‘I was made to sweep the room’.
and (iii) nokɬʻ +Ø+Ø ekʻ +Ø+a +s a+Ø+w
room(IV) SG NOM sweep.IV.sweep INF do.IV.do-PERF
lit. ‘Room to.undergo.sweeping underwent.causation’. =
‘The room was made to be swept’.
154 3 Syntactic subject: syntactic relations, once again

– Parameter 3. The SyntSubj is positioned immediately before the MV after all

other its actants.
– Parameter 4a. The SyntSubj controls—almost exclusively—the noun-class
agreement not only of the MV, as seen in (27) – (28), but also of circumstan-
tials and even of certain actants, as in (27f), where the adverb ditabu ‘soon’,
the particle ejbu ‘very’ and the actant dez ‘to.me’ agree in noun class with the
SyntSubj.
– Parameter 4b. Archi has a “converse” voice16 that promotes the AgCo to the
SyntSubj, while demoting the former SyntSubj to the DirO (and turning the
verb into a transitive one):

(29) a. Buwa +Ø+mu χ̄walli +Ø+Ø b +a +r +ši b +i

mother(II) SG INSTR bread(III) SG NOM III do IMPF CONV III be-PRES
lit. ‘By.motherAgCo breadSyntSubj doing is’. = ‘Mother is baking bread’.

vs. b. Buwa +Ø+Ø χ̄walli +Ø+Ø b +a +r +ši d+i

mother(II) SG NOM bread(III) SG NOM III do IMPF CONV II be-PRES
lit. ‘MotherSyntSubj breadDirO doing is’. = ‘Mother is baking bread’.

NB The two sentences in (29) contrast: (29a) answers the question “What is happen-
ing?”, while (29b) constitutes an answer to the question “What about Mother?”; in a
sentence of this type, the SyntSubj must be Thematic.

– Parameter 5. The SyntSubj is always marked by the nominative.

As we see, Archi does not have an ergative construction, since its SyntSubj is
always in the nominative; it does not have an ergative case, either: its AgCo is in
the instrumental. But like Lezgian, Archi is an ergative language.

16 The “passive” in Archi. It is to some extent similar to the “passive” of Dyirbal, see Note 5,
p. 129. On voice in Archi, see Kibrik 1975 and 2003: 352–354 (however, Kibrik himself does not
treat this transformation as voice; he speaks simply of a binominative construction). Testelec
1979 was probably the first to explicitly insist on the voice-like character of this verbal “alterna-
tion” in Daghestanian languages and draw a parallel with Dyirbal.
The passive, or converse, voice in Archi has two characteristic properties:
– As in several other Daghestanian languages (Avar, Bezhta, Gunzib, Tsez), this voice is possible
only in one of the imperfective aspects: in the durative, the habitual, the progressive and the
frequentative.
– In this voice, the Archi MV receives a DirO in the nominative, so that both the SyntSubj and the
DirO are in the nominative, which is a kind of anathema for an ergative language. Moreover,
the MV agrees with this DirO—along with the SyntSubj, so that the MV becomes bipersonal,
and, so to speak, transitive.
 3.4 Establishing the syntactic subject in a language 155

3.4.3 Polypersonal agreement of the Main Verb

Language type 3. In language L, the MV agrees simultaneously with (at least)

two actants, using two sets of agreement markers. In some Type 3 languages, the
MV can simultaneously agree with three or even four actants. However, in order
to simplify, I consider here the MV’s agreement just with two actants—L1 and L2,
one of which is thus the SyntSubj and the other one, the DirO. This introduces
into our inquiry an additional dimension: the necessity to distinguish between
SyntSubjs and DirOs.
For a Type 3 language L, two situations must be considered: either a monoac-
tantial MV in L uses exclusively one set of agreement markers, or it uses alterna-
tively both (as a function of the lexical unit playing the role of MV).

Subtype 3a. In L the transitive MV agrees simultaneously with two actants, but
an intransitive (≈ monoactantial) MV features only one type of agreement.17
In this case, the only actant of an intransitive MV is its SyntSubj, so that the
researcher has to decide exclusively between the two actants of a transitive biac-
tantial MV. Such a situation is found in many languages; I select Georgian and
Basque for an examination.
Georgian. In contrast to Lezgian and Archi, a transitive Georgian MV agrees—
in person and number—simultaneously with two of its actants, which are, there-
fore, the SyntSubj and the DirO.18 We have to settle accounts between these two:
which one is the boss—i.e. the SyntSubj?
A transitive verb has two sets of agreement markers: Set I and Set II. Only the
markers of Set I are exclusively used for the actant of a monoactantial MV, which
stands in most cases in the nominative, cf. (30a); as I just said, it is the SyntSubj.
But this fact by itself is not sufficient to consider Set I markers as exclusively
subject markers, since on a transitive verb they can in principle cross-reference
the DirO: precisely this, as we will see, happens in Basque. One has to compare
both these actants of a transitive MV as to their case-marking and mutual linear

17 In a given L, an intransitive verb Vintr can feature one of the two agreement scenarios:
– Vintr has the same set of SyntSubj-agreement markers as one of the two agreement marker sets
of Vtrans (e.g., Georgian and Basque, see below);
– Vintr has a special set of SyntSubj-agreement markers, as in Yimas, which has three different
agreement marker sets: for an intransitive SyntSubj, for a transitive SyntSubj, and for a DirO
(Foley 1991). Cf.:
(i) Ama +wa ‘I.go’. ~ Pu +ka +tay ‘I.see.them’. ~ Pu +ŋa +tay ‘They.see.me’.
1.SGSUB go 3.PLOBJ 1.SGSUB see 3.PLSUB 1.SGOBJ see
18 I leave out the agreement with the IndirO (rather than with the DirO), possible with some verbs.
156 3 Syntactic subject: syntactic relations, once again

order. In the least marked transitive clause, with the MV in a tense of the present
series, the actant cross-referenced by Set I markers is in the nominative and pre-
cedes the MV, just as the SyntSubj of an intransitive MV precedes it. The other
actant, which is in the dative, in a communicatively neutral sentence either
follows the MV, or precedes it, while following the nominative actant. Therefore,
the first—nominative—actant is the SyntSubj of the transitive MV, so that Set I
markers must be considered to be subject markers. As a result, the SyntSubj in a
Georgian clause is the element cross-referenced by subject markers; it is boxed in
(30), and the subject markers on the MV are boldfaced.

(30) Georgian (“T.E.” stands for thematic element, a semantically empty suffix
used to form a verbal stem in several tenses19)

a. Kʻac+Ø+i Ø +muša+ob +s ‘[The] man works’. ~

man SG NOM 3SUB work T.E. PRES.3.SGSUB
Kʻac+eb+i Ø +muša+ob+en ‘[The] men work’.
man PL NOM 3SUB work T.E. PRES.3.PLSUB

b. Me v +muša+ob+Ø ‘I work’. ~
I-NOM 1SUB work T.E. PRES.SGSUB
Čven v +muša+ob+t ‘We work’.
we-NOM 1SUB work T.E. PRES.PLSUB

c. Kʻac+Ø +i m +xatʻ +av +s me ~

man SG NOM 1.SGOBJ draw T.E. PRES.3.SGSUB I-DAT
g+xatʻ+av+s šen ~ Ø+xatʻ+av+s mas/mat
2OBJ youSG-DAT 3OBJ he-DAT/they-DAT
‘[The] man draws me ~ youSG ~ him/them’.

d. Kʻac+eb+i m +xatʻ +av +en me ~

man PL NOM 1.SGOBJ draw T.E. PRES.3.PLSUB I-DAT
g+xatʻ+av+en šen ~ Ø+xatʻ+av+en mas/mat
2OBJ youSG-DAT 3OBJ he-DAT/they-DAT
‘[The] men draw me ~ youSG ~ him/them’.

19 As indicated to me by Z. Baratashvili, in Old Georgian the thematic element was linked to

the “punctual aspect ~ durative aspect” opposition, being the marker of the durative. — In our
Georgian examples all verbs are in the indicative mood; to simplify the presentation, IND is not
shown in the glosses.
 3.4 Establishing the syntactic subject in a language 157

e. Kʻac+Ø +ma da +m +xatʻ +a me ~

man SG ERG PERF 1.SGOBJ draw AOR.3.SGSUB I-NOM
da+g+xatʻ+a šen ~ da+Ø+xatʻ+a is/isini
2OBJ youSG-NOM 3OBJ he-NOM/they-NOM
‘[The] man drew me ~ youSG ~ him/them’.

f. Kʻac+eb+ma da +m +xatʻ +es me ~

man PL ERG PERF 1.SGOBJ draw AOR.3.PLSUB I-NOM
da+g+xatʻ+es šen ~ da+Ø+xatʻ+es is/isini
2OBJ youSG-NOM 3OBJ he-NOM/they-NOM
‘[The] men drew me ~ youSG ~ him/them’.

g. Me v +xatʻ +av +Ø mas /mat ‘I draw him/them’.

I-NOM 1SUB draw T.E. PRES.SGSUB he-DAT/they-DAT

h. Me da +v +xatʻ +e +Ø is /isini ‘I drew him/them’.

I-ERG PERF 1SUB draw AOR.1.SGSUB SGSUB he-NOM/they-NOM

The Georgian SyntSubj has five privileges:

– Parameter 1. It depends only on the MV.

– Parameter 2. It is not omissible.
– Parameter 3. In a communicatively neutral sentence, it precedes the MV and
all its other actants.
– Parameter 4:
a. If the transitive MV is in one of the present series tenses, the SyntSubj is in
the nominat­ive: (30a–d); the DirO is in the dative: (30c–d). This is the most
common nominative construction, such as seen in SAE and many other
languages.
If the transitive MV is in an aorist series tense, the case marking of the
SyntSubj and the DirO changes to, respectively, the ergative and the nomi-
native, as in (30e–h), although their syntactic status does not change. A
transitive Georgian MV in an aorist series tense and its two main actants
form, of course, an ergative construction. (Just like Hindi, Georgian mani-
fests split ergativity: the ergative construction appears only with aorist
series tense forms; elsewhere we have the nominative construction.) In
accordance with the convention concerning SyntSubj property violations,
the appearance of the ergative in­stead of the “canonical” nominative does
not make the definition of the SyntSubj in Georgian any more problematic.
b. Georgian has a passive (Harris 1981: 103ff), which confirms the subject-
hood of the ergative SyntSubj:
158 3 Syntactic subject: syntactic relations, once again

(31) a. Gogi+m es stʻatʻia+Ø+Ø /stʻatʻi+eb+i da +Ø +cʻer +a

Gogi ERG this paper SG NOM/paper PL NOM PERF 3SUB write AOR.3.SG
‘Gogi wrote this paper/these papers’.
b. Es stʻatʻia+Ø+Ø /stʻatʻi+eb+i da +cʻer +il +Ø +i
this paper SG NOM /paper PL NOM PERF write PASS.PART SG NOM
Ø +iq+o Gog +is mier
3SUB be AOR.3.SG Gogi GEN by.means.of
‘This paper/These papers was/were written by Gogi’.
NB In Georgian, the MV does not reflect the plural of an inanimate SyntSubj; that is
why dacʻerili ‘written’ and iqo ‘was’ are in the singular for both SyntSubjs ‘paper’ and
‘papers’.

Thus, Georgian has the ergative construction and the ergative case, but it is not
an ergative language.20

20 Two problematic cases of subjecthood in Georgian. In this connection, the evidential and
the affective verbs, known also as “inverse,” should be mentioned.
Georgian has a group of verb forms (currently called “perfect forms,” or “III series forms”),
which carry the meaning ‘by hearsay’ ≈ ‘this being second-hand testimony’ ≈ ‘apparently’ (in
the gloss, ALTR stands for the gram­meme ‘for the other’ of the inflectional category of version):
(i) a. Bičʻ+Ø+s gamo+u +gzavni +xar +Ø šen
boy SG DAT PERF ALTR send-PASS.PART be-PRES.2SUB SGSUB youSG-NOM
‘Apparently, the boy sent youSG’.
b. Bičʻ+Ø+s gamo+u +gzavni +xar +t tkven
boy SG DAT PERF ALTR send-PASS.PART be-PRES.2SUB PLSUB youPL-NOM
‘Apparently, the boy sent youPL’.
c. Bičʻ+Ø+s gamo+u +gzavni +a +Ø gogo+Ø+Ø
boy SG DAT PERF ALTR send-PASS.PART be-PRES.3.SGSUB SGSUB girl SG NOM
‘Apparently, the boy sent [a] girl’.
d. Bičʻ+Ø+s gamo+u +gzavni +a +Ø gogo+eb+i
boy SG DAT PERF ALTR send-PASS.PART be-PRES.3.SGSUB SGSUB girl PL NOM
‘Apparently, the boy sent girls’.
e. Bičʻ+eb+s gamo+u +gzavni +a +t gogo+Ø+Ø
boy PL DAT PERF ALTR send-PASS.PART be-PRES.3.SGSUB PLOBJ girl SG NOM
‘Apparently, the boys sent [a] girl’.
These forms express an evidential, which requires, as is typologically natural, a perfective form.
Traditionally, the noun in the dative is considered to be the SyntSubj. However, judging by ag-
reement, it is the NNOM that is the SyntSubj, the NDAT being an IndirO, which is again typologically
quite plausible. But there is a wrinkle: as seen in (i‑d) and (i-e), in the evidential, the MV does not
agree with the animate SyntSubj of the 3rd person in number as expected, since in (i-d) instead
of *gamougzavni+arian3.PLSUB we have gamougzavni+a3.SGSUB. This minor irregularity, however,
should not change our treatment of the SyntSubj.
Georgian also has a significant class of verbs expressing feelings and attitudes whose basic
diathesis is converse with respect to the corresponding English verbs—that is, ‘I like you’ is in
 3.4 Establishing the syntactic subject in a language 159

Basque. The Basque transitive MV also agrees simultaneously with at least

two of its actants.21 As in Georgian, there are two sets of agreement affixes, the
prefixes INOM, cross-referencing the NNOM, and the affixes IIERG, cross-referencing
the NERG; for the single—nominative—actant of an intransitive MV only the pre-
fixes of set INOM are used. But here comes the important difference with Georgian:
with a transitive MV, one of its two actants is always in the ergative; there is no
tense-induced ergative split—that is, no nominative construction that helps us
identify the Synt­Subj. In Basque, we cannot know which affixes are subjectival.
Therefore, in the following examples, the boldfaced agreement affixes are speci-
fied by the noun they cross-reference: NNOM vs. NERG; for instance, “3NOM” as a gloss
of a marker m means ‘m cross-references the NNOM’, etc. (In the examples of (32)
the clause element that will be eventually identified as SyntSubj is boxed.)

(32) Basque (s = /ś/, tx = /č/, z = /s/)

a. (i) Gizon+a +Ø+Ø etorri d +Ø +a
man DEF SG NOM22 come-PERF.PART 3NOM SGNOM be
‘The man has [lit. ‘is’] come’.
(ii) Gizon+a +k +Ø etorri d +ir +a
man DEF PL NOM come-PERF.PART 3NOM PLNOM be
‘The men have come’.
(iii) Ni+Ø etorri n +aiz
I NOM come-PERF.PART 1NOM be
‘I have come’.

Georgian ‘YouSG are.likable to.me’ (verbs similar to the Georgian feeling verbs exist in Russian,
French, German etc., where, however, they are rather exceptional):
mo+m+cʻon+xar+Ø ‘YouSG are.likable to.me’ = ‘I like you’.
(mo- ⇔ ‘towards.me’, m- ⇔ 1.SGOBJ, cʻon ⇔ ‘be.likable’, ‑xar ⇔ ‘you.are’, -Ø ⇔ SGSUB);
m+i+qvar+s ‘He/They is/are.lovable to.me’. = ‘I love him/them’.
(m- ⇔ 1.SGOBJ, i- ⇔ IPSE ‘for oneself’ [grammeme of version], qvar ⇔ ‘be.lovable’, -s ⇔ 3.SGSUB);
Ø+u+qvar+t ‘He/They is/are.lovable to.them’ = ‘They love him/them’.
(Ø- ⇔ 3OBJ, u- ⇔ ALTR ‘for the other’ [grammeme of version], qvar ⇔ ‘be.lovable’, -t ⇔ PLOBJ);
v+u+qvar+var ‘I am.lovable to.him’. = ‘He loves me’.
(v- ⇔ 1.SGSUB, u- ⇔ ALTR ‘for the other’, qvar ⇔ ‘be.lovable’, -var ⇔ ‘I.am’);
g+ʒul+t ‘He/They is/are.hatable to youPL’. = ‘YouPL hate him/them’.
(g- ⇔ 2OBJ, ʒul ⇔ ‘be.hatable’, -t ⇔ PLOBJ).
In traditional view, the SyntSubj is the NDAT. But, as in the case of the evidential, the MV’s agree-
ment clearly indicates the NNOM as the SyntSubj (with the same complications).
21 Again, for simplicity’s sake, agreement with the IndirO and the so-called allocutive form are
not considered.
22 The morphic structure of a Basque nominal wordform is a controversial topic. Here I am
using my own description.
160 3 Syntactic subject: syntactic relations, once again

b. (i) Gizon+a +Ø +k kotxe+a +Ø+Ø saldu

man DEF SG ERG car DEF SG NOM sell-PERF.PART
d +Ø +u +Ø +Ø
3NOM SGNOM have 3ERG SGERG
‘The man has sold the car’.

(ii) Gizon+a +Ø+k kotxe+a +k+Ø saldu

man DEF SG ERG car DEF PL NOM sell-PERF.PART
d +it +u +Ø +Ø
3NOM PLNOM have 3ERG SGERG
‘The man has sold the cars’.

(iii) Gizon+e +k kotxe+a +Ø+Ø saldu

man PL.DEF ERG car DEF SG NOM sell-PERF.PART
d +Ø +u +Ø +te
3NOM SGNOM have 3ERG PLERG
‘The men have sold the car’.

(iv) Gizon+e +k kotxe+a +k +Ø saldu

man PL.DEF ERG car DEF PL NOM sell-PERF.PART
d +it +u +Ø +zte
3NOM PLNOM have 3ERG PLERG
‘The men have sold the cars’.

(v) Ni+k kotxe+a +Ø+Ø saldu d +Ø +u +t

I ERG car DEF SG NOM sell-PERF.PART 3NOM SGNOM have 1.SGERG
‘I have sold the car’.

(vi) Ni+k kotxe+a +k +Ø saldu d +it +u +t

I ERG car DEF PL NOM sell-PERF.PART 3NOM PLNOM have 1SGERG
‘I have sold the cars’.

In Basque, a transitive MV cross-references its DirO by the same markers as an

intransitive MV cross-references its SyntSubj. For this reason, in Basque, the exis-
tence of actant-shuffling modifications of the verb is crucial. The language has
two such modifications (Rebuschi 1978: 76–77, 82–83; Rebuschi 1981: 92, 1982:
299ff, 1986; Rebuschi’s data are quoted with drastic sim­plifications): a passive
and two detransitivizations, which target the SyntSubj’s syntactic position.
Passive: a diathetic permutation “DirOʹ ⇒ SyntSubj, SyntSubjʹ ⇒ AgCo” [the
prime means ‘initial’].
The Basque passive is illustrated in (33), where the sentences semantically
correspond to the sentences in (32b):
 3.4 Establishing the syntactic subject in a language 161

(33) (i) Kotxe+a +Ø+Ø gizon+a +Ø+k saldu +a +Ø

car DEF SG NOM man DEF SG ERG sell-PERF.PART DEF SG
d +Ø +a
3NOM SGNOM be
‘The car is sold by the man’.
(ii) Kotxe+a +k+Ø gizon+a +Ø+k saldu +a +k
car DEF PL NOM man DEF SG ERG sell-PERF.PART DEF PL
d +ir +a
3NOM PLNOM be
‘The cars are sold by the man’.
(iii) Kotxe+a +Ø+Ø gizon+e +k saldu +a +Ø
car DEF SG NOM man DEF.PL ERG sell-PERF.PART DEF SG
d +Ø +a
3NOM SGNOM be
‘The car is sold by the men’.
(iv) Kotxe+a +k+Ø gizon+e +Ø+k saldu +a +k
car DEF PL NOM man DEF SG ERG sell-PERF.PART DEF PL
d +ir +a
3NOM PLNOM be
‘The cars are sold by the men’.
(v) Kotxe+a +Ø +Ø ni+k saldu +a +Ø d +Ø +a
car DEF SG NOM I ERG sell-PERF.PART DEF SG 3NOM SGNOM be
‘The car is sold by me’.
(vi) Kotxe+a +k+Ø ni+k saldu +a +k d +ir +a
car DEF PL NOM I ERG sell-PERF.PART DEF PL 3NOM SGNOM be
‘The cars are sold by me’.

The auxiliary verb ‘be’ and the participle agree—in definiteness and number—
only with the subject.

Detransitivizations: they result in “SyntSubjERG ⇒ SyntSubjNOM” 23

23 The category of transitivization. DETRANS(itivizer) is a grammeme of transitivization, an in-

flectional category of the verb—similar to, but different from, voice. It resembles voice in that
it impacts the verb central actants, the SyntSubj and the DirO; it differs from voice in that it
does not permute the DSyntAs of the verb with the respect to its SemAs, but only modifies their
surface realization (see Mel’čuk 2006a: 231ff). This category includes at least three grammemes:
NEUTRAL ~ DETRANS ~ TRANS(itivizer). Tongan, examples (21)–(22), features the pair NEUTRAL ~
TRANS. Chukchi has even two detransitivizers: DETRANS-1 and DETRANS-2. DETRANS-1, expressed
162 3 Syntactic subject: syntactic relations, once again

Basque has two detransitivizations: a progressive construction and a resulta-

tive construction.

– The progressive construction is marked by the adjective ari ‘being in the

process of, doing’ and uses the auxiliary izan ‘be’ even for transitive verbs,
which become eo ipso intransitive (since a transitive verb uses as its auxiliary
only ukan ‘have’): the SyntSubj, instead of the ergative, takes the nomina-
tive, as an intransitive SyntSubj should; the former DirO remains in the nomi-
native, but loses its status as a DirO, since the verb becomes intransitive; the
MV agrees only with the SyntSubj:

by the prefix ine-/ena‑, lowers the Synt-rank of the DirO (which becomes an IndirO); DETRANS-2
(the suffix -tku/-tko) not only lowers the Synt-rank of the DirO, but it also makes its appearance
in the clause undesirable and, at the same time, blocks the expression of all other objects and
complements, which are allowed both with the basic form and with the DETRANS-1 form.
(i) Chukchi
a. Γəm+nan tə +ret +ərkən+Ø kimitʕ+ən (tomγ+etə)
I INSTR 1.SGSUB transport PRES 3.SGOBJ load SG.NOM friend SG/PL.DAT
‘IX⇔I transport a.loadY⇔II (to.a.friend/to.friendsZ⇔III)’.
b. Γəm+Ø t +ine +ret +ərkən kimitʕ+e (tomγ+etə)
I NOM 1.SGSUB DETRANS-1 transport PRES load SG.INSTR friend SG/PL.DAT
‘IX⇔I transport a.loadY⇔II (to.a.friend/to.friendsZ⇔III)’.
c. Γəm+Ø tə +ret +ətku +rkən (?kimitʕ+e ?
tomγ +etə)
I NOM 1.SGSUB transport DETRANS-2 PRES load SG.INSTR friend SG/PL.DAT
‘IX⇔I transport (a.loadY⇔II) (to.a.friend/to.friendsZ⇔III)’.
Sentence (i-a) presents an ergative construction, obligatory in Chukchi for any transitive verb:
the SyntSubj ‘I’ is in the instrumental, and the DirO ‘[a] load’, in the nominative. In (i-b), we find
a nominative construction, possible only for an intransitive verb: the SyntSubj, which remains
‘I’, is in the nominative; the DirO ‘[a] load’ has become an OblO in the instrumental, thus losing
its salience; the two OblOs are optional. Finally, (i-c) is again a nominative construction: the two
OblOs—‘load’ and ‘friends’—are incompatible with each other and even less salient than in the
preceding sentence; their omission is preferred.
Roughly, sentence (i-a) answers the question ‘What are you transporting and to whom?’, (i-
b), the question ‘What are you doing?’, and (i-c), the question ‘What is your occupation?’
Degrees of transitivization/detransitivization, related to the degree of the impact of the deno-
ted action upon the object, are not a rarity; here is another example—from Warlpiri (Australia):
(ii) Warlpiri
a. Maliki+ḷi ka +Ø +Ø ŋarka+Ø yaḷki+ṇi
dog ERG PRES 3.SGSUB.3.SGOBJ NEUTRAL man NOM bite NON-PAST
‘The dog is biting the man’.
b. Maliki+ḷi ka +ḷa +ǯinta ŋarka+ku yaḷki+ṇi
dog ERG PRES 3.SGSUB.3.SGOBJ DETRANS man DAT bite NON-PAST
‘The dog is biting at the man’.
In Warlpiri, DETRANS lowers the transitivity of the verb, turning its DirO into an IndirO; but the
verb remains transitive: it still presents an ergative construction, with the SyntSubj in the erga-
tive case.
 3.4 Establishing the syntactic subject in a language 163

(34) a. (i) Gizon+a +Ø +Ø kotxe+a +Ø/k +Ø saltzen ari d +Ø +a

man DEF SG NOM car DEF SG/PL NOM sell-GER doing 3NOM SGNOM be
‘The man is selling the car/s’.
(ii) Gizon+a +k +Ø kotxe+a +Ø/k +Ø saltzen ari d +ir +a
man DEF PL NOM car DEF SG/PL NOM sell-GER doing 3NOM PLNOM be
‘The men are selling the car/s’.

– The other detransitivization (called “antipassive” in Rebuschi 1981: 92) pro-
duces a result­ative construction, in which not only the auxiliary, but also the
participle agrees with the SyntSubj (in the nominative) according to defi-
niteness and number; the MV is also intransitive, so that the “former” DirO
becomes a Quasi-DirO:

b. (i) Gizon+a +Ø +Ø kotxe+a +Ø/k+Ø saldu +a +Ø

man DEF SG NOM car DEF SG/PL NOM sell-PAST.PART DEF SG
d +Ø +a
3NOM SGNOM be
‘The man is having.sold the car/s’.

(ii) Gizon+a +k +Ø kotxe+a +Ø/k+Ø saldu +a +k

man DEF PL NOM car DEF SG/PL NOM sell-PAST.PART DEF PL
d +ir +a
3NOM PLNOM be
‘The men are having.sold the car/s’.

The four Basque SyntSubj’s privileges are as follows:

– Parameter 1. It depends exclusively on the MV.

– Parameter 2. It is non-omissible.
– Parameter 3. It tends to precede the MV and other actants.
– Parameter 4b. Its role is targeted by the passive and is confirmed by detransi-
tivizations.

In conclusion, Basque is a non-ergative language, but it does have an ergative

construction (without split) and an ergative case.

Subtype 3b. In L the MV can agree simultaneously with two actants, and a mono-
actantial MV features both types of agreement.
Probably the best-known example here comes from Acehnese (Malayo-Poly-
nesian).
164 3 Syntactic subject: syntactic relations, once again

Acehnese (Durie 1985, 1987, 1988). Acehnese has no syntactic processes: no

voices, no raisings, no detransitivization, no switch-reference, etc.; word order
is extremely flexible. The only reliable syntactic property of actants of the Main
Verb that amounts to a privilege is verb agreement—cross-referencing of actants
on the MV. The Acehnese MV cross-references two of its actants (only if they are
animate): one by a prefixal marker, the other by a suffixal marker. However, in the
simplest clauses, which feature a semantically monoactantial verb having just
one syntactic actant, both types of agreement occur, which means that in (35a)
and (35b), featuring intransitive verbs, we see two different types of actant—one
controlling prefix agreement and the other controlling suffix agreement.

(35) Acehnese (ê = /e/, ô = /o/, eu = /ɯ/, ë = /ɤ/, j = /ǯ/)

a. Lôn+lôp ‘I enter’. and Geu+lôp ‘He enters’.

1.SG enter 3.SG enter

b. Rhët+lôn ‘I fall’. and Rhët+geuh ‘He falls’.

fall 1.SG fall 3.SG

Thus, both these types of actant are privileged in Acehnese, since they, and only
they, control the agreement of the MV. Therefore, one of these actants must be the
SyntSubj and the other, the DirO. To decide which one of the two is more privi-
leged than the other and thus is the SyntSubj, we need to consider a biactantial
verb in a sentence where both types of actant are expressed:

c. Lôn+ngieng+geuh ‘I see him/her’.

1.SG see 3.SG
and Geu+ngieng+lôn ‘He/She sees me’.
3.SG see 1.SG

Examining sentences with two privileged syntactic actants simultaneously

present, we find that:

– The prefixal marker on the verb is obligatory and cannot be linearly sepa-
rated from the verb (35d-i), while the suffixal marker is not obligatory and
can migrate to the outer edge of the verbal phrase (35d-ii):

d. (i) Gopnyan lôn+ngieng ‘Him I.see’.

he 1.SG see
and Lôn geu+ngieng ‘Me he.sees’.
I 3.SG see
 3.4 Establishing the syntactic subject in a language 165

vs. *Lôn ngieng+geuh ‘I see.him’.

I see 3.SG
and *Gopnyan ngieng+lôn ‘He sees.me’.
he see 1.SG

(ii) Ka +leupah+lôn u keude baroe. ≡ Ka+leupah u keude baroe+lôn.

PAST reach 1.SG to town yesterday  ‘I reached the town yesterday’.

– The imperative requires the prefixal marker and does not allow the suffixal
one:

e. (i) Neu+peumeu’ah! ‘Forgive me!’ ~ *Peumeu’ah! ~ *Neu+peumeu’ah+lôn!

2.SG forgive
(ii) Neu+peujêt ie nyoe keu jih! ‘Make him drink this water!’
2.SG make.drink water this to he

To put it differently, only the prefix-referenced actant can be the Addressee of an

imperative utterance.

– The prefix-referenced actant, and only this actant, can be introduced by the
preposition lê, when following the Main Verb:

f. Gopnyan lôn+tët +rumoh lê lôn ‘I burned down his house’. =

he 1.SG burn house I lit. ‘He I.burned.house by I’.

Let me add, as icing on the cake, that only the prefix-referenced actant con-
trols its own obliga­tory Equi-Deletion with the verb TÊM ‘want’, no matter whether
the governed verb is intran­sitive or transitive:

g. (i) Gopnyan geu+têm jak. ~ *Gopnyan geu+têm geu+jak ‘He wants to go’.
he 3.SG want go
and Gopnyan geu+têmtaguen bu. ~
he 3.SG want cook rice
*Gopnyan geu+têm geu+taguen bu ‘He wants to cook rice’.
vs. (ii) *Gopnyan geu+têm rhët ‘He wants to fall’.
he 3.SG want fall
The sentence (35g–ii) is incorrect since the verb rhët requires the suffix-referencing of
its only actant.
166 3 Syntactic subject: syntactic relations, once again

Therefore, the prefix-referenced actant is more privileged in Acehnese: it is

the SyntSubj. The other one, suffix-referenced, is the DirO. This simply means that
in (35b) a literal gloss should be rather ‘It.falls me/him’. M. Durie himself does not
name these two actants in this way: he calls them Agent and Undergoer, since 35
years ago the notions of SyntSubj and SyntObj were too vague to be of any use;
Durie 1985: 190–191 correctly indicates that none of Acehnese clause elements
corresponds to the characteristics of the “syntactic pivot,” a moot concept used
at the time instead of SyntSubj. However, Durie makes it absolutely clear that
“Agent” and “Undergoer” are not genuine semantic relations, but clearly syntac-
tic ones (see especially Durie 1987). There­fore, it can be safely concluded that, by
calling the prefix-referenced actant the SyntSubj and the suffix-referenced one
the DirO, I simply sharpen and, at the same time, generalize the terminology.
What is special about the Acehnese SyntSubj and DirO is their more direct
link to semantic roles. In many languages such as English or Russian, a SyntSubj
can fulfill various semantic roles: it can express an Agent (John beat up Paul.),
a Patient (John got a beating.), a Cause (John really worries us.), an Experiencer
(John likes boiled potatoes.), a Property Carrier (John is intelligent.), Time (The
next morning saw John in Nevada.), and so on; to a lesser extent, the same is true
of the DirO. But in Acehnese, the SyntSubj expresses only the volitional Actor,
and the DirO only the non-volitional Undergoer. The semantic opposition of voli-
tionality is extremely important in Acehnese; the language has special deriva-
tional means to change the volitionality of a verb (Durie 1988: 7): jak ‘go, walk’
~ teu+jak ‘walk without volition’ or seunang ‘be happy’ ~ meu+seunang ‘make
oneself happy, enjoy oneself’. However, such an alignment of syntactic relations
to semantic roles by no means diminishes the importance of syntactic relations.
With the proposed terminological change, one can draw an interesting paral-
lel between the Acehnese sentences of the (35b) type—that is, sentences with a
verb that has a DirO only, but no SyntSubj—and Russian impersonal construc-
tions, in which the only semantic actant of the verb is expressed by a DirO (the
verb in these constructions expresses an incontrollable state):

(36) MenjaACC tošnit/rvët lit. ‘[It] nauseates/vomits me’. = ‘I am nauseated/I vomit’.

MenjaACC znobit lit. ‘[It] chills me’. = ‘I have a chill’.
MenjaACC trjasët lit. ‘[It] shakes me’. = ‘I shake’.
MenjaACC proneslo lit. ‘[It] diarrhea-ed me’. = ‘I had diarrhea’.
MenjaACC skrjučilo lit. ‘[It] completely.bent me’. = ‘I was doubled up [in pain]’.
MenjaACC razneslo lit. ‘[It] expanded me’. = ‘I got fat’.

I do not see any substantive difference between Acehnese Sakêt-lôn lit. ‘[It]
hurts/sicks me’. = ‘I am hurting/sick’, which is an impersonal construction, and
 3.5 Syntactic subject problems related to impersonal construction 167

the Russian impersonal construction of the type Menja lixoradit lit. ‘[It] fevers
me’. = ‘I have fever’. The difference is quantitative: Russian has a handful of such
impersonal verbs, while in Acehnese there are hundreds of them.24

3.5 S
 yntactic subject problems related to impersonal
constructions
On several occasions, a dubious treatment of an actant as the SyntSubj is due to
the failure to recognize the presence of a zero dummy subject, a lexeme similar to
the expletive and meteoro­logical it of English, but having an empty signifier. Let
me consider two cases, in Icelandic and in Amele.
Icelandic (Andrews 2001). Icelandic has a common type of sentences of the
form in (37):

(37) Icelandic
a. Bát +Ø +inn /Bát+a +na rak á land
boat SG.ACC DEF /boat PL.ACC DEF drift-PAST.3.SG to shore
lit. ‘[It] drifted the.boat/s to shore’. = ‘The boat/s drifted to shore’.

b. Bát +i +num /Bát +u +num hvolf +di

boat SG.DAT DEF /boat PL.DAT DEF capsize PAST.3.SG
lit. ‘[It] capsized the.boat/s’. = ‘The boat/s capsized’.

c. (i) Hann kasta+ði stein +i +num/stein+u +num

he-NOM throw PAST.3.SG stone SG.DAT DEF /stone PL.DAT DEF
‘He threw with.the.stone/s’.
(ii) Stein+i +num/Stein+u +num var kasta+ð
stone SG.DAT DEF /stone PL.DAT DEF be-PAST.3.SG throw PAST.PART
‘The stone/s were thrown’. lit. ‘[It] was thrown with.the.stone/s’.

According to Andrews 2001, the boldfaced element in the sentences of (37) is

the SyntSubj, since its behavior shows at least 13 features that it shares with the

24 There is also a semantic difference, irrelevant in the present context: in Russian, such
verbs denote mostly harmful or at least unpleasant physiological states and processes, while
in Acehnese they cover a much larger area of non-volitional properties, states, events, and pro-
cesses. A formal difference should also be mentioned: the Russian impersonal construction
has a dummy zero SyntSubj, which imposes on the verb the agreement in 3.SG.NEU(ter), while
Acehnese has no dummy subject (Durie 1985: 180), since the verb does not require automatic
subject agreement.
168 3 Syntactic subject: syntactic relations, once again

behavior of the “canonical” SyntSubjs of Icelandic: it controls coreference with

the “subject” of an infinitive and the choice of the reflexive possessive pronoun
sinni ‘self’s’ (Rus. svoj), it can appear between an auxiliary and the past participle
of the lexical verb (where only SyntSubjs are admitted), etc. However, “not only
are they not nominative in case, but the verb does not agree with them” (Andrews
2001: 93), while normal SyntSubjs in Icelandic control the agreement of the MV
and are marked by the nominative. Therefore, I conclude that these suspicious
clause elements are not SyntSubjs—even though they behave in many respects as
prototypical SyntSubjs sometimes do under specific conditions. Otherwise, it is
not clear what Andrews and many others who share his perspective on this issue
understand by a subject: by all means, not a clause element that is the depending
mem­ber of a particular SSynt-relation.
In reality, the sentences in (37a–b) and (37c-ii) manifest an impersonal con-
struction with a zero subject: in (37a–b), this is the lexeme Ø«ELEMENTS»
(3, sg) , denoting
some slightly mysterious natural forces; in (37c-ii), this is the zero dummy subject
lexeme Ø(3, sg), which is semantically empty. These zeroes are equivalent to Eng.
it, Ger. es and Fr. il. (Spanish and Russian also have, in such contexts, a zero
dummy: for instance, Sp. Se lee muchas novelas lit. ‘[It] itself reads many novels’
and Rus. Zdes′ mnoj siženo ‘Here by.me [it is] sat’.) The correct glossing of (37a),
(37b) and (37c-ii) would be ‘It drifted the boat/s to shore’, ‘It capsized the boat/s’
and ‘It was thrown with the stone/s’. That is exactly how all these constructions
are described in an elementary manual of Icelandic for non-natives (Glendenning
1983: 49–50).
As for the SyntSubj’s coreferential deletion in Icelandic, it does not tell us
anything about the subjecthood of the boldfaced clause elements:

(38) Icelandic (Pouplier 2003: 367)

a. Þeim líkar maturinn og borða mikið

they-DAT please-PRES.3.SG food-SG.NOM.DEF and eat-PRES.3.PL much
lit. ‘To.them pleases the food and [they] eat much’. =
‘They like the food and eat much’.

b. Okkur vantaði peninga og vorum svangir

we-DAT lack-PAST.3.SG money-PL.ACC and be-PAST.1.PL hungry
lit. ‘To.us [it] lacked money and [we] were hungry’. =
‘We lacked money and were hungry’.

In (38), the deletion of the SyntSubj in the second coordinate clause is controlled
by semantic properties of the controlling element (a human Experiencer) rather
than by its syntactic nature.
 3.5 Syntactic subject problems related to impersonal construction 169

Amele (Roberts 1987, 1988, 2001). In Amele, the MV can simultaneously agree
with four types of actant (quadri-personal agreement). Agreement affixes are differ-
ent for each type of actant; the agreement in the simplest clauses—an intransitive MV
and only one actant—allows the researcher to establish the Subject Agreement affix
set and thus to identify the SyntSubj without problems: the SyntSubj in an Amele
sentence is the noun that imposes the use of these particular agreement affixes.
A problem concerning the SyntSubj in Amele comes from the category of
switch-reference: in case a sentence includes two (or more) verbs, the preceding
being subordinated to the following (V1←synt–V2), a switch-reference grammeme
on V1 is supposed to indicate whether V2 has a SyntSubj referentially identical to
that of V1. (For instance, in John came in and sat down both verbs have the same
SyntSubj; in John came, and I sat down the verbs have different SyntSubjs.) Cf. (39):

(39) Amele (j = /ǯ/, q = /gb͡/)

a. Ija hu +f +ig mad +ig +en
I come if-SAME-SUB 1.SGSUB speak 1.SGSUB FUT
‘If I come, [I] will.speak’.
vs. Uqa ho +oʔ +b fi ija mad +ig +en
he come if-DIF-SUB 3.SGSUB if I speak 1.SGSUB FUT
‘If he comes, I will.speak’.
b. Ege wen +Ø +g +en
we hunger give 1.PLOBJ 3.SGSUB.REMOTE.PAST
‘We became hungry’. lit. ‘[It] us hunger gave’.
c. Ege ʔo +ʔob +ob
we REAL-GER walk 1.PLSUB.SAME-SUB
wen +Ø +g +en
hunger give 1.PLOBJ 3.SGSUB.REMOTE.PAST
lit. ‘We walking, [it] us hunger.gave’. = ‘As we walked, we became hungry’.25

25 Two interesting details about Amele are worth mentioning.

– The semi-auxiliary (≈ light) verb ‘give’ has a zero stem, so that its observable forms consist
solely of suffixes; see Mel’čuk 2006a: 474–475.
– The meaning ‘want to Y’ is expressed in Amele by means of an impersonal construction with
the light verb ‘give’ and the imperative form of Y; the literal rendering of ‘X wants to Y’ is ‘It
gives to.X: «Y!»’ (Roberts 2001: 221):
(i) a. Ija ma j +ag +a Ø +t +ena ‘I want to eat taro’. =
I taro eat 2.SGSUB IMPER give 1.SGOBJ.3.SGSUB PRES lit. ‘[It gives] me «EatSG taro!»’.
b. Ege ma j +eig +a Ø +g +ena ‘We want to eat taro’. =
we taro eat 2.PLSUB IMPER give 1.PLOBJ.3.SGSUB PRES lit. ‘[It gives] us «EatPL taro!»’.

Both sentences in (i) have a dummy zero subject.

170 3 Syntactic subject: syntactic relations, once again

d. Eu jagel November na uqa odo+ʔo +b

this month in he do DIF-SUB 3.SGSUB
ʔul +g +en
leave 1.PLOBJ 3.SGSUB.REMOTE.PAST
lit. ‘This in November he had.done, left.[it.]to.us.he’.
[Roberts 1987: 304, (620)]

Now, Amele has an impersonal construction, which expresses physiological

and psychological states of a person; the construction has a dummy zero subject
Ø(3, sg) with which the verb agrees; this is shown by the Subject agreement gram-
memes 3, SG on the verb: see (39b). The Experiencer appears as DirO (also identified
by verb agreement), and the designation of the state itself (a noun or an adjective)
is a quasi-object, very much like the quasi-object noun in Persian verbal colloca­
tions; it is not cross-referenced on the verb.26 As a result, what is found in (39b)
seems to be an unproblematic construction similar to Russian impersonal con-
structions of the type NasDirO trjasët lit. ‘[It] shakes us’. = ‘We are shaking’ or NasDirO
klonit v son lit. ‘[It] pushes us into sleep’. = ‘We are sleepy’, with a dummy zero
SyntSubj, cf. above, (12) and (36). So far, so good. But in a two-clause sentence,
such as that in (39c), the verb of the first clause, where the SyntSubj is ege ‘we’, is
marked as having the same SyntSubj as the verb of the second clause, while this
latter has a dummy zero subject. This fact makes Roberts remark that, although
ege ‘we’ in the second clause is a DirO, it has some SyntSubj properties, in the
first place—controlling the feature “same/ different subjects” (Roberts 2001: 204).
But why do we have to say that the suffix ‑ob indicates the same SyntSubj in the
next clause? Roberts states himself (1988) that the switch-reference in Amele may
track the sequence of Themes (“same Theme/different Theme”) rather than that
of SyntSubjs. The detailed examples given in Roberts 1987: 292–305 also point to
rather semantic character of Amele switch-reference: thus, in (39d), the SyntSubj

26 These nominal-verbal collocations are also known as complex, or periphrastic, verbs.

Here are a few Persian examples (Samvelian 2012; the quasi-direct-object is in boldface):
(i) Maryäm otaq+raDirO ǯaru zäd
lit. ‘Myriam room broom hit’. = ‘Myriam swept the room’.
(ii) Maryäm Omid+raDirO šekast däd
lit. ‘Myriam Omid defeat gave’. = ‘Myriam defeated Omid’.
(iii) Maryäm Omid+raDirO dust dār+äd
lit. ‘Myriam Omid friend has’. = ‘Myriam loves Omid’.
(iv) Maryäm Omid+raDirO gušt kärd
lit. ‘Myriam Omid ear did’. = ‘Myriam listened to Omid’.
The collocations of this type are extremely widespread in Persian: in fact, most verbs of SAE
languages correspond to V+N collocations in Persian.
 3.6 A difficult case: the syntactic subject in Lushootseed 171

is, of course, the same, but the marker of DIF-SUB signals the change of world
setting—a new situation obtains. Therefore, if we accept that switch-reference in
Amele marks the preservation/change of Themes (or maybe of situations), the
problem disappears: it suffices to replace the names of grammemes SAME-SUB
and DIF-SUB in (39c–d) by SAME-THEME and DIF-THEME.

3.6 A difficult case: the syntactic subject in Lushootseed

However, things in general are not as beautiful and well-behaved as the preced-
ing exposition might imply. Let us consider Lushootseed, which has repeatedly
been claimed not to have a SyntSubj, but—as Beck 1996 and 2000 convincingly
argued—has one after all, although not with­out a complication.
Lushootseed. Lushootseed is a “kind of” Type 2 language—it features
monopersonal agree­ment, but only for bi-actantial (≈ transitive) verbs: its intran-
sitive verbs do not show agreement with its actant. Therefore, Lushootseed cannot
be considered in parallel with genuine monopers­onal agreement languages. Here
is a relevant dataset for Lushootseed (borrowed from Beck 2000).

(40) Lushootseed (PUNCT stands for ‘punctual aspect’)

a. An intransitive verb
(i) ʔu +ʔəƛʻ čəd ‘I came’.
PUNCT come I
(ii) ʔu +ʔəƛʻ čəxw ‘YouSG came’.
PUNCT come youSG
(iii) ʔu +ʔəƛʻ ti čʻačʻas ‘The child came’.
PUNCT come the child
(iv) ʔu +ʔəƛʻ Ø(3rd person) ‘He/She/They came’.
PUNCT come he/she/they

The only actant of an intransitive verb is expressed by a second-position pronom-

inal clitic, which is a zero lexeme for the 3rd person (singular and plural).

b. A transitive verb (LC stands for ‘limited control’)

(i) ʔu +ʔəyʻ+dxw+Ø čəd ti čʻačʻas
PUNCT find LC ACT I the child
‘I found the child’.
(ii) ʔu +ʔəyʻ+dxw+Ø čəxw ti čʻačʻas
PUNCT find LC ACT youSG the child
‘YouSG found the child’.
172 3 Syntactic subject: syntactic relations, once again

(iii) ʔu +ʔəyʻ+dxw+Ø Ø(3rd person) ti čʻačʻas

PUNCT find LC ACT he/she/they the child
‘He/She/They found the child’.
(iv) ʔu +ʔəyʻ+du+Ø +bš ti čʻačʻas
PUNCT find LC ACT 1.SGOBJ the child
‘The child found me’.
(v) ʔu +ʔəyʻ+du+Ø +bicid ti čʻačʻas
PUNCT find LC ACT 2.SGOBJ the child
‘The child found youSG’.
(vi) ʔu +ʔəyʻ+du+Ø +bš čəxw
PUNCT find LC ACT 1.SGOBJ youSG
‘YouSG found.me’.
(vii) ʔu +ʔəyʻ+du+Ø +bicid čəd
PUNCT find LC ACT 2.SGOBJ I
‘I found.youSG’.
(viii) ʔu +ʔəyʻ+du+Ø čəd/čəxw Ø(3rd person)
PUNCT find LC ACT I /youSG he/she/they
‘I/YouSG found him/her/them’.

The meaning ‘The child/I/YouSG found him/her/them’ is impossible to express

in Lushootseed as is, since there is no physical (= non-zero) pronominal clitic
expressing the 3rd person object; and a sentence of the form in (40b-iii) means
‘He/She/They found the child’. The only way to verbalize the meaning ‘The child
found him/her/them’ is to turn the Main Verb into the passive:

(ix) ʔu +ʔəyʻ+du+b Ø(3rd person) ʔə ti čʻačʻas

PUNCT find LC PASS he/she/they by the child

The SyntSubj in Lushootseed has five syntactic privileges out of six privileges
possible (since the language has no cases, case marking—Parameter 4a—is irrel-
evant):

– Parameter 1. The SyntSubj depends exclusively on the MV.

– Parameter 2. It is non-omissible from the SyntS of the sentence.
– Parameter 3. It follows the MV, preceding another prepositionless actant,
which is a DirO.
– Parameter 4b. Its syntactic role is targeted by the passive:

(41) ʔu +ʔəyʻ+dxw+Ø čəd ti čʻačʻas ‘I found the child’.

PUNCT find LC ACT I the child
3.7 The syntactic subject: its syntactic role vs. its semantic and communicative roles 173

vs. ʔu +ʔəyʻ+du+b čəd ʔə ti čʻačʻas ‘I was.found by the child’.

PUNCT find LC PASS I by the child

Parameter 6. It is the only actant that can be expressed by special subject clitics.
However, it is not the SyntSubj that controls the person-number agreement of
the Lushootseed bi-actantial (i.e. transitive) MV, but the DirO, and this only if the
DirO is of the 1st or 2nd person (cf. Awa Pit, example (4), p. 127). This fact creates
an additional difficulty for the declared principle that the MV, if it agrees only
with one of its actants, must agree with the SyntSubj, Subsection 3.4.2, p. 149; but
otherwise, it does not undermine our approach to the definition of the SyntSubj
as a cross-linguistic phenomenon.

3.7 T
 he syntactic subject: its syntactic role vs. its semantic
and communicative roles

The problem of defining the SyntSubj has arisen in part as a result of the failure
to strictly separate, on the one hand, the purely syntactic properties that define a
syntactic element of the clause, and, on the other hand, some semantic and com-
municative properties characterizing that element. It is true that syntactic clause
elements encode—in the ultimate analysis—semantic roles of the correspond­ing
meanings and are tightly controlled by communicative factors. This, however, is
not a reason for abandoning syntactic relations; and, by all means, this is impos-
sible. We should simply keep in mind that in some languages the alignment of
syntactic relations to semantic roles is very intricate; thus, in English, as illus-
trated above, a SyntSubj can correspond to a large variety of semantic roles. In
other languages such alignment is more straightforward: thus, in a basic clause of
Archi the SyntSubj cannot be an Agent, an Experiencer or a Cause. But even if in
some cases there is a one-to-one correspondence between syntactic and semantic
roles, this should not lead to confusing them (remember that a one-element set
is essentially different from an element). In some languag­es, the correspondence
between syntactic, semantic and communicative roles is close to one-to-one.
Thus, speaking of Lushootseed, Beck (2000: 310) states “that although there is
an unusually close ‘fit’ between the semantic structure of an utterance and the
syntactic role that each particip­ant … is assigned by the grammar, this fit is not
one-hundred percent and so the invo­cation of a syntactic role … seems justified.”
This close fit is not at all astonishing: the SyntSubj as the most privileged syn-
tactic actant tends to express the most privileged semantic role of Agent and the
most privileged communicative role available to a nominal—that of the Theme,
which in its turn, tends to be Given, referential and definite.
174 3 Syntactic subject: syntactic relations, once again

3.8 The direct object

Having dealt with the SyntSubj, I can briefly turn to the DirO—in order to round
up my presentation and to show how the proposed parameters work on a larger
scale. As the reader was warned, this section is limited to a bare minimum and
does not include examples (for good linguistic data concerning the DirO, see
Plank, ed. 1984). Without further ado, I will sketch below a universal definition
of Direct Object.

Definition 3.2 – direct object

The direct object is the second most privileged SSynt-actant of a transitive

verb in a non-er­gative L; what exactly are syntactic privileges in L has to be
indicated by a list of specific DirO privileges (= properties) elaborated for L.

It is immediately clear that the DirO and, respectively, the direct-objectival SSyn-
tRel are not cross-linguistically universal: according to our definition of ergative
language, they are present only in non-ergative languages (barring some excep-
tional, derived constructions, such as, for instance, the progressive construction
in Archi, p. 154) and only in clauses with a transitive MV. The DirO’s syntactic
privileges are determined using the same definitional parameters as those of the
SyntSubj, but, of course, by different values thereof:

1. The DirO exclusively depends on a verb (the finite form, the infinitive/masdar,
the parti­ciple, the converb), but not necessarily on the MV.
2. The DirO can be non-omissible from the syntactic structure of the clause.
3. The DirO’s linear position is specified with respect to the governing verb and/
or with respect to its other actants.
4a. The DirO can have morphological impact on the verb (= agreement LDirO–
morph→V): in many languages, a transitive verb agrees with its DirO.
4b. The DirO is involved in actant shuffling: as a result of the verb’s inflection,
the DirO can be promoted to SyntSubj status (the SyntSubj being demoted to
Agent Complement or Oblique Object).
5. The verbs can also have morphological impact on the DirO (= government
LDirO←morph–V); the DirO is quite often marked by a special case: as a rule,
the accusative or the nominative (the latter, in an ergative construction).
6. The DirO can pronominalize in a particular way.
 3.9 Summing up 175

3.9 Summing up
The conclusions concern two main aspects of the discussion: establishing and
defining particular SSyntRels in particular languages (3.9.1) and their cross-lin-
guistic universality (3.9.2).

3.9.1 Defining surface-syntactic relations

A SSyntRel r in the phrase Li–r→L of language L is defined exclusively by a set of

parame­ters concerning the following properties of Li and L:

1) the possible syntactic types of Li;

2) omissibility/non-omissibility of L from the Sem- and Synt-structures;
3) possible linear arrangements of Li and L;
4) L’s morphological properties;
5) Li’s morphological properties;
6) prosodic properties of the of the Li–r→L phrase.

In this way, all available linguistic expressive means are taken into account. To
put it differ­ently, any SSyntRel r is defined strictly by statements necessary to
implement the abstract phrase Li–r→L in an utterance.
A list of parameters necessary for the definition of SyntSubj—that is, of
the subjectival SSyntRel—is presented in Subsection 3.2.3 above. Each SSyntRel
requires, of course, its own set of definitional parameters: thus, prosody is irrel-
evant for SyntSubj, but it is important for different appositive SSyntRels in Russian
(gorod-sad ‘city [which is a] garden’ vs. gorod Sad ‘city [which is named] Sad’).
All other properties of a SSyntRel (and of its depending and governing
members) are characterizing, or descriptive: they specify the syntactic behavior
of the SSyntRel r in language L, r itself being previously established based on its
definitional properties.

3.9.2 Cross-linguistic universality of particular surface-syntactic relations

Surface-syntactic relations as a linguistic phenomenon are cross-linguistically

universal in the following sense: they are present in all languages, appear in all
sentences of a language, and involve all lexical units in a sentence.
The universality of SSyntRels is similar to the universality of parts of speech
or phonemes (Dryer 1997: 116–119). Major word classes, known as parts of speech,
necessarily exist in any language, but this, of course, does not mean that lan-
guage L1 has the same parts of speech as L2: some languages do not have a class
176 3 Syntactic subject: syntactic relations, once again

of adverbs, some others lack adverbs and adjectives. Now, all languages seem
to have verbs and nouns, but this is simply due to the universalist definitions of
‘verb’ and ‘noun’: verbs are lexemes that can, without special modification, fulfill
the role of the syntactic head of a clause; nouns are lexemes that can, without
special modification, fulfill the role of actants of a verb. Here the similarity
between the parts of speech of different languages ends: a semantic equivalent of
a noun of L1 is not necessarily a noun in L2, since, for instance, L1 can allow for
event and property nouns (such as ‘arrival’ and ‘beauty’), while in L1 such mean-
ings are implemented exclusively by verbs. The same is true about the verbs, etc.
An actual part of speech of L1 must be defined strictly within L1, but this does
prevent L2 from having the same part of speech, albeit with different elements!
Thus, nouns of L1 do not always semantically correspond to nouns in L2 (and vice
versa), but there is a heavy overlap, and in both languages a noun can fulfill the
same general syntactic role—be an actant of a verb.
Another instructive parallel with a particular SSyntRel is the nominative case.
It is the case of nomination, and with such a universalist definition, any language
that has cases has a nominative. But the nominatives of two languages are, as a
rule, by no means fully equivalent: they can play different syntactic roles; however,
for them to be called nominatives it is enough that both are used for nomination.
In the same vein, the datives of two languages normally are different in their syn-
tactic behavior (consider the Georgian or Hindi dative, which is used to mark the
DirO, with the dative of classical or Slavic languages), yet they are correctly called
datives because both mark the Receiver actant of the verbs of giving. By the way,
the dative is not cross-linguistically universal; even more than that: no other case
but the nominative is, since some case languages have only two cases: the nomina-
tive and the oblique; such are, for instance, Old French and Kurdish.
It is easy to multiply the examples: the Russian present tense does not have
the same senses and the same syntactic uses as the present tense in German or
Japanese, but this does not prevent us from calling all the corresponding gram-
memes PRESENT, since all of them can designate the coincidence with the moment
of speech; the singular in Russian and English does not fully correspond to the
Hungarian or Turkish singular; etc.
It is in this sense that some SSyntRels are cross-linguistically universal.

NB Deep-syntactic relations (Chapter 2, 2.2.1, pp. 32–35) are not considered here. The set of
DSyntRels is postulated deductively for all languages; it is universal in the sense that it is cross-
linguistically valid—it is sufficient for the description of deep-syntactic structures in any lan-
guage (of course, under the condition that we allow for the use of fictitious lexemes: Chapter 2,
2.2.2, pp. 36ff ). A particular DSyntRel may not appear in all languages. The DSyntRels I, II, IIdir.sp,
III, ATTR, ADDRESS, APPEND and COORD seem to be universal; the DSyntRels IV, V and VI are absent
from some languages; I have no solid evidence about ATTRdescr and PSEUDO-COORD.
 3.9 Summing up 177

A set of SSyntRels is established empirically for each given language L (for a ten-
tative list of SSyntRels of English, see Mel’čuk & Pertsov 1987: 85–156 and Mel’čuk
2016: 184–194). A particular SSyntRel may or may not be cross-linguistically
universal. Thus, the subjectival SSyntRel, examined in this chapter, is universal,
while the Russian approximate-quantitative SSynt­Rel (present in the phrase knig–
approximate-quantitative→dvadcat′ lit. ‘books twenty’ = ‘may­be twenty books’) is
found in Russian, but not in most other languages. The direct-objectival SSyntRel
is widespread, but not universal: ergative languages do not have it. I don’t know
whether the indirect-objective SSyntRel is universal or not; I have doubts about the
modificative SSyntRel.
The cross-linguistic universality of the subjectival SSyntRel is due, of course,
to its universalist definition as the most privileged SSyntRel of L.

27

A corrected version of Mel’čuk 2014b.

4 “Multiple subjects” and “multiple direct objects”
in Korean
4.1 Introductory remarks
4.2 The problem stated:
Is the same-case noun string a sequence of multiple subjects/multiple objects?
4.3 The prolepsis
4.3.1 The notion of prolepsis
4.3.2 Prolepses in Korean
4.4 The nominative vs. the subjective case
4.4.1 The Korean nominative
4.4.2 The Korean subjective
4.5 What are a syntactic subject and a direct object—in general and in Korean?
4.5.1 The syntactic subject
4.5.2 The direct object
4.6 Multiple same-case nouns in a Korean clause
4.6.1 “Multiple subjects” in a Korean clause
4.6.1.1 Rhematic prolepsis + subject
4.6.1.2 Subject + agentive complement/subject of a non-finite verb form
4.6.1.3 Subject + pseudo-conjunct
4.6.2 “Multiple direct objects” in a Korean clause
4.6.2.1 Indirect object + direct object
4.6.2.2 Direct object + quasi-direct object
4.6.2.3 Direct object of the Main Verb + direct object of a non-finite form
4.6.2.4 Affected object + direct object
4.6.2.5 Direct object + pseudo-conjunct
4.6.2.6 “Quadruple direct objects”
4.6.3 Other “multiple objects” in a Korean clause
4.7 Conclusions

4.1 Introductory remarks

For many years I have been fascinated by statements found in numerous refer-
ence books and manuals that Korean has multiple subjects [Subjs] and multi-
ple direct objects [DirOs] in one clause; see, for instance, a detailed descriptive
grammar Sohn 1994 (e.g., pp. 235 and 237), the paper MacDonald & Welch 2009
or the PhD thesis Cho 2011. It is commonly said that a simple Korean clause—that
is, a clause with just one finite verb—can contain several non-coordinated sub-
jects and several non-coordinated DirOs. Here are standard examples, one-clause
sentences (1) and (2), in which sequences of “Subjs” and, respectively, of “DirOs”
are shaded (the names of grammatical cases are used in these examples in the
traditional way):

https://doi.org/10.1515/9783110694765-005
180 4 “Multiple subjects” and “multiple direct objects” in Korean

(1) Nay+ka paym+ii musep +Ø +ta

I NOm snake NOm be.fearful PReS DeCL(arative)
lit. ‘I snake is.fearful’. = ‘I am afraid of the snake’.

(2) Kay+ka John+ɨl son +ɨl mul+ess +ta

dog nom ACC hand ACC bite PaSt DeCL
lit. ‘Dog John hand bit’. = ‘The dog bit John on the hand’.

Sentence (1) presents a sequence of two nouns in the case currently called nomi-
native (marked by the suffix -ka after a vowel and -i after a consonant); both
nouns are traditionally considered to be SyntSubjs. Sentence (2) presents two
accusative nouns (suffixes -lɨl/-ɨl), both considered to be DirOs.
NB There is no full parallelism between multiple nominative and multiple accusative construc-
tions; they receive different treatments and different descriptions, as we will see in Subsections
4.6.1 and 4.6.2. In particular, sentence (1) is communicatively not neutral—unlike its English
gloss; it is translated more precisely as ‘It is me who is afraid of the snake’ (cf. example (6a) in
Subsection 4.3.2).

Longer sequences of nominative and accusative nouns are possible, but, for sim-
plicity’s sake, the discussion will be at first limited to sequences of two nomina-
tive or accusative nouns.

Transliteration and pronunciation

In this chapter, Yale Romanization of the Korean script is used—with some modi-
fications aimed at a better one-to-one correspondence of our transliteration
symbols with letters and digraphs of Korean alphabet (hankɨl). Here are some
elementary pronunciation rules for the adopted transliteration:
– An unaspirated lax voiceless consonant is automatically voiced between
vowels and semi-voiced in the word-initial position, so that Maryka is pro-
nounced as [mæriga], hata ‘do’ as [hada], etc.
– A doubled consonant letter indicates “tenseness”: kk = /k̄/, ss = /s̄ /, etc. A
tense, or strong, consonant is never voiced.
– /l/ has an allophone [r] in an intervocalic position.
– The letter e represents [ə] or [ʌ]; the digraphs ay, ey and oy stand for [ε], [e]
and [œ].

Sequences of non-coordinated nouns that are in the same grammatical case and
are considered to play the same syntactic role (–r→N1-CASE-x + –r→N2-CASE-x + …) are
not such a rarity cross-linguistically: a similar situation, although with respect
to multiple nominatives only, is observed in Japanese (for instance, Kuno 1973:
34, 62ff). Several caseless languages allow for sequences of non-coordinated
 4.1 Introductory remarks 181

nouns in the same syntactic role—for instance, Mandarin and quite a few lan-
guages across the linguistic board, such as Totonac (Beck 2016) and Kinyarwanda
(Kimenyi 1980, Dryer 1983). Multiple NCASE-xs are well known in ancient languages
(Ancient Greek, Sanskrit, Biblical Hebrew, Old Church Slavonic): ‘The God his
voice entered the room’, ‘People admired the King the face’, ‘He was killed by
elephants by their legs’, etc. Here is an actual example of triple accusatives from
the “Iliad”:

(3) Idomene +ùs ùs Oinóma +on bále gastér+a méssē+n

Idomeneus nom Oinomaos ACC struck belly ACC middle ACC
lit. ‘Idomeneus Oinomaos struck belly middle’. =
‘Idomeneus struck Oinomaos in the middle of the belly’.

However, in order to simplify my task, I will leave out any attempt at typological
generalizations, limiting myself to Korean.
There is no shortage of studies dedicated to multiple same-case noun se-
quences in Korean; here I cannot even try a review of the literature.
The goal of this chapter is to establish for Korean some clause elements and
their case encodings that would ensure a straightforward transition from a depen-
dency syntactic structure of a Korean sentence to the sentence itself. Therefore, the
legitimate question for the examples that are given below is not “Why this expres-
sion is described as such and such clause element?”; each example is intended to
illustrate the following implication: “If this expression is described as such and
such clause element, then the passage from the syntactic structure to this sen-
tence is simple and consistent with General Syntax.” The idea is to put Korean
“multiple subjects” and “multiple objects” in the perspective of General Syntax.

Three specific problems with the following discussion

The nature of the present chapter leads to (at least) three complications that a
reader has to deal with: the first is related to the character of the Korean language,
the second is brought in by the topic itself, and the third one is very general,
almost philosophical.

1) Due—at least in part—to the agglutinative character of Korean, the case

suffixes can be omitted in various contexts; it is often the case that a given
Korean sentence, especially stripped of its spoken prosody, allows for several
readings. On the other hand, a given meaning can be expressed by means of
different syntactic structures, which results in different, but more or less syn-
onymous sentences. However, these additional options are, as a rule, logi-
cally irrelevant for the exposition and can be ignored.
182 4 “Multiple subjects” and “multiple direct objects” in Korean

2) The study of multiple clause elements in Korean requires the consideration

of the quasi-total­ity of Korean grammar: the morphology and semantics
of grammatical cases, some inflectional categories of the verb (different
gerunds), the communicative structure (Rheme ~ Theme, Focus, Contrast,
etc.), word order, syntactic intonation and phrasing, and the inventory of
the clause ele­ments. It goes without saying that it is impossible to seriously
tackle all these outstanding tasks here. I simply have to be less than precise
and leave out several details that, with all their impor­tance, are again logi-
cally irrelevant for the goals of this chapter.
3) Last, but not least, the proposed description of an interesting syntactic phe-
nomenon of Korean is carried out within the framework of the Meaning-Text
perspective, based on general typological considerations, dependency syn-
tactic representation and a formal system of linguistic notions and terms.

4.2 T
 he problem stated: Is the same-case noun string a sequence
of multiple subjects/multiple objects?
The tendency to interpret a grammatical case as a marker of a specific syntactic
role is quite understandable. In conformity with this tendency, many Korean-
ists conclude that a sequence of the same-case nouns is a sequence of the same
clause elements. As a result, they speak of multiple subjects and multiple direct
objects in Korean. However, general linguistics tells us that a given Main Verb in a
clause cannot have more than one subject or more than one direct object (without
counting, of course, coordinated Subjs and DirOs). The Subj and the DirO are syn-
tactic actants of a lexical unit L; L’s syntactic actants correspond to L’s semantic
actants. But:

Semantic and main syntactic actants of a lexical unit L—that is, the subject,
the direct object and the indirect object—are not repeatable with L. In other
words, the Main Verb L of a clause can have just one Synt-actant of each of
these three types.

This is so for an obvious semantic reason: each semantic actant saturates a

specific semantic slot of L, implementing one of the arguments of the predicate
‘L’, and a given semantic slot cannot receive, by its very nature, more than one
element within a given utterance. Since the three main syntactic actants express
semantic actants, the same is true of L’s main syntactic slots.
NB The element filling a semantic slot in ‘L’ can be either one semantic entity or one list of
semantic entities. In the latter case, the corresponding syntactic element (= a phrase) that fills
a corresponding syntactic slot of L is a chain of conjoined sentence elements, so that a sentence
 4.3 The prolepsis 183

can actually have several conjoined Subjs or conjoined DirOs ( John, Peter and Mary arrived or I
saw John, Peter and Mary ). But a given syntactic actant slot can never have a multiple expression
by non-coordinated actants.

Therefore, based on general linguistic knowledge, it is possible to state the fol-

lowing:

There cannot be multiple Subjs or multiple DirOs in Korean.

Having said this, I have to solve the contradiction between the most Korean
scholars’ state­ments and the corresponding general linguistic statements. In
order to do this, I will examine strings of Korean same-case nouns and explain
what they are in reality. And for this, I need to answer two questions:

– What is the Subj and the DirO—in general and in Korean?

– What are the elements of a “suspect” Korean same-case noun sequence that
are neither sub­jects nor direct objects?

But before these questions can be attacked, two auxiliary notions absolutely
needed for the discussion have to be introduced: prolepsis (3) and nominative case
(4).

4.3 The prolepsis

4.3.1 The notion of prolepsis

A clause element illustrated by the French sentence in (4), which manifests three
such (boldfaced) elements, is well-known in linguistics, but strangely has no
accepted name:

(4) French
Jacqueline, son père, le frigo, elle le lui a refilé
lit. ‘Jacqueline, her father, the fridge, she it to.him has passed’.

This clause element can be called prolepsis. More than 65 years ago, A. Xolodovič
(1954: 253–254) described this clause element in Korean, calling it “a complement
of a special kind.”
184 4 “Multiple subjects” and “multiple direct objects” in Korean

Definition 4.1 – prolepsis (Mel’čuk 2001: 130ff)

A lexical unit L (with its syntactic dependents) appearing in a clause is a pro-

lepsis if and only if L satisfies simultaneously the following four conditions:
1. Syntactically, L is only loosely linked to the rest of the clause: it is neither
a syntactic actant of the Main Verb nor one of its circumstantials.
2. Linearly, L is positioned clause-initially.
3. Prosodically, L is “insulated” from the clause by a pause, a stress and a
special intonation contour.
4. Morphologically (in a language with cases), L is most often—but not
exclusively!—in the nominative, the least marked case. (In Korean, a pro-
lepsis can be marked also by the subjective case, see Subsection 4.4.2.)

A prolepsis L normally serves to express a communicative value assigned to the

meaning ‘L’—more precisely, the Rheme or the Theme of the clause.
The Rheme (also known as comment) is the part of the clause that states what
the Speaker wants the clause to communicate; the Theme (topic) is the part of
the clause about which the Rheme is stated (e.g., Mel’čuk 2001 and 2012–2015:
vol. 1, 306ff). In a number of languages, a meaning selected by the Speaker to be
presented as the Rheme or the Theme can or must be implemented as a prolepsis.
NB The communicative organization of a clause is quite complicated and cannot be properly
dealt with here. However, it is useful to indicate its two following properties.
– The Comm-organization is essentially semantic, that is, the distribution of Rhemes and
Themes con­cerns primarily the meaning of the clause; the marking of Comm-organization on
the syntactic level (and in the clause itself) reflects its surface implementation, which does
not stand in one-to-one correspondence with its semantic source.
– The Comm-organization is recursive, that is, a thematic or rhematic area can have within it
another Rheme ~ Theme division of a lower level.
Thus, the sentence in (6a) has the following Comm-organization on the semantic level:
‘[[what is long]Theme2 [is the trunk]Rheme2]Theme1 [at the elephant]Rheme1’

Prolepses, which are widespread in colloquial French, are also possible in

English, albeit used rather sparingly: John, he is a nice guy. But South-East Asian
languages abound in prolepses, and Korean is no exception.

4.3.2 Prolepses in Korean

Korean prolepses are characterized by the following four features:

1. A rhematic prolepsis is marked by the subjective case (-ka/-i), see Subsection

4.4.2 below. A thematic prolepsis carries a special Theme marker -nɨn/-ɨn,
 4.3 The prolepsis 185

which is most frequently added to a bare noun stem—that is, to the nomina-
tive form; however, the -nɨn/-ɨn marker can also attach to a non-nominative
case form, an adverb and a converb:

(5) a. Seoul+eyDAT+nɨn salam+Ø manhta

lit. ‘In.SeoulTHEME person be.many’. = ‘In Seoul there are many people’.
b. Usen+ɨn nay+ka sakwa+lɨl mekessta
lit. ‘FirstTHEME , I apple ate’.
c. Nay+ka sakwa+lɨl mek+ko+nɨn siptta
lit. ‘[I apple eating]THEME , [I] want’. = ‘To eat an apple, I want it’.

NB Note an interesting asymmetry in that rhematicity is expressed by a grammatical case,

while thematicity has a special marker -nɨn/-ɨn, which can combine with the marker of a
case form. For more on the meaning and use of -nɨn/-nɨn, see Lee & Ramsey 2000: 163–166.

2. Prolepses can be multiple, so that a clause can have several thematic and/or
rhematic prolepses (Sohn 1994: 203; Chang 1996: 200); for simplicity’s sake,
I limit the examples to two prolepses (shaded):

(6) a. [K
Khokkili+ka]Rheme [kkho +ka]Rheme kil +Ø +ta
elephant SUBJ trunk SUBJ be.long PReS DeCL(arative)
‘It is the elephant [such that] it is [his] trunk [that] is long’.
b. [K
Khokkili+Ø +nɨn]Theme [kkho +Ø +nɨn]Theme kil +Ø +ta
elephant nom th trunk nom th be.long PReS DeCL
‘As for the elephant, as for [its] trunk, [it] is long’.
c. [K
Kho +Ø +nɨn]Theme [kkhokkili +ka]Rheme kil +Ø +ta
trunk nom th elephant SUBJ be.long PReS DeCL
‘As for trunk, it is the elephant [whose trunk] is long’.
d. [K
Khokkili+Ø +nɨn]Theme [kkho +ka]Rheme kil +Ø +ta
elephant nom th trunk SUBJ be.long PReS DeCL
‘As for the elephant, it is trunk [that] is long’.
NB Korean is a strong Pro-Drop language; no pronouns coreferential with prolepses can
appear in the clause in the roles of SyntSubj, DirO, Possessor, etc.

3. Thematic and rhematic prolepses appear mainly in Theme – Rheme linear

order, as seen in (6c–d).

4. A Korean prolepsis can follow a regular clause element, which is fronted

for communicative purposes (boxed; SUBJ is the subjective case, see next
section):
186 4 “Multiple subjects” and “multiple direct objects” in Korean

(7) Kay+hantey [John+ɨn]

[
[John+ɨn]Theme son +i mul+li +ess +ko
dog Dat th hand SUBJ bite PaSS PaSt ConV(erb)
[Mary+nɨn]Theme tali+ka mul+li +ess +ta
th leg SUBJ bite PaSS PaSt DeCL
lit. ‘By.dog John hand being.bitten Mary leg was.bitten’. =
‘It is by the dog that John had his hand bitten and Mary, her leg’.

In (7), the thematic prolepses Johnɨn and Marynɨn follow the Agent Complement
kayhantey ‘by.dog’.
NB Korean has over 50 converbs—non-finite verbal forms used as modifiers of the Main Verb.
These converbs express various meanings: manner, purpose, intention, reason, result, concomi-
tance, etc. However, since this is irrelevant for the present discussion, the type of the converb
will not be indicated.

4.4 The nominative vs. the subjective case

Now an important correction has to be introduced: it concerns the name of what

is traditionally called the nominative case in Korean grammar.

4.4.1 The Korean nominative

The case in -ka/-i is systematically called the nominative in Korean grammar,

since it is used to mark the syntactic subject (as in Latin!). However, this use of
the term is incorrect.

Definition 4.2 – nominative case (Mel’čuk 1988: 208, 255–256, 2006a: 110ff,
especially 152–153)

The nominative is the case of the form of the noun used for nomination.

The genuine nominative case exists, of course, in Korean and has the zero marker
-Ø, which is quite typical of the nominative in languages of the world: na+Ø ‘I’,
kay+Ø ‘dog’, namu+Ø ‘tree’, salam+Ø ‘person’; Korean grammarians refer to it as
the “basic form” of a noun or—as in Xolodovič 1954: 54—the “basic case.” The
nominative is used in Korean dictionaries as the lexicographic form, as it should
be; it appears in texts in various syntactic roles:
 4.4 The nominative vs. the subjective case 187

(8) Possible syntactic roles of a nominative noun form

Subject John+Ø kanta ‘John goes (somewhere)’.

DirO Na+Ø John+Ø ponta ‘I John see’.
IndirO Ne+Ø John+Ø kɨke ču+ess+ni? ‘You John this gave?’
Copular attribute Na+Ø sensayŋ+Ø ita ‘I teacher am’.
Adnominal attribute salam+Ø moksoli ‘person voice’ = ‘human voice’
Direction circumstantial Seoul+Ø kanta ‘[I/You/He/… to] Seoul go/goes’.
Address Sensayŋnim+Ø, ili osipsio! ‘Respected.teacher,
come here!’
Thematic prolepsis Na+Ø+nɨn kanta ‘I [= As for me] [I] go’.
NB 1. As one can see, the Theme marker -nɨn/-ɨn is added to the form of the nominative, which is
the expected grammatical case of a prolepsis. (This was clearly stated in Xolodovič 1954: 57.)
2. The zero marked nominative can replace the subjective in -ka/-i (see immediately below),
the accusative in -lɨl/-ɨl and the genitive in -ɨy without affecting the meaning (especially,
in colloquial speech).

Therefore, the case in -ka/-i is not a nominative. Since it is used to mark all types
of SyntSubj, it can be called the subjective.

4.4.2 The Korean subjective

Definition 4.3 – subjective case (Mel’čuk 1988: 263, 2006a: 153)

The subjective is the case used first and foremost for marking the syntactic
subject of any type, but which cannot serve for nomination.

The Korean subjective marks, of course, the SyntSubj, this being its main, but not
only, function. It also marks at least the following three secondary syntactic roles:

1. The attribute of the copula or of a copula-like verb, as in (9):

(9) a. John+i sensayŋ+i ani +ta ‘John is not a teacher’.

SUBJ teacher SUBJ be.not DeCL

b. John+i sensayŋ+i toy +ess +ta ‘John became a teacher’.

SUBJ teacher SUBJ become PaSt DeCL

2. The agentive complement [AgCo] of a nominalized, adjectivalized or adverbi-

alized verb (≈ converb) in (at least) two constructions.
188 4 “Multiple subjects” and “multiple direct objects” in Korean

– First, the subjective case marks the AgCo of the manner converb in an analyti-
cal causative construction «VCONV(erb) + hata ‘make’»; this AgCo semantically
is the Causee Actor, as in (10):

(10) a. John+i Mary+ka ←agent-compl–[čhayk+ɨl]—ilk +ke hay +ss +ta

SUBJ SUBJ book aCC read ConV make PaSt DeCL
lit. ‘John made Mary reading book’, where Mary is an AgCo of the converb
ilkke ‘reading’.

This causative construction has two further “case frames”—with different

cases of the Causee Actor noun:

b. John+i Mary+lɨl čhayk+ɨl ilk +ke hay +ss +ta

SUBJ ACC book aCC read ConV make PaSt DeCL
c. John+i Mary+eykey čhayk+ɨl ilk +ke hay +ss +ta
SUBJ DAT book aCC read ConV make PaSt DeCL

The subjective on the Causee Actor = agentive complement (here, MaRy) alter-
nates with the accusative and the dative (the dative indicates voluntary agentiv-
ity of the Causee Actor). The use of an NSUBJ as an AgCo with a manner converb in
Korean is similar to the use of an NNOM as an AgCo of an infinitive in Portuguese
(as in Ter euNOM saúde é bom lit. ‘To.have I health is good’. = ‘It is good that I have
health’.) or of a gerund in Spanish (as in Nos casamos hace 50 años estando yo
sin trabajo lit. ‘We married 50 years ago, being I without work’.)

– Second, the subjective case marks the AgCo of a nominalized or adjectival-

ized verb in what corresponds to a completive or a relative clause in an SAE
language, as in:

(11) a. John+i čhayk+ɨl ssu +m

SUBJ book aCC write NOmIN(alizer)
‘John book write.fact’ = ‘that John wrote a book’
b. John+i ssu +n čhayk
SUBJ write ADJ(ectivalizer) book
‘John written book’ = ‘book that John wrote’

3. The oblique object of a parametric verb (‘weigh’, ‘be.long’, ‘cost’), as in (12):

(12) I čhayk+i paek kram+i naka +n +ta

this book SUBJ hundred gram SUBJ weigh PReS DeCL
‘This book weighs 100 grams’.
 4.4 The nominative vs. the subjective case 189

The most important property of the Korean subjective, which it shares with
the Japanese sub­jective case in -ga, is its use to mark the Rheme (or the Rhematic
Focus) of the clause (Chang 1996: 200); two cases have to be distinguished.
In the simplest case, we have a rhematic subject in the subjective case (boxed):

(13) [Khokkili +Ø +nɨn]Theme kho +ka Rheme kil +Ø +ta

elephant NOM TH trunk SUBJ be.long PRES DECL
‘As for the elephant, it is his trunk that is long’. ≈
‘With elephants, what is long is their trunk’.

Sentence (13) is good as an answer to the question As for elephants, what is long
with them? The NSUBJ is syntactically the SyntSubj, and communicatively the
Rheme.
A more complex situation obtains when the subjective marks a rhematic pro-
lepsis (boxed):

(14) Nay+ka John+i čoh +ta

I SUBJ SUBJ be.likable DECL
‘I like John’. = lit. ‘[It is] I [to whom] John is.likable’.

— Korean allows for an even more complex picture: the subjective case can
mark as rhematic a clause element that is different from the SyntSubj, is
not a prolepsis and is already marked by another case; the result is what is
known as “case stacking” (boxed; Schütze 2001: 194):

(15) Na+eykey+ka paym+i musep +Ø +ta

I DAT SUBJ snake SUBJ be.fearful PRES DECL
‘I am afraid of snakes’. = lit. ‘[It is] to.me [that] snake is.fearful’, where
naeykey ‘to.me’ is an IndirO.

The phenomenon of case stacking led some researchers to say that -ka and -i suf-
fixes are homophonous: each marks either the subjective case or the Rheme (see,
for instance, Schütze 2001). However, the subjective has still another suffix—
namely, -kkeyse, which is honorific; it also can be stacked in a corresponding
situation. Thus, not only -ka and -i, but -kkeyse as well should be considered
homophonous, which is jarring. In addition, the accusative suffix -lɨl/-ɨl is also
used to express the Rheme, thus again producing case stacking (Sohn 1994: 184):

(16) John+ɨn Mary+eykey+lɨl ka+ss+ta ‘As for John, it is to Mary that he went’.

Should we see the homophony “accusative vs. rhematization” in this suffix, too?
190 4 “Multiple subjects” and “multiple direct objects” in Korean

On the one hand, the use of grammatical cases for the expression of commu-
nicative values and referentiality is well-known cross-linguistically (Tibetan,
Yukaghir, Daghestanian langua­ges); on the other, the Korean subjective and
accusative carry the nuance of focusing (emphasis, contrast) even when used in
their genuine syntactic function. Therefore, I prefer to consider the correspond-
ing markers to be case suffixes, allowing for rhematizing behavior.
Fortunately (for me), the solution of this additional problem is irrelevant to
my topic here.
Summing up, the “suspect” same-case noun strings are not N1-NOM N2-NOM
… Nn‑NOM, but N1-SUBJ N2-SUBJ … Nn‑SUBJ. This correction does not, however, affect the
essence of the problem considered here—namely, the question whether such a
string is a string of surface-syntactic subjects. It was implied above and will be
shown below that it is not.
NB Some Korean grammarians speak of the use of the subjective to mark a direct object (e.g.,
Sohn 1994: 237; the boxing is mine—IM):
(i) Nay+ka kohyaŋ+i kɨlip+ess+ta
lit. ‘As.for.me, hometown lacked’. = ‘I missed [my] hometown’.
(ii) Nay+ka sensayŋnim+i musep+ess+ta
lit. ‘As.for.me, respected.teacher was.frightening’. = ‘I was.afraid of [the] respected.teacher’.

The boxed NSUBJs are described in Sohn 1994 as DirOs. This is, however, a simple
misunderstanding pro­voked by the English translation. Korean kɨlip means ‘X
lacks to Y’ rather than ‘Y misses X’; in this respect, Korean is like French: Ma ville
natale me manque lit. ‘My hometown to.me lacks’. Analogously, musepta means
‘X is fearful for Y’ rather than ‘Y is afraid of X’. The boxed nouns are quite regular
Subjects.

4.5 W
 hat are a syntactic subject and a direct object—in general
and in Korean?
4.5.1 The syntactic subject

Definition 4.4 – syntactic subject (Chapter 3, Definition 3.1, p. 135)

The syntactic subject [SyntSubj] is the most privileged surface-syntactic

actant in of the Main Verb language L.

Although the definition of SyntSubj is language-universal, its privileges must be

specified for each language. Based on the list of universal definitional parameters
of a SSynt-clause element (Chapter 3, Table 3.1, p. 124), it is possible to formulate
the five privileges of the SSynt-subject in Korean:
 4.4 The nominative vs. the subjective case 191

1) The SyntSubj can depend only on the Main Verb [MV], which is a genuine finite
verb or a predicative adjective (definitional parameter 1 in Chapter 3, Table 3.1).
A clause that underwent adjectivalization or nominalization of its Main
Verb ceases to be a clause and cannot have a SyntSubj; the main SSynt-actant
of an adjectivalized or nominalized verb is its agentive complement.
2) The SyntSubj cannot be omitted from the SSyntS of the sentence (defini-
tional parameter 2 in Chapter 3, Table 3.1). In a one-word sentence of the type
John+i+ta ‘There is John’ the noun John appears in the predicative form (lit.
‘John is’), and it is difficult to call it Synt-subject; but it is an obvious SyntSubj
in the SSyntS of this sentence (cf. Chapter 3, 3.2.4.2, NB in Comment 2, p. 126).
3) The SyntSubj linearly precedes all other MV’s actants—with the exception of
rhematic elements, which can be fronted; this corresponds to the definitional
parameter 3 in Chapter 3, Table 3.1. (The thematically-loaded clause elements
that can precede the SyntSubj are prolepses and, therefore, not actants.)
4) The SyntSubj is affected by the valence-changing inflection of the MV.
Namely, it express­es the “endpoint” of passivization of the MV: the DirO of
the active form of the MV becomes the SyntSubj of its passive (definitional
parameter 4b in Chapter 3, Table 3.1).1

1 Clause elements mistakenly taken for SyntSubjects in Korean

1. Sometimes the dative IndirO of a verb or of an adjective of affection/possession is called Synt-
Subj (Kim & Sells 2010: 609):
(i) Sensayŋ+nim+kkey čhayk+i manh +ta
teacher HON DAT.HON book SUBJ be.many DECL
lit. ‘To.respected.teacher books are many’. = ‘The teacher has many books’.
However, the boxed clause element is the subject only in English translation. It is a typical
IndirO fronted for communicative and pragmatic reasons. Note that it does not obligatorily
impose honorification on the MV (although the honorific form manh+si+ta makes the sentence
more acceptable).
2. A noun in the locative is not the subject in (ii), either (although some consider it to be a subject):
(ii) Hoysa +eyse Ø na+Ø hanthey phosaŋkɨm+ɨl ču +ess +ta
company LOC «they» I NOM to award ACC give PAST DECL
lit. ‘In company, «they» gave to me an award’.
The SyntSubj here is a zero lexeme, meaning ‘indefinite people’—in conformity with Han’s (2004,
2006) proposal: Ø(pl)
«people»
. It corresponds nicely to the Rus. zero pronoun Ø(3,
«people»
pl) (V kompanii mne
dali premiju ‘In company, «they» gave me award’.) and to the Fr. and Ger. non-zero pronouns on
and man. This zero lexeme has a clearly human reference (as it is to be expected):
(iii) a. Toŋmulwen+eyse Ø(pl) «people»
halu+ey tu +kki meki +lɨl ču +n +ta
zoo LOC «they» day DAT two times fodder ACC give PRES DECL
‘In the zoo, «they» give fodder [to animals] twice a day’.
vs. b. *Toŋmulwen+eyse Ø(pl) «people»
halu+ey tu +kki meki +lɨl mek+nɨn +ta
zoo LOC «they» day DAT two times fodder ACC eat PRES DECL
‘In the zoo, *«they» [= animals] eat fodder twice a day’.
192 4 “Multiple subjects” and “multiple direct objects” in Korean

(17) Koyaŋi+ka čwi +lɨl mek+Ø +ess +ta

cat SUBJ mouse ACC eat ACT PAST DECL
‘The cat ate the mouse’.
vs. Čwi +ka koyaŋi+hantey mek +hi +ess +ta
mouse SUBJ cat DAT eat PASS PAST DECL
‘The mouse was.eaten by the cat’.

5) The SyntSubj is the only MV’s actant that accepts the honorific suffix -kkeyse
≈ ‘highly respected’ of the subject-marking subjective case (definitional
parameter 5 in Chapter 3, Table 3.1):2

(18) Eme +nim+kkeyse ka+si +ess +ta

mother HON SUBJ go HON PAST DECL
‘Mother went (somewhere)’.

2 Honorification in Korean
Honorification imposed on the MV by an actant is sometimes considered to be another privilege
of the Korean SyntSubj. This is, however, incorrect:
– On the one hand, a SyntSubj does not impose honorification on some verbs (O’Grady 1991:
102):
(i) John+eykey sensayŋ+nim+i *philyoha +si +ta  [the correct form is philyoha+ta]
DAT teacher HON SUBJ be.needed HON DECL
lit. ‘To.John the.respected.teacher is.needed’. = ‘John needs a respected teacher’.
– On the other hand, other actants of the MV (or even their Possessors) and prolepses can impo-
se honorification on it (Gerdts & Youn 1989: 3 and Jang 1997: 36):
(ii) Sensayŋ+nim+ɨy elkul+ey paykmuk+i mut +ɨsi +ess +ta
teacher HON GEN face DAT chalk SUBJ smudge HON PASS DECL
‘The chalk respectfully.smudged the respected.teacher’s face’.
(iii) Sensayŋ+nim+kkeyse son +i čaku+si +ta
teacher HON SUBJ.HON hand SUBJ small HON DECL
 lit. ‘It is the respected.teacher whose hands are respectfully.small’. =
‘The respected.teacher has small hands’.
(iv) John+i sensayŋ+nim+ɨl aphu+si +ta +ko mit +ess +ta
SUBJ teacher HON ACC sick HON DECL CONV believe PAST DECL
‘John believes the respected.teacher to be respectfully.sick’.
In (ii), honorification is imposed on the Main Verb by the Possessor of an OblO, in (iii) by a rhe-
matic prolepsis, and in (iv), the DirO of the MV imposes honorification on a converb!
Honorification (as well as reflexivization) is controlled in Korean by the semantic role of the
corresponding sentence elements. Cf.: “Phenomena such as reflexive interpretation and hono-
rific agreement are sensitive to the most ‘prominent’ of a verb’s semantic arguments” (O’Grady
1991: 105; emphasis added—IM).
 4.6 Multiple same-case nouns in a Korean clause 193

4.5.2 The direct object

Definition 4.5 – d
 irect object (Chapter 3); this is a simplified version of Definition
3.2 (p. 174)

The direct object is the second most privileged surface-syntactic actant in a

non-ergative language L.

The DirO exists only in non-ergative languages (Chapter 3, Subsection 3.2.6, p. 134);
its definition is language-universal, but its privileges must be specified for each
language individually. In Korean, the privileges of a DirO are:

1) The DirO tends to linearly follow all other MV’s actants—that is, to be placed
immediately before the MV (barring a Quasi-DirO).
NB A quasi-direct object (and a pseudo-subject) are clause elements different from the DirO
and the SyntSubj; on the quasi-direct-objectival-2 surface-syntactic relation in Persian, see
Chapter 2, Section 2.4, No. 9, p. 55, and Mel’čuk 2012–2015: vol. 3, 431; on the pseudo-
subjectival surface-syntactic relation in English, see Chapter 2, Section 2.4, No. 6, p. 51,
and Mel’čuk 2012–2015: vol. 3, 445.

2) The DirO is the only MV’s actant that accepts the accusative case which does
not alternate with any other case except for the nominative.
3) The DirO is the only MV’s actant that can be promoted to SyntSubj status by
MV’s passivi­zation.
NB Тhe IndirO is defined in the same way: it is the third most privileged actant, whose
privileges have to be specified for each individual language; etc.

4.6 Multiple same-case nouns in a Korean clause

After lengthy and fairly cumbersome preparations, the ground is ready for
answering the main question of this chapter (Section 4.2, p. 182):
What actually are, from a syntactic viewpoint, sequences of same-case
nouns in Korean?

Several linguists in the past took steps towards a correct analysis of N1-CASE N2-CASE …
Nn‑CASE sequences in Korean. Thus, O’Grady 1991: 235–242 proposes a fine analysis
of the NACC that is in a collocational link with the verb hata ‘make’, insisting on
its special syntactic role (which I propose to call quasi-direct object). In a similar
way, Sohn 1994: 204 explicitly says that in sentence (19) “the predicate is directly
related to the last NP which is its subject [boxing is added—IM]. The other preced-
ing … NPs are best considered topics.”
194 4 “Multiple subjects” and “multiple direct objects” in Korean

(19) Nay+ka čha+ka thaie+ka kumeŋ+i na +ss +ta

I SUBJ car SUBJ tire SUBJ hole SUBJ occur PAST DECL
lit. ‘I car tire hole occurred’. = ‘I have a hole in a tire of my car’.

However, I don’t know of a systematic overview and formal description of same-

case noun sequences in Korean. I undertake such overview here, beginning with
so-called “multiple sub­jects” (4.6.1), then considering “multiple direct objects”
(4.6.2), and finishing with a few remarks about “other multiple objects” (4.6.3).

4.6.1 “Multiple subjects” in a Korean clause

A Korean clause can contain several consecutive nouns in the subjective case, but
only one of them is the surface-syntactic subject of the clause’s Main Verb. Let us
consider two consecutive NSUBJs; three situations are to be examined.

1) N1-SUBJ and N2-SUBJ are syntactically not linked

a) N1-SUBJ and N2-SUBJ depend in parallel on the Main Verb: N1-SUBJ is a rhematic
prolepsis and N2‑SUBJ is the subject (4.6.1.1).
b) N1-SUBJ is the SyntSubj of the Main Verb, and N2‑SUBJ is the AgCo (or the
SyntSubj?) of a non-finite verbal form, i.e., of a converb (4.6.1.2).

2) N1-SUBJ and N2-SUBJ are syntactically linked

c) N2-SUBJ is a pseudo-conjunct of N1-SUBJ, this latter being the SyntSubj of the

Main Verb (4.6.1.3).

4.6.1.1 Rhematic prolepsis + subject

The N2-SUBJ is the SyntSubj, the N1-SUBJ being syntactically a prolepsis that expresses
the Rheme of the clause. This description was explicitly proposed in O’Grady
1991: 121ff. Lee & Ramsey 2000: 144 say that N1-SUBJ is, so to speak, a “subject” of
the whole following clause rather than that of its MV; see also Kim & Sells 2010:
607, where several important references are given that buttress the treatment of
N1-SUBJ as a prolepsis, although they do not use this term.
The NSUBJs that compose the string under consideration have the following
important semantic property: the N2-SUBJ and the N1-SUBJ are linked by a metonymic
semantic relation. For instance, ‘N2‑SUBJ is a part of N1-SUBJ’, or ‘N2-SUBJ is located in/on
N1-SUBJ’, or else ‘N2-SUBJ happens during N1‑SUBJ’:
 4.6 Multiple same-case nouns in a Korean clause 195

(20) a. John+i kho +ka kil +Ø +ta

SUBJ nose SUBJ be.long PRES DECL
‘[It is] John [whose] nose is long’.
b. Thakča+ka čhayk+i manh +ta
table SUBJ book SUBJ be.many DECL
‘[It is the] table [where] books are many’.
c. Pom +i kkočh +i manh +ta
spring SUBJ flower SUBJ be.many DECL
‘[It is the] spring [when] flowers are many’.

The N2-SUBJ can also represent a rhematic prolepsis; cf. (21), where the nominal
clause elements have the same forms as in (13):

(21) [Khokkili+Ø +nɨn]Theme [kho +ka]Rheme kil +Ø +ta

elephant NOM TH trunk SUBJ be.long PRES DECL
‘As for the elephant, it is his trunk, [it] is long’.

In (21), the SyntSubj is a pronominal lexeme ‘it’, which does not appear in the
sentence.
In (22), the first two NSUBJs are rhematic prolepses, the last one being the
SyntSubj:

(22) John+i enehak +i koŋpu +ka toy +ess +ta

SUBJ linguistics SUBJ study(N) SUBJ be.made PAST DECL
lit. ‘[It is] John [and] linguistics [that] study was.made’. =
‘A study was done of linguistics by John’.
NB Sentence (22) can be also obtained from a different SSyntS. Namely, John is a rhematic
prolepsis; enehak ‘linguistics’ is the SyntSubj; and koŋpu ‘study(N)’ is the PseudoSubj of
the light verb toyta, parallel to the Quasi-DirO of the light verb hata (4.6.2.2): ‘[It is by]
John [that] linguistics was study done’.

4.6.1.2 Subject + agentive complement/subject of a non-finite verb form

This situation is found in a phrasal causative construction with the verb hata
‘make’ and the converb (= gerund) in -ke of the lexical verb, see (10a), repeated
here as (23a), as well as in constructions with other non-finite forms, see (23b):

(23) a. John+i Mary+ka čhayk+ɨl ilk +ke hay +ss +ta

SUBJ SUBJ book ACC read CONV make PAST DECL
‘John made Mary read a book’.
196 4 “Multiple subjects” and “multiple direct objects” in Korean

b. John+i Mary+ka čhayk+ɨl ilk +ɨn +ta +ko mit +nɨn +ta
SUBJ SUBJ book ACC read PRES DECL CONV believe PRES DECL
‘John believes Mary is.reading a book’.

In both sentences, Johni is the SyntSubj of the Main Verb, and Maryka is the AgCo
(or the SyntSubj) of the gerund (ilkke and ilkɨntako).
NB The agentive complement can probably be considered to be the syntactic subject of a non-
finite verb form; based on available data, I cannot solve this dilemma. However, it is irrelevant to
my point, since whatever the answer, there will be no multiple subjects of the Main Verb.

4.6.1.3 Subject + pseudo-conjunct

The N1-SUBJ is the SyntSubj, the N2-SUBJ being a pseudo-conjunct dependent element.

Definition 4.6 – pseudo-conjunct (Chapter 2, Section 2.4, No. 122, p. 111)

L2 is a pseudo-coordinate dependent, or a pseudo-conjunct, of L1, iff L2 follows

L1 imme­­diately and can play the same surface-syntactic role as L1, but does
not allow for a coord­inating conjunction linking L2 to L1.

The semantic load of a pseudo-conjunct L2 of L1 is to express an elaboration of L1:

‘L1, more precisely L2’; for instance:

(24) Leo lives in Spain,–pseudo-coord→in Barcelona,–pseudo-coord→on 4th May

Street,–pseudo-coord→in a big building,–pseudo-coord→on the fifth floor.

Thus, take sentence (25), which is the passive version of sentence (2):

(25) Kay+hantey John+i son +i mul+li +ess +ta

dog DAT SUBJ hand SUBJ bite PASS PAST DECL
lit. ‘By.dog John hand was.bitten’.

John is the SyntSubj, and the son ‘hand’ is its pseudo-conjunct: ‘John, more pre-
cisely [his] hand, was bitten by the dog’.

Sentence (25) formally corresponds also to two further syntactic structures:

– John is a rhematic prolepsis, while son ‘hand’ is the SSynt-subject: ‘It was
John whose hand was bitten by the dog’ (see 4.6.1.1).
– Both John and son are rhematic prolepses: ‘It was John and his hand, it was
bitten by the dog’. This is possible because, as indicated above, Korean is a
Pro-Drop language, and the resumptive pronouns such as ‘it’ or ‘his’ do not
appear in the sentence.
 4.6 Multiple same-case nouns in a Korean clause 197

In sentence (26), both nouns in the subjective—Mary and John—can be consid-

ered rhematic prolepses: ‘It was Mary and John, she put him to sleep’ (4.6.1.1).

(26) Mary+ka John+i ča +ke hay +ss +ta

SUBJ SUBJ sleep CONV make PAST DECL

Additional interpretations of (26) are also possible: John can be the AgCo/the
SyntSubj of the converb čake, while Mary can be the SyntSubj of hata ‘make’
(4.6.1.2) or a rhematic prolepsis.
A string of consecutive NSUBJs containing more than two components is easily
described in proposed terms: one of these NSUBJs can be the SyntSubj of the Main
Verb, one can be the CoAg of a non-finite verb form (= of a converb), and all the
others are rhematic prolepses.

4.6.2 “Multiple direct objects” in a Korean clause

The situation with strings of accusative nouns is slightly more complex. If we

consider a sequence of two NACCs, the following five (rather than three, as for the
NSUBJ) cases have to be dis­tinguished.

N1-ACC and N2-ACC are syntactically not linked

a) N1-ACC and N2-ACC depend in parallel on a ditransitive Main Verb: N1-ACC is an
IndirO, and N2-ACC the DirO (4.6.2.1).
b) N1-ACC and N2-ACC depend in parallel on a light Main Verb: N1-ACC is its DirO, and
N2-ACC is its Quasi-DirO (4.6.2.2).
c) N1-ACC is the DirO of the Main Verb, and N2-ACC is the DirO of the gerund in a
phrasal (= analytical) causative (4.6.2.3).
d) N2-ACC is the DirO of the Main Verb, and N1-ACC is the affected object of the same
verb (4.6.2.4).
N1-ACC and N2-ACC are syntactically linked
e) N2-ACC is a pseudo-conjunct to N1-ACC (4.6.2.5).

4.6.2.1 Indirect object + direct object

(27) a. Mary+ka John+ɨl čhayk+ɨl ču +Ø +ess +ta

SUBJ ACC book ACC give ACT PAST DECL
‘Mary gave John a book’.
198 4 “Multiple subjects” and “multiple direct objects” in Korean

On N1-ACC, but not on N2-ACC, the accusative freely alternates with the dative (without
changing its syntactic role):

b. Mary+ka John+eykey čhayk+ɨl ču +Ø +ess +ta

SUBJ DAT book ACC give ACT PAST DECL
‘Mary gave John a book’.

N1-ACC does not passivize, while N2-ACC does:

c. *John+i Mary+ey ɨyhay čhayk+ɨl ču +eči +ess +ta

SUBJ DAT by book ACC give PASS PAST DECL
‘John was given a book by Mary’.
vs. Čhayk+i Mary+ey ɨyhay John+eykey ču +eči +ess +ta
book SUBJ DAT by DAT give PASS PAST DECL
‘The book was given by Mary to John’.

In this respect, Korean is different from English and Japanese, which both have
indirect passives.
Another example of the same construction, where N1-ACC implements an
IndirO:

(28) Mary+ka lobotɨ+lɨl phal+ɨl ta +Ø +ass +ta

SUBJ robot ACC arm ACC attach ACT PAST DECL
‘Mary attached the arm to the robot’.

In Latin, Serbian and German, N1-ACC turns out to be an IndirO or OblO, while
N2-ACC is a genuine DirO. Consider also Ger. WasACC fragt er michACC ? lit. ‘What asks
he me?’, where was ‘what’ is an OblO: this was alternates with worüber ‘about
what’; only was and das ‘this’ are possible in the accusative in this position,
while any semantically convenient noun can replace mich.
NB On multiple accusatives in various languages, see Mel’čuk 2009a: 96, endnote [3].

4.6.2.2 Direct object + quasi-direct object (O’Grady 1991: 236, 1992)

(29) a. John+i enehak +ɨl koŋpu +lɨl hay +ss +ta

SUBJ linguistics ACC study(N) ACC make PAST DECL
lit. ‘John makes [a] study linguistics’. = ‘John studies linguistics’.

b. John+i enehak +ɨy koŋpu +lɨl hay +ss +ta

SUBJ linguistics GEN study(N) ACC make PAST DECL
‘John does a study of linguistics’.
 4.6 Multiple same-case nouns in a Korean clause 199

c. John+i enehak +Ø koŋpu +lɨl hay +ss +ta

SUBJ linguistics NOm study(N) ACC make PaSt DeCL
‘John does a linguistics study’

Sentence (29a) presents a well-known phenomenon—a so-called transitive “peri-

phrastic (= compound, or complex) verb.” This is a collocation whose base is a
predicative noun Npredic (koŊPU ‘[a] study’), and the collocate is a support verb
Vsupport (hata ‘make’); as a whole, the collocation “Npredic Vsupport” is syntactically
equivalent to a transitive verb having a regular DirO N (here, enehak ‘linguis-
tics’), something like “[makeVsupport [a] studyNpredic]Vtrans linguisticsN=DirO” ≈ “study
linguistics”. Inside this collocation, the Npredic must be encoded as a Quasi-DirO-2
of the Vsupport hata, since the latter cannot have two DirOs:

‘linguistics’ dir-obj–hata ‘make’–quasi-dir-obj-2→koŊPU ‘study’

enehak ‘linguistics’←
‘make [a] study linguistics’ = ‘study linguistics’

Passivization of (29a) produces (22), in which koŊPU ‘study’ can be considered as

a pseudo-subject (4.5.2).
The same meaning can be expressed by different syntactic structures, where
koŊPU is a regular DirO of the verb hata, and it takes enehak as its adnominal
attribute (in the genitive or the nominative), see (29b–c).
The quasi-dir-objectival-2 surface-syntactic relation is necessary for many lan-
guages. The best-known among these is, perhaps, Persian, where the role of a
transitive verb is played, most of the time, by a phrase “support verb Vsupport +
deverbal noun S0” (“[N-ra]” stands for the DirO, -ra being a postposition that
obligatorily marks a definite DirO):

(30) Persian
‘[to] end [N]’ = tämäm kärdän [N-ra] lit. ‘ending do [N]’
‘[to] begin [N]’ = aġaz kärdän [N-ra] lit. ‘beginning do [N]’
‘[to] light up [N]’ = ateš kärdän [N-ra] lit. ‘fire do [N]’
‘[to] beat [N]’ = kotak zädän [N-ra] lit. ‘beating hit [N]’
‘[to] show [N]’ = nešän dadän [N-ra] lit. ‘sign give [N]’
‘[to] learn [N]’ = yad gereftän [N-ra] lit. ‘memory take [N]’
‘[to] congratulate [N]’ = tabrik goftän [N-ra] lit. ‘congratulation say [N]’

Since the quasi-direct-objectival-2 SyntRel is not commonly accepted, it seems

worthwhile to indicate four properties of the Quasi-DirO-2 in Korean that illus-
trate its status.
200 4 “Multiple subjects” and “multiple direct objects” in Korean

(31) a. A Quasi-DirO-2 does not accept an adjectival modifier:

Johni enehakɨl *simtoissnɨn koŋpulɨl hayssta lit. ‘John linguisticsACC deep
study did’.
b. A Quasi-DirO-2 cannot be pronominalized with kukes ‘that thing’:
Johni enehakɨl koŋpulɨl hayko, Maryka suhakɨl *kukesɨl hayssta
lit. ‘John linguisticsACC study having.done, Mary mathematicsACC the.same
did’.
c. A Quasi-DirO-2 should not be linearly separated from the verb hata (other­
wise, the sentence is judged awkward by some speakers):
?
Johni koŋpulɨl enehakɨl hayssta.
d. A Quasi-DirO-2 cannot undergo relativization:
Johni ha+nɨn koŋpu ‘by.John made study’ vs.
*Johni enehakɨl ha+nɨn koŋpu ‘by.John [of] linguistics made study’.

A Quasi-DirO-2 is more constrained than a regular DirO; it seems to “coalesce”

with hata.

4.6.2.3 Direct object of the Main Verb + direct object of a non-finite verb form
The sentence in (10b), reproduced here as (32), contains two DirOs (boxed), which
depend on two different clause elements: Marylɨl is the DirO of the MV HATA
‘make’, while čhaykɨl is the DirO of the converb ilkke ‘reading’; cf. (32):

(32) a. John+i Mary+lɨl čhayk+ɨl ilk +ke hay +ss +ta

SUBJ ACC book ACC read CONV make PAST DECL
‘John made Mary read a book’.

As is normal for the DirO of an MV, Mary can be promoted to the SyntSubj in a
periphrastic passive construction (similar to the English GET-passive):

b. Mary+ka John+ey ɨyhay čhayk+ɨl ilk +ke toy +ess +ta

SUBJ DAT by book ACC read CONV become PAST DECL
‘Mary was made by John to read a book’.

This situation obtains with the phrasal (= analytical) causative construction.

 4.6 Multiple same-case nouns in a Korean clause 201

4.6.2.4 Affected object + direct object

(33) a. Mary+ka John+ɨl sačin +ɨl ččič +ess +ta
SUBJ ACC picture ACC tear.up PAST DECL
‘Mary tore up John’s picture’.

It is the picture that Mary tore up, not John: sačin ‘picture, photo’ is the DirO of
the Main Verb. And what about John? This clause element is known as an affected
object, referring to the entity affected by the event. The meaning is roughly like
this: “What Mary did to John was tear up his picture.” In Mandarin Chinese, the
affected object is introduced by the preposition bǎ and is called “retained object”
(Li & Thompson 1981: 470–471; see Chapter 2, Section 2.5, SSyntRel No. 13, p. 56,
and also Chapter 8, Section 8.4, p. 305):

b. Mali bǎ Juhen bǎng-le liǎngzhi jiǎo

Mary John tie.up PERF two foot
‘Mary tied up John’s feet’.
NB “Affected object” is simply the name of a surface-syntactic clause element; it should
not be construed as a semantic characterization. In this technical sense, wolf in John
killed the wolf is not an affected object, but simply a DirO.

4.6.2.5 Direct object + pseudo-conjunct

(34) a. Kay+ka John+ɨl son +ɨl mul+ess +ta
dog SUBJ ACC hand ACC bite PAST DECL
‘The dog bit John on the hand’. (O’Grady 1991: 3)

N1-ACC is the DirO, and each of the following Ni-ACCs is a pseudo-coordinate conjunct
of the pre­ceding NACC. Ni-ACC do not easily allow permutation—(34b-i), but can be
omitted without affect­ing the grammaticality of the sentence—(34b-ii):

b. (i) ?/*Kay+ka son+ɨl John+ɨl mulessta.

(ii) Kay+ka John+ɨl mulessta
‘The dog bit John’.

This is exactly what is to be expected from a pseudo-conjunct since it expresses

an elaboration of the preceding element.
NB The meaning of sentence (34a) can be also expressed by a different sentence with a differ-
ent syntactic structure, in which SON ‘hand’ is a DirO and JOHN is a possessor attribute in the
genitive:
Kay+ka John+uy son+ɨl mulessta ‘The dog bit John’s hand’.
202 4 “Multiple subjects” and “multiple direct objects” in Korean

4.6.2.6 “Quadruple direct objects”

Sentence (35) features four consecutive NACCs (O’Grady 1991: 77):

(35) John+i kɨ mune +lɨl tali+lɨl kkɨt pupun+ɨl čokɨm+ɨl čal+ass+ta

SUBJ the octopus ACC leg ACC end part ACC bit ACC cut PAST DECL
lit. ‘John the octopus leg end part bit cut’. =
‘John cut a bit from the end part of a leg of the octopus’.

The first NACC mune ‘octopus’ is the DirO, and the following NACCs are pseudocon-
juncts:

čalɨta ‘cut’–dir-obj→mune–pseudo-coord→tali–pseudo-coord→kkɨt+pupun
–pseudo-coord→čokɨm

The situation is, however, furthermore complicated by the fact that (roughly)
the same semantic content can be expressed by different syntactic structures,
which determine different distributions of case suffixes. Thus, we can have sen-
tences in which the noun mune remains in the accusative, but some of other
nouns obtain the nominative (these are boldfaced):

(36) a. John+i kɨ mune +lɨl tali+lɨl kkɨt pupun+Ø čokɨm+ɨl čal+ass+ta

SUBJ the octopus ACC leg ACC end part NOM bit ACC cut PAST DECL

b. John+i kɨ mune +lɨl tali+Ø kkɨt pupun+ɨl čokɨm+ɨl čal+ass+ta

SUBJ the octopus ACC leg NOM end part ACC bit ACC cut PAST DECL

c. John+i kɨ mune +lɨl tali+Ø kkɨt pupun+Ø čokɨm+ɨl čal+ass+ta

SUBJ the octopus ACC leg NOM end part NOM bit ACC cut PAST DECL

In sentences (36) an NNOM is an adnominal complement/attribute of the following

N; all NACCs keep their syntactic role. Formally:

kkɨt+pupunNOM←subj-adnom–čokɨm ‘end.part bit’,

taliNOM←subj-adnom–kkɨt+pupun ‘leg end.part’, etc.

(For the subj-adnom SSyntRel, see Chapter 5, 5.3.1/2, p. 211ff.) An adnominal depen-
dent can also be in the genitive (the suffix -ɨy): kkɨt+pupun+ɨy čokɨm+ɨl ‘bit of
end.part’ or tali+ɨy kkɨt+pupun+ɨl ‘end.part of the leg’.
Due to optional “subjective ~ nominative” and “accusative ~ nominative”
alternations, Korean allows for sequences of NNOMs:
 4.7 Conclusions 203

(37) John+Ø kɨ mune+Ø tali+Ø kkɨt pupun+Ø čokɨm+Ø čalassta

nom nom nom nom nom
lit. ‘John the octopus leg end part bit cut’.

However, such a sequence does not present new problems. The wordform
sequence in (37) implements one of the two syntactic structures:

– Either John is the SyntSubj and mUne ‘octopus’ is the DirO; each of the
following NNOMs is a pseudo-conjunct to the preceding noun (‘John cut the
octopus, on the leg, the end part, a bit’).
– Or John is the SyntSubj and čokɨm ‘bit’ is the DirO; each of the NNOMs that
precede čokɨm is an adnominal attribute to the following N (‘John cut a bit
of the end part of the leg of the octopus’).

The syntactic ambiguity of sequence (37), as well as of all such sequences, is in

fact resolved by prosody, which is not considered in this chapter.

4.6.3 Other “multiple objects” in a Korean clause

As it can be expected, Korean allows for other “multiple cases”; thus, it has
sequences of NDATs (Maling & Kim 1992):

(38) a. Nay+ka Mary+eykey kwi+ey pimil +ɨl soksaki+ess +ta

I SUBJ DAT ear DAT secret aCC whisper PaSt DeCL
lit. ‘I to.Mary to.ear secret whispered’. =
‘I whispered the secret into Mary’s ear’.

b. Koŋčaŋ+ey čhaŋko +ey pul+i na +ss +ta

factory DAT storeroom DAT fire SUBJ occur PaSt DeCL
lit. ‘In.factory in.storeroom fire occurred’. =
‘A fire broke out in the factory’s storeroom’.

Such examples do not add anything new to the discussion: the first NDAT is an
IndirO or a circumstantial, and the second is its pseudo-conjunct.

4.7 Conclusions

1. Korean has neither “multiple subjects” nor “multiple direct objects”: what
is theoretically not possible is impossible in any of the possible worlds (≈ in
204 4 “Multiple subjects” and “multiple direct objects” in Korean

any language). Korean does have, however, multiple subjective case nouns
and multiple accusative case nouns—that is, strings of NSUBJs and NACCs in one
clause.
2. The noun form commonly called “nominative” in Korean grammar is in fact
the subjective case (in -ka/-i/-kkeyse); the nominative exists as well and is
marked by a zero suffix: -Ø.
3. The Korean subjective marks the syntactic subject, the attribute of a copula-
like verb, the agent complement of a non-finite verb form, a rhematic prolep-
sis, and the oblique object of a parametric verb.
4. A string of NSUBJs represents one of three syntactic possibilities:
– either the last NSUBJ is the SyntSubj, all the preceding ones being rhematic
prolepses;
– or the N1-SUBJ is the SyntSubj and the N2-SUBJ is an AgCo (or the SyntSubj) of a
non-finite verb form;
– or else the N1-SUBJ is the SyntSubj, each of the following NSUBJs being a pseudo-
conjunct to the previous NSUBJ.
5. A string of NACCs corresponds to four syntactic possibilities:
– either the N1-ACC is the indirect object, the N2-ACC being the direct object;
– or the N1-ACC is the direct object, the N2-ACC being a quasi-direct object with a
light verb;
– or the N1-ACC is the direct object of the Main Verb, while the N2‑ACC is the direct
object of the lexical converb in the periphrastic causative;
– or else the N1-ACC is the direct object, each of the following NACCs being a
pseudo-conjunct to the previous NACC.

A corrected version of Mel’čuk 2015b.

5 G
 enitive adnominal dependents in Russian:
­surface-syntactic relations in the N→Ngen phrase
5.1 The problem stated
5.2 Criteria for distinguishing surface-syntactic relations within N→NGEN phrases
5.3 The problem solved
5.3.1/2 Genitivus Subjectivus vs. Genitivus Objectivus: the subjective-adnominal-­completive
and objective-adnominal-completive SSyntRels
5.3.3 Genitivus Qualitatis: the qualificative-adnominal-attributive SSyntRel
5.3.4 Genitivus Possessivus: the possessive-adnominal-attributive SSyntRel
5.3.5 Genitivus Attributivus: the characterizing-adnominal-attributive SSyntRel
5.3.6 Genitivus Metaphoricus: the metaphorical-adnominal-attributive SSyntRel
5.3.7 Genitivus Phrasemicus: no special SSyntRel
5.4 Overview of the six SSyntRels proposed
5.5 Closing remarks: NGEN cannot be a personal pronoun
5.6 Conclusions

5.1 The problem stated

Russian has several types of N→NGEN phrases: a noun N and its syntactic nominal
dependent NGEN in the genitive case without preposition (smert′ IvanaGEN ‘death
of.Ivan’, prosmotr fil′movGEN ‘watch­ing of.films’, čelovek udivitel′noj sud′byGEN ‘man
of.amazing destiny’). For the convenience of an overview, the NGEN dependents
in these phrases can be grouped according to the type of the semantic relation
between N and NGEN. (By “semantic relation” is meant here the relation between
the mean­ings ‘N’ and ‘NGEN’—that is, between the sources of N and NGEN in the
underlying semantic structure; ‘X’ stands for “meaning of а linguistic entity X.”)
Four major cases are logically possible:

– NGEN implements one of N’s semantic actants,

N expressing a semantic predicate (or quasi-predicate) ‘N(NGEN)’
– NGEN expresses a semantic predicate,
N implementing one of NGEN’s semantic actants ‘NGEN(N)’
– NGEN is semantically linked to N by a predicate
(or a configuration of predicates) ‘σ’ ‘σ(N, NGEN)’
NB The predicate ‘σ’ typically has no segmental (= phonemic) expression in the sentence;
the correspond­ing meaning is either carried by the surface-syntactic relation that links N
and NGEN or remains unexpressed, to be accessed by the Addressee through the context.
(One exception, leading to a semantic-syntactic mismatch, is presented below, in Subsec-
tion 5.3.5.)

https://doi.org/10.1515/9783110694765-006
206 5 Genitive adnominal dependents in Russian

– NGEN and N are semantically not linked, since neither NGEN nor N have
separate semantic sources: they form together a semantic unit; in
other words, the N→NGEN phrase is a non-compositional phraseme,
that is, an idiom or a nomineme (Mel’čuk 2012–2015: vol. 3, 293–362)1‘N_NGEN’

All these cases are represented among Russian N→NGEN phrases; the respective
examples follow. For the ease of reference, each group of examples is given a
conventional Latin name; all glosses are literal.

(1) ‘N(NGEN)’
a. Genitivus Subjectivus: NGEN-subj expresses DSynt-actant [DSyntA] I of N, e. g.:
zasedanie komitetaI ‘meeting of.committee’
otsutstvie [neskolʹkix] licI ‘absence of.several persons’
steny tualetaI ‘walls of.bathroom’
b. Genitivus Objectivus: NGEN-obj expresses DSyntA II of N (or, in some rather
infrequent cases, N’s DSyntA III, see Raxilina 2010: 253), e. g.:
sozdanie komitetaII ‘creation of.committee’
arest [neskolʹkix] licII ‘arrest of.several people’
pokupatelʹ rybyII ‘buyer of.fish’ ~
pokupatelʹ FediIII ‘buyer of.Fedya’ = ‘buyer from.Fedya’
NB NGEN-obj-ii and NGEN-obj-iii do not cooccur with the same syntactic governor (*pokupatelʹ rybyII
Fediiii vs. pokupatelʹ rybyII u FediIII ‘buyer of.fish from Fedya’), which allows us to not distin-
guish them at the surface-syntactic [SSynt-]level — that is, to use the same SSynt-relation
for both.

(2) ‘NGEN(N)’
Genitivus Qualitatis, e.g.:
ploščadka [nebolʹšogo] razmera ‘area [of.small] size’
čelovek [redkogo] uma ‘man [of.rare] intelligence’
suščestvitelʹnoe [množestvennogo] čisla ‘noun [of.plural] number’
devuška [moej] mečty ‘girl [of.my] dream’

(3) ‘σ(N, NGEN)’

a. ‘N←1–σ–2→NGEN’: the noun N is semantic actant [SemA] 1 of the predi-
cate ‘σ’

1 The components of a compositional phraseme—a collocation or a cliché—have their inde-

pendent semantic sources; NGEN in these phrasemes is subordinated to N by the charact-adnom
­SSyntRel: see Section 5.4, Item 5, p. 228.
 5.1 The problem stated 207

i. Genitivus Possessivus:
‘N←1–belong–2→NGEN’ [‘belong’ = ‘be owned’], e.g.:
igruški Miši ‘toys of.Misha’, fabrika otca ‘factory of.father’
alʹbom Anny ‘album of.Anna’
ii. Genitivus Characteristicus:
‘N←1–σ–2→NGEN’, e. g.:
vozdux Pariža ‘air of.Paris’ = ‘air existing.in Paris’
žëny [šaxskogo] garema ‘wives [of.Shah’s] harem’
= ‘wives being.elements.of the Shah’s harem’
putešestvija prošlogo veka ‘travels of.past century’
= ‘travels that.took.place.in the past century’
Mefistofelʹ Šaljapina ‘Mephisto of.Shalyapin’
= ‘Mephisto as.interpreted.by Shalyapin’
Saskija Rembrandta ‘Saskia of.Rembrandt’
= ‘Saskia as.painted.by Rembrandt’
kontinent lʹvov i žirafov ‘continent of.lions and of.giraffes’
= ‘continent inhabited.by lions and giraffes’
b. ‘N←2–σ–1→NGEN’: the noun N is SemA 2 of the predicate ‘σ’
Genitivus Metaphoricus:
‘N←2–similar–1→NGEN’ [‘NGEN is similar to N’ = ‘as if NGEN were N’], e. g.:
sutany dyma ‘soutanes of.smoke’, okean tajgi ‘ocean of.taïga’
čaša utra ‘cup of.morning’ (F. García Lorca in M. Cvetaeva’s translation)
raduga [jarkostrekočuščix] kryl ‘rainbow [of.brightly.chirping] wings’
[the title of an article about a congress of entomology]

(4) ‘N_NGEN’
Genitivus Phrasemicus: no semantic link between N and NGEN, both
forming together a semantic unit (the phrase N→NGEN is a non-composi-
tional phraseme: an idiom or a nomineme); N and NGEN-phras have no sepa-
rate semantic sources in the underlying semantic structure.
a. NGEN-phras in an idiom (the top corners ˹…˺ enclose idioms), e. g.:
˹džentelʹmen udači˺ ‘gentleman of.fortune’ ≈ ‘bandit’
˹trubka mira˺ ‘pipe of.peace’, ˹čaška Petri˺ ‘cup of.Petri’ = ‘Petri dish’
 ˹kapli [datskogo] korolja˺ ‘drops [of.Danish] king’ = ‘expectorant cough
syrup’
b. NGEN-phras in a nomineme, e. g.:
Ostrova [Zelënogo] Mysa ‘Islands [of.Green] Cape’
Mys [Dobroj] Nadeždy ‘Cape [of.Good] Hope’
ploščadʹ Puškina ‘Square of.Pushkin’
 korifej [vsex] vremën i narodov ‘corypheus [of.all] times and peoples’
[Comrade Stalin]
208 5 Genitive adnominal dependents in Russian

The first two NGENs—Genitivus Subjectivus and Genitivus Objectivus—are adnominal

complements; the following five types of NGEN—Genitivus Qualitatis, Genitivus
Possessivus, Genitivus Characteris­ticus, Genitivus Metaphoricus, and Genitivus
Phrasemicus—are adnominal attributes.
As far as I know, there is no in-depth description of SSyntRels linking an NGEN
to its syntactic governor N in Russian. In Mel’čuk 1974: 224, all Russian N→NGEN
phrases (and a host of others) were described by three SSyntRels:
– the agentive SSyntRel (priezd ministra ‘arrival of.minister’), corresponding to
deep-syntactic relation [DSyntRel] I;
– the 1st completive SSyntRel (provody ministra ‘send-off of.minister’), corre-
sponding to DSyntRel II or III; and
– the attributive SSyntRel (malʹčik [vysokogo] rosta ‘boy [of.tall] height’)—with
the admission that the attributive SSyntRel is a “dump ground” for all non-
agreeing postmodifiers of a Russian noun N that do not correspond to N’s
DSynt-actants.

This tripartite division—two actantial and one “general-attributive” NGENs—

was retained in Iomdin 2010c: 26–43 and then in Mel’čuk 2012b: 137–140 (differ-
ent names of SSyntRels being used). But today I think that the time is ripe for a
substantive linguistic analysis of Russian N→NGEN phrases, which must allow me
to better determine their SSynt-description.
The question asked in this chapter is straightforward:
How many different surface-syntactic relations—and, of course, which ones—
are needed to describe N→NGEN phrases in Russian?

The problem of acceptability—that is, of linguistic correctness—of particular

N→NGEN phrases is left out of consideration (see Raxilina 2010 and Borščëv &
Parti 2011, as well as many other studies mentioned in these titles). Only correct
N→NGEN phrases are considered here.

5.2 D
 istinguishing surface-syntactic relations within N→NGEN
phrases
To establish an inventory of SSyntRels in a language the linguist has to observe
two types of requirements (Iordanskaja & Mel’čuk 2009a).

– Linguistic requirements: all dependents of an SSyntRel must exhibit identi-

cal (or quite similar) syntactic properties relevant in the given language.
– Formal requirements: an SSyntRel must satisfy formal Сriteria A–C of the
definition of SSyntRel (Chapter 2, Section 2.4, pp. 40ff).
 5.2 Distinguishing surface-syntactic relations within N→NGEN phrases 209

In our particular case—that is, the Russian N→NGEN phrases—the linguistic

requirements are satisfied trivially: all phrases considered are of the same struc-
ture, and all dependent NGENs have the same syntactic properties except for their
mutual ordering; this latter property is used for dis­tinguishing the SSyntRels
involved. As far as Criteria A–C are concerned, Criteria A (presence of a syntactic
dependency relation between two lexemes in an utterance) and B (orientation
of the syntactic dependency relation between two lexemes in an utterance) are
irrelevant for the present discussion. Only Criteria C1 and C3 need to be used
for the definition of SSyntRels within Russian N→NGEN phrases; for the reader’s
convenience, these criteria are reproduced here. (Criterion C2, that of syntactic
substitutability, is not relevant, either: it is satisfied for all N→NGEN phrases.)

Criterion C1 (presence of semantic contrast)

Notation: wi(L) is a wordform of lexeme L.

One and the same hypothetical SSyntRel r should not describe two phrases
w1(L1)–r→w2(L2) and w3(L1)–r→w4(L2)
if Conditions 1 and 2 are simultaneously satisfied:
Condition 1
These phrases contrast semantically, the contrast being manifested
either in the form of the phrases themselves or in the syntactic behavior
properties of their members.
Condition 2
If these phrases differ in their form, they differ only by some syntactic­
means of expression—by word order of their elements, syntactic prosody,
or syntactic grammemes.

If Criterion C1 is satisfied—that is, if Conditions 1 and 2 are both satisfied, r should

be split into two different SSyntRels, r1 and r2, r1 ≠ r2.
Since this chapter only deals with the phrases of the same form (namely,
N→NGEN), Condition 2 of Criterion C1 is irrelevant (just like Criterion C2), because
it is always satisfied. Our reasoning is thus based on a semantic contrast that
manifests itself “outside” the phrase in question; and in this case, the semantic
contrast can manifest itself only by its syntactic behavior with respect to cosubor-
dinated other N→NGEN phrases—in particular, in their mutual ordering.

Criterion C3 (no limited repeatability)

An SSyntRel r must be either unlimitedly repeatable or non-repeatable—

that is, it cannot be limitedly repeatable.
Criterion C3 is actively exploited in the following reasoning.
210 5 Genitive adnominal dependents in Russian

Now we are fully equipped to take on the problem formulated in Section 5.1: What
are the SSyntRel ri in a Russian phrase of the N–ri→NGEN form?

5.3 The problem solved

Following the indications of Criteria C1 (Condition 1) and C3, the description of
the N→NGEN phrases requires six SSyntRels, which will be introduced below.
Each pair of N→NGEN phrases being contrasted must be compared strictly
under the “everything else being equal” condition, the latter understood in the
following sense:

1) The cosubordinated noun phrases being mutually ordered are of the same
weight—roughly, of the same number of stressed syllables and of the same
syntactic complexity. As is known (see, for instance, Wasow & Arnold 2003),
in a string of cosubordinated phrases postposed to their governor, heavier
phrases tend to follow lighter ones. Thus, the dubious expression ?perevod
Bunina “Gajja­vaty” ‘translation of.Bunin of.Hiawatha’ becomes perfect with
a heavier NGEN-obj phrase: perevod Bunina zamečatelʹnoj poèmy Longfello
‘translation of.Bunin of.brilliant poem of.Longfellow’.
2) No communicative factors intervene (such as topicalization, focalization,
emphasis, etc.). This means, among other things, that all the examples are
considered under neutral prosody; emphatic intonation can make acceptable
otherwise ungrammatical expressions.
3) All cosubordinated noun phrases considered below are restrictive modifiers,
since descriptive modifiers, characterized by special prosody, can violate the
standard ordering: kovry nebolʹšogo raz­mera ètogo perioda ‘carpets of.small
size of.this period’ ~ *kovry ètogo perioda nebolʹšogo razmera [restrictive
modifier], but kovry ètogo perioda, nebolʹšogo razmera, … [descriptive modi-
fier].
4) No ambiguity is created by the given linear arrangement.
 5.3 The problem solved 211

5.3.1/2 Genitivus Subjectivus vs. Genitivus Objectivus:

The subjective-adnominal-completive and objective-adnominal-completive
SSyntRels

Criterion C1, Condition 1:

(5) a. Semantic contrast between NGEN-subj and NGEN-obj

perevod Bunina ‘translation of.Bunin’:
either Bunin translated somebody/something, or somebody translated
Bunin.
b. Different syntactic behavior of NGEN-subj and NGEN-obj
i. perevod “Gajavaty”NGEN-obj BuninaNGEN-subj
‘translation of.Hiawatha of.Bunin’ vs.
?
perevod Bunina “Gajavaty”
ii. portret devočkiNGEN-obj SerovaNGEN-subj ‘portrait of.young.girl of.Serov’ vs.
?
portret Serova devočki

NGEN-subj (Genitivus Subjectivus) that corresponds to N’s SemA 1 and NGEN-obj (Geni-
tivus Object­ivus) that corresponds to N’s SemA 2 semantically contrast, see (5a).
Everything else being equal, NGEN-obj precedes NGEN-subj, that is, it is positioned
closer to their common governor N than NGEN-subj, see (5b). The word order differ-
ence in these phrases’ syntactic behavior is the manifestation of their semantic
contrast.
The semantic contrast of NGEN-subj and NGEN-obj is rather limited in scope—it is
possible only in the context of a handful of governing nouns. However, in typo-
logical perspective it is important. On the one hand, the same contrast is found
in Russian modificative adjectives: repinskie ženskie portrety ‘Repin women’s
portraits’ ~ ??ženskie repinskie portrety, where the “objectival” adjective must be
closer to the governor than the “subjectival” one. On the other hand, the linear
precedence of NGEN-obj with respect to NGEN-subj in Russian N→NGEN phrases corre-
sponds to a universal typological feature of natural languages: the direct object
manifests closer semantic ties to its governor than the subject. Two well-known
examples suffice to illustrate this point:

– The wide-spread ergative construction, where the DirO is marked by the nom-
inative case and controls the agreement of the Main Verb, while the Subject
is in oblique case and—in some languages and/or some contexts—does not
affect the form of the Main Verb.
– V–dir-obj→N collocations, whose base N is the direct object of the support
verb, like launch an attack or pay attention, are the most frequent among
verbal collocations.
212 5 Genitive adnominal dependents in Russian

Following Criterion C1, Condition 1 (the NGEN-subj and NGEN-obj phrases do not
differ in their form, but show a semantic contrast manifested in different syntac-
tic behavior—namely, different word order), NGEN-subj and NGEN-obj must be subordi-
nated to their governor N by two different SSyntRels: subjectival-adnominal(-com-
pletive) and objectival-adnominal(-completive). (The names subjectival and objectival
are meant strictly as conventional labels, without any semantic load. Thus, in
the phrases stakan–subj-adnom→moloka ‘glass of.milk’, člen–subj-adnom→partii
‘member of.party’, serdce–subj-adnom→materi ‘heart of.mother’ or pjatoe–subj-
adnom→janvarja ‘[the] fifth of.January’ the subj-adnom SSyntRel shows only that
the NGEN-subj expresses DSyntA I of N, whatever its semantic role.)
Criterion C3 confirms the proposed solution: the subj-adnom and obj-adnom
SSyntRels are both non-repeatable; if subj-adnom and obj-adnom SSyntRels are
not distinguished, the dependent NGEN will be repeatable exactly twice, which is
forbidden by Criterion C3.
In traditional descriptions of Russian, the proper semantic representation of
predicate nouns is, as a rule, lacking. Thus, the genitive peska ‘of.sand’ in kuča
peska ‘pile of.sand’ is treated as Genitivus Quantitatis, while brat Ivana ‘brother
of.Ivan’ is said to manifest Genitivus Possessivus. In point of fact, pesok ‘sand’
expresses SemA 1 (DSyntA I) of kuča (our Genitivus Subjectivus), and Ivan,
SemA 2 (DSyntA II) of brat (our Genitivus Objectivus). The overwhelming major-
ity of Russian adnominal genitives turn out to be Genitivus Subjectivus or Objecti-
vus.2 (For more on semantic predicates and semantic/deep-syntactic actants, see
Mel’čuk 2012–2015: vol. 1, 215 ff; vol. 3, 4 ff.)
The subj-adnom SSyntRel describes only N→NGEN phrases; semantically close
phrases with the instrumental case or with a preposition are represented in
the SSynt-structure in a different way: by the agentive SSyntRel (rassmotrenie–
agentive→komitetom ‘study by.committee’; dogovor–agent­ive→meždu stranami
‘treaty between countries’).
The obj-adnom SSyntRel also describes only N→NGEN phrases; the Ni that
depends on N and is not in the genitive is subordinated to N by the indir-objectival
or oblique-objectival SSyntRel (podarok–indir-objectival→IvanuDAT ‘gift to.Ivan’;

2 Mel’čuk 2016 proposes a slightly different syntactic description of Russian N→NGEN-subj and
N→NGEN-obj phrases. Namely: 1) The present subj-adnom SSyntRel was called agentive-attributive;
this agentive-attributive SSyntRel covered also N→NINSTR phrases, for which I reserve now the agen-
tive SSyntRel. 2) The present obj-adnom SSyntRel was called patientive-attributive. 3) There was the
actantial-attributive SSyntRel, designed to describe the N→NGEN phrases in which NGEN expresses
N’s DSyntA I or II not corresponding to the syntactic subject or the direct object. Now I believe
that this description is too semantic and replace it.
 5.3 The problem solved 213

zanjatija–oblique-objectival→matematikojINSTR lit. ‘doing with.mathe­

matics’ =
‘studying/working in mathematics’).

5.3.3 Genitivus Qualitatis: the qualificative-adnominal-attributive SSyntRel

Criterion C1, Condition 1:

(6) a. Semantic contrast between NGEN-qual and NGEN-subj/NGEN-obj

portret neobyčnoj formy ‘portrait of.extraordinary form’:
either the form of the portrait is extraordinary (NGEN-qual),  or
the portrait was painted by somebody called “Extraordinary Form”
­(NGEN-qual),  or
else the portrait represents somebody/something called “Extraordinary
Form” (NGEN-obj).
b. Different syntactic behavior of NGEN-qual with respect to both NGEN-subj
and NGEN-obj
i.  portret neobyčnoj formyNGEN-qual Adeli BloxNGEN-obj blestjaščego KlimtaNGEN-subj
‘portrait of. extraordinary form of.Adel Bloch of.brilliant Klimt’
neobyčnoj formy portret Adeli Blox blestjaščego Klimta  and
vs. *portret Adeli Blox blestjaščego Klimta neobyčnoj formy
??portret Adeli Blox neobyčnoj formy blestjaščego Klimta  and
ii. tovary vysšego sortaNGEN-qual našego magazinaNGEN-obj
‘products of.highest class of.our store’
vs. *tovary našego magazinaNGEN-obj vysšego sortaNGEN-qual

As in the preceding case, the semantic contrast in (6a) is manifested through

different syntactic behavior of NGEN-qual with respect to NGEN-subj/NGEN-obj, see (6b):
everything else being equal, NGEN‑qual precedes NGEN-subj and NGEN-obj. Similarly:

(7) statuja ogromnogo razmera Aleksandra Tretʹego Paolo Trubeckogo

‘statue of.huge size of.Alexander III of.Paolo Trubetzkoy’ and
ogromnogo razmera statuja Aleksandra Tretʹego Paolo Trubeckogo
vs. ??statuja Aleksandra Tretʹego Paolo Trubeckogo ogromnogo razmera and
*statuja Aleksandra Tretʹego ogromnogo razmera Paolo Trubeckogo

Deviation from the standard ordering NGEN-qual + NGEN-subj

If N denotes a set or a quantity that measures the denotation of NGEN-subj, then NGEN-subj
precedes NGEN-qual, see (8).
214 5 Genitive adnominal dependents in Russian

(8) a. kuča morskogo peska ogromnogo razmera ‘pile of.sea sand of.huge size’
vs. ??kuča ogromnogo razmera morskogo peska
b. tolpa studentov-fizikov ogromnogo razmera
‘crowd of.students physicists of.huge size’
vs. ??tolpa ogromnogo razmera studentov-fizikov

In what follows, we will see other cases where the meaning of N or of NGEN plays a role
in determining the mutual ordering of different NGENs, see Subsection 5.3.5, p. 220.

The indicated standard ordering can be violated, for instance, by the weight of
the phrase under consideration:

(9) a. statuja Friny Praksitelja neobyčajnogo izjaščestva

‘statue of.Phryne of Praxiteles of.extraordinary elegance’
vs. *statuja neobyčajnogo izjaščestva Friny Praksitelja
b. fragmenty DNK fiksirovannogo razmera ‘fragments of.DNA of.fixed size’
vs. fragmenty fiksirovannogo razmera različnyx dezoksiribonukleinovyx kislot
‘fragments of.fixed size of.various desoxyribonucleic acids’

In this construction, N typically expresses NGEN’s SemA 1: ‘N←1–NGEN’, as, e.g., in

portretN neoby­čajnoj krasotyNGEN ‘portrait of.extraordinary beauty’; less frequently,
N can implement SemA 2 of NGEN: ‘N←2–NGEN’, as, e. g., in devuška moej mečty
‘girl of.my dream’ [= ‘a girl of whom I dream’], lososʹ xolodnogo kopčenija ‘salmon
of.cold smoking’ [= ‘salmon that has been smoked cold’] or sumka ručnoj raboty
‘bag of.handiwork’ [= ‘the bag that has been manufactured manually’].

Criterion C3:
NGEN-qual is repeatable; we can have, for instance, three cosubordinated NGEN-quals:
neobyčajnoj krasotyNGEN-qual šarfik jarkogolubogo cvetaNGEN-qual
nebolʹšogo razmeraN GEN-qual
‘of.extraordinary beauty little.scarf of.bright.light.blue color of.small size’

Note that the problem of mutual ordering of different cosubordinated NGENs is

not considered in this chapter; however, since the main tool for establishing
different SSyntRels in N→NGEN phrases is exactly their mutual ordering, this
problem cannot be completely avoided. Thus, the repeatability of the qualificat­ive-
adnominal SSyntRel raises the following question. Suppose a noun N has two or
more qual-adnom dependents; what should be their mutual linear arrangement?
How do we specify it, since a particular order may be preferable? Thus, sumka
krasnogo cveta sovremennogo dizajna ‘handbag of.red color of.modern design’
 5.3 The problem solved 215

is OK, while ??sumka sovremennogo dizajna krasnogo cveta ‘handbag of.modern

design of.red color’ is not. Shouldn’t this force us to distinguish the SSyntRel sub-
ordinating cvet ‘color’ from the SSyntRel subordinating dizajn ‘design’:

N–r1→cvet and N–r2→dizajn, where r1 ≠ r2?

The answer is no, and the reason is as follows:

The linear position of a qual-adnom dependent NGEN-1 with respect to another

qual-adnom de­pendent NGEN-2 is determined by the meaning of these depen-
dent NGENs: an NGEN that denotes the color (of N’s denotation) tends to
precede an NGEN denoting its design, etc.

The situation is identical to what holds for many codependent (= cosubordinated)

adjectives modifying the same noun: as shown in Iordanskaja 2000 for Russian
and in Iordanskaja & Mel’čuk 2017: 221–237 for French (based on the classic work
Vendler 1968), a string of anteposed codepend­ent adjectives is linearized accord-
ing to their meanings;3 roughly:

“ subjective estimate” > “size” > “spacial position” > “form” > “color” > “material” > “kind” N
(udivitelʹnaja ogromnaja vnešnjaja kruglaja krasnaja kirpičnaja protivola­vinnaja stena
‘amazing enormous external round red brick anti-avalanche wall’)

NB Note that the order of anteposed modifiers is (roughly) a mirror image of that of postposed
modifiers. In point of fact, we deal here with the proximity of different modifiers to the noun
modified.

The mutual ordering of codependent NGENs in Russian is considered in Chapter 11,

pp. 381ff.
Different qual-adnom NGENs are linearly ordered between themselves based on
the same principle, viz. according to their meanings. It must, however, be empha-
sized that this rule works, of course, only under the condition “everything else
being equal”—that is, if the cosubordinated genitive-noun phrases being mutu-
ally ordered are of the same weight, etc.

Genitivus Qualitatis in Russian has at least three relevant particularities:

– NGEN-qual requires a modifying adjective: *portret krasoty ‘portrait of.beauty’;

some NGEN-quals (lexically marked) allow — instead of an adjective — a modifying

3 English has so-called Royal Order of Adjectives; see, for instance,

https://theeditorsblog.net/2014/04/08/keeping-adjectives-in-line/ or
https://www.ucl.ac.uk/internet-grammar/adjectiv/ordering.htm
216 5 Genitive adnominal dependents in Russian

genitive noun or an apposition: statuètka rabotyNGEN-qual ČelliniGEN ‘statu-

ette of.work of.Cellini’, traktor zavoda “Krasnyj Molot” ‘tractor of.plant «Red
Hammer»’. NGEN-qual can also be modified by an idiom: gostinica ˹srednej ruki˺
‘hotel of.middle hand’ = ‘hotel of mediocre quality’ ~ ˹srednej ruki˺ gostinica.
– NGEN-qual can be anteposed with respect to N under certain conditions (see
Chapter 11, Section 11.1, p. 371).
– Not every noun can appear as NGEN-qual:
devuška neobyčajnoj sudʹby ‘girl of.extraordinary destiny’ vs.
*devuška neobyčajnoj učasti ‘girl of.extraordinary fate’.

This constraint seems to be lexical (rather than semantic); therefore, all nouns
that can be NGEN-qual (or those that cannot?) must be supplied with a special syn-
tactic feature.
The same considerations as in Subsection 5.3.2 (based on Criteria C1 and C3)
allow for postulating the third SSyntRel for Russian N→NGEN phrases: qualifica-
tive-adnominal(-attributive).
The qual-adnom SSyntRel describes not only the N→NGEN phrases, but also
three other con­structions:

– N→[N1(parameter)INSTR + v ‘in’ + NUM←N2(measure)ACC]; for instance:

mostN–qual-adnom→širinojN1-INSTR v 10 metrovN2 ‘bridge by.width in 10 meters’
– N→[N1(parameter)INSTR + NUM←N2 (measure)NOM]; for instance:
mostN–qual-adnom→širinojN1-INSTR 10 metrovN2 ‘bridge by.width 10 meters’;
– N→[NUM←N2(measure)NOM] + V ‘in’ + N1(parameter)ACC ; for instance:
mostN–[10 metrov]–qual-adnom→v širinuN1-ACC ‘bridge 10 meters into width’.

In the perspective of text synthesis, the choice between these constructions

and –qual-ad­nom→NGEN-qual is done according to the dependent of N1(parameter): if
N1(parameter) has a dependent of the form –qual-adnom→N2(measure)→NUM (= express-
ing a numerical value), then N1(parameter) has the instrumental case or is introduced
by the preposition v ‘in’ and cannot be anteposed; otherwise, N1(parameter) is in the
genitive and can be anteposed. Cf.:
most širin+oj (v) 10 metrov ~ *širin+oj (v) 10 metrov most vs.
 ost neobyčajnoj širin+y ‘bridge of.extraordinary width’ ~
m
neobyčajnoj širin+y most.

In all the remaining types of the N→NGEN phrase, ‘NGEN’ is semantically not
linked to ‘N’ directly by a predicate-argument relation: either ‘NGEN’ and ‘N’ are
linked indirectly—via an additional pre­dicate (or a configuration of predicates),
or they are semantically not linked at all, forming a non-compositional phraseme.
 5.3 The problem solved 217

5.3.4 Genitivus Possessivus: the possessive-adnominal-attributive SSyntRel

NGEN-poss and N are semantically linked indirectly—via the predicate ‘belong.to’ [=

‘be.owned.by’]:

(10) sad otca ‘garden of.Father’

= ‘garden belonging.to Father’
derevʹja soseda ‘trees of.neighbor’
= ‘trees belonging.to the neighbor’
stadion universiteta ‘stadium of.University’
= ‘stadium belonging.to the University’
bolʹnica ministerstva ‘hospital of.ministry’
= ‘hospital belonging.to the ministry’
[zamorskie] territorii Francii ‘[overseas] territories of.France’
= ‘territories belonging.to France’

Consequently, NGEN-poss denotes a person in the broadest sense: an individual, an

organization, a country, etc., and N, an entity that can be owned. This means
that ‘N’ can be only a semantic name or a quasi-predicate, so that the possibility
of a semantic contrast between NGEN-poss and NGEN-subj/obj is limited, although not
excluded.

Criterion C1, Condition 1:

(11) a. Semantic contrast between NGEN-poss and NGEN-subj/NGEN-obj

skulʹptura Nikolaeva ‘sculpture of.Nikolaev’:
either skulʹptura NikolaevaNgen-poss ‘sculpture belonging to Nikolaev’;
or skulʹptura NikolaevaNgen-subj ‘sculpture created by Nikolaev’;
or else skulʹptura NikolaevaNgen-obj ‘sculpture representing Nikolaev’.

b. Different syntactic behavior of NGEN-poss with respect to NGEN-subj/NGEN-subj

and NGEN-qual
i. skulʹptury NikolaevaNgen-subj [Omskogo] muzejaNgen-poss
‘sculptures created by Nikolaev belonging to Omsk Museum’ ~
*skulʹptury [Omskogo] muzeja Nikolaeva
ii. fabriki [kuxonnoj] mebeliNgen-obj [našego] gorodaNgen-poss
‘factories of kitchen furniture belonging to our town’ ~
*fabriki [našego] goroda [kuxonnoj] mebeli
218 5 Genitive adnominal dependents in Russian

iii. park [ogromnogo] razmeraNgen-qual [našego] gorodaNgen-poss

‘park [of.huge] size [of.our] town’ ~
*park [našego] goroda [ogromnogo] razmera (for the meaning ‘huge-
size park’)

As far as linear ordering is concerned, NGEN-poss follows all other NGENs.

Since NGEN-poss expresses the semanteme ‘belong.to’, it is quite natural to intro-
duce the correspond­ing SSyntRel: possessive-adnominal-attributive.4

Criterion C3:

The poss-adnom SSyntRel is non-repeatable, just as the subj-adnom and obj-adnom

SSyntRels.

At the DSynt-level, the NGEN-poss is marked by the fictitious lexeme «prinadležatʹ»

[= «belong»].
The “possessive” syntactic relation—interpreting “possession” in the most
liberal way possible— occupies a place of honor in linguistic typology (see, e. g.,
Aikhenvald 2013). On the one hand, all actual uses of the genitive case developed
out of its possessive use (in the strict sense of ownership); on the other hand,
languages manifest a multitude of formal means to express “possession.” This is
a weighty argument in favor of introducing the genitive-possessive SSyntRel.
The possessive-adnominal SSyntRel describes exclusively N→NGEN phrases.

5.3.5 Genitivus Attributivus: the characterizing-adnominal-attributive SSyntRel

NGEN-attr and N are semantically linked by an “additional” predicate ‘σ’:

‘N←i–σ–j→NGEN-attr’.
This can be almost any general binary predicate, as is seen in (12):5

4 The possessive SSyntRel was proposed for English (Mel’čuk & Pertsov 1987: 139–140; Mel’čuk
2016: 97) to describe N’s←N phrases (Dad’s arrival, a whole month’s work).
5 “Almost” is necessary since some constraints do exist. First, this ‘σ’ is different from ‘belong.to’
(the possessive SSyntRel) and ‘similar.to’ (the metaphorical SSyntRel). Second, as Raxilina (2010:
253) noted, the predicative semanteme ‘X prednaznačen dlja Y-a’ ≈ ‘X is for Y’ cannot be expressed
by NGEN but requires an explicit expression: ‘book that is.for Pete’ ⇔ kniga dlja Peti/*kniga Peti.
And, of course, there can be other such cases.
 5.3 The problem solved 219

(12) vozdux gor ‘air of.mountains’

= ‘air that exists.in the mountains’
životnye savanny ‘animals of.savannah’
= ‘animals that live.in the savannah’
filosofija dvadcatogo veka ‘philosophy of.twentieth century’
= ‘philosophy practiced.in 20th century
strana lʹvov ‘country of.lions’
= ‘country that is.inhabited.by lions’
dela [minuvšix] dnej [Puškin] ‘events [of.past] days’
= ‘events that took.place.in the past’
krik boli ‘cry of.pain’
= ‘cry caused.by pain’
[dva] časa dnja/noči ‘[two] o’clock of.day [PM] / of.night [AM]’
= ‘two o’clock during the day/the night’
KADEŠ [drevneegipetskix] xronik ‘KADESH [of.ancient.Egyptian] chronicles’
= ‘KADESH that is.mentioned.in ancient Egyptian chronicles’

Criterion C1, Condition 1:

(13) a. Semantic contrast between NGEN-attr and NGEN-subj

i. lob borcaNGEN-attr ‘forehead of.wrestler’ [A. Žolkovskij]
= ‘forehead typical.for a wrestler’ vs.
ii. lob [ètogo] borcaNGEN-subj forehead [of.this] wrestler’
iii. polovina [18-go] vekaNGEN-attr ‘half of.18th century’
= ‘a half [of an artifact] manufactured.in the 18th century’
(while the other half was manufactured in a different century) vs.
iv. polovina [18-go] vekaNGEN-subj ‘[one] half of.18th century’

b. Different syntactic behavior of NGEN-attr with respect to NGEN-subj/ NGEN-obj,

NGEN-qual and NGEN-poss
lob borcaNGEN-attr [našego] polkovnikaNGEN-subj ‘forehead of.wrestler [of.our]
i. 
colonel’ ~ *lob [našego] polkovnika borca
ii. bjust karrarskogo mramoraNGEN-attr molodoj ženščinyNGEN-obj
‘bust of.Carrara marble of.young woman’ = ‘bust made.of Carrara
marble’ ~ ?bjust molodoj ženščiny karrarskogo mramora
NB The construction in (13b-ii) is quite similar to the constructions described by the qual-
adnom SSyntRel, see Subsection 5.3.3. However, in spite of this similarity, there are two
essential differences:
220 5 Genitive adnominal dependents in Russian

– 
Semantic difference: NGEN-qual expresses a predicate denoting a property and takes N
as its Sem-actant, while NGEN‑attr in (13b-ii) denotes a substance—material of which the
denotation of N is made.
Syntactic difference: NGEN-qual can precede N, but NGEN‑attr cannot: neobyčajnoj krasoty
– 
bjust molodoj ženščiny ‘of.extraordinary beauty bust of.young woman’ vs. *karrarskogo
mramora bjust molodoj ženščiny.

iii. vozdux [neobyčajnoj] čistotyNGEN-qual [gimalajskix] vysotNGEN-attr

‘air [of.extraordinary] purity [of.Himalayan] heights’ [= ‘air existing.
in Himalayan heights’] ~
?
vozdux gimalajskix vysot neobyčajnoj čistoty
mašina [moskovskogo] avtozavodaNGEN-attr ètogo] general-majoraNGEN-poss
iv. 
‘car [of.Moscow] automaker [of.this] major-general’ [= ‘car manufa-
tured.by the Moscow automaker belonging.to this major-general’] ~
*mašina ètogo general-majora moskovskogo avtozavoda

Deviations from the standard ordering NGEN-attr + NGEN-obj and NGEN-qual + NGEN-attr
1. If NGEN-attr denotes localization (of N), it follows the cosubordinated NGEN-obj,
see (14a).
2. If NGEN-attr denotes material or kind (of N), it precedes a cosubordinated NGEN-qual,
see (14b–c).

(14) a. fabriki obuviNGEN-obj ItaliiNGEN-attr ‘factories of.shoes of.Italy’ = ‘factories situ-

ated.in Italy’ ~ *fabriki Italii obuvi
b. stol krasnogo derevaNGEN-attr ogromnyx razmerovNGEN-qual
‘table of.red wood of.huge dimensions’ ~
*stol ogromnyx razmerovNGEN-qual krasnogo derevaNGEN-attr

The versatility of the predicate ‘σ’, which semantically underlies the N→NGEN-attr
phrase, reminds one of nominal compounds, e. g. in English. The attempts at
describing semantic relations between the members of an English nominal com-
pound—that is a phrase of the N1 + N2 type—are numerous; suffice it to indicate,
for instance, the classic Hatcher 1960 and Levi 1978, as well as more recent Weis-
kopf 2007 ones. The researchers specify a couple dozen meanings, insisting,
however, that their inventory is not and cannot be exhaustive. That is what I think
as well; but in this chapter I will not try to circumscribe more precisely the range
of possible ‘σ’ in the Russian semanteme configuration ‘N←i–σ–j→NGEN‑attr’.
The N→NGEN-attr phrase is described by the characterizing-adnominal-attributive
SSyntRel.
The charact-adnom SSyntRel is opposed to the subj-adnom, obj-adnom, qual-
adnom, and poss-adnom SSyntRel.
 5.3 The problem solved 221

Criterion C3:

Since ‘σ’ is so variegated, the NGEN-attr is repeatable:

(15) kolʹco [dutogo] zolotaNGEN-attr [šestnadcatogo] vekaNGEN-attr6

‘ring [of.filled] gold [of.sixteenth] century’

The description of the N→NGEN-attr phrase requires two additional remarks.

– Not every NGEN semantically fit for the N→NGEN-attr construction can be freely
used in it. First, several NGEN-attrs are subject to semantic constraints (Raxi-
lina 2010); for instance, in the N→NGEN-attr phrase with the underlying predi-
cate ‘be.in’ the noun N must be used generically: devuški Moskvy or každaja
devuška Moskvy, but not *èta devuška Moskvy. Second, there are also lexical
constraints: thus, mebelʹ [krasnogo] dereva ‘furniture [of.red] wood’ = ‘of
mahogany’ is perfectly OK, while *mebelʹ [karelʹskoj] sosny ‘furniture [of.
Karelian] pine’ is impossible (the correct expression is mebelʹ iz [‘from’]
karelʹskoj sosny). Similarly, kolonny [čërnogo] mramora ‘columns [of.black]
marble’ vs. *kolonny [zolotistogo] pesčanika ‘columns [of.golden] sandstone’
(the correct expression is kolonny iz [zolotistogo] pesčanika). Therefore, the
nouns that can be used as dependents of the qual-adnom SSyntRel must be
lexically marked—that is, they must be supplied with a special syntactic
feature. This applies at least to the names of materials.
– Since the charact-adnominal SSyntRel is so “loose,” it can cover cases of
semantic-syntactic mis­matches in which an NGEN participates; here is one
such case, linked to particular lexical units (or classes of lexical units).

The semantic-syntactic mismatch linked to lexical units of ljubimyj ­‘favorite’

type:

‘X←1–ljubimyj–2→Y’ ⇔ ljubimyj←ATTR–L(‘X’)–ATTR→L(‘Y’) ⇔
ljubimyj←modif–L(‘X’)–charact-adnom→L(‘Y’) ⇔
ljubimyj šokoladX PetiY ‘favored chocolate of.Petya’

The noun NY, which semantically depends on ‘ljubimyj’ (it is its Sem-actant 2),
depends syn­tactically (as an NGEN-attr) on the noun NX, modified by ljubimyj. (Cf.
Partee & Borschev 2000 on the similar behavior of the English adjective favorite.)
The adjective rodnoj ‘native’ [= ‘where someone was born’] behaves in the
same way:

6 NGEN-attr denoting material requires an adjectival modifier: *kolʹco zolotaNGEN-attr ‘ring of.gold’.
222 5 Genitive adnominal dependents in Russian

‘X←1–rodnoj–2→Y’ ⇔ rodnoj←ATTR–L(‘X’)–ATTR→L(‘Y’) ⇔
rodnoj←modif–L(‘X’)–charact-adnom→L(‘Y’) ⇔
rodnoj gorodX PetiY ‘native town of.Petya’

To this we have to add all superlatives:

‘X←1–samyj.znamenityj–2→Y’ ⇔ znamenityjSUPERL←ATTR–L(‘X’)–ATTR→L(‘Y’)
samyj znamenityj xokkeistX KanadyY ‘the.most famous hockey-player of.Canada’

The corresponding representations are given in Section 5.4. This mismatch is due
to the fact that the predicate ‘σ’ linking N and NGEN is, in this case, expressed by a
lexeme that does not accept NGEN as syntactic dependent.
Along with the N→NGEN-attr phrases, the charact-adnom SSyntRel describes as
well all PREP→N phrases functioning as adnominal attributes:

(16) a. knigi s poželtevšimi stranicami v kožanyx pereplëtax

books with yellowish pages in leather bindings
b. prestuplenija kommunizma protiv čelovestva v točnom smysle
crimes of.Communism against humanity in proper meaning
slova
of.the.word

5.3.6 Genitivus Metaphoricus: the metaphorical-adnominal-attributive SSyntRel

NGEN-metaph and N are semantically linked indirectly — via the predicate ‘be.similar.
to’, but with what is known as head-switching:

‘X←1–similar–2→Y’ ⇔ L(‘Y’)–ATTR→«predstavljatʹ»–II→L(‘X’)
zvëzdy, poxožie na iskry ‘stars similar to sparks’ ⇒ iskry zvëzd ‘sparks of.stars’

This construction is marked in the DSynt-structure by the fictitious lexeme

«predstavljatʹ» [= «represent»], which marks the expression as metaphoric.
NGEN-metaph expresses the basis of a metaphoric transfer to N. Suppose that the
Speaker wants to compare the Moon—the basis of a metaphoric transfer—to a
cold eye, which is the metaphor: the Moon is similar to a cold eye; and he says
[xolodnyj] glaz luny ‘[cold] eye of.moon’ (see, for instance, Mixeev 2000).

(17) iskry zvëzd ‘sparks of.stars’; sutany dyma ‘soutanes of.smoke’

lenta dorogi ‘ribbon of.road’; bacilly straxa ‘bacilli of.fear’
poluxleb ploti ‘halfbread of.flesh’ [O. Mandelštam]
[opozdavšie] pticy gazet [R. Roždestvenskij] ‘[belated] birds of.newspapers’
 5.3 The problem solved 223

bljudečki-očki [spasatelʹnyx] krugov [V. Majakovskij] ‘saucers-eyeglasses

of.life.buoys’

Criterion C1, Condition 1:

(18) a. Semantic contrast between NGEN-metaph and NGEN-subj/NGEN-obj/NGEN-qual/

NGEN-attr/NGEN-poss
i. kolesnica solnca ‘chariot of.Sun’:
kolesnica solncaNGEN-metaph ‘Sun as if it were a chariot’ vs.
kolesnica SolncaNGEN-subj /NGEN-poss
‘chariot of somebody [e. g., a god] called Sun’
ii. pytka ljubvi ‘torture of.love’:
pytka ljubviNGEN-metaph ‘love as if it were a torture’ vs.
pytka ljubviNGEN-obj/NGEN-attr ‘torture applied to love/induced by love’
iii. požar cvetov ‘fire of.colors’:
požar cvetovNGEN-metaph ‘colors as if they were a fire’ vs.
požar [raznyx] cvetovNGEN-qual ‘fire [of.different] colors’
b. Different syntactic behavior of NGEN-metaph with respect to NGEN-qual and NGEN-attr
(with all other NGENs, NGEN-metaph cannot cooccur for semantic reasons)
i. iskry [neobyčajnoj] jarkostiNGEN-qual [takix dalëkix] zvëzdNGEN-metaph ~
‘sparks [of.extraordinary] brightness [of.so faraway] stars’
*iskry [takix dalëkix] zvëzd neobyčajnoj jarkosti
[for the meaning ‘sparks are extraordinarily bright’]
poluxleb plotiNGEN-metaph MandelʹštamaNGEN-attr
ii. 
‘half-bread of.flesh of.Mandelstam’
[= ‘expression “half-bread of flesh” used by Mandelstam’] ~
*poluxleb Mandelʹštama ploti

The SSyntRel for NGEN-metaph can be called metaphorical-adnominal-attributive.

Criterion C3:

The metaphorical-adnominal SSyntRel is non-repeatable.

Note that the metaph-adnom SSyntRel is used in the collocations with the LF
Figur:7

7 Lexical function Figur returns for a lexical unit L the lexical unit L′ that expresses the stan-
dard metaphor for L:
Figur(tuman ‘fog’) = pelena [tumana] ‘curtain of.fog’ or
Figur(gnev ‘anger’) = plamja [gneva] ‘flame of.anger’.
224 5 Genitive adnominal dependents in Russian

stena–metaph-adnom→doždja plamja–metaph-adnom→strasti
‘wall of.rain’ ‘flame of.passion’
červʹ–metaph-adnom→somnenija luč–metaph-adnom→nadeždy
‘worm of.doubt’ ‘ray of.hope’
grad–metaph-adnom→pulʹ znamja–metaph-adnom→borʹby
‘hail of.bullets’ ‘banner of.fight’

The metaph-adnom SSyntRel describes only N→NGEN phrases.

5.3.7 Genitivus Phrasemicus: no special SSyntRel

NGEN-phrass appear within phrasemes and come in two major types: an NGEN-phras
being part of a com­positional phraseme (a collocation or a termeme) and an
NGEN-phras being part of a non-compositional phraseme (an idiom or a nomineme).
This difference is relevant since in the deep-syntactic structure, a compositional
phraseme is represented by its complete subtree (so that the NGEN must be present
already at this level), while a non-compositional phraseme appears as a single
node (and the NGEN enters the scene only in the surface-syntactic structure).

(19) a. Collocations
i.  čelovek dela 〈dolga, slova, česti〉
‘man of.business 〈of.duty, of.word, of.honor〉’
dom [našix] grëz ‘house [of.our] dreams’
roman veka ‘novel of.century’
ii. gvardii seržant ‘of.Guards sergeant’
ordena [Lenina] zavod «Molot» ‘of.Order [of.Lenin] factory «Hammer»’ =
‘«Hammer» factory decorated with the Order of Lenin’
b. Termemes
dvigatelʹ [vnutrennego] sgoranija ‘engine [of.internal] combustion’
zakon Oma ‘law of.Ohm’; boleznʹ Alʹcgejmera ‘disease of.Alzheimer’

(20) a. Idioms
˹čaška Petri˺ ‘cup of.Petri’ = ‘Petri dish’
˹koktejlʹ Molotova˺ ‘cocktail of.Molotov’ = ‘Molotov cocktail’
˹roza vetrov˺ ‘rose of.winds’ = ‘compass rose’
˹krik duši˺ ‘scream of.soul’ = ‘verbal expression of very strong emotions’
˹pojas vernosti˺ ‘belt of.fidelity’ = ‘chastity belt’
˹dama serdca˺ ‘lady of.heart’ = ‘beloved woman’
 5.3 The problem solved 225

˹pir Valtasara˺ ‘feast of.Belshazzar’

˹lico [kavkazskoj] nacionalʹnosti˺ ‘person [of.Caucasian] ethnicity’ =
‘[a] native of the Caucasus region’
b. Nominemes
More Laptevyx ‘Sea of.Laptevs’
sozvezdie Gončix Psov ‘Constellation of.Greyhounds’
Ostrova Zelënogo Mysa ‘Islands [of.Green] Cape’

An NGEN-phras is special only in that it is an element of a phraseme, since from a

purely syntactic viewpoint it is like any other, not phraseologized NGEN. However,
exactly because of its phraseologic­al nature, it linearly precedes all other NGENs
and, in some cases (specified lexically), it can or must be anteposed to its N. These
particularities of an NGEN-phras can be indicated in the surface-syntactic struc­ture
in one of two ways: either by a special SSyntRel or by special syntactic features
of the NGEN-phras. Postulating for NGEN-phrass a genitive-phrasemic-adnom SSyntRel
seems, at first blush, an easy solution, but, unfortunately, it cannot be accepted.
The reason is simple: There are lots of phraseologized clause elements that show
unusual word order and other “deviations.” Thus:

– Phraseologized modifiers, with obligatory postposition of the modifying

adjective, e. g.: papa–modificative→Rimskij ‘Pope Roman’ or xmyrʹ–modi­fi­
cative→bolotnyj ‘douche.bag swampy’ ≈ ‘insi­gnificant, despicable man’.
– Phraseologized subjects, with obligatory anteposition/postposition of the
subject, e. g., the idioms ˹Čërt←subjectival–[ego]–poberi!˺ ‘Devil him take!’
= ‘Let the devil take him!’, with obligat­ory anteposition of the subject, vs.
˹Ne daj–subjectival→Bog!˺ ‘Not allow God!’ = ‘God forbid!’, with obligatory
postposition of the subject.
– Phraseologized direct objects with obligatory anteposition of the DirO, e. g.,
the idioms ˹palʹčiki←dir-obj–obližešʹ˺ ‘fingers you.will.lick’ = ‘this is very
tasty’ or ˹sobaku←dir-obj–sʹʹestʹ˺ ‘dog [to] eat’ = ‘be very experienced’.
– Phraseologized circumstantials, with obligatory anteposition of the circum-
stantial (and of the DirO), e. g., the idioms
direct-obj

Moju ustalostʹ ˹kak←circumstantial–[rukoj]–snjalo˺

‘My tiredness as with.hand [it] took.away’. = ‘My tiredness vanished in a trice’.
direct-obj

Ivana ˹kak←circumstantial–[vetrom]–sdulo˺
‘Ivan as by.wind [it] blew away’. = ‘Ivan disappeared in a trice’.
226 5 Genitive adnominal dependents in Russian

– Phraseologized preposition complements, with obligatory postposition of

the PREP, e. g., the idiom ˹ne korysti←prepositional–radi˺ ‘not gain for’ = ‘not
for gain’.

If we systematically apply this solution—that is, if we introduce special X-phrase-

mic SSyntRels, we will have to double all SSyntRels whose dependents can be
phraseologized and as a result acquire “exotic” syntactic properties within
phrasemes. Therefore, we are forced to accept the opposite solution: a lexeme
that manifests deviant behavior because it is part of a phraseme must receive the
corresponding syntactic features at the moment where the node of this phraseme
is expanded into its subtree. The conclusion: there is no special phrasemic-adnom
SSyntRel; an NGEN-phras is subordinated to its governing N by the charact-adnom
SSyntRel. For instance:

koktejlʹ–charact-adnom→Molotova ‘cocktail of.Molotov’ =

‘incendiary weapon—a glass bottle with flammable liquid …’

5.4 Overview of the six SSyntRels proposed

Six SSyntRels are proposed for the description of the Russian N→NGEN phrases:
1) subjectival-adnominal-completive 4) possessive-adnominal-attributive
2) objectival-adnominal-completive 5) characterizing-adnominal-attributive
3) qualificative-adnominal-attributive 6) metaphorical-adnominal-attributive

For each of these six SSyntRels the corresponding formal representations are
given: the semantic subnetwork—its semantic source, the deep-syntactic subtree,
the surface-syntactic subtree, as well as an example.

1) Subjectival-adnominal-completive SSyntRel 8

Sem ‘Y–1→X’ 8
DSynt L(‘Y’)(N)–I→L(‘X’)(N)
SSynt L(‘Y’)(N)–subj-adnom→L(‘X’)(N)
Example ‘spatʹ‘Y’–1→Ivan‘X’’ [= ‘Ivan sleeps’]
son–I→Ivan ‘Ivan’s sleep’
son–subj-adnom→Ivan: son Ivana

8 A reminder: underscoring of a semanteme ‘σ’ in a semantic structure ‘S’ shows its communica-
tively dominant status: ‘σ’ is a minimal paraphrase of the whole ‘S’, such that ‘S’ can be reduced
to ‘σ’ with loss, but without distortion, of information. Thus, ‘dog←1–sleeps’ represent A/The
dog sleeps, and ‘dog←1–sleeps’ underlies a/the sleeping dog and a/the dog that is sleeping.
 5.4 Overview of the six SSyntRels proposed 227

2) Objectival-adnominal-completive SSyntRel

Sem ‘Y–2→X’

DSynt L(‘Y’)(N)–II→L(‘X’)(N)

SSynt L(‘Y’)(N)–obj-adnom→L(‘X’)(N)

Example ‘nagraditʹ‘Y’–2→Ivan‘X’’ [= ‘[Somebody] decorates Ivan [with a medal]’]

nagraždenie–II→Ivan ‘decorating of Ivan’
nagraždenie–obj-adnom→Ivan: nagraždenie Ivana

3) Qualifying-adnominal-attributive SSyntRel

Sem ‘X←1–Y←1–Z’

DSynt L(‘X’)(N)–ATTR→L(‘Y’)(N)–ATTR→L(‘Z’)(ADJ)

SSynt L(‘X’)(N)–qual-adnom→L(‘Y’)(N)–modif→L(‘Z’)(ADJ)

Example ‘dom‘X’←1–krasivyj‘Y’←1–očenʹ-očenʹ‘Z’’ [= ‘an extraordinarily beautiful house’]

DOM–ATTR→krasota–ATTR→Magn [Magn ⇒ neobyčajnyj ‘extraordinary’]
‘house of extraordinary beauty’
DOM–qual-adnom→krasota–modificative→neobyčajnyj:
dom neobyčajnoj krasoty / neobyčajnoj krasoty dom

Only the case where N implements SemA 1 of NGEN is presented. Similarly:

(21) a. ‘čelovek←1–duša←1–kristalʹnejšij’ ⇔
‘human.being←1–soul←1–crystal.purest’
čelovek–ATTR→duša–ATTR→kristalʹnejšij ⇔
čelovek–attr-adnom→duša–modif→kristalʹnejšij ⇔
Kristalʹnejšej duši čelovek! ‘Of.crystal.purest soul human.being!’
[V. I. Lenin about his wife, N. K. Krupskaja, in a risqué political joke]

4) Possessive-adnominal-attributive SSyntRel

Sem ‘X←1–prinadležatʹ–2→Y’ [= ‘X←1–belong–2→Y’]

DSynt L(X)(N)–ATTR→«prinadležatʹ»–II→L(Y)(N)
[«prinadležatʹ» ‘belong’ is a fictitious lexeme marking the possessive relationship]

SSynt L(X)(N)–possessive-adnom→L(Y)(N)

Example ‘dom‘X’←1–prinadležatʹ–2→ministr‘Y’’ [= ‘house that belongs to the minister’]

dom–ATTR→« prinadležatʹ»–II→ministr ‘house of.minister’
dom–possessive-adnom→ministr: dom ministra
228 5 Genitive adnominal dependents in Russian

5) Characterizing-adnominal-attributive SSyntRel

Sem ‘X←1–naxoditʹsja–2→Y’ [= ‘X←1–be.located–2→Y’]

DSynt L(‘X’)(N)–ATTR→«naxoditʹsja»–II→L(‘Y’)(N)
[«naxoditʹsja» ‘be.located’ is a fictitious lexeme marking the localization relationship]

SSynt L(‘X’)(N)–charact-adnom→L(‘Y’)(N)

Example ‘mosty‘X’←1–naxoditʹsja–2→Pariž‘Y’’ [= ‘bridges that are in Paris’]

mostPL–ATTR→«naxoditʹsja»–II→pariž ‘bridges of.Paris’
mostPL–charact-adnom→pariž: mosty Pariža

Only the case where ‘σ’ = ‘be.located’ is presented. Similarly:

(22) a. ‘putešestvija←1–proisxoditʹ←1–vremja–2→vek←1–vosemnadcatyj’  ⇔
‘travels←1–happen←1–time–2→century←1–eighteenth’
putešestviePL–ATTR→«proisxoditʹ»–II→vek–ATTR→vosemnadcatyj⇔
putešestviePL–charact-adnom→vek–modif→vosemnadcatyj  ⇔
putešestvija vosemnadcatogo veka ‘travels of the 18th century’
b. ‘opyt←1–[priobretënnyj]–v_tečenie–2→nedelja←1–ètot’  ⇔
‘experience←1–[acquired]–during–2→week←1–this’
opyt–ATTR→«v_tečenie»–II→nedelja–ATTR→ètot  ⇔
opyt–charact-adnom→nedelja–modificative→ètot  ⇔
opyt ètoj nedeli ‘experience of this week’

Special Cases of NGEN-attr

The semantic-syntactic mismatch caused by the adjective LJUBIMYJ ‘favorite’ (Section 5.3.5)

Sem ‘X←1–ljubimyj–2→Y’ [= ‘X←1–favorite–2→Y’]

DSynt ljubimyj←ATTR–L(‘X’)(N)–ATTR→L(‘Y’)(N)

SSynt ljubimyj←modificative–L(‘X’)(N)–charact-adnom→L(‘Y’)(N)

Example ‘filʹm←1–ljubimyj–2→Petja’ [= ‘film that is favorite of Petya’]

ljubimyj ←ATTR–filʹm–ATTR→Petja ‘favorite film of.Petya’
ljubimyj←modificative–filʹm–charact-adnom→Petja: ljubimyj filʹm Peti
 5.4 Overview of the six SSyntRels proposed 229

NGEN-attr in a non-standard collocation

Sem ‘X←1–Y’
The meaning ‘Y’ corresponds to a non-standard collocational LF Ψ in the lexical
entry for L(‘X’)

DSynt L(‘X’)(N)–ATTR→L(phras, synt1, ...)

[The DSynt-structure contains the lexeme L, which is the value of the non-standard
LF Ψ(L(‘X’)); L is taken, together with additional syntactic features, from the lexical
entry for L(‘X’).]

SSynt L(‘X’)–r→L(phras, synt1, ...)

[The SSyntRel r is also taken from the lexical entry for L(‘X’) — together with L.]

‘seržant←1–služitʹ–2→gvardija’ [= ‘sergeant that serves in the Guards’]

Example seržantL(‘X’)–ATTR→gvardija(phras, antepos)SG-L(‘Y’) ‘sergeant of.Guards’
seržant–charact-adnom→gvardija(phras, antepos)SG: gvardii seržant ‘of.Guards sergeant’

NGEN-attr in a termeme

Sem
‘X←1–Y’ | Tlexical
he meaning ‘Y’ corresponds to a non-standard termemic LF Ψ in the
entry for L(‘X’)

DSynt L(X)(N)–ATTR→«termin»–II→L
[The lexeme L, which is the value of Ψ(L(‘X’)) is taken, together with additional
syntactic features, from the lexical entry for L(‘X’); the fictitious lexeme «termin»
means ‘term’.]

SSynt L(X)(N)–r→L
[The SSyntRel r is also taken from the lexical entry for L(‘X’)(N)—together with L.]

Example ‘boleznʹ←1–affecting the brain of older people…’ [= ‘Alzheimer’s disease’]

boleznʹL(‘X’)–ATTR→«termin»–II→AlʹcgejmerL ‘disease of.Alzheimer’
boleznʹ–charact-adnom→Alʹcgejmer: boleznʹ Alʹcgejmera

NGEN-attr in an idiom

Sem ‘X’

DSynt ˹L1_L2_...˺ [one node]

SSynt L1–r→L2(phraseological)
[The SSyntRel r is specified in the lexical entry for the idiom ˹L1_L2_...˺ — in its SSynt-tree.]

Example ‘ničtožnaja ličnostʹ’ [= ‘totally unimportant person’, ‘a nobody’]

˹otstavnojL3 kozyL2 barabanščikL1˺ [one node] ‘retired goat’s drummer’
barabanščik–charact-adnominal→koza(phras, antepos)SG: otstavnoj kozy barabanščik
230 5 Genitive adnominal dependents in Russian

NGEN-attr in a nomineme

Sem ‘X’

DSynt L1_L2_... [one node]

SSynt L1–r→L2(phraseological)

Example ‘Zemlja_Franca-Iosifa’ [a polar archi­pelago] [= ‘Franz-Joseph Land’]

Zemlja_Franc-Iosif [one node] ‘Land of.Franz-Joseph’
Zemlja SGL1–charact-adnominal→FrancL2–name-junctive-3→Iosif: Zemlja Franca-Iosifa

6) Metaphorical-adnominal SSyntRel

Sem ‘X←1–poxožij–2→Y’ [= ‘X←1–similar–2→Y’]

DSynt L(‘Y’)(N)–ATTR→«predstavljatʹ»–II→L(‘X’)(N)
[«predstavljatʹ» ‘represent’ (a fictitious lexeme) marks a metaphoric relationship.]

SSynt L(‘Y’)(N)–metaph-adnom→L(‘X’)(N)

Example ‘zvëzdy←1–poxožij–2→iskry’ [= ‘stars similar to sparks’]

iskraPL–ATTR→« predstavljatʹ»–II→zvezdaPL ‘sparks of.stars’
iskraPL–metaph-adnom→zvezdaPL: iskry zvëzd

5.5 Closing remarks: NGEN cannot be a personal pronoun

The SSyntRels proposed above for the description of the Russian N→NGEN phrases
have the following syntactic property: their dependent, i. e. NGEN, cannot be a
nominal personal pronoun (ja ‘I’, ty ‘youSG’, on ‘he’, …). For the qualif-adnom
SSyntRel this is obvious, since the NGEN-qual must have a dependent adjective, and
this is impossible for a personal pronoun. The other five adnominal SSyntRels
could in principle allow for NGEN being a pronoun, but they do not (with one
exception, to be mentioned right away):

(23) son–subj-adnom→Ivana ‘sleep of.Ivan’ vs.

*son–subj-adnom→menja ‘sleep of.me’
portret–obj-adnom→Ivana ‘portrait of.Ivan’ vs.
*portret–obj-adnom→menja ‘portrait of.me’
dom–poss-adnom→Ivana ‘house of.Ivan’ vs.
*dom–poss-adnom→menja ‘house of.me’
mosty–charact-adnom→Pariža ‘bridges of.Paris’ vs.
*mosty–charact-adnom→menja ‘bridges of.me’
[Paris is speaking, e.g. in a fantastic tale]
 5.6 Conclusions 231

iskry–metaph-adnom →zvëzd ‘sparks of.stars’ vs.

*iskry–metaph-adnom→nas ‘sparks of.us’
[stars are speaking, e.g. in a fantastic tale]

At the same time, the replacement of NGEN by a pronominal possessive adjective

remains possible: moj son ‘my sleep’, moj portret ‘my portrait’, moj dom ‘my
house’, moi mosty ‘my bridges’, naši iskry ‘our sparks’.9 Therefore, the impossi-
bilities in (23) have to be blocked by the following general rule of Russian:
A nominal personal pronoun in the genitive case cannot syntactically
depend on a noun; the corresponding adjectival possessive pronoun must
be used instead.

There is, however, an interesting exception: with some governors an obj-adnom

NGEN can be pronominalized by a nominal personal pronoun in the genitive, cf.:

(24) vključenie–obj-adnom→menja v sostav komiteta

‘inclusion of.me in body of.committee’
presledovanie–obj-adnom→menja policiej ‘persecution of.me by.police’
otpravka–obj-adnom→menja obratno ‘sending of.me back’

These expressions are highly constrained—both semantically (process-denoting

nouns accept personal pronouns in the genitive more easily) and/or lexically (the
capacity of having a genitive actant pronominalized has to be specified in the Gov-
ernment Pattern of the corresponding nouns.10 (See relevant remarks in A ­ presjan
2010: 12–14.)

5.6 Conclusions
Three important conclusions can be drawn from the above discussion.

1. The six adnominal SSyntRels proposed for the SSynt-description of the

Russian N→NGEN phrases are necessary (barring my possible mistakes), but
not sufficient for this task—not be­cause more SSyntRels are needed, but
simply because establishing the necessary SSyntRels for a particular type of

9 The pronominal possessive adjectives are used here in the 1st person because in the 3rd person
the forms of the nominal personal pronoun in the genitive and those of the corresponding posses-
sive pronominal adjective are homophonous: ego ‘of.him/ his’, eë ‘of.her/her’, IX ‘of.them/their’.
10 Once again, we see the special nature of direct objects, which was mentioned in Subsection
5.3.1/2, p. 211.
232 5 Genitive adnominal dependents in Russian

phrase is but a very first step. To ensure a proper treatment of Russian NGENs,
and in the first place, their correct linear ordering (with respect to other NGENs
as well as to different dependents of the modified noun) we need a set of syn-
tactic features for nouns that allow/disallow their appearance in particular
construction of the N→NGEN type. As the next step, the following three sets of
rules must be elaborated:
– The SSynt-rules for the N→NGEN phrases; these rules stipulate how the
actual phrases (strict­ly speaking, their deep-morphological representa-
tions) are obtained from their SSynt-representa­tions and positioned with
respect to their governor and other cosubordinated NGEN phrases. These
rules need a thorough description of linear ordering of Russian cosubordi-
nated NGENs; such a description is presented in Chapter 11, pp. 369ff.
– The DSynt-rules for the N→NGEN phrases; these rules stipulate how their
SSynt-representa­tions are obtained from their DSynt-representations.
– The Sem-rules for the N→NGEN phrases; these rules stipulate how their
DSynt-representa­tions are obtained from their Sem-representations.
2. Linear ordering of cosubordinated NGEN phrases must be studied within a
much broader frame of mutual ordering of all types of cosubordinated modi-
fiers, in the first place—cosubord­inated adjectives. Various semantic, ref-
erential, communicative, and phonological factors play a role and must be
taken into account.
3. Since this chapter aims at a linguistically and typologically valid justification
for the SSyntRels proposed, it is necessary to widen its linguistic base—that
is, to compare our solution to adnominal dependents in other languages.

11

Previously published as Mel’čuk 2018a.

Part III: H
 ard Nuts in Syntax –
Cracked by Dependency Description
6 Relative clause: a typology
6.1 The problem stated
6.2 What is a relative clause
6.2.1 Some basic terms
6.2.2 The notion of relative clause
6.2.3 Restrictive vs. descriptive relative clauses
6.3 What is not a relative clause: constructions often confused with relative clauses
6.3.1 Constructions isofunctional with relative clauses: non-clausal modifiers
6.3.1.1 Modifying construction headed by a verb adjectivalization
[= participial phrase]
6.3.1.2 Modifying construction headed by a verb nominalization
[= appositional phrase]
6.3.1.3 Clause introduced by an expression meaning ‘such that’
6.3.2 Constructions isostructural with relative clauses: clausal non-modifiers
6.3.2.1 Introductory remarks
6.3.2.2 Pseudo-relative clauses
6.3.2.3 Indirect-interrogative clauses
6.3.2.4 Cleft clauses
6.3.2.5 Nominalized clauses [= “internally-headed relative clauses”]
6.4 Typology of the restrictive relative clause
6.4.1 Relative clauses at the semantic, deep-syntactic and surface-syntactic levels
6.4.2 Deep- and surface-syntactic relations subordinating a relative clause
6.4.3 Syntactic constraints on relative clauses
6.4.4 Syntactic parameters characterizing a relative clause construction
6.4.5 Illustrations of different types of RC constructions (L→RC)
6.4.5.1 The RC’s governor is a non-pronominal noun (L(N, non-pron)→RC)
6.4.5.2 The RC’s governor is a nominal pronoun (L(N, pron)→RC)
6.4.6 Deep-morphological parameters characterizing the implementation of an RC
6.4.6.1 Word order parameters characterizing an RC
6.4.6.2 Inflectional parameters characterizing an RC

The present chapter has been conceived under the impact of the trail-blazing paper Z ­ aliznjak &
Padučeva 1975: I wanted to take the next step and bring the typology of relative clause to the
modern level, 45 years later. However, many obstacles stood in my way—most than anything,
the lack of many necessary notions and formalisms. As a result, my journey took years… Finally,
in 2017 I was close to completion and looking forward to submitting my modest contribution to
the evaluation and criticism of the two close friends, Andrej Zaliznjak and Elena Padučeva, who
had introduced me to the domain. And then the tragedy struck: on 24th December 2017 Andrej
Zaliznjak passed away, and on July 16th 2019 Elena Padučeva left us.

I dedicate this text to the loving memory of these two extraordinary human beings.

https://doi.org/10.1515/9783110694765-007
236 6 Relative clause: a typology

Confusion is worse than error.

Francis Bacon (1561–1626)

6.1 The problem stated

The relative clause [RC] is a hot topic in today’s linguistics: Google offers hun-
dreds of thousands of titles dealing with the RC, one way or another. It is abso-
lutely out of the question to offer even a cursory overview of the relevant litera-
ture; I will limit myself to indicating just some major publications from which I
took the data and some ideas for my own study: Benveniste 1957–58, Peranteau et
al. (eds.) 1972, Zaliznjak & Padučeva 1975[2002], Keenan & Comrie 1977, Comrie &
Keenan 1979, Lehmann 1984, Keenan 1985, Givón 1990: 645–698, Alexiadou et al.
(eds.) 2000, Andrews 2007, Comrie & Estrada-Fernández (eds.) 2012, Xolodilova
2014, Matsumoto et al. (eds.) 2017; to this one has to add Dixon 2010: 313–369,
a detailed typological study of relative clauses and similar constructions in the
world’s languages. In this connection, I can only repeat the famous statement
by Newton: “If I have seen further, it is by standing on the shoulders of Giants.”
The RC fully deserves this heightened attention: it is a very special type of
subordinate clause. It is unique in that, as will be shown (6.4.4, p. 263), in the
world’s languages, RCs come in 36 syntactic types, each featuring a different
syntactic structure and allowing—in principle—15 deep-morphological (= linear/
inflectional) realizations. And all this without taking in considera­tion innumer-
able minor particularities found in RCs of various languages.
The task of this chapter is twofold:

1. To propose a rigorous definition of the RC.

2. To propose a rigorous typology of the RC—that is, a deductively organized
calculus of its logically possible major types.

At the same time, the corresponding terminology must be stabilized and refined,
where needed. However, no new linguistic facts are introduced: my account is
100% based on data already available.
NB I take linguistic facts and their analyses as they are presented in the original source; the
relatively full bibliography on the RC is found in De Vries 2018.

In the present context, what matters most is the logic of the discussion. This
chapter is, in the first place, an exercise in elaborating and perfecting the concep-
tual apparatus of linguistics.
 6.2 What is a relative clause 237

6.2 What is a relative clause

6.2.1 Some basic terms

The first step must be sketching out, if only in broad lines, what exactly will be
defined. (On principles for formulating linguistic definitions, see Chapter 8, Sub-
section 8.2.1, pp. 291ff.)

The relative clause is treated here as a lexeme-modifying subordinate clause.

Thus, the RC is considered strictly as a syntactic phenomenon with a well-

defined semantic function, and therefore it receives a combined semantic-
syntactic definition: it is a particular syntactic con­struction implementing a
particular subclass of modifiers.
NB I do not follow the path blazed in Keenan & Comrie 1977: “It is necessary to have a largely
syntax-free way of identifying RCs in an arbitrary language. Our solution to this problem is to use
an essentially semantically based definition of RC” (p. 63; emphasis added—IM).

Definition 6.1 – modifier

L̃ stands for a lexical expression (= a phrase).

L̃ is a modifier of L in utterance U if and only if in the semantic structure of

U, L depends on L̃ semantically (= ‘L’ is a semantic actant of ‘L̃’), and, in the
deep-syntactic structure of U, L̃ directly depends on L syntactically:

L̃ –←sem →
synt– L

This happens if in U the expression ‘L̃’ semantically governs ‘L’, but communica-
tively ‘L̃’ is dominated by ‘L’, which is shown by the underscoring of L: ‘L̃–sem→L’;
see 6.2.2, p. 242, Figure 6.1). As is obvious, communicative dominance plays a crucial
role in specification of syntactic modifiers.
NB Definition 6.1 uses two shorthand formulations:
1. “L depends on L̃ semantically” means “L semantically depends on a lexeme Lʹ in L̃”—that is,
‘L’ is a semantic actant of ‘Lʹ’.
2. “L̃ directly depends on L syntactically” means “L̃’s top node Lʹ—that is, L̃’s head—syntacti-
cally directly depends on L.”

The second step is sharpening the terminology: giving more precise meanings
to the terms clause and relative, while banning the misleading, but frequently
appearing terms head/top node of the relative clause and antecedent of the rela-
tive clause.

Clause. In the current literature on RCs, the noun clause is applied to different
types of phrase, provided the phrase is formed and controlled by a semantically
238 6 Relative clause: a typology

predicative word—a finite or non-finite verb form (reading a fascinating book; …

[agreement] obtained from Mazda after long negotiations), an adjective with its
actants (equal to the previous group) or even a noun with its actants (the march of
military units along Pennsylvania Avenue). In this monograph only the finite-verb
phrases are considered as clauses: the expression *non-finite clause is not allowed.

Relative. This term needs two refinements: what it is applied to and what it
denotes.
• In the literature, the adjective relative is applied to the names of three differ-
ent syntactic units:
– to a subordinate clause that functions as a modifier, marked in (1) as A;
– to the noun phrase formed by a noun and a subordinate clause that modi-
fies it, marked as B;1
– to the whole sentence that contains an RC: *relative sentence, marked as C.2

(1) Intellectuals [who had mastered higher mathematics] wanted to apply

A
their knowledge.
B
C

In this chapter, only the usage A is allowed: the term relative clause will denote
a particular type of subordinate clause in the strict sense, without inclu-
sion of the lexeme L it modifies. For B, following Lehmann (1984 and 1986:
664), a more convenient term will be used: relative construc­tion, that is
L–synt→RC. The usage C is banned altogether.
• In the literature, the term relative clause is often used to denote a clause
having a specific internal structure (in particular, containing a WH-word) no
matter what its external semantic and syntactic functions in the sentence are;
as a result, one can see in linguistic texts such terminolo­gical monstrosities
as *object relative clause (as in [John brought] what Mary had cooked.) or
*independent relative clause (as in Fr. Qui ne dit mot [consent] ‘Who does not
say word agrees’: Sandfeld 1965: 85ff). For me, an RC can only be a modifier; it
would be logical to drop the use of the adjective relative and speak of modifier
(subordinate) clauses—as we do about subject, object, circum­stantial, etc.
subordinate clauses. This is not done here simply in order to avoid a clash
with the universally accepted term of relative clause, extremely frequent and
well rooted in practice.

1 “The term RC is used to apply to the collocation of the head NP and the restricting clause”
(Keenan & Comrie 1977: 64).
2 The search for the phrase “relative sentence” returned over 27,000 hits on Google (2021-02-23).
 6.2 What is a relative clause 239

Head. In the current literature, the name head/top node is often applied to the
lexeme—mostly a noun—modified by an RC; thus, in the book that John is reading
the noun book is called *“the head of the relative clause that John is reading.” This
usage is banned: the syntactic head of a phrase (= its top node) is an element of
the phrase—that is, it is structurally inside the phrase (in the case of an RC, it is its
Main Verb); the external lexical unit that syntactically subordinates the phrase is
its syntactic governor, not its head. Thus, bcd is a phrase whose head is c, while
a is its governor:

a b c d

As a consequence, the term *headless relative clause is absurd: no clause (and no

phrase, for this matter) can be headless.

Antecedent. The lexeme L modified by an RC is often called its antecedent, that

is, ‘the antecedent of the relative clause,’ which is quite infelicitous. This term is
normally used for the noun whose repeated occurrence is replaced by a substitute
(= anaphoric) pronoun, as in (2a), where the noun country is the antecedent of
the pronoun it, or (2b), where the noun bombings is the antecedent of the relative
pronoun that3, but not of the relative clause that kill both police and civilians:

(2) a. The country must behave as it has behaved in the past.

b. They carry out bombings that kill both police and civilians.
NB To avoid confusion, here are different lexemes of the vocable THAT.
THAT1 demonstrative adjectival pronoun (that decision, those decisions)
THAT2 demonstrative nominal (= substitute) pronoun (Those who want
to go must sign up tomorrow.)
THAT3 relative nominal pronoun (a decision that was made in a hurry)
THAT4 degree adverb (not that intelligent)
THAT5  empty subordinating conjunction (He wrote that he would be absent.)

In this chapter, a lexeme modified by an RC is called simply the lexeme modified

by the RC or the RC’s (syntactic) governor.

6.2.2 The notion of relative clause

An RC must be defined by two conditions:

1) Its semantic function, which specifies the closest superset—the set of all
modifiers; it corresponds to genus proximum, or the closest kind, in an Aris-
totelian definition.
240 6 Relative clause: a typology

2) Its syntactic organization, which specifies the necessary subset—that of mod-

ifiers of a particular clausal type; this is Aristotle’s differentiae specificae.

Definition 6.2 – relative clause

C is a full-fledged clause.

A subordinate clause C is called relative iff C is a modifier of an LU L.

C is a subordinate clause, Cʹ being its superordinate (= matrix) clause; L is C’s
syntactic governor.

– If ‘C’ does not semantically bear on ‘L’ (= ‘L’ is not a Sem-actant of ‘C’), then
C is not a rela­tive, but a completive clause (it semantically depends on L and
expresses one of L’s Sem-actants):

(3) a. the caseL [where/when a definition introduces a technical term]C

the factL [that/*which Johnson was able to succeed]C
b. the reasonL [why Johnson was unable to succeed]C
the requirementL [that airline advertising be more transparent]C
the fearL [that his life was in danger]C

In (3), C is a completive subordinate clause that expresses a Sem-actant of L and is

L’s DSynt-actant [DSyntA] rather than its modifier: in (3a), C is L’s DSynt-actant I,
and in (3b), L’s DSynt-actant II.3 (Cf. the case of a definition introducing a technical
term and the fact of Johnson being able to succeed, where the actants of case and
fact are expressed by of-phrases.)

– If, however, ‘C’ semantically bears on ‘L’, but C does not syntactically depend
on L, then C is not a re­lative, but an direct-object clause or an object-attribu-
tive-completive clause (see Chapter 2, Section 2.5, Nos. 7 and 24):

dir-obj
(4) a. John says→about lifeL [that it is not easy]C.
NB (4a) is perceived by some speakers as not natural, even if grammatical; but its
Russian equivalent, (4b), is perfectly OK:
dir-obj
b. Džon govorit→pro žizn′L[, čto ona nelegka]C.

3 A completive clause of this type can be synonymous with an RC, which we see in (i) and (ii):
(i) the case [in which a definition introduces a technical term] or
(ii) the requirement [according to which airline advertisement must be more transparent]
This should not, of course, prevent us from formally distinguishing them.
 6.2 What is a relative clause 241

obj-attr-compl
c. Fr. Je leL←vois [qui←traverse la rue]C
lit. ‘I him see who is.crossing the street’.

In (4), the clause C is not a modifier. In (4a–b), it is a direct-object clause that

semantically and syntactically depends on the verb say/govorit′—it expresses its
Sem-actant 2; in (4c), ‘C’ semantically bears on the source of the pronoun LE ‘him’
(in a way quite similar to Je le vois endormi ‘I him see asleep’ or Je le vois traverser
la rue ‘I him see cross the street’), while syntactically, C depends on the verb voir
‘see’ and is an object-attributive clause (see Sandfeld 1965: 139–159 and van der
Auwera 1985).
NB Sentences such as (4c) are typical of Romance languages; see а detailed review in Graffi
2017. The subordinate clause such sentences contain is commonly called pseudo-relative, a ter-
minological usage I cannot accept: I reserve the term pseudo-relative for a completely different
type of clause: see Subsection 6.3.2.2, pp. 249ff.
Note that the governor of an RC is not necessarily a noun; it can be a lexical
unit of any part of speech: see below, Subsection 6.2.3, (8), p. 244.
I do not know of an expression that intuitively is not an RC, but would be
accepted by Definition 6.2.
The diagrams in Figure 6.1 below present the schematic structures of a pro-
totypical relative construction, that is, an RC together with its syntactic governor,
as seen in (5):

(5) Husby put on the book a bindingL-RC’s governor [ that was designed to be easily
taken apart]RC.

These structures are given at three levels of representation: semantic [Sem-],

deep-syntactic [DSynt-], and the surface-syntactic [SSynt-]; they are abstract in
that they are formulated regard­less of any specific language.

3 The general schema of the RC construction’s syntactic structures foresees the

image of the governor (boxed in cells b, c-i and c-ii of Figure 6.1), while it is
known that in many lan­guages the governor’s image cannot physically appear
in the RC at the morphological surface (see 6.4.4, Item 4, p. 263). However, at
a syntactic level, the presence of the govern­or’s image in the RC’s structure
is necessary: it is its surface-syntactic role that decides whether, in the given
language, the relativization of the given phrase is possible or not. Thus, in a
language that only allows for relativization of the SSynt-subject (like Mala­
gasy), the configuration –i→ L(‘X’) , where i ≠ subjectival, in the SSyntS of the
sentence to be synthesized blocks the production of an RC. This is one of filter
rules mentioned in 6.4.3 below, p. 258.
242 6 Relative clause: a typology

At the SSynt-level, the governor’s image undergoes obligatory pronominalization

in a broad sense: 1) either it remains as is—as a duplicate of the governor (zero
pronominalization); 2) or it is elided (= deleted, or gapped); 3) or else it is replaced
by a pronoun—personal or relative.

a. b. c.-i c.-ii
L(‘X’)(N, non pron) L(‘X’)(N, non pron) L(‘X’)(N, NON PRON)

ATTRrestr restrictive-relative
restrictive-relative
‘Y’ L(CONJ, subord)
(

(
L(V)FIN L(V)FIN
ρ subord-conjunctional
(

(
(
L(‘Y’) L(V)FIN L(‘Y’)
‘X’ R(ρ) r(R)

( L(‘Y’)
L(‘X’) r(R) PRON(L(‘X’))

PRON(L(‘X’))

Sem-structure DSynt-structure SSynt-structure

1. Underscoring of 1. L(‘Ψ’) is a lexeme ex- 1. r(R) means “SSyntRel r corresponding to DSyntRel R.”
a node shows its pressing the meaning 2. PRON(Ψ) denotes the result of appropriate pronomi-
communicative ‘Ψ’. nalization of a lexical expression Ψ—including ellipsis
dominance. 2. R(ρ) means “DSyntRel (= gapping).
2. ρ stands for a R corresponding to
Sem-relation SemRel ρ.”
(ρ = 1, 2, …)
3. «non pron(ominal)» is the negation of a syntactic feature that characterizes nominal
L(CONJ, subord)

subord-conjunctional

pronouns (it, he, …, somebody, etc.); in this case we deal with a genuine noun.
L(‘X’)(N, non pron)

restrictive-relative

4. The bidirectional dashed arrow represents coreferentiality.

L(‘Y’)
L(V)FIN

PRON(L(‘X’))

5. The horizontal parentheses enclose an optional fragment of the syntactic tree:

an arbitrary chain of syntactic dependencies: . ( (
r(R)

This fragment is taken into account only if L(‘Y’) is not a finite verb—that is, not the
Main Verb of the relative clause; for instance:

the houseL(‘X’) whose L(‘X’)←ownerL(‘Y’)←livesL(V)FIN→in→France

Otherwise, the fragment in parentheses is ignored.
6. Boxing indicates the governor’s image.

Figure 6.1 Schematic structures of a prototypical relative clause

The diagrams in (b–c) show:
– The lexeme L(‘X’) modified by the RC; L(‘X’) is the RC’s governor.
– The lexeme L(‘X’)—a duplicate—within the RC which is coreferential with the gover-
nor; this is the image of the RC’s governor.
– The chain of syntactic links between the two occurrences of L(‘X’).
The diagram in (c-i) represents an RC with an explicit subordinator—L(CONJ, subord), while
that in (c-ii) an RC without an explicit subordinator.
 6.2 What is a relative clause 243

(6) Illustrations of Figure 6.1

a. An RC with a subordinator
Hebrew

ha+yeledL [še ani←dibarti→ alav [⇐ al+hu]]

the boy that I talked about.him
‘the boy about whom I talked’ [the subordinator is the conjunction še ‘that’]
b. An RC without a subordinator

the sea-level increaseL [to→ which the climate system is→committed]

• At the semantic level, the L→RC construction has the same structure as any
modificative construction of any language (including the adjective←noun con-
struction). More specifically, it has a communicatively dominating meaning
‘X’L semantically dominated by meaning ‘Y’C: ‘Y–sem→X’. This schema is a
linguistically universal semantic description of any modifier Y of X.
NB Our semantic characterization of the RC does not include the requirement of narrowing
the reference of its governor; this property distinguishes two subclasses of RCs—restrictive
vs. descriptive RCs, see the next subsection.

• At the deep-syntactic level, an RC construction has quite a specific structure:

all and only RC constructions have this form, which is valid cross-linguisti-
cally. To put it differently, in any language, any RC construction has this DSynt-
form, and any construction having this DSynt-form is an RC construction. The
DSynt-form of an RC construction is specified by the following two properties:

– The dependency of the DSynt-head of the RC on L (= its governor) is direct.

– The RC contains an image L of its governor L, this image being anaphori-
cally related to L (Padučeva 1974: 130–133).

The DSynt-relation [DSyntRel] subordinating the RC—more precisely, the RC’s

head—to its governor L is ATTRrestr (i.e., restrictive-attributive; a descriptive RC is
subordinated to its governor by ATTRdescr, see 6.2.3).

• At the surface-syntactic level, an RC construction has one of two possible

structures, de­pending on the language—with or without an explicit subordi-
nator, that is, a marker of the sub­ordinate character of the clause. Schema
(c-i) in Figure 6.1 covers Persian-type RCs, see (18), p. 256; schema (c-ii)
describes more familiar RCs of European types.

•  he behavior of an RC at the deep-morphological level of sentence repre-

T
sentation is irre­levant for its definition: the linear ordering of the elements
244 6 Relative clause: a typology

of the RC and of the RC itself with re­spect to its governor in the superordi-
nate clause, as well as concomitant morphological markings, characterize
the surface implementation of an RC in a particular language, rather than
the RC itself.

3 The schemata in Figure 6.1 do not cover all possible types of RC: they describe
only the RC whose governor is a genuine noun, while there are RCs with pro-
nominal governors, see below, 6.4.5.2.

6.2.3 Restrictive vs. descriptive relative clauses

As is well known, RCs—like all syntactic modifiers of nouns or pronouns—fall into

two major classes according to their impact on the reference of the lexeme modified:

(7) a. He approved a lawL [ that strips parliamentarians of their legal

immunity].
b. He approved this lawL, [ which 〈?that〉 strips parliamentarians of their
legal immunity].

– In (7a), a restrictive [= defining] RC “enriches” the meaning of L—that is, more

precisely, it narrows its reference: the sentence talks about a particular law
such that it strips parliamentarians of immunity; this RC ensures the identi-
fication of the law for the Addressee.
– In (7b), a descriptive [= qualifying] RC does not impact the reference of L, but
adds some additional information about the law that is already known to the
Addressee (“this law”).

Restrictive and descriptive RCs show many differences, the best known being that
in English the descriptive RC requires the relative pronoun which rather than
that3 (see Carlson 1977 and Platzak 2000, where other differences are discussed;
however, according to D. Beck, the prohibi­tion against that3 in descriptive RCs is
dying out in contemporary English).
NB Not all languages make a distinction between restrictive and descriptive RCs. Thus, it does
not exist in Japanese (Kuno 1973: 235).

An RC modifying a verb or a non-pronominal adjective can be only descrip-

tive, cf. (8):

(8) But Johnson objectedL, [which amazed everybody]C.

Johnson asked to be releasedL, [to which everyone agreed]C.
Johnson called our friend Peter lazyL, [which he is not]C.
 6.3 What is not a relative clause 245

(This type of RC is what is called in Quirk et al. 1991: 1118–1120 sentential relative
clause, opposed to phrasal relative clause.)
In this chapter, descriptive RCs are left out of consideration.4

6.3 W
 hat is not a relative clause: constructions often confused
with relative clauses
The description of RCs in modern linguistics is often impeded by the confusion
between RCs and other complex constructions that are either isofunctional with
RCs, playing the same semant­ic and syntactic role as an RC (being semantic and
syntactic modifiers; 6.3.1), or isostructural with RCs, having the same—or, at
least, similar—syntactic structure (among other things, containing relative or
similar pronouns; 6.3.2). To ensure a rigorous treatment of RCs, it is necessary to
specify the syntactic formations that are not RCs, but tend to be perceived as RCs.

6.3.1 C
 onstructions isofunctional with relative clauses: non-clausal
modifiers

According to the part of speech of the syntactic head (= top node) of a noun-modify-
ing con­struction, three types of this construction are possible: 1) finite-verb headed
constructions—that is, our RCs; 2) adjectivalized-verb headed constructions; and
3) nominalized-verb headed con­structions. The constructions of types 2 and 3 can
fulfill the same semantic and syntactic roles as an RC: they can modify a lexical
unit both semantically and syntactically; however, they are not relative clauses—
because they are not clauses in the strict sense of the term (see 6.3.1.1 and 6.3.1.2).

6.3.1.1 M
 odifying construction headed by a verb adjectivalization
[= participial phrase]
English stock examples of noun-modifying non-clausal constructions are [pas-
sengers] wanting to go to Glasgow and [the General Assembly] immobilized in those
troubled days.

4 On the descriptive relatives that depend on a lexeme different from a noun, see Chapter 2,
No. 74, p. 90. — Some researchers propose to distinguish another class of RC: “relative clauses
of the third kind” (Carlson 1977 and Grosu & Landman 1998), of the type of I took the three books
(that) there were on the table.
246 6 Relative clause: a typology

Since “non-finite clauses” are not clauses, noun-modifying constructions

headed by an adjec­tival form of a verb—that is, by a participle—are not RCs.5

(9) a. Russian
čitajušč+ego knig+u mal′čik+a ~
reading MASC.SG.GEN book SG.ACC boy(masc) SG.GEN
‘[of] reading book boy’
mal′čik+a, čitajušč+ego knig+u
‘of [a] boy reading book’

udivlënn+ye podark+om devočk+i ~

amazed PL.NOM gift SG.INSTR girl(fem) PL.NOM
‘amazed by.gift girls’
devočk+i, udivlënn+ye podark+om
‘girls amazed by.gift’

b. Kolyma Yukaghir (Maslova 2003: 418)

purk+in šoromo lē+je šoromo
seven ATTR person eat ATTR person
‘seven people having.eaten person’ = ‘person who ate seven people’

c. Turkish (ğ is not pronounced; ş = /š/)

Ada +Ø+da gör+düğ +üm kişi +Ø+Ø
island SG LOC see PAST_PART 1.SG person SG NOM
lit. ‘on.island my.having.seen person’ = ‘person whom I saw on the island’

The bold-faced expressions in (9) are typical participial phrases, not clauses.6

5 “English also has nonfinite participial relative clauses, as in the man reading the book. Nonfi-
nite relative clauses are sometimes not considered as relative clauses; however, since there are
many languages where relative clauses are all nonfinite and since these constructions mean the
same thing as finite relative clauses in English, such participial constructions are considered as
relative clauses” (Dryer 2013, after (8)). Curiously, it is exactly for this reason that I do not want to
consider participial constructions to be RCs.
6 Participial phrases have an interesting variety, infelicitously called “Possessive Relative” (Ack-
ermann & Nikolaeva 2013):
(i) Western Khanty
xans+ǝm nēpǝk+ēm lit. ‘written my.book’ = ‘book written by me’
write PARTICIPLE book 1SG
In such a construction, the Agent of the participle is marked not on the participle (as, for in-
stance, in Turkish: (9c)), but on the modified noun.
 6.3 What is not a relative clause 247

6.3.1.2 M
 odifying construction headed by a verb nominalization [=
appositional phrase]
Actual English examples are unavailable, since this construction does not exist in
that language; artificial examples could be as follows:
passengers to-go-to-Liverpool wanters
‘passengers wanting to go to Liverpool’
boy my yesterday meetee
‘boy whom I met yesterday’
NB In spite of the fact that the Patient noun suffix -ee is not fully productive in English, it
will be used in the glosses here and further, so that meetee means ‘that whom [somebody]
meets/met’.

(10) a. Seri (Marlett 2012; c, qu = /k/, cö = /kw/, j = /χ/, x = /x/)

hapxa cö +c +aasitim quij ‘rabbit his/her.deceiver the’ =
rabbit 3.SG SUBJ.NOMINAL deceive the ‘the rabbit who deceived him/her’
hapxa h +oco +aasitim quij ‘rabbit my.deceivee the’ =
rabbit 1.SG OBJ.NOMINAL deceive the ‘the rabbit whom I deceived’

As can be seen from (10a), Seri has two prefixal nominalizers, the subjectival c- [≈
Eng. -er] and the objectival oco- [≈ Eng. -ee] (underscored); they correspond to
the meanings of the lexical functions S1 (≈ Agent noun) and S2 (≈ Patient noun)
and turn the verb into an agent or patient noun.

b. Yaqui (Álvarez González 2012; ch = /č/, j = /χ/)

Joan uka chu’u+ta Maria+ta ke’e+ka +m +ta
Juan the-ACC dog ACC Maria ACC bite PERF SUBJ.NOMINALIZER ACC
me’a+k
kill PERF
‘Juan the dog Maria biterN killed’. = ‘Juan killed the dog that bit Maria’.
Joan uka bachi+ta em jinu+ka +’u +ta
Juan the-ACC corn ACC youSG-GEN buy PERF OBJ.NOMINALIZER ACC
bwa’a+k
eat PERF
lit. ‘Juan the corn your buyN ate’. = ‘Juan ate the corn that you bought’.
Joan uka kari +ta em tomte +ka +’apo
Juan the-ACC house ACC youSG-GEN be_borne PERF LOC.NOMINALIZER
jinu+k
buy PERF
lit. ‘Juan the house your birthplaceN bought’. =
‘Juan bought the house where you were born’.
248 6 Relative clause: a typology

Like Seri, Yaqui also has clausal nominalizers, but they differ from those of
Seri in that, first, they are suffixal and, second, there are three of them. (Several
interesting details are omitted: thus, the Seri objectival nominalizer oco- is used
only for a DirO, while Yaqui objectival nominalizer -’u can be used for all types of
object—direct, indirect and oblique; etc.)
The picture presented for Seri and Yaqui is typical of many other American
Indian languages: they do not have an RC, but use a nominal apposition in the
same semantic role—as a modifier.

6.3.1.3 Clause introduced by an expression meaning ‘such that’

Subordinate clauses introduced by an adjective with the meaning of ‘such ... that …’
are semantically very close to RCs:

(11) a. … and with brutalityL such that the rest of the world cannot stand idly by.
b. It was such a cold afternoonL that we stopped playing.
c. It is a differential operatorL such that its restriction on T vanishes.

A such.that-clause bears semantically on the governor (= modified) noun L,

just like an RC, and sometimes it alternates freely with an RC; thus, (11c) is syn-
onymous with a sentence containing an RC: It is a differential operatorL [whose
restriction on T vanishes]. However, in a such.that-clause, the that-introduced
clause C syntactically depends on such and is its DSyntA II: L–ATTR→such–
II→C. Therefore, Definition 6.2 is violated (C does not directly depend on L), and
therefore C is not an RC. Moreover, in such.that-clauses we have the subord­inate
conjunction that5 rather than a relative pronoun that3 (see 6.2.1 above, p. 239).
SUCH.THAT-clauses should not be considered in a discussion of RCs.7

6.3.2 Constructions isostructural with relative clauses:

clausal non-modifiers

6.3.2.1 Introductory remarks

Two major types of isostructurality with RCs can be distinguished: internal and
external isostructurality.

7 In French, a noun-modifying clause introduced by the conjunction comme ‘as’ constitutes a

special case similar to such.that-clauses:
(i) gâteaux–modif→comme [je les aime] ‘cakes as I like them’
The comparative conjunction comme is semantically full and appears in the DSyntS; therefore,
the subordinate clause proper (in this case, je les aime) does not depend on gâteau directly, and
thus Definition 6.2 is violated.
 6.3 What is not a relative clause 249

• Internal isostructurality consists in the presence of a WH-pronoun. In prac-

tically all Euro­pean languages an RC normally includes a WH-pronoun: in
English it is which, who, where, …, and that3 (seen, e.g., in moving screens
that bob up and down); this pronoun constitutes the image of the RC’s gover-
nor. WH-pronouns feature many semantic and syntactic particularities that
have always attracted, and still attract, linguists’ attention; unfortunately,
there exists a clear tendency to identify the set of relative clauses and that of
subordinate clauses including WH-pronouns [WH-clauses]. However, this is
an unwarranted step:

– On the one hand, a large number of the world’s languages do not have
relative pronouns at all, while having RCs—that is, they have RCs without
relative WH-pronouns (we will see this in 6.4.3 below).
– On the other hand, there are various types of WH-clause that are by no
means RCs.

Therefore, we have to accept the following statement:

The set of relative clauses and the set of WH-clauses have a large intersec-
tion—there are RCs that are also WH-clauses (for instance, in familiar Euro-
pean languages), but there are RCs that are not WH-clauses and there are
WH-clauses that are not RCs.
NB At least four subclasses of WH-pronouns have to be distinguished:
– interrogative (What/Who is this?)
– indirect-interrogative (I know who is this boy.)
– relative (The person who brought the letter is here.)
– pseudo-relative (What [= ‘that.thing.which’] he did is a crime.)

It would be unwise to undertake here an overview of possible WH-clauses; I

will limit myself to three most common types: pseudo-relative clauses (6.3.2.2),
indirect-interrogative clauses (6.3.2.3) and cleft clauses (6.3.2.4); cf. also object-
attributive subordinate clauses, illustrated in 6.2.2, (4c), p. 241.

• External isostructurality of a clause with RCs consists in its role in the sen-
tence: semant­ically it is a modifier while being, at the syntactic level, not a
modifier, but an actant of the Main Verb. This is the notorious “internally-
headed RC” (6.3.2.5).

6.3.2.2 Pseudo-relative clauses

In (12), a particular type of subordinate WH-clause is illustrated that is commonly
called free/ headless relative:
250 6 Relative clause: a typology

(12) a. You can buyY [what these people sellZ]C.

b. I likeY [how she dancesZ]C.
c. His decisionY depends on [where Mary isZ now]C.

However, the term free/headless relative cannot be accepted:

– What is called a free/headless relative clause is in no sense relative—syntacti-

cally it does not modify anything, but is an actant or a circumstantial.
– It is not free—syntactically it necessarily depends on something, being a sub-
ordinate clause.
– It is not headless—as all clauses, it necessarily has a syntactic head (its Main
Verb).8
– The lexeme what seen in (12a) is not а relative pronoun: a relative pronoun
is semantically empty (= has no source in the Sem-structure), while this
what has a meaning, although very general—‘that.something’. Moreover,
what cannot be used as a relative pronoun in a genuine RC: *the house what
we bought. Pronouns of this type could be called demonstrative-relative, or
pseudo-relative.

Therefore, the clauses what these people sell, who she loves and where Mary is now
in (12) are not RCs. Since, however, they are superficially similar to RCs, they can
be called pseudo-relative clauses [pseudo-RCs].
NB In linguistic literature, the term pseudo-relative clause is often used in several different
senses. Among other things, object-attributive clauses in Romance languages are regularly called
pseudo-relatives; see Subsection 6.2.2 above, example (4c), p. 241.

A pseudo-RC is syntactically equivalent to a noun phrase, is used as an

SSynt-actant or an SSynt-circumstantial and is subordinated to its governor by
the subjectival, direct-objectival, …, prepositional or circumstantial SSyntRels. While
a genuine RC modifies a noun, a pseudo-RC “incor­porates” the meaning it seems
to modify: in (12a), the pronoun what actually means ‘that.thing.which’.9 We can
indicate this by ascribing to this what the syntactic feature «pseudo-­rel(ative)».
Figure 6.2 (next page) presents the structures of typical pseudo-RCs at the
three levels.
The Sem-, the DSynt- and the SSynt-structures of the sentences of the type
(12b) and (12c) are the same, with ‘the.way.that’ and ‘the.place.where’ instead of
‘something’ and, respectively, how and where instead of what.

8 Already in Kručinina 1968 such clauses were called antecedentless, which is by far more logical.
9 A similar description of this what was already proposed in 1960s (Kuroda 1968: 246). Pseudo-
relatives were treated as noun phrases in Mel’čuk & Pertsov 1987: 486–487.
 6.3 What is not a relative clause 251

A particular type of pseudo-RCs are pseudo-RCs that form compound indefi-

nite pronouns of the type boldfaced in sentences such as I met yesterday you will
never guess whom or John lives where he wants (Chapter 9, Subsection 9.2.1.1.1,
p. 312; see also Chapter 2, Section 2.5, No. 75, p. 91).
Pseudo-RCs can easily be and often are confused with RCs whose governor
is a pronoun.
NB For more on pseudo-RCs, see, among others, Horvath & Grosu 1987 and Caponigro et al. 2013.

a. b. c.
L(‘Y’)(V) L(‘Y’)(V)
‘Y’ ‘Z’
Ri(ρi) r(Ri)
ρi ρj
L(‘Z’)(V)FIN L(‘Z’)(V)FIN
Rj(ρj) r(Rj)
‘something’

WHAT(pseudo-rel) WHAT(pseudo-rel)

Sem-structure DSynt-structure SSynt-structure

aʹ. bʹ. cʹ.

BUY(V) BUY(V)
‘buy’ ‘sell’
II direct-objectival
2 2
SELLFIN SELLFIN
II direct-objectival
‘something’

WHAT(pseudo-rel) WHAT(pseudo--rel)

Figure 6.2 Schematic semantic, deep-syntactic and surface-syntactic structure of sentences

with a prototypical pseudo-relative clause of (12a) type

Here are a couple of slightly more complex pseudo-relatives, where the interroga-
tive-relative pronoun is adjectival.

(13) a. Russian, highly colloquial or archaic

[Kotor+uju knig +u najd +ëšʹ,]
which Fem.SG.aCC book(fem) SG.aCC find-PeRF.FUt 2.SG
252 6 Relative clause: a typology

čitaj+Ø tut že
read ImPeR.2.SG right.away
lit. ‘[Which book you.will.find,] read (it) right away’.

b. Latin (Caesar)
Sabin+us [qu +os tribun +os
Sabin SG.nom which maSC.PL.aCC tribune(masc) PL.aCC
habe+ba +t] se sequi iube +Ø +t
have ImPF 3.SG himself follow-InF order(V) PReS 3.SG
lit. ‘Sabin, [which tribunes he.had,] himself follow [them] orders’. =
‘Sabin orders the tribunes he had to follow him’.

6.3.2.3 Indirect-interrogative clauses

(14) I know [who/where this boy is 〈when this boy arrived〉].

The subordinate clauses in (14) are indirect questions; they do not modify any-
thing and are DSyntAs (in this case, II) of the verb knoW. They contain an inter-
rogative WH-pronoun and thereby differ from pseudo-RCs (with their pseudo-
relative WH-pronoun).
Interestingly, in a particular language, an indirect-interrogative clause (15a)
can contrast semantically with a genuine RC having a pronominal governor, i.e. a
correlative to [a form of tot II.1] as in (15b):

(15) Russian
a. Menja interesuet, [čto Ivan pišet]INDIR-INTERR
lit. ‘I am interested in.what Ivan is writing’. =
‘I don’t know what Ivan is writing, and I am interested to know it’.
b. Menja interesuet to, [ čto Ivan pišet]RC
‘I am.interested.in the.stuff that Ivan is writing’. =
‘I know what Ivan is writing something, and I am interested in it’.

6.3.2.4 Cleft clauses

A cleft sentence contains a subordinate WH-clause:

(16) It is me [who pushed John].

However, this WH-clause does not modify anything: it is a pseudo-subject of the

verb Be, as seen in Figure 6.3. (On pseudo-subjects, see Chapter 2, Section 2.5,
SSyntRel No. 6, p. 51.)
 6.3 What is not a relative clause 253

BEIND, PRES
subjectival pseudo-subjectival
copular-attribu-
tive-objectival
PUSHIND, PAST
6
IT I subjectival
direct-objectival

WHO JOHN
Figure 6.3 The surface-syntactic structure of sentence (16). — it stands for the empty—dummy—it.
6

Therefore, this WH-clause is not an RC: it is a cleft.

6.3.2.5 Nominalized clauses [= “internally-headed relative clauses”]

A special case was already mentioned above: a phrase that is formed by a clause
nominalized as a whole (which is, of course, different from a phrase formed by
a nominalized verb) and is used as a bizarre hybrid: a semantic modifier, but a
syntactic actant.
NB When only the Main Verb of a clause is nominalized, all its syntactic dependents change
their form, adapting it to the nominal governor, as in John reads the book ⇒ John’s reading of the
book; but under the nominalization of the whole clause all dependents of the Main Verb remain
as they are—a nominalizer is added to the clause and nothing else changes:
John reads the book ⇒ [John reads the book]+nominalizer.

Clause nominalizations of this kind are known as internally headed relative

clauses (see Xolodovič 1971, Grosu 2002, 2012 and Hiraiwa et al. 2017, with a vast
bibliography). The most discussed type of this clause nominalization is found in
Japanese and Korean; Lakhota supplies a different variety.

(17) a. Japanese
Keikan +wa dorobō+ga ginkō+kara
policeman theme robber SUBJ bank aBL
deteki +ta +no +o tukamae+ta
come.out PaSt «fact» aCC arrest(V) PaSt
lit. ‘Policeman robber from.bank came.out «fact» arrested’.
= ‘The policeman arrested the robber who came out from the bank’.10

10 For a detailed discussion of Japanese nominalized clauses, see Grosu & Hoshi 2009, where it
is shown, among other things, that a -no-clause can be 4-way ambiguous.
254 6 Relative clause: a typology

b. Korean
Kyoŋčhal +ɨn kaŋto+ka unhaŋy+eye nao +nɨn
policeman theme robber SUBJ bank aBL come.out PaRtICIPLe
kes +ɨl čap +ass +ta
«fact» aCC arrest(V) PaSt DeCLaR(ative)
lit. ‘Policeman robber from.bank coming.out «fact» arrested’. =
‘The policeman arrested the robber who came out from the bank’.
c. Lakhota (Williamson 1987)
Mary owiza wã kaġe ki he ophewalthu
Mary quilt a make the that.one I.buy.it
lit. ‘Mary quilt a make the that.one I bought’.
= ‘I bought the quilt that Mary had made’.
NB Japanese and Korean both have special markers on the lexical item that expresses
the semantic-communicative Theme of the main clause, the suffixes -wa and -ɨn, respec-
tively. subj(ective) stands for the subjective case, which marks the syntactic subject
(and is different from the nominative, the case of nomination, having the -Ø suffix).

All the three languages use overt clause nominalizers (underscored):

– Japanese has a morphological clause nominalizer—the suffix -no. It is added

to the last lexeme of the clause, but since in Japanese the Main Verb is obliga-
torily clause-final, -no attaches to it and turns the clause into a noun phrase
that accepts case suffixes required by its external context. All dependents
of the Main Verb retain their case marking and remain verbal dependents:
Dorobōga ginkōkara detekita lit. ‘Robber from.bank came.out’ is a complete
finite clause that can be used as such.
– Korean recurs to a lexical clause nominalizer—the auxiliary noun keS, which
is used either with a determiner in the sense of ‘thing’, or—as in (17b)—as an
empty nominalizer governing a participial phrase. Both -no and keS can be
approximately glossed as ‘fact’.
– In Lakhota, the clause is nominalized by the definite article kI ‘the’.

In (17a) and (17c), the bold-faced clause is not an RC since it is not a modifier; in
(17b), the bold-faced expression is not even a (finite) clause.
Sentences in (17) illustrate Case 9 in the general calculus of possible com-
binations of semantic, syntactic and morphological dependencies between two
lexemes in an utterance (Mel’čuk 1988: 123–124 and 2012–2015: vol. 3, 457–458,
480): a case of head-switching, or reversal of semantic vs. syntactic dependen-
cies. Syntactically, Japanese and Korean say that the policeman “arrested the
coming_out–synt→[of the robber],” but semantically, this means, of course, that
 6.3 What is not a relative clause 255

he arrested the robber who was coming out of the bank. Similarly, in the Lakhota
example, I bought not the Mary’s making–synt→[of a quilt], but the quilt that was
made by Mary.
The bold-faced phrases in (17) are currently called relative clauses only
because they are used as (approximate) equivalents of European RCs. However,
they have a completely different syntactic structures: in the first place, they do
not syntactically modify a lexical unit L and they do not feature L’s image. They
should not be called relative. A possible term could be nominalized clauses.
NB For a detailed discussion of nominalized clauses, see, for instance, Grosu 2012.
Figure 6.4 below gives the schematic structures of prototypical nominalized
clauses at the three representation levels.

a. b. c.-i (Japanese) c.-ii (Korean) c.-iii (Lakhota)

‘Y’ ‘Z’ L(‘Y’)(V)FIN L(‘Y’)(V)FIN L(‘Y’)(V)FIN L(‘Y’)(V)

ρi ρj Ri(ρi) r(Ri) r(Ri) r(Ri)

L(N, nominalizer)
L(‘Z’)(V)FIN L(‘Z’)(V)FIN, L(‘Z’)(V)FIN
NOMIN(alized auxiliary
‘X’ r(Rj) clause)
Rj(ρj) L(‘Z’)(V)PART r(Rj) determinative

L(‘X’)(N) r(Rj)
L(‘X’)(N) L(‘X’)(N) KI(DET)
L(‘X’)(N)

Sem-Structure DSynt-Structure SSynt-Structure

aʹ. bʹ. cʹ.-i (Japanese) cʹ.-ii (Korean) cʹ.-iii (Lakhota)

‘arrest’ ‘come_ “ARREST(V)” TUKUMAE ‘arrest(V)’ ČAP ‘arrest(V)’ OPHETHUNG ‘buy’

out’
II direct-objectival direct-objectival
objectival direct-objectival
2 1
KES(nominalizer)
“COME_ OUT” DETEKI ≈ ‘fact’ KAĠE ‘make’
‘come_out’ auxiliary
‘robber’ I direct- determinative
subjectival NAO+NƗN objectival
‘coming_out’
“ROBBER”
DOROBŌ ‘robber’ subjectival OWIZA KI(DET)
‘quilt’

KAŊTO ‘robber’

Figure 6.4 Partial semantic, deep-syntactic and surface-syntactic structures of a nominalized

clause. — The DSyntSs of the sentences (17a–b) are strictly isomorphic and can be
presented as three versions: with a morphological nominalizer (Japanese-style, cʹ.-i)
and with a lexical nominalizer (Korean-style, cʹ.-ii; and Lakhota-style, cʹ.-iii).
256 6 Relative clause: a typology

6.4 Typology of the restrictive relative clause

The typology of RCs will be sketched in six steps: a couple of prototypical RCs
with their semantic and syntactic structures (6.4.1); the DSyntRel and the SSynt­
Rel subordinating an RC (6.4.2); syntactic constraints on RCs (6.4.3); parameters
that specify structural types of RC (6.4.4); illustrations of different types of RCs
(6.4.5); and deep-morphological parameters that specify possible linearizations
and morphologizations of RCs (6.4.6).

6.4.1 R
 elative clauses at the semantic, deep-syntactic and surface-syntactic
levels

To add some linguistic flesh to our abstract RC skeleton, here are RC-construction
structures at the three levels, in Persian and in Russian.

(18) Persian
pesär+i [ke be u ketab+Ø+ra dad +änd]
boy DEF that to s/he book SG DirO give-PAST 3.PL
lit. ‘the boy that to him a book [«they»] gave’
= ‘the boy to whom they gave a book’

PESÄR(N)SG, DEF

PESÄR(N)SG restrictive-relative
KE(CONJ, subord)
‘dadän’ [‘give’] ATTRrestr subord-conjunctional

DADÄN(V)PAST DADÄN(V)PAST
3
ρ III oblique-objectival
BE(PREP)
‘pesär’ [‘boy’]
prepositional
PESÄR(N)SG

USG ‘s/he’

Sem-Structure DSynt-Structure SSynt-Structure

Figure 6.5 Partial semantic, deep-syntactic and surface-syntactic structures of the Persian
phrase (18)
 6.4 Typology of the restrictive relative clause 257

(19) Russian
mal′čik(masc)+Ø, [ kotor+omu da +l +i knig+u]
boy SG.NOM which MASC.SG.DAT give PAST PL book SG.ACC
‘boy to.whom [«they»] gave a book’

MALʹČIK(N, masc)SG MALʹČIK(N, masc)SG

‘datʹ’ [‘give’] ATTRrestr restrictive-relative

DATʹ(V)PERF, PAST DATʹ(V)PERF, PAST

3
ρ III indirect-objectival

‘malʹčik’ [‘boy’]
KOTORYJ(N)MASC, SG
MALʹČIK(N, masc)SG
[‘which’]

Sem-Structure DSynt-Structure SSynt-Structure

Figure 6.6 Partial semantic, deep-syntactic and surface-syntactic structures of the Russian
phrase (19)

6.4.2 Deep- and surface-syntactic relations subordinating a relative clause

The situation with the DSyntRel that subordinates a restrictive RC is unproblem-

atic: it is ATTRrestr, as is the case with all restrictive modifiers. However, as far as
the corresponding SSyntRel is concerned, the situation becomes more compli-
cated. In the first place, it is different in different languages, so that what is said
below describes, strictly speaking, English and perhaps some other structurally
similar languages. The following question arises: Is a special SSyntRel to sub-
ordinate RC—the relative SSyntRel,11 as proposed, e.g., in Iomdin 2010c: 33 and
Mel’čuk 2016: 190—needed or one can do simply with the modificative SSyntRel,
the same as for adjectival modifiers? In all other cases the SSyntRels do not dis-
tinguish between LUs and clauses as their dependents: thus, the subjectival and
direct-objectival SSyntRels subordinate an LU and a clause the same way; prob-
ably, we should not make the distinction for modifiers, either? However, such
a conclusion would be wrong: in English (and other European languages), the

11 “Relative” means by default “restrictive-relative”; a descriptive RC is subordinated to its gov-

ernor by the descriptive-relat­ive SSyntRel.
258 6 Relative clause: a typology

modificative SSyntRel is repeatable (a noun can have a theoretically unlimited

number of cosubordinated—not coordin­ated!—adjectives, as in large expensive
square black table), while the relative SSyntRel is not repeatable: normally, a noun
can have only one restrictive RC.12 To unify these two SSyntRels—the relative and
the modificative—would mean violating Criterion C3 for postulating SSyntRels,
see p. 44.

6.4.3 Syntactic constraints on relative clauses

Languages differ as to their ability to relativize particular elements of the clause.

In other words, the possibility of constructing an RC in a given language is sensi-
tive to the SSynt-role to be played by the image of the RC’s governor within the
RC itself. This fact was established by Keenan and Comrie (Keenan & Comrie
1977, Comrie & Keenan 1979) and formulated as the fundamental Keenan-Comrie
Accessibility Hierarchy:13

Subject > DirO > IndirO > OblO > Gen(itival)Co(mplement) > Compar(ative)Co(mplement)

This hierarchy stipulates that, if the language under consideration allows for RCs
where the SSynt-image of the RC’s governor is, say, an IndirO, then in this lan-
guage an RC can contain the governor’s image in any SSynt-role placed higher in
the hierarchy—in this case, in the role of DirO and Subject.
This means that the DSynt-rules of a language must include a filter  rule that
checks the SSyntS of each RC under synthesis and rejects those that violate the
options allowed for the governor’s image in this language. For an example of such
a filter rule, see the 3 remark after (5), p. 241.

12 Interestingly, a restrictive RC and a descriptive RC can cooccur with the same governor:
(i) The book [THAT you hold now]restrictive [, WHICH belongs to John,]descriptive is precious.
In fact, even two restrictive RCs can exceptionally cooccur (in a very particular context):
(ii) Fr. la seule FEMME [QUE je connaisse] [QUI ne croit pas en Dieu]
‘the only woman whom I know who does not believe in God’ (Sandfeld 1965: 216).
13 The Keenan-Comrie hierarchy was proposed with an eye to relativization, but it turned out to
be valid in regard to other properties of clause elements. In particular, this hierarchy manifests
itself in the ability of a clause element to control reflexive and reciprocal anaphora (Pollard & Sag
1992: 266) and to allow extraction (Abeillé 1997: 23). Cf. also an over-all hierarchy of the syntactic
privileges of clausal elements in French obtained by using a set of relevant syntactic properties
(Iordanskaja & Mel’čuk 2009a).
 6.4 Typology of the restrictive relative clause 259

6.4.4 Syntactic parameters characterizing a relative clause construction

The surface-syntactic structure of an RC construction is specified by the following

five logically inde­pendent parameters:

– Syntactic class of the RC’s governor

– Presence (in the RC) of a subordinator
– Presence (in the RC) of a relative pronoun
– Presence and form of the RC’s governor’s image
– Form of the RC’s governor

Systematic illustration of different types of RC is offered below, in Subsection 6.4.5.

1) Syntactic class of the RC’s governor

This parameter has two possible values:

the RC’s governor is a genuine noun — the RC’s governor is a pronoun

(20) Russian
A Russian pronoun that can function as an RC’s governor is a nominal or
adverbial pronoun of one of the following four types:
– Demonstrative—of the type tot ‘that.somebody’, tam ‘in that place’,
togda ‘at that time’, …; see below, (23d).
– Indeterminate—of the type nekto ‘somebody’, nečto ‘something’,
koe-kto ≈ ‘somebody’, kto-to ≈ ‘somebody’, kto-nibud′ ‘anybody’,
gde-nibud′ ‘anywhere’, koe-kogda ≈ ‘some time’, …
– Negative—of the type nikto ‘nobody’, ničto ‘nothing’, nigde ‘nowhere’,
nikogda ‘never’, …
– Quantifier-like—of the type vse ‘everybody’, ljuboj ‘anybody’, každyj
‘each’, vsjudu ‘everywhere’, vsegda ‘always’, ...

(21) a. English
The country [ that controls the sea] controls the land. vs.
HE [ WHO controls the sea] controls the land.
b. German
Nur DER verdient sich Freiheit wie das Leben,
only that.person deserves oneself freedom as.well.as the life
[ DER täglich sie erobern muß]
who daily these conquer must
260 6 Relative clause: a typology

‘He only earns both freedom and existence / Who must reconquer them
each day’ [W. Goethe].
NB Note the homonymy of the two der: the first (the RC’s governor), a wordform of
der1, is a nominal demonstrative pronoun meaning ‘that person’, while the second (the
governor’s image), a wordform of der2, is a relative pronoun ‘which/who’.

RCs whose governor is a pronoun were formally introduced into the discus-
sion of RC typology by Zaliznjak & Padučeva 1975 [2002: 663], to be later analyzed
by Lehmann 1984: 299–304 and Citko 2004; this type of RC is widespread in Slavic
and Romance languages. A stock example is given in (22a):

(22) Polish (Citko 2004; c = /c/, cz = /č/, ż = /ž/, ó = /u/, ą = /õ/, ę = /ẽ/, y = /ɨ/)
a. Jan czyta TO, [ CO Maria czyta]
Jan reads that(N)ACC whatACC Maria reads
‘Jan reads that what Maria reads’.
b. Jan czyta KSIĄŻK +Ę, [ KTÓRĘ Maria czyta]
Jan reads book(fem) SG.ACC thatFEM.SG.ACC Maria reads
‘Jan reads the book that Maria reads’.

Citko wittily called such RCs “light-headed relatives,” since the demonstrative
pronoun TO ‘that.something(N)’ is semantically almost empty (= “light”) in com-
parison with, e.g., książka ‘book’. However, RCs with a pronominal governor
cannot be “light-headed,” since to is the governor of the RC, but by no means its
head. However, it is true that such RCs are special in many respects.

Let me indicate some prominent facts about RCs with a pronominal governor in
Russian: examples (23a–d).

(23) Russian: RCs where the governor is a correlative

a. These RCs are synthesized from pseudo-RCs (underscored). Special syn-
tactic rules introduce into the SSyntS of a pseudo-relative clause the
semantically empty pronoun to, called correlative,—in order to ensure the
grammatical and/or stylistic correction of the sentence (with the correla-
tive the RC becomes more formal):
Kto xočet poexat′ na èkskursijupseudo-RC, dolžen zapisat′sja zdes′. ~
who wants to.go for excursion, must sign.up here
TOT, [ KTO xočet poexat′ na èkskursiju,]RC dolžen zapisat′sja zdes′.
b. For the same reason, the correlative tot can also be introduced in the
superordinate clause, which gives two more versions of the sentence
being synthesized:
 6.4 Typology of the restrictive relative clause 261

Kto xočet poexat′ na èkskursijupseudo-RC, tot dolžen zapisat′sja zdes′. ~

who wants to.go for excursion, must sign.up here
TOT, [ KTO xočet poexat′ na èkskursiju,]RC tot dolžen zapisat′sja zdes′.

The last sentence features a double expression of the subject: by the RC with tot
and by another tot. However, the duplication of a clause element by an empty
pronoun is known in several Slavic and Romance languages, as well in Modern
Greek and Alban­ian.

c. A syntactically introduced “duplicating” tot can be easily mistaken for

the governor of the pseudo-relative; thus, in the following sentence togo
is not the governor of the pseudo-relative:
Čem serdce bednoe polno, togo ne vyrazjat slova
with.what heart poor is.full that.thing not will.express words
‘Words cannot express what fills [my] poor heart’. [D. Merežkovskij]

The syntactically primary form of this sentence is Čem serdce bednoe polno, ne
vyrazjat slova.

d. An RC with a pronominal governor does not use the “regular” relative
pronoun koto­ryj ‘that/which’, but uses instead a special relative pronoun
corresponding to the pro­nominal governor. As a result, there are the fol-
lowing pairs:
built on the T-stem built on the K-stem
to ‘that.somehing’ ~ čto ‘what
tot ‘that.somebody’ ~ kto ‘who’
TOGDA ‘then’ ~ kogda‘when’
tak ‘this.way’ ~ kak ‘how’
tam ‘there’ ~ gde ‘where’
tuda ‘to.there’ ~ kuda ‘to.where’
ottuda ‘from.there’ ~ otkuda ‘from.where’
Etc.

2) Presence of a subordinator
This parameter has two possible values:

the subordinator is present — the subordinator is absent

A subordinator (a.k.a. complementizer, or—in our case—a relativizer), either invari-

able or mor­phologically agreeing with the RC’s governor, is part of the RC—it con-
stitutes its syntactic head. It can be either a lexeme—a subordinating conjunction,
262 6 Relative clause: a typology

as in (24a); or a morphological marker—an affix, as in (24b); or else a specific

prosody, as in (24c–d). In the examples the subordinator is underscored:

(24) a. Hebrew
HA+IŠA [še Roni natan LA et ha+sefer]
the woman that Roni gave to.her DirO the book
lit. ‘the.woman that Roni gave to.her the.book’
= ‘the woman to whom Roni gave the book’
b. Swahili (ch = /č/, j = /ǯ/; Roman numerals in the glosses stand for noun
classes)
Ni +li +nunua ki+tabu jana ‘I.bought [a] book yesterday’. ~
1.SGsubj PAST buy VII book yesterday
KI+TABU [ni+li+ch+o+nunua jana] ‘[the] book I.bought yesterday’
NB The subordinator prefix o- marks the clause as subordinated; the prefix
ch- [⇐ ki- |_/Vowel/]
shows the agreement in noun class (VII) with the governor, kitabu.

c. Chichewa (Downing & Mtenje 2011; th = /th/, the acute accent indicates
the high tone)
M+BALÁ i +ná +bá n+dalámá z+àángá  ‘A.thief stole my money’. ~
IX thief IXsubj PAST steal X money X my
M+BALÁ [í +ná +bá n+dalámá z+àángá] i +ku +tháawa
IX thief IXsubj-REL PAST steal X money X my IXsubj PRES  run.away
‘The thief who stole my money is running away’.
NB In (24c), the subordinator is expressed by the high tone on the subject prefix i- of
the Main Verb of the RC. This tonal subordinator of an RC can cooccur with the suffixal
subordinator -o and with the relative pronoun méné ‘which’:

d. Chichewa
M+BALÁ [ I +MÉNÉ í +ná +bá n+dalámá z+angáa+y+o]
IX thief IXsubj which IXsubj-REL PAST steal X money X my IX REL
i +ku +tháawa
IXsubj PRES run.away
‘The thief who stole my money is running away’.

If the subordinator is a lexeme, in the SSyntS this lexeme is the syntactic governor
of the RC’s Main Verb, i.e. the head of the RC. It can, but need not, cooccur with
the RC’s governor’s image. Schematically (an artificial English example):

the MAN→[that WHOM I←met yesterday…]

 6.4 Typology of the restrictive relative clause 263

3) Presence of a relative pronoun

This parameter has three values:

the relative pronoun is present (in the RC) —

the relative pronoun is present (in the RC) and
a correlative pronoun is present (in the superordinate clause) —
the relative pronoun is absent (from the RC)

A relative pronoun present in the RC may require a correlative demonstrative

pronoun in the superordinate clause (underscored in (25)); this correlative pronoun
syntactically depends on the RC’s governor.

(25) Hindi
muǯhe vah ĀDMĪ [ ǯo Sītā+ko ačč hā lagtā hɛ] pasand nahī̃ hɛ~
I-DAT that man which Sita DAT nice seeming is likable not is
lit. ‘To me that man which to.Sita nice seeming is likable not is’.
= ‘I don’t like the man who seems nice to Sita’.
[ ǯo Sītā+ko ačč hā lagtā hɛ] muǯhe vah ĀDMĪ pasand nahī̃ hɛ
which Sita DAT nice seeming is I-DAT that man likable not is
lit. ‘Which to.Sita nice seeming is, to.me that man likable not is’.
= ‘I don’t like the man who seems nice to Sita’.

Hindi has fairly complex rules that allow/require the introduction of the correla-
tive demon­strative, as well as the use of the correlative instead of the governor or
its image (Mahajan 2000); these rules do not concern us here.

4) Presence and form of the governor’s image

This parameter has three possible values:

the governor’s image is a duplicate of the governor —

the governor’s image is elided (= gapped) —
the governor’s image is replaced with a pronoun

NB In a more detailed study, the type of the replacing pronoun—relative or personal—should

be taken into account. Interestingly, relative pronominalization characteristic of European lan-
guages is, according to Comrie 1998, a typological rarity.

See examples (27)–(32) below.

Combining the values of Parameters 1–4 (2 × 2 × 3 × 3) produces 36 logically pos-

sible syn­tactic types of restrictive RC. As will be shown, deep morphology—word
order and inflection—multiplies this number by 15; as a result, we have 540 types
264 6 Relative clause: a typology

all in all. Small wonder, then, that the description of RC is so difficult! Of course,
not all logically possible combinations are actually allowed in all languages,
but there still are too many possible syntactic-morphological types of RC, which
defies their systematic exhaustive overview in this chapter. Consequently, only
some types of RC are illustrated in Subsection 6.4.5.

6.4.5 Illustrations of different types of RC constructions (L→RC)

The RC and its gloss are boldfaced and put in square brackets; the RC’s governor
is printed in italic caps; the governor’s image is also printed in boldfaced italic
caps and boxed.

6.4.5.1 The RC’s governor is a non-pronominal noun (L(N, non-pron)→RC)

A subordinator is absent; a relative pronoun is absent; the governor’s image is a
duplicate of the governor

(26) Vietnamese
a. Ðứa CON [đứa CON trễ] khóc
CLASS child CLASS child be.late cry
lit. ‘The child [the child is.late] is.crying’. = ‘The child who is late is crying’.
b. Ðứa CON [đứa CON khóc] trễ
CLASS child CLASS child cry be.late
lit. ‘The child [the child is.crying] is.late. = ‘The child who is crying is late’.

A subordinator is absent; a relative pronoun is absent; the governor’s image is

elided

(27) a. Japanese (Kuno 1973: 249, (12a))

[sin+da +no +de minna +ga kanasin +da] HITO
die PAST NOMINALIZER LOC everybody SUBJ be.saddened PAST person
lit. ‘[died.fact.because.of everybody saddened.was] person’
= ‘person because of whose death everybody was saddened’.

Japanese allows for relativization of any clause elements and does not require
explicit expression of the semantic relation between the RC and its governor:

atama+ga yoku naru hon

lit. ‘head good becomes [because of it] book’
= ‘a book that makes you smart’
 6.4 Typology of the restrictive relative clause 265

toire+ni ikenai komāsyaru

lit. ‘to.toilet do.not.go [because of it] commercial’
= ‘a commercial that won’t let you go to the bathroom’
hon+o katta oturi
lit. ‘bookACC have.bought change’ = ‘change from buying the book’

(See Shibatani 2009: 167–168 on the obligatory ellipsis of the governor’s image as
one of the means of marking the subordinate character of the RC in Japanese; the
other one is word order: Japanese being a strict verb-final language, the linear
position of a finite verb before a noun marks this verb as the Main Verb of an RC.)

b. Ancash Quechua (Cole 1987)

[Nuna+Ø ranti+shqa+n] BESTYA+Ø alli ka+rqo +n
man NOM buy PERF 3 horse NOM good be PAST 3
lit. ‘[Man bought] horse good was’. = ‘The horse the man bought was good’.

c. Mohave (Lehmann 1984: 111)

ʔaqwaq+ny ʔ +akyā+k +e ʔUTIS [m +ūmač]+ny +m
deer DEF 1.SG shoot REAL EMPH.3 gun 2.SG find DEF with
lit. ‘Deer.the I.did.shoot gun [you.find.the.with]’.
= ‘I did shoot the deer with the gun you had found’.

A subordinator is absent; a relative pronoun is absent; the governor’s image is

present—as a personal pronoun

(28) Arabic, with an indefinite governor

KITAB +ĀNI [a +qraʔ+u = HUMĀ ] hunā
book(masc) DUAL.NOM 1.SG read 1.SG they-MASC.DUAL.ACC here
lit. ‘Two.books [I.read.them.two] [are] here’. = ‘Two books that I read are here’.

NB 1. The equality sign (=) shows the adjunction of a clitic rather than that of an affix.
2. If the RC ‘s governor is definite, the subordinator is necessary; cf. (32b).

A subordinator is absent; a relative pronoun is present; the governor’s image is

present—as this relative pronoun

(29) Russian
KNIG +A, [ KOTOR+UJU ja čitaj+u,] naxoditsja zdes′
book(fem) SG.NOM which FEM.SG.ACC I-NOM read PRES.1.SG is here
lit. ‘Book [which I am.reading] is here’.
266 6 Relative clause: a typology

A subordinator is absent; a relative pronoun is present; the governor’s image is

present—as the governor’s duplicate subordinating this relative pronoun

(30) a. Latin (Livius)

[Qu +ibus die +bus Cum +ae
which MASC.PL.ABL day(masc) PL.ABL Cumae PL.NOM
liberat +ae sunt,] i +s +dem die +bus
liberated FEM.PL.NOM are this MASC.PL.ABL same day(masc) PL.ABL
Semproni +us prospere pugna+Ø +t
Sempronius SG.NOM successfully fight PRES 3.SG
lit. ‘[Which days Cumae are liberated,] the.same days Sempronius suc-
cessfully fights’. = ‘The same days that Cumae are liberated, Sempronius
fights successfully’.
b. Burushaski (Tiffou & Patry 1995: 371, (76))
[Ámen híra xat čía,] khéne híre
which man-SG.DAT letter-SG.NOM I.to.him.gave this man-SG.ERG
sía aúlum bái
read-INF unable is
lit. ‘[Which to.man letter I sent,] this man read [it] unable is’.

A subordinator is present; a relative pronoun is absent; the governor’s image is

elided

(31) Basque: the suffixal subordinator (more precisely, a relativizer) -n is invariable

[gizon+a +k libur+a +Ø eman dio+n] emakume+a +Ø
man the ERG book the NOM given has RELATIVIZER woman the NOM
lit. ‘[the man the book given has.REL] the woman’
= ‘the woman to whom the man has given the book’

A subordinator is present; a relative pronoun is absent; the governor’s image is

present—as a personal pronoun

(32) a. Persian: the lexemic subordinator KE ≈ ‘that’ (a subordinating conjunction)

is invariable14
ketab+e xub +i [ke an+ra mi +xan+äm] inǯast
book IZAFET good DEF that it DirO PRES read 1.SG here.is
lit. ‘Book good [that it I.am.reading] here.is’.
= ‘The good book that I am reading is here’.

14 If the governor’s image is the SSynt-subject in the RC it is obligatorily elided:

(i) ketab+i ke inǯast ‘the.book that is.here’ vs. *ketab+i ke u inǯast ‘the.book that it is.here’
 6.4 Typology of the restrictive relative clause 267

b. Arabic, with a definite governor: the lexemic subordinator ALLAÐ- ≈ ‘such

that…’ (an adjective) agrees with the governor in gender, number and case
AL+KITĀB +ĀNI [allað +āni a +qraʔ+u
the book(masc) DUAL.NOM which MASC.DUAL.NOM 1.SG read 1.SG
= HUMĀ] hunā
they-MASC.DUAL.ACC here
lit. ‘The.two.books [two.which I.read them.two] [are] here’.
= ‘The two books that I am reading are here’.
NB If the RC’s governor is indefinite, the subordinator is absent; cf. (28).

6.4.5.2 The RC’s governor is a nominal pronoun (L(N, pron)→RC)

A subordinator is absent; a relative pronoun is present; the governor’s image is
present—as this relative pronoun

(33) Russian; the governor is a correlative pronoun

a. Without duplication of the governor
TOT, [ KTO išč +et,] vsegda najd +ët~
that.person-MASC.SG.NOM who-NOM seek PRES.3.SG always find-FUT 3.SG
lit. ‘That.person [= ‘he’] who seeks always will.find’.
= ‘One who seeks finds’ [The Gospel of Matthew].
b. With duplication of the governor
TOT, [ KTO išč +et,]
that.person-MASC.SG.NOM who-NOM seek PRES.3.SG
tot vsegda najd +ët
that.person-MASC.SG.NOM always find-FUT 3.SG
lit. ‘That [= ‘he’] [who seeks,] that.person always will.find’.

(34) Latin (Cicero; Zaliznjak & Padučeva 1975 [2002: 670])

ILL +O iugul +e +t [qu +em tradidit
this SG.ABL jugulate SUBJUNCT.PRES 3.SG which MASC.SG.ACC transferred
ens +em ]pseudo-RC
sword(masc) SG.ACC
lit. ‘With.this let.him.jugulate [which he.transferred sword]’.
= ‘Let him jugulate [somebody] with the sword that he transferred’.
268 6 Relative clause: a typology

3 Let it be emphasized that sentences like the following one do not contain an RC:

[ KTO išč +et,] (tot) vsegda najd +ët

who-NOM seek PRES.3.SG (that.person-MASC.SG.NOM) always find-FUT 3.SG
lit. ‘Who seeks, (that.person [= ‘he’]) always will.find’.
The boldfaced clause in brackets is not an RC, but a pseudo-RC, which is the
subject of the verb najdët ‘will.find’. The optional correlative tot in the matrix
clause is a resumpt­ive pronoun duplicate of the subject expressed by the pseudo-
relative.

An identical situation obtains with any other correlative pronoun as the RC’s gov-
ernor, for instance:

(35) Russian
a. Without duplication of the governor
OTTUDA, [ OTKUDA on prišël,] k nam dobiraj+utsja mašinoj
from.there from.where he came to us reach PRES.3.PL by.car
lit. ‘From.there [from.where he came] [people] reach us by car’.
b. With duplication of the governor
OTTUDA, [ OTKUDA on prišël,] ottuda k nam
from.there from.where he came from.there to us
dobiraj+utsja mašinoj
reach PRES.3.PL by.car
lit. ‘From.there from.where he came from.there [people] reach us by car’.

However, in the sentence [ OTKUDA on prišël,] (ottuda) k nam dobirajutsja

mašinoj the clause otkuda on prišël is not an RC, but a pseudo-RC. (Cf. 6.4.4, (23),
p. 260.) Here is another example of pseudo-RCs of the same kind:

(36) Old Russian (Zaliznjak & Padučeva 1975 [2002: 652])

[A čto brat +ъ tvoi dējalъ NASILI +E
and which brother SG.NOM your was.doing violence(neu) SG.ACC
na Novēgorodē] a t +ogo sja otstupi
in Novgorod and that.one NEU.SG.GEN REFL renounce-IMPER
lit. ‘And [which violence [your] brother committed in Novgorod,] and
from.that renounce!’ = ‘The violenceACC that your brother committed in
Novgorod, renounce thatGEN!’
NB 1. The conjunction A ≈ ‘and’ at the beginning of each clause is irrelevant for our illus-
tration. Such a conjunction is a typical means to mark the beginning of a clause in
many languages.
2. In (36), the pseudo-RC functions syntactically as a prolepsis.
 6.4 Typology of the restrictive relative clause 269

6.4.6 D
 eep-morphological parameters characterizing the implementation
of an RC
The deep-morphological representation [DMorphR] of a clause is a linear
sequence, i.e. a string, of the DMorphRs of all its lexemes; this string is supplied
with the information on phras­ing (pauses, stresses, intonation contours). This
subsection is concerned with word order (6.4.6.1) and morphological marking
(6.4.6.2) in and around an RC.

6.4.6.1 Word order parameters characterizing an RC

1) Linear position of the RC (with respect to its superordinate clause)
The RC can occupy one of the five possible linear positions with respect to its
superordinate clause (Lehmann 1984: 49ff; Dryer 2013):

and precedes the governor

(= antenominal)
and follows the governor
anteposed to — postposed to — is inside the superordinate clause
(= postnominal)
and surrounds the governor
(= circumnominal)

NB If the RC is ante- or post-posed to its superordinate clause, it can linearly not be in contact
with the governor (and its dependents). This is what is known as an extraposed RC.

(37) a. The RC is anteposed to its superordinate clause

Latin (Ovidius; Zaliznjak & Padučeva 1975[2002: 678], with a slight modi-
fication of the word order)
[ QU +AE praeteriit,] nec iterum revocabitur
which FEM.SG.NOM passed and.not back will.be.called.back
UND +A
wave(fem) SG.NOM
lit. ‘[Which flowed.past,] and.not back will.brought the.wave’.
b. The RC is postposed to its superordinate clause
An aeroelastic SYSTEM is considered [ THAT includes the cubic stiffness].
c. The RC is inside its superordinate clause and precedes its governor
Latin (Ovidius; Zaliznjak & Padučeva 1975 [2002: 673])
Alter +a, [ QU +AS orien+s hab+u +it],
another FEM.SG.NOM which FEM.PL.ACC East SG.NOM have PERF 3.SG
prelat +a PUELL+IS
preferred FEM.SG.NOM girl PL.DAT
lit. ‘Another, [which East had,] preferred to.girls’. = ‘Another [girl is] pre-
ferred to the girls East had’.
270 6 Relative clause: a typology

d. The RC is inside its superordinate clause and follows its governor

Reviews of this BOOK, [which was published last year,] are mostly positive.
e. The RC is inside its superordinate clause and surrounds its governor (in
other words, a discontinuous RC is wrapped around its governor)
(i)
Hittite (Lehmann 1986: 6)
[KASKAL +z +a KW +IT ]… ASSU +Ø 
campaign SG.ABL and which NEU.SG.ACC goods(neu) SG.NOM
… [utaḫḫun] n +at apedanda halissiyanun
brought-1.SG and it-ACC this-INSTR decorated-1.SG
lit. ‘[And.from.campaign which] GOODS [I.brought,] and.it with.this
I.decorated’. = ‘And the booty which I brought from the campaign, I
decorated it [something mentioned before] with this’.
(ii) Bambara (Keenan & Comrie 1977: 65)
Cὲ ye [ye]… so … [ min ye] san
man PAST PAST horse which see buy
lit. ‘Man has [I have]… horse …[which seen] bought’. =
‘The man bought the horse which I had seen’.
NB The discontinuous RC [n ye … min ye] ‘I have … which seen’ = ‘which I have seen’
is “wrapped” around its governor so ‘horse’.
(iii) Kabiyé (Hiraiwa et al. 2017: 24, (88))15
[Ɛso tɪ ya]… mangu … [ngʊ yɔ] ki +we ɖeu
Eso not bought mango which the 3.SG be good
lit. ‘[Eso not bought] mango [which the] is good’.
= ‘The mango which Eso did not buy is good’.

Tagalog allows for all the three linear positions of the RC with respect to its gov-
ernor:

(38) Tagalog (the oblique case marker ng is pronounced /naŋ/)

a. The RC precedes the governor
[nagbigay ng bigas sa bata] na LALAKI
gave OBL rice DAT child LINKER man
lit. ‘[gave rice to.child] LINKER man’ = ‘the man who gave rice to the child’
NB The linker is an auxiliary (= structural) lexeme that marks an expression as a modi-
fier to the following or to the preceding noun; thus, X na Y(N) and Y(N) na X both mean ‘Y
which is X’.

15 The proposed syntactic description of this Kabiyé sentence is buttressed by the possibility of
sentence (i), absolutely synonymous with it, but showing a different word order:
(i) Mangu ngʊ Ɛso tɪ ya yɔ kiwe ɖeu lit. ‘Mango which Eso not bought the is good’.
 6.4 Typology of the restrictive relative clause 271

b. The RC follows the governor

LALAKI na [nagbigay ng bigas sa bata]
man LINKER gave OBL rice DAT child
lit. ‘man LINKER [gave rice to.child]’
= ‘the man who gave rice to the child’
c. The RC surrounds the governor
[nagbigay]… na LALAKI …[ng bigas sa bata]
gave LINKER man OBL rice DAT child
lit. ‘[gave] LINKER man [rice to.child]’
= ‘the man who gave rice to the child’

2) Linear position of the relative pronoun (inside the RC)

The relative pronoun can occupy one of the following three linear positions:

the first — the last — in situ (that is, where it is required by the syntactic structure of the RC)

Deep-morphological parameters 1 and 2 specify 5 × 3 = 15 major morphological

types of RC, of which the following six are illustrated in (39).

(39) Hindi (Koul 2008: 187–194)

Hindi features much flexibility in linear placement of the RC and of the
relative pronoun within it. The sentences meaning ‘The boy who lives in
Delhi is my brother’ can be implemented as follows:
[ ǯo
a. (i) dillī +mẽ rahtā hε] vah LAṚKĀ+Ø
which-SG Delhi in live-PRES_PART.SG.MASC is that boy NOM
merā bhāī +Ø hε
my brother NOM is
lit. ‘[Which in.Delhi living is] that boy my brother is’.

(ii) [dillī+mẽ ǯo rahtā hε] vah LAṚKĀ+Ø merā bhāī+Ø hε

vah LAṚKĀ+Ø
b. (i) merā bhāī +Ø hε [ ǯo dillī +mẽ
that boy NOM my brother NOM is which-SG Delhi in
rahtā hε]
live-PRES_PART.SG.MASC is
lit. ‘That boy my brother is [which in Delhi living is]’.

(ii) vah LAṚKĀ+Ø merā bhāī+Ø hε [dillī+mẽ ǯo rahtā hε]

272 6 Relative clause: a typology

vah LAṚKĀ+Ø [ ǯo
c. (i) dillī +mẽ rahtā hε]
that boy NOM which-SG Delhi in live-PRES_PART.SG.MASC is
merā bhāī +Ø hε
my brother NOM is
lit. ‘That boy [which in Delhi living is] my brother is’.

(ii) vah LAṚKĀ+Ø [dillī+mẽ ǯo rahtā hε] merā bhāī+Ø hε

6.4.6.2 Inflectional parameters characterizing an RC

Inflectional parameters specific to RCs are 1) the inflection of the subordinator,
2) the “rela­tive” inflection of the RC’s Main Verb (that is, the inflection character-
izing the Main Verb as part of an RC), and 3) the inflection of the RC’s governor
and the governor’s image. Only the RC-related inflection of the governor and its
image will be considered here.

1) Inflection of the governor’s image: progressive case attraction

The expression “progressive case attraction” refers to the deviant grammatical
case of the governor’s image due to the impact of the governor’s case; cf. regres-
sive case attraction in (42).

(40) a. Latin (Horatius; Bianchi 2000: 58)

notant +e IUDIC + E [QU +O nost +i]
judging MASC.SG.ABL judge(masc) SG.ABL who MASC.SG.ABL know 2.SG
lit. ‘judging judge [by.which youSG.know]’
= ‘while the judge you know was judging’
b. Ancient Greek (Xenophon; Grimm 2007: 140)
ándr+es áksio +i t +ēs ELEPHTERÍ+ ĀS
man PL.NOM worthy PL.NOM the FEM.SG.GEN freedom(fem) SG.GEN
[H +ĒS kektḗ +sthe]
which FEM.SG.GEN possess PRES.2.PL
lit. ‘men worthy of.the freedom [of.which youPL.possess]’
= ‘men worthy of the freedom which you possess’

In (40a), the expected case of the relative pronoun quis ‘which, who’ is the accu-
sative qu+em ‘MASC.SG.ACC’, required by the transitive verb ‘know’; the devia-
tion—i.e., the ablative qu+o—is caused by the ablative case of the governor, the
noun iudex ‘judge’ in the Ablativus Absolutus construction. In (40b), the genitive
of the governor noun, elephtería ‘freedom’, is the cause of the deviant genitive
h+ēs of the relative pronoun hos ‘who, which’, instead of the “legitimate” accu-
sative h+ēn.
 6.4 Typology of the restrictive relative clause 273

2) Inflection of RC’s governor

a. Definiteness of the governor

Persian requires an RC-modified N (= the governor of the RC) to have an
obligatory defi­niteness marker, the suffix -i:

(41) Persian
zän +i [ke män u +ra dust dar +äm] inǯast
woman DEF that I s/he DirO friend have 1.SG here.is
(˹DUST DAR˺ is an idiom meaning ‘[to] love’)
lit. ‘The.woman [that I her love] here.is’.

b. Regressive case attraction

“Regressive case attraction” (also known as inverse attraction) refers to
the deviant grammatical case of the governor due to the impact of the
governor image’s case; cf. progressive case attraction above, (40).

(42) a. Latin (Vergilius; Bianchi 2000: 59)

urb +em [ qu +am statu+o]
city(fem) SG.ACC which FEM.SG.ACC found IND.PRES.1.SG
vestr+a est
yours FEM.SG.NOM is
‘The cityACC [whichACC I found] is yours’.
b. Ancient Greek (Sophocles; Grimm 2007: 140)
T +on andr +ón [H +ON
the MASC.SG.ACC man(masc) SG.ACC which MASC.SG.ACC
pálai zḗt +eis] éstin entháde
for.long.time look.for PRES.2.SG is here
lit. ‘The manACC [whichACC youSG look.for for.a.long.time] is here’.

The case of the noun urbs ‘city’, which is the subject of the sentence (42a), should
be the nomin­ative: urb+s; the deviation is caused by the influence of the accusa-
tive of the governor’s image quam. Similarly, in (42b) the subject noun anér ‘man’
must be in the nominative, but receives the accusative by “assimilating” its own
case to the accusative of the relative pronoun.
274 6 Relative clause: a typology

c. Ingrian (= Izhorian); Kholodilova 2013)

ihmis+tä [ ke +tä siä nä+i +t] ja miu+n
man SG.PART who SG.PART you.NOM see PAST 2SG and I GEN
veiko +in o +vat üstäv +i +ä
brother.NOM 1SG be PRES.3PL friend PL PART
‘The man whom you saw and my brother are friends’.

The head of the subject phrase of the matrix clause, ihmine ‘man’, is supposed
to be in the nominative case; but it can get the partitive case under the impact of
the relative pronoun ke, which obtains its partitive as the direct object of the verb
nä- ‘see’.
7 esli …, to … ‘if …, then …’
Syntax of binary conjunctions in Russian
7.1 The problem stated: the binary conjunction esli …, to … ‘if …, then …’
in the syntactic structure of a sentence
7.2 The problem solved
7.3 Conjunctions: a small typology
7.4 Binary conjunctions in Russian
7.5 Phraseological nature of binary conjunctions

7.1 T
 he problem stated: the binary conjunction esli …, to …
‘if …, then …’ in the syntactic structure of a sentence
The object of this chapter is the set of binary conjunctions (Rus. parnye sojuzy),
subordinating and coordinating; they are also known in the English-language lit-
erature as correlative subordinators/coordin­ators (Quirk et al. 1991: 935–941, 999–
1001). A typical example can be the Russian binary subordinating conjunction
esli …, to … ‘if …, then …’; and the question asked is as follows:

What is the syntactic structure of a complex sentence including this

conjunction?

For instance:

(1) Rus. Esli ty←somnevaeš′sja, to ja←mogu→proverit′

‘If you doubt, then I can verify’.

All syntactic links in sentence (1) are obvious, except for the particle to2, the
second component of the binary conjunction under analysis.1 The problem with
to2 stems from the fact that this lexeme cannot be used alone—i.e. without esli
‘if’ (unlike the English then in the English binary conjunction if …, then …). As a
result, the first idea that comes to mind is to make to2 syntactically dependent on
esli: esli–r→to2; all the more so, because to2 is linearly positioned with respect
to esli, namely, necessarily after it (see immediately below, Criterion A for the
presence of a syntactic link, Condition 2). Then the binary conjunction esli …,

1 to1 is a component of three Russian repeated conjunctions to1 …, to1 …, to1 ... ‘now …, now …,
now …’, ˹to1 li˺…, ˹to1 li˺…, ˹to1 li˺ … ‘whether…, or…, or…’ and ˹ne to1˺ …, ˹ne to1˺ …, ˹ne to1 ˺ …
‘maybe …, or …, or …’.

https://doi.org/10.1515/9783110694765-008
276 7 Syntax of binary conjunctions in Russian

to2 … can be stored in the lexicon exactly in the form of this syntactic subtree.
Such a description—launched, probably, by myself—has been tacitly accepted
and applied for almost half a century:

–– In Mel’čuk 1974: 231, No. 31, (e), the surface-syntactic relation [SSyntRel] r
between esli and to2 was called “1st auxiliary.”
–– In Mel’čuk & Pertsov 1987: 331, No. 19.1, it was rebaptized “binary-junctive.”
–– In Iomdin 2010c: 43, 1.2.4.5, it appears under the name of “correlative.”
–– In Mel’čuk 2012b: 143, No. 51, it is “correlative-auxiliary.”

The name of this SSyntRel is, of course, not important: what really matters is the
syntactic dependency of to2 on esli.
However, the syntactic description of the conjunction under analysis as­
ESLI–r→TO2 contradicts the definition of syntactic dependency! More precisely, I am
referring to the definition of surface-syntactic relation that was advanced in Mel’čuk
1988: 130–144 and has been used as such since (Mel’čuk 2009a: 25–40 and 2012–
2015: vol. 3, 411–433). For the ease of reading, I will reproduce here the first part of
this defi­n­ition: Criterion A of the presence of a syntactic dependency between two
lexemes in a sentence. (Criteria B and C are not relevant to the following discussion.)

Criterion A: Presence of a surface-syntactic relation between lexemes L1 and L2

(Chapter 2, Section 2.4, p. 40)
Prosodic unity and linear arrangement in the configuration L1–r–L2
In a given utterance U of L, the lexemes L1 and L2 can have a direct Synt-
dependency link, that is, they can form a configuration L1–r–L2, if and only
if both of Conditions 1 and 2 are simultaneously satisfied:
Condition 1
(a) General case
L1–r–L2 can be implemented by a phrase of language L, such as N—V,
V—N, ADJ—N, PREP—N, ADV—ADJ, NUM—N, etc.
(b) Special case
L1–r–L2 alone cannot be implemented by a phrase of L, but taken together
with a convenient configuration of lexemes from the set {Li} appearing in
the same utterance it can, such that the following three configurations
are implementable by phrases of L:
1) L1—{Li-1} L2—{Li-2}, 2) L1—{Li-1} and 3) L2—{Li-2}.
Condition 2
The linear position of one of the lexemes L1 and L2 in the phrase L1–r–L2 is
specified with respect to the other.
 7.2 The problem solved 277

Examples

The Special case (b) of Condition 1 covers two types of expressions:

(i) L1—L2—L3(N), as in oneL1—ofL2—themL3. Here, *one of is not a phrase, while the
utterances of them and one of them are phrases; consequently, the configura-
tion one—of is allowed for.

(ii) L1—{Li-1} L2(CONJ)—Li-2},

as in It becameL1—{obvious}{Li-1} thatL2—{he wasn’t there}{Li-2}.

Here, *became that is not a phrase, while became obvious and that he wasn’t there
are phrases, with became and that as their heads; therefore, the configuration
became—that is accepted as legitimate.

7.2 The problem solved

Rus. esliL2 Y, to2L1 X ‘if Y, then X’ (a binary subordinating conjunction):
–– The expression *esli to is not a phrase; Condition 1 does not allow for the
configuration *esli—to2.
–– esliL2 is a subordinating conjunction; it forms a phrase with the subordinate
clause Y{Li-2} it introduces. The particle to2L1 forms a phrase with the superor-
dinate clause X{Li-1}.
–– esliL2 syntactically subordinates the Main Verb of Y and is itself subordinated
to the Main Verb of X: MV(X){Li-1}→esliL2→MV(Y){Li-2}.
–– to2L1 syntactically depends on the Main Verb of X.2
As a result, we have the following SSynt-structure: esli→Y, to2←X. The second com­
ponent of a binary conjunction—in this case, the particle to2—is subordinated to
the MV of the super­ordinate clause by the auxiliary SSyntRel (Chapter 2, Section 2.5,
SSyntRel No. 46, p. 76).
For readers acquainted with the dependency syntactic descriptions in the
Meaning-Text frame­work the proposed updating must seem quite natural. What
is surprising is the fact that it took so long to see the problem. I am correcting
here a mistake that has been being perpetrated for many years; it concerns all the
binary conjunctions and a motley set of expressions similar to them.

2 TO2 changes the syntactic combinability of the clause C it introduces: a TO2-clause C can be
used only if it has a subordinate clause Cʹ introduced by the conjunction esli. This, however,
is a feature of the active SSynt-valence of the clause, not of its passive valence, which remains
unchanged.
278 7 Syntax of binary conjunctions in Russian

7.3 Conjunctions: a small typology

Before an inventory of Russian binary conjunctions can be offered, I need to
sketch a general typology of conjunctions—in order to give the discussion a
certain depth.

• According to their meaning/function, conjunctions are divided in two major

families: subordinating vs. coordinating.
• According to their form, conjunctions are classified along two independent axes:
–– the number of components: single (just one component) vs. binary (two
components) vs. repeated (built by a theoretically unlimited repetition of
the first component);
–– the structure of components: simple (all its components are monolexemic)
vs. compound (at least one component is plurilexemic).

A binary or repeated conjunction is necessarily discontinuous: its components

cannot be in linear contact within an utterance.
Since repeated conjunctions can be only coordinating, there are 10 logically
possible classes of conjunctions, exemplified with Russian conjunctions in the fol-
lowing table (KĂК and ТĂК stand for unstressed particles), see Table 7.1, next page.

Comments

1. Consider the expressions of the form v svjazi s tem, čto… lit. ‘in connection with
the.fact that …’ (cf. v svjazi s ètim rebënkom ‘in connection with this child’), vsledst-
vie togo, čto… lit. ‘as.consequence of.the.fact that …’ (cf. vsledstvie ètogo rešenija ‘as.
consequence of.this decision’), v silu togo, čto… lit. ‘in virtue of.the.fact that …’ (cf.
v silu ètoj teoremy ‘in virtue of.this theorem’), etc. In spite of often repeated state-
ments, such an expression is not a compound conjunction, although syntactically
it is equivalent to one. An expression of this type consists of a preposition that syn-
tactically subordinates a nominal correlative pronoun: either TOTII.1 ≈ ‘this. one’
or TOTII.2 ≈ ‘the.fact’; this, in turn, leads to one of the following two specific cases:

— TOTII.1 subordinates a relative clause introduced by the relative nominal pro­

noun čto2 ‘what’; both are seen in sentence (2a), whose SSyntS is given in (2b):

(2) a. Oni pošli tuda nesmotrja na to, čto im skazala mat′

‘They went there, despite of this what to.them had.said Mother’.
relative
b. nesmotrja–obl-obj→na–prepos→TOTII.1 čto2←dir-obj–SKAZAT′ …
despite of this.one what tell
 7.3 Conjunctions: a small typology 279

Table 7.1 Universal types of conjunctions

simple: compound:
monolexemic components plurilexemic components

subordinating coordinating subordinating coordinating

1 2 3 4

edva ‘as soon as’ i ‘and’, a ‘≈ and’ ˹kak tol′ko˺ ˹to est′˺ ‘that is’
‘as soon as’
single

esli ‘if’, raz ‘≈ if’ ili ‘or’, ˹potomu čto˺ ˹a takže˺ ‘as well as’
libo ‘or [exclusive]’ ‘because’

kogda ‘when’ no ‘but’ ˹ták kăk˺ ‘since’

xotja ‘although’ ˹v to vremja kak˺

‘in the time as’

5 6 7 8

esli …, (to2) … i …, i … ˹kak tol′ko˺ …, ˹kak …, ták i˺ …

‘if …, then …’ ‘and …, and …’ (tăk) … ‘as soon ‘both …, and …’
as …’

edva …, (KĂK) … ili …, ili … ˹ne tak (on xoroš), ˹ne stol′ko
‘no sooner …, than ‘or…, or…’ kak˺ (my v nëm (slyšimye), skol′kо˺
binary

…’ nuždaemsja) (ugadyvaemye)
‘not so.much he [is] ‘not so.much (heard),
good, as we need him’ but.rather guessed’

tol′ko ..., (KĂK) … ni …, ni …

‘as soon …, then ...’ ‘nor …, nor …’

xotja …, (no) …
‘although …, but …’

9 10

i …, i …, i … ˹to1 li˺…, ˹to1 li˺…,

‘and …, and …, and …’ ˹to1 li˺ …
‘whether…, or…, or…’
repeated

to1 …, to1 …, to1 … ˹ne to1˺…, ˹ne to1˺

— —
‘now …, now …, now …’ …, ˹ne to1˺…
‘maybe …, or …, or …’

libo …, libo …, libo …

‘either …, or…, or…’
280 7 Syntax of binary conjunctions in Russian

TOTII.1 and Čto2 are declined independently, each one in conformity with the
government by its own syntactic governor: vopreki tomu, o čëm im govorila mat′
lit. ‘despite this about what to.them was saying Mother’. TOTII.1 accepts ves′ ‘all’
as modifier: nesmotrja na vsë to, čto im ska­zala mat′ lit. ‘despite of all this what
to.them had.said Mother’.

— The pronoun TOTII.2, which has the surface forms only in neuter singular,
subordinates the invariable semantically empty complementizer (subordi-
nating con­junction) čto1 ‘that5’, which introduces a completive clause. Thus,
sentence (3a) has the SSyntS, (partially) shown in (3b):

(3) a. Oni pošli tuda nesmotrja na to, čto mat′ zapretila im èto
‘They went there despite of this that Mother had.forbidden them this’.
b. pojti–circum→nesmotrja–obl-obj→na–prepos→TOTII.2–correl
go despite of the.fact
→čto1–subord-conjunct→zapretit′ …
that5(CONJ) forbid

TOTII.2 does not, of course, accept any modifier.

2. The parentheses around the second component of a binary conjunction indicate

that this component is optional, which means that the given conjunction is but an
“expanded” version of the basic—single—conjunction. Thus, esli …, (to2) … is an
optional expansion of esli, and so forth.
NB The omissibility of the optional component of a binary conjunction is not absolute, but de­-
pends on contextual conditions. Thus, in the Russian sentence Edva ja vstal, kak upal snova­
‘No.sooner I stood.up than [I] fell again’ the component kak cannot be left out. On the other
hand, the optional component of a binary conjunction can be excluded in a particular context:
see Section 7.5, the lexical entry for ESLI ‘if’, p. 285.

3. All plurilexemic components of compound conjunctions are idioms, which is

shown, as always, by top corners: ˹potomu čto˺ ‘because’, ˹kak tol′ko˺ …, (tăk)
... ‘as soon as …, ...’, etc.

4. The initial component of a binary or repeated coordinating conjunction—shown

in (4) in boldface—behaves not as a conjunction, but as a particle syntactically
subordinated to the head of the first coordinated phrase by the restrictive SSyntRel:

(4) a. Ja ne xoču ni←restr–est′,–coordinative→ni–coord-conjunct→pit′

‘I don’t want neither eat nor drink’.

b. Ja xoču to_li←restr–est′,–coordinative→to_li–coord-conjunct→pit′
‘I want or.maybe eat or.maybe drink’.
 7.4 Binary conjunctions in Russian 281

Similarly, the second component of a binary subordinating conjunction is not a

conjunction, either, but also a particle:
circumstantial
auxil
(5) a. Edva ja priotkryl dver′, (kak) kot vyskočil na ploščadku
‘No.sooner I had.slightly.opened the.door than the.cat jumped onto the.
landing’.
circumstantial
auxil
b. Esli ja uedu, (to) kto budet polivat′ cvety?
‘If I leave, (then) who will water the.plants?’
NB The auxiliary SSyntRel links the second component of a binary subordinating conjunction—
in this case, kak and to2—to the syntactic head of the superordinate clause. 

5. Russian has syntactic constructions that express the conjunction-like meaning

‘as soon as…’/‘no sooner…, than…’ and play the role of binary conjunctions:

–– stóit/stóilo X-u YINF, PERF, kak…  Stoilo mne pojavit′sja, kak Ivan uxodil
‘As soon as I appeared, Ivan would leave’.
–– ne uspelASPECT X YINF, PERF, kak… MVASPECT Ne uspelaPERF ja pojavit′sja, kak Ivan
ušëlPERF ‘As soon as I [female] appeared,
Ivan left’. ~ Ne uspe­valaIMPF ja poja-
vit′sja, kak Ivan uxodilIMPF ‘As soon as I
[female] was appearing, Ivan would be
leaving’.

For simplicity’s sake, these constructions are ignored here, since they do not add
any theoretical difficulty.

7.4 Binary conjunctions in Russian

Here is an illustrative inventory of Russian binary conjunctions (probably, incom-
plete):

1. ˹čem – tem˺ ‘the A1COMPAR …, the A2COMPAR …’

2. edva – (kăk) ‘no sooner …, than …’
3. esli – (to2)/(tăk) ‘if …, then …’
4. eželi – (to2)/(tăk) ‘if …, then …’
5. ˹kak – tak i˺1 ‘both … and …’
6. ˹kak – tak i˺2 ≈ ‘from the moment that …, then …’
7. kogda – (to2) ‘when …, then …’
282 7 Syntax of binary conjunctions in Russian

8. ˹kol′ skoro˺ – (to2) ‘if …, then …’

9. ˹liš′ tol′ko˺ – (kăk) ‘as soon …, then …’
10. ˹ne stol′ko – skol′ko˺ ‘not so.much …, but.rather …’
11. ˹ne tak – kak˺ ‘not so.much …, as …’
12. ne tol′ko – no i ‘not only …, but also …’
13. raz – (to)/(tăk) ≈ ‘if …, then …’
14. ˹tak kak˺ – (to2) ‘since …, then …’
15. tol′ko – (kăk) ‘as soon …, then …’
16. xotja – (no) ‘although …, but …’

Interestingly, a half of these conjunctions are expanded versions of the corres­

ponding single conjunctions.

A couple of examples will be helpful.

(6) ˹ČEM – TEM˺ ‘the …, the …’ (binary subordinating conjunction; an idiom)

comparative
subord-conjunct
Čem bol′še my uglubljaemsja v prošloe, tem←restr–žëstče←stanovitsja
a. 
granica mužskogo i ženskogo mirov
‘The more we go.deeper into past, the rigider becomes the.border [between]
male and female worlds’.

It is worthwhile to show the same SSyntS for this sentence, but with a different
ordering of the superordinate and subordinate clauses:

b. Granica mužskogo i ženskogo mirov subord-conjunct

stanovitsja tem←restr–žëstče,–compar→čem bol′še my uglubljaemsja v prošloe
‘The.border [between] male and female worlds becomes the rigider the
more we go.deeper into past’.

Unlike other Russian binary conjunctions, ˹čem – tem˺ allows for both the ante-
position and the postposition of the subordinate clause.

(7) ˹NE StoL′KO – SKOL′KO˺ ‘not so.much …, but.rather …’

(coordinating binary conjunction; an idiom)

a. On ne stol′ko←restr–sražalsja,–coord→skol′ko–coord-conjunct→byl sražaem
‘He not so.much was.battling, but.rather was being.battled’.

Here the second component, which corresponds to the interrogative-relative pro-

nominal adverb skol′ko ‘how much’, is a coordinating conjunction, to similar to
a/no ‘but’; cf.:
 7.5 Phraseological nature of binary conjunctions 283

b. On ne←restr–sražalsja,–coord→a/no–coord-conjunctive→byl sražaem
‘He not was.battling, but was being.battled’.

Several expressions are sometimes listed among binary conjunctions, while in

fact they are not. For instance, ˹ne_to ⟨ne_tak, ne_skazat′⟩_čtoby˺ lit. ‘not this
⟨not so, not to.say⟩ that’ – a ⟨no⟩ ‘but’ is not a binary conjunction, because it is not
a conjunction at all. Its first component is an idiom ˹ne_to⟨ne_tak, ne_skazat′⟩_
čtoby˺ ≈ ‘not quite’, which is a kind of adverb that can appear alone—without the
second component a ⟨no⟩ ‘but’; while the second component is a simple coordi-
nating conjunction, which can also appear alone, without the first component:

(8) a. Ne_to_čtoby ja ustal,→a prosto vremja isteklo

‘Not this that I got.tired, but simply time ran.out’.
b. Ja ne ustal,→a prosto vremja isteklo
‘I am not tired, but simply time ran.out’.

7.5 Phraseological nature of binary conjunctions

A binary conjunction is by its very nature a plurilexemic expression that is not

free: it is a phraseme (Mel’čuk 2012d, 2012–2015: vol. 3, 263–362, 2015a). However,
it is quite an uncom­mon phraseme: its components are not syntactically linked
in a direct way. As far as I know, such syntactically discontinuous phrasemes
have not been considered before. Indeed, a phraseme is “a phrase that…,” while
esli – to2 or EDVA – KAK are obviously not phrases of Russian. The solution to
this difficulty is simple: one has to consider these expressions together with their
actantial variables: esli Y, to2 X and EDVA Y, KAK X are bona fide phrases. It is
under this form that they must be stored in a dictionary.
Now, if binary conjunctions are phrasemes, what type of phraseme are they?

–– Eight of 16 Russian binary conjunctions are idioms, since they are non-com-
positional (for the idioms ˹kak Y, tak i X˺1/2, see Mel’čuk 2017):

˹čem Y, tem Х˺ ‘the Y, the X’

˹kak X, tak i Y˺1 ‘both X and Y’
˹kak Y, tak i X˺2 ≈ ‘from the moment that Y, then X’
˹kol′ skoro Y˺, (to2) X ‘if Y, then X’
˹liš′ tol′ko Y˺ – (kăk) X ‘as soon …, then …’
˹ne tak Y, kak Х˺ ‘not so.much X, as Y’
˹ne stol′ko X, skol′ko Y˺ ‘not so.much X, but.rather Y’
˹tak kak˺ – (to2) ‘since …, then …’
284 7 Syntax of binary conjunctions in Russian

–– One coordinating binary conjunction—ne tol′ko – no i ‘not only …, but also

…’ (Dlja èto­go važny ne tol′ko finansy, no i političeskaja volja lit. ‘For this [are]
important not only finances, but also political will’.)—is a formuleme (a sub-
class of clichés; see Mel’čuk 2015a), since it is compositional, but fixed: there
is no *ne liš′ – a takže, where each component is substituted by its synonym.
–– The remaining seven Russian binary conjunctions are collocations, although
of an unusual type: there is no direct syntactic link between the base and the
collocate. The base is the first component, which controls the use of the second
component (collocate); the latter is semantically empty, optional, must follow
the base and occupies the initial linear position in the superordinate clause.

It seems worthwhile to show how Russian binary conjunctions could be pre-

sented in a lexicon of the ECD type (our illustrations are, of course, rough and
incomplete sketches of lexical entries.

A binary conjunction that is an idiom is lexicographically described in a separate

lexical entry as follows.

˹čem Y, tem X˺ ‘the Y, the X’, idiom; conjunction, binary, subordinating

Definition
comparative
˹Čem–subord-conj→MV(Y), tem←restr–ACOMPAR/ADVCOMPAR←⋯–MV(X)˺: ≈ ‘X is
a function of Y’
Government pattern
Y ⇔ II
1. VFIN | VFIN –⋯→ A′COMPAR/ADV′COMPAR

Čem bol′še ljudej budetMV(Y) učastvovat′, tem vyše budetMV(X) verojatnost′ uspexa
‘The more people will participate, the higher will.be probability of.success’.
Syntactic properties
The clause Y precedes or follows the clause X.

The conjunction ˹čem – tem˺ has as DSynt-actant II a VFIN—the Main Verb of the
subordinat­e clause; this VFIN has as its (maybe indirect) dependent an adjective/
an adverb of comparative degree. The conjunction depends itself on the an adjec-
tive/an adverb of comparative degree in the superordinate clause.
 7.5 Phraseological nature of binary conjunctions 285

˹ne stol′ko X, skol′ko Y˺ ‘not so.much X, but.rather Y’, idiom; conjunc-

tion, binary, coordinating

Definition
˹ ne←restr–stol′ko←restr–X,–coord→skol′ko–coord-conj→Y˺: ‘not so.much
X, but rather Y’
Government pattern
Y ⇔ II
L stands here for a lexical unit of any part of speech
1. L

 akoe videnie buduščego javljaetasja ne stol′ko vozvyšennymX, skol′ko

T
spravedlivymY
‘Such a.vision of.future is not so.much noble, but.rather just’.
On sidel ne stol′ko zaX stolom, skol′ko naY stole
‘He was.sitting not so.much at the.table, but.rather on the.table’.
Syntactic properties
The clause Y follows the clause X.

A binary conjunction that is a formuleme appears in the lexical entry of its lexical
anchor (see the Glossary); it is described by means of a non-standard lexical func-
tion. In ne tol′ko – no i ‘not only …, but also …’ the lexical anchor is no ‘but’.

no ‘but’, conjunction, coordinating

Definition
X, no Y: ‘X, but Y’.
…

Lexical Functions
… coord-conj
«not only X, but also Y» : ne←restr–tol′ko←restr–X,–coord→no i←restr–Y


Finally, a binary conjunction that is a collocation of the basic (single) conjunction

is lexico­graphically described in the entry for its base also by a non-standard LF.
In esli – to2 ‘if …, then …’, the base is esli.

esli ‘if’, conjunction, subordinating

Definition
Esli Y, X: ‘if Y, X’.
286 7 Syntax of binary conjunctions in Russian

Government Pattern
Y ⇔ II
1. VFIN

Esli bol′še ljudej učastvuet, verojatnost′ uspexa vyše

‘If more people participate, the probability of success [is] higher’.

Lexical Functions
…
1) esli + ... + to2;
to introduce the 2) not tol′ko/liš′←esli (Esli Ivan pridët, to ja ujdu ‘If Ivan
superordinate clause Y: comes, I’ll leave’; Tol′ko esli Ivan pri­dët, (*to) ja ujdu.);
3) not Y1 + esli + …+ to2 + Y1 (Ivan, esli ty pridëš′,
to2 ←restr–MV(Y) (*to) pokažet tebe vsë ‘Ivan, if you come, will.show you
everything’; see Testelec 2001: 263)

As for a repeated conjunction, only its initial component shows any specificity
(see Section 7.3 above, Comment 4): it is not a conjunction, but a particle depend-
ing on the syntactic head of the first coordinated phrase and indicating the begin-
ning of a repeated conjunction. Thus:

(9) restrictive coord-conj coord-conj

razrabotat′ ili nadëžnyj fil′tr,–coord→ili novuju krasku,–coord→ili xorošee ograždenie

‘develop either a reliable filter, or a new paint, or a good fence’

The second component of a repeated conjunction forms a collocation with the

first one; the second component is the base, the first being its collocate, while the
third, the fourth, etc. components are free repetitions of the second one.
Binary conjunctions are characterized by their “discontinuous” character:
they form phrases only together with their actants, since their own components
are syntactically not directly linked. In this, they are unlike almost all other
phrasemes. However, they share this feature with a few idioms, which I would
like to quote here:

Rus. ˹poka I.2b ne←X˺

‘before X [takes place]’ (Iordanskaja & Mel’čuk 2009b): Poka on ne ušëlX, …
‘Before he leaves, …’, with an expletive ne ‘not’, which is obligatory and
does not negate.
 7.5 Phraseological nature of binary conjunctions 287

Rus. ˹pri vsëm←X-e˺

‘despite X’ (V. Apresjan 2014):
pri vsëm ego talanteX ‘despite [lit. with all] his talent’
r

Rus. ˹to_li eščë←X(V)˺

‘I signal that X(V) will take place with an extraordinary (very bad or very good)
actant r’: To li ty togda eščë uvidiš′X!
‘This whether you then still will.see!’ =
‘I signal that you will then see something very bad’.

Rus. ˹čto za←X˺? ≈

‘What kind of X?’: Čto Ivan za čelovek? ‘What kind of human being is Ivan?’

Eng. ˹nothing→if not←X(ADJ)˺ ≈

‘extremely X’: Barbara was nothing if not feminineX.

Fr. ˹en tout←X(N)˺

‘while being completely ADJ(X)’: Tu le feras en toute libertéX lit. ‘You this will.
do in all freedom’. = ‘You will do this while being fully free’ (Anscombre
2001).

An expanded and corrected English version of Mel’čuk 2017.

8 T
 he East/Southeast Asian answer to the
European passive
8.1 The problem stated
8.2 The passive voice
8.2.1 Requirements on a scientific definition
8.2.1.1 Substantive requirements on a scientific definition
8.2.1.2 Formal requirements on a scientific definition
8.2.2 Prototypical passives
8.2.3 Diathesis
8.2.4 Grammatical voice
8.2.5 Passive
8.3 Chinese “passive construction”
8.4 Affected-subject construction in Chinese
8.5 Affected-subject constructions in Southeast Asian languages
8.6 The problem solved

To Vitja Xrakovskij
Kak vse my znaem, on takovskij,
Naš slavnyj Viktor Ès. Xrakovskij!
Svoj nežnyj šlju emu privet,
Želaju žitʹ sto dvadcatʹ let!1

8.1 The problem stated

This chapter tries to answer a seemingly simple question:
Is there a passive voice in Mandarin Chinese?

1 ‘As we all know, he’s like that, / Our glorious Viktor Es. Khrakovskij! / I am sending him my ten-
der greetings, / And wishing that he live hundred twenty years!’ — However, while formulating
this desire, I feel slightly embarrassed. The fact is that way back, in 1948, the Soviet Union was
celebrating the 70-th anniversary of the Greatest Leader of all times and peoples, the Most Fa-
mous General and the Most Beloved Father of scientists and athletes, as well as the First Linguist
and the First Gynecologist, Comrade Joseph V. Stalin. In the middle of the festivities, students of
the Moscow Foreign Languages “Maurice Thorez” Institute prepared, as was the custom, a hand-
written wall newspaper dedicated to the event. It featured a poem created by a local Homer, in
which the author wished that “the Great Stalin might live hundred thousand years.” Everything
seemed perfect, when, suddenly, the school’s Communist Party Bureau ordered the paper re-
moved from the wall and destroyed. The editor got a Party reprimand. “Why should we limit our
greatest Leader’s longevity?”, he was told. Many years later, I have my doubts: Maybe the Party
Bureau was right after all? Maybe we really should not limit?

https://doi.org/10.1515/9783110694765-009
290 8 The East/Southeast Asian answer to the European passive

A number of descriptive grammars, reference books, manuals and special studies

speak about the passive voice in Chinese, indicating, however, its particularities
with respect to what is called passive in many European languages. I will start
with presenting my blunt answer:

No, there is no such a thing as passive voice in Mandarin Chinese.

The goal that “our” languages achieve by using the passive is reached in Chinese—
and several Southeast Asian languages—in quite a different way. The rest of the
chapter demonstrates what exactly this way is.
What is needed to establish whether there is or is not the passive voice in
Chinese? I will try to answer by following the example of a Soviet-era military
medical cadet of the last cen­tury, who was asked at the final examination at
his Academy: “What do you need to give an enema?” He became famous for his
prompt answer: “First, you need an enema; second, you need an anus; third, you
need to apply the first to the second.” In the same vein, I need, first, a definition
of the passive voice; second, a precise description of the relevant Chinese facts;
third, I have to apply the first to the second—and bingo!
Consequently, the chapter is organized in an obvious way: Section 8.2 pres-
ents a definition of the passive voice; Section 8.3 describes the construction called
passive in Chinese; Section 8.4 applies the definition of the passive to the Chinese
facts in order to achieve and buttress the conclusion that what we see is not the
passive voice, but an essentially different phenomenon; Section 8.5 sketches the
situation in a couple of Southeast Asian languages facing a similar problem; and
Section 8.6 presents the conclusion.
It is not by chance that the first version of this text was written as a con-
tribution for Viktor Xrakovskij’s Festschrift: Xrakovskij is one of those scholars
who pioneered intensive and extensive investigations into the problems of voice
in many languages. He also published several important and influential studies
in the domain: Xrakovskij 1974, 1975, 1981, to name but a few (they were repub-
lished, with corrections and additions, in Xrakovskij 1999); see also Xrakovskij
(ed.) 1981 and the “summing-up” paper Xrakovskij 2004. Therefore, this chapter
is a tribute to his long-standing and fruitful work in the domain.

8.2 The passive voice

To propose a rigorous definition of the passive voice, we need to formulate the

principles on which such a definition must be based (8.2.1) and then give the
definitions of diathesis (8.2.2), of voice (8.2.3), and of the passive (8.2.4).
 8.2 The passive voice 291

8.2.1 Requirements on a scientific definition

Since this chapter is essentially based on the definitions of the concepts used, it
is worth­while to dwell on the concept of scientific definition itself. I will consider
definitions of lin­guistic concepts, although what is said might well apply in other
fields (see Mel’čuk 2006b).

8.2.1.1 Substantive requirements on a scientific definition

A scientific definition of a concept X must satisfy the following three substantive
requirements related to the question “What exactly is to be defined?”

1) A definition of the phenomenon X one wants to define must be oriented

towards proto­typical cases of X; “deviant” cases are to be covered by addi-
tional special conditions.
2) The phenomenon X must be defined as a particular case, or a subclass, of
a more general phenomenon Y. In other words, a definition must be strictly
deductive—that is, to be an Aristotelian/Boetian analytical definition of the
form “X is a Y that is Z,” where Y and Z have been defined previously.
3) Specific differences Z—properties that define X as a subclass of Y—must be
reduced to the simplest2 defining features possible, so that they ensure a
systematic hierarchic­al class inclusion.

For instance, consider a commonly used definition of ergative construction

(which does not satisfy one of the above requirements and will be shown to be
incorrect):

Definition *8.1 – *ergative construction

Ergative construction [= X] is a transitive verb predicative construction of

the form
Subj — Verb — DirO [= Y]
such that [= Z]:
(i) its direct object is marked in the same way as the subject of an intran-
sitive verb in this language;
(ii) its direct object is marked by the nominative case;
(iii) its subject is marked by a special case different from the nominative.

2 ‘Simplest’ is to be construed liberally enough: simplest, but such that allows for a sensible
classifica­tion.
292 8 The East/Southeast Asian answer to the European passive

This definition satisfies our first and second substantive requirements above, but
fails the third one: its Z is not the simplest possible—it consists of three indepen-
dent properties each of which can be absent. Taking out one of them produces a
new definition that defines … what? Something for which there is no name and
which does not belong to a previously defined subclass. What is indeed a tran-
sitive verb predicative construction for which Condition (i) is not satisfied—its
direct object is not marked (under specific circumstances) in the same way as
the subject of an intransitive verb, but Conditions (ii) and (iii) are satisfied? Such
a situation is found, for in­stance, in Motu (Lister-Turner & Clark 1931: 28–30), in
common sentences of type (1b):

(1) a. Sisia+na vada e +la

dog PATH PERF 3.SGSUB go
‘The dog has gone’.
b. Sisia+ese boroma+Ø vada e +kori+a
dog ERG pig NOM PERF 3.SGSUB bite 3.SGOBJ
‘The dog has bitten the pig’.
c. Sisia+ese mero+na vada e +kori+a
dog ERG boy PATH PERF 3.SGSUB bite 3.SGOBJ
‘The dog has bitten the boy’.
NB In Motu, the DirO is marked by the nominative, except for human nouns: with them,
it is marked by the pathetive case (a special case found also in some Malayo-Polynesian
and Australian lan­guages, see Mel’čuk 1988: 180–181). The intransitive subject in Motu is
marked by the pathetive (≠ nominative!), and the transitive subject, by the ergative case.

According to the letter of Definition *8.1, (1b) is not an ergative construction

because its DirO is not marked the same way as an intransitive subject. But what
is it? This type of construction clearly belongs to a subclass of transitive verbal
constructions that also includes the ergative construction in the sense of Defini-
tion *8.1; however, this subclass has no defini­tion and no name. The construction
in (1c), which is very close to (1b), belongs nevertheless to a different subclass
(of transitive verbal constructions), to which we even cannot refer: it does not
have a name, either. Worse, these two subclasses do not form a common subclass
within the class of transitive constructions. In order to avoid such violations of
step-wise consistent hierarchical classification, I propose to define first the most
general subclass of verbal predicative constructions that includes the Definition
*8.1 ergative construction as a particular case (see Mel’čuk 1988: 182, 251, 258ff;
Mel’čuk 2006a: 269ff).
 8.2 The passive voice 293

Definition 8.1 – ergative construction

Ergative construction [= X] is a verbal predicative construction [= Y] whose

SSynt-subject is marked by a case different from the nominative [= Z].

Proceeding from Definition 8.1, one would say that in (1a) we have an intransitive
ergative construction, in (1b) a transitive ergative construction with a nominative
DirO (the most current type), and in (1c) a transitive ergative construction with a
non-nominative DirO. The ergative construction in the sense of Definition *8.1 is
then a transitive ergative construc­tion with a nominative DirO that coincides with
the intransitive nominative Subject—a very particular case.

The proposed way of defining guarantees a systematic inclusion of concepts in

the corresponding subclasses, without missing important intermediate classes.

8.2.1.2 Formal requirements on a scientific definition

A definition must also be formally correct in the three respects, related to the
question “How do we define what we define?”:

4) A definition must be formal—that is, it should be applicable verbatim

(= mechanically).
5) A definition must be rigorous—that is, it should contain only concepts which
either have been defined prior to it or else are indefinable (and previously
enumerated as such).
6) A definition must be adequate—that is, sufficient and necessary, covering all
the phenomena that are perceived as subsumable under the corresponding
notion, and nothing but such phenomena.

In this chapter these six requirements on scientific definitions are accepted as

postulates.
294 8 The East/Southeast Asian answer to the European passive

8.2.2 Prototypical passives

As prototypical cases of passives, I take the passive in Latin, Armenian and Swahili:

(2) a. Latin
(i) Serv+i reg+em porta+Ø +nt +Ø
slave PL.NOM king SG.ACC carry PRES.IND 3.PL ACT
‘The slaves are carrying the king’.
(ii) Rex (a) serv+is porta+Ø +t +ur
king-SG.NOM by slave PL.ABL carry PRES.IND 3.SG PASS
‘The king is being carried by the slaves’.
b. Armenian
(i) Ašot+Ø +Ø +ə namak+er+Ø +ə gr +Ø +ec +Ø
Ashot SG NOM DEF letter PL NOM DEF write ACT AOR IND.3.SG
‘Ashot wrote the letters’.3
(ii) Namak+er+Ø +ə gr +v +ec +in Ašot +Ø +i +Ø
letter PL NOM DEF write PASS AOR IND.3.PL Ashot SG DAT NON-DEF
koγmic
from.side
‘The letters were written by Ashot’.
c. Swahili (Roman numerals stand for noun classes)
(i) Wa+tanzania wa+na +sem +Ø +a Ki+swahili
II Tanzania II PRES speak ACT DECLAR VII Swahili
‘Tanzanians speak Swahili’.
(ii) Ki+swahili ki+na +sem +w +a na Wa+tanzania
VII Swahili VII PRES speak PASS DECLAR with/by II Tanzania
‘Swahili is spoken by Tanzanians’.

Here is what can be stated about these prototypical passives:

– There is no propositional semantic difference between sentences (i) and (ii).

They show, of course, a communicative semantic difference: in sentence (i),
the Actor is the Sem-Theme (= topic) of the sentence, while in sentence (ii),
the Patient is. The passive is used to express communicative information.

3 Note the following particularity of the nominal case system in Modern Armenian: it does not
have an accusative, so that the subject and the DirO are both marked by the nominative. A similar
situation is found in other languages, for instance, in Romanian and Nivkh. Here, the DirO is
case-marked the same way as the subject of an intransitive verb. However, the transitive predica-
tive construction of these languages is never called ergative!
 8.2 The passive voice 295

– The crucial syntactic difference between sentences (i) and (ii) is as follows:
–In sentence (i), the Actor is expressed as the deep-syntactic actant I/the
surface-syntactic subject, and the Patient, as the DSyntA II/the direct object.
–In Sentence (ii), the Actor is expressed as the deep-syntactic actant II/the
surface-syntactic agent complement, and the Patient, as the DSyntA I/the
SSynt-subject.
– The crucial morphological difference between sentences (i) and (ii) con-
sists in the dif­ference between the forms of the Main Verb: in sentence (ii) it
has a special suffix (opposed to a zero suffix in sentence (i)), which marks
the communicative and syntactic modification, stated above; this suffix is
the marker of the passive. As a result, we obtain the opposition of active vs.
passive forms. All other morphological differences observed in the verb and
the actantial nouns are automatic consequences of that difference.

Based on the active ~ passive opposition observed in prototypical cases, we must

call passive such verbal forms that are semantically (= propositionally) identical
to active forms, but syntactically entail the transformation characterized above
and are formally marked for this. To describe this transformation, the concept of
diathesis is needed.

8.2.3 Diathesis

Each lexeme that expresses a predicative meaning has actants at the three
levels: SemAs, DSyntAs and SSyntAs. What interests us here is the correspon-
dence between the SemAs and DSyntAs of a lexeme. For instance, the noun joy
(X’s joy over Y) has two SemAs: ‘X’, who experiences the feeling, and ‘Y’, which
is the cause and the object of that feeling; joy also has two DSyntAs: DSyntA I
(implemented on the surface by a possessive form or by a phrase with of), which
expresses ‘X’, and DSyntA II (implemented by a prepositional phrase with over),
which denotes ‘Y’.4

Definition 8.2 – diathesis

The diathesis of a lexeme L is the correspondence between L’s SemAs and

DSyntAs.

4 There is a huge literature on the concepts of diathesis and voice, which I cannot survey even
cursorily. I base this exposition on my own work—in particular, Mel’čuk 2004 and 2006a: 181–262.
296 8 The East/Southeast Asian answer to the European passive

The noun joy has the following diathesis: X ⇔ I, Y ⇔ II; it can also be written as

X Y

I II

In many languages, a verbal lexeme (e.g., a transitive verb) can have more than
one diathesis: such is exactly the case in Latin, Armenian, and Swahili. One of
the diatheses corresponds to the basic, lexicographic form of the verb, while the
other one corresponds to the passive—that is, a form “derived” from the basic
form by the corresponding affix. This other diathesis can be written as X ⇔ II,
Y ⇔ I, or as

X Y

II I

Now we can say that passivization is the following modification of the basic dia-
thesis:

X Y X Y
⇔
I II II I

Three operations are possible on diatheses—permutation of DSyntAs with respect

to SemAs, suppression of DSyntAs, and referential identification of SemAs (with vio-
lation of the correspondence between SemAs and DSyntAs in the basic form).
These operations produce, for a binary basic diathesis, 12 possible distinct modi-
fications, including the zero one (see, e.g., Mel’čuk 2006a: 184–191, 2006b: 194–
209). The zero modification of the basic diathesis corresponds to the active, and
the simplest permutation produces the diathesis that corresponds to the passive.

8.2.4 Grammatical voice

At this juncture, the definition of grammatical voice seems straightforward.

Definition 8.3 – grammatical voice

Grammatical voice is an inflectional verbal category whose grammemes

(= particular voices) mark the modification of the basic diathesis of the verb
and are themselves formally marked on the verb.
Note that formally marked on the verb does not necessarily mean ‘marked on
the verb by an affix’: a modification of the basic diathesis can be marked by a
 8.2 The passive voice 297

structural word such as an auxiliary verb, as in (3a), or an invariable particle, as

in (3b).

(3) a. English/French/German
The letter was written by John. ≡
La lettre a été écrite par Jean
‘The letter has been written par Jean’. ≡
Das Brief wurde von Hans geschrieben
‘The letter became by Hans written’.
b. Albanian, the forms of the aorist indicative (ë = /ə/)
Active vs. Passive
‘I opened’ hapa ~ ‘I was opened’ u hapa
‘YouSG opened’ hape ~ ‘YouSG were opened’ u hape
‘He opened’ hapi ~ ‘He was opened’ u hap
‘We opened’ hapëm ~ ‘We were opened’ u hapëm
‘YouPL opened’ hapët ~ ‘YouPL were opened’ u hapët
‘They opened’ hapën ~ ‘They were opened’ u hapën

Thus, voices can have analytical forms, just like any other inflectional category.
But can there be a change of the basic diathesis that is not marked on the
verb, but on one of its actants? Yes, such a situation is logically possible, and it
exists, for instance, in Ancient Chinese (Jaxontov 1965: 47, 1974: 201):

(4) a. Sha ren ‘[He] killed [a] man’. ~ Sha yu ren ‘[He] was.killed by [a] man’.
b. Cheng bao min, de bao cheng
city.walls protect people virtue protect city.walls
‘City walls protect people, the virtue protects city walls’. ~
Min bao yu cheng, cheng bao yu de
people be.protected by city.walls city.walls be.protected by virtue
‘People are protected by city walls, city walls are protected by virtue’.

The diathesis of the verbs sha ‘kill’ and bao ‘protect’ in the first members of the
pairs of sentences in (4) changes in their occurrences in the second members of
these pairs. However, since this change of the basic diathesis is not marked on
the verb, second sentences of the pairs in (4) do not represent grammatical voice.
These sentences manifest the pseudo-passive construction; a genuine passive con-
struction needs a passive verbal form.
298 8 The East/Southeast Asian answer to the European passive

8.2.5 Passive

Based on Definitions 8.2 and 8.3, passive voice can be now defined.

Definition 8.4 – passive voice

Passive is a (grammatical) voice that marks a modification of the basic dia-

thesis such that it involves the permutation of DSyntA I.

In other words, a passive voice necessarily entails demotion of DSyntA I.

Summing up:

The passive voice of verb V is a synthetic or analytical form of the lexe-

me V that expresses a modification of the basic diathesis of V such that
consists in permuting at least V’s DSyntA I.

We have the first piece of our puzzle. Let’s move to the second one.

8.3 Chinese “passive construction”

Unfortunately (for me), I do not know Chinese, and in what follows I proceed
strictly from the data available in printed sources. In the first place, these are Li &
Thompson 1981, Hashimoto 1988, Ren 1993, Paris 1998, Huang 1999, and Huang
et al. 2008, from which I take my examples; see also Liu 2016, where a rich bibli-
ography of the question is found. (I modified these examples a bit, to make them
easier to understand; among other things, I replaced the Chinese human names
by English ones.) Here is a typical example of what is currently called passive
sentence/passive construction in Chinese:

(5) Mary bèi tǔfěi shā-le bàba

bandits kill PERF father
‘Mary bèi bandits killed father’. = ‘Mary lost her father to bandits’.
(The lexeme bèi cannot be properly glossed before its meaning and syntactic
function are clarified.)

Huang 1999, following in some respects Hashimoto 1988, demonstrates that the
lexeme bèi, commonly called “the passive marker,” is in fact an auxiliary verb
with a very vague meaning ≈ ‘[to] undergo [that ...]’ or ‘[to] be affected by ...’, and
a rather syntactic function. Generally speaking, what follows bèi is a normal full
clause with its own syntactic subject. As a result, sentence (5) is best literally trans-
 8.3 Chinese “passive construction” 299

lated as ‘Mary «underwent.that» bandits killed father’. Here are Huang’s four argu-
ments for this description (again, I slightly reformulated and rearranged them).

1) bèi is not a preposition. In spite of many traditional approaches that classify

bèi as a preposition (e.g., Alleton 1973: 121–122, Li & Thompson 1981: 365, Ren
1993: 127ff),5 the presumed passive marker bèi is by no means a preposition intro-
ducing an agent noun complement. Consider, for instance, (6a), a very common
type of sentence con­taining the lexeme bèi:

(6) a. Mary zuótiān bèi John dǎ-le

yesterday underwent hit PERF
‘Mary was hit by John yesterday’.

One cannot say that here bèi forms a prepositional phrase with John, for at least
three reasons:

– The presumed prepositional phrase *bèi John cannot be positioned in any

other slot in the sentence, while normal prepositional phrases can appear
in all these slots—except for the position between bèi and the subject of the
subordinate clause, see (6b) vs. (6c):

(6) b. *Bèi John Mary zuótiān dǎ-le.

*Mary bèi John zuótiān dǎ-le.
*Mary zuótiān dǎ-le bèi John.
*Mary bèi John dǎ-le zài jiālǐ.
c. Zài jiālǐ Mary bèi John dǎ-le.
at home
Mary zài jiālǐ bèi John dǎ-le.
Mary bèi John zài jiālǐ dǎ-le.
NB But not *Mary bèi zài jiālǐ John dǎ-le; see below, before (14), on the particularities of
the clause introduced by bèi.

– Very often (actually, more often than not) bèi is not followed by a noun, but
by the subjectless verb:

(6) d. Mary bèi dǎ-le ‘Mary was beaten’.

5 M.-C. Paris (1998: 363, footnote 7) says that she considers bèi to be a verb; however, she calls
it “Agent Marker” and describes it as forming a phrase with the Agent noun, that is, technically
treats it as a genuine preposition. — Let me emphasize that, as far back as more than half a cen-
tury ago, Solnceva 1962: 66–67 clearly stated that bèi cannot be con­sidered a preposition, but
rather represents a defective verbal form.
300 8 The East/Southeast Asian answer to the European passive

In this case, it is customary to speak of an ellipsis of the noun; however, the ellip-
sis of a noun between a preposition and a verb would be an exceptional situation
in Chinese. Moreover, this is not a contextual ellipsis: the missing indication of
the agent is absent from the syntactic (and the semantic) structure of the sen-
tence: it cannot be “restored” from previous context.

– bèi normally is not repeated in coordination, although this repetition is near-

obligatory for prepositions in Chinese (Hashimoto 1988: 331–332):

(7) a. Mary bèi qīnrén huáiyí wàirén zhǐzé

relative suspect stranger criticize
‘Mary is suspected by relatives [and] criticized by strangers’.
b. Mary zài xuéxiào xuéxí, zài jiālǐ xiūxi
in school study in home rest
vs. ?Mary zài xuéxiào xuéxí jiālǐ xiūxi.

2) Syncategorematic (= subject-oriented) lexemes in the bèi-governed clause.

If bèi is not a preposition, it is a verb, and the phrase following it is a subordinate
clause—an object of bèi. Let us see what type of clause it is.

– The bèi-governed clause can contain the adverb gùyì ‘intentionally’, as in (8):

(8) Mary bèi John gùyì dǎ-le

intentionally hit PERF
‘Mary underwent.that John intentionally hit [her]’.

Since gùyì can semantically bear only on the syntactic subject and in (8) it bears
on John, the use of gùyì shows that John is the syntactic subject in the bèi-gov-
erned clause (bèi-clause for short).

– The bèi-clause can also contain the reflexive pronoun zìjǐ ≈ ‘(one)self’, which
can be coreferential only with the syntactic subject:

(9) Zhè fēng xìn bèi John dàihuí zìjǐ-de jiā qù-le
this CLASS letter bring.back self ’s home go PERF
‘This letter was brought back by Johni to hisi house’.

This gives another indication to the effect that John is the subject of the bèi-
clause (rather than the complement of the “preposition” bèi).

3) Syntactic phenomena characteristic of normal clauses. A bèi-clause can

feature various constructions that show it to be a normal clause, with a subject, a
Main Verb, and even—if its MV is transitive—with a DirO:
 8.3 Chinese “passive construction” 301

– Coordination with gapping, illustrated by (10):

(10) Mary bèi John mà -le liǎng-cì, Peter dǎ-le sān-xià
scold PERF two times hit PERF three times
‘Mary was scolded twice by John and hit three times by Peter’.

– So-called “long-distance dependencies,” where the element coreferential

with the subject of the bèi-clause syntactically depends on the Main Verb of
this clause not directly, but through a string of subsequent dependencies:
(11) Mary bèi zhèngfǔ pài jǐngchá zhuāzǒu-le
government send police arrest PERF
‘Mary underwent.that government sent police have.arrested’. =
‘Mary was arrested by the police on government’s orders’.
As one can see, the understood (but not expressed) DirO ‘her’ in the bèi-clause
depends on ‘arrest’, which in its turn depends on the MV ‘send’.
– The verb in a bèi-clause can have its own DirO (Chappell 1986: 274 and 277,
where the conditions imposed on this DirO are specified):
(12) a. Tā bèi dírén dǎshāng -le tuǐ
he enemy hit.wound PERF leg
‘He underwent.that enemy hit.wounded leg’. =
‘He had his leg wounded by the enemy’s fire’.
b. Tā bèi péngyou kāi -le yī ge wánxiào
he friend play PERF a CLASS joke
‘He underwent.that friends played a joke’. =
‘He had a joke played on him by his friends’.
c. Yīfu bèi shāo-le yī ge dòng
cloths burn PERF a CLASS hole
‘Cloths underwent.that [fire] burnt a hole’. =
‘The fire burnt a hole in his cloths’.

4) Resumptive pronoun in the bèi-clause. A bèi-clause can contain a resump-

tive pro­noun fulfilling the role of its DirO and coreferential with the subject of the
sentence—in at least two types of context:
– If the resumptive pronoun is not sentence-final, but is followed by a lexical
expression, as in (13a):
(13) a. Mary bèi John dǎ-le tā sān-xià
hit PERF she three times
‘Mary underwent.that John hit her three times’. =
‘Mary was hit by John three times’.
302 8 The East/Southeast Asian answer to the European passive

– If the resumptive pronoun is turned into an affected object, introduced by the

preposition bǍ and preceding the Main Verb, see (13b):

(13) b. Mary bèi John bǎ tā dǎ-le

she hit PERF
‘Mary underwent.that John her hit’. = ‘Mary was hit by John’.

This would not be possible if the Main Verb of the bèi-clause were (in any sense)
passive.6
Taking all this into consideration, we have to accept Hashimoto’s and Huang’s
proposal, summarizing it in the following two points:

– Syntactically, bèi is the Main Verb of the whole sentence; it is a bivalent auxiliary
verb and roughly means something like ‘undergo that ...’ = ‘be affected by ...’.
– The bèi-clause is a normal active clause with its own Main Verb. However, this
clause has a few special properties: it can lack an overt subject not in a contex-
tual controlled ellipsis; if its MV is transitive, its own DirO most often—although
by no means always!—semantical­ly corresponds to the subject of bèi; if this
DirO occupies the last linear position in the sen­tence, it cannot be expressed by
a resumptive pronoun, but otherwise it can; its subject cannot be preceded by a
prepositional phrase, which otherwise is quite common, cf. (6c) above; etc.7

Therefore, a Chinese “passive” sentence is, if literally glossed, something like this:

(14) a. Mary bèi dǎ-le

‘Mary underwent.that [some.people] hit [her]’.
b. Mary bèi John dǎ-le
‘Mary underwent.that John hit [her]’.
c. Mary bèi John dǎ-le tā sān-xià
‘Mary underwent.that John hit her three.times’.

6 A resumptive pronoun cannot appear in a bèi-clause as a DirO (coreferential with the subject
of the sentence) if this pronoun turns out to be clause-final element:
(i) Mary bèi John dǎ-le *tā lit. ‘Mary bèi John hit her’. = ‘Mary was hit by John’.
There is still another argument supplied by Hashimoto 1988: 335 against the sentences with the
bèi-sentence being “passive” constructions: the possible absence of the “active” counterpart.
Thus, for (ii) there is no corres­pondent sentence without bèi:
(ii) Kānshǒu bèi fànrén pǎo-le
‘[The] jailer underwent [= suffered because of] [the] criminal’s escape’.
This argument is, however, invalid: “passives without actives” are not a rarity at all. Recall the
Japanese passive, let alone verba deponentia of Classical languages.
7 Huang 1999: 11:
(i) John bèi Mary zài xuéxiào dǎ-le ‘John was hit by Mary at school’.
vs. (ii) *John bèi zài xuéxiào Mary dǎ-le ‘John was hit at school by Mary’.
 8.3 Chinese “passive construction” 303

bèi is the MV of the whole sentence;8 the SSynt-structure of (14c) is as follows:

bèi ‘by’

direct-
subjectival objectival
dǎ-le ‘have hit’
Mary circumstantial
direct-
objectival subjectival

Mary John sān-xià ‘three times’

Figure 8.1 Surface-syntactic structure of sentence (14c)9 (The bèi-clause is subordinated to bèi
by the direct-objectival SSyntRel.)

8 It is true that bèi does not have several properties of normal Chinese verbs: thus, it does not ac-
cept aspect suffixes -le, ‑guo, -zhe, etc. But then, several Chinese lexemes that are traditionally
accepted as auxiliary verbs—for instance, shǏ ‘let’, néng ‘can’ or yào ‘want’—do not have these
properties, either (Li & Thompson 1981: 172ff; Hashimoto 1988: 339–340). Also bèi alternates with
genuine verbs gěi ‘give’, jiào ‘be called’ and ràng ‘let, allow’ (Li & Thompson 1981: 506):
(i) Wǒ gěi/jiào/ràng tā tōu -le liǎng kuài jīn
I got he steal PERF two dollar money
‘I underwent/allowed that he stole two dollars [from me]’.
On the other hand, Hashimoto 1988: 340 indicates that the continuous aspect suffix ‑zhe is
found on bèi in texts, even if rarely.
It is interesting to mention a linguistic phenomena happening in Chinese now: bèi begins
to be increasingly used to introduce such verbs as zìshā ‘commit suicide’ or zìyuàn ‘volunteer’,
for instance, Tā bèi zìshā lit. ‘He underwent committing.suicide’, to mean that he was murdered,
the murder disguised as a suicide; Tā bèi zìyuàn lit. ‘He underwent volunteering’—that is, he was
forced to volunteer. No matter how sporadic such usages, they show that bèi is not perceived by
speakers as a preposition.
9 The SSyntS presented in Figure 8.1 needs two additional comments.
– If bèi is not recognized as the MV of the sentence, but is considered to be a passive-agent
marking preposition, the SSyntS of (14c) appears as follows:
dǎ-le ≈ ‘have hit’
circumstantial
subjectival passive-agentive

bèi ‘by’
Mary sān-xià ‘three times’
prepositional

John
The alternative SSyntS of (14c)—with bèi as an agentive preposition
In this structure, the MV dā-le has to be interpreted as ‘have been hit’, i.e., as having an “in-
verted” diathesis without any overt marking.
– As indicated to me by R. Niu (p. c.), some Chinese linguists—for instance, Dexi Zhu, Yufa
304 8 The East/Southeast Asian answer to the European passive

The discussion in Section 8.3 gives us the second piece of the puzzle: the bèi-
construction is not passive; it is a normal “active” clause introduced by the aux-
iliary verb bèi, which has very little semantic content in Modern Chinese and is
used mostly for syntactic and communicative purposes. The schematic form of a
bèi-clause is as follows:

‘X undergoes.that Y Z-s (X)’ ⇔ X bèi Y Z(V) (X)

Now I have to apply the first piece of the puzzle to the second one. If something
does not walk like a duck and does not quack like a duck, why call it duck? Perhaps
simply because it reminds us of a duck (for instance, a small goose). Yet this is not
a compelling reason. The bèi-construction is not passive at all: it is built around
the transitive verb bèi that does not change its only diathesis and, quite naturally,
has no marking of a diathesis change; nor does it somehow change the diathesis
of its subordinate verb Z(V). But it does resemble a passive construction in some
essential respects, which will be presented in the next subsection.

8.4 Affected-subject construction in Chinese

The bèi-construction is used when the Speaker wants to speak about X and say
about X that Z done by Y happened to X or at least somehow affected X. Suppose
that Y Z-ed (e.g., offended) X, and the Speaker chooses to communicate this while
talking about X; since in Chinese, as a general rule, the syntactic subject must
express the Sem-Theme of the sentence, he has to say X bèi Y Z. In this way, the
bèi-construction does two things with one blow: communicatively, it turns ‘X’ into
the Sem-Theme; syntactically, it turns L(‘X’) into an affected subject. This is what
makes this construction similar to the passive of many lan­guages: both the bèi-
construction and the passive construction fulfill (almost) the same communicative
and syntactic roles. However, the similarity, even identity, of roles fulfilled by two
linguistic phenomena does not entail the similarity, let alone identity, of the phe-
nomena themselves. Should we consider English prepositions to be case markers
simply because they often play the same role as cases (marking syntactic depen-
dencies)?
The Chinese construction in question should by no means be called passive;
I would suggest for it the straightforward name of affected-subject construction. In

jiangyi [Lec­tures on grammar], 1982, Beijing: Commercial Press, while considering bèi as a
verb, propose to describe it as trivalent rather than bivalent: thus, bèi in (14c) would have as
its DSyntA I the noun Mary, as its DSyntA II, the noun John, and as its DSyntA III, the verb
dā-le. However, for the present discussion this distinction is irrelevant.
 8.4 Affected-subject construction in Chinese 305

addition to its being explicit and clear, this term has another advantage: it forms
a pair with the name of another Chinese construction, described, e.g., in Li &
Thompson 1981: 463ff: the b̌A‑construction, or affected object construction. Let
us start with an example:

(15) Wǒ bǎ chábēi nòngpò-le

I as.for tea.cup break PERF
‘I as.for tea.cup broke’. = ‘I broke the tea cup’.

If a nominal expression N (here, chábēi ‘tea cup’) is intended to be the direct

object, it must follow the Main Verb (here, nòngpò ‘break’), as all DirOs do. But,
if supplied with the preposition b̌A ≈ ‘as for’, it ceases to be a DirO and must
precede the verb. (The b̌A-construc­tion must satisfy several constraints: the verb
in this construction must express an action that really affects the referent of N; N
itself must be definite or generic, but not indefinite specific; etc. This, however, is
not relevant for my purposes here.) The sentence element implemented by b̌A + N
cannot be considered a DirO: first, it is difficult to call a nominal introduced by a
preposition a direct object;10 and second, more importantly, a sentence with the
bA-construc­
̌ tion can contain a genuine DirO (boxed in (16)):

(16) a. Wǒ bǎ júzi bō -le pí

I tangerine remove PERF skin
‘I as.for tangerine removed skin’. = ‘I removed the skin from the tangerine’.
b. Wǒ bǎ John bǎng-le liǎngzhi jiǎo
I tie.up PERF two foot
‘I as.for John tied up two feet’. = ‘I tied up John’s feet’.

Therefore, from a syntactic angle, the b̌A + N phrase must be given a special name;
it is known as affected object. The surface-syntactic relation that links it to the
Main Verb also cannot be called direct-objectival; I propose affected-objectival
(see Chapter 2, Section 2.5, No. 13, p. 56).
The affected-object construction also thematizes the nominal expression
involved, just as the affected-subject construction does; this enhances its simi-
larity with the latter. Huang et al. 2008: 155–162 emphasizes the parallelism of

10 A few cases where a preposition marks a DirO are known. Such is, for instance, the preposi-
tion et in Hebrew, which marks exclusively definite DirOs and does nothing else; the preposition
z- ‘as’ marking the DirO in Classical Armenian; or the prepositions a in Spanish and pe in Ro-
manian, which are necessary for animate DirOs under particular conditions of referentiality and
specificity. However, in none of these cases is the pre­position-marked DirO compatible within
the clause with another DirO.
306 8 The East/Southeast Asian answer to the European passive

both constructions in several respects. But of course this parallelism is not com-
plete: to begin with, b̌A in the b̌A-construction is a prepo­sition, while bèi in the
bèi-construction is a verb; b̌A does not carry any propositional mean­ing, and bèi
does (even if not much); there are other differences as well. But this is beyond the
limits of our discussion.

8.5 Affected-subject constructions in Southeast Asian

languages
To drive the nail home, I will consider what is called “passive” in Vietnamese,
based on Truong 1970 and Tam 1976. Here are examples borrowed from Tam 1976:

(17) a. Nga đánh Nam ‘Nga beat Nam’. ~

Nam bị 11 Nga đánh ‘Nam was beaten by Nga’.
b. Nga khen Nam ‘Nga congratulated Nam’. ~
Nam được Nga khen ‘Nam was congratulated by Nga’.

bị and được are auxiliary verbs with meanings, respectively, ‘undergo, suffer’ and
‘receive, benefit from’; accordingly, they produce sentences with adversative or
beneficial/neutral meaning.
The same situation obtains in Lao, Khmer and Thai (Tam 1976: 442):

(18) a. Lao
Khacaw khaa muu khoi ‘They killed my friend’. ~
they killed friend I
Muu khoi thyyk khacaw khaa ‘My friend was killed by them’.
friend I undergo they killed
b. Khmer
Kee bɔmbaek kbaal kñom ‘They broke my head’. ~
they break head I
Kñom trəw kee bɔmbaek kbaal (kñom) ‘I got my head broken by them’.
I undergo they break head I
c. Thai
Dek tii maa ‘The child hit the dog’. ~
child hit dog
Maa thuuk dek tii ‘The dog was hit by the child’.
dog undergo child hit

11 The auxiliary verb bị stems from the Chinese bèi, borrowed into Vietnamese.
 8.6 The problem solved 307

A similar situation is observed in Burmese:

(19) Cuŋdɔ yaiʻ+ði ðu+go ‘I hit him’. ~

I hit DECLAR he DirO
Ðu cuŋdɔi əyaiʻ+go khang +ya +ði
he my blow DirO experience receive DECLAR
‘He experience.received my blow’. = ‘He was hit by me’.

The difference with the preceding three languages is that what corresponds to
the bèi-clause is nominalized in Burmese: ‘I hit’ ⇒ ‘my blow’; however, in the
present context this is irrelevant.

8.6 The problem solved

Summing up, the Chinese construction with the lexeme bèi is not passive; the
category of voice does not exist in Chinese. bèi is a verb, and the bèi-construc-
tion can be called the affected-subject construction. The same recommendation
applies to similar constructions in Vietnamese, Thai, Lao, Khmer, and Burmese.

12

A corrected version of Mel’čuk 2014c.

9 P
 ronominal idioms with a blasphemous noun in
Russian and syntactically similar expressions
9.1 The problem stated
9.2 К-expressions: their surface-syntactic structure and lexicographic description
9.2.1 Ξ+К-expressions
9.2.1.1 Phraseologized Ξ+К-expressions
9.2.1.1.1 ˹Ξ←К˺-expressions: idioms with a blasphemous noun (Tam živët
˹čërt znaet kto˺ ‘There lives devil knows who’)
9.2.1.1.2 Ξ–⋯→К-expressions: collocations governing an indirect-interro­
gative clause (Čërt znaet, kto on takoj ‘Devil knows who he [is]’)
9.2.1.2 Free Ξ←К-expressions
9.2.2 К+Ξ-expressions
9.2.2.1 Phraseologized К→Ξ-expressions: collocations (Maša poedet
kuda→ugodno ‘Masha will.go to.where pleasant’)
9.2.2.2 Free К←⋯–Ξ-expressions (Maša poedet kuda ty zaxočešʹ ‘Masha will go
to.where youSG will.want’)
9.3 Blasphemous idioms are not syntactic phrasemes

9.1 The problem stated

The object of this chapter is the set of Russian expressions such as [Ona poexala]
čërt znaet kuda lit. ‘She went devil knows to.where’ vs. [Ona poexala] kuda eë
otpravili lit. ‘She went to.where [«they»] her had.sent’ vs. Čërt (eë/ego) znaet,
kuda [ona poexala] lit. ‘Devil (her/it) knows to.where [she went]’. The character-
istic component of such an expression is an inter­rogative-relative pronoun, or a
К-word (to use an Isačenko-Apresjan-Iomdin term, analogous to WH-word): kuda
‘to.where’, kto ‘who’, kak ‘how’, skolʹko ‘how much/many’, … These expres-
sions will be called К-expressions and represented by means of the symbols К (for
the К‑word itself ) and Ξ (for the remaining part of the expression).
In connection with Russian К-expressions two important questions have to
be asked:

– What is the surface-syntactic structure of a К-expression and what must be

its lexicographic description? The answer is given in Section 9.2.
– К-expressions are often called syntactic phrasemes, but I think that this is con-
ceptually incor­rect. Section 9.3 considers the notion of syntactic phraseme
and characterizes К-expressions from the standpoint of their phraseological
status.

https://doi.org/10.1515/9783110694765-010
310 9 Pronominal idioms with a blasphemous noun in Russian

The proposed description of Russian К-expression is based on the data pre-

sented and analyz­ed in several studies by L. Iomdin: Iomdin 2005, 2007 and
2010а,b.
All glosses supplied to Russian examples below are literal.

9.2 К-expressions: their surface-syntactic structure and

­lexicographic description
Russian К-expressions are divided (Iomdin 2010b: 148–149) into two major classes:

– those that end with a К-word, that is, Ξ+К, e.g., [Zdesʹ možno vstretitʹ] čërt
znaet kogo ‘Here [is] possible to.meet devil knows whom’;
– those that begin with a К-word, that is, К+Ξ, e.g., [Zdesʹ možno vstretitʹ] kogo
tolʹko xočešʹ ‘Here [is] possible to.meet whom only you.want’.

The mutual linear position of К and Ξ is invariable in each of these classes.

I follow Iomdin’s division: Ξ+К-expressions are considered in Subsection
9.2.1, and К+Ξ-expressions, in Subsection 9.2.2.

9.2.1 Ξ+К-expressions

Ξ+К-expressions come in two types: phraseologized, i.e., containing a phraseolo-

gized part Ξ (9.2.1.1), and free, i.e., where Ξ is not phraseologized (9.2.1.2).

9.2.1.1 Phraseologized Ξ +К-expressions

A phraseologized Ξ+К-expression consists of two major components: a colloca-
tion Ξ, and an interrogative-relative pronoun—a К-word, which can be introduced
by a preposition. (On colloca­tions, see Chapter 1, Section 1.2.4.2, p. 23; see also
below, Section 9.3.)
In its turn, the collocation Ξ has:

– As its base, the verb znatʹ ‘know’ in the present indicative.

– As its collocate, a noun N(devil), which is the SyntSubj of the verb znatʹ.

The noun N(devil) belongs to a set that was called by the Russian linguist Aaron Dol-
gopolsky (1930–2012) about 50 years ago imja čertyxatelʹnoe (more than approxi-
mately, ‘blasphemous/swearing noun’).1 This set includes the names of:

1 The adjective čertyxatelʹnyj, invented by Dolgopolsky, is derived from the verb čertyxatʹsja
‘[to] swear uttering the name of Devil (= čërt)’ = ‘[to] blaspheme’.
 9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 311

1) unholy power (čërt ‘devil’, bes ≈ ‘demon’, dʹjavol ‘devil’, lešij ‘wood goblin’,
old ljad ≈ ‘devil’);
2) holy power (bog ‘God’, allax ‘Allah’);
3) a plant (xren ‘horseradish’; actually, a euphemism for “penis”);
4) an animal (pës ‘male dog’);
5) a human (kto ‘who’, šut ‘buffoon’);
6) a gesture (fig ‘fig sign’, šiš ‘fig sign’; also a euphemism for “penis”);
7) the old Russian letter “x” (xer; another euphemism for “penis”); and
8) the penis (x - - [extremely vulgar and therefore offensive]).2

As one sees, the set of blasphemous nouns is a syntactic rather than a semantic
class: it includes such a semantically neutral (pro)noun as kto ‘who’ and such
divine nouns as Bog ‘God’ and Allax ‘Allah’.

2 Here are some diagnostic contexts for a blasphemous noun N(devil) (note that not every N(devil)
can appear in all of these contexts):
1) N(devil) znaet ‘knows’, … (Čërt znaet, gde ona sejčas! ‘Devil knows where she [is] now!’)
2) Na koj ‘For what’ N(devil) Х Y-u? (Na koj ljad èto mne? ‘For what hell this [is needed] to.me?’)
3) Na ‘For’ N(devil)-á Y-u Х? (Na xrená koze bajan? ‘For [what] horseradish to.a goat
an.accordeon?’)
4) Kakogo ‘Of.what’ N(devil)-а …?’  (Kakogo čërta on prišël? ‘Of.what devil [= What for] did he
come?’)
5) Ná ‘To’ N(devil) Y! (Ná fig ètu drjanʹ! ‘To fig.sign [= Down with] this shit!’)
6) do ‘up.to’ N(devil)-á Х-а (Piva/Deneg u nix do figá ‘Of.beer/Of.money at them [is] up.to
fig.sign [= very much]’.)
7) Ni ‘No’ N(devil)-á (Ni figá emu ne sdelaetsja ‘No fig.sign [= Nothing] to.him will.
happen’.)
8) N(devil) Х-u! (Šiš/Xren emu! ‘A.fig.sign/Horseradish [= Nothing] to.him!’)
9) N(devil) s ‘with’ Х-om (Bog/Allax s nim… ‘God/Allah with him…’ ≈ ‘I don’t care
about him’.)
10) K ‘To’ N(devil)-u Y! (K čërtu vse zaboty! ‘To devil [= Down with] all problems!’)
An expression of this form belongs to one of the following three families of linguistic units:
• Idioms, for instance, ˹do čertá˺ [Х-a/-ov] ‘very much Х/very many Xs’, ˹na čertá˺ [Y-uDAT ХNOM]/
˹na koj čërt˺ [Y-uDAT ХNOM] ‘Y does not need X at all’ and ˹na fig˺/˹k čërtu˺ [Y-аACC]!˺ ≈ ‘I wish Y to
cease to be in my personal domain’ [a very rude rejection]. For a detailed description of these idioms,
see Iomdin 2010b: 174–190.
• Collocations, for instance, A čërt eë znaet, gde Maša živët ‘But devil her knows where Masha
lives’. = ‘I don’t know where Masha lives, and I don’t want to know’ [the utterance expresses a strongly
negative attitude of the Speaker toward Masha].

• Lexemes (namely, clausatives), for instance, interjections such as čërt [s Y-om]! ‘Devil with
Y!’ ≈ ‘I don’t care what happens with Y, who annoys me’, fig [Y-uDAT]! ‘Fig.sign to Y!’ ≈ ‘Y won’t
get anything!’, etc.
312 9 Pronominal idioms with a blasphemous noun in Russian

In a Ξ+К-expression the blasphemous noun N(devil) appears always in the sin-

gular as the syntactic subject of the verb znatʹ ‘know’.
The Russian Ξ+К expressions are of two subtypes according to their syntactic
structure: either the collocation Ξ depends directly on the К-word, forming with
it an idiom (˹Ξ←К˺; 2.1.1.1), or the К-word indirectly depends on (the head of) the
collocation Ξ (Ξ–⋯→К; 2.1.1.2).
NВ Russian has other similar Ξ+К expressions, for instance: odnomu bogu izvestno kto/kogda/
skolʹko/... ‘to.only God is.known who/when/how.many/…’; jasnoe delo kto/kogda/skolʹko/...
‘obvious thing who/when/how.many/…’; malo (li) kto ... ‘few (whether) who …’; etc. These
expressions will not be con­sidered in order not to encumber the presentation, since whatever is
said below applies, mutatis mutan­dis, to them as well.

9.2.1.1.1 ˹ Ξ←К˺-expressions: idioms with a blasphemous noun

(Tam živët ˹čërt znaet kto˺ ‘There lives devil knows who’)
A kogda nočʹju svetit mesjac,
To on svetit čërt znaet kak…
‘And when at.night shines the.moon,
Then it shines devil knows how…’
S. Esenin

˹Ξ←К˺-expressions can be illustrated by the sentences in (1):

(1) a. [Sjuda prixodit] ˹čërt znaet kto˺

‘To.here comes devil knows who’.
b. [On otpravilsja tuda] ˹bog znaet začem˺
‘He went there God knows for.what’.
c. ˹Xren znaet skolʹko˺ [on spal]
‘Horseradish knows [for] how.long he slept’.
d. ˹Pës znaet kuda˺ [my idëm ]
‘Dog knows to.where [we are.walking]’.
e. [Do goroda my dojdëm] ˹fig znaet kogda˺
‘To city we will.arrive fig.sign knows when’.

The ˹Ξ←К˺-expression in each of (1) sentences is an idiom—a non-compositional

multiword lexical unit that needs to be entered in a lexicon as a whole (for a defi-
nition of idiom, see Section 9.3; idioms are indicated by top corners: ˹…˺).
Semantically, all idioms in (1) are quite similar—up to the degree of expressiv-
ity and the language register: they express negative evaluation of the situation by
 9.2 К-expressions: their surface-syntactic structure and ­lexicographic description 313

the author of the utterance.3 But they are not synonymous: each one of these
˹Ξ←К˺ idioms has its own meaning. However, since what interests us here is the
syntactic structure of these idioms (rather than their meaning), I limit myself to a
lexicographic description of only three of these.

˹čërt znaet kto˺ ‘devil knows who’, colloquial, idiom, N, pronominal

˹čërt znaet kto˺: ‘person whom the author of the utterance evaluates very negatively’
NB In the surface realization of the construction PREP→˹čërt znaet kto˺ the preposition
is placed before kto or (less frequently) before čërt znaet: On pišet čërt znaet dlja kogo ‘He
writes devil knows for whom’. ~ On pišet dlja čërt znaet kogo ‘He writes for devil knows whom’.

Na ètom sajte čërt znaet kto registriruetsja ‘On this site devil knows who
is.registered’. | Opasenija bolʹnogo popastʹ v ruki čërt znaet k komu ne
bezosnovatelʹny ‘The patient’s apprehensions to.find.himself in the.hands of
devil knows who [are] not groundless’. | Ja obedal čërt znaet s kem vo frake ‘I
was.dining devil knows with whom in tails’ [I. Brodskij]. | On izvestnyj pevec,
a Džastin Timberlejk prosto čërt znaet kto ‘He [is] a well.known singer, but
Justin Timberlake [is] simply devil knows who’.
˹čërt znaet gde˺ ‘devil knows where’, colloquial, idiom, ADV, pronominal
˹čërt znaet gde˺: ‘in a very faraway or wrong place’
… esli by ne odin strannyj čelovek, živšij čërt znaet gde, a točnee – v gorode
Ure Xaldej­skom ‘… if it weren’t for one strange person, who lived devil knows
where, and more.pre­cisely—in the.city of Ur of.Chaldees’ [I. Guberman & А. Okunʹ]. |
No čto delatʹ, esli ty sam živëšʹ čërt znaet gde? ‘But what to.do if you yourself
live devil knows where?’
˹čërt znaet skolʹko˺ ‘devil knows how.much/many’, colloquial, idiom, ADV, pro-
nominal, (+ NX-GEN)
˹čërt znaet skolʹko˺: ‘a very large quantity [of Ξ]’
Kopajusʹ čërt znaet skolʹko vremeni i nikak ne mogu najti ošibku ‘I am.looking.
into.this [for] devil knows how.much time and still [I] cannot find [the] error’. |
Vy uže čërt znaet skolʹko drug s drugom znakomy, začem tratitʹ vremja na prelju-
dii? ‘You already devil knows how.much.time with each other [are] acquainted,
why waste time on preludes?’ | Za period trenirovok ja probežal čërt znaet

3 The author of the utterance U is either the Speaker or a person to whom the Speaker ascribes U.
Consider the utterance U1 = Puškin writes that there is no happiness in the world; U1’s author is me,
Igor Mel’čuk, that is, the Speaker. But the utterance U2 = there is no happiness in the world was
produced by Puškin, U2’s author. In other words, the Speaker is the author of the primary utter-
ance. (On the contrast “the Speaker vs. author of the utterance,” see, in particular, Iordanskaja
& Mel’čuk 1995.)
314 9 Pronominal idioms with a blasphemous noun in Russian

skolʹko kilometrov krossov ‘During the training period I have.run devil knows
how.many kilometers of.cross-country races’. | Da ja tam uže čërt znaet skolʹko
raz byval! ‘But I there already devil knows how.many times have.been!’

All ˹Ξ←К˺ idioms can include an expletive personal pronoun of 3sg on ‘he/she/it’
in the accusative: Ona pišet čërt eë/ego←znaet čto ‘She writes devil her/it knows
what’; this phenom­enon is described below, in Subsection 9.2.1.1.2.
An ˹Ξ←К˺ idiom can contain only some of the blaspemous nouns: On živët
*kto4 ⟨*šiš, *bes, *ljad⟩ znaet gde. At the same time, some N(devil)s can alternate in
these idioms, with a more or less significant semantic and/or stylistic difference:
čërt/bog/xren/fig/... znaet gde ‘devil/God/ horseradish/fig.sign/… knows where’.
However, even in case where two such idioms are semant­ically and syntactically
fully equivalent, each one of them is a separate linguistic sign, that is, a separate
lexical unit.
The surface-syntactic structure of an ˹Ξ←К˺ idiom is obvious: its surface-
syntactic head is the К-word (an interrogative-relative pronoun L(pron, interr-rel)): the
phrase ˹čërt←znaet←kto˺ is syntactically similar to koe-←kto ≈ ‘someone’,
aby←kto ‘no.matter who’ or kto→‑nibudʹ ≈ ‘somebody’ (koe-kto, koe-gde; aby
kto, aby gde; kto-nibudʹ, gde-nibudʹ). In an ˹Ξ←К˺ idiom the pronoun К satisfies
all three Criteria B1 – B3, which determine the SSynt-head of a phrase (Chapter
2, Section 2.4).
For the convenience of the reader, here are these criteria.

Criteria B – SSynt-dominance
Criterion B1 – The passive SSynt-valence of a phrase
In the phrase L1–r–L2, the lexeme L1 is the Synt-governor of L2 if the following
condi­tion is satisfied:
The passive SSynt-valence of the L1–r–L2 phrase is determined by the passive
SSynt-valence of L1 to a greater extent than by that of L2; then we have L1–r→L2.
Criterion B2 – The morphological links between an element of a phrase and
its external context (in a language that has inflectional morphology)
If the phrase L1–r–L2 in which the passive SSynt-valence of its components
does not allow one to establish the SSynt-head is such that
—either L1 controls the inflection of lexemes external to the phrase
—or L1’s inflection is controlled by such lexemes,
then L1 is the SSynt-head of the phrase: L1–r→L2.

4 Cf. He lives who knows where, which is quite normal in English.

 9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 315

Criterion B3 – The semantic content of a phrase

If the phrase L1–r–L2 in which neither the passive SSynt-valence nor the
morphology allows one to establish the SSynt-governor means ‘a kind/an in-
stance of L1’ rather than ‘a kind/an instance of L2’, then L1 is the SSynt-head
of the phrase: L1–r→L2.

In conformity with these criteria, the SSynt-head of an idiom of the ˹čërt←znaet

←kto˺ type is the К-word:

– Syntactically, the К-word determines the passive syntactic valency of a ˹Ξ←К˺

idiom: the idiom appears in all surface-syntactic slots where the К-word can
appear; thus, pisatʹ komu ‘to.write to.whom’ ~ pisatʹ čërt znaet komu ‘to.write
devil knows to.whom’; exatʹ v kakoj gorod ‘to.go to what city’ ~ exatʹ čërt znaet
v kakoj gorod ‘to.go devil knows to what city’; poterjatʹ gde ‘to.lose where’ ~
poterjatʹ čërt znaet gde ‘to.lose devil knows where’).
NB Strictly speaking, Criterion B1 being satisfied is already sufficient to recognize the
К-word as the syntactic head of a ˹Ξ←К˺ idiom. But the two other criteria, B2 and B3,
buttress this result; and even if their indications are redundant, it seems interesting to con-
sider them, too.

– Morphologically, the К-word is the morphological contact point of a ˹Ξ←К˺

idiom: it is the К-word that is declined in conformity with the syntactic
context; thus, On est čërt znaet čtoACC ‘He eats devil knows what’. ~ On pišet
čërt znaet komuDAT ‘He writes devil knows to.whom’. ~ On zanimaetsja čërt
znaet čemINSTR ‘He is.busy devil knows with.what’.
– Semantically, the К-word expresses the generic notion in the meaning of the
whole ˹Ξ←К˺ idiom: kto ⇔ ‘a person’, gde ⇔ ‘in a place’, skolʹko ⇔ ‘a
quantity’, etc.

At the SSynt-level, (the head of) the Ξ component (Ξ = N(devil)←subjectival–

ZNATʹPRES), that is, the verb znatʹ, is subordinated to К-word by means of the WH-
pronominal SSyntRel (Chapter 2, Section 2.5, SSyntRel No. 75, p. 91): the Ξ com-
ponent is an incomplete clause depending on the К-word. For instance:

On est čërt znaet←WH-pronominal–čto ‘He eats devil knows what’.

At the deep-syntactic level, the ˹Ξ←К˺ idiom is represented, as any idiom, by one
node.
The technique of joining to a К-word a dependent collocation engenders in
Russian an open class of syntactically formed indefinite pronouns: the whole ˹Ξ←К˺
expression is syntactically equivalent to a “compound” indefinite pronoun, of
316 9 Pronominal idioms with a blasphemous noun in Russian

the same type as kto-nibudʹ ≈ ‘anybody’, kto by to ni byl(o) ‘whoever’, malo

kto ‘few who’ = ‘few people’, koe-kto ≈ ‘somebody’, aby kto ≈ ‘no matter who’.
(For a detailed description of this class of indefinites in Russian, see Testelets &
Bylinina 2005.)
This phenomenon is well known in many other languages. M. Haspelmath
(1997) indicates four major types of such pronouns in more than hundred struc-
turally very different languages:

1. ‘Dunno’ type: On rabotaet bog znaet gde

‘He works God knows where’.
2. ‘No matter’ type: On rabotaet dlja nas nevažno gde
‘He works for us irrelevant where’.
3. ‘Pleases’ type: On rabotaet gde xočet
‘He works where [he] wants’.
4. ‘Maybe’ type: On rabotaet gde emu predlagajut
‘He works where to.him [«they»] propose’.

Ξ+К-expressions, discussed in Subsection 9.2.1, are of the first and the second
type, and К+Ξ-expressions (Subsection 9.2.2), of the third and the fourth.

The DSynt- and SSynt-structures of a sentence containing an ˹Ξ←К˺ idiom are as

follows:

(2) а. On žil čërt znaet gde ‘He lived devil knows where’.
b. The DSyntS of (2а):
˹čërt_znaet_gde˺←II–žitʹPAST–I→on
c. The SSyntS of (2а):
čërtSG←subj–znatʹPRES←WH-pronominal–gde←circumstantial–žitʹPAST–subj→on

9.2.1.1.2 Ξ
 –⋯→К-expressions: collocations governing an indirect-interro­
gative clause (Čërt znaet, kto on takoj ‘Devil knows who he [is]’)
I kto ego znaet, čego on morgaet?
‘And who him knows why he is.blinking?’
[M. Isakovskij; a popular Soviet song].

Typical examples of Ξ–⋯→К-expressions are given in (3):

(3) a. Čërt znaet, kto [sjuda xodit]

‘Devil knows who to.here comes’.
 9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 317

b. Bog znaet, za čem [on otpravilsja]

‘God knows for what he went’.
c. Xren znaet, skolʹko [on spal]
‘Horseradish knows [for] how.much.time he slept’.
d. Pës znaet, kuda [nam idti]
‘Dog knows to.where we must.go’.
e. Šut znaet, kakogo [goda èta krepostʹ]
‘Buffoon knows of.which year [is] this fortress’.
f. Fig znaet, kogda [my dojdëm do goroda]
‘Fig.sign knows when we will.reach [the] town’.

Here, the component Ξ, as stated above, is a collocation of the verb znatʹ ‘know’,
which governs a completive clause that contains a К-word. The SSynt-structure of
this completive clause is quite normal: it depends, via its top Main Verb [MV], on
znatʹ (being its DirO), for instance:

(4) а. Čërt znaet, gde [on žil] ‘Devil knows where he was.living’.
b. DSyntS of (4а): čërtSG←I–znatʹPRES–II→žitʹPAST–I→on
II→gde
c. SSyntS of (4а): čërtSG←subj–znatʹPRES–dir-obj→ žitʹPAST–subj on
circum→gde

The collocation of the verb znatʹ with N(devil) is described by means of a non-stan-
dard lexical function «I don’t [know] ...», which means ‘the Speaker signals
that he does not know the answer to the question “Ψ?” about a situation P(Х, …,
Ψ), to which he is indifferent or which he evaluates negatively’ (for lexical func-
tions, see Chapter 1, Subsection 1.2.4.2, p. 23). Ψ stands for an indefin­ite pronoun
that means ‘somebody’, ‘something’, ‘somewhere’, etc., and Ψ? denotes the
corres­ponding interrogative pronoun: ‘who?’, ‘what?’, ‘where?’.
Let it be emphasized that the meaning of this type of collocation includes the
semanteme ‘the Speaker’ (i.e., ‘I’) and ‘signaling’ (in the communicative sense):
in this way is reflected the fact of “insubordinability,” or “non-quotability,” of
this collocation, see immediately below.5
An idiom of the ˹čërt znaet gde˺ type and a collocation of the čërt znaet
[, gde …] type, both containing a blasphemous noun, differ in several aspects.

5 On ‘signaling’ vs. ‘communicating’, see Mel’čuk 2001: 242–260.

318 9 Pronominal idioms with a blasphemous noun in Russian

– Semantically, a ˹Ξ←К˺ idiom expresses a negative attitude of the utterance’s

author toward the referent of the meaning ‘˹Ξ←К˺’ (5а); however, the corre-
sponding collocation simply states his ignorance of the information ‘К’ (plus
indifference or perhaps annoyance of the utterance’s author), cf. (5b):

(5) а. Ivan zivët čërt znaet gde – v Kryžopole!

‘Ivan lives devil knows where—in Kryzhopol!’
[The author knows where Ivan lives, but condemns his place of residence:
the town of Kryzhopol.]
b. – Gde zivët Ivan?
‘Where does Ivan live?’ – Čërt (ego) znaet ‘Devil (him/it) knows’.
[The author does not know where Ivan lives and does not want to know.]

– Communicatively, a ˹Ξ←К˺ idiom is a descriptive expression (although it is

expressive), while a Ξ—⋯→К expression is a signalative (Mel’čuk 2001: 354).
Therefore, only a ˹Ξ←К˺ idiom can be subordinated to the verb of a superor-
dinate clause, and a Ξ–⋯→К collocation, as is typic­al of signalatives, cannot:

(6) а. Ivan zajavil, čto Maša zanimaetsja čërt znaet čem

‘Ivan declared that Masha was.doing devil knows what’.
b. Ivan zajavil, čto *čërt znaet, čem Maša zanimaetsja
‘Ivan declared that devil knows what Masha was.doing’.

– Syntactically, a ˹Ξ←К˺ idiom includes the К-word; the string čërt znaet kto
‘devil knows who’ is a phrase. In a Ξ–⋯→К expression only the collocation
Ξ—that is, čërt znaet, …—is a phrase; the К-word (= kto ‘who’, čem ‘with.
what’, gde ‘where’, …) is part of the completive clause governed by the col-
location. Тhis К-word depends directly on the MV of the completive clause
via one of the …-obj or circum SSyntRels, and indirectly on the verb znat′,
the head of the Ξ collocation: čërt znaet,–⋯→kto. The completive clause
that contains the К-word is an indirect question and, therefore, allows for
the interrogative particle li ‘whether’ and the conjunction ili ‘or’, which is
excluded in the case of the idiom:

(7) а. Čërt znaet, priexala li ona ‘Devil knows whether she arrived’.
b. Čërt znaet, priexala ona ili net ‘Devil knows [whether] she arrived or not’.

With a Ξ collocation, the completive clause can be omitted altogether in an answer:

(8) [– A čem ona zanimaetsja? ‘And what [is] she doing?’]

– Čërt (eë) znaet (čem) ‘Devil (her) knows (what)’.
9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 319

– Prosodically, the idiom carries a continuous intonation contour, while the

collocation can have a pause after znatʹ.
– Lexically, the idiom allows for the replacement of čërt znaet by the expres-
sion čërt-te (where ‑te is possibly a reduction of tebe ‘to.you’, a Dativus
Ethicus; cf. polnote ≈ ‘enough, stop it’ ⇐ polno tebe ‘enough to.you’), but
the collocation does not (“ | ” stands for a minor pause):

(9) а. Ona zanimaetsja | čërt-te čem

‘She does devil-te what’. ~ Čërt-te čem | ona zanimaetsja.
b. *Čërt-te, | čem ona zanimaetsja.

Both phrasemes of the Ξ+К type, that is, the idiom and the collocation, can
include the ex­pletive personal pronoun of 3rd person on ‘he/she/it’ in the accusa-
tive, depending on znaet ‘knows’. Two constructions, semantically equivalent,
are possible:

– Either on is used non-referentially, in the neuter singular form ego ‘it’: (10a, b)
and (11a, b).
– Or on is coreferential to the SSynt-subject of the superordinate clause, which
contains the idiom, or to the SSynt-subject of the completive subordinate
clause, which depends on the collo­cations; in this case, on agrees with this
SyntSubj in gender and number: (10c, d) and (11c, d).

(10) а. Maša zanimaetsja čërt ego znaet čem

‘Masha does devil it knows what’.
b. Deti zanimajutsja čërt ego znaet čem
‘Children do devil it knows what’.
c. Maša zanimaetsja čërt eë znaet čem
‘Masha does devil her knows what’.
d. Deti zanimajutsja čërt ix znaet čem
‘Children do devil them knows what’.

(11) а. Čërt ego znaet, čem Maša zanimaetsja

‘Devil it knows what Masha does’.
b. Čërt ego znaet, čem deti zanimajutsja
‘Devil it knows what children do’.
c. Čërt eë znaet, čem Maša zanimaetsja
‘Devil her knows what Masha does’.
d. Čërt ix znaet, čem deti zanimajutsja
‘Devil them knows what children do’.
320 9 Pronominal idioms with a blasphemous noun in Russian

There are no semantic differences between (a, b) and (c, d) variants; the on
pronoun reinforces the colloquial character of both expressions, but does not
impact their meaning. Nevertheless, the idiom and the collocation differ in the
frequency of the expletive on: on is more typical of the collocation, where, in
some specific contexts, it becomes nearly obligatory, cf. (12):

(12) [– Čem že ona zanimaetsja? ‘What is she doing?’]

– A čërt ego/eë znaet! ‘But devil her/him knows!’ ⟨??A cërt znaet!⟩

The addition of the expletive onACC raises the question of the syntactic role of this
pronoun inside the idiom/collocation. Its accusative points at a DirO; but the col-
location under analysis has already a DirO: the completive clause subordinated
to znatʹ. Moreover, the expletive on ‘it’ carries no meaning and is optional. This
allows us to consider it as a quasi-direct object. A stock example of Quasi-DirO
is the noun in the Persian verb-noun collocations and idioms that are transitive
verbal expression—that is, as a rule, they have a genuine DirO:

(13) a. Reza bäčče-ra tänbih kärd

Reza child DirO punishment did
‘Reza punished the.child’.
b. Reza bäčče-ra färib dad
Reza child DirO deception gave
‘Reza deceived the.child’.
c. Reza bäčče-ra dust dašt
Reza child DirO friend got
‘Reza befriended the.child’.

The noun in boldface is the Quasi-DirO-2 of the verb. In the SSyntS, it is sub-
ordinated to the verb by the quasi-direct-objectival-2 SSyntRel, see Chapter 2,
Section 2.5, SSyntRel No. 9, p. 55.
Now, let us see a lexicographic description of collocations of the «I don’t
[know]» type in the lexical entry for the verb znatʹ ‘know’.
 9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 321

ZNATʹ ‘know’
···
'
«I don’t [know]» : N(devil)←subjectival–znatʹPRES (coll. –quasi-dir-objectival-2→on ‘he’) ( | )

Conditions
1) znatʹ ‘know’–direct-objectival→∆CLAUSE | L←subject–L2–⋯→L1(pronoun, interr-rel).
2) N(devil) = {bog ‘God’, coll. čërt ‘devil’, substandard fig ‘fig sign’, kto ‘who’, substandard
pës ‘dog’, substandard šut ‘buffoon’, vulg. xren ‘horseradish’}.
3) on ‘he’: either onSG, NEU, or [g, n](on) ⇐ [g, n](L); ONSG, NEU must be positioned immedi-
ately after N(devil).

☛ 1. ( | ) is an optional minor pause; ' represents a very strong stress.

2. 
∆CLAUSE stands for the SSynt-tree of a completive clause that depends on the verb znatʹ
‘know’ in conformity with the government pattern of this verb; this clause has the SSynt-
structure of the form: L←subject–L2–⋯→L1(pronoun, interr-rel).
3. g and n stand, respectively, for “gender” and “number.”

Russian also has other phraseologized Ξ+К-expressions, for instance, malo/

redko kto ‘few/rare [are] who’ [Malo kto èto čitaet ‘Few [are] who read this’]
or aby kto ≈ ‘no.matter who’ [Ja ne pojdu aby k komu ‘I will.not visit no.matter
whom’]. These are as well non-standard colloca­tions of some (by no means all)
pronouns L(pron, interr-rel). An example:

kto
···
«few» malo(antepos)←intraphrasemic–kto ‘few [are] who’,
redko(antepos)←intraphrasemic–kto ‘rare [are] who’
«no matter» aby(antepos)←intraphrasemic–kto ‘no.matter who’,
čërt-te(antepos)←intraphrasemic–kto ‘no.matter who’

The particles/adverbs malo, redko, aby and čërt-te are subordinated to the
interrogative-relative pronoun by the intraphrasemic SSyntRel, which links a
dependent component of a “quasi-compound word,” including, in particular, the
particle koe-, to its central component—the relative pronoun. (In the inventory
of Russian SSyntRels in Mel’čuk 1974: 221–235 this SSynt­Rel—№ 31, p. 231—was
called 1st auxiliary.)

9.2.1.2 Free Ξ←К-expressions

oblique-objectival
(14) а. Maša zanimaetsja ne pripomnju←WH-pronominal–čem
‘Masha does I.don’t remember what’.
b. Maša zanimaetsja nikto v Moskve ne znaet čem
‘Masha does nobody in Moscow doesn’t know what’.
322 9 Pronominal idioms with a blasphemous noun in Russian

c. Maša zanimaetsja vse v Moskve znajut čem

‘Masha does everybody in Moscow knows what’.
d. Maša zanimaetsja matʹ Petra daže predpoložitʹ ne možet čem
‘Masha does mother of.Peter even presume cannot what’.
е. Maša zanimaetsja ty nikogda ne ugadaešʹ čem
‘Masha does you never won’t guess what’.

This type of expressions in English was studied in depth in Lakoff 1974.6

Free Ξ←К-expressions do not contain a blasphemous noun, but, syntacti-
cally, they are similar to phraseologized Ξ+К-expressions, which contain one.
Namely, the SSyntS of a free Ξ+К-expression is the same as that of its phraseolo-
gized counterpart: the SSynt-head is an L(pron, interr-rel) (i.e. К-word), which subordi-
nates—by the WH-pronominal SSyntRel—the top node MV of the rest.
Let us consider in more detail the production of sentence (14е). Its (simpli-
fied) SemS, DSyntS and SSyntS are presented in Figure 9.1 [‘what?’ represents the
meaning of the interro­gative pronoun čto?]:

‘čto? = what?’ ‘ugadatʹ = guess’

a. ‘zanimatʹsja ≈ do’
2
1 2 1
1
2 1 ‘nikogda =
1 never’
‘Maša’ ‘čto-to ≈ ‘ty = you’
‘bytʹ = be’
something’ ‘ne = not’

b.
ZANIMATʹSJA ≈ ‘do’ c. ZANIMATʹSJA ≈ ‘do’

subj obl-obj
I II
ČTO? ‘what?’
ČTO? ‘what?’
MAŠA MAŠA
ATTR WH-pronominal

ATTR UGADATʹ ‘guess’ restr UGADATʹ ‘guess’

NE ‘not’ NE ‘not’
I II subj circum

ATTR
ČTO? ‘what?’ NIKOGDA ‘never’
TY ‘you’ TY ‘you’

NIKOGDA ‘never’

Figure 9.1 Semantic, deep-syntactic and surface-syntactic structures of sentence (14е)

6 Here is one of Lakoff’s examples (p. 322):

(i) John invited you’ll never guess how many people to you can imagine what kind of a party at it
should be obvious where with God only knows what purpose in mind, despite you can guess what
pressure.
9.2 К-expressions: their surface-syntactic structure and ­lexicographic description 323

The SemS in Figure 9.1(a) can be implemented more or less literally:

(15) а. Maša zanimaetsja čem-to, i ty nikogda ne ugadaešʹ, čtó èto

‘Masha does something, and you will never guess, what is this’.
b. Maša zanimaetsja čem-to, i ty nikogda ne ugadaešʹ, čem (imenno) (ona zani­­
maetsja)
‘Masha does something, and you will never guess, what (exactly) (she does)’.

The sentences (15a, b) are produced from the SemS in Figure 9.1(a) according to
standard Sem-, Synt- and Morph-rules. However, free Ξ+К-expressions, in par-
ticular, that in (14е), have quite specific deep- and surface-SyntSs; consequently,
a special Sem-rule—or, more precisely, a schema for an open set of Sem-rules—is
needed:

‘byt′ = be’
‘σ’ L(‘σ’)
i 1 2
R(i)
1 1 ‘Y?’ ⇔
‘Y’ L(‘Y?’)
2
ATTR

‘σ (‘sčitat′2 =
believe’)’ L(‘σ(‘sčitat′2’))’

Conditions

1) ‘Ψ’ : ‘čtо-to’ ‘gde-tо’ ‘kuda-to’ ‘kogda-to’ kak-to’ ‘začem-to’ ‘počemu-to’ ‘skol′ko-to’

‘something’ ‘somewhere’ ‘to.somewhere’ ‘some-time’ ‘somehow’ ‘for some purpose’ ‘for some reason’ ‘some.how.many’

‘Ψ?’: čtо? gde? kuda? kogda? kak? začem? počemu? skol′ko?

The inventory of LUs that can appear in this Sem-rule is only illustrative, that is, not exhaustive.
2) Sem-actant 1 of the predicate ‘σ̃(‘sčitat′2 = believe’) is:
‘ja = I’, ‘ty = youSG’, ‘nikto = nobody’, ‘vse = everybody’ or ‘even←N’.
3) The verb L(‘σ̃ (‘sčitat′2’)) is either in the imperfective present, or in the perfective future, or else in the
conditional-subjunctive.

Notations
L(‘σ’) stands for an LU L that has the meaning ‘σ’;
‘σ̃ (‘sčitatʹ2 = believe’)’ is a meaning whose central component is ‘sčitatʹ2 = believe’; for instance:
‘σ̃’ = ‘znatʹ = know’ ‘Х believes that Y holds, and this is true’;
‘σ̃’ = ‘dogadatʹsja = have guessed’ ‘caused1 by Х’s intuition, Х began to believe that Y holds, and this is true’;
‘σ̃’ = ‘predpolagatʹ = presume’ ‘caused1 by Х’s logical reasoning, Х believes that Y holds’.

Figure 9.2 Sem-rule ‘Ψ←1–be–2→’ ⇔ –ATTR→ (so to speak, a semantic ellipsis)

324 9 Pronominal idioms with a blasphemous noun in Russian

9.2.2 К+Ξ-expressions

The К+Ξ-expressions, just like thе Ξ+К-expressions, are also of two types: phrase-
ologized, in which Ξ is a lexeme or an idiom (9.2.2.1), and free (9.2.2.2).

9.2.2.1 Phraseologized К→Ξ-expressions: collocations (Maša poedet

kuda→ugodno ‘Masha will.go to.where wanted’)
Interrogative-relative pronouns form a series of collocations. Thus, čto ‘what’ has
two non-standard collocations of the К→Ξ-type, where čto is the base: in the one
the collocate is the idiom ˹bog poslal/pošlët˺ ‘God has.sent/will.send’, which
constitutes an incomplete clause depending on the pronoun); in the second the
collocate is the adverb ugodno lit. ‘pleasant’ ≈ ‘any’:

čto ‘what’
···
«what is available» : ˹bog poslal/pošlët˺(postposed)←intraphrasemic–čto
«anything» : ugodno(postposed)←intraphrasemic–čto

čto has some more non-standard collocations of this type, the collocates being
the idioms ˹bog na dušu položit˺ ‘God will.put on [your] soul’ ≈ ‘whatever’ or
˹pód ruku popadëtsja˺ ‘under.hand will.come’ ≈ ‘whatever’ and the adverbs
ni_popadja ≈ ‘-ever’ or popalo ≈ ‘-ever’. In the SSynt-structure these collocates
are subordinated to the К-word by the intraphrasemic SSyntRel.
Other interrogative-relative pronouns—kto ‘who’, kuda ‘to.where’, gde
‘where’, kogda ‘when’, začem ‘what.for’, etc.—also have these collocates, which
gives a few dozens of К→Ξ-collocations.

9.2.2.2 Free К←⋯– Ξ-expressions (Maša poedet kuda ty zaxočešʹ ‘Masha

will go to.where youSG will.want’)

Examples (16) and (17) present free К←⋯– Ξ-expressions.

subjectival
(16) а. Maša vsegda xorošo ispolʹzuet čto k nej popadët v ruki
‘Masha always uses well what to her falls into hands’.
b. Maša poedet kuda eë pošljut
‘Masha will.go to.where [«they»] her will.send’.
oblique-objectival
(17) а. Maša zanimalasʹ čem Vanja xotel[, čtoby ona zanimalasʹ]
‘Masha was.doing what Vanya wanted [that she were.doing]’.
9.2 К-expressions: their surface-syntactic structure and l­ exicographic description 325

b. Maša vstretitsja s kem vam budet ugodno[, čtoby ona vstretilasʹ]

‘Masha will.meet with whom to.you will.be pleasant [that she were.
meeting]’.
☛ T he struck-through fragment in (17a–b) is a part of the sentence that is present in the sur-
face-syntactic structure, but can be (and most often is) elided in the sentence itself.

The free К+Ξ-expressions in the sentences of (16) and (17) are pseudo-relative
clauses (the current name being headless/free relatives: see, for instance, Bresnan
& Grimshaw 1978; this type of clause is analyzed in Chapter 6, Section 6.3.2.2).
Such a clause can always be “reconstructed” to become a genuine relative clause
by adding to it the correlate pronoun tot/to ‘this’, tuda ‘to.there’, togda ‘then’,
etc.: kto javitsja k nej ‘who comes to her’ ≡ tot, kto javitsja k nej ‘the.person who
comes to her’ or kuda eë pošljut ‘to.where [«they»] her will.send’ ≡ tuda, kuda eë
pošljut ‘to.there to.where [«they»] her will.send’.
The SSynt-structure of a pseudo-relative clause must be the same as that of
a genuine relative clause and as that of its independent equivalent—that is, of an
interrogative clause. Namely, the К‑word, i.e. the L(pron, rel), of the pseudo-relative
clause is an object or a circumstantial of the MV of this clause, and the MV itself
depends, also as the subject, an object or as a circumstantial, on the MV of the
superordinate clause, see (16).
In a pseudo-relative clause, the L(pron, rel) pronoun plays, metaphorically speak-
ing, a double role: from the angle of its meaning it appears to be simultaneously
an element of the two clauses—that is, of both the pseudo-relative and its super-
ordinate. Syntactically, this is, of course, impossible; but the pressure from the
meaning is so high that, for instance, in Russian the surface (= morphological)
form of L(pron, rel) must satisfy the government pattern of the MVs of both these
clauses—otherwise, the result is ungrammatical. (In the sentences of the (17)
type, by “MV of the pseudo-relative clause” is meant the elided verb shown in
square brackets.) Let it be emphasized that just a formal coincidence of the signi-
fiers is required: thus, in (16а), the wordform čto ‘what’ corresponds to two dif-
ferent DMorphSs, i.e. to different cases—čtoACC (in the superordinate clause) and
čtoNOM (in the pseudo-relative).7

7 This constraint is known in many languages. Thus, in German the К‑word can “combine” two
different cases, but only if both these cases have identical signifiers (Dalrymple & Kaplan 2000:
759, (2) ff; I slightly modified the examples for better clarity):
(ii) a. Ich habe gewaschen was übrig blieb ‘I have washed what extra remained’.
vs. b. *Ich habe gewaschen wen/wer übrig blieb ‘I have washed whom/who extra remained’.
Both superordinate clauses in (ii) must have a К‑word in the accusative, and both correspond-
ing pseudo-relatives require a К‑word in the nominative. However, the pronoun was ‘what’ has
326 9 Pronominal idioms with a blasphemous noun in Russian

The SemS, the DSyntS and the SSyntS of sentence (17а) are presented, respec-
tively, in Figure 9.3:

а. ‘zanimatʹsja’ ≈ ‘do’ ‘Vanja = Vanya’

‘zanimatʹsja ≈ do’
1
1 2 2
2 1
1 1
‘Maša = Masha’ ‘čto-to = ‘xotetʹ = want’
something’

b. ZANIMATʹSJA ≈ ‘do’

I II
ČTO-TO ‘something’
MAŠA 1
‘Masha’
ATTR

XOTETʹ ‘want’

I II

1 ZANIMATʹSJA ≈ ‘do’
VANJA ‘Vanya’ I II

MAŠA ČTO-TO ‘something’

‘Masha’

identical forms of the accusative and the nominative, while the forms of the accusative and the
nominative of the pronoun wer ‘who’ are different; therefore, sentence (ii-а) is correct, and sen-
tence (ii-b) is not.
Spanish allows for the following construction (Suñer 1984):
(iii) Escribí a quien viste ayer ‘I.wrote to whom you.saw yesterday’.
In the superordinate clause, the preposition a introduces DSyntA III of the verb escribir ‘write’
(i.e. it corres­ponds to the dative), and in the subordinate clause it introduces its DSyntA II (i.e. it
corresponds to the accusative).
The question of the presence and the nature of constraints on government patterns of the MVs
of the superordinate and the subordinate clauses in the case of pseudo-relatives used to be very
popular in linguistics; see, for instance, Grosu 1988.
9.2 К-expressions: their surface-syntactic structure and ­lexicographic description 327

c. ZANIMATʹSJA ≈ ‘do’

subjectival oblique-objectival

ČTO ‘what’
MAŠA
‘Masha’
relative

XOTETʹ ‘want’

subjectival direct-objectival

ZANIMATʹSJA ≈ ‘do’

VANJA
‘Vanya’ subjectival oblique-objectival

MAŠA ‘Masha’ ČTO ‘what’

Figure 9.3 Semantic, deep-syntactic and surface-syntactic structures of sentence (17а).

☛ The struck-through lexemes in the SSyntS in Figure 9.3(c) represent the clause ele-
ments to be elided: they do not appear in the DMorphS of the sentence.

The semantic rule (SemS ⇔ DSyntS) that ensures the transition between the struc-
tures in (а) and (b) of Figure 9.3 looks as follows:

L(‘σ1’)

II

‘σ1’ ‘σ2’ ‘σ1’

2 L(‘Y?’)
2 2
⇔ ATTR

‘Y’ L(‘σ2’)
II

L(‘σ1’)

Figure 9.4 Sem-rule ‘–2→’⇔ –ATTR→

☛ ‘Ψ’ and ‘Ψ?’ have here the same meaning as in Sem-rule of Figure 9.2 above: ‘Ψ’ rep-
resents the meaning of an indefinite pronoun (‘čto-to ≈ something’), and ‘Ψ?’ is the
meaning of the corresponding interrogative pronoun (‘čto? = what?’).

The Sem-rule in Figure 9.4 describes the “translation” of Sem-relation 2 by DSynt-

relation ATTR in the given—quite specific!—context; this translation presupposes
328 9 Pronominal idioms with a blasphemous noun in Russian

head-switching: DSynt-dependencies are inverted with respect of the correspond-

ing Sem-dependencies.
The transition between the structures in (b) and (c) of Figure 9.3 is specified
by a DSynt-rule that includes the following condition on the resulting SSynt-tree:

form(GP[r](L1)) = form(GP[r](L2)), & L1–r→L(‘Ψ?’), L2–r→L(‘Ψ?’), L(‘Ψ?’) L(‘Ψ?’)

(“Line r of the government patterns of lexemes L1 and L2, which directly syntactically govern,
via SSyntRel r, coreferential occurrences of lexeme L(pron, rel), i.e. L(‘Ψ?’), specifies the identi-
cal inflectional forms of lexeme L(pron, interr-rel).”)

This conditions is the notorious “matching constraint” on both coreferential

occurrences of the К-word that appear in the DSynt-structure. Our formulation is,
of course, incomplete: we also need an inventory of all specific cases where the
given language allows for grammatical non-identity of the coreferential occur-
rences of the К-word under the condition of identity of their signifiers.

9.3 Blasphemous idioms are not syntactic phrasemes

Blasphemous idioms of the type ˹čërt znaet kto˺ ‘devil knows who’ are syn-
tactically quite particular, which has lead to them being often called “syntactic
phrasemes,” or “syntactic idioms.” I do not think that this is a felicitous usage,
and I will try here to put some order into the linguist’s notional system in what
concerns phrasemes (see Mel’čuk 2012-2015: vol. 3, 293–386).
Syntactic phrasemes are, of course, a subclass of phrasemes; it is natural to
see them as oppos­ed to lexemic and morphemic phrasemes. But first of all, what
is a phraseme tout court?

Definition 9.1 – phraseme

A phraseme is a complex linguistic sign (= a configuration of no less than two

linguistic signs) that is not free—that is, it is such that at least one sign in it
cannot be freely selected by the Speaker
1) according to this sign’s meaning and particular combinatorial properties,
2) following general rules of the language, and
3) independently of all other individual signs being part of the complex
sign under consideration.
NB In this definition, the term (linguistic) sign is used as a shorthand for ‘sign or appropri-
ate -emic set of signs’. Lexemes and morphemes making up phrasemes are not signs but
sets of signs; the term phra­seme can itself apply both to an actual sign ([They] kill two birds
with one stone, killed two birds with one stone, killing two birds with one stone, etc.) and to
a set of signs ˹kill two birds with one stone˺.
 9.3 Blasphemous idioms are not syntactic phrasemes 329

The best known type of phrasemes are lexemic phrasemes (for more, see Mel’čuk
2012d and 2015a).

Definition 9.2 – lexemic phraseme

A lexemic phraseme is a phraseme consisting of no less than two lexemes.

In other words, a lexemic phraseme consists of several (two or more) minimal

lexical units, i.e. lexemes.
There are three major classes of lexemic phrasemes.

– Non-compositional fully constrained phrasemes:

1) 
Idioms, such as ˹do čertá˺ lit. ‘up to devil’ = ‘a lot of’; ˹kot naplakal˺
lit. ‘tomcat cried’ = ‘very little of’; ˹datʹ po šapke˺ lit. ‘give on hat’ = ‘fire
[an employee]’.

– Compositional phrasemes:
A. Half constrained, where one component is selected freely according to its
meaning and corresponding linguistic rules, while the second is selected
as a function of the first one:
Collocations, such as zakljatyj vrag ‘sworn enemy’, nanesti udar ‘deal
2) 
a blow’, sdatʹ èkzamen ‘pass an exam’, where the collocate (boldfaced)
is selected as a function of the base.
B. Fully constrained:
3) 
Clichés, such as Menja zovut ... lit. ‘[«They»] call me …’ = ‘My name is
…’; Skolʹko vam let? lit. ‘How.many to.you years?’ = ‘How old are you?’;
Naskolʹko ja mogu suditʹ, … lit. ‘As.far.as I can judge, …’

The borderlines between these three classes of lexemic phrasemes are deter-
mined exclusively on the basis of their internal properties (compositionality and
constrained character) and are absolute: intermediate cases are logically impos-
sible.
Morphemic phrasemes are defined in an analogous way (Mel’čuk 1964 and
1992–2000: vol. 4, 398–403; Beck & Mel’čuk 2011).
330 9 Pronominal idioms with a blasphemous noun in Russian

Definition 9.3 – morphemic phraseme

A morphemic phraseme is a phraseme consisting of morphemes inside one

wordform—that is, either a phraseologized complex stem, or a phraseolo-
gized complex affix, or else a phraseologized combination of a stem with
an affix.

A morphemic phraseme also consists of minimal morphological units.

Morphemic phrasemes fall in the same three major classes as lexemic phrase-
mes:

1)Morphemic idioms:
kon+ëk lit. ‘horse+DIMINUTIVE’ = ‘hobby-horse’
[˹sest′ na svoego kon′ka˺ ‘get on one’s hobby-horse’]
na+kaz(-at′) lit. ‘[to] up+show’ = ‘[to] punish’
o+grom(-nyj) lit. ‘about+thunder(-ous)’ = ‘enormous’
2) Morphemic collocations (the collocates are boldfaced):
tul+jak lit. ‘Tul(-a)+-er’ = ‘inhabitant of Tula’
vs. moskv+ič lit. ‘Moskv(-a)+-er’ = ‘inhabitant of Moscow’
vs. kaluž+anin lit. ‘Kalug(-a)+-er’ = ‘inhabitant of Kaluga’
3) Morphemic cliché:
From pomogatʹ ‘[to] help’ pomošč+nik ‘helper’ ~ *pomog+atelʹ
From razbiratʹ ‘[to] dismantle’ razbor+k(a) ‘dismantling’ ~ *razbir+ani(e)

A syntactic phraseme must be defined in the same way—as a phraseme consist-

ing of minimal syntactic units. A minimal syntactic unit is a minimal syntactic
structure L1–r→L2, where L1 and L2 are lexemic variables, r being a surface-syn-
tactic relation [SSyntRel].

Definition 9.4 – syntactic phraseme (Mеl’čuk 1987)8

A syntactic phraseme is a phraseologized complex linguistic sign that con-

sists of at least two minim­al syntactic trees such that its signifier is non-
segmental, that is, contains prosody or a bound lexemic variable, e.g., L(X),
symbolizing the operation of duplication of the phraseme’s actant X.

8 In Mel’čuk 2012-2015: vol. 3, Ch. 16 syntactic phrasemes are called constructional phrasemes.
 9.3 Blasphemous idioms are not syntactic phrasemes 331

The signifier of a lexemic phraseme is segmental: a lexemic phraseme contains

only lexemes of the language, but no special prosody and no bound lexemic vari-
ables; the opposition “lexemic phrasemes ~ syntactic phrasemes” is obvious.
All syntactic phrasemes are idioms—i.e., they are non-compositional. Stock
examples of syn­tactic idioms are, for instance (like lexemic idioms, syntactic
idioms are also indicated by top corners ˹ ˺):

slon ˹(byl) kak slon˺

lit. ‘elephant (was) as elephant’ = ‘It is/was quite an ordinary elephant’.
nos ˹(byl) kak nos˺
lit. ‘nose (was) as nose’ = ‘It is/was quite an ordinary nose’.

vojna ˹(byla) kak vojna˺

lit. ‘war (was) as war’ = ‘It is/was quite an ordinary war’.

The semantic and deep-syntactic rules for this syntactic idiom are as follows:

‘quite’ ○–1→○–1→○ ‘X’ ⇔ X ○←I–○–II→○ ˹QUITE ORDINARY˺ ⇔

‘ordinary’ BYTʹ ‘be’
X ○←subjectival–○–copular-completive→○–comparative-conjunctional→○ LX
BYTʹ ‘be’ KAK ‘as’

☛ X stands for «a lexical unit that expresses the meaning ‘X’»; LX means «a duplicate of the
lexical unit X».

The meaning ‘quite ordinary’ is expressed by a whole SSynt-subtree and cannot

be expressed piecemeal.

Here are a few more examples of Russian syntactic idioms (expressions in square
brackets are actants of the idiom; the lexemic variable LX stands for a duplicate
of X).

[YDAT] ˹i [XNOM] bytʹ ne v LX-ACC˺

lit. ‘To.Y even X is not as X’ = ‘X is even unable to enjoy Y’
MašeY i prazdnikX ne v prazdnikX
lit. ‘To.Masha even feast [is] not as feast’. = ‘Masha even cannot enjoy the feast’.
˹[XNOM u YGEN do-Z-itsja]˺
lit. ‘X at Y will.up-Z’ = ‘X will suffer at the hands of Y because of X’s excessive Z-ing’
MašaX u negoY dokritikuetsjaZ!
lit. ≈ ‘Masha at him will.suffer.from.criticizing’. = ‘Masha will suffer at his hands
for criticizing too much’.
332 9 Pronominal idioms with a blasphemous noun in Russian

˹[ХNOM u YGEN ZFUT]˺

lit.
 ‘X at Y will.Z’ = ‘X will suffer at the hands of Y because of X’s Z-ing’
TyX u negoY poguljaešʹZ!
lit. ‘You at him will.go.for.a.walk!’ = ‘Don’t even try to go for a walk on him!’
[ХINF] ˹tak LX-INF˺!
lit.
 ‘X so X!’ = ‘If we must X, let’s do X intensely!’
SpatʹX tak spatʹX!
lit. ‘Sleep so sleep!’ = ‘If I must sleep now, let me do it with gusto!’

[ХNOM] ˹estʹ LX-NOM˺

lit. ‘X is X’
ŽenščinaX estʹ ženščinaX
lit. ‘A woman is a woman’. = ‘The person in question is a woman and behaves
as a typical woman does’. (On reduplicative syntactic idioms in several
languag­es, see Wierzbicka 1991: 391–452.)

Auxiliary, or structural, lexemes that appear in syntactic phrasemes are the

copula bytʹ ‘be’, the particles i ≈ ‘also’ and ne ‘no’, the indefinite adverb (unac-
cented) tăk ‘so’, the relative pro­noun čto ‘what’, prepositions and conjunctions.

The expressions ˹kakogo čërta?˺, ˹na koj čërt?˺, ˹čërta s dva!˺ etc. should not
be called syntactic phrasemes: they are quite regular lexemic idioms, since they
include at least two specific lexemes. Neither are syntactic phrasemes the expres-
sions of the type čërt znaet[, čto/gde/začem...], discussed above, in 9.2.1.1.2:
these are normal non-standard lexemic collocations, since the verb znat′ ‘know’
retains in them its full meaning. And, finally, the expressions of the form Čërt s
X‑om lit. ‘Devil with X’ = ‘I don’t care at all about X’ are not phrasemes at all: they
contain one of the lexemes of the vocable čërt—an interjection (a clausative)
with its governed complem­ent [s ‘with’ Ninstr] (such as Doloj terroristov! ‘Down
with terrorists’ or T ′fu na tebja! ‘I spit on you!’).

A reworked English version of Mel’čuk 2012d.

Part IV: W
 ord Order – Linearizing Dependency
Structures
10 Word order in Russian
10.1 The problem stated
10.2 The input and output for linearization rules
10.3 Linearization rules (illustrated with Russian data)
10.3.1 Three major types of linearization rules
10.3.2 Linearization rules for Russian
10.3.2.1 Type I: Syntactic linearization (+ morphologization) rules
10.3.2.2 Type II: Adjusting linearization rules
10.3.2.3 Type III: Filtering linearization rules
10.4 Conclusions
Appendix: Communicative differences that determine the six word arrangements in Motto 2

To Jura Apresjan, with admiration.

1. Ju. Apresjan – kak mnogo v ètom zvuke!
lit. ‘Ju. Apresjan, how much there is in this sound!’
2. ‘I love you!’ in Russian:
Ja tebja ljublju!
Ja ljublju tebja!
Ljublju ja tebja!
Ljublju tebja ja!
Tebja ja ljublju!
Tebja ljublju ja!

10.1 The problem stated

The words in a sentence necessarily follow each other in a particular order—

speech has a strictly linear character, which is physiologically determined. But
meaning expressed by a sentence is not organized linearly. Therefore, at some
point in the process of sentence production, the Speaker (or a model of the Speak-
er’s language—for instance, an automatic device) has to line­arize the lexical units
that are selected to construct the sentence. Therefore, linearization of a structure
built from lexical units is actually the operation to examine while discussing
word order. At least since Tesnière 1959: 17–20, linearization is recognized as one
of the main linguistic operations, cross-linguistically universal: the expression of
non-linear meaning by linear sentences.
The description of linearization in language L can be broken down into three
steps:

1) Define the input and output representations, i.e., specify the two sets of
structures: the structures of the first set (these are non-linear, i.e. arborescent,

https://doi.org/10.1515/9783110694765-011
336 10 Word order in Russian

input structures) must be processed by linearization rules, and the structures

of the other set (these are linear output structures) must be arrived at.
2) Define the relevant linguistic factors, i.e., determine the types of linguistic
phenomena of L that affect linearization and have to be accounted for in lin-
earization rules.
3) Define the set of linearization rules such that any valid input is properly
matched by them to some valid output.

The present chapter addresses only two of these steps: it describes (partially)
the input and output representations needed for linearization in Russian and
sketches out the major classes of linearization rules, their form and their interac-
tion. An in-depth discussion of linguistic factors affecting word order in Russian
is left out, although such factors are used in the rules proposed.
Due to its importance in the process of speech and its universality, linear-
ization occupies a place of honor in linguistics. One certainly cannot complain
about the scarcity of publications dealing with word order in the most diverse
languages; if anything, they are too numerous to be reviewed.

Apology No. 1 Since even a short list of selected references would be impossibly long, I abstain
from giving any general references concerning word order as such or word order
in Russian in particular.

In spite of this overwhelming wealth of texts on word order, there are, to the best
of my knowledge, no word order studies in which the input syntactic structure
is defined in terms of dependencies and the linearization rules are formally pre-
sented. Only two exceptions are known to me:

– My own sketch of linearization rules for Russian (Mel’čuk 1965 and 1974:
260–290; see also the book Mel’čuk & Pertsov 1987, which presents local rules
for word order in English), based on a dependency syntactic structure and
the step-by-step strategy of linearization (see Section 10.3). Published more
than 50 years ago, this proposal produced no echo in general or Russian lin-
guistics; therefore, it seems permissible to take these rules up, using newer
knowledge and newer skills, and present them in an improved form. That is
what will be done in this chapter.1
– The work of Kim Gerdes, Sylvain Kahane and their collaborators on word
order (Gerdes & Kahane 2001, 2004, 2007; Gerdes 2002; El-Kassas & Kahane

1 Within the framework of a pilot project on the automatic rephrasing the claim sentences in
patents, S. Mille and L. Wanner (2008) developed some linearization rules for English (drawing
upon the rules implemented by F. Lareau in the scope of the MARQUIS Project).
 10.1 The problem stated 337

2004; Bohnet 2007). It is also based on a dependency syntactic structure but

uses an essentially different technique of linearization: the strategy of pre-
defined full-sentence pattern (Subsection 10.3.1, p. 343), also known as “topo-
logical model.” Therefore, I cannot build directly on their findings and will
limit myself to this brief remark.
NB There is, however, a special case that needs to be mentioned. The last decade has
seen a series of computer programs that are “trained,” by using statistical methods, to
linearize dependency structures: the program is given a set of dependency structures, each
with a corresponding sentence, and by trial and error method it finds the patterns of cor-
respondences—the typical linearization for a given configuration of dependencies. A good
example of such a program for Chinese (a language where linear order is almost a unique
means for expressing the dependencies in the sentence) is found in He et al. 2009. But
these programs, no matter how well they carry out the linearization, do not give us the lin-
guistic rules describing word order, while it is exactly such rules that interest us here. See
also Filippova & Strube 2009.

The description of word order in this chapter proceeds in the Meaning-to-Text,

i.e. linguistic synthesis, direction and appears as a set of rules for linearization of a
starting non-linearized dependency structure.

3 Although the dominant philosophy in linguistics has been and still is to analyze

texts and represent the results of the analysis, word order has always been
studied by linguists in the synthesis direction.

Russian is chosen as the object language of the description not only because it is
my mother tongue, but also because it is ideal as the target of a word order study.
There are at least two reasons for this.
On the one hand, word order in Russian is extremely flexible in two senses.

– First, almost every permutation of words—or, more precisely, of “saturated”

phrases—in a Russian sentence is grammatical, see Motto 2. However, these
different word arrangements are not optional or arbitrary: they are controlled
by subtle, but strict communicative conditions, so that actually they are not
anarchically free, but well regulated (see Appendix, p. 366).
– Second, the syntactic structure of a Russian sentence is highly independent
of its commu­nicative structure: in Russian almost any element of the clause
can play almost any communicative role. Because of that Russian is rich in
variegated and complex word order phenomena. In sharp contrast, English
requires, for instance, that, as a general rule, the Synt-Theme of a sentence
be expressed by the Synt-Subject; to achieve this, English often has recourse
to the passive:
338 10 Word order in Russian

Rus. StatʹjuACC-Synt-T, DirO popravil LeoNOM lit. ‘The articleACC has.corrected LeoNOM’.
vs. Eng. The paperSynt-T, Subj was corrected by Leo.

On the other hand, word order in Russian is well studied (although rather infor-
mally), enabling the researcher to draw data from many sources; unfortunately,
only few main references can be mentioned here: Sirotinina 1965, Kovtunova
1976, 1980, and Yokoyama 1985, 1986: 171ff.

Apology No. 2 The description of linearization rules, no matter how sketchy and approximate,
requires a huge number of concepts and formalisms from surface syntax. To
explain all these would amount to writing a thick volume. Therefore, I use in this
chapter what I need without warning; I ask my readers for forgiveness and hope
that examples and minimal explanations will prove sufficient.

Apology No. 3 Linearization is intimately linked to prosodization of the deep-morphological

structure obtained: the word groups that undergo linearization must feature an
appropriate prosody; in fact they do not exist without this prosody (see Yokoyama
1985, Gerdes & Kahane 2007 and Zimmerling 2008 for well-justified insistence
on this relationship). However, in order to simplify my task, I omit everything
concerning prosodization.

The main goal of the present chapter is to outline, in a very rough way, the lin-
earization rules for a natural language, stated on the basis of Russian data,
but in a relatively general form.

In Russian, as well as in other languages where inflectional morphology is used

along word order to express syntactic dependencies, linearization rules are inti-
mately related to morphologization. (In syntax, morphologization is an operation
that computes the syntactic, or surface, grammemes, or syntactically-conditioned
morphological values, such as the grammatical case of the noun, the person and
number of the finite verb, the gender, number and the case of the adjective—based
on the information contained in the surface-syntactic structure.) Although I am
interested in linearization only, it is convenient to include in the picture relevant
data on morphologization, since this does not require special effort.
Formally speaking, linearization rules proper should be kept distinct from
morphologization rules. However, two factors interfere with such an approach, a
theoretical and a practical one.

– Theoretically, in many cases morphologization is inextricably intertwined

with linearization: thus, languages (for instance, Arabic) often have different
type of agreement of the Main Verb with the SyntSubj depending on the linear
position of the latter with respect to the verb. This is quite understandable:
 10.2 The input to and output of linearization rules 339

morphological markers constitute, together with linear arrangement of words,

one complex signifier for a surface-syntactic relation in a given context.
— Practically, mixed linearization + morphologization rules are more familiar
for a not-too-form­ally minded reader and, therefore, easier to grasp. In what
follows word order rules are presented in this mixed form, together with
the indication of corresponding syntactic grammemes they impose on the
members of a SSynt-dependency relation.

The remainder of the chapter is naturally divided into three sections: Section 10.2
describes the input and output for word order rules, while Section 10.3 presents a
fragment of surface-syntactic rules—linearization + morphologization rules—for
Russian; Section 10.4 contains some conclusions.

10.2 The input to and output of linearization rules

Two linguistic entities are generally assumed to be the main sources of informa-
tion that determines the linear arrangement of words in sentences: the syntactic
structure [SyntS] and the syntactic-communicative structure [Synt-CommS] of the
sentence to be produced. Here, the two are considered to be the necessary and
sufficient input to the linearization rules. They will be introduced in very general
terms as postulates, without detailed explanations or justifications.
The rules for linearization of the lexemes of sentence S, or linearization rules,
have as their in­put two structures of S’s surface-syntactic representation: the
surface-syntactic structure [SSyntS] and the surface-syntactic-communicative
structure [SSynt-CommS]. They produce, as their output, the deep-morphological
structure [DMorphS] of S, which, generally speaking, must be supplied with deep-
morphological prosodic structure [DMorph-ProsS]; however, in conformity with
the convention adopted (see Apology No. 3 above), the latter is not considered.

The input to linearization rules

Definition 10.1 – the surface-syntactic structure of a sentence

The SSyntS of sentence S is an unordered dependency tree where each

lexeme of S is represented by a node (of which this lexeme is the label) and
whose branches represent language-specific surface-syntactic relations
[SSyntRels], which link these lexemes; the names of the SSyntRels are labels
on the branches.

(On SSyntRels, see Chapter 2.)

340 10 Word order in Russian

Definition 10.2 – the surface-syntactic-communicative structure of a sentence

The SSynt-CommS of sentence S is a division of S’s SSyntS into communica-

tive areas (= subtrees) such that each has its Comm-dominant node specified
and is labeled with a value of a Synt-Comm-opposition.

(On Comm-Dominance and Comm-oppositions, see Mel’čuk 2001.)

The SSynt-CommS (as well as the DSynt-CommS) uses fewer communicative
oppositions than the semantic-communicative structure [Sem-CommS], namely—
the following five:

1. SSynt-Thematicity
2. SSynt-Givenness (not relevant for linearization in article languages)
3. SSynt-Focalization (not relevant for linearization in languages with lexical
expression of Focalization)
4. SSynt-Perspective
5. SSynt-Emphasis

The remaining Sem-Comm-oppositions—those of Assertivity, Unitariness and

Locutionality—are fully transcoded at the syntactic level into lexical units, gram-
memes and syntactic constructions; they, so to speak, disappear from the scene.
Moreover, the Synt-Comm-oppositions are different with respect to their Sem-
Comm-sources (Mel’čuk 2001: 64-66). In this chapter, only the Synt-Comm-oppo-
sition of Thematicity is taken into account. This reduces, of course, the power of
the proposed description, which misses several word arrangements that are pos-
sible in Russian for the expression of Focalization, Perspective, and Emphasis; yet
this simplifies the presentation a great deal.

Output of linearization (+ morphologization) rules

Definition 10.3 – the deep-morphological structure of a sentence

The DMorphS of sentence S is the linear sequence of all S’s lexemes supplied
with all relevant grammemes.

As a basic example, let me consider Russian sentence (1), its SSyntS with the
superposed partial SSynt-CommS in Figure 10.1, and its DMorphS in Figure 10.2.

(1) Metodom gravitacionnoj razvedki byla otkryta neft′ v Kazaxstane

lit. ‘By.method of.gravitational exploration was discovered oil in Kazakhstan’.≈
‘The method of gravitational exploration led to the discovery of oil in Kazakh-
stan’.
 10.2 The input to and output of linearization rules 341

This sentence comes from an elementary physics manual, where the preceding
paragraph is dedicated to the characterization of gravitational technology in geo-
physical explorations; the sentence corresponds to the underlying question What
else about gravitational exploration? In the diagram, TSSynt stands for the surface-
syntactic Theme (= topic), and RSSynt, for the surface-syntactic Rheme (= comment).

BYT′ ‘be’IND, PAST

subjectival passive-analytical
OTKRYT′PERF, PART, PASS, PAST, SHORT
‘discover’
NEFT′SG
‘oil’ circumstantial circumstantial
METOD ‘method’SG, INSTR
VI ‘in’ TSSynt
objectal-adnominal
prepositional
modificative
KAZAXSTANSG RSSynt RAZVEDKASG
‘exploration’
GRAVITACIONNYJ
‘gravitational’

Figure 10.1 Partial surface-syntactic representation of sentence (1)

The corresponding deep-morphological structure of sentence (1) is shown in

Figure 10.2:

metodSG, INSTR gravitacionnyjSG, FEM, GEN razvedkaSG, GEN

byt′PAST, SG, FEM otkryt′PERF, PART, PASS, PAST, SHORT, SG, FEM
neft′SG, NOM v KazaXstanSG, LOC

Figure 10.2 Deep-morphological structure of sentence (1)

Word order rules—that is, linearization + morphologization rules, schematically

represented in (2),—are responsible for producing the linear arrangement of fully
inflected lexemes, that is, the DMorphS of the sentence (without prosodic orga-
nization, as stated above). Linearization and morphologization rules constitute
a submodule of the SSynt-module of the Meaning-Text model (this module also
includes rules for Prosodization, omitted here):

(2) SSyntS
SSynt-CommS  Morphologization
Linearization 
rules DMorphS

SSyntS and SSynt-CommS, which essentially determine linearization, are repre-

sented as input structures for word order rules.
342 10 Word order in Russian

There are, to be sure, other factors that affect linearization:

– Semantic factors: e.g., semantically different circumstantials may be posi-

tioned differently; the position of a circumstantial can depend on its seman-
tic scope; etc.
– Rhetorical factors: e.g., a particular arrangement may be highly colloquial or
poetic.
– Stylistic factors: e.g., longer word groups preferably follow shorter ones in
postposition to the common governor; some languages encourage the word
order that produces chaining dependencies rather than embedded ones:

a→b→c→d is preferred over a c→d b

– Lexical factors: e.g., the adverb enough is exceptional in that it is placed after
its adjectival governor; pronouns may be positioned with respect to their gov-
ernor differently than the corresponding non-pronominal lexemes.
– The clarity of the text produced: e.g., a particular word arrangement can be
chosen to avoid ambiguity or else to reduce the number of embedded depen-
dencies.

Word order rules must in principle account for all these factors.

10.3 Linearization rules (illustrated with Russian data)

Linearization rules are presented here in general form, but are illustrated with
specific Russian examples. The rules given are sufficient to carry out the transi-
tion from the SSyntS of Figure 10.1 to the DMorphS of Figure 10.2: they describe
word order in the Russian sentence (1).

Apology No. 4 My characterization of word order in Russian concerns but a small fragment of the
possible arrangements in written texts of a scientific-official nature—that is, the
most neutral word arrangements, whatever that means.

10.3.1 Three major types of linearization rules

The general architecture of the linearization submodule of the SSynt-module of a

Meaning-Text linguistic model depends on the researcher’s choice of the overall
linearization strategy; there are two basic possibilities.
The first linearization strategy is top-down: it uses a Predefined Full-Sentence
Pattern [PFSP]. First, the researcher constructs a general pattern (or patterns) of
 10.3 Linearization rules (illustrated with Russian data) 343

whole sentences in language L—a sequence of hierarchically embedded slots,

which represent linear positions of sentence elements. The sentence word-order
pattern is divided into fields, each of which is filled by boxes, themselves divided
into fields, and so on; a box is provided for a particular element of the sentence.
Under the PFSP-strategy, linearization rules compute for each element of the
starting dependency structure its place in the PFSP and the box it opens, with
appropriate fields for its dependents. In the course of this operation, all other
word-order factors, such as communicative structure, special lexical properties,
etc. are equally taken into account. Such an approach seems to be especially good
for German—see the above-mentioned work by Gerdes & Kahane and Bohnet.
The second linearization strategy is step-by-step, or bottom-up: linearization
rules compute, by stages, the mutual disposition of the elements of the starting
dependency SSyntS; they first build (≈ linearize) simple phrases, then unite them
in complete phrases, then build clauses out of complete phrases, and, finally,
unite clauses to produce the sentence. Here, again, two approaches are possible:
a one-stage or multiple-stage approach.

– One-stage approach would be to formulate the rules for each SSynt-relation,

supplying every rule with all the conditions necessary for a correct place-
ment of its dependent member. For instance, a rule for the direct object [DirO]
would say that a DirO follows the governing verb and precedes its other Synt-
actants, if 1) this DirO is not part of the Synt-Theme, 2) it is not part of Synt-
Rheme-focus, 3) it is not a pronoun, 4) it is not a very long (“heavy”) word
group followed by another shorter group, which is another Synt-actant of the
same verb, etc.

– The other approach, which I favor, stipulates that all such conditions are for-
mulated separately (from purely syntactic rules) in very general terms: Synt-
communicative linearization rules, Pronoun linearization rules, Word-group
heaviness-based linearization rules, etc. This is possible to do since these
rules are logically and linguistically (at least in Russian!) independent from
syntactic rules. Such a presentation allows for a more compact and better
organized set of rules, avoiding unnecessary repetitions of the same condi-
tions in several rules. But the price to be paid for this advantage is to separate
the linearization process into stages: first you do approximate linearization,
according only to general syntactic rules; then you reshuffle the preliminary
arrangement thus obtained, pressing into service all additional rules; finally,
you check the result for inadmissible sequences and reshuffle once more to
avoid these. It is this—step-by-step—strategy that is adopted here.
344 10 Word order in Russian

 Let me emphasize, lest confusion arises, that I do not mean here a real proce-
3
dure separated into three consecutive stages. I am talking only about a way
of representing things that better agrees with a linguist’s intuition. All the
constraints introduced in different blocks of rules below can in fact be applied
together. This problem, however, exceeds the limits of the present chapter.

In accordance with the above strategy, linearization—that is, roughly speaking,

the transition from an SSyntS to the corresponding DMorphS—is performed using
three major types of rules:

I. Syntactic linearization + morphologization rules, which are responsible for

Stage 1. Based exclusively on the SSyntS, they produce the preliminary “frame”
of the linearized sentence —an arrangement of wordforms that may be good
if it weren't for the impact of other factors, see immediately below. This is the
most neutral, unmarked word order, or the default case.
II. Adjusting linearization rules, which, at Stage 2, apply to the output of syn-
tactic linearization rules—that is, to the preliminary frame of the sentence—
and alter it to reflect all additional factors: requirements of the SSynt-Comm-
structure, obligatory extractions (WH-words) and inversions (Is he busy?),
pronominalization (especially cliticization, where it exists), reshuffling of
word groups according to their heaviness, etc. These rules lead to serious
modifications of the neutral word order.
III. Filtering linearization rules work at Stage 3 and apply to the output of Stage 2.
They constitute, in fact, a list of undesirable word sequences. Each “bad”
sequence is assigned a “fine,” which is an empirically established negative
number; the sentence-to-be receives a cumulative fine and is processed—
that is, undergoes possible permutations of word groups—in such a way as
to reduce to zero or at least to minimize the overall fine “slapped on” the
arrangement under consideration.

It is assumed that these three major types of linearization rules are sufficient for
the description of word order in many (if not most) languages.2
Although the types of linearization rules are presented in sequence, one type
after another, this is only a manner of speaking: these rules are not externally

2 Some languages feature additional factors perturbing neutral word order. However, I think
that any such additional difficulty can be treated as a new subtype of Adjusting rules. Thus,
for instance, languages featuring second-position clitics, such as Serbian/Croatian, need special
rules to position the clitic cluster—after all other elements of the sentence are linearly arranged
(Milićević 2009). These rules are part of our Type 2 rules.
 10.3 Linearization rules (illustrated with Russian data) 345

ordered—that is, there are no special indications concerning the order of their
application. Each rule is formulated in such a way that it effectively applies only
when this application produces a correct result. The linguistic model can try to
apply them in any order, in parallel and/or repeatedly; I assume that this will
never lead to incorrect linearization.

10.3.2 Linearization rules for Russian

Some illustrative linearization rules for Russian are presented below in three sub-
sections corresponding to the three above types.

10.3.2.1 Type I: Syntactic linearization + morphologization rules

Syntactic linearization rules fall into five groups: I.A – I.E.

I.A: SSyntS ⇔ DMorphS correspondences

These rules do two things:

– they establish the preliminary, or the most unmarked, arrangement of word-

forms in the sentence, according only to the SSyntRels that link them;
– they produce the necessary syntactic grammemes, i.e., they carry out mor-
phologization.

A rule stating an SSyntS ⇔ DMorphS correspondence has the general form

L1x–r→L2z ⇔ 1) L1xʹ +(…+) L2zʹ | C1

⇔ 2) L2zʹ +(…+) L1xʹ | C2

Here:
– The left-hand side contains a minimal SSynt-subtree with the SSyntRel r; L1
and L2 are lexemes, x and z being the sets of appropriate semantic (= deep)
grammemes.
– The right-hand side contains one or two possible strings made up of the same
lexemes, with xʹ and zʹ being x and z with addition of all necessary syntactic
grammemes.
– “+” indicates the linear sequence, while “…” shows a possible gap between
lexemes L1 and L2, that is, the presence of other lexemes separating L1 and L2
in the sentence; parentheses mark optionality.
– C is the set of conditions that are essentially constraints on L1 and L2; among
other things, they describe the context in which the particular subrule applies
346 10 Word order in Russian

and thus determine different linear arrangements of L1 and L2. C also may
include additional indications concerning the placement of L1 and L2 into par-
ticular positions in linearization patterns; see below.

Such a rule specifies the linear—and, when needed, morphological—expression

of the SSyntRel r. (Stating that the dependent element of r can be positioned both
after or before its governor also constitutes an indication of the linear expression
of r.)
From the viewpoint of linearization, there are three types of SSyntRels: local,
semi-local, and non-local.

1) Local SSyntRels control linearization within rigidly organized minimal word

groups [MWGs], such as a nominal minimal word group [MWGN], an adjectival
minimal word group [MWGA], etc.; see MWG linearizing patterns below, pp. 355–
356. An MWG represents a sequence of wordforms with a completely fixed order
(within the framework adopted in this chapter—neutral academic-register texts—
no element of a MWG can change position).
A local SSyntRel r has the following three properties:

– r specifies the only possible linear disposition for its members. (Within MWGs
there are no options for different word arrangements, and neither the Synt-
CommS nor other factors have significant impact here.)

– r is unique, or not repeatable: r’s governor can have only one immediate
r-dependent.

An important exception
The modificative SSyntRel allows for several parallel adjectival modifiers
depending on the same governor, as in the phrase
modif
modif
English non-significative phonological←modif–alternations.

Such adjectives have to be ordered in a unique position in the linearizing

pattern according to special rules such as those discussed in Vendler 1968
and Iordanskaja 2000.3

– The mutual disposition of all local co-dependents is fixed: they are assigned
pre-established positions in the corresponding linearizing pattern.

3 Rules that are referred to here as special do not fall into any of major rule types introduced in
the present chapter; they are really special. Special rules are organized in blocks; such a block is
attached to one or several “normal” rules.
 10.3 Linearization rules (illustrated with Russian data) 347

Local SSyntRels include cross-linguistically the prepositional SSyntRel (PREP→N),

the determinative SSyntRel (DET←N), the quantitative SSyntRel (NUM←N), the modi­
ficative SSyntRel (A←N), the coordinative-conjunctional SSyntRel (CONJ(coord)→Ψ),
etc. One important local SSyntRel is coordinative (Y→X): its dependent always
follows the governor, as in

John–coord→and Mary or
attacked,–coord→advanced–coord→and captured, etc.

An MWG corresponds to a “very compact” phrase, which practically cannot be

torn apart by any factors and is moved around as a whole.

2) Semi-local SSyntRels control linearization of MWGs within complete word

groups [CWGs], that is, linearly ordered word sequences that roughly correspond
to complete clause elements: the subject CWG, the direct-object CWG, the dura-
tion-circumstantial CWG, etc. Within CWGs, word order is more flexible than
within MWGs, yet it still remains rather constrained: the co-dependent MWGs can
be arranged differently among themselves, but all of them are allowed to occupy
only one position with respect to their governor: all of them either precede the
governor or all of them follow it. Note that:

– the dependencies between MWGs are, of course, those between their heads;
– semi-local SSyntRels also control the linear disposition of clauses within the
sentence.

The properties of a semi-local SSyntRel r are:

– r also specifies just one linear disposition for its members.

– r is also near-unique: r’s governor can have, in most cases, only one immedi-
ate r-dependent.
– The mutual disposition of semi-local co-dependents is not fixed, and special
rules are needed to compute the arrangement into one CWG of several MWGs
that “semi-locally” depend in parallel on the same governor.

Semi-local SSyntRels include several adnominal-completive/attributive (N1→N2-gen)

SSyntRels, the relative and descriptive-relative SSyntRels (N→CLAUSErel), the cir-
cumstantial SSyntRel (N→PREP), etc.; see Chapter 2, Section 2.5, Nos. 48–56, Nos.
73–74 and No. 33. (Note that with a verb the circumstantial SSyntRel is non-local,
see immediately below.)
A CWG represents a phrase that traditionally corresponds to an element of
the clause: the subject, the direct object, a circumstantial, etc. Such a phrase is
also relatively compact, but less so than a phrase corresponding to an MWG: it
348 10 Word order in Russian

can be cut in two parts that exchange their linear positions following the require-
ments of the Synt-CommS.

3) Non-local, or global, SSyntRels, responsible for the mutual arrangement of

CWGs within a clause, link the top node of the clause SSyntS, that is, the finite
verb (in Standard Average European type languages), to its immediate depen-
dents (i.e., actants and circumstantials). A non-local SSyntRel r is opposed to
local and semi-local SSyntRels:

– A non-local r normally does not specify a unique order of its members. Even
in languages with a rather rigid word order various inversions and/or permu-
tations are possible between the Main Verb and the SyntSubj, the Main Verb
and the DirO, etc.
– A non-local r is not necessarily unique: r’s governor can have several imme-
diate r‑depend­ents (several oblique objects or several circumstantials).
– The mutual disposition of non-local dependents is not fixed; it depends on
numerous, very complex and sometimes even contradictory factors.

An SSyntRel can simultaneously be of more than one type as a function of its

governor. Thus, all circumstantial SSyntRels are non-local if their governor is a
finite verb, but semi-local otherwise.

In accordance with the three types of SSyntRels, four further groups of lineariza-
tion rules are needed:

I.B—for local SSyntRels (= for the construction of MWGs by means of linearizing

patterns);
I.C—for semi-local SSyntRels (= for the construction of CWGs);
I.D—for non-local SSyntRels (= for the construction of clauses);
I.E—for arranging the clauses within the sentence.

Remark
The moment seems ripe for a short theoretical digression: dependency vs. constituency (see,
e.g., Mel’čuk 2009a: 89–95). As is well known, in modern linguistic literature the dependency
description of the syntactic structure of a sentence is opposed to the phrase-structure, or con-
stituent, description. The dependency approach categorically rejects constituents as a means for
representing the syntactic structure of a sentence. This is due to the fact that a syntactic struc-
ture written in terms of constituents combines—or, if we do not mince our words, confuses—two
very different relations between lexical units: syntactic dependency (governor ~ dependent)
and linear order (precedes ~ follows), a distinction that cannot be ignored following Tesnière
1959. Linear order is the most important means that natural languages use for expressing (=
marking) syntactic relations. In languages without inflectional morphology (like Chinese or Viet-
namese) it is the only means (plus, of course, prosody); in languages like English it is the central
 10.3 Linearization rules (illustrated with Russian data) 349

and most exploited means; and even in Russian, where on many occasions the word order seems
irrelevant (cf. Motto 2), it still plays a leading role: for a number of constructions it remains the
basic marker of syntactic relations. Therefore, constituency logically cannot be used as a formal
means to represent the syntactic structure of a sentence. It is not necessary to have long discus-
sions about two ways of representing syntactic structure: constituency in syntactic structures is
a logical absurdity. However, constituents, or phrases, do of course exist in language and have
to be modeled in any linguistic description. But their legitimate place is 1) in the DMorphS of the
sentence, where they appear as prosodic phrases, and 2) in the linearization rules, where they
serve as building blocks in the process of the linearization of the SSyntS. These two types of
constituents do not stand in one-to-one correspondence with each other. The above-mentioned
MWGs and CWGs are nothing but constituents of the second kind; each represents a projec-
tion of the corresponding subtree. These constituents change during linearization: they can be
united, cut in two, have their parts moved around separately, etc., and finally they emerge as
phonological phrases, or constituents of the first kind (which can be different from projections of
the corresponding subtrees). And now, back to linearization rules.

Since the SyntS of sentence (1) contains 7 different SSyntRels, at least 7 SSynt-
rules for Russian must be presented. The following symbols and conventions are
used:
– AGREEV(N) (“Verb-Noun agreement operator”) is a set of rules that describe
agreement of the Main Verb [MV] with the noun that in most cases is its
subject; AGREEA(N) (“Adjective-Noun agreement operator”) is a set of rules
that describe agreement of a modifying adjective with the modified noun.
– («Σ») is a feature of the syntactics of a lexeme; e.g.: the feature («subj-gen») marks
in the lexicon a verb that requires its subject to be in the genitive, like XVatatʹ
‘be sufficient’: GorjučegoGEN xvatit ‘Fuel will.be.sufficient’.
– “L ➡ No.n(MWGX)” means that the lexeme L must go into the n-th position in
the corresponding linearizing pattern (these patterns for MWGs are described
below, I.B, p. 356).
– “g(L)” stands for the syntactic word group of the lexeme L (the “projection”
of the full subtree having L as the top node); units of g(L) type do not appear
at any representation level: they are used only in syntactic rules.
– In order to save space, the syntactics of the elements in the left-hand side of
the rule are not repeated in the right-hand side. (The syntactics, which are
elements of the lexicon, actually constitute the context of the application of
the rule and could have been indicated in the Condition part.)
– Shading indicates the context—that is, the elements that are not affected by
the rule, but whose presence is necessary for the rule to apply.

The SSyntS ⇔ DMorphS rules for Russian presented below are approximate; their
conditions are simply hinted at. In fact, each of these rules is just a placeholder
for a serious study of all contextual factors.
350 10 Word order in Russian

Actantial dependency rules

SSynt-rule I.A-1: SSynt-subject [a non-local SSyntRel]

L1(ΔVP)FIN L2-NOM L1
1) (…) not A

subjectival
AGREE (L1; L2)
V(N)

L2-GEN L1 [L1→NE, and Llex = («exist»)]
2) (…) A=
L
L2(N)
2(N) or
Llex = («subj-gen»)
AGREEV(N) (L1; L2)

☛ ΔVP stands for ‘verbal standard subtree’ (on standard subtrees, see Mel’čuk & Pertsov 1987:
485ff): a chain of subsequently dependent lexemes of the form L1-fin→Li-1→Li-2→…→Llex,
where possible Li have to be specified by a list (for instance, couldL1-fin haveLi-1 begunLi-2 toLi-3
separateLlex; Llex is the lexical verb whose combinatorial properties concerning the subject
percolate to the top node of this standard subtree. This is what was called the verbal
nucleus by S. Kahane (Kahane 1997 and 2001, Kahane & Mel’čuk 1999). ΔVP is used only in
the formulation of syntactic rules and does not appear as such in the SSyntS of a sentence.
The notation “L(ΔVP)” means ‘lexeme L that is the syntactic head of ΔVP.’

Comments

1. Rule I.A-1 does not uniquely specify the linear position of the subject with
respect to the MV, since, generally speaking, both positions are possible; the
appropriate one must be established by rules I.DI, see below, p. 358.
2. The condition A is intended to capture the use of the genitive on the subject
with some verbs marked “(«exist»)” that are negated and with other verbs that
always have the subject in the genitive and are marked “(«subj-gen»)”.
3. Rule I.A-1 ignores the following two frequent cases of the implementation of
the subjectival SSyntRel:
– Agreement of the copula MV. If the subject is èto ‘this’, the copula agrees
with its complement (rather than with the subject): Èto bylMASC Ivan(masc)
‘This was Ivan’. ~ Èto bylaFEM Marija(fem) ‘This was Mary’.
– Agreement of the MV if the subject is an infinitive or a subordinate čto-
clause; the MV must be in the 3sg, NEU: UčitʹsjaINF mne nravilosʹSG.NEU
lit. ‘To.study me pleased’; Čto Ivan otsutstvuet, udivljaet3.SG vsex lit. ‘That
Ivan is.absent amazes everyone’.
Such cases require additional SSynt-rules.

Examples (the glosses here and below are literal)

moglaL1 bytʹ obnaružena neftʹL2 ‘could be discovered oil’

~ neftʹL2 moglaL1 bytʹ obnaružena ‘oil could be discovered’
 10.3 Linearization rules (illustrated with Russian data) 351

MyL2, kak uže jasno, možemL1 sčitatʹ… ‘We, as [it is] already clear, can believe ...’
~ MožemL1 li myL2 sčitatʹ… ‘Can whether we believe ...’

Nikakix novyx ulikL2 ne pojavilosʹL1 = Llex = («exist»)

‘Of.no new evidence not appeared’.

Ètoj neftiL2-gen xvatiloL1= Llex = («subj-gen») ‘Of.this oil was.sufficient’.

~ Ètoj neftiL2-gen ne xvatiloL1 ‘Of.this oil was.not sufficient’.

SSynt-rule I.A-2: Complement of a preposition [a local SSyntRel]

Notation: “L(II[case])” means ‘DSyntA II of L is expressed by the grammatical case case.’

L1(Prep, II[case])

L1 L2-CASE
prepositional  + (… +) and L1  No.3(MWGN), L2  No.10(MWGN)

L2(N)

Examples

vL1 našem rasporjaženiiL2; voL1 vsex nixL2; dljaL1 togoL2, čtoby…;

at our disposal’ in all them for this that...
ni sL1 kemL2
no with body = ‘with nobody’

Comments

There are several complications in the use of Russian prepositions, not covered
by this rule; for instance:

– Some prepositions can or must be postposed to the nouns they introduce:

Boga radi! ‘God’s sake!’, rassudku vopreki ‘to.reason in.spite’, mesjac tomu
nazad/spustja ‘month ago/back’. For such prepositions/postpositions,
another SSynt-rule of type I.A is needed.
– A preposition is always postposed in approximate-quantitative and approx-
imate-ordinal constructions: dnja na tri ‘days for three’ = ‘for approximately
three days’, denʹ na tretij ‘day on third’ = ‘approximately on the third day’.
This linear arrangement is ensured by the SSynt-rules for approximate-
quantitative and approximate-ordinal SSyntRels (Chapter 2, Section 2.5,
Nos. 67 and 69, p. 87f ).
– Some special choices have to be made: the choice of the vocalic form of a
consonantal prim­ary preposition (vo instead of v, ko instead of k, etc.); the
 352 10 Word order in Russian

choice of the n-form of a substitute pronoun governed by the preposition (nix

instead of ix, nemu instead of emu, etc.); the choice of the split form of such
negative pronouns as nIkto, nIčto, etc. (*s nikem ⇒ ni s kem, etc.). These
and similar complications are taken care of by deep-morphological rules that
realize the corresponding radical morpheme of the preposition, of the substi-
tute pronoun and of the negative pronoun.

Adjunctial dependency rules

SSynt-rule I.A-3: Instrumental circumstantial

[a non-local SSyntRel, if L1 is a finite verb; semi-local otherwise]

L1
L1 L2(N)-INSTR
circumstantial  (…)

L2(N)-INSTR

Examples

PolučajutL1 sok sledujuščim obrazomL2

‘[«They»] obtain juice in.following way’.
L2 L1
sostavlen
Tekst1) byl+(… +) and LL1 izvestnym
1 ➡ No.10(MWG metodomL2
N); and:

‘Text wasif Lconstructed

2 = («quant»), thenby.known
L2 ➡ No.4(MWG method’.
N); 1) L2 has ≤ 3 postposed dependent
if L2 = («dem»), then L2 ➡ No.5(MWGN);
full lexemes;
L1(N) if L2 = («poss»), then L2 ➡ No.7(MWGN);
if L2 = («ord»), then L2 ➡ No.8(MWGN); 2) L2 ≠ («only postpos»)
SSynt-rule I.A-4: Adjectival
if L2 ≠ modifier
(«quant»), («dem»), («poss»), («ord»),
modificative ⇔ then L2 ➡ No.9(MWGN)
[a local SSyntRel, if L2 is a single ADJ; semi-local otherwise]4
AGREE (L2; L1)
A(N)

L2(ADJ)LONG L1 L2 L1 L2 has > 1 dependent full lexemes

2) +(… 1)
+) +(…and +)LL1 ➡ No.10(MWG
and N),
L1 ➡ No.10(MWG or
N ); and:
2 ➡ No.5(MWG A)
AGREE A(N) (Lif2;LL2 1=) («quant»), then L2 ➡ No.4(MWG
L2 = («postpos»)
N); 1) L2 has ≤ 3 postposed dependent
if L2 = («dem»), then L2 ➡ No.5(MWGN);
full lexemes;
L1(N) if L2 = («poss»), then L2 ➡ No.7(MWGN);
if L2 = («ord»), then L2 ➡ No.8(MWGN); 2) L2 ≠ («only postpos»)
if L2 ≠ («quant»), («dem»), («poss»), («ord»),
modificative ⇔ then L2 ➡ No.9(MWGN)
AGREE A(N) (L2; L1)

L2(ADJ)LONG L2 L1 L2 L2 has > 1 dependent full lexemes

1) +(…2)+) +(… and L+) and L ➡ No.10(MWGN),
1 ➡ No.10(MWG1N); and: or
L2 ➡ No.5(MWG A)
if L2 = («quant»),
AGREE then (LL2 2➡ 1)
; LNo.4(MWG
A(N) N); 1) L2 has = («postpos»)
≤L32 postposed dependent
if L2 = («dem»), then L2 ➡ No.5(MWGN);
full lexemes;
L1(N) if L2 = («poss»), then L2 ➡ No.7(MWGN);
if L2 = («ord»), then L2 ➡ No.8(MWGN); 2) L2 ≠ («only postpos»)
if L2 ≠ («quant»), («dem»), («poss»), («ord»),
then L2 ➡ No.9(MWGN)
4 The⇔position of a complex adjectival modifier depends in fact on many subtle and closely intertwined
modificative

1963. (L2; L1)

factors; see SannikovAGREE A(N)

L2(ADJ)LONG L1 L2 L2 has > 1 dependent full lexemes

2) +(… +) and L1 ➡ No.10(MWGN), or
L2 ➡ No.5(MWG A)
AGREE A(N) (L 2; L 1) L2 = («postpos»)
 10.3 Linearization rules (illustrated with Russian data) 353

L2 can be only a long-form adjective—i.e., an adjective bearing the grammeme long.5

Examples

važnajaL2 ocenkaL1; ètaL2ʹ isključitelʹno važnajaL2ʹʹ ocenkaL1;

important estimate this extraordinarily important estimate

važnajaL2 dlja vsex nas ocenkaL1 ≡ ocenkaL1, važnajaL2 dlja vsex nas, …
important for all us estimate

ocenkaL1, važnajaL2 dlja vsex nas v svjazi s processom

estimate important for all us in connection with process
formirovanija novyx grupp, …
of.formation of.new groups

polnyjL2 tragizma periodL1 ≡ periodL1, polnyjL2 tragizma

full of.tragic.events period

papaL1 rimskijL2 = («only postpos») ‘Pope Roman’ ~ *rimskijL2 papaL1

SSynt-rule I.A-5: Adnominal complement/attribute [a semi-local SSyntRel]

L1(N)

L1 L2-GEN
...-adnominal-...  + (… +)

L2(N)

This rule processes several adnominal SSyntRels: see Chapter 2, Nos. 48–53, p. 77ff.

Examples

razdačaL1 bogatstvL2; važnost′L1 ètoj zadačiL2; granicyL1

distribution of.wealth importance of.this problem borders
evropejskix stranL2
of.European countries

5 The Russian adjective has the inflectional category of PREDICATIVITY: as a general rule, a quali-
tative adjective has a long, or attributive, form and a short, or predicative, form: e.g., the adjec-
tive vysokij ‘high’ has a set of 48 long forms (vysokij, vysokogo, …, vysokaja, vysokoj, …, vysokoe,
..., vysokie, …) and a set of 4 short forms (vysok, vysoka, vysoko, vysoki). An adjective in the long
form is used in all syntactic roles possible for an adjective—as a modifier and as the attributive
complement of a copula verb; an adjective in the short form is used in Contemporary Russian
only as the attributive complement of a copula verb.
354 10 Word order in Russian

židkost′L1 golubogo cvetaL2; kollekciiL1 ètogo millioneraL2

liquid of.light-blue color collections of.this millionaire

SSynt-rule I.A-6: Prepositional/Adverbial circumstantial of Time or Location

[a non-local SSyntRel, if L1 is a finite verb; a semi-local SSyntRel otherwise]

L1
L1 L2
circumstantial  (…) L2 = («temporal»/«local»)

L2(PREP)/ADV

Examples

polučennoeL1 vL2 Moskve pisʹmo; VL2-1 1989 godu Ivan rabotalL1 vL2-2 Moskve.
received in Moscow letter in 1989 year Ivan worked in Moscow
VčeraL2 šëlL1 doždʹ ‘Yesterday was.falling rain’.

Ancillary dependency rules

SSynt-rule I.A-7: The analytical form of the passive [a non-local SSyntRel]

L3 … L1(BYTʹ)
subjectival
L3 L1 (…) L2-PASS.PART, PERF, PAST, SHORT
passive-analytical  … subjectival

AGREEA(N) (L2; L3)

L2

Comment

The passive-analitycal SSyntRel is introduced as a special SSyntRel because both

members of the passive-analytical SSyntRel are fixed. In contrast, for instance, to
the copular-completive SSyntRel, it admits as its governor only the verb Bytʹ ‘be’,
but no other copula; and as its dependent, only a short perfective passive past
participle:

(3) a. Passive-analytical SSyntRel

bylL1 polučenL2 ‘was received’ ~ *okazalsja polučen ‘proved received’,
*stal polučen ‘became received’,
*kažetsja polučen ‘seems received’
 10.3 Linearization rules (illustrated with Russian data) 355

b. Copular-completive SSyntRel
byl cennym ‘was valuable’ ~ okazalsja cennym ‘proved valuable’,
stal cennym ‘became valuable’,
kažetsja cennym ‘seems valuable’
byl vračom ‘was a.doctor’ ~ okazalsja vračom ‘proved a.doctor’,
stal vračom ‘became a.doctor’,
kažetsja vračom ‘seems a.doctor’

Examples

OcenkaL3 bylaL1 by nemedlenno polučenaL2. | IvanL3 ØBYT′L1 uvolenL2.

Estimate would have.been immediately obtained Ivan is fired
BudučiL1 prinjatyL2, èti studentyL3 polučajut stipendiju.
Being admitted these students receive scholarship

I.B: linearization patterns for MWGs

The linearization of the dependents of a local SSyntRel is done by means of linear-

izing patterns [LPs], describing the word order in MWGs. An LP is a rigid sequence
of positions, each of which admits one syntactic element L; these positions cor-
respond to local SSyntRels.

Exception P
 osition No. 9 for the modificative SSyntRel admits several co-dependent adjectives;
their mutual order is established by special rules attached to this position. (For a
sketch of such rules, based on semantic and syntactic properties of Russian adjectives
modifying the same noun, see Iordanskaja 2000 and 2003.)

If an L that is meant to occupy a position in a linearizing pattern LP1 has its own
dependents, L is not put into the MWG under construction: another linearizing
pattern LP2 ensures the construction of L’s own MWG, which, at the next stage, is
united with the MWG specified by LP1. Thus, if a NUM(eral), which is supposed
to go into Position No. 6 of a nominal MWG [MWGN], is a compound NUM (e.g., tri
milliona šestʹsot sorok semʹ tysjač dvesti tridcatʹ odin ‘3 647 231’), then a numeral
MWG [MWGNUM] is first built, and it is embedded into MWGN as a whole at the stage
of uniting MWGs into CWGs (see example below).
There are several MWG linearizing patterns for a language; they correspond to
MWGs of different word classes: for instance, MWGN, MWGADJ, MWGNUM, MWGADV,
MWGV-INF and MWGV‑FIN in Russian. A SSyntS ⇔ DMorphS rule indicates for both
members of the SSyntRel described the positions in the corresponding pattern. Not
all of the positions in a linearizing pattern have to be filled: the pattern represents a
maximal possible string associated with an MWG, i.e., a potential minimal phrase.
356 10 Word order in Russian

Two linearizing patterns of MWGs are cited here for Russian: MWGN and MWGA.

Pattern of a minimal nominal word group

(in a broad sense: including the prepositional groups)

1 2 3 4 5 6 7 8 9 10 11
CONJ(coord) PARTICLE PREP A(quant) A(dem) NUM A(poss) A(ord) A N Ψinvar
no lišʹ dlja vsex ètix semi našix vtoryx važnyx čisel li
‘but’ ‘only’ ‘for’ ‘all’ ‘these’ ‘seven’ ‘our’ ‘second’ ‘important’ ‘numbers’

☛ Ψinvar stands for any non-Russian expression: a technical symbol, a number, a formula, a
foreign-lan-guage word/phrase, etc.

Pattern of a minimal adjectival word group

1 2 3 4 5
CONJ(coord) PARTICLE ADJ(pron) ADV ADJ
ili tolʹko takoj očenʹ tëmnyj
‘or’ ‘only’ ‘such’ ‘very’ ‘dark’

For better readability, both patterns are simplified. Thus, the position for the neg-
ative particle ne ‘not’ is not shown (ne can precede practically any element of a
pattern, except for the first one), nor is the position for a contrastive particle such
as že ≈ ‘as for’ (Naturalʹnye čisla že ras­smatrivatʹsja ne mogut ≈ ‘As for natural
numbers, they cannot be considered’.) or už ≈ ‘very’ (ne takoj už tëmnyj ‘not so
very dark’). Such particles are “squeezed” into MWGs by the corresponding lin-
earization rules of type I.A.
The MWGN admits the embedding of the MWGADJ and MWGNUM: for instance,
MWGNUM pjatʹ tysjač šestʹsot sorok semʹ ‘five thousand six.hundred forty seven’
and MWGA ne nastolʹko už važnyx ≈ ‘not so very important’ can be introduced as
wholes into positions Nos. 6 and 9 of an MWGN, respectively. In a similar way,
practically all positions admit embedding of coordinated WGs: for instance, ètix
ili tex ‘these or those’ must be embedded in position No. 5, or vtoroj, četvërtyj i des-
jatyj ‘[the] second, fourth and tenth’, in position No. 8, etc. This type of embed-
ding is carried out by rules of I.C group.
Let me emphasize the following interesting fact: what is known in the Russian
grammatical tradition as a “complex verbal predicate” (auxiliary bytʹfut ‘be’ + the
infinitive; auxiliary bytʹ + passive past participle; copula bytʹ ‘be’ + attributive
noun/adjective; etc.) does not correspond to an MWG or even to a CWG. From the
viewpoint of linearization, the complements/attributives of auxiliary and copular
 10.3 Linearization rules (illustrated with Russian data) 357

verbs behave like any other SSynt-actants: they are non-local dependents and are
ordered at the stage of linearizing CWGs inside the clause.
For the SSyntS of Figure 10.1, Rules I.B (= linearizing patterns) produce six
MWGs, listed in alphabetical order:
1. byla; 2. gravitacionnoj razvedki; 3. metodom; 4. neftʹ; 5. otkryta; 6. v Kaza-
xstane

I.C: Arranging word groups within a complete word group

I.C I: Positioning of the dependent word group Ψ with respect to the governing
word group Ξ

In these rules, the following writing convention is used: the dependency shown
between two word groups [WGs] represents the dependency between their top
nodes. The WGs that form a complete WG can be themselves both minimal and
complete.

– r = coord:
I.C I-1. Ξ–coord→Ψ ⇔ Ξ + (…+) Ψ

– r ≠ coord:
I.C I-2. WGNʹ→WGN * ⇔ WGNʹ + (…+) WGN
I.C I-3. WGN→WGADJ ** ⇔ 1) WGN + (…+) WGADJ
⇔ 2) WGADJ ➡ No. 9 (WGN)
I.C I-4. WGN→WGINF ⇔ WGN + (…+) WGINF
I.C I-5. WGN→WGADV ⇔ WGN + (…+) WGADV
I.C I-6. WGA→WGN ⇔ WGADJ + (…+) WGN
I.C I-7. WGADJ→WGADV** ⇔ 1) WGADJ + (…+) WGADV
⇔ 2) WGADV + (…+) WGADJ
* Exception: NP of the form takogo roda ‘of such a type’ can be anteposed—by another
rule of I.CI type not given here.
** The choice between the two subrules is made according to Conditions not specified here.

Examples

I.C I-2 [
maksimalʹnaja verojatnostʹ]WG-Nʹ [vsex podobnyx raspredelenij]WG-N
maximal probability of.all such distributions
I.C I-3(1) [ili veličina]WG-N [ne polnostʹju opredelënnaja]WG-ADJ
‘or [a] magnitude not fully determined’
I.C I-3(2) [ili [ne polnostʹju opredelënnaja]WG-ADJ veličina]WG-N
‘or [a] not fully determined magnitude’
358 10 Word order in Russian

I.C I-4 [ego udivitelʹnaja sposobnostʹ]WG-N [spatʹ]WG-inf

‘his amazing ability to.sleep’
I.C I-5 [vse simvoly]WG-N [sleva]WG-ADV
‘all symbols on.the.left’
I.C I-6 [polnostʹju lišeny]WG-ADJ [neobxodimyx sredstv]WG-N
‘[are] fully deprived of.necessary means’
I.C I-7(1) [vstrečajuščiesja]WG-ADJ [lišʹ izredka]WG-ADV
‘encountered only rarely’
I.C I-7(2) [lišʹ izredka]WG-ADV [vstrečajuščiesja]WG-ADJ
‘only rarely encountered’

I.C II: Mutual arrangement of codependent word groups Ψ

either Ξ + WGagent + WGadnom.compl + WGadnom.compl + WGCirc, non-manner + WGobl-obj+ WGobl-obj

+ WGCirc-manner
or Ξ + WGCirc-manner + WGadnom.compl + WGagent + WGCirc-manner

In our test sentence, Rule I.C II unites MWGs 2 and 3, which gives us five CWGs:
1. byla, 2. metodom gravitacionnoj razvedki, 3. neftʹ, 4. otkryta, 5. v Kazaxstane

I.D: Arranging complete word groups within a clause

The arrangements proposed here are valid only without taking into account the
SSynt-Comm-structure and other “perturbing” factors—that is, for the word order
traditionally called neutral. The neutral word order obtains in cases where the
SSyntS and the SSynt-CommS are not in conflict (the subject is (in) the SSynt-
Theme, there is no Focalization, etc.). Rules I.D perform three operations:

I.D I  linearizing the elements of the verbal nucleus (referred to below as MV,
~
MV being the lexical verb, the last element of the nucleus)
I.D II
~
linearizing the actants with respect to MV
~
I.D III linearizing the circumstantials with respect to MV and the actants

I.D I: Building the verbal nucleus

~
Within an MV, with a neutral word order, a dependent follows its governor:

Ξ→Ψ ⇔ Ξ + Ψ. In our case, Rule I.D I gives [byla + otkryta]MV

~.
 10.3 Linearization rules (illustrated with Russian data) 359

I.D II: Linearizing the actants

The SSynt-actants are numbered: A1 is the subject; A2 is the DirO, the strongest
OblO with an intransitive verb or the complement of a copula; A3 is the IndirO or
an OblO; and A4 is another OblO. Different arrangements of the actants Ai with
~
respect to the MV are mostly determined by the nature of the MV: the MV is a
copula, the MV is an existence verb, or the MV is neither. An additional case is a
non-finite verb that governs actants.

Governor = MV
MV = («copula»)

additional conditions arrangement examples

1. A1 ≠ VINF
~ +...+ A2
A1 +...+ MV SašaA1 –MV naš voždʹA2 ‘Sasha [is] our leader’.
or SašaA1 bylMV našim voždëmA2 ‘Sasha was our leader’.
A2 = VINF SašaA1 (budet) dovolenA2 ‘Sasha (will be) happy’.
VesA1 okazalsja ravnymA2 1,008 kg
‘The.weight turned.out.to.be equal [to] 1.008 kg’.
PostupitʹA1 tak označalo by poterjatʹA2 kontrolʹ nad situaciej
‘To.act like.this would.mean lose control over the.situation’.

2. A1 = VINF /CONJ(«subord»)
~ +...
A2(A2) +...+ MV ØBYTʹ-MV neobxodimo/BudetMV neobxodimoA2 učestʹA1 vse
and + A2 +...+ A1 faktory ‘[It is/will.be] necessary to.account for.all factors’.

A2 = ADJ/ADV(«pred-inf») [“A2(A2)” means VamA2(A2) budetMV nadoA2 učestʹA1 vse faktory

‘The second actant of ‘To.you [it] will.be necessary to.account for.all factors’.
the second actant’] Osobenno važnoA2, čtoA1 učteny vse faktory ‘[It is] espe-
cially important that [are] accounted.for all factors’.

MV = («exist»)

additional conditions arrangement examples

3. – ~ +...+ A1 Šël silʹnyj doždʹ ‘[It] was.falling [a] heavy rain’.

A2 +...+ MV
Na doroge pojavilsja vsadnik ‘On road [it] appeared [a] rider’.
U nas imeetsja veličinaA1 c > 0, zavisjaščaja ot P
At us, [there] is magnitude c > 0, depending on P’.

MV ≠ («copula»), ≠ («exist»)

additional conditions arrangement examples

4. MV ≠ («pred-inf»),
~
A1 +...+ MV +...+ A2 +...+ A3 Ètot operator sopostavljaet čislo A čislu B s pomo­
≠ («aux») + ...+ A4 +...+ A5 +...+ A6 ščʹju funckcii f
‘This operator associates number A to number B
by using function f’.

5. MV = («pred-inf») A2 +...+ A1 MneA2 xočetsja žitʹA1 lit. ‘To.me is.desire to.live’.

360 10 Word order in Russian

6. MV = («aux»), ~
MV +…+ A1 V Avstralii byl predotvraščënA2 krupnyj teraktA1 ‘In
A2 = part, pass
~]
[A2(MV) is inside MV Australia, was thwarted [a] serious act.of.terror’.
and A2(A2) = —

Governor ≠ MV

Governor = (V)non-fin

additional conditions arrangement examples

7. – Governor + A2 + A3 + A4 svestiG zadačuA2 kA3 predyduščej

‘to.reduce [the] pro­blem to.the previos [one]’;
svjazyvavšixG indejcevA2 sA3 francuzskimi perekupšči­kami
‘that.were.connecting Indians with French merchants’

The application of Rule I.D II-4 results in the following arrangement of actantial
groups:

[byla + otkryta] MV
~ + neftʹA1

I.D III: Linearizing the circumstantials

Linear disposition of circumstantials is controlled mainly by their semantic

nature. Thus, Time and Location circumstantials tend to occupy the left edge of
~
the clause, while Direction circumstantials mostly follow the MV; manner circum-
stantials behave differently—as a function of their own structure: a simple adverb
~ ~
precedes the MV, while a prepositional phrase follows MV. Therefore, the rules
for circumstantial linearization need a list of all circumstantial types (in the rules
below only 8 such types are given). The semantic type of a Circum is identified:

– Either by the subordinating SSyntRel, such as durative-circumstantial (Rus. Ivan

spal–[dva]–dur-circum→časa lit. ‘Ivan slept two hours’.), Chapter 2, Section 2.5,
No. 34, p. 72.
– Or by the lexicographic features, semantic and syntactic; for instance, “L =
Circumtime” means that L is an L(«temporal»), as, e.g., late, week, after, etc.
– Or by the full (= semantic) case of L; thus, “L = Circuminstr” means that L is an
L(N)instr, etc.

In the table below, numbers attached to an arrangement indicate mutual order

of two codepen­dents; negative numbers specify the distance from the governor
to the left, and positive numbers specify the distance from it to the right. The
number after the indication “leftmost” is to be understood as follows: 0—the
 10.3 Linearization rules (illustrated with Russian data) 361

very first element of a clause (a conjunction), +1—the second element (a non-con-

junctional connector: A0, sledovatelʹno+1, ... ‘And, consequently, ...’), etc. Thus,
Rules I.D III-4/5 stipulate that Circummanner is positioned closer to the verb than the
­Circumquant.

the type of Circum arrangement examples

1. Circumtime left of [MV, Ai], –3 V 1932 godu on pereexal v Moskvu

‘In year 1932 he moved to Moscow’.

2. Circumloc left of [MV, Ai], –3 V Moskve on rabotal nad knigoj

‘In Moscow he worked on [the] book’.

3. Circumdur right of [MV, Ai], +2 On rabotal nad knigoj vsju nedelju

‘He worked on [the] book [the] whole week’.

4. Circummanner left of G, –1 and right of A1 On mog by tščatelʹno proveritʹMV zamki

‘He could.have carefully checked [the] locks’.

5. Circumquant left of G, –2 and right of A1 On tri raza tščatelʹno proveril zamki

‘He three times carefully checked [the] locks’.

6. Circumway, On pereskočil čerez lužu odnim pryžkom

Circuminstr, right of [MV, Ai], +1 ‘He jumped over [the] puddle in.one leap’.

Circumcomit On poexal v Moskvu s dvumja druzʹjami

‘He went to Moscow with two friends’.

7. Kak izvestno, on poexal v Moskvu s dvumja druzʹjami

Circumparenth left of MV ‘As [is] known, he went to Moscow with two friends’. ~
On, kak izvestno, poexal v Moskvu s dvumja druzʹjami
‘He, as [is] known, went to Moscow with two friends’.

8. Circumconnect leftmost, +1 Sledovatelʹno, my dokazali naše predpoloženie

‘Therefore, we have.proven our assumption’.

According to Rules I.D III-2 and I.D III-6, the locative circumstantial v Kazaxstane
and the instrumental circumstantial metodom gravitacionnoj razvedki are posi-
tioned as follows:

[V + Kazaxstane]C-loc + [byla + otkryta] MV

~ + [neftʹ]A1 + [metodom gravitacionnoj razvedki]C-instr

I.E: Arranging clauses within a sentence

[for more details, see Iordanskaja & Mel’čuk 2015]

The linear position of a subordinate clause inside the sentence depends on the
type of the clause (that is, on the SSyntRel that subordinates its top node) and on
the conjunction that introduces it, being its top node.
☛ TN stands for the top node of a clause; “clause(L)★” means ‘the clause headed by L minus
the clause headed by TN’.
362 10 Word order in Russian

I.E-1. L–subject→TN ⇔ 1) clause(TN) + clause(L)★ | L = (V)

⇔ 2) clause(L)★ + clause(TN) | L = (A/ADV)
I.E-2. L–object→TN ⇔ clause(L)★ + clause(TN)
I.E-3. L–circum→TN ⇔ 1) clause(TN) + clause(L)★ | TN = ESLI, KOGDA
⇔ 2) clause(L)★ + clause(TN) | TN = ČTOBY
I.E-4. L–relative→TN ⇔ L + clause(TN) | clause(TN) becomes part
of CWG(L)

The linear arrangements indicated here are valid only for neutral word order and
may be changed by communicative and other factors.

Examples
I.E-1(1) ČtoTN on ušël, nikogo ne udiviloL ‘That he left nobody not astonished’.
I.E-1(2) Bylo očevidnoL, čtoTN on ušël ‘[It] was obvious that he left’.
I.E-2 Ja znajuL, čtoTN on ušël ‘I know that he left’.
I.E-3(1) EsliTN on pridët, ja ujduL ‘If he comes I will.leave’.
I.E-3(2) On pridëtL, čtobyTN ja mog ujti ‘He will.come that I could leave’.
I.E-4 professorL, k kotoromu ja prišëlTN ‘professor to whom I came’

The output sentence (1) consists of just one clause, so that Rules I.E do not apply.

10.3.2.2 Type II: Adjusting linearization rules

Adjusting rules account for special linear arrangements determined by such
factors as Synt-Comm-organization, WH-words, pronouns and similar phenom-
ena; these rules can be thought of as transformations defined over established
preliminary arrangements.

☛ The number associated in a rule with a word group manipulated by the rule characterizes the
mutual arrangement of elements claiming the same position. Thus, in Rule II.1-2, number
+3 associated with the group Ψ means that Ψ may be preceded only by the elements with
numbers 0 (conjunctions), +1 (WH-words), and +2 (Comm-Specifiers); these numbers are
associated with conjunctions and WH-words in the corresponding rules for their positioning.

II.1. Expressing Synt-Comm-organization

II.1-1. If Ψ ⊆ Rheme in a declarative clause, then Ψ must be rightmost in the clause.

II.1-2. If Ψ ⊆ Theme in a declarative clause, then Ψ must be leftmost in the clause,
that is, Ψ: +3.
 10.3 Linearization rules (illustrated with Russian data) 363

The presentation of rules II.1 is approximate: other Synt-Comm-rules are needed

that ensure the expression of Focalization and Emphasis; all Synt-Comm-rules
must also take care of the corresponding prosody, etc. (Eight linear-prosodic
transformations of Russian presented in Zimmerling 2008: 560 correspond to our
Type II.1 rules.)

II.2. Extractions

WG(Ξ(wh)) stands for a word group consisting of a WH-word (a relative or inter-

rogative pronoun) and the string of its successive governors—up to, but with the
exclusion of, the MV. This is what is called nominal nucleus in Kahane 1997 and
2001.
WG(Ξ(wh)) must be second leftmost in the clause—it can be preceded only by a
conjunction; WG(Ξ(wh)): +1.
* i ja prišël k kotoromu ‘and I came to which’ ⇒
i [k kotoromu]WG(Ξ(wh)) ja prišël ‘and [to which] I came’

II.3. Nominal pronouns

V + A2(N, pronominal) ⇒ A2(N, pronominal) + V
Èto udiviloV vsex issledovatelejA2 ⇒ Èto vsexA2(N, pron) udiviloV
‘This astonished all researchers’. ‘This all [= everybody] astonished’.
Maša možet ljubitʹV IvanaA2 ⇒ Maša možet egoA2 ljubitʹV
‘Masha may love Ivan’. ‘Masha may him love’.

II.4. Interrogative inversion

A1 + MV ⇒ MV + A1 | in a general question

Since in our test example the CWG1 [= Circuminstr] expresses the SSynt-T, every-
thing else belonging to the Synt-R, SSynt-Comm-Rules II.1-1 and II.1-2 give the
prefinal arrangement (4):

(4) Prefinal arrangement:

[metodom gravitacionnoj razvedki]C-instr + [v + Kazaxstane]C-loc + [byla + otkryta]MV

~ + [neftʹ]A
1
364 10 Word order in Russian

10.3.2.3 Type III: Filtering linearization rules

These rules identify bad word sequences in the sentence under construction and
slap on them numerical “fines.” Then a special mechanism carries out permuta-
tions of CWGs in order to minimize the cumulative fine of the sentence. Such per-
mutations should not be applied to the ele­ments that belong to the SSynt-Theme
and the Rhematic Focus.

Table 10.1 F iltering linearization rules6 7

Notation: l(X) stands for ‘length of the word group X in terms of the number of
stressed wordforms.’

situation to avoid “fine” examples

1. Relative heaviness of adjacent CWGs6
CWG1 + CWG2 | CWG2 ⊄ Rheme: ?
On peredal [knigu v krasnom pereplëte]CWG1 Ivanu CWG2
l(CWG1) – l(CWG2) > 0 and ≤ 3 –2 ‘He passed [the book in a red binding] to.Ivan’.
??
On peredal [ètu tolstuju knigu v krasnom pereplëte i obe
tetradi]CWG1 IvanuCWG2 ‘He passed [this thick book in a red
>3 and ≤ 6 –6 binding and both notebooks] to.Ivan’.
???
On peredal [ètu tolstuju knigu v krasnom pereplëte i obe
>6 tetradi, kotorye byli najdeny našimi sotrudni­kami,]CWG1
–15 IvanuCWG2 ‘He passed [this thick book in a red binding and
both notebooks that were found by.our collabora­tors] to.
Ivan’.
2. Unbalanced distribution of CWGs around MV
On one side of MṼ there are –10 ??
[V Moskve]SSynt-T, CWG1 [semʹja Ivana]CWG2 živëtMV
≥ 2 CWGs and on the other side ‘In Moscow Ivan’s family lives’.
there is none ??
BudetMV [možno]CWG1 [svesti]CWG2 [priznaki fonem]CWG3
[k dvoičnym]CWG4 ‘[It] will.be possible to.reduce features
of.phonemes to binary [ones]’.
3. Misplacement of a non-manner circumstantial
MṼ + Cnon-manner + A1 –3 ??
Ètu kniguA2 pročëlMV vC 2005 godu IvanA1
‘This book read in year 2005 Ivan’.
4. Cnon-manner + MṼ | C ≠ –2 ?
Kolumbom vC 1492 godu bylaMV otkryta Amerika
a SSynt-Speci­fier7 ‘By.Columbus in year 1942 was discovered America’.
5. Misplacement of the agent of a passive verb
A2 + A1 + MVpass –3 ??
Kolumbom Amerika byla otkryta v 1492 godu posle
dolgogo plavanija ‘By.Columbus America was discovered
in year 1942 after [a] long sea.voyage’.

6 On the role of relative heaviness of word groups to be linearized (and other interesting factors),
see Abeillé & Godard 2000.
7 A Comm-Specifier is a part of a Sem-CommS/Synt-CommS that is outside of its Communicative
Core (= Rheme + Theme) and semantically bears on this core, specifying some details about it;
Comm-Specifiers are divided in Comm-circumstantials, Comm-characterizers and Comm-con-
nectors (Mel’čuk 2001: 96–100).
 10.4 Conclusions 365

Another bad arrangement is typical of article languages: thus, in French (Abeillé

& Godard 2000) a phraseologized complement that has no determiner cannot be
separated from its verb by another complement:

(5) a. *Cela donne à Marie faim

lit. ‘This gives to Mary hunger’. ~ Cela donne faim à Marie.
b. Cela donne à Marie une grande faim
lit. ‘This gives to Mary a big hunger’. ~
Cela donne une grande faim à Marie
lit. ‘This gives a big hunger to Mary’.
c. Jean donne à Marie une pomme ‘John gives to Mary an apple’. ~
Jean donne une pomme à Marie ‘John gives an apple to Mary’.

However, in Russian such a situation is impossible because of the absence of

determiners. But being part of a collocation or an idiom may impact the linear
position of a clause element.
In our example, the prefinal arrangement in (4) gets the fine of –2 (by Rule
III-4: the Cir­cumloc v Kazaxstane is not marked as a Specifier: it is part of the
SSynt-R. To reduce the fine to zero the following permutation can be used (the
permuted element is boxed):

~ ~
Circuminstr + Circumloc + MV + A1 ⇒ Circuminstr + MV + A1 + Circumloc

The result is a good linear arrangement:

(6) Metodom gravitacionnoj razvedki byla otkryta neftʹ v Kazaxstane.

Sentence (6) coincides with our test sentence (1).

10.4 Conclusions

This chapter presents a rough sketch of how linearization (+ morphologization)

of the surface-syntactic structure of a sentence can be captured for Russian, a
language with extremely flexible word order, within the framework of the Mean-
ing-Text approach. The overview presented here partially defines the input and
output representations needed for linearization in general and sketched out the
major classes of linearization rules, their form and their interaction. One example
Russian sentence has been worked through in some detail, showing how these
366 10 Word order in Russian

linearization + morphologization rules function to yield a good linear arrange-

ment of words inside a sentence.
The next step seems to be obvious: to elaborate a more or less exhaustive set
of word order rules for Russian. In the process, the researchers must consider and
describe systematically prosodic aspects of linearization.

Appendix: C
 ommunicative differences that determine the six
word arrangements in Motto 2

First, two important remarks.

– Each of the linear arrangements of words in (i) – (vi) can be associated with a
particular intonation contour that expresses the communicative organization
of the utterance (see Yokoyama 1985, where the importance of the relation-
ship between word order and intonation—especially in Russian—is properly
emphasized). These contours are shown here in an approximate way.
– Each of the arrangement of words (i) – (vi) admits several other patterns of
prosodization expressing different communicative structures, of which only
one is chosen to illustrate my point.

↗ ↘ ↘
(i) Ja tebja ljublju,
uttered with neutral, or level (i.e., unmarked) intonation, is a simple declara-
tive utterance—a logical, non-emotional statement of fact; the whole utter-
ance is Rhematic.

(ii) Ja ljubljú tebja
is an emphatic utterance, with strong stress on the verb and uninterrupted
falling contour; ja is the Synt-T, and the rest, the Synt-R.

(iii) Ljubljú ja tebja

is also an emphatic utterance, as well with strong stress on the verb and unin-
terrupted falling contour; the whole utterance is Rhematic and much more
colloquial than (ii).

(iv) Ljubljú tebja ja

is the same as (iii).
 10.4 Conclusions 367

↗
(v) Tebja | já ljublju
has ty as the Synt-T, everything else being the Synt-R—with ja as Rhematic
Focus (contrasting with another candidate, understood, but not named: ‘but
not …!’).
↗
(vi) Tebja ljublju já is the same as (v).

Chapter 10 is an expanded and corrected version of Mel’čuk 2011.

11 Linear ordering of genitive adnominal dependents
cosubordinated to a noun in Russian
11.1 The problem stated
11.2 Rules for ordering cosubordinated NGENs
11.3 Illustrations of NGEN ordering rules
11.4 Ordering of cosubordinated NGENs vs. ordering of cosubordinated ADJs
11.4.1 Ordering of cosubordinated ADJs
11.4.2 Comparison of both orderings: similarities and differences

To Anna Wierzbicka, a closest friend for 55 years

Przyjaźń jest rzeczą diabelnie trudną ‘Friendship is a devilishly difficult thing’
(Wierzbicka 1971: 83). Yes, Anna, generally speaking, this is so; but with you
friendship is the easiest thing!

11.1 The problem stated

This chapter constitutes a natural continuation of Chapter 5 (pp. 205ff ), where six
surface-syntactic relations [SSyntRels] that are necessary for the description of
N→NGEN phrases in Russian are proposed: it considers the linear ordering of geni-
tive nouns N1-GEN, N2-GEN, N3-GEN, … cosubordinated to the same noun N in Russian.
NB In fact what is being ordered are the whole phrases headed by these NGENs.

Example (1) shows that this order is not free:

(1) a. glagolyN napravlennogo dviženijaN1-GEN soveršennogo vidaN2-GEN russkogo

verbs of.directed movement of.perfective aspect of.Russian
jazykaN3-GEN
language
b. *glagolyN soveršennogo vidaN2-GEN russkogo jazykaN3-GEN napravlennogo
dviženijaN1-GEN

Therefore, the object of this chapter is, schematically, correspondence (2):

(2) Surface-syntactic structure Deep-morphological structure

N

⇔ N + N1-GEN + N2-GEN + N3-GEN + …

r1 r2 r3

…
N1 N2 N3

https://doi.org/10.1515/9783110694765-012
370 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

More specifically, I will present some surface-syntactic [SSynt-]rules that estab-

lish the corres­pondence between a SSynt-subtree and the deep-morphological
[DMorph-]string implementing it. The SSynt-subtree under consideration has the
following three properties:

(i) It is headed by a noun N on which syntactically depend several nouns N1, N2,
N3, … (each with its own dependents, if any).

(ii) The SSynt-relations ri that subordinate these Nis to N impose on them the
genitive case (in the DMorph-string). These SSynt-relations are six in number
(Chapter 5, pp. 226–230):

N—subjectival-adnominal-completive→NGEN-subj
priezd—subj-adnom→otca ‘coming of.Father’
stakan—subj-adnom→vody ‘glass of.water’
NGEN-subj expresses N’s deep-syntactic [DSynt-]actant I.

N—objectival-adnominal-completive→NGEN-obj
osvoboždenie—obj-adnom→otca ‘liberation of.Father’
portret—obj-adnom→rebënka ‘portrait of.child’
NGEN-obj expresses N’s DSynt-actant II.

N—qualificative-adnominal-attributive→NGEN-qual
balka—[nedostatočnoj]—qual-adnom→dliny ‘beam [of.insufficient] length’
NGEN-qual denotes a predicate whose Sem-actant 1 or 2 is expressed by N (‘dlina/
length—1→balka/ beam’: balka dvuxmetrovoj dliny ‘beam of.two.meter length’;
‘mečta/dream—2→devuška/girl’: devuš­ka moej mečty ‘girl of.my dream’). In
Russian, an NGEN-qual must necessarily have a syntactic depend­ent, normally
an adjective.

N—characterizing-adnominal-attributive→NGEN-charact
krik—charact-adnom→boli ‘scream of.pain’
živopisʹ—charact-adnom→Vozroždenija ‘painting of.Renaissance’
NGEN-charact and N are semantically related not as a predicate and its argument,
but by means of an “additional” predicate ‘σ’, which is not explicitly expressed
in the sentence: ‘krik, vyražaju­ščij‘σ’ bolʹ’/‘scream expressing‘σ’ pain’ or
‘živopisʹ vo.vremja‘σ’ Vozroždenija’/‘painting during‘σ’ Renaissance’.

N—possessive-adnominal-attributive→NGEN-poss
stadion—poss-adnom→universiteta ‘stadium of.University’
NGEN-poss and N are semantically related by means of the predicate ‘σ’ = ‘belong’:
‘stadion, prinadležaščij‘σ’ universitetu/stadium belonging.to‘σ’ the University’.
 11.1 The problem stated 371

N—metaphorical-adnominal-attributive→NGEN-metaph
lenta—metaph-adnom→dorogi ‘ribbon of.road’
NGEN-metaph is the lexeme whose metaphor is N: lentaN ‘ribbon’ is the metaphor
of doro­gaNGEN‑metaph ‘road’.

(iii) The six SSyntRels in question require the postposition of their dependent
NGENs with respect to the modified N, with one exception: the qual-adnom
SSyntRel allows the anteposition of its NGEN‑qual, if this NGEN-qual 1) has N as its
Sem-actant 1, 2) does not have itself a depending noun phrase and 3) is lexi-
cally marked as allowing for anteposition;1 for instance:

pojas golubogo cveta ‘belt of.light.blue color’ ~ golubogo cveta pojas

The anteposition of NGEN-qual is left out of consideration in this chapter.

 The basic order of postposed cosubordinated NGENs is determined syntacti-

3
cally—by the above SSyntRels: for each pair of these SSyntRels the mutual
order of their dependent NGENs is indicated. As a result, we obtain a general
six-position template (Table 11.1 in Section 11.2 below) that specifies the correct
position for each type of NGEN.

Such a template is possible because of the following essential fact:

Generally speaking, a dependent NGEN-i can occupy different linear posi-

tions with respect to its governing N as a function of the SSyntRel ri in the
N—ri→NGEN-i phrase.

1 The three cases of impossibility of NGEN-qual’s anteposition can be illustrated by the following
examples:
1) moego razmera tufli ‘of.my size shoes’ (‘razmer—1→tufli’) vs.
*moej mečty devuška ‘of.my dream girl’ (‘mečta—2→devuška’)
2) golubogo cveta lenta ‘of.light.blue color ribbon’ vs.
cveta morskoj volny lenta ‘of.color of.sea wave ribbon’ = ‘aquamarine ribbon’; the correct
*
expression: lenta cveta morskoj volny
3) neobyčajnoj krasoty portret ‘of.extraordinary beauty portrait’ vs.
prošedšego vremeni glagol ‘of.past tense verb’ (vremja ‘tense’ is not lexically marked as
*
allowing for anteposition)
In cases 1) and 2) the anteposition of an NGEN-qual can be possible under additional communicative
and/or syn­tactic conditions.
372 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

Thus:

(3) a. For the meaning ‘statue representing Athena and carved by Phidias’:
statuja AfinyNGEN FidijaNGEN ‘statue of.Athena of.Phidias’ ⟨*statuja Fidija Afiny⟩;
but for the meaning ‘statue representing Phidias and carved by Athena’:
statuja FidijaNGEN AfinyNGEN ‘statue of.Phidias of.Athena’ ⟨*statuja Afiny Fidija⟩
b. For the meaning ‘poet’s soul of this philosopher’:
duša poètaNGEN ètogo filosofaNGEN ‘soul of.poet of.this philosopher’ ⟨*duša
ètogo filosofa poèta⟩;
but for the meaning ‘philosopher’s soul of this poet’:
duša filosofaNGEN ètogo poètaNGEN ‘soul of.philosopher of.this poet’ ⟨*duša
ètogo poèta filosofa⟩

There are 15 logically possible pairs of NGENs (the number of combinations

from 6 by 2 without repetitions): NGEN-subj – NGEN-obj, NGEN-subj – NGEN-qual, etc. Three of
these pairs are semantical­ly impossible: NGEN‑metaph does not combine with NGEN‑obj,
NGEN‑subj and NGEN‑poss (it is difficult to imagine a metaphorically used noun that has
a subject/object actant or a possessor). As a result, there appear 12 SSyntRel pairs.
On the other hand, the qual-adnom and charact-adnom SSyntRels are repeatable,
so that we end up with 14 SSyntRel pairs to consider.
However, the use of SSyntRels alone for the linear ordering of cosubordinated
NGENs is not sufficient: for some SSyntRel pairs, the order of NGENs depends also on
the meaning of N and/or on that of NGENs. Thus, in the phrase krik boliNGEN-charact
­PetiNGEN-subj ‘scream of.pain of.Pete’ the NGEN‑charact can only precede the NGEN-subj (*krik
Peti boli), but in proizvedenija vosʹmidesjatyx go­dovNGEN-charact Lʹva TolstogoNGEN-subj
‘works of.1880s of.Leo Tolstoy’ ~ proizvedenija Lʹva Tolstogo vosʹmidesjatyx godov
the NGEN-charact can both precede or follow the NGEN‑subj—if the NGEN-charact de­notes the
temporal coordinate of the fact denoted by the governor N. As a consequence, the
proposed rules have to account for semantic factors as well.
Before proceeding to the formulation of NGEN-ordering rules, the following
principle has to be stated:

Each of the rules is valid only everything else being equal.

This means that the two cosubordinated NGENs being compared and ordered are
of the same weight (the corresponding phrases contain the same number of sylla-
bles and are of the same syntactic complexity) and there are no discourse factors
intervening.
The expression “discourse factors” should be understood very broadly. It
covers communic­ative and referential phenomena that can disturb the word
order observed in discourse-neutral contexts. The following discussion ignores:
 11.1 The problem stated 373

• The impact of the communicative structure. For instance, contrastive empha-

sis on one of cosubordinated NGENs can change their habitual linear order.
Thus, the neutral order is N + NGEN«material» + NGEN«color»:2 stol krasnogo dereva
bolʹšogo razmera ‘table of.mahogany of.big size’ ~ ?stol bolʹšogo razmera
krasnogo dereva; however, under emphasis, the dispreferred order is quite
normal:

(4) Ja išču stol bolʹšogo razmera KRASNOGO DEREVA, a ne iz karelʹskoj berëzy

‘I am.looking.for a.table of.big size OF.MAHOGANY, and not of Karelian birch’.

• The impact of the referential structure.

– A modifier either specifies a subclass of possible referents of the modi-
fied lexeme L (a restrict­ive modifier) or characterizes L’s referents without
specifying a subclass of these (a qualifying modifier). In what follows we
consider only restrictive modifiers. Thus, we exclude from our description
the situation where one of the cosubordinated NGENs is used as a qualifying
modifier (in dashes):

(5) Stoly malogo razmera – krasnogo dereva – u nas imejutsja v bolʹšom količestve
‘We have tables of.small size—of.mahogany—in a large quantity’.

– A
 restrictive modifier specifies a subclass of possible referents of the modi-
fied lexeme L; cosubordinated restrictive modifiers specify subsequent
subclasses of L’s possible referents. In a discourse neutral context, the
order of isolating these subclasses is irrelevant for the Speaker—different
characteristics of the L’s referents are, so to speak, informationally equal
for him. In this case, the linear order of cosubordinated modifiers is deter-
mined by their own properties—syntactic and/or semantic. This is the situ-
ation studied in the present chapter. However, we exclude the situation
where the Speaker first selects a particular subclass of L’s referents and
then introduces a subclass of this subclass. For instance, the dispreferred
order ?stol malogo razmera krasnogo dereva is quite OK if one speaks about
tables of small size and specifies a subclass of these from the viewpoint of
their material; sentence (6) is absolutely correct because of the referential
and communicative effects:

(6) Stoly malogo razmera krasnogo dereva u nas imejutsja v bolʹšem količestve, čem
takie že stoly iz karelʹskoj berëzy ‘We have tables of.small size of.mahogany in
a larger quantity than such tables of Karelian birch’.

2 Here and below an expression in small caps in « » quotes stands for a semantic label, whose
formal status is left vague.
374 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

11.2 Rules for ordering cosubordinated NGENs

The linear order of cosubordinated NGENs postposed with respect to their common
governor N is described by rules of three types:

1) Rule for the standard linear order of different-type NGENs, represented by their
maximal template (Table 11.1 below).

By “standard linear order” is meant here the order conditioned exclusively by

surface-syntactic relations that subordinate NGENs to their governor N, without
recourse to the semantic properties of the nouns involved. These properties are
taken into account by Rules 2.

2) Rules specifying semantic factors that affect standard linear order of differ-
ent-type NGENs.

Rules 2 are, in a sense, stronger than Rule 1: they impose deviations from the
standard order of NGENs determined by Rule 1.

3) Rule for the linear order of same-type NGENs, represented by their semanti-
cally-conditioned hierarchy (Table 11.2).

Rules 1–3 are part of word order, or linearization, rules for Russian (Chapter 10);
more precisely, they are a subset of the quasi-local word order rules.

1) Standard linear order of different-type NGENs

Table 11.1 Linear order of different-type postposed cosubordinated NGENs in Russian

1 2 3 4 5 6

–qual-adnom→N –metaph-adnom→N –obj-adnom→N –charact-adnom→N –subj-adnom→N –poss-adnom→N

2) Semantic factors of the linear ordering of different-type cosubordinated NGENs

Semantic properties of NGEN
1. If NGEN denotes a kind of N (rather than characterizing an individual N),
then this NGEN precedes all other cosubordinated NGENs.3
2. If NGEN denotes the material of which N is made,
then this NGEN precedes all other cosubordinated NGENs except for NGEN
­denoting kind.

3 Fairly often, the kind of N is expressed by an actant of N: zavod boepripasov ‘ammunition plant’,
škola tancev ‘dancing school’, detskaja bolʹnica ‘children’s hospital’; see Section 11.3, (9b), p. 377.
 11.2 Rules for ordering cosubordinated NGENs 375

3. If NGEN-charact denotes the time of N,

then NGEN-charact precedes or follows another NGEN-charact, NGEN-subj, NGEN‑obj and
NGEN-poss.

4. If NGEN-charact denotes the causer of N,

then this NGEN-charact precedes or follows NGEN-poss.

Semantic properties of N
5. If N denotes the quantity of NGEN or a set of NGENs,
then this NGEN precedes all other cosubordinated NGENs.

3) Linear order of the same-type cosubordinated NGENs

 everal same-type cosubordinated NGENs are possible only for two repeatable
S
SSyntRels: qual-adnom and charact-adnom. The mutual order of same-type
NGENs is determined by the following semantic hierarchy:

Table 11.2 Semantic hierarchy of same-type NGENs *

N < «KIND» < «MATERIAL» < «COLOR» < «SHAPE» < «SIZE»/«WEIGHT»/«ORIGIN»
< external characteristics < «(subjective) EVALUATION»

 n internal property of a real-world entity is its inherent property, inseparable from it: e.g., kind,
* A
material, color, form, texture, size, weight, etc. An external property of an entity is its position in
space and time, characteristics related to its functioning, its social role, etc.

This hierarchy, based on Vendler 1968: 128,4 is underlain by the Inherence prin-
ciple:
The modifiers of an N cosubordinated to this N by the same SSyntRel are lin-
early arranged according to the degree of their semantic “inherence” with
respect to the N: a more inherent characterization stands closer to the N.

The “degree of semantic inherence” of modifiers cannot be formally defined, but

one may think that the proposed hierarchy reflects this degree quite well. Thus,
the “objective” character­istics are more inherent than the “subjective” ones, the
internal properties are more inherent than the external ones, and «KIND» is the
most inherent characteristic.
Let it be emphasized that, although this hierarchy is introduced for the same-
type NGENs, it is partially valid also for the different-type NGENs. More precisely,

4 Vendler’s study (1968), following, as he indicates, Ziff 1960, considers English anteposed co-
subordinated adjectives with respect to their mutual linear ordering.
376 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

Rules 2 are based on the same Inherence principle: thus, the NGEN expressing
«KIND» precedes all other NGENs, etc.

11.3 Illustrations of NGEN ordering rules

The above rules will be illustrated proceeding as follows.

– The SSyntRels are considered pairwise, one after the other, from left to right
(in conform­ity with the template in Table 11.1).
– Each pair of SSyntRels is illustrated by phrases featuring the standard order
of the two NGEN nouns.
– Each deviation from this standard order is explicitly indicated.
– Each of the two repeatable SSyntRels—that is, qual-adnom and charact-adnom—
is also consi­dered in combination with itself.
– More than three co-subordinated NGENs are practically unacceptable.

The rules in question specify the best ordering possible. Deviations from it can
be charac­terized by different degrees of ill-formedness, of which three are distin-
guished: ungrammatical (*), hardly acceptable (??), jarring (?). We are aware that
our judgments of grammaticality can be challenged; however, for the purposes
of this chapter it is sufficient to perceive a difference in the degree of correctness.

—qual-adnom→N

This SSyntRel is repeatable.

With —qual-adnom→N

(7) a. tort domašnego prigotovlenija gigantskogo razmera

cake of.domestic preparation«origin» of.giant size«size» and
tort gigantskogo razmera domašnego prigotovlenija

b. kovër pëstryx cvetov nebolʹšogo razmera neobyčajnoj

carpet of.different colors«color» of.small size«size» of.extraordinary
krasoty
beauty«evaluation»
vs. ?kovër nebolʹšogo razmera pëstryx cvetov neobyčajnoj krasoty and
*kovër neobyčajnoj krasoty pëstryx cvetov nebolʹšogo razmera

The order of NGEN-quals in (7b) corresponds to the semantic hierarchy in Table 11.2.
 11.3 Illustrations of NGEN ordering rules 377

With —metaph→N: NGEN-qual precedes NGEN-metaph

(8) minarety strelʹčatoj formyNGEN-qual zavodskix trubNGEN-metaph
minarets of.arrow shape of.mill chimneys
vs. *minarety zavodskix trubNGEN-metaph strelʹčatoj formyNGEN-qual

With —obj-adnom→N: NGEN-qual precedes NGEN-obj

(9) a. portret nebolʹšogo razmeraNGEN-qual molodoj zenščinyNGEN-obj
portrait of.small size of.young woman
vs. portret molodoj zenščinyNGEN-obj nebolʹšogo razmeraNGEN-qual
?

b. sistema raspredelenijaNGEN-obj toka vysokoj nadëžnostiNGEN-qual

system of.distribution«KIND» of.current of.high reliability
vs. *sistema vysokoj nadëžnosti raspredelenija toka

The NGEN raspredelenija [toka] is an NGEN-obj (being DSynt-actant II of the noun

sistema); according to the standard template (Table 11.1), it should follow an
NGEN-qual—as in (9a). How­ever, a semantic factor perturbs the standard order: this
NGEN-obj identifies a kind of system (≈ a particular device), not an individual
system, so that in conformity with Rule 2.1 it must precede the NGEN-qual.

With —charact-adnom→N: NGEN-qual precedes NGEN-charact

(10) a. voda kristalʹnoj čistotyNGEN-qual ètogo ozeraNGEN-charact
water of.crystal purity of.this lake
vs. *voda ètogo ozeraNGEN-charact kristalʹnoj čistotyNGEN-qual

b. stol krasnogo derevaNGEN-charact ogromnyx razmerovNGEN-qual

table of.red wood«material» of.huge dimensions
vs. ??stol ogromnyx razmerovNGEN-qual krasnogo derevaNGEN-charact

(10b) demonstrates again the impact of a semantic factor: according to Rule 2.2
the NGEN denoting material precedes all other NGENs (except the one denoting kind).

With —subj-adnom→N: NGEN-qual precedes NGEN-subj

(11) a. kartina nebolʹšogo razmeraNGEN-qual neizvestnogo xudožnikaNGEN-subj
painting of.small size unknown artist
vs. ?kartina neizvestnogo xudožnikaNGEN-subj nebolʹšogo razmeraNGEN-qual

b. kuča morskogo peskaNGEN-subj bolʹšogo razmeraNGEN-qual

pile«quantity» of.sea sand of.big size
vs. kuča bolʹšogo razmeraNGEN-qual morskogo peskaNGEN-subj
??
378 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

c. rjumka krasnogo vina pričudlivoj formy

wine.glass«quantity» of.red wine of.bizarre shape
vs. *rjumka pričudlivoj formy krasnogo vina5

The deviation from the standard order in (11b–c) is imposed by Rule 2.5.

With —gen-poss→N: NGEN-qual precedes NGEN-poss

(12) kartiny nebolʹšogo razmeraNGEN-qual ètogo kollekcioneraNGEN-qual

paintings of.small size of.this collector
vs. *kartiny ètogo kollekcioneraNGEN-poss nebolʹšogo razmeraNGEN-qual

—metaph-adnom→N

This SSyntRel is non-repeatable and combines only with an NGEN-qual (see above)
and with an NGEN-charact.

(13) minarety zavodskix trubNGEN-metaph Leonida SolovʹëvaNGEN-charact

minarets of.mill chimneys of.Leonid Solovyov
vs. *minarety Leonida SolovʹëvaNGEN-charact zavodskix trubNGEN-metaph

—obj-adnom→N

This SSyntRel is non-repeatable.

With —charact-adnom→N: NGEN-obj precedes NGEN-charact

(14) a. zavody boepripasovNGEN-obj UralaNGEN-charact

plants of.ammunition of.the.Urals
vs. *zavody UralaNGEN-charact boepripasovNGEN-obj

b. issledovanija dvux poslednix letNGEN-charact processovNGEN-obj aromatizacii

studies of.two last years«time» of.processes of.aromatization
and
issledovanija processovNGEN-obj aromatizacii dvux poslednix letNGEN-charact

5 This is an interesting case, since it represents a “superposition” of two lexemes: RJUMKA1a

‘wine glass with a thin stem…’ (rjumka pričudlivoj formy ‘wine glass of bizarre shape’) and
RJUMKA1b­ ‘quantity of liquid…’ (rjum­ka vina ‘glass of wine’): Xozjajka postavila peredo mnoj
rjumku krasnogo vina pričudlivoj formy ‘The hostess put in front of me a glass of red wine of a
­bizarre shape’. However, this superposition is not possible in all contexts: *On vypil rjumku
­krasnogo vina pričudlivoj formy ‘He drank a glass of wine of bizarre shape’.
 11.3 Illustrations of NGEN ordering rules 379

The variation of the placement of the NGEN-charact denoting time is allowed by Rule 2.3.

With —subj-adnom→N: NGEN-obj precedes NGEN-subj

(15) a. portret devočkiNGEN-obj SerovaNGEN-subj vs. ?portret SerovaNGEN-subj devočkiNGEN-obj

portrait of.girl of.Serov

b. talant ljubviNGEN-obj poètaNGEN-subj vs. *talant poètaNGEN-subj ljubviNGEN-obj

talent of.love«kind» of.poet
NB The violation of the standard order in (15b) is worse than that in (15a) because of Rule 2.1:
NGEN-obj in (15b) denotes a kind of talent. In other words, if (15a) violates only a syntactic rule,
(15b) violates both a syntactic rule and a semantic constraint.

With —gen-poss→N: NGEN-obj precedes NGEN-poss

(16) portret devočkiNGEN-obj s serʹgoj amsterdamskogo muzejaNGEN-poss 

portrait of.girl with earring of.Amsterdam museum
vs. *portret amsterdamskogo muzejaNGEN-poss devočkiNGEN-obj s serʹgoj

—charact-adnom→N

This SSyntRel is repeatable.

With —charact-adnom→N

(17) a. pisateli Vostočnoj EvropyNGEN-charact devjatnadcatogo vekaNGEN-charact

writers of.Eastern Europe«location» of.19th century«time» and
pisateli devjatnadcatogo vekaNGEN-charact Vostočnoj EvropyNGEN-charact

The freedom of the placement of the NGEN-charact denoting time is ensured by Rule 2.3.

(17) b. klinok damasskoj staliNGEN-charact semnadcatogo vekaNGEN-charact

blade of.Damascus steele«material» of.17th century«time»
izvestnogo Abu-VaxbaNGEN-charact
of.known Abu-Wahb«causer» and
 klinok damasskoj staliNGEN-charact izvestnogo Abu-VaxbaNGEN-charact semnadcatogo
vekaNGEN-charact
vs. *klinok izvestnogo Abu-VaxbaNGEN-charact damasskoj staliNGEN-charact ­semnadcatogo
vekaNGEN-charact
380 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

The impossibility of the last phrase is also determined by semantic hierarchy:

the NGEN-charact denoting «MATERIAL» should precede other NGENs (except «KIND»).

With —subj-adnom→N: NGEN-charact precedes NGEN-subj

(18) a. krik užasaNGEN-charact rebënkaNGEN-subj vs. *krik rebënkaNGEN-subj užasaNGEN-charact

scream of.horror of.child

b. grudʹ mysliteljaNGEN-charact moego drugaNGEN-subj

chest of.thinker of.my friend
vs. *grudʹ moego druga myslitelja [ungrammatical in the intended meaning]

c. bjust karrarskogo mramoraNGEN-charact velikogo MikelandželoNGEN-subj

bust of.Carrara marble of.great Michelangelo
vs. *bjust velikogo MikelandželoNGEN-subj karrarskogo mramoraNGEN-charact

d. rasskazy vosʹmidesjatyx godovNGEN-charact Antona Pavloviča ČexovaNGEN-subj

short.stories of.1880s of.Anton Pavlovitch Chekhov
and
rasskazy Antona Pavloviča ČexovaNGEN-subj vosʹmidesjatyx godovNGEN-charact

The freedom of placement of the NGEN-charact that denotes time corresponds to Rule 2.3.

With —gen-poss→N: NGEN-charact precedes NGEN-poss

(19) a. kulinarnye knigi srednevekovoj ItaliiNGEN-charact našej bibliotekiNGEN-poss

cook books of.medieval Italy of.our library
vs. *kulinarnye knigi našej bibliotekiNGEN-poss srednevekovoj ItaliiNGEN-charact

b. knigi vosemnadcatogo vekaNGEN-charact našej bibliotekiNGEN-poss

books of.18th century of.our library and
knigi našej bibliotekiNGEN-poss vosemnadcatogo vekaNGEN-charact
c. knigi izdatelʹstvaNGEN-charact Muton našej bibliotekiNGEN-poss
books of.publisher Mouton of.our library and
knigi našej bibliotekiNGEN-poss izdatelʹstvaNGEN-charact Muton

The freedom of placement of the NGEN-charact that denotes time, as in (19b), or the
causer, as in (19c), corresponds, respectively, to Rules 2.3 and 2.4.
 11.4 Ordering of cosubordinated NGENs vs. ordering of cosubordinated ADJs 381

—subj-adnom→N

This SSyntRel is non-repeatable.

With —gen-poss→N: NGEN-subj precedes NGEN-poss

(20) bjust MikelandželoNGEN-subj ÈrmitažaNGEN-poss

bust of.Michelangelo of.Hermitage.museum
vs. *bjust ÈrmitažaNGEN-poss MikelandželoNGEN-subj

To demonstrate how the rules proposed can be applied, let us return to example
(1), repeated here as (21):

(21) glagoly napravlennogo dviženijaNGEN-charact soveršennogo vidaNGEN-qual

verbs of.directed movement of.perfective aspect
russkogo jazykaNGEN-subj
of.Russian language

– First, the mutual arrangement of cosubordinated NGENs is specified by the

standard template (Table 11.1) for different-type NGENs: NGEN-qual precedes
NGEN-subj. The phrase russkogo jazyka ‘of.Russian language’ is an NGEN-subj that
expresses DSyntA I of glagoly ‘verbs’, which are ele­ments of the set ‘Russian
language’; according to the NGEN order template, it must follow the phrase
soveršennogo vida ‘of.perfective aspect’ (an NGEN-qual).
– Second, the mutual arrangement of NGEN-charact and NGEN-qual is specified by
Rule 2.1: in the standard case (= according to the template), NGEN-qual precedes;
but if NGEN-chatact denotes the kind of N, then NGEN-qual follows. And in (21), the
phrase napravlennogo dviženija denotes a particular kind of verbs.

 rdering of cosubordinated NGENs vs. ordering of

11.4 O
cosubordinated ADJs

It is interesting to compare the ordering of Russian postposed cosubordinated

NGENs with the ordering of Russian anteposed cosubordinated adjectives. As is
to be expected, NGENs and adject­ives, both being noun modifiers and on multiple
occasions synonymous, show significant paral­lelism in their ordering. First the
rules for the ordering of cosubordinated adjectives are presented (11.4.1) and then
they are compared with the corresponding rules for NGENs (11.4.2).
382 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

11.4.1 Ordering of cosubordinated ADJs

The papers Iordanskaja 2000 and 2003 propose a hierarchical semantic classifi-
cation of Russian adjectives that determines their mutual linear ordering—more
precisely, their relative closeness to the modified noun.6 Table 11.3 below presents
this classification. The higher in the table an adjective semantic class is (i.e., the
higher its rank), the closer its instance must be to the modified noun. This is so
since an adjective’s rank corresponds to the degree of inherence of the character-
istic the adjective expresses: the more inherent the characteristic, the closer to
the noun is the adjective.

Table 11.3 Hierarchical semantic classification of adjectives (Iordanskaja 2003: 161–162)

I Objective characteristics: properties

A Qualitative (non-measurable) properties

1. Permanent properties

1) Internal properties

a) Kind (kofejnaja [čaška] ‘coffee [cup]’)

b) Material (farforovaja [čaška] ‘china [cup]’)
c) Color (golubaja [čaška] ‘light blue [cup]’)
d) Shape (vytjanutaja [čaška] ‘elongated [cup]’)
e) Other internal properties (prozračnaja [čaška] ‘transparent [cup]’)

2) External properties (dešëvaja [čaška] ‘cheap [cup]’)

a) Functioning/using characteristics (udobnaja [čaška] ‘convenient [cup]’)

2. Temporary properties (čistaja [čaška] ‘clean [cup]’)

B Quantitative (measurable) properties (kroxotnaja [čaška] ‘tiny [cup]’)

II Subjective characteristics: evaluation (zamečatelʹnaja [čaška] ‘remarkable [cup]’)

And now some examples.

– Adjectives that express an objective characteristic are closer to the modified

noun than adject­ives expressing a subjective characteristic:

(22) zamečatelʹnaja vysokaja ëlka vs. ?

vysokaja zamečatelʹnaja ëlka
remarkable tall fur.tree

6 On the topic of ordering cosubordinated adjectives, see Svenonius 2008.

 11.4 Ordering of cosubordinated NGENs vs. ordering of cosubordinated ADJs 383

– Adjectives that express a qualitative characteristic are closer to the modified

noun than than adjectives expressing a quantitative characteristic:

(23) malenʹkaja srednevekovaja bašnja vs. srednevekovaja malenʹkaja bašnja

?

small medieval tower

– Adjectives that express a permanent characteristic are closer to the modified

noun than adjectives expressing a temporary characteristic:

(24) razbitoe uglovoe okno vs. uglovoe razbitoe okno

?

broken corner window

– Adjectives that express an internal property are closer to the modified noun
than adjectives expressing an external property:

(25) dešëvye gorjačie bubliki vs. gorjačie dešëvye bubliki

?

cheap hot bagels

– Hierarchy of internal property adjectives: for instance, «material» adjectives

are closer to the modified noun than «color» adjectives; «kind» adjectives
are closer to the modified noun than any other adjectives; etc.

(26) a. krasnyj aljuminievyj čajnik vs. aluminievyj krasnyj čajnik

?

read aluminum teapot

b. farforovaja kofejnaja čaška vs. ??
kofejnaja farforovaja čaška
china coffee cup

To sum up: The linear ordering of cosubordinated adjectives is determined

semantically, i.e., by their meaning—of course, everything else being equal, the
same as with NGENs (see the end of Section 11.1): that is, the weight of the NGEN’s
phrase and discourse factors.
However, this is not true for Russian possessive adjectives, such as MAMIN
‘Mom’s’ or PETIN ‘Pete’s’: their mutual linear arrangement is determined by their
syntactic role, cf.:

(27) a. Petin[subj]/[poss] mamin[obj] portret

‘Mom’s portrait by Pete’/‘Mom’s portrait belonging to Pete’
b. mamin[subj]/[poss] Petin[obj] portret
‘Pete’s portrait by Mom’/‘Pete’s portrait belonging to Mom’

(28) a. Petin[poss] mamin[subj] portret devočki

‘a girl’s portrait by Mom belonging to Pete’
b. mamin[poss] Petin[subj] portret devočki
‘a girl’s portrait by Pete belonging to Mom’
384 11 Linear ordering of genitive adnominal dependents cosubordinated to a noun

To account for this fact, in addition to the modificative SSyntRel, three more SSyn-
tRels for possessive adjectives in Russian are needed: possessive-modificative, sub-
jectival-modificative, and objectival-modificative. As can be seen from (27)–(28), the
order of possessive adjectives with respect to the modified N is as follows:

ADJ←poss-modif— + ADJ←subj-modif— + ADJ←obj-modif— + N.

For instance:

Petin repinskij mamin portret lit. ‘Pete’s Repin’s Mom’s portrait’ =

‘Mom’s portrait by Repin owned by Pete’
NB The cooccurrence of two or more possessive adjectives is rare, so that, generally speaking,
it could be ignored. However, this case is interesting from a theoretical viewpoint.

The cooccurrence of possessive adjectives with “normal” ones is determined by

two general enough rules:

1) The possessive-modificative ADJ precedes all “normal” ADJs

(29) maminy[poss] dovoennye[external] poželtevšie[color] semejnye[kind] fotografii
Mom’s pre-war yellowed family photographs

2) The subjectival-modificative and objectival-modificative ADJs follow all “normal”

ADJs
(30) a. dovoennye[external] poželtevšie[color] maminy[subj] fotografii našego doma
pre-war yellowed Mom’s photographs of.our house
b. dovoennye[external] poželtevšie[color] maminy[obj] fotografii, sdelannye Petej
pre-war yellowed Mom’s photographs taken by.Pete

11.4.2 Comparison of both orderings: similarities and differences

Recall that cosubordinated NGENs follow the governing N, while cosubordinated

ADJs precede it. Therefore, the order of NGENs is a mirror image of that of ADJs. This
means that comparing both orders we speak in fact of the degree of closeness of
an NGEN or an ADJ to its governor N.
The ordering of cosubordinated NGENs and that of cosubordinated ADJs in
Russian are similar in the following two respects:

– The mutual ordering of Russian possessive ADJs (ADJ←poss-modif— + ADJ

←subj-modif— + ADJ←obj-modif— + N) is the same (of course, mirror-wise) as
the mutual ordering of the corres­ponding NGENs (that is, N + —obj-adnom→
 11.4 Ordering of cosubordinated NGENs vs. ordering of cosubordinated ADJs 385

NGEN+ —subj-adnom→NGEN + —poss-adnom→NGEN; Table 11.1). The possessive

adjectival modifier is the outermost, and the objectival adjectival modifier is
closer to the modified noun than the subjectival one. This is only natural,
since possessive ADJs are simply adjectivalizations of NGENs.
– The mutual ordering of repeatable NGENs (that is, qual-adnom and charact-
adnom NGENs) is the same as the mutual ordering of non-possessive ADJs,
since it is determined by the same hierarchical semantic classification of
the corresponding lexical units. This is also natural, since the closeness of
a modifier to its governor N is determined by the degree of semantic inher-
ence of the characteristic expressed: a more inherent characterization stands
closer to N.

The ordering of cosubordinated NGENs and that of cosubordinated ADJs in Russian

are different in the following respects:

– The cosubordinated NGENs are ordered syntactically—according to different

SSyntRels that link them to the governor, with several semantic “corrections”
imposed by their meaning.
– The cosubordinated ADJs are ordered semantically—according to their
meaning, with several syntactic “corrections” concerning possessive ADJs,
which are positioned in conformity with the governing SSyntRels.7

Chapter 11 is a slightly modified version of Iordanskaja & Mel’čuk 2019.

References
Abeillé, Anne. 1997. Fonction ou position objet ? Au gré des langues 11. 8–29.
Abeillé, Anne & Danièle Godard. 2000. French word order and lexical weight. In Robert Borsley
(ed.), The nature and function of syntactic categories [Syntax and Semantics 32], 325–360.
Ackerman, Farrell & Irina Nikolaeva. 2013. Descriptive typology and linguistic theory: the study
in the morphosyntax of relative clauses. Stanford, CA: CSLI Publications.
Aikhenvald, Alexandra. 2013. Possession and ownership: a cross-linguistic per­spective. In Alex-
andra Aikhenvald & Robert M. W. Dixon (eds.), Possession and ownership: A cross-linguis-
tic typology, 1–64. Oxford: Oxford University Press.
Aikhenvald, Alexandra, Robert M. W. Dixon & Masayuki Onishi (eds.). 2001. Non-canonical
marking of subjects and objects. Amsterdam/Philadelphia: John Benjamins.
Alexiadou, Artemis, Paul Law, André Meinunger & Chris Wilder (eds.). 2000. The syntax of relative
clauses. Amsterdam/Philadelphia: John Benjamins.
Alleton, Viviane. 1973. Grammaire du chinois. Paris: Presses Universitaires de France.
Álvarez Gonzales, Albert. 2012. Relative clauses and nominalizations in Yaqui. In Comrie &
Estrada-Fernández (eds.), Relative clauses in languages of the Americas. A typological
overview, 67–95. Amsterdam/Philadelphia: John Benjamins.
Andrews, Avery. 2001. Noncanonical A/S marking in Icelandic. In Aikhenvald et al. (eds.), Non-
canonical marking of subjects and objects, 85–111. Amsterdam/Philadelphia: John Ben-
jamins.
Andrews, Avery. 2007. Relative clauses. In Timothy Shopen (ed.), Language typology and syn-
tactic description, Vol. 2: Complex construction, 206–236. Cambridge: Cambride University
Press.
Anscombre, Jean-Claude. 2001. L’analyse de la construction En tout N par D. Leeman : quelques
remarques. Travaux de linguistique 42–43. 183–197.
Apresjan, Jurij. 2010. Vvedenie [Introduction]. In Apresjan et al., Teoretičeskie problemy russkogo
sintaksisa. Vzaimodejstvie grammatiki i slova­rja, 11–20. Moskva: Jazyki slavjanskix kulʹtur.
Apresjan, Jurij, Igorʹ Boguslavskij, Leonid Iomdin & Vladimir Sannikov. 2010. Teoretičeskie
problemy russkogo sintaksisa. Vzaimodejstvie grammatiki i slova­rja [Theoretical problems
of Russian syntax. Interaction between grammar and lexicon]. Moskva: Jazyki slavjanskix
kulʹtur.
Apresjan, Valentina. 2014. Syntactic idioms across languages: corpus evidence from Russian
and English. Russian Linguistics 38 (2). 187–203.
Auger, Pierre. 1965. Les modèles dans la science. Diogène 52. 3–15.
Beck, David. 1996. Subjecthood, agency, and topicality in Lushootseed. Toronto Working
Papers in Linguistics 15 (1). 1–29.
Beck, David. 2000. Semantic agents, syntactic subjects, and discourse topics: how to locate
Lushootseed sentences in space and time. Studies in Language 24 (2). 277–317.
Beck, David. 2016. Uniqueness and grammatical relations in Upper Necaxa Totonac. Linguistics
54 (1). 59–118.
Beck, David & Igor Mel’čuk. 2011. Morphological phrasemes and Totonacan verbal morphology.
Linguistics 49 (1). 175–228.
Benveniste, Émile. 1957–58. La phrase relative, problème de syntaxe générale. Bulletin de la
Société de Linguistique de Paris 53(1). 39–53. [Reprinted In Benveniste, Émile. 1966. Pro-
blèmes de linguistique générale, Paris: Gallimard, 208–222.]

https://doi.org/10.1515/9783110694765-013
388 References

Bhaskararao, Peri & Venkata Subbarao (eds.). 2004. Non-nominative subjects, vols. 1–2.
Amsterdam/Philadelphia: John Benjamins.
Bianchi, Valentina. 2000. Some issues in the syntax of relative determiners. In Artemis Alex-
iadou, Paul Law, André Meinunger & Chris Wilder (eds.), The syntax of relative clauses,
53–81. Amsterdam/Philadelphia: John Benjamins.
Bickel, Balthasar. 2011. Grammatical relations typology. In Jae Jung Song (ed.), The Oxford hand-
book of linguistic typology, 399–444. Oxford: Oxford University Press.
Bohnet, Bernd. 2007. The induction and evaluation of word order rules using cor­pora based on
the two concepts of topological models. In Language Generation and Machine Translation
[UCNLG+MT] workshop at MT Summit XI, Copen­hagen, 38–45.
Borščëv, Vladimir & Barbara Parti. 2011. Genitiv mery v russkom jazyke, tipy i sorta [The genitive
of measure in Russian: types and sorts]. In Igorʹ Boguslavskij, Leonid Iomdin & Leonid
Krysin (eds.), Slovo i jazyk. Sbornik statej k 80-letiju akademika Ju. D. Apresjana, 95–137.
Moscow: Jazyki slavjanskix kulʹtur.
Bresnan, Joan & Jane Grimshaw. 1978. The syntax of free relatives in English. Linguistic Inquiry
9 (3). 333–391.
Burgess, Clifford, Katarzyna Dziwirek & Donna Gerdts (eds.). 1995. Grammatic­al relations: theo-
retical approaches to empirical questions. Stanford, CA: CSLI Publications.
Caponigro, Ivano, Harold Torrence & Carlos Cisneros. 2013. Free relative clauses in two Mixtec
languages. International Journal of American Linguistics 79 (1). 41–96.
Carlson, Gregory. 1977. Amount relatives. Language 53 (3). 520–542.
Chang, Suk-Jin. 1996. Korean. Amsterdam/Philadelphia: John Benjamins.
Chappell, Hilary. 1986. The passive of bodily effect in Chinese. Studies in Lan­guage 10 (1). 271–
296.
Cho, Yongjoon. 2011. Multiple nominative construction in Korean and its second language
acquisition by adult English speakers. PhD Thesis. Los Angeles, CA: University of Southern
California.
Citko, Barbara. 2004. On headed, headless, and light-headed relatives. Natural Language &
Linguistic Theory 22 (1). 95–126.
Cole, Peter. 1987. The structure of internally headed relative clauses. Natural Language and Lin-
guistic Theory 5 (2). 277–302.
Cole, Peter & Jerrold Sadock (eds.). 1977. Grammatical relations [Syntax and Semantics 8]. New
York et al.: Academic Press.
Comrie, Bernard. 1998. Rethinking the typology of relative clauses. Language Design 1. 59–86.
Comrie, Bernard & Zarina Estrada-Fernández (eds.). 2012. Relative clauses in languages of the
Americas. A typological overview. Amsterdam/Philadelphia: John Benjamins.
Comrie, Bernard & Edward Keenan. 1979. Noun phrase accessibility revisited. Language 55 (3).
649–664.
Cooreman, Ann. 1994. A functional typology of antipassives. In Barbara Fox & Paul Hopper
(eds.), Voice: form and function, 49–87. Amsterdam/Philadelphia: John Benjamins.
Croft, William. 1994. The semantics of subjecthood. In Marina Yaguello (ed.), Subjecthood and
subjectivity, 29–75. Paris: Ophrys/Institut français du Royaume-Uni.
Dalrymple, Mary & Ronald Kaplan. 2000. Feature indeterminacy and feature resolution.
­Language 76 (4). 759–798.
Dalrymple, Mary & Irina Nikolaeva. 2011. Objects and information structure. Cambridge: Cam-
bridge University Press.
Danlos, Laurence, Benoît Sagot & Rosa Stern. 2010. Analyse discursive des incises de citation.
In Franck Neveu, Valelia Muni Toke, Jacques Durand, Thomas Klingler, Lorenza Mondada &
 References 389

Sophie Prévost (eds.), Congrès Mondial de Linguistique Française – CMLF2010, 2237–2254.

Paris: Institut de linguistique française. See also: https://www.youtube.com/
watch?v=JOHw7lX3Gu4
Davies, William & Stanley Dubinsky (eds.). 2001. Objects and other subjects. Grammatical func-
tions, functional categories and configurationality. Dordrecht et al.: Kluwer.
De Vries, Mark. 2018. Relative clauses. In Oxford Bibliographies in Linguistics, Oxford University
Press.
Dixon, Robert M. W. 1972. The Dyirbal language of North Queensland. Cam­bridge: Cambridge
University Press.
Dixon, Robert M. W. 2010. Basic linguistic theory, Vol. 2. Grammatical topics. New York: Oxford
University Press.
Downing, Laura & Al Mtenje. 2011. Prosodic phrasing of Chichewa relative clauses. Journal of
African Languages and Linguistics 32. 65–112.
Dryer, Matthew. 1983. Indirect objects in Kinyarwanda revisited. In David Perl­mutter (ed.), Rela-
tional grammar 1, 129–140. Chicago/London: The University of Chica­go Press.
Dryer, Matthew. 1997. Are grammatical relations universal? In Joan Bybee, John Haiman & Sandra
Thompson (eds.), Essays on language function and language type (dedicated to T. Givón),
115–143. Amsterdam/Philadelphia: John Benjamins.
Dryer, Matthew. 2013. Order of relative clause and noun. In Matthew Dryer & Martin Haspelmath
(eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evo-
lutionary Anthropology. See also http://wals.info/chapter/90.
Durie, Mark. 1985. A grammar of Acehnese on the basis of a dialect of North Aceh. Dordrecht/
Cinnaminson: Foris.
Durie, Mark. 1987. Grammatical relations in Acehnese. Studies in Language 11 (2). 365–399.
Durie, Mark. 1988. Preferred argument structure in an active language. Lingua 74. 1–25.
El Kassas, Dina & Sylvain Kahane. 2004. Modélisation de l’ordre des mots en arabe standard.
In JEP-TALN 2004, Fès.
Falk, Yehuda. 2006. Subjects and Universal Grammar. Cambridge: Cambridge University Press.
Farrell, Patrick. 2005. Grammatical relations. Oxford/New York: Oxford Univer­sity Press.
Filippova, Katja & Michael Strube. 2009. Tree linearization in English: improving language
model based approaches. In Proceedings of NAACL and HLT 2009, 225–228. Boulder, CO:
Association for Computational Linguistics.
Foley, William. 1991. The Yimas language of New Guinea. Stanford, CA: Stan­ford University Press.
Foley, William & Robert Van Valin. 1977. On the viability of the notion of sub­ject in Universal
Grammar. In Proceedings of the Third Annual Meeting of the Berkeley Linguistics Society
[BLS 3], 293–320.
Foolen, Ad. 2004. Expressive binominal NPs in Germanic and Romance lan­guages. In Günter
Radden & Klaus-Uwe Panther (eds.), Studies in Linguistic Motivation, 75–100. Berlin/New
York: Mouton de Gruyter.
Gaatone, David. 1988. Cette coquine de construction : remarques sur trois struc­tures affectives
du français. Travaux de linguistique 17. 159–176.
Gerdes, Kim. 2002. Topologie et grammaires formelles de l’allemand. PhD Thesis. Paris: Paris-7
University.
Gerdes, Kim, Bruno Guillaume, Sylvain Kahane & Guy Perrier. 2018. SUD or Surface-Syntactic
Universal Dependencies: An annotation scheme near-isomorphic to UD. Universal Depen-
dencies Workshop 2018, Brussels.
Gerdes, Kim & Sylvain Kahane. 2001. Word order in German: a formal depen­dency grammar
using a topological hierarchy. In Actes d’ACL 2001, Toulouse.
390 References

Gerdes, Kim & Sylvain Kahane. 2006. L’amas verbal au cœur d’une modélisation topologique
de l’ordre des mots. In Kim Gerdes & Claude Muller (eds.), Ordre des mots et topologie de
la phrase française [= Linguisticæ Investigationes] 29 (1). 75–89.
Gerdes, Kim & Sylvain Kahane. 2007. Phrasing it differently. In Leo Wanner (ed.), Selected lexical
and grammatical issues in the Meaning-Text theory, 297–335. Amsterdam/Philadelphia:
John Benjamins.
Gerdts, Donna & Cheong Youn. 1989. Non-nominative subjects in Korean. In Susumu Kuno et al.
(eds.), Harvard studies in Korean linguistics III: Proceed­ings of the 1989 Workshop on
Korean linguistics, 235–248. Seoul: Hanshin Publishing Company.
See also:http://www.sfu.ca/~gerdts/papers/GerdtsYounNonNominativeSubjects.pdf
Gil, David. 1994. The structure of Riau Indonesian. Nordic Journal of Linguistics 17. 179–200.
Givón, Talmy. 1990. Syntax. A functional typological introduction. Vol. II. Am­sterdam/Philadel-
phia: John Benjamins.
Givón, Talmy. 2012. Toward a diachronic typology of relative clause. In Bernard Comrie & Zarina
Estrada-Fernández (eds.), Relative clauses in languages of the Americas. A typological over-
view, 3–25. Amsterdam/Philadelphia: John Benjamins.
Glendenning, Peter Josef. 1983. Icelandic. London: Hodder and Stoughton.
Goldberg, Adele & Ray Jackendoff. 2004. The English resultative as a family of constructions.
Language 80 (3). 532–568.
Graffi, Giorgio. 2017. What are ‘Pseudo-Relatives’? In Roberta D’Alessandro, Gabriele Iannàccaro,
Diana Passino, Anna M. Thornton (eds.), Di tutti i colori. Studi linguistici per Maria Gross-
mann, 115–131. Utrecht: Utrecht University Repository.
Grimm, Scott. 2007. Case attraction in Ancient Greek. In Balder Ten Cate & Henk Zeevat (eds.),
Logic, Language, and Computation. 6th International Tbilisi Symposium on Logic, Lan-
guage, and Computation [TbiLLC 2005], 139–153. Berlin/Hei­delberg: Springer.
Grosu, Alexander. 1988. Pied Piping and the matching parameter. The Linguistic Review 6. 41–58.
Grosu, Alexander. 2002. Strange relatives at the interface of two millennia. GLOT International
6 (6). 145–167.
Grosu, Alexander. 2012. Towards a more articulated typology of internally headed relative con-
structions: the semantics connection. Language and Linguistics Compass 6 (7). 447–476.
Grosu, Alexander & Koji Hoshi. 2016. Japanese internally headed relatives: their distinctness
from potentially homophonous constructions. Glossa: A Journal of General Linguistics 1 (1).
1–31.
Grosu, Alexander & Fred Landman. 1998. Strange relatives of the third kind. Natural Language
Semantics 6. 125–170.
Han, Na-Rae. 2004. Korean null pronouns: classification and annotation. In Proceedings of the
ACL 2004 Workshop on Discourse Annotation, 33–50.
Han, Na-Rae. 2006. Korean zero pronouns: analysis and resolution. PhD Thesis. Philadelphia,
PA: University of Pennsylvania. See also: http://citeseerx.ist.psu.edu/viewdoc/download?
doi=10.1.1.126.3973&rep=rep1&type=pdf
Handschuh, Corinna. 2014. The typology of marked-S languages. Berlin: Lan­guage Science Press.
Harris, Alice. 1981. Georgian syntax. A study in relational grammar. Cambridge et al.: Cambridge
University Press.
Hashimoto, Mantaro. 1988. The structure and typology of the Chinese passive con­struction.
In Masayoshi Shibatani (ed.), Passive and voice, 329–354. Amsterdam/ Philadelphia: John
Benjamins.
 References 391

Haspelmath, Martin. 1991. On the questions of deep ergativity: the evidence from Lezgian.
Papiere zur Linguistik 44/45 (1/2). 5–27.
Haspelmath, Martin. 1993. A grammar of Lezgian. Berlin/New York: Mouton de Gruyter.
Haspelmath, Martin. 1997. Indefinite pronouns. Clarendon Press: Oxford.
Haspelmath, Martin. 2016. The serial verb construction: comparative concept and cross-linguistic
generalizations. Language and Linguistics 17 (3). 291–319.
Hatcher, Anna. 1960. An introduction to an аnalysis of English noun compounds. Word 16 (3).
356–373.
He, Wei, Haifeng Wang, Yuqin Guo & Ting Liu. 2009. Dependency based Chin­ese sentence reali­
zation. In Joint Conference of the 47th Annual Meeting of the Association for Computational
Linguistics and 4th International Joint Confer­ence on Natural Language Processing (ACL-
IJCNLP 2009), Singapore, 809–817.
Hiraiwa, Ken, George Akanlig-Pare, Samuel Atintono, Adams Bodomo, Komlan Essizewa &
Fusheini Hjudu. 2017. A comparative syntax of internally-headed relative clauses in Gur.
Glossa: a journal of general linguistics 2 (1), 27. 1–30, DOI: https://doi.org/10.5334/gjgl.40
Horvath, Julia & Alexander Grosu. 1987. On the notion “head”: evidence from free relatives and
interrogatives. Theoretical Linguistics 14 (1). 35–64.
Huang, C.-T. James. 1999. Chinese passive in comparative perspective. Tsing Hua Journal of
Chinese Studies 29. 423–509.
Huang, C.-T. James, Y.-H. Audrey Li & Yafei Li. 2008. The syntax of Chinese. Cambridge et al.:
Cambridge University Press.
Iomdin, Leonid. 2005. A hypothesis of two syntactic starts. In Jurij Apresjan & Leonid Iomdin
(eds.), Vostok – Zapad: Vtoraja meždunarodnaja konferencija po modeli Smysl – Text, 165–
175. Moskva: Jazyki slavjanskoj kulʹtury.
Iomdin, Leonid. 2007. Russian idioms formed with interrogative pronouns and their properties.
In Kim Gerdes, Tilmann Reuther & Leo Wanner (eds.), Pro­ceedings of the Third International
Conference on Meaning-Text Theory, 179–189. Mün­chen/Wien: Wiener Slawistischer Alma-
nach [Sonderband 69].
Iomdin, Leonid. 2010a. Gipoteza o dvux sintaksičeskix načalax [The hypothesis of two syntactic
starts]. In Jurij Apresjan, Igorʹ Boguslavskij, Leonid Iomdin & Vladimir Sannikov. 2010.
Teoretičeskie problemy russkogo sintaksisa. Vzaimodejstvie grammatiki i slova­rja [Theo-
retical problems of Russian syntax. Interaction between grammar and lexicon], 129–140.
Moskva: Jazyki slavjanskix kulʹtur.
Iomdin, Leonid. 2010b. Sintaksičeskie frazemy: meždu leksikoj i sintaksisom [Syntactic phrase-
mes: between lexicon and syntax]. In Jurij Apresjan, Igorʹ Boguslavskij, Leonid Iomdin &
Vladimir Sannikov. 2010. Teoretičeskie problemy russkogo sintaksisa. Vzaimodejstvie
grammatiki i slova­rja [Theoretical problems of Russian syntax. Interaction between
grammar and lexicon], 141–190. Moskva: Jazyki slavjanskix kulʹtur.
Iomdin, Leonid. 2010c. Sintaksičeskie otnošenija [Syntactic relations]. In Jurij Apresjan, Igorʹ
Boguslavskij, Leonid Iomdin & Vladimir Sannikov. 2010. Teoretičeskie problemy russkogo
sintaksisa. Vzaimodejstvie grammatiki i slova­rja [Theoretical problems of Russian syntax.
Interaction between grammar and lexicon], 21–43. Moskva: Jazyki slavjanskix kulʹtur.
Iordanskaja, Lidija. 1986. Propriétés sémantiques des verbes promoteurs de la négation en
français. Lingvisticæ Investigationes 10 (2). 345–380.
Iordanskaja, Lidija. 2000. Sopodčinenie prilagatelʹnyx v russkom jazyke (po sle­dam Vendlera)
[Cosubordination of adjectives in Russian (following in Vend­ler’s footsteps)]. In Leonid
392 References

Iomdin & Leonid Krysin (eds.), Slovo v tekste i v slovare. Sbornik statej k 70-letiju aka-
demika Ju. D. Apresjana, 379–390. Moskva: Jazyki russkoj kulʹtury.
Iordanskaja, Lidija. 2003. L’ordonnancement des adjectifs co-dépendants en russe. In Sylvain
Kahane & Alexis Nasr (eds.), Proceedings of the First Inter­national conference on Meaning-
Text theory, 159–169. Paris: École Normale Supérieure.
Iordanskaja, Lidija. 2007. Lexicographic definition and lexical cooccurrence: pre­suppositions as
a ‘No-Go’ zone for the meaning of modifiers. In Kim Gerdes, Tilmann Reuther & Leo Wanner
(eds.), Meaning-Text Theory 2007 [MTT’07], 209–218. München/Wien: Wiener Slawistischer
Almanach.
Iordanskaja, Lidija & Igor Mel’čuk. 1981. On a class of Russian verbs which can introduce Direct
Speech. In Per Jakobsen and Helen Krag (eds.), The Slavic Verb, 51–66. Copenhague:
Rosenkilde & Bagger.
Iordanskaja, Lidija & Igor Mel’čuk. 1995. Traitement lexicographique de deux connecteurs­
­textuels du français contemporaIn ˹en fait˺ vs ˹en réalité˺. In Hava Bat-Zeev Shyldkrot &
Lucien Kupferman (eds.), Tendances récentes en linguistique française et générale (volume
dédié à David Gaatone), 211–236. Amster­dam/Philadelphia: John Benjamins.
Iordanskaja, Lidija & Igor Mel’čuk. 2009a. Establishing an inventory of surface-syntactic rela-
tions: valence-controlled surface dependents of the verb in French. In Alain Polguère & Igor
Mel’čuk (eds.), Dependency in linguistic descrip­tion, 151–234. Amsterdam/Philadelphia:
John Benjamins.
Iordanskaja, Lidija & Igor Mel’čuk. 2009b. Semantics of the Russian conjunction poka ‘while,
before, until’. In Tilman Berger, Markus Giger, Sibylle Kurt & Imke Mendoza (eds.), Von
grammatischen Kategorien und sprachlichen Welt­bildern – Die Slavia von der Sprachge-
schichte bis zur Politsprache. Festschrift für Daniel Weiss zum 60. Geburtstag, 233–262.
München/Wien: Wiener Slawistischer Almanach.
Iordanskaja, Lidija & Igor Mel’čuk. 2015. Ordering of simple clauses in an English complex sen-
tence. Rhema/Rema 4. 17–59.
Iordanskaja, Lidija & Igor Mel’čuk. 2017. Le mot français dans le lexique et dans la phrase. Paris:
Hermann.
Iordanskaja, Lidija & Igor Mel’čuk. 2019. Semantics in syntax: linear ordering of genitive adnomi­
nal dependents cosubordinated to a noun in Russian. Vopropsy jazykoznanija № 4. 33–46.
Iordanskaja, Lidija & Alain Polguère. 1988. Semantic processing for text genera­tion. In Proceed-
ings of the First International Computer Science Confer­ence ’88. Artificial Intelligence:
Theory and Applications, 310–318. Hong Kong.
Iordanskaja, Lidija & Alain Polguère. 2005. Hooking up syntagmatic lexical functions to lexico-
graphic definitions. In Jurij Apresjan and Leonid Iomdin (eds.), East-West Encounter,
Second International Conference on Meaning-Text Theory [MTT’05], 170–186. Moskva:
Jazyki slavjanskoj kulʹtury.
Jackendoff, Ray. 1984. On the phrase the phrase ‘the phrase’. Natural Language and Linguistic
Theory 2. 25–37. [A more recent corrected version: Jackendoff, Ray. 2010. Meaning and the
Lexicon. The Parallel Architecture. 1975–2010, 327–341. Oxford: Oxford University Press.]
Jackendoff, Ray. 2002. English particle constructions, the lexicon, and the auto­nomy of syntax.
In Nicole Dehé, Ray Jackendoff, Andrew McIntyre & Silke Urban, Verb-Particle Explorations,
67–94. Berlin: Mouton de Gruyter. [A more recent corrected version: Jackendoff, Ray. 2010.
Meaning and the Lexicon. The Parallel Architecture. 1975–2010, 226–249. Oxford: Oxford
University Press.]
 References 393

Jang, Youngjun. 1997. Multiple subject construction in Korean: a functional explanation. Kansas
Working Papers in Linguistics 22 (1). 25–40.
Jaxontov, Sergej. 1965. Drevnekitajskij jazyk [Ancient Chinese]. Moskva: Nauka.
Jaxontov, Sergej. 1974. Nekotorye passivnye konstrukcii v kitajskom jazyke [Some passive con-
structions in Chinese]. In Aleksandr Xolodovič (ed.), Tipologija passivnyx konstrukcij.
Diatezy i zalogi, 195–202. Leningrad: Nauka.
Jones, Michael. 1993. Sardinian Syntax. London/New York: Routledge.
Kachru, Yamuna. 2006. Hindi. Amsterdam/Philadelphia: John Benjamins.
Kahane, Sylvain. 1997. Bubble trees and syntactic representations. In Tilmann Becker & Hans-
Ulrich Krieger (eds.), Proceedings of the 5th Meeting of the mathematics of language (MOL 5),
70–76. Saarbrücken.
Kahane, Sylvain. 2001. A fully lexicalized grammar for French based on Mean­ing-Text theory. In
Alexander Gelbukh (ed.), Computational linguistics and intelligent text processing, Second
International conference, CICLing 2001, Mexico-City, Mexico, February 18–24, 2001, Pro-
ceedings [Lecture notes in computer science 2004], 18–31. Berlin/Heidelberg: Springer.
Keenan, Edward. 1976. Towards a universal definition of “subject.” In Charles Li (ed.), Subject
and topic, 305–333. New York et al.: Academic Press.
Keenan, Edward. 1985. Relative clauses. In Timothy Shopen (ed.), Language typology and syn-
tactic description. II. Complex constructions, 141–170. Cambridge et al.: Cambridge Univer-
sity Press.
Keenan, Edward & Bernard Comrie. 1977. Noun phrase accessibility and Univer­sal Grammar.
Linguistic Inquiry 8 (1). 63–99.
Kholodilova, Maria. 2013. Inverse attraction in Ingrian Finnish relative clauses. Linguistica
Uralica 49 (2). 96–116.
Kibrik, Aleksandr. 1975. Nominativnaja/èrgativnaja konstrukcii i logičeskoe uda­renie v arčinskom
jazyke [Nominative/ergative construction and logical accent in Archi]. In Issledovanija po
strukturnoj i prikladnoj lingvistike, 54–62. Moskva: Izdatelʹstvo Moskovskogo universiteta.
Kibrik, Aleksandr. 1977. Opyt strukturnogo opisanija arčinskogo jazyka. III. Dinamičeskaja gram-
matika [Towards a structural description of Archi. III. A dynamic grammar]. Moskva:
Izdatelʹstvo Moskovskogo universiteta.
Kibrik, Aleksandr. 1997. Beyond subject and object: towards a comprehensive relational typology.
Linguistic Typology 1. 279–346.
Kibrik, Aleksandr. 2001. Grammatical relations. In Neil Smelser & Paul Baltes (eds.), International
encyclopedia of the social and behavioral sciences, 6342–6348. Amsterdam: Pergamon
Press.
Kibrik, Aleksandr. 2003. Konstanty i peremennye jazyka [Constants and variables in language].
Sankt-Peterburg: Aletejja.
Kim, Hong-Bok & Peter Sells. 2010. Oblique case marking on core arguments in Korean. Studies
in Language 34 (3). 602–635.
Kimenyi, Alexandre. 1980. Relational grammar of Kinyarwanda. Berkeley, CA: UCLA Press.
Koul, Omkar. 2008. Modern Hindi grammar. Springfield, VA: Dunwoody Press.
Kovtunova, Irina. 1976. Sovremennyj russkij jazyk. Porjadok slov i aktualʹnoe čle­nenie
predloženija [Modern Russian. Word order and “actual division” of a sentence]. Moskva:
Prosveščenie.
Kovtunova, Irina. 1980. Porjadok slov [Word order]. In Natalʹja Švedova (ed.), Russkaja gramma-
tika II. Moskva: Nauka.
394 References

Kozinskij, Icxak. 1983. O kategorii “podležaščee” v russkom jazyke [“Syntactic subject” category
in Russian], Predvaritelʹnye publikacii Problemnoj gruppy po èksperimentalʹnoj i prikladnoj
lingvistike, vypusk 156. Moskva: Institut rus­skogo jazyka AN SSSR.
Kroeger, Paul. 1993. Phrase structure and grammatical relations in Tagalog. Stanford, CA: CSLI
Publications.
Kroeger, Paul. 2007. McKaughan’s analysis of Philippine voice. In Loren Billings & Nelleke
Goudswaard (eds.), Piakandatu ami Dr. Howard P. McKaughan, Manila: Linguistic Society
of the Philippines and SIL Philippines, 41–46. See http://gamma.sil.org/acpub/repository/
Kroeger-McK-voice-final.pdf
Kručinina, Irina. 1968. Konstrukcija s mestoimeniem kotoryj v sovremennom russkom jazyke
[Construction with the pronoun kotoryj ‘which’ in Modern Russian]. Voprosy jazykoznanija
№ 2. 82–88.
Kuno, Susumu. 1973. The structure of the Japanese language. Cambridge, MA/London: The MIT
Press.
Kuroda, Sige-Yuki. 1968. English relativization and certain related problems. Language 48 (2).
244–266.
Lakoff, George. 1974. Syntactic amalgams. In Chicago Linguistic Society: Papers from the 10th
Regional Meeting, 321–344.
Lazard, Gilbert. 1994. L’actance. Paris: Presses Universitaires de France.
Lee, Iksop & S. Robert Ramsey. 2000. The Korean Language. Albany, NY: SUNY Press.
Lehmann, Christian. 1984. Der Relativsatz. Tübingen: Gunter Narr.
Lehmann, Christian. 1986. On typology of relative clause. Linguistics 24 (4). 663–680.
Lepschy, Giulio. 1994. The status of subject in linguistic theory. Closing remarks. In Marina
Yaguello (ed.), Subjecthood and subjectivity, 275–279. Paris: Ophrys/Institut français du
Royaume-Uni.
Levi, Judith. 1978. The syntax of complex nominals. New York: Academic Press.
Li, Charles & Sandra Thompson. 1981. Mandarin Chinese. A functional refer­ence grammar.
Berkeley, CA et al.: University of California Press.
Li, Charles (ed.). 1976. Subject and topic. New York et al.: Academic Press.
Lister-Turner, Robert & J. B. Clark. 1931. A grammar of the Motu language of Papua. Sydney:
Pettifer.
Liu, Na. 2016. The structures of Chinese long and short bei passives revisited. Language and
Linguistics 17 (6). 857–889.
MacDonald, Danica & Nicholas Welch. 2009. Topic, focus and double subjects in Korean. In Pro-
ceedings of the 2009 Annual Conference of the Canadian Linguistic Association, 1–14.
Mahajan, Anoop. 2000. Relative asymmetries and Hindi correlatives. In Artemis Alexiadou, Paul
Law, André Meinunger & Chris Wilder (eds.), The syntax of relative clauses, 201–229.
Amsterdam/Philadelphia: John Benjamins.
Maling, Joan & Soowon Kim. 1992. Case assignment in the inalienable posses­sion construction
in Korean. Journal of East Asian Linguistics 1. 37–68.
Marlett, Stephen. 2012. Relative clauses in Seri. In Bernard Comrie & Zarina Estrada-Fernández
(eds.), Relative clauses in languages of the Americas. A typological overview, 213–241.
Amsterdam/Philadelphia: John Benjamins.
Marneffe, Marie-Catherine de, Timothy Dozat, Natalia Silveira, Katri Haverinen, Filip Ginter,
Joakim Nivre & Christopher Manning. 2014. Universal Stanford Depen­dencies: a cross-­
linguistic typology. In Proceedings of the Ninth International Conference on Language
resources and Evaluation, 4585–4592. European Languages Resources Association (ELRA).
 References 395

Marneffe, Marie-Catherine de & Christopher Manning. 2008/2015. Stanford typed dependencies

manual. See http://nlp.stanford.edu/software/dependencies_manual.pdf
Maslova, Ekaterina. 2003. A grammar of Kolyma Yukagir. Berlin: Mouton de Gruyter.
Matsumoto, Yoshiko, Bernard Comrie & Peter Sells (eds.). 2017. Noun-modifying clause con-
structions in languages of Eurasia. Rethinking theoretical and geo­graphical boundaries.
Amsterdam/Philadelphia: John Benjamins.
Mel’čuk, Igor. 1964. Obobščenie ponjatija frazeologizma (morfologičeskie fraze­ologizmy) [A gen-
eralization of the concept “phraseologism” (morphological phraseologisms)]. In Leonid
Rojzenzon (ed.), Materialy konferencii “Aktualʹnye voprosy saovremennogo jazykoznanija
i lingvističeskoe nasledie E. D. Poliva­nova”, tom I, 89–90. Samarkand: Samarkandskij
­Gosudartsvennyj Universitet.
Mel’čuk, Igor. 1965. Porjadok slov pri avtomatičeskom sinteze russkogo teksta (predvaritelʹnoe
soobščenie) [Word order in automatic synthesis of Russian texts (preliminary communica-
tion)]. Naučno-texničeskaja informacija № 12. 36–44. [A French translation: I. Mel’čuk,
1967: T.A. Informations 2. 65–84.]
Mel’čuk, Igor. 1974. Opyt teorii lingvističeskix modelej «Smysl⇔Tekst». Seman­tika, sintaksis
[Outline of a theory of linguistic Meaning-Text models. Semant­ics, syntax]. Moskva: Nauka.
[Reprint: 1995.]
Mel’čuk, Igor. 1982. Lexical functions in lexicographic description. Proceedings of the VIIIth
Annual Meeting of the Berkeley Linguistics Society [BLS 8], 427–444.
Mel’čuk, Igor. 1987. Un affixe dérivationnel et un phrasème syntaxique du russe moderne. Essai
de description formelle. Revue des études slaves 53 (3). 631–648.
Mel’čuk, Igor. 1988. Dependency syntax: theory and practice. Albany, NY: State University of New
York Press.
Mel’čuk, Igor. 1992–2000. Cours de morphologie générale, vol. 1–5. Montréal: Les Presses de
l’Université de Montréal/Paris: C.N.R.S.
Mel’čuk, Igor. 2001. Communicative organization in natural language. The semantic-communi-
cative structure of sentences. Amsterdam/Philadelphia: John Benjamins.
Mel’čuk, Igor. 2004. Actants in semantics and syntax. I/II: Actants in seman­tics/in syntax.
­Linguistics 42 (1). 1–66; 42 (2). 247–291.
Mel’čuk, Igor. 2006a. Aspects of a theory of morphology. Berlin/New York: Mouton de Gruyter.
Mel’čuk, Igor. 2006b. Calculus of possibilities as a technique in linguistic typology. In Felix
Ameka, Alan Dench & Nicolas Evans (eds.), Catching lan­guage. The standing challenge of
grammar writing, 171–205. Berlin/New York: Mouton de Gruyter.
Mel’čuk, Igor. 2007. Lexical functions. In Harald Burger, Dmitry Dobrovol’­skij, Peter Kühn & Neal
Norrick (eds.), Phraseology. An international handbook of contemporary research, 119–131.
Berlin/New York: Mouton de Gruyter.
Mel’čuk, Igor. 2009a. Dependency in natural language. In Alain Polguère & Igor Mel’čuk (eds.),
Dependency in linguistic descrip­tion, 1–110. Amsterdam/Philadelphia: John Benjamins.
Mel’čuk, Igor. 2009b. Functional linguistic models: a step forward in the study of Man. BULAG
[= Bulletin de la linguistique appliquée et générale] 33 [= Inter­national symposium on
data and sense mining, machine translation and con­trolled languages, and their applica-
tion to emergencies and safety critical domains], 1–10. Besançon: Presses Universitaires
de Franche-Compté.
Mel’čuk, Igor. 2011. Word order in Russian. In Igorʹ Boguslavskij, Leonid Iomdin & Leonid Krysin
(eds.), Slovo i jazyk (Sbornik statej k vosʹmidesjatiletiju akademika Ju. D. Apresjana), 499–
525. Moskva: Jazyki slavjanskix kulʹtur.
396 References

Mel'čuk, Igor. 2012a. Bi-nominative sentences in Russian. In Veronika Makarova (ed.), Russian
Language Studies in North America. New Perspectives from Theoretical and Applied Lin-
guistics, 85–105. London: Anthem.
Mel’čuk, Igor. 2012b. Jazyk: ot smysla k tekstu [Language: from meaning to text]. Moscow: Jazyki
slavjanskoj kulʹtury.
Mel’čuk, Igor. 2012c. Mestoimennye vyraženija s imenem čertyxatelʹnym tipa [Ona uexala] čërt
znaet kuda i im podobnye v russkom jazyke [Pronominal expressions with a blasphemous
noun of the type [Ona uexala] čërt znaet kuda ‘She left Devil knows where’ and similar ones
in Russian]. Russkij jazyk v naučnom osveščenii 2 (24). 5–22.
Mel’čuk, Igor. 2012d. Phraseology in the language, in the dictionary, and in the computer. Year-
book of Phraseology 3. 31–56.
Mel’čuk, Igor. 2012–2015. Semantics: from meaning to text. Vols 1–3. Amster­dam/Philadelphia:
John Benjamins.
Mel’čuk, Igor. 2014a. Dependency in language. In Kim Gerdes, Eva Hajičová & Leo Wanner
(eds.), Dependency linguistics. Recent advances in linguistic theory using dependency
structures, 1–33. Amsterdam/Philadelphia: John Benjamins.
Mel’čuk, Igor. 2014b. Syntactic subject: syntactic relations, once again. In Vladi­mir Plungjan
with Mixail Danièlʹ, Ekaterina Ljutikova, Sergej Tatevosov & Olʹga Fëdorova (eds.), Jazyk.
Konstanty. Peremennye. Pamjati Aleksandra Ev­genʹeviča Kibrika [Language. Constants.
Variables. To the memory of Alexan­der Kibrik], 169–216. Sankt-Peterburg: Aletejja.
Mel’čuk, Igor. 2014c. The East/South-East Asian answer to the European passive. In Sergej
­Dmitrenko & Natalʹja Zaika (eds.), Acta Linguistica Petropolitana, 2014/10 (3) [= Festschrift
Xrakovskij], 451–472. Sankt-Peterburg: Nauka.
Mel’čuk, Igor. 2015a. Clichés, an understudied subclass of phrasemes. Yearbook of Phraseology 6.
55–86.
Mel’čuk, Igor. 2015b. Multiple “subjects” and multiple “objects” in Korean. Lan­guage Research
51 (3). 485–516.
Mel’čuk, Igor. 2015–2016. A general inventory of surface-syntactic relations in world languages.
Part one. Moscow Journal of Linguistics 17(2). 75–103; Part two, Moscow Journal of Linguis-
tics 18 (1). 94–120.
Mel’čuk, Igor. 2016. Language: from meaning to text. Moskva/Boston, MA: Academic Studies
Press.
Melʹčuk, Igorʹ. 2017. KAK…, TAK I…: čto èto za? [KAK…, TAK I… ‘as..., so also…’: what kind of stuff
is this?]. Russkij jazyk v naučnom osveščenii 1 (33). 67–85.
Mel’čuk, Igor. 2018a. Genitive adnominal dependents in Russian: surface-syn­tactic relations in
the N→NGEN phrase. Voprosy jazykoznanija № 4. 25–46.
Mel’čuk, Igor. 2018b. Les prépositions zéro en français. Lingvisticæ Investiga­tiones 41 (2). 269–
283.
Mel’čuk, Igor. 2018c. “Wordlets”: One of Zholkovsky’s major contributions to the notion of deep-
syntactic structure. In Dennis Ioffe, Marcus Levitt, Joe Peschio & Igor Pilshchikov (eds.),
A/Z: Essays in Honor of Alexander Zhol­kovsky, 350–360. Boston: Academic Studies Press.
Mel’čuk, Igor & Nikolaj Pertsov. 1987. Surface syntax of English. A formal model within the
Meaning-Text framework. Amsterdam: John Benjamins.
Mel’čuk, Igor & Alain Polguère. 2007. Lexique actif du français. L’apprentissage du vocabulaire
fondé sur 20 000 dérivations sémantiques et collocations du français. Brussels: De Boeck.
Mel’čuk, Igor & Alexander Zholkovsky. 1984. Explanatory Combinatorial Dictionary of Modern
Russian, Wien: Wiener Slawistischer Almanach.
 References 397

Mel’čuk, Igor & Aleksandr Žolkovskij. 2016. Tolkovo-kombnatornyj slovarʹ rus­skogo jazyka. 2-oe
izdanie, ispravlennoe [Explanatory and combionatorial dictionary of Russian. 2nd edition,
corrected]. Moskva: Jazyki slavjanskix kulʹtur.
Mel’čuk, Igor et al. 1984–1999. Dictionnaire explicatif et combinatoire du français contemporain :
Recherches lexico-sémantiques I–IV, Montréal: Les Presses de l’Université de Montréal.
Melis, Ludo. 2000. L’infinitif de narration comme prédication seconde. Langages 127. 36–48.
Milićević, Jasmina. 2007. La paraphrase. Modélisation de la paraphrase langagière. Bern: Peter
Lang.
Milićević, Jasmina. 2009. Linear placement of Serbian clitics in a syntactic dependency frame-
work. In Alain Polguère & Igor Mel’čuk (eds.), Dependency in linguistic descrip­tion, 235–
276. Amsterdam/Philadelphia: John Benjamins.
Mille, Simon & Leo Wanner. 2008. Making text resources accessible to the reader: the case of
patent claims. In Proceedings of the 6th Intern. Conference on Language Resources and
Evaluation [LREC’08], 1394–1400. Marrakesh. See also http://www.lrec-conf.org/proceed-
ings/lrec2008/
Mixeev, Mixail. 2000. Žizni myšʹja begotnja ili toska tščetnosti? (O metaforičes­koj konstrukcii s
roditelʹnym padežom) [Mouse scurrying of life or pining of vanity? The metaphorical con-
struction with the genitive]. Voprosy jazykozna­nija № 2. 47–69.
Nagel, Ernest, Patrick Suppes & Alfred Tarski (eds.). 1962. Logic, methodo­logy and philosophy
of science. Stanford, CA: Stanford University Press.
Nichols, Johanna, Gilbert Rappaport & Alan Timberlake. 1980. Subject, topic, and control in
Russian. In Bruce Caron et al. (eds.), Proceedings of the Sixth Annual Meeting of the Berke-
ley Linguistics Society [BLS 6], 372–386.
Niu, Ruochen & Timothy Osborne. 2019. Chunks are components: A dependency grammar
approach to the syntactic structure of Mandarin. Lingua 224. 60–83.
O’Grady, William. 1991. Categories and case. The sentence structure of Korean. Amsterdam/
Philadelphia: John Benjamins.
O’Grady, William. 1992. On the status of ha-ta in the multiple complement structures. In 1992
Seoul International Conference on Linguistics [SICOL’92], 238–247. Seoul: The Linguistic
Society of Korea.
Otsuka, Yuko. 2000. Ergativity in Tongan. PhD Thesis. Oxford: University of Oxford.
See http://www2.hawaii.edu/~yotsuka
Otsuka, Yuko. 2010. DP ellipsis in Tongan: is syntactic ergativity real? Natural Language and
Linguistic Theory 28. 315–342.
Padučeva, Elena. 1974. O semantike sintaksisa [On semantics of syntax]. Mos­kva: Nauka.
Palmer, Frank. 1994. Grammatical roles and relations. Cambridge: Cambridge University Press.
Paris, Marie-Claude. 1998. Syntaxe et sémantique de quatre marqueurs de transiti­vité en chinois
standard: ba, bei, jiao et rang. In André Rousseau (ed.), La transitivité, 357–370. Villeneuve
d’Ascq: Presses Universitaires de Septentrion.
Partee, Barbara & Vladimir Borschev. 2000. Possessives, favorite, and сoercion. In Rebecca Daly
& Anastasia Riehl (eds.), Proceedings of ESCOL 99, 173–190. Ithaca, NY: Cornell University.
Patejuk, Agnieszka & Adam Przepiórkowski. 2019. Coordination of unlike gramma­tical functions.
In https://syntaxfest.github.io/syntaxfest19/proceedings/papers/paper_3.pdf
Peranteau, Paul, Judith Levi & Gloria Phares (eds.). 1972. The Chicago Which Hunt. Papers from
the Realtive Clause Festival. Chicago: Chicago Linguistic Society.
Plank, Frans (ed.). 1984. Objects. Towards a theory of grammatical relations. London et al.: Aca-
demic Press.
398 References

Platzak, Christer. 2000. A Complement-of-Nº account of restrictive and non-restrictive relatives.

In Artemis Alexiadou, Paul Law, André Meinunger & Chris Wilder (eds.), The syntax of relative
clauses, 265–308. Amsterdam/Philadelphia: John Benjamins.
Poiret, Rafaël & Haitao Liu. 2019. Les dépendants adnominaux prépositionnels en français :
Relations syntaxiques de surface dans le syntagme N→SP. Français Moderne 2. 259–280.
Polguère, Alain. 2014. Rection nominale : retour sur les constructions évalua­tives. Travaux de
linguistique 68. 83–102.
Polguère, Alain & Igor Mel’čuk (eds.). 2009. Dependency in linguistic descrip­tion. Amsterdam/
Philadelphia: John Benjamins.
Pollard, Carl & Ivan Sag. 1992. Anaphors in English and the scope of binding theory. Linguistic
Inquiry 23 (2). 261–303.
Pouplier, Marianne. 2003. Referential subject and object gaps in Modern Iceland­ic. Nordlyd 31 (2).
356–371.
Quirk, Randolph, Sydney Greenbaum, Geoffrey Leech & Jan Svartvik. 1991. A comprehensive
grammar of the English language. London/New York: Long­man.
Raxilina Ekaterina. 2010. Konstrukcija s russkim roditelʹnym i eë formalʹnaja interpretacija [The
construction with the Russian genitive and its formal interpretation]. In Ekaterina Raxilina
(ed.), Lingvistika konstrukcij [Construc­tion linguistics], 247–285. Moskva: Azbukovnik.
Rebuschi, Georges. 1978. Cas et fonction sujet en basque. Verbum 1 (1). 69–98.
Rebuschi, Georges. 1981. Quelques problèmes de syntaxe anglaise pour la gram­maire dite
“relationnelle”. Verbum 4 (1). 85–119.
Rebuschi, Georges. 1982. Structure de l’énoncé en basque. [Collection ERA 642, numéro spécial.]
Paris: Université Paris 7.
Rebuschi, Georges. 1986. Diathèse et (non-)configuralité : l’exemple du basque. Actance 2.
175–208.
Ren, Xiaobo. 1993. Syntaxe des constructions passives en chinois. Paris: Lan­gages Croisés.
Roberts, John. 1987. Amele. London et al.: Crook Helm.
Roberts, John. 1988. Switch-reference in Papuan languages: a syntactic or extra­syntactic
device? Australian Journal of Linguistics 8 (1). 75–117.
Roberts, John. 2001. Impersonal constructions in Amele. In Alexandra Aikhenvald, Robert
M. W. Dixon & Masayuki Onishi (eds.), Non-canonical marking of subjects and objects,
201–250. Amsterdam/Philadelphia: John Benjamins.
Samvelian, Polett. 2012. Grammaire des prédicats complexes. Paris: Lavoisier.
Sandfeld, Kristian. 1965. Syntaxe du français contemporain. Les propositions subordonnées.
Genève: Droz.
Sannikov, Vladimir. 1963. Mesto rasprostranënnogo opredelenija po otnošeniju k opredeljae-
momu slovu v russkoj fraze [The position of an expanded adjectival modifier with respect
to the modified word in a Russian sentence]. Voprosy jazykoznanija № 1. 124–130.
[Reprinted 2008 in V. Sannikov, Russkij sintaksis v semantiko-pragmatičeskom prostrans-
tve, 585–598. Moskva: Jazyki slavjanskix kulʹtur.]
Sannikov, Vladimir. 2010a. Proleptičeskie i blizkie k nim konstrukcii [Proleptic and similar con-
structions]. In Jurij Apresjan, Igorʹ Boguslavskij, Leonid Iomdin & Vladimir Sannikov. 2010.
Teoretičeskie problemy russkogo sintaksisa. Vzaimodejstvie grammatiki i slova­rja [Theo-
retical problems of Russian syntax. Interaction between grammar and lexicon], 113–129.
Moskva: Jazyki slavjanskix kulʹtur.
Sannikov, Vladimir. 2010b. Konstrukcii s toždestvennymi slovoformami [Con­structions with
identical wordforms]. In Jurij Apresjan, Igorʹ Boguslavskij, Leonid Iomdin & Vladimir
 References 399

Sanni­kov, Teoretičeskie problemy russkogo sintaksisa. Vzaimodejstvie grammatiki i slova­rja

[Theoretical problems of Russian syntax. Interaction between grammar and lexicon], 191–208.
Moskva: Jazyki slavjanskix kulʹtur.
Schütze, Carson. 2001. On Korean “case stacking”: The varied functions of the particles ka and lul.
The Linguistic Review 19. 193–292.
Schwartz, Arthur. 1976. On the universality of subjects: the Ilocano case. In Charles Li (ed.),
Subject and topic, 519–543. New York et al.: Academic Press.
Shibatani, Masayoshi. 2009. Elements of complex structures, where recursion isn’t. The case of
relativization. In Talmy Givón & Masayoshi Shibatani (eds.), Syntax complexity. Diachrony,
acquisition, neuro-cognition, evolution, 163–198. Amster­dam/Philadelphia: John Benjamins.
Siewierska, Anna & Dik Bakker. 2012. Three takes on grammatical relations. A view from the
languages of Europe and North and Central Asia. In Pirkko Suihkonen, Bernard Comrie &
Valery Solovyev (eds.), Argument structure and grammatical relations. A crosslinguistic
typology, 295–324. Amsterdam/ Philadelphia: John Benjamins.
Sirotinina, Olʹga. 1965. Porjadok slov v russkom jazyke [Word order in Russian]. Saratov:
Izdatelʹstvo Saratovskogo universiteta.
Sohn, Ho-Min. 1994. Korean. London/New York: Routledge.
Solnceva, Nina. 1962. Stradatelʹnyj zalog v kitajskom jazyke (problemy morfo­logii) [The passive
voice in Chinese (morphological problems)]. Moskva: Izda­telʹstvo vostočnoj literatury.
Suihkonen, Pirkko, Bernard Comrie & Valery Solovyev (eds.). 2012. Argument structure and gram-
matical relations. A crosslinguistic typology. Amsterdam/ Philadelphia: John Benjamins.
Suñer, Margarita. 1984. Free relatives and the matching parameter. The Linguistic Review 3.
363–387.
Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decompo-
sition of DP. In Louise McNally and Christopher Kennedy (eds.), Adjectives and Adverbs:
Syntax, Semantics, and Discourse, 16–42. Oxford University Press: Oxford.
Tam, Duy Le. 1976. Vietnamese passive. Papers from the 12th Regional Meeting of the Chicago
Linguistics Society, 438–449.
Tchekhoff, Claude. 1979. La construction ergative en avar et en tonguien. Paris: Klincksieck.
Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Klincksieck.
Testelec, Jakov. 1979. Ob odnom javlenii diatezy v dagestanskix jazykax [On a diathesis phenom-
enon in Daghestanian languages]. In Materialy IX mežvuzov­skoj studenčeskoj konferencii po
strukturnoj i prikladnoj lingvistike, 50–53. Moskva: Izdatelʹstvo Moskovskogo universiteta.
Testelec, Jakov. 2001. Vvedenie v obščij sintaksis [Introduction to general syntax]. Moskva:
Izdatelʹstvo RGGU.
Testelets, Yakov & Elizaveta Bylinina. 2005. Sluicing-based indefinites in Russian. In Steven
Franks, Frank Gladney & Mila Tasseva-Kurkchieva (eds.), Formal approaches to Slavic linguis-
tics 13: The South Carolina meeting, 355–364. Ann Arbor, MI: Michigan Slavic Publications.
Tiffou, Étienne & Richard Patry. 1995. La relative en bourouchaski du Yasin. Bulletin de la Société
de linguistique de Paris 90 (1). 335–390.
Trương Van Chình. 1970. Structure de la langue vietnamienne. Paris: Paul Geuth­ner.
Vajs, Danièlʹ [= Weiss, Daniel]. 2000. Russkie dvojnye glagoly: kto xozjain, a kto sluga?
[Russian double verbs: who is the governor, and who is the dependent?] In Leonid Iomdin
& Leonid Krysin (eds.), Slovo v tekste i v slovare. Sbornik statej k semidesjatiletiju aka-
demika Ju. D. Apresjana, 356–378. Moskva: Jazyki russkoj kulʹtury.
Van der Auwera, Johan. 1985. The predicative relatives of French perception verbs. In Machelt
Bolkenstein, Caper de Groot & Lachlan Mackenzie (eds.), Predicate and terms in Functional
Grammar, 219–234. Dordrecht, Holland/Cinnaminson, USA: Foris.
400 References

Van Valin, Robert. 1981. Grammatical relations in ergative languages. Studies in Language 5 (3).
361–394.
Vendler, Zeno. 1968. Adjectives and nominalizations. The Hague/Paris: Mouton.
Vincze, Orsolya & Margarita Alonso Ramos. 2011. A proposal for a multi­level linguistic represen-
tation of Spanish personal names. In Kim Gerdes, Eva Hajičová & Leo Wanner (eds.), Depling
2011 proceedings. Barcelona. 85–93.
Wanner, Leo (ed.). 1996. Lexical functions in lexicography and natural lan­guage processing.
Amsterdam/Philadelphia: John Benjamins.
Wechsler, Stephen. 1995. The semantic basis of argument structure. Stanford, CA: CSLI Publica-
tions.
Weiskopf, Daniel. 2007. Compound nominals, context, and compositionality. Synthese 156 (1).
161–204.
Wierzbicka, Anna. 1971. Kocha, lubi, szanuje. Medytacje semantyczne [(S/he) loves, likes,
respects (you). Semantic meditations]. Warszawa: Wiedza Pow­szechna.
Wierzbicka, Anna. 1991. Cross-cultural pragmatics. The semantics of human interaction. Berlin/
New York: Mouton de Gruyter.
Williamson, Janis. 1987. An indefiniteness restriction for relative clauses in Lakhota. In Eric
Reuland & Alice G. B. ter Meulen (eds.), The representation of (in)definiteness, 168–190.
Cambridge, MA: MIT Press.
Wittgenstein, Ludwig. 1922. Tractatus logico-philosophicus. London: Kegan Paul, Trench, Trubner
& Co./New York: Harcourt, Brace & Co. See http://gutenberg.org/ebooks/5740
Xolodilova, Marija. 2014. Otnositelʹnoe pridatočnoe [Relative subordinate clause]. In Materialy
dlja proekta korpusnogo opisanija russkoj grammatiki (http://rusgram.ru).
Xolodovič, Aleksandr. 1954 [2013]. Očerk grammatiki korejskogo jazyka [Outline of Korean
grammar]. Moskva: URSS.
Xolodovič, Aleksandr. 1971. Nekotorye voprosy upravlenija v japonskom jazyke [Some problems
of government in Japanese]. In Igorʹ Vardulʹ (ed.), Voprosy japonskogo jazyka, 113–132.
Moskva: Nauka.
Xrakovskij, Viktor. 1974. Passivnye konstrukcii [Passive constructions]. In Alek­sandr Xolodovič
(ed.), Tipologija passivnyx konstrukcij. Diatezy i zalogi, 5–45. Leningrad: Nauka.
Xrakovskij, Viktor. 1975. Isčislenie diatez [Diathesis calculus]. In Diatezy i zalo­gi, 34–51. Lenin-
grad: Leningradskoe otdelenie Instituta jazykoznanija AN SSSR.
Xrakovskij, Viktor. 1981. Diateza i referentnostʹ [Diathesis and referentiality]. In Viktor Xrakovskij
(ed.), Zalogovye konstrukcii v raznostrukturnyx jazykax [Voice constructions in languages
of various structure], 5–38. Leningrad: Nauka.
Xrakovskij, Viktor. 1999. Teorija jazykoznanija. Russistika. Arabistika [Linguistic Theory. Russian­
­istics. Arabistics]. Sankt-Peterburg: Nauka.
Xrakovskij, Viktor. 2004. Koncepcija diatez i zalogov (isxodnye gipotezy – ispytanie vremenem)
[Concepts of diathesis and grammatical voice (starting hypotheses and the test of time)].
In Viktor Xrakovskij, Andrej Malʹčukov & Sergej Dmitrenko (eds.), 40 let Sankt-Peterburgskoj
tipologičeskoj škole, 505–519. Moskva: Znak.
Xrakovskij, Viktor (ed.). 1981. Zalogovye konstrukcii v raznostrukturnyx jazykax [Voice construc-
tions in languages of various structure]. Leningrad: Nauka.
Yaguello, Marina (ed.). 1994. Subjecthood and subjectivity. Paris: Ophrys/Institut français du
Royaume-Uni.
Yokoyama, Olga. 1985. A diversified approach to Russian word order. In Michael Flier & Richard
Brecht (eds.), Issues in Russian Morphosyntax, 187–208. Columbus, OH: Slavica.
 References 401

Yokoyama, Olga. 1986. Discourse and word order [= Pragmatics and Beyond 6]. Amsterdam/
Philadelphia: John Benjamins.
Zaliznjak, Andrej & Elena Padučeva. 1975. K tipologii otnositelʹnogo predlože­nija [Toward typol-
ogy of the relative clause]. Semiotika i informatika 6. 51–102. [Reprinted in Andrej Zaliznjak,
2002, “Russkoe imennoe slovoizmenenie” s priloženiem izbrannyx rabot po sovremennomu
russkomu jazyku i obščemu jazykoznaniju, 648–698. Moskva: Jazyki slavjanskoj kulʹtury.]
Zangenfeind, Robert. 2012. Towards a system of syntactic dependencies of Ger­man. Computa-
tional Linguistics and Intellectual Technologies 11 (18), 706–715. See also http://www.
dialog-21.ru/digests/dialog2012/materials/pdf/106.pdf.
Ziff, Paul. 1960. Semantic analysis. Ithaca, NY: Cornell University Press.
Zimmerling, Anton. 2008. Lokalʹnye i globalʹnye pravila v sintaksise [Local and global rules in
syntax]. In Aleksandr Kibrik et al. (eds.), Computational lin­guistics and intellectual technol-
ogies 7 (14), 551–562.
Zimmerling, Anton. 2012. Nekanoničeskie podležaščie v russkom jazyke [Non-canonical sub-
jects in Russian]. In Maria Voejkova (ed.), Ot značenija k forme, ot formy k značeniju:
sbornik statej v čestʹ 80-letija A.V. Bondarko, 568–590. Мoskva: Jazyki slavjanskix kulʹtur.
Žolkovskij, Aleksandr & Igor Melʹčuk. 1967. O semantičeskom sinteze [On semantic synthesis],
Problemy kibernetiki 19. 177–238.
Index of definitions
Definition 1.1 – functional model (p. 8)
A system of symbolic expressions M(E) created by the researcher to descri-
be the function­ing of E is a functional model of E if and only if it associates
with the given inputs the same outputs as E does.

Definition 3.1 – syntactic subject (p. 135)

The syntactic subject in a clause of L is the most privileged SSynt-actant of

this clause's Main Verb.

Definition 3.2 – direct object (p. 174)

The direct object is the second most privileged SSynt-actant of a transitive

verb in a non-er­gative L.

Definition 4.1 – prolepsis (p. 184)

A lexical unit L (with its syntactic dependents) appearing in a clause is a
prolepsis if and only if L satisfies simultaneously the following four condi-
tions:
1. Syntactically, L is only loosely linked to the rest of the clause: it is neither
a syntactic actant of the Main Verb nor one of its circumstantials.
2. Linearly, L is positioned clause-initially.
3. Prosodically, L is “insulated” from the clause by a pause, a stress and a
special intonation contour.
4. Morphologically (in a language with cases), L is most often—but not
exclusively!—in the nominative, the least marked case.

Definition 4.2 – nominative case (p. 186)

The nominative is the case of the form of the noun used for nomination.

Definition 4.3 – subjective case (p. 187)

The subjective is the case used first and foremost for marking the syntactic
subject of any type, but which cannot serve for nomination.

Definition 4.4 – syntactic subject (p. 190), cf. Definition 3.1

The syntactic subject [SyntSubj] is the most privileged surface-syntactic

actant in language L.

https://doi.org/10.1515/9783110694765-014
404 Index of definitions

Definition 4.5 – d
 irect object (p. 193), cf. Definition 3.2

The direct object is the second most privileged surface-syntactic actant in a

non-ergative language L.

Definition 4.6 – pseudo-conjunct (p. 196)

L2 is a pseudo-coordinate dependent, or a pseudo-conjunct, of L1, iff L2 follows

L1 imme­­diately and can play the same surface-syntactic role as L1, but does
not allow for a coord­inating conjunction that would link L2 to L1.

Definition 6.1 – modifier (p. 237)

L̃ stands for a lexical expression (= a phrase).

L̃ is a modifier of L in utterance U if and only if, in U, L depends on L̃ seman-

tically (= ‘L’ is a semantic actant of ‘L̃’), and L̃ directly depends on L syntacti-
cally: ̃ –sem→
L ←synt– L

Definition 6.2 – relative clause (p. 240)

C is a subordinate full-fledged (= finite-verb) clause, Cʹ is its superordinate
(= matrix) clause, and L is a lexeme in Cʹ.

A subordinate clause C is called relative iff C is a modifier of an LU L.

Definition 8.1 – ergative construction (p. 293)

Ergative construction is a verbal predicative construction whose SSynt-

subject is marked by a case different from the nominative.

Definition 8.2 – diathesis (p. 295)

The diathesis of a lexeme L is the correspondence between L’s SemAs and

DSyntAs.

Definition 8.3 – grammatical voice (p. 296)

Grammatical voice is an inflectional verbal category whose grammemes

(= particular voices) mark the modification of the basic diathesis of the
verb and are themselves formally marked on the verb.
 Index of definitions 405

Definition 9.1 – phraseme (p. 328)

A phraseme is a complex linguistic sign (= a configuration of no less than two
linguistic signs) that is not free—that is, it is such that at least one sign in it
cannot be freely selected by the Speaker
1) according to this sign’s meaning and particular combinatorial properties,
2) following general rules of the language, and
3) independently of all other individual signs being part of the complex
sign under consideration.

Definition 9.2 – lexemic phraseme (p. 329)

A lexemic phraseme is a phraseme consisting of no less than two lexemes.

Definition 9.3 – morphemic phraseme (p. 330)

A morphemic phraseme is a phraseme consisting of morphemes inside one
wordform—that is, either a phraseologized complex stem, or a phraseolo-
gized complex affix, or else a phraseologized combination of a stem with
an affix.

Definition 9.4 – syntactic phraseme (p. 330)

A syntactic phraseme is a phraseologized complex linguistic sign that con-
sists of at least two minim­al syntactic trees such that its signifier is non-
segmental, that is, contains prosody or a bound lexemic variable, e.g., L(X),
symbolizing the operation of duplication of the phraseme’s actant X.

Definition 10.1 – the surface-syntactic structure of a sentence (p. 339)

The SSyntS of sentence S is an unordered dependency tree where each lexeme
of S is represented by a node (of which this lexeme is the label) and whose
branches represent language-specific surface-syntactic relations [SSyntRels]
that link these lexemes; the names of the SSyntRels are labels on the branches.

Definition 10.2 – the surface-syntactic-communicative structure of a

­sentence (p. 340)
The SSynt-CommS of sentence S is a division of S’s SSyntS into communica-
tive areas (= subtrees) such that each has its Comm-dominant node specified
and is labeled with a value of a Synt-Comm-opposition.

Definition 10.3 – the deep-morphological structure of a sentence (p. 340)

The DMorphS of sentence S is the linear sequence of all S’s lexemes supplied
with all relevant grammemes.
Index of notions and terms, supplied with a glossary
The glossary contains brief explanations of linguistic notions appearing in this
book. However, the formulations found therein are not necessarily precise and/
or complete.
Boldfaced page numbers refer to spots where the definition or a substantive
discussion of the term in question is found.

Ablativus absolutus
Absolute construction that consists of a noun in the ablative case and a parti-
ciple dependent on it and fulfills the SSynt-role of a circumstantial; e.g.:
Lat. Cen+ā→parat+ā cuncti triclinium intrant
‘[The] supper prepared, all dining-room enter’.
See pp. 71, 85, 272

absolute construction
Adverbial expression without a finite verb, semantically bearing on the whole
clause, but syntactically linked to it loosely; e.g.:
Once home, he met Mary. | John was working, the child asleep at his side.
See pp. 71, 104, 105

absolutive
Case used to mark the Synt-subject of an intransitive verb and the DirO of a
transitive verb (but not used for nomination).
See pp. 141, 145, 147

accusativus cum infinitivo

Syntactic construction of a transitive verb where the DirO expres­ses the Agent
of the infinitival object; e.g.:
Mary knows him to be treacherous (‘he is treacherous’). |
We saw John enter the café (‘John entered the café’).
See p. 61

accusativus cum participio

Syntactic construction of a transitive verb where the DirO expresses the Agent
of the participial complement of the verb; e.g.:
We saw John entering the café (‘John was entering the café’).
See pp. 61, 65

https://doi.org/10.1515/9783110694765-015
 Index of notions and terms, supplied with a glossary 407

actant (of an LU L)
— (of L), deep-syntactic
Lexical unit Lʹ whose presence in the utterance is predicted (= implied) by
the signified of the lexical unit L and which depends on L semantically and
syntactically:

E.g.: John adores Mary.

Cf. modifier/circumstantial (of an LU L).
See pp. 33, 47, 66, 135, 140, 182, 205, 237, 240, 253, 286, 295, 359

— (of ‘σ’/L(‘σ’)), semantic

– Either the semanteme ‘σʹ’ that depends on the semanteme ‘σ’ and corre-
sponds to a semantic actant slot in ‘L’; e.g.:
‘John←1–love–2→Mary’, where ‘John’ and ‘Mary’ are, respectively, SemA
1 and 2 of ‘love’
– Or the lexical unit L(‘σʹ’) that semantically depends on the lexical unit L(‘σ’).
See pp. 33, 106, 182, 205, 221, 237, 323, 402

— (of L), surface-syntactic

Lexical unit Lʹ that syntactically depends on the lexical unit L and either is L’s
syntactic subject/direct object or shares several relev­ant syntactic properties
with these clause elements; e.g., indirect object:
give–indir-objectival→John [the permission to leave]
See pp. 39, 47, 66, 83, 111, 122, 135, 174, 190, 191, 193

adjunct, free (of an LU L)

Modifier/Circumstantial of the lexical unit L.
See p. 352

adjunct, verbal: see verb, phrasal

affix
Morph that is not a radical; e.g.: -s in finger+s, -ing in formul­at+ing, re- in
re+formulate, etc. Cf. radical.
See pp. 27, 159, 169, 262, 296, 330
408 Index of notions and terms, supplied with a glossary

agreement
One of the two types of morphological dependency (the other one being gov-
ernment): the wordform w1 is said to agree with the wordform w2 if and only if
some grammemes of w1 are determined by:
1) Some grammemes of w2:
thisw1 stickw2 ~ thesew1 sticksw2
2) The agreement class of w2:
Fr. beauMASC-w1 palais(masc)w2 ‘beautiful palace’ ~
belleFEM-w1 maison(fem)w2 ‘beautiful house’
3) Some semantemes in the signified of w2:
Rus. Vrač(masc)w2 prišëlMASC-w1 ‘The doctor [male] arrived’. ~
Vrač(masc)w2 prišlaFEM-w1 ‘The doctor [female] arrived’.
See pp. 12, 18, 120, 126–128, 132, 137, 140–173, 349, 350

analysis, linguistic (= speech understanding)

Operation whereby the Addressee of a speech act goes from the text received to
the linguistic meaning expressed by it: Text ⇒ Meaning; cf. synthesis, linguistic.
See pp. 10, 19, 26, 337

analytical form (of a lexical unit L)

Complex linguistic expression in which a grammeme of the lexeme L is real-
ized by a separate lexeme Lʹ; e.g.: willLʹ stay, where the FUTURE grammeme is
expressed by an auxiliary verb; cf. synthetic form.
See pp. 96, 97, 297, 354

apophony
Meaningful alternation; e.g.: A/ɪ/⇒/æ/
past , as in sing ~ sang.

See p. 19

approximate-quantitative syntactic construction (in Russian)

Construction “N + NUM”, in which the anteposition of the noun with respect
to the numeral expresses the meaning ‘the Speaker is uncertain about the
number’ (the neutral order is NUM + N) ; e.g., tonn desjatʹ lit. ‘tons ten’ =
‘maybe ten tons’ (desjatʹ tonn means ‘ten tons’). In the deep-syntactic struc-
ture, this construction is encoded by the fictitious lexeme «PRIMERNO» (lit.
‘approximately’) = «MAYBE».
See pp. 87, 88, 177, 351
 Index of notions and terms, supplied with a glossary 409

arborization
Semantic operation whereby the branches of a deep-syntactic structure are
constructed under synthesis. Cf. lexicalization and morphologization.
See p. 15

asyndetic
Without conjunction; e.g.: the sentence John entered, Mary left features an
asyndetic coordination of two clauses.
See p. 115

attribute
Noun, prepositional phrase or adverb characterizing a noun: e.g.: days of
happiness, a man of integrity, the formula above.
See pp. 86, 93, 140, 143, 187, 199, 201, 203, 208, 222, 353

Base (of a collocation)

Component of a collocation that is selected by the Speaker freely and that
controls the selection of the collocate (see); e.g.: in pay attention, attention
is the base; in black coffee, coffee is the base.
See pp. 23, 55, 199, 284–286, 310, 329

basic structure (of a linguistic representation)

Structure on which other structures of a linguistic representation (= the
peripheral ones) are superimposed.
See p. 11

Characterization (= modification)
Syntactic phenomenon whereby a lexical unit L′ depends on a lexical unit L
syntactically, but is its semantic governor:

E.g.: longLʹ stickL, girlL withLʹ umbrella, runningL fastLʹ. Cf. complementation.
clause (simple)
Phrase that contains a VFIN and all its direct and indirect dependents—except
for another phrase of the same type; e.g.:
John told Mary the news. | that I know the truth | which we found yesterday
See pp. 238, 253
410 Index of notions and terms, supplied with a glossary

clause element
Lexical expression (= a lexeme with its dependents) that can be a direct syn-
tactic dependent of the head of a full-fledged clause; e.g.: the Synt-subject,
the DirO and other objects, a circumstantial, etc.
See pp. 52, 57–59, 93, 122ff, 327, 347, 365

clausative
Part of speech an element of which is syntactically equivalent to a clause; e.g.:
Down [with terrorists!] |Wow! | Not at all.
See pp. 54, 311

cleft
Syntactic construction used to express Focalization:
IT←BE→(PREP→)N THAT/WHO-CLAUSE
E.g.: It was from JohnFOCALIZED that Mary learnt the news.
See pp. 52, 252

cliché
Compositional conceptual-lexemic phraseme; e.g.:
Rome was not built in one day. | Everybody makes mistakes. | No parking.
See pp. 206, 329, 330

clitic
Lexeme L that carries no stress (nor tone) and in the text phonetically “leans”
on a normal wordform, called L’s host.
See pp. 77, 84, 146, 171–173, 265, 344

—, resumptive
Pronominal clitic that repeats (= “resumes”) a clause element according to
the rules of language; e.g.: Sp. Le creo a Juán ‘Him I believe to Juan’, where
the clitic le repeats, or “doubles,” the IndirO a Juan.
See pp. 49, 50, 58, 67

—, second position
Clitic that must be linearly positioned after the first phrase of the clause; e.g.:
Serb. Supu sam pojeo ‘Soup am having.eaten’ = ‘As for soup, I ate it up’ or
Pojeo sam supu ‘Having.eaten am soup’, where the form sam (1.SG present of
biti ‘be’) is a second-position clitic.
See pp. 171, 344
 Index of notions and terms, supplied with a glossary 411

cognate object
Surface-syntactic object N of a verb V such that N’s meaning is the same as
that of V, which otherwise cannot have an object; e.g.: die a heroic death or
Rus. umeret´ gerojskoj smert´juINSTR [idem].
See pp. 54, 72

collocate
Component of a collocation that is selected by the Speaker as a function of
the collocation’s base; e.g.: in pay attention, PAY is the collocate.
See pp. 23, 25, 55, 199, 284, 286, 310, 324

collocation
Compositional lexemic phraseme one component of which—the base—is
selected by the Speaker freely (according to its meaning and combinatorial
properties), while the second component—the collocate—is chosen as a func-
tion of the base; e.g.: pay ATTENTION, heavy INVOLVEMENT, under CONSTRUC-
TION, black COFFEE, leap YEAR.
See pp. 23, 55, 88, 170, 193, 199, 223, 224, 229, 284–286, 310, 311, 316–321, 324, 329, 330

communicate
Express meanings by clauses that implement propositions describing situa-
tions the Speaker targets and that have such a form that they can be negated
or questioned. Cf. signal(V).

communicatively dominant semanteme (in a configuration of semantemes)

In a configuration of semantemes ‘σ1—σ2’ the semanteme ‘σ1’ is communica-
tively dominant if and only if the configuration ‘σ1—σ2’ can be reduced to ‘σ1’
such that the meaning conveyed is simply impoverished, but not distorted
(Iordanskaja & Polguère 1988); the communicative dominance of ‘σ1’ is shown
by underscoring. E.g.: in ‘A bird is.singing’ the semanteme ‘sing’ is communi-
catively dominant, since the utterance is about singing; in ‘a singing bird’, the
semanteme ‘bird’ is communicatively dominant, since this utterance is about
a bird. For a rigorous definition, see Mel’čuk 2012–2015: vol. 1, Ch. 6, 315ff.
See pp. 226, 237, 242, 243

comparand (in a comparative construction)

Element to which the comparate is compared; e.g.: in X is heavier than Y the
element Y is the comparand.
See pp. 66, 103
412 Index of notions and terms, supplied with a glossary

comparate (in a comparative construction)

Element which is compared to the comparand; e.g.: in X is heavier than Y the
element X is the comparate.
See p. 103

complement
Element L′ in a complementation construction.
See pp. 17, 62, 63, 65, 68, 77, 78, 107, 130, 143, 174, 186–188, 195, 196, 202, 208, 226, 299,
332, 350–353, 365

complementation (of L)
Syntactic phenomenon whereby a lexical unit L′ depends on a lexical unit L
syntactically and semantically: ; e.g.: cutL [a] logL′, John’sL′ arrivalL,
overL the city L′. Cf. characterization.
See pp. 33, 46

complementizer
Semantically empty subordinating conjunction that introduces a subordi-
nate clause; e.g.: that in I saw that John was sick.
See pp. 102, 261, 280

complex verb: see periphrastic verb

compositional (complex linguistic sign s)

Complex linguistic sign s that can be represented as a regular “sum” of signs
s1 and s2: s = s1 ⊕ s2.
See pp. 206, 207, 217, 224, 283, 284, 312, 329, 331

conceptual representation: see representation, conceptual

conceptics
Logical device (= set of rules) responsible for the correspondence between
conceptual repre­sentations and semantic representations:
{ConceptRl}⇐conceptics⇒{SemRi}
Conceptics is part of a general model of human linguistic behavior.
See p. 19

conjunction (in logic and semantics)

Logical operator “∧”: A ∧ B is true if and only if both A and B are true.
See p. 12
 Index of notions and terms, supplied with a glossary 413

conjunction (in syntax), binary

Conjunction that consists of two components that are not syntactically linked
between themselves.
See pp. 76, 275ff

context (of a linguistic rule)

Part of a rule that is not manipulated by the rule, but whose presence (in the
rule’s input) is necessary for the rule to apply.
See pp. 15, 349

conversion, morphological
Morphological operation of replacing a feature of syntactics of a wordform;
e.g. OIL(N) ⇒ OIL(V).
See p. 27

coordination
One of the two major types of semantic/syntactic dependency (the other
one being subordination), which unites several semantemes/lexical elements
playing the same semantic/syntactic role; e.g.: The dresses were red, blue,
yellow. | John and Mary travel together. | John awoke, but stayed in bed.
See pp. 34, 110, 300, 301

coreference
Relation that holds between two LUs L1 and L2 in an utterance if and only if
L1 and L2 have the same referent, i.e. they are coreferential. Coreference is an
equivalence relation and is represented by a dashed double-headed arrow:
L1 L2.
See pp. 119, 138, 139, 168

criteria for SSynt-relations: consult Ch. 2

See pp. 40–44, 50, 58, 67, 209–212, 276, 314, 315

Deep- (sublevel of a linguistic representation)

Sublevel that is closer to meaning.
See p. 11
414 Index of notions and terms, supplied with a glossary

deep-syntactic relation [DSyntRel]

One of 13 cross-linguistically universal dependency relations introduced for
the description of deep-syntactic structures of sentences in any language. Cf.
surface-syntactic relation.
See pp. 32ff, 176

deep-syntactic representation: see representation, deep-syntactic

definition, lexicographic (of an LU L)

Formal description of L’s meaning by an expression (of the same language)
that is an exact paraphrase of L satisfying five special rules (Mel’čuk 2012–
2015: vol. 2, 283–293).
See pp. 25

denotation (of a linguistic sign)

Set of all facts or entities of the extralinguistic world that the sign can describe
(= all potential referents of this sign).
See pp. 34, 35, 42, 213, 220

dependency relation (semantic or syntactic)

Binary relation between two semantemes or two lexical units in an utterance:
‘σ1→σ2’ or L(‘σ1’)→L(‘σ2’); this relation is antireflexive and antisymmetric,
and can be non-transitive (semantic dependen­cies) or antitransitive (syntactic
dependencies).
See pp. 120, 209, 339

dependency tree
Formalism for representing the syntactic structure of a sentence; a network
satisfying the following two conditions:
1. Each node receives no more than one entering arc.
2. There is one and only one node that does not receive any arc; this node is
the top node of the tree.
See pp. 12, 13, 38, 226, 242, 275, 321, 328, 331, 339, 345, 350

derivational means
Expressive means (affix, reduplication, conversion or auxiliary lexeme) that
expresses a derivateme; e.g.: the suffix -er (‘one who…’: swimmer), the redu-
plication R ⇒ R, shmR (≈ ‘R which is ludicrous’: data, shmata!), the morpho-
logical conversion N ⇒ V (‘apply N to …’: [to] hammer), etc.
See pp. 37, 166
 Index of notions and terms, supplied with a glossary 415

descriptive expression
Expression used to communicate a meaning. Cf. signalative expression.
See p. 318

descriptive modifier: see modifier, restrictive

diathesis (of an LU L)
correspondence between L’s Sem-actants and its DSynt-actants (specified in
L’s government pattern).
See pp. 118, 129, 143, 158, 295, 296, 298, 303

dictionary article (of an LU L) = lexical entry (of L)

Systematically presented full information about L.
See pp. 25, 229, 284, 285, 321

disjunction
Logical operator “∨”: A ∨ B is true if and only if at least A or B is true.

distinctive number: see lexicographic number

dominance, communicative: see communicatively dominant semanteme

double verbs
Particular type of a verb series: two verbs of which the second is a pseudo-
conjunct of the first; e.g.:
Rus. Ivan sidit smeëtsja ‘Ivan is.seated. is.laughing’. |
Pojdi kupi xleba! ‘Go buy some.bread!’
See p. 113

Ellipsis
Syntactic operation whereby some repeated occurrences of a phrase in the
SSyntS are deleted in the DMorphS of the sentence; e.g.: John travelled to
England, and Mary [travelled] to Spain. | John can play the guitar, and Mary
[can play the guitar] too.
See pp. 103, 125, 242, 265, 300, 302, 323

‘entity’
Class of semantemes denoting objects, living beings, substances, places, etc.;
e.g.: ‘Sun’, ‘sand’, ‘boy’, ‘water’, ‘ravine’, ‘city’. All these are semantic names.
Cf. ‘fact’.
See pp. 75, 182, 201, 375
416 Index of notions and terms, supplied with a glossary

equivalence relation
Binary relation that is reflexive, symmetric and transitive.
See pp. 9, 119

equivalent (semantic representations)

SemR1 and SemR2 are equivalent if and only if one can be transformed into
the other (without affecting the meaning represented) by some rules of the
language.
See pp. 22, 105, 176, 314, 319

ergarive case
Case used to mark the Synt-subject of a transitive verb (but not used for nomi-
nation).
See pp. 134, 141–143, 145, 149, 157, 158, 162

ergarive construction
Syntactic predicative construction where the syntactic subject is not marked
by the nominative case.
See pp. 134, 147, 149, 150, 154, 157, 158, 163, 211, 291, 293

ergative language
Language in which verbs that correspond semantically to transitive verbs
of European languages have as the central component of their meaning the
semanteme ‘undergo’ (rather than then semanteme ‘cause1/2’).
See pp. 134, 143, 154, 177

evaluative construction
Syntactic construction of the type your foolN1 of a bossN2, where an evaluative
noun N1 that semantically qualifies N2 is used as its syntactic governor.
See p. 80

explanatory combinatorial dictionary [ECD]

Dictionary proposed by the Meaning-Text theory.
See pp. 25–27, 284

‘Fact’
Class of semantemes denoting states, processes, properties, actions, events,
etc.; e.g.: ‘grief’, ‘be.located [somewhere]’, ‘sick’, ‘expensive’, ‘write’, ‘explode’,
‘five’. All these semantemes are semantic predicates. Cf. ‘entity’.
See pp. 254, 255
 Index of notions and terms, supplied with a glossary 417

factive verb
Verb that accepts the complement clause that P and whose mean­ing includes
a presupposed component ‘⟦P being true⟧’; e.g.: regret(V) being a factive
verb, the sentences He regrets John’s having left and He does not regret John’s
having left both imply that John has left.
See p. 20

fictitious lexeme
Lexeme that does not exist in the language but is introduced by the linguist in
order to represent a meaningful syntactic construction in the DSyntS; e.g.: Rus.
«primerno» (lit. ‘approximately’) = «maybe», which represents the approxi-
mate-quantitative construction.
See pp. 12, 32, 36, 37, 47, 50, 51, 56, 57, 66, 67–72, 80ff, 108, 176, 218, 222, 227, 228, 229, 230

filter rule: see rule, filter

finite (verbal form): see verbal form, finite

flexible (word order): see word order, flexible

Focalization
One of communicative values: Focalized is the semantic configuration ‘σ̃ ’ that
the Speaker presents as logically prominent; e.g.: It is a pen that I need ~
What I need is a pen.
See pp. 210, 340, 358, 363

functional model: see model, linguistic functional

Government
One of the two types of morphological dependency (the other one being
agreement): the wordform w1 is said to be governed by the wordform w2 if
and only if some grammemes of w1 are determined by some features of the
syntactics of w2; e.g.:
Fr. leACC-w1 remercierw2 lit. ‘him thank’ or
Ger. ihmDAT-w1 dankenw2 lit. ‘to.him thank’,
where the verb determines the case of the object.
See pp. 12, 118, 120, 121, 130, 174, 280
418 Index of notions and terms, supplied with a glossary

government pattern [GP] (of lexical unit L)

Table that describes the actants of the headword L in a lexical entry: L’s dia-
thesis, the surface form of L’s SSynt-actants, their combinability, etc.
See pp. 46, 59–61, 83, 231, 284, 285, 321, 326, 328

governor, syntactic (of an LU L)

LU Lʹ on which the LU L depends syntactically; e.g.:
some←synt–grammemes; Chapter–synt→11; John←synt–is–synt→working.
See pp. 41–46, 73, 85, 97, 123, 135, 206, 210–212, 232, 241ff, 259ff, 280, 314, 315, 346–348,
354, 359, 360, 372, 374

grammeme
Value of an inflectional category, e.g.:
past is a grammeme of the category “verbal tense”;
pl(ural) is a grammeme of the category “nominal number”;
pladj(ural) is a grammeme of the category “adjectival number”.
See pp. 12, 14, 38, 43, 96, 97, 122, 126, 142, 146, 158, 159, 161, 169–171, 209, 296, 338, 340,
345, 353

—, deep (= semantic)
Semantically full grammeme, which has a source in the semantic structure; a
deep gram­meme appears in the deep- and surface-syntactic structures.
See pp. 13, 87, 345

—, surface (= syntactic)
Semantically empty grammeme, which has no source in the semantic struc-
ture; it is imposed by government or agreement and appears in the deep-
morphological structure.
See pp. 12, 13, 338

Head, syntactic (of a phrase P)

LU L which is part of phrase P and on which all other LUs of P depend syntac-
tically—directly or indirectly; e.g.:
South Korean warships conducted live-fire exercises. | Hold Infinity in the
palm of your hand [W. Blake]. | what wives and children say.
See pp. 34, 41, 42, 49, 53, 56, 63, 73, 74, 89, 125, 136–138, 176, 237, 239, 243, 245, 247, 261,
274, 277, 280, 286, 314, 315, 322, 347, 350, 370
 Index of notions and terms, supplied with a glossary 419

homonymy (of linguistic expressions E1 and E2)

Relation of homonymy holds between linguistic expressions E1 and E2
whose signifiers are identical and signifieds do not share a semantic bridge;
e.g.: BOX(N)1 ‘container that …’ ~ BOX(N)2 ‘sport that…’. Homonymy is indicated
by superscripts.
See p. 260

Idiom
One of the two types of lexical units (the other being a lexeme)—a multiword
expression. An idiom is a non-compositional phraseme; e.g.:
˹ALL THUMBS˺ ‘very awkward’ or ˹HIT THE ROAD˺ ‘[to] leave’.
See pp. 12, 89, 152, 206, 207, 216, 224, 225, 229, 273, 280, 282, 283, 284, 286, 309ff

—, lexemic
Idiom that consists of at least two full lexemes; e.g.:
˹ALL THUMBS˺ ‘very awkward’ or ˹HIT THE ROAD˺ ‘[to] leave’.
See pp. 96, 99, 100, 107

—, morphemic
Idiom that consists of morphemes inside one wordform; e.g.:
{FOR} + {GET} ⇔ FORGET
See p. 330

—, syntactic
Idiom that is a constrained complex sign s if and only if its signifier /s/ is
non-segmental, that is, /s/ contains prosody or a bound lexemic variable,
e.g., L(X), symbolizing the operation of duplication of the phraseme’s actant
X; e.g.: [X] ˹OR NO L(X)˺ ‘no matter whether there is X or not’ (Rain or no rain,
we are going.) | [Xs] ˹WILL BE L(Xs)˺ ‘Xs have a typical feature, and that’s what
you have to expect from a particular X’ (Girls will be girls.).
See pp. 70, 110, 331, 332

inflectional category
Set of mutually opposed grammemes; e.g.:
nominal number = {SG, PL}; verbal tense = {PRES, PAST, FUT}.
See pp. 126, 147, 158, 161, 353
420 Index of notions and terms, supplied with a glossary

Keenan-Comrie Accessibility Hierarchy

Hierarchy of clause element types from the viewpoint of their ability to
undergo different syntactic operations (such as relativization, etc.)
See pp. 123, 135, 258

Lexeme
One of the two types of lexical unit (the other one being an idiom)—a one-word
expression. A lexeme is a set of wordforms and analytical-form phrases that
differ only by inflectional significations; e.g.:
I ={I, me}; see = {see, sees, saw, seeing, have seen, am seen, will see, ... }.
For a rigorous definition of lexeme, see Mel’čuk 2012–2015: vol. 1, p. 59, Defi-
nition 1.21.
See pp. 13, 14, 15–18, 20, 26, 33, 49, 96, 97, 142, 176, 191, 295, 311, 328

lexical function
Function f which is associated to a meaning ‘σ’ and which, when applied to
a lexical unit L, returns the lexical unit Lʹ that expresses ‘σ’ in the context of
L: f‘σ’(L) = Lʹ; e.g.:
‘do attention’ ⇔ pay attention, ‘do a step’⇔ take [a] step,
‘do a favor’ ⇔ do [a] favor.
See pp. 12, 16, 17, 22, 23, 24, 55, 223, 247, 285, 286, 317

lexical unit [LU]

A lexeme or an idiom.
See pp. 11, 12, 15, 20, 23, 25, 182, 312, 335, 348

—, full
Lexical unit that has its source in the semantic structure.
See pp. 12, 96, 97, 248, 352

—, structural (= grammatical, or auxiliary)

Lexical unit that either has no source in the semantic structure ( is semanti-
cally empty) and is introduced into the syntactic structure by a syntactic rule
(compare John to Mary; ask whether John is there), or expresses a grammeme
(We will follow you.).
See pp. 12, 13, 46, 130, 270, 297, 332
 Index of notions and terms, supplied with a glossary 421

lexicalization
Semantic operation whereby the lexical nodes of a deep-syntactic structure
are constructed under synthesis. Cf. arborization and morphologization.
See p. 15

lexicographic number
Code used to identify a particular sense of a polysemous lexical item and to
indicate the semantic distance between senses; e.g.: BACK(N)I.1 ‘body part …’
(My back hurts.) vs. BACK(N)I.2 ‘part of clothing covering the backI.1’ (back of
a vest) vs. BACK(N)I.3 ‘part of a seat designated to support the back I.1 of the
sitting person’ (back of a chair), etc.

light verb: see verb, light

linguistic dependency: see dependency relation

linguistic model: see model, linguistic

linguistic sign: see sign, linguistic

linker
Part of speech whose elements are semantically empty and fulfill a purely
syntactic role—they mark the syntactic dependency between a noun and its
characterizer.
See pp. 93, 94, 143, 270

Meaning, linguistic (of an expression E)

Invariant of all paraphrases of E.
See pp. 1, 9, 10, 11, 20, 23, 27, 184, 243

meaning-bearing syntactic construction: see syntactic construction, meaningful

metaphor (of ‘σ1’)

Relation of metaphor holds between two meanings ‘σ1’ and ‘σ2’ such that
‘σ2’ contains ‘σ1’ and the denotation of ‘σ2’ is similar to the denotation of ‘σ1’;
within ‘σ2’, the meaning ‘σ1’ is introduced by a semanteme that indicates
its role—such as ‘˹as if˺’. E.g.: ‘heartII.1’ (of the problem) is a metaphor of
‘heartI.1’ (of John), since ‘heartII.1 of X’ = ‘central point of X —˹as if˺ it were
the heartI.1 of X’. Cf. metonymy.
See pp. 80, 207, 218, 222, 223, 230, 371
422 Index of notions and terms, supplied with a glossary

metonymy (of ‘σ1’)

Relation of metonymy holds between two meanings ‘σ1’ and ‘σ2’ such that ‘σ2’
contains ‘σ1’ and the denotation of ‘σ2’ is contiguous to the denotation of ‘σ1’;
e.g.: ‘heartI.2’ (He pressed his hands to his heart.) is a metonymy of ‘heartI.1’
(of John), since ‘heartI.2 of X’ = ‘part of X’s chest were X’s heartI.1 is’. Cf.
metaphor.
See pp. 16, 194

model, linguistic functional (of language L)

A logical device—that is, a set of rules—that simulates the linguistic activity of
speakers of L (i.e., speech production and speech understanding). A linguistic
model is necessarily function­al, in the following two senses: 1) it represents
the functioning, rather than the structure, of L; 2) it models L as a mathemat-
ical function, i.e., a mapping from L’s meanings to L’s texts, and vice versa.
See pp. 7–9, 19, 27, 28

modification: see characterization

See pp. 33, 46, 81, 88

modifier/circumstantial (of an LU L)
Lexical unit Lʹ that syntactically depends on L, but semantically bears on L:

E.g.: red apple; run fast

Cf. actant (of an LU L).
See pp. 73, 81, 93, 143, 200, 215, 225, 237, 240, 243, 245, 257, 270, 280, 346, 352, 353, 373,
375, 385

—, descriptive
Modifier of an LU L that does not define a subset of entities specified by L, but
only adds a non-definitional characterization to ‘L’; e.g.:
These booksL[, sold in our bookstore,]L’s Descr.Modif. are affordable.
See pp. 34, 35, 86, 87, 88, 89, 90, 92, 210, 243, 244, 245, 258

—, restrictive
Modifier of an LU L that defines a subset of entities specified by L; e.g.:
BooksL [sold in our bookstore]L’s Restrict.Modif. are affordable.
See pp. 34, 76, 87, 88, 90, 210, 242, 243, 244, 256–258, 263, 373
 Index of notions and terms, supplied with a glossary 423

module (of a linguistic model)

Component of a linguistic model: a set of rules ensuring the transi­tion between
two adjacent levels of representation of utterances (foreseen by the linguistic
model).
See pp. 11, 27

—, deep-syntactic
Module ensuring the transition between the deep-syntactic and surface-syn-
tactic representations of utterances.

—, morphological
Module ensuring the transition between the morphological and phonological
representa­tions of utterances.
See p. 14

—, phonological
Module of a linguistic model ensuring the transition between the phonologi-
cal and phonetic representations of utterances.

—, semantic
Module ensuring the transition between the semantic and deep-syntactic
representations of utterances.

—, surface-syntactic
Module ensuring the transition between the surface-syntactic and deep-mor-
phological representations of utterances.
See pp. 2, 341

mood
Inflectional category of the verb whose grammemes indicate the way the
designated fact is viewed/reported by the Speaker: as objective and real (the
indicative mood), as hypothetical (the conditional mood), as possible or
wished for (the subjunctive mood), as an injunction (the imperative mood),
and so on.
See pp. 12, 98, 156

morphological module: see module (of a linguistic model), morphological

morphological representation: see representation, morphological

424 Index of notions and terms, supplied with a glossary

morphologization
Semantic operation whereby the inflectional values for lexical units (of the
syntactic structure under synthesis) are constructed. Cf. arborization and lexi-
calization.
See p. 15

Name (semantic)
Meaning denoting an entity and having no slots for other mean­ings; e.g.:
‘sand’, ‘Moon’, ‘girl’, ‘rhinoceros’, ‘ravine’.
See p. 217

network, semantic
Graph that is fully connected, fully directed and fully labeled; is used to rep-
resent the meaning of linguistic expressions.
See pp. 11, 12, 226

nominative
Case of nomination; e.g.:
Lat. aquil+aSG.NOM ‘eagle’
Rus. malʹčik+iPL.NOM ‘boys’
Basque begi+ØSG+ØNOM ‘eye’
See pp. 38, 128, 133, 134, 138, 141, 142, 143, 145, 152, 154, 155–157, 159, 162, 168, 176, 184,
186, 187, 202, 211, 254, 273, 291, 292, 293, 294, 325, 326

nomineme
Non-compositional conceptual-lexemic phraseme, that is, a phraseme con-
strained with respect to its conceptual representation (= a compound proper
name); e.g.: Medicine Hat (a Canadian city), Brown shirts (a paramilitary wing
of the Nazi party), Saint-Bartho­lomew’s Day (the massacre of Protestants by
Catholics in Paris in 1572).
See pp. 206, 207, 224, 225, 230

non-finite (verbal form): see verbal form, non-finite

Object, direct
Second most privileged element of the clause.
See pp. 18, 39, 44, 52, 58, 118, 130, 174, 190, 193, 200, 201, 211, 241, 291, 305, 343
 Index of notions and terms, supplied with a glossary 425

—, indirect
Third most privileged element of the clause.
See pp. 36, 57, 58, 197

oblique case
Case either opposed to the nominative in a two-case system, or marking cumu-
latively several heterogeneous syntactic roles.
See pp. 143, 176

Paraphrase (of sentence S)

Sentence Sʹ that is synonymous with sentence S; e.g.:
(1) S: ‘Two brothers of Egyptian origin were arrested in France while prepar-
ing to commit an attack’. =
(2) Sʹ: ‘The French police captured two brothers, originally from Egypt, who
were getting ready to perpetrate an attack’.
See pp. 10, 226

peripheral structure: see structure, peripheral (of a linguistic representation)

periphrastic verb
Verbal phrase of the form V→N where the verb V is semantically empty and
the essential meaning of the phrase is expressed by the noun N; e.g.: pay
attention or fall in love. Such a phrase is a verbal collocation.
See pp. 170, 199

phoneme (of L)
The set of all phones of L whose articulatory/acoustical differ­ences are never
used in L to distinguish signs; e.g.:
Eng. /t/ = {[t] (stick), [th] (tick)
/d/ = {[d] (seed), [ɾ] (seeded)}
See p. 38

phonemic representation: see representation, phonemic

phonetic representation: see representation, phonetic

phonological module: see module, phonological

phrasal verb: see verb, phrasal

426 Index of notions and terms, supplied with a glossary

phrase
Utterance that consists of syntactically linked wordforms, features a prosodic
unity, but is not necessarily a unit of communication; e.g.: the report of the
arrival of new shipment of trucks, the report of the arrival, the report, the arrival
of the new shipment, etc. A phrase can be pronounced and understood outside
of a particular context; it is perceived by speakers as existing in their language.
See pp. 1, 18, 40–44, 120, 237, 276, 277, 314, 315, 343, 347–349

phraseme
Phrase that is not free: the selection of its components by the Speaker is con-
strained; four major classes of phrasemes are idioms, nominemes, collocations
and clichés.
See pp. 206, 224–226, 283, 309, 310, 319, 324, 328

—, conceptual-lexemic
Phraseme constrained with respect to its conceptual representation; concep-
tual-lexemic phrasemes come in two varieties: nominemes and clichés.

—, discontinuous
Phraseme that forms a phrase only taken together with its actants; e.g.:
˹NOTHING IF NOT˺ [X].
See p. 286

—, lexemic
Phraseme consisting of lexemes.

—, morphemic
Phraseme consisting of morphemes that appear within one wordform.
See p. 330

—, semantic-lexemic
Phraseme constrained with respect to its meaning (= its semantic representa-
tion); semantic-lexemic phrasemes come in two varieties: idioms and colloca-
tions.
See pp. 329, 331

—, syntactic: see idiom, syntactic

See pp. 330–332

predicate (semantic)
Meaning denoting a fact and having “slots” for other meanings without
which it is incomplete; e.g.:
 Index of notions and terms, supplied with a glossary 427

‘intelligent(X)’ [X is intelligent], ‘love(X,Y)’ [X loves Y], ‘under(X, Y)’ [X is

under Y], ‘order(X, Y, Z)’ [X orders Y to do Z], ‘buy(X, Y, Z, W)’ [X buys Y
from Z for W], etc.
See pp. 11, 182, 205–207, 217, 218, 220–222, 370

prefix
Affix that precedes the radical; e.g.: re+consider or un+constitutional.
See pp. 37, 96, 99, 100, 132, 140, 159, 162, 164–166, 247, 262

presupposition
Part ‘⟦σʹ⟧’ of the meaning ‘σ’ that is not negated or questioned when the
whole ‘σ’ is negated or questioned; in other words, ‘⟦σʹ⟧’ is not accessible
to negation or interrogation. E.g.: the sentence John knows that Mary is in
town presupposes ‘Mary is in town’; this presupposed meaning remains not
affected when the sentence is negated or questioned: both sentences John
does not know that Mary is in town and Does John know that Mary is in town?
presuppose that Mary is in town.
See pp. 21, 22

Pro-Drop language (= pronoun-dropping language)

Language in which a personal pronoun in a particular syntactic role is nor-
mally elided; e.g.:
1) Sp. ¿Cómo estás? ‘How are you?’
or No tiene sentido ‘It does.not make sense’.
2) Jap. Kono hon, watasiga yonda ‘This book, I read it’.
See pp. 125, 126, 145, 185, 196

prolepsis: see Definition 4.1, p. 184

See pp. 148, 183, 184, 185, 187, 189, 192, 194–197, 268

pronominalization
Syntactic operation whereby some repeated occurrences of nouns in the
DSyntS are replaced by substitute pronouns in the SSyntS. E.g.:
John←see→father→John ⇔ John←see→father→his
See pp. 124, 126, 131, 138, 140, 143, 144, 242, 263, 344

pronoun, empty
Pronoun that is semantically empty and is introduced by syntactic rules in
order to insure syntactic well-formedness of the clause; e.g.: It is known that
John has left.
See pp. 49, 128, 136, 167–168, 181, 253, 260, 261
428 Index of notions and terms, supplied with a glossary

—, indefinite
Pronoun referring to a non-specified entity or fact of a particular type: a
person, e.g., anybody, somebody, nobody; a thing: anything, something,
nothing; a place: anywhere, somewhere, nowhere; etc.
See pp. 49, 91, 109, 315–319, 327

—, substitute
Pronoun used instead of a noun, which is its source; e.g.: HE, SHE, THEY, IT,
WHICH, etc.
See pp. 67, 239

prosody
Suprasegmental expressive means of language: stresses, tones. intonation
contours, pauses.
See pp. 43, 111, 175, 210, 262, 338, 363

pseudo-conjunct
Element of the clause that is linked to its syntactic governor as a conjunct,
but does not allow for a coordinating conjunction and semantically repre-
sents an elaboration of the governor.
See pp. 55, 112, 194, 196, 201, 203

pseudo-relative clause
Subordinate clause that has the structure of a relative clause (contains a WH-­
word), but is not a modifier of a noun (as a genuine relative is): a pseudo-­
relative is syntactically equiva­lent to a noun phrase and functions as a syn-
tactic actant; e.g.: We will eat what you brought.
See pp. 49, 53, 108, 249ff, 268, 325, 326

pseudo-X
Element Xʹ that is similar to X, but not enough to be confounded with X in a
linguistic description. Cf. quasi-X.
See p. 33

Quasi-predicate
Meaning denoting entities (as a semantic name), but having “slots” for other
meanings (as a semantic predicate); e.g.: ‘brother of personY’, ‘head of personX’,
‘roof of buildingX’, etc.
See p. 205
 Index of notions and terms, supplied with a glossary 429

quasi-X
Element Xʹ that is similar enough to X so that it is possible to confound it with
X. Cf. pseudo-X.
See p. 33

Radical
Morph that is obligatorily contained in any wordform1 and whose syntactics
1) is similar to the syntactics of the majority of morphs of the language and 2)
contributes the majority of features to the syntactics of the wordform to which
it belongs; e.g.: finger- in finger+Ø and finger+s, fast in fast, formulat(e)- in
formulat+ing, etc. Cf. affix.

raised possessor
Adnominal complement of the noun N syntactically depending on the verb V
becomes an object of V; e.g.: Fr. Ce livre lui a gâché sa carrière lit. ‘This book
to.him has destroyed his carrier’, where lui, the dative form of elle/il ‘s/he’,
is introduced into the clause as a SSynt-actant—the IndirO—of the Main Verb,
although the verb gâcher does not have a corresponding Sem-actant.
See pp. 47, 57

referent (of a linguistic sign s)

Fact or entity in the extralinguistic world (real or imaginary) to which the
linguistic sign s refers in the given utterance. Cf. denotation.
See pp. 37, 57, 87, 128, 130, 133, 305, 318, 373

reflexivity (of a binary relation)

Property of a binary relation R:
R(a, b) ⇒ R(a, a); e.g.: R = ‘be of the same size’.

relation, syntactic
Relation of syntactic dependency between two lexical units; e.g.:
two←synt–units or love–synt→John
representation (linguistic)
Formal object designed to represent a particular aspect of linguist­ic entities;
consists of several structures whose character depends on the level of repre-
sentation.
See pp. 10, 11, 27

1 This formulation leaves out megamorphs—amalgamated realizations of strings of morphemes,

such as: me ⇔ {I}⊕{OBL} or am ⇔{BE}⊕{IND.PRES}⊕{1.SG}.
430 Index of notions and terms, supplied with a glossary

—, conceptual
Representation of the informational content of a sentence at a pre­linguistic
level: a network composed of discrete concepts (designations of elements of
extra-linguistic reality by means of lexical units of natural language, “freed”
as much as possible from their linguistic peculiarities) and the relations be­­
tween them.
See p. 19

—, deep-syntactic
Representation of the formal organization of sentences at the deep-syntactic
level.
See p. 232

—, morphological (of sentences)

Representation of the linear organization of sentences in terms of fully
inflected lexemes.
See pp. 232, 269

—, phonemic
Representation of texts in terms of phonemes and prosodemes.

—, phonetic
Representation of texts in terms of allophones and alloprosodies.
See p. 9

—, semantic
Representation of the meaning of a set of synonymous sentences.
See pp. 9, 27, 232

—, surface-syntactic
Representation of the formal organization of sentences at the surface-syntactic
level.
See pp. 232, 341

restrictive (modifier)
Modifier L of Lʹ is restrictive if and only if it semantically specifies a sub-
class of Lʹ’s denotation; e.g.: in a French book the restrictive modifier French
reduces the denotation of ‘book’ to only ‘French book’.
See pp. 34, 87, 90, 210, 243, 244, 257, 258, 373
 Index of notions and terms, supplied with a glossary 431

resumptive clitic: see clitic, resumptive

rule (linguistic)
Formal expression specifying a correspondence between linguistic objects.
Consult Chapter 1, 1.2.3.
See pp. 1, 2, 15ff, 27, 36–38

—: equivalence rule
Rule specifying the equivalence between two linguistic objects of the same
level of representation: X ≡ Y | C.

—: filter rule
Rule specifying the ill-formedness of a linguistic object on a given level of
represent­ation: *XY [“the configuration XY is ill-formed”].
See pp. 241, 258, 344, 364

—: transition rule
Rule specifying the transition between two linguistic objects of two adjacent
levels of representation: X ⇔ Y | C.
See pp. 15, 39, 45, 139, 344

Semanteme
Meaning (≈ signified) of a lexical unit of the language; e.g.: ‘fence1’ (a wooden
fence), ‘ugly1’ (an ugly face), ‘ugly2’ (an ugly incident), ‘hesitate’, ‘˹sit on the
fence˺’, etc.
See pp. 11, 20, 63, 218, 220

semantic component (of a lexicographic definiens)

Configuration of semantemes in a definiens that plays a particular structural
role in this definiens.

—, central
Component (of the definiens) that expresses the generic part of the meaning
of the LU under description.
See pp. 134, 151

—, generic see component, central

432 Index of notions and terms, supplied with a glossary

semantic decomposition
Representation of a linguistic meaning in terms of simpler linguistic meanings.
See pp. 20–22

semantic derivation
Semantic relation between LUs L and Lʹ such that their semantic difference
‘σ’ = ‘L’ - ‘Lʹ’ is regular in the language in question (that is, it is found in many
lexical pairs), but there is no regular formal difference between L and Lʹ; e.g.,
LION ~ DEN, MOVIE ~ (MOVIE) THEATER, SURGERY ~ OPERATION ROOM, etc., ‘σ’
= ‘place where L’.
See p. 22

semantic primitive/prime
Simple meaning (= semanteme) of language L that cannot be decomposed in
terms of other meanings of L; e.g.: ‘no’, ‘time’, ‘speak’, ‘feel1’, ‘good’, ‘this’, etc.
See p. 20

semantic representation: see representation, semantic

sentence
Maximal utterance that typically consists of clauses (one or more) and is a
complete unit of communication.

sign, linguistic
Triplet 〈X; Y; Z〉, where X is the signified, Y the signifier, and Z the syntactics;
e.g.:
page(N)1 = 〈‘one side of a piece of paper in …’; /peiǯ/; Σ = noun, countable, …〉

signification, linguistic
Any type of information carried by a linguistic sign: a genuine meaning, a
syntactic feature, a semantically empty grammeme, a stylistic characteristic,
etc.

signal(V)
Express meanings without using clauses which implement logical proposi-
tions describing the situations the Speaker targets and which have such a
form that they can be negated or questioned; signaling expressions cannot
be negated or questioned. Cf. communicate.
Consult Mel’čuk 2001: Ch. 3.
See pp. 148, 286, 317
 Index of notions and terms, supplied with a glossary 433

signalative expression
Expression that serves to signal a meaning.
See p. 318

simpler, semantically
Meaning ‘σ1’ is simpler than the meaning ‘σ2’ if and only if ‘σ2’ can be decom-
posed using ‘σ1’, but not vice versa.
See p. 20

source (of a pronoun L)

LU in the DSyntS that is replaced by the pronoun L in the SSyntS.
See pp. 125, 241, 250

Speaker, the
The initiator of the given speech act; the person who says I in this speech act.
See pp. 20, 22, 23, 37, 126, 184, 222, 304, 313, 317, 373

split ergativity
Property of a language in which one set of the MV forms requires the ergative
construction and the other set of the MV forms is used with the nominative
construction.
See pp. 151, 157

stem
Radical taken together with derivational affixes; e.g.:
swimmer- is the stem of the wordforms swimmer, swimmers and swimmer’s;
unlucky- is the stem of the wordforms unlucky, unluckier and unluckiest.
See pp. 99, 330

string
A linear sequence of elements.
See pp. 14, 19, 38, 190, 215, 269, 318, 345, 355, 370, 429

structural lexical unit: see lexical unit, structural

structure
Component of a linguistic representation.
434 Index of notions and terms, supplied with a glossary

—, basic
Autonomous structure of a linguistic representation upon which all peripheral
structures are superimposed.
See p. 11

—, deep-morphological
Basic structure of a deep-morphological representation.
See pp. 14, 124

—, deep-syntactic
Basic structure of a deep-syntactic representation.
See pp. 12, 13, 32, 33, 45, 176, 222

—, peripheral
Non-autonomous structure of a linguistic representation that is superposed
on the basic structure and specifies some of its essential properties.
See pp. 11, 13

—, semantic
Basic structure of a semantic representation.
See pp. 11, 12

—, surface-syntactic
Basic structure of a surface-syntactic representation. Cf. Definition 10.1.
See pp. 13, 14, 38, 39, 45, 124–126, 145, 225

—, surface-syntactic anaphoric
One of the peripheral structures of a surface-syntactic representation: a
system of corefe­rentiality links between lexemes of the sentence; it is shown
by bidirectional dashed arrow .
See p. 39

—, surface-syntactic communicatve
One of the peripheral structures of a surface-syntactic representation. Cf.
Definition 10.2, p. 340.

subject, syntactic
The most privileged element of the clause.
See pp. 39, 117ff, 135, 186, 187, 190, 191, 192, 194–197, 241, 254, 266, 293, 298, 300, 304,
312, 219, 350
 Index of notions and terms, supplied with a glossary 435

subjective
Case used to mark the Synt-subject (but not used for nomination). Cf. Defini-
tion 4.3, p. 187.
See pp. 55, 112, 184ff

subordination
One of the two major types of syntactic dependency (the other one being
coordination), which unites two lexical units L1 and L2 to build a phrase that,
as a whole, has the passive syntactic valence of one of these LUs—the gover-
nor (in bold); e.g.: the dresses; were red; John travels a lot.

substitutability test
Test that allows the researcher to see whether two signs/expressions can be
included into the same unit of a higher level or be described by a common
representation at some level: these signs/expressions must be mutually sub-
stitutable at least in some contexts.
See pp. 44, 209

suffix
Affix that follows the radical; e.g.: chair+s, read+ing, read+er.
See pp. 19, 128, 129, 132, 146, 156, 162, 164–166, 169, 170, 180, 189, 192, 202, 204, 247,
254, 262, 266, 273, 295, 303

superentry, lexical (in a dictionary)

Set of lexical entries that describes a vocable.
See p. 26

surface- (sublevel of a linguistic representation)

Sublevel that is closer to text. Cf. deep-.
See pp. 2, 11

surface-syntactic relation [SSyntRel]

One of several dozen language-specific dependency relations introduced for
the description of surface-syntactic structures of sentences in a particular
language. Cf. deep-syntactic relation.
See pp. 36, 38, 45ff, 175, 193, 205, 208, 241, 276, 305, 330, 339, 369, 374

symmetry (of a binary relation)

Property of a binary relation R: R(a, b) ⇒ R(b, a); e.g.: R = ‘be close to’.
436 Index of notions and terms, supplied with a glossary

synonymous (linguistic expressions E1 and E2)

Two linguistic expressions E1 and E2 are synonymous iff their meanings are
identical; e.g.:
– Sentences
(1) We are short of booze. and (2) We don’t have enough alcohol. are syn-
onymous.
– Sentence (3) and phrase (4)
(3) John has been absent for three years. and (4) John’s three-year absence
are synonymous.
– Lexemes MINISTER and MINISTERIAL are synonymous (but by no means
synonyms).
See pp. 1, 10, 20, 59, 240, 248, 270, 313

synonyms, exact (= full)

Lexical units 1) that have identical signifieds and different signifiers, 2)
whose syntactic actants (if any) correspond one-to-one and 3) that belong to
the same deep part of speech; e.g.:
SOFA ~ COUCH, BEHEAD ~ DECAPITATE, CRAZY ~ NUTS.
See p. 284

synonymy (of linguistic expressions E1 and E2)

1) Identity of the meanings of two linguistic expressions E1 and E2:‘E1’ = ‘E2’.
2) Relation between two lexical units L1 and L2 that are synonyms (e.g.,
FILM ~ MOVIE).
See p. 10

syntactic construction, meaningful

Syntactic construction that carries meaning of lexical type; in the surface-
syntactic structure such a construction is described by a special surface-syn-
tactic dependency relation, and in the deep-syntactic structure, by a fictitious
lexeme. E.g.: Rus. approximate-quantitative construction, where the anteposi-
tion of the quantified noun expresses the meaning ‘maybe’ (pjatʹ metrov ‘five
meters’ ~ metrov pjatʹ ‘maybe five meters’).
See pp. 27, 36

syntactic feature (of a lexical unit)

Indication of a cooccurrence property of an LU; e.g.: «postposed» is a syn-
tactic feature of the adjectives that can follow the modified noun (notary
 Index of notions and terms, supplied with a glossary 437

public, secretary general, [in] matters military, times immemorial). The same
as feature of the syntactics of the LU.
See pp. 66, 68, 69, 73, 77, 93, 109, 216, 221, 225, 232, 242, 250

syntactic idiom: see idiom, syntactic

syntactics
One of the three components of a linguistic sign (along with the signified and
the signifier) that contains information on the sign’s cooccurrence with other
signs in the form of a set of features; e.g.: the syntactics of the noun scissors
contains the following features:
“noun”, “plural only”, “quantification by Num pair(s) of”.
See pp. 18, 38, 100, 136, 349

synthesis, linguistic (= speech production)

Operation whereby the Speaker goes from a meaning he wants to convey to
the text that expresses this meaning: Text ⇒ Meaning; cf. analysis, linguistic.
See pp. 10, 19, 26, 216, 260, 337

synthetic form
Expression in which a grammeme of the lexical unit L is realized by a morpho-
logical means; e.g.: Fr. pardonne+r+a [s/he] ‘will pardon’, where the FUTURE
grammeme is expressed by the suffix -r. Cf. analytical form (phrase).
See pp. 156, 298

Termeme
Conceptual-lexemic phraseme that is an established term.
See p. 224

text (in the technical sense)

Physical (= superficial) expression of a meaning, in terms of speech sounds
or graphic symbols.
See pp. 1, 9, 10, 19, 26

transition rule (= correspondence rule): see rule, transition

transitivity (of a binary relation)

Property of a binary relation R: R(a, b) ∧ R(b, c) ⇒ R(a, c); e.g.: R = ‘be bigger
than’.
438 Index of notions and terms, supplied with a glossary

tree, syntactic: see dependency tree

Underlying question
Question Q formulated by the linguist in order to elicit the semantic-commu-
nicative structure of sentence S; this is a question to which S can be an appro-
priate answer. E.g.:
Q = “What about John?” allows for identification of the semantic Theme;
Q = “What did John do?” identifies the semantic Rheme:
[John]TSem [left for the South Pole]RSem .
See p. 341

Valence, syntactic (of a lexical unit L)

—, active
Set of syntactic elements whose presence in the clause and the surface form
controlled by L; e.g.: the active syntactic valence of the noun SeCRetaRy is
the phrase “to N” (secretary to the company’s president) alternating with a
possessive phrase (John’s secretary, his secretary).
See pp. 33, 46–47, 130, 146, 191, 277

—, passive
Set of syntactic constructions in which L can appear as a dependent element;
e.g.: the passive syntactic valence of an English adjective is 1) being a modi-
fier to a noun, 2) being the attribute of a copula verb, etc.
See pp. 42, 115, 314

verb, light
Collocational verb that is semantically empty in the context of its base; e.g.:
Pay in pay attention or LIe in the responsibility lies with N. Light verbs are ele-
ments of the value of the lexical functional verbs Operi, Funci and Laborij.
See pp. 55, 169, 195, 197, 204

—, phrasal
Verbal phrase that is an idiom and consists of a verb and an adverb of a par-
ticular kind (= verbal adjunct); e.g.: ˹BRInG UP˺, ˹DRoP In˺, ˹ShoW oFF˺.
See pp. 96, 99
 Index of notions and terms, supplied with a glossary 439

verbal form, finite

Verbal form that expresses mood and, as a result, can constitute the syntactic
head of a clause; e.g.: reads, am, read!
See pp. 39, 47, 138, 191, 238, 245, 265, 348, 352, 354

—, non-finite
Verbal form that does not express mood and, as a result, cannot constitute
the syntactic head of a clause; e.g.: reading, [to] be, written.
See pp. 186, 194–197, 200, 238, 246, 359

verbal interjection
Russian interjection that denotes a punctual event and can be used as the
Main Verb of a clause (with all corresponding actants); e.g.: A Ivan vžik emu
po ruke ‘And Ivan quickly.cut to.him on hand’, where the interjection vžik
refers to the sound produced by quickly cutting something with a sharp tool.
In many cases, although not always, a verbal interjection corresponds to the
stem of a verb, such as pryg! from prygnutʹ ‘jump’, and is considered as the
form of “ultramomentaneous aspect.”
See pp. 48, 136

version
Inflectional category whose grammemes mark on a verb for whose benefit the
action in question is performed (e.g., ‘for oneself’ ~ ‘for the other’ ~ — [neutral]).
See p. 158–159

vocable
Set of lexical units (= lexemes or idioms) related by polysemy. In the diction-
ary, a vocable is described by a superentry.
See pp. 26, 148, 239, 332

voice
Verbal inflectional category whose grammemes (= particular voic­es) mark the
modifications of the basic diathesis of the verb and are themselves formally
marked on the verb.
See pp. 129, 130, 141, 143, 147, 148, 154, 161, 164, 289ff, 296, 298

Weight (of a phrase)

The phrase’s length calculated in terms of the number of syllables.
See pp. 210, 214, 215, 372, 383
440 Index of notions and terms, supplied with a glossary

WH-word
Pronoun belonging to one of the four subclasses:
– interrogative (Who is this boy? | Where is meat?)
– indirect-interrogative (I know who is this boy. | I know where is meat.)
– relative (The hill that we see there is called Sugarloaf. | The boy who is
reading the book is John.)
– pseudo-relative (Who wants can come. | I like what you have bought.)
See pp. 238, 309, 344, 362, 363

word order: see Chapter 10

—, flexible
Word order that is more or less independent of the communicative structure
and allows for various permutations of wordforms and phrases.
See pp. 124, 145, 147, 164, 337, 347

wordform
Segmental sign that is more or less autonomous and not representable in
terms of other (previously established) wordforms.
See pp. 10, 36, 43, 48, 64, 108, 126, 325, 330, 344–346
Index of languages
For each language we indicate: family, branch/sub-branch; geographic location.

Acehnese [Austronesian, Malayo-Polynesian; Enga [Trans-New-Guinea; Papua-New Guinea]

Indonesia, Sumatra] 132, 164, 166, 167 149
Albanian [Indo-European, Albanian; Albania] English [Indo-European, Germanic; United
94, 261, 297 Kingdom, USA, Canada, Australia, New
Amele [Trans-New Guinea; Papua-New Zealand] 22, 25, 31, 32, 36–38, 51, 52,
Guinea] 167, 169–171 56, 62, 74, 77, 78, 83, 85, 89, 93, 96, 99,
120, 121, 126, 128, 132, 143, 148, 158, 166,
Ancient Chinese (dead) [Sino-Tibetan, Sinitic;
167, 173, 176, 177, 180, 184, 190, 191, 193,
China] 297
198, 200, 215, 218–221, 244–247, 249,
Ancient Greek (dead) [Indo-European, Hellenic; 257, 259, 262, 275, 287, 297, 298, 304,
Ancient Greece] 65, 71, 181, 272, 273 314, 322, 332, 336, 337, 346, 348, 375
Arabic [Afro-Asiatic, Semitic; several Arabic- Ewe [Niger-Congo; Ghana] 113
speaking countries] 31, 54, 72, 126, 265, Finnish [Uralic, Finnic; Finland] 65, 102, 103,
267, 338 133
Archi [Nakh-Daghestanian; Russia, Caucasus] French [Indo-European, Italic/Romance; France,
134, 147, 152–155, 173, 174 Switzerland, Belgium, Canada] 31, 32,
Armenian [Indo-European, Armenian; 34, 50–53, 57, 59–61, 63–66, 70, 71, 74,
Armenia] 294, 296, 305 75, 77–82, 85, 86, 88, 89, 95, 99, 104,
Awa Pit [Barbacoan; Columbia and Ecuador] 127 107, 108, 122, 132, 159, 176, 183, 184, 190,
215, 248, 258, 297, 360, 365
Bambara (= Bamana) [Mandé; Mali] 270
Georgian [Kartvelian; Georgia] 134, 147, 149,
Basque [language isolate; Spain and France]
155–159, 176
38, 134, 147, 155, 159, 160, 162, 163, 266,
424 German [Indo-European, Germanic; Germany,
Austria, Switzerland] 31, 32, 49, 52, 54,
Biblical Hebrew (dead) [Afro-Asiatic, Semitic;
66, 81, 83, 96, 99, 103, 105, 108, 159,
Ancient Palestine] 128, 181
176, 198, 259, 297, 325, 343
Bulgarian [Indo-European, Slavic; Bul-
Hebrew [Afro-Asiatic, Semitic; Israel] 128,
garia] 58, 67, 84, 98
130, 181, 243, 262, 305
Burushaski [language isolate; Pakistan] 266
Hindi [Indo-European, Indo-Iranian/Indo-
Catalan [Indo-European, Italic/Romance; Spain] Aryan; India] 109, 128, 134, 143, 147,
92, 101 149–152, 157, 176, 263, 271
Chichewa [Niger-Congo, Benue-Congo/Bantu; Hittite (dead) [Indo-European, Anatolian;
Malawi] 262 Ancient Hittite Empire (Modern Turkey)]
Chinese (Mandarin) [Sino-Tibetan, Sinitic; 270
China] 3, 38, 56, 147–149, 181, 201, 289, Hungarian [Uralic, Ugric; Hungary] 59, 60, 96,
290, 297, 298, 300, 302–307, 337, 348 99, 100, 103, 120, 121, 176
Chukchi [Chukotko-Kamchatkan; Russia, Icelandic [Indo-European, Germanic/Scandi-
North-Eastern Siberia] 134, 161, 162 navian; Iceland] 167, 168
Dyirbal [Pama-Nyungan; Australia] 128–130, Ilocano [Austronesian, Malayo-Polynesian;
134, 154 Philippines] 130, 132

https://doi.org/10.1515/9783110694765-016
442 Index of languages

Ingrian (= Izhorian) [Uralic, Finnic; Russia] 274 Romanian [Indo-European, Italic/Romance;

Italian [Indo-European, Italic/Romance; Italy] Romania] 54, 67, 130, 294, 305
66 Russian [Indo-European, Slavic; Russia] 3, 31,
Kabiyé [Niger-Congo, Atlantic-Congo/Gur; 32, 38, 43, 44, 48–51, 53, 54, 56–58, 60,
Togo] 270 63, 64, 66–69, 71, 72, 74–80, 82, 84–93,
98, 101, 103, 106–113, 121, 125, 126, 128,
Kazakh [Turkic; Kazakhstan] 63
132, 133, 136–140, 159, 166–170, 175–177,
Khanty, Western (= Ostyak) [Uralic, Ugric; 205, 206–212, 215, 216, 220, 226, 230–
Russia] 246 232, 241, 246, 251, 252, 256–260, 265,
Khmer [Austroasiatic; Cambodia] 306, 307 267, 268, 275, 278, 280–284, 309–312,
315, 316, 321, 325, 331, 335–340, 342,
Kinyarwanda [Niger-Congo, Benue-Congo/
343, 345, 348, 349, 351, 353, 355, 356,
Bantu; Rwanda] 127, 181
363, 365, 366, 369, 370, 374, 381–385
Korean [Koreanic; South Korea, North Korea]
Russian, Old (dead) [Indo-European, Slavic;
2, 55, 57, 63, 75, 76, 112, 126, 179–194,
Old Russia] 268, 311
196–199, 202–204, 253–255
Sanskrit (dead) [Indo-European, Indo-Aryan;
Lakhota [Siouan; USA] 253–255
Ancient India] 181
Lao [Tai-Kadai; Laos] 306, 307
Sardinian [Indo-European, Italic/Romance;
Latin (dead) [Indo-European, Italic; Ancient Italy, Sardinia] 82
Rome] 61, 65, 71, 85, 186, 198, 206, 252,
Serbian [Indo-European, Slavic; Serbia] 53,
266, 267, 269, 272, 273, 294, 296 57, 97, 101, 198, 344
Lezgian [Nakh-Daghestanian; Russia, Cauca- Seri [language isolate; Mexico, South California]
sus] 38, 128, 134, 135, 141–143, 147, 247, 248
154, 155
Spanish [Indo-European, Italic/Romance;
Lushootseed [Salishan; Canada, West Coast] Spain and Latin America with the excep-
117, 171–173 tion of Brazil] 31, 49, 54, 57, 58, 62, 79,
Malagasy [Austronesian, Malayo-Polynesian; 90, 95, 97, 101, 106, 125, 130, 168, 188,
Madagascar] 126, 130–132, 136, 147, 241 305, 326
Maori [Austronesian, Malayo-Polynesian; New Swahili [Niger-Congo, Benue-Congo/Bantu;
Zealand] 97 Tanzania, Kenya and several other coun-
Mohave [Yuman; USA, Arizona] 265 tries of Eastern Africa] 101, 262, 294, 296

Motu [Austronesian, Malayo-Polynesian; Tagalog [Austronesian, Malayo-Polynesian;

Philippines] 100, 132, 143–147, 270
Papua-New Guinea] 292
Thai [Tai-Kadai; Thailand] 306, 307
Navajo [Na-Dené (= Athabaskan); USA] 125
Tongan [Austronesian, Malayo-Polynesian;
Paamese [Austronesian; Papua-New Guinea]
Tonga] 131, 141, 143, 145–147, 161
113
Turkish [Turkic; Turkey] 35, 62, 63, 176, 246
Persian [Indo-European, Indo-Iranian/Iranian;
Iran] 55, 143, 170, 193, 199, 256, 266, Vietnamese [Austroasiatic, Vietic; Vietnam]
273, 320 147, 148, 264, 306, 307, 348
Polish [Indo-European, Slavic; Poland] 92, 260 Warlpiri [Pama-Nyungan; Australia] 134, 162

Portuguese [Indo-European, Italic/Romance; Yaqui [Uto-Aztecan; Mexico] 247, 248

Portugal and Brazil] 62, 188 Yimas [Sepik; Papua-New Guinea] 155
Quechua [Quechuan; Perú, Ecuador, Bolivia] Yukaghir, Kolyma [Yukaghiric; Russia, Eastern
265 Siberia] 190, 246
Index of semantic and lexical units
applaud(V) — p. 25 it6 — empty (= dummy) it, as in It is John who
left early. — p. 253
bǍ ≈ ‘as for’, Mandarin Chinese — a preposi-
tion that introduces a fronted affected direct ljubimyj ‘favorite’, Russian — pp. 221, 228
object — pp. 302, 305, 306
ne stol′ko – skol′ko — ‘not so.much,
bèi ‘undergo’, Mandarin Chinese — a structural but.rather’, Russian — pp. 279, 282, 284
(= auxiliary) verb that introduces a clause — no ‘but’, Russian — pp. 283, 256
pp. 298–303
‘setand’ — semanteme representing logical
‘cause1’ — the semanteme of involuntary conjunction — p. 12
causation (‘be the cause’) — p. 16
such … that … — p. 248
‘cause2’ — the semanteme of voluntary
­causation (‘be the causer’) — p. 16 sure — pp. 20–22

comme ‘as — like’, French — p. 248 -te (in čërt-te), Russian — particle, an abbre-
viation of tebe ‘to.you’, used as a Da­tivus
˹čem – tem˺ ‘the – the’, Russian — pp. 281– ­Eticus — pp. 319, 321
284
that1 — demonstrative adjectival pronoun,
˹čërt znaet gde˺ — ‘devil knows where’, as in that decision, those decisions — p. 239
­Russian — pp. 311, 313, 315–318
that2 — demonstrative substitute nominal
˹čërt znaet kto˺ — ‘devil knows who’, pronoun, as in Those who want to go must
­ ussian — pp. 312, 314, 316, 318, 328
R sign up tomorrow — p. 239

˹čërt znaet skolʹko˺, Russian — ‘devil knows that3 — relative nominal pronoun, as in a
how much/many’ — pp. 313, 314 ­decision that was made in a hurry — pp. 239,
244, 248, 249
čto1 ‘that5’, Russian — an unvariable seman-
tically empty complementizer (subor­dinating that4 — degree adverb, as in not that intelligent
conjunction) that introduces a completive that5 — empty subordinating conjunction,
clause — pp. 102, 138, 241, 280, 350 i.e. complementizer, as in It meant that the
čto2 ≈ ‘which’, Russian — nominal relative interest in me was aroused. — pp. 239, 248
pronoun — pp. 278, 280 ‘«they»’ — semanteme of the indefinite per-
der1, der2, German — p. 260 sonal pronoun, equival­ent to Fr. on and Ger.
man — pp. 49, 57, 137, 191, 256, 309, 316, 324,
«DERISION», English — p. 37 329, 352
doubt(V) — p. 20 to1, Russian — component of several Russian
repeated conjunctions — pp. 275, 279
esli – to ‘if – then’, Russian — pp. 275–280,
285, 286 to2, Russian — empty particle that is the
­second (optional) component of several binary
èto ‘this’, Russian — presentative particle —
conjunctions: esli …, to2 … ‘if …, then …’,
pp. 75, 116
kogda …, to2 … ‘when …, then …’, etc. —
flow(N) — pp. 15–17 pp. 275–277, 279–283, 286

https://doi.org/10.1515/9783110694765-017
444 Index of semantic and lexical units

totI ‘that [tree]’, Russian — adjectival demon- forms and governs a completive clause —
strative pronoun, as in tot dom, gde on rotilsja pp. 90, 107, 108, 252, 280
‘that house where he was.born’
u ‘at’, Russian — preposition widely used to
totII.1 ≈ ‘that.one’, Russian — nominal cor- introduce oblique objects and cir­cumstan­tials
relative pronoun, as in Tot, kto ustal, možet of a very general meaning, as in U Ivana
otdoxnutʹ lit. ‘That.one who is.tired can sluči­losʹ nesčastʹe lit. ‘At Ivan happened
take.a.rest’, which has all morphological forms accid­ent’ — p. 84
and governs a relat­ive clause — pp. 90, 108,
179, 180 Ø(3,
People
pl)
— zero lexeme that expresses the
meaning ‘indeterminate people’ (≈ ‘«they»’)
totII.2 ≈ ‘the.fact’, Russian — nominal cor-
— p. 49
relative pronoun—a nominalizer, as in To, čto
on ušël, udivilo menja lit. ‘That that he left Ø(3,
METEO
sg, neu)
— zero lexeme that expresses the
amazed me’. totII.2 has only neuter singular meaning ­‘elements/forces of nature’ — p. 49