Noun Valency (PDFDrive) PDF

Noun Valency
Studies in Language Companion Series (SLCS)

This series has been established as a companion series to the periodical
Studies in Language.
For an overview of all books published in this series, please see
http://benjamins.com/catalog/slcs
Editors
Werner Abraham Elly van Gelderen
University of Vienna / Arizona State University
University of Munich
Editorial Board
Bernard Comrie Christian Lehmann
Max Planck Institute, Leipzig University of Erfurt
and University of California, Santa Barbara
Marianne Mithun
William Croft University of California, Santa Barbara
University of New Mexico
Heiko Narrog
Östen Dahl Tohuku University
University of Stockholm
Johanna L. Wood
Gerrit J. Dimmendaal University of Aarhus
University of Cologne
Debra Ziegeler
Ekkehard König University of Paris III
Free University of Berlin
Volume 158
Noun Valency
Edited by Olga Spevak
Noun Valency
Edited by
Olga Spevak
University of Toulouse 2
John Benjamins Publishing Company

Amsterdam / Philadelphia
TM
The paper used in this publication meets the minimum requirements of
8
the American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data
Noun Valency / Edited by Olga Spevak.

p. cm. (Studies in Language Companion Series, issn 0165-7763 ; v. 158)
Includes bibliographical references and index.
1. Grammar, Comparative and general--Noun phrase. 2. Grammar, Comparative and
general--Nominals. 3. Grammar, Comparative and general--Verb. 4. Grammar,
Comparative and general--Syntax. 5. Dependency grammar. I. Spevak, Olga.
P271.N6796 2014
415’.5--dc23 2014013631
isbn 978 90 272 5923 3 (Hb ; alk. paper)
isbn 978 90 272 6998 0 (Eb)
© 2014 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands
John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of content
Abbreviations vii
Editor’s foreword ix
Contributors xv
chapter 1
Contribution of valency to the analysis of language 1
Jarmila Panevová
chapter 2
Special valency behavior of Czech deverbal nouns 19
Veronika Kolářová
chapter 3
Nominalizations of Spanish perception verbs at the
syntax–semantics interface 61
Elisa Bekaert & Renata Enghels
chapter 4
Case assignment, aspect, and (non-)expression of patients: A study
of the internal structure of Czech verbal nouns 89
Věra Dvořák
chapter 5
A data-driven analysis of the structure type ‘man–nature relationship’
in Romanian 113
Ana-Maria Barbu
chapter 6
Classifier noun phrases of the type N1N2 in Bulgarian 141
Petya Osenova
chapter 7
Noun phrasal complements vs. adjuncts 161
Rossella Resi
chapter 8
Noun valency in Latin 183
Olga Spevak
Index 211
Abbreviations
abl ablative
acc accusative
act actor (agent)
addr addressee
adj adjective
adj-poss possessive adjective
app appurtenance
ben beneficiary
cc content clause
cp complementizer phrase
dat dative
dir direction
do direct object
dp determiner phrase
eff effect
fa first argument
fm free modification
gen genitive
id identity
inf infinitive
ins instrument
ins instrumental
ip inner participant
ip inflectional phrase
ipfv imperfective
loc locative/location
mann manner
mat material
n noun
nom nominative
np noun phrase
nrrc non-restrictive relative clause
nv deverbal noun
opt optional
 Noun valency
orig origin
p phrase
pat patient
pfv perfective
pn perception nominalization
poss possessive
pp prepositional phrase
ppc propositional character
pron pronoun
pv perception verb
qm quasivalency modifier
qp quantifier phrase
rc relative clause
rcc relational coordination construction
rrc restrictive relative clause
refl reflexive
sa second argument
v verb
vf valency frame
vm valency member
vp verb phrase
For interlinear glosses The Leipzig Glossing Rules were used.

Editor’s foreword
From Tesnière (1959) onward,1 the concept of valency, “the capacity a verb has
for combining with particular patterns of other sentence constituents” (Gilbert
1994: 4878), is mainly applied to verbs and serves to determine their valency
frames. Complements required by the valency of a verb are called “arguments”
(Fr. “actants”, Germ. “Ergänzungen”); complementation that is not required by
the valency of a verb is called “satellite” (or “adjunct”, Fr. “circonstant”, Germ.
“freie Angabe”). Later scholars dealing with valency extended the same concept
to nouns – and adjectives – which, due to their semantic values, may also require
arguments (Sommerfeldt & Schreiber 1996, among others).
Despite several recent publications (van Durme 1997; Alexiadou & Rathert
2010; Rathert & Alexiadou 2010), the valency of nouns is a topic that still remains
in the shadow of the valency of verbs. Additionally, approaches to noun valency
are rather diverse. Some scholars attribute the capacity of taking arguments
only to nominalizations or deverbal nouns (e.g. Grimshaw 1990), some focus
on support verb constructions (especially within French linguistics), some even
deny noun valency as such (e.g. Mackenzie 1997). Nevertheless, the emergence
of valency lexicons of verbs, nouns, and adjectives, which started in the 1970s
(Helbig & Schenkel 1969 being the first), testifies to the usefulness of this concept
in a description of a language.
This volume aims to contribute to the discussion of noun valency – under-
stood here in a broader sense as a capacity of requiring complements – not only
from a theoretical point of view, as is often the case, but also from an empirical
one by presenting a series of studies focusing on particular questions and based
on data-driven research. The current volume explores properties of valency nouns
in a variety of languages, including Bulgarian, Czech, German, Latin, Romanian,
and Spanish. The specificity of this book consists in the diversity of the meth-
odological approaches used. It is not embedded within one particular linguistic
theory, either. Rather, it explores different theoretical frameworks: Head-driven
. The invention of the concept of valency is usually credited to Tesnière. His book first
appeared in 1953, the 1959 edition is posthumous; its composition dates back to the 1930s
and 1940s. However, Gilbert (1994: 4878) reports that the notion of valency had also been
expounded by de Groot in his Structurele syntaxis (1949) and is even hinted at by Bühler
in 1934.
 Noun valency
Phrase Structure Grammar (HPSG), the Minimalist Program within Genera-

tive Grammar, F unctional Generative Description (FGD), and Construction
Grammar. In the empirical studies, different types of electronic corpora are used,
the details of which are presented below. Special attention is paid to deverbal
nouns, but nouns expressing quantity and “compound-like” constructions involv-
ing relationship and interactivity are also dealt with.
The chapters in this volume grew out of a workshop “Noun valency” organized
during the 45th annual meeting of the Societas Linguistica Europaea in Stockholm
(29th August – 1st September 2012). This lively and engaging workshop revealed
the importance of the exploitation of corpora if we want to advance our knowl-
edge of the behavior of valency nouns. The building of electronic databases and
their annotation is without doubt hard work; however, the fruits collected from
them constitute an irreplaceable source of information.
There are a number of electronic corpora available for modern languages,
especially annotated corpora making it possible to formulate precise research
questions. In this context, it is worth mentioning the Prague Dependency Tree-
bank (PDT), which is elaborated within the framework of Functional Generative
Description (Sgall et al. 1986). Its specificity resides in the interaction between the
theoretical framework and the annotation of texts. In particular, the syntactic and
semantic analysis of actual sentences, which is essential in providing the annota-
tions, brought about the necessity of refinement, adaptation, and complementa-
tion of some theoretical notions. For example, annotators cannot simply ignore
the fact that an obligatory – and therefore valencial – complement is not expressed.
This raises the question of how this or a similar situation should be treated meth-
odologically and leads to the consideration of new concepts (contextual ellipsis,
incorporation of participants, etc.) that have to be elaborated theoretically and
implemented in an existing methodological framework. Jarmila Panevová’s study
(Contribution of valency to the analysis of language) deals with theoretical issues
of valency on a general level and with the specificity of noun valency as opposed
to verb valency. Among other things, she presents the criteria for establishing
valency frames as well as a survey of the semantic participants and their surface
forms. Her comprehensive account is based on her rich experience with annota-
tion of the PDT corpus on the one hand, and with the creation of PDT-Vallex on
the other. PDT-Vallex is a valency lexicon of verbs, nouns, adjectives, and adverbs
〈http://ufal.mff.cuni.cz/PDT-Vallex〉, which arose as a by-product of PDT corpus
annotation. Moreover, Panevová herself (1974 and 1975) provided a theoretical
starting point for treating and annotating valency complements.
Within the same framework, Veronika Kolářová (Special valency behavior
of Czech deverbal nouns) exploits the PDT corpus, which contains 1.5 million
words with syntactic annotation and 0.8 million words with complex semantic
Editor’s foreword 
a nnotation, and also the much larger, lemmatized, and morphologically annotated
Czech National Corpus 〈http://ucnk.ff.cuni.cz〉. She is concerned with valency
properties of Czech deverbal nouns that can exhibit typical or item-specific syn-
tactic behavior. She focuses on deverbal nouns situated on the boundary between
action nouns and result nouns and gives an account of their valency frames. She
pays special attention to the potential relationship between special forms of com-
plementation and changes in the meaning of deverbal nouns.
Elisa Bekaert and Renata Enghels (Nominalizations of Spanish perception
verbs at the syntax–semantics interface) focus on perception nominals, which have
received less attention than true action nouns. More particularly, they examine
the relationship between the complementation of the source verbs and their cor-
responding nominalizations, based on a large amount of empirical data provided
by the (non-annotated) Corpus de Referencia del Español Actual 〈www.rae.es〉.
Some preliminary distinctions are necessary in order to understand the behavior
of perception nominals, for instance between eventive nouns such as visión, and
referential nouns such as vista. This semantic difference is shown to have a clear
impact on the syntax of perception nominals: the first category typically expresses
the first argument (perceiver), whereas the second occurs with the second argu-
ment (object of perception).
Věra Dvořák (Case assignment, aspectual properties, and (non-)expression
of patients: A study of the internal structure of Czech verbal nouns) concentrates
on one type of Czech deverbal event noun, those corresponding to the English
“ing-of ” nominals. Working in the framework of Generative Grammar, she exam-
ines the surface realization of their participants (agents, patients, and goals) with
respect to the source verbs. The deverbal nouns under examination are sensitive to
aspect, just like their corresponding verbs, and can take (im)perfectivity-marking
aspectual affixes. This point is closely related to the issue of non-overt or null
patients, which are allowed with imperfective verbs but usually not with their per-
fective counterparts. On the basis of the data elicited from native speakers, Dvořák
claims that deverbal event nouns, irrespective of their aspectual value, admit a null
patient if it is inferrable from the previous context. Otherwise – that is, without
sufficient contextual information – they behave just like the corresponding verbs
in not allowing null patients in combination with perfective stems.
Ana-Maria Barbu’s contribution (A data-driven analysis of the structure type
‘man–nature relationship’ in Romanian) is about juxtaposition constructions con-
taining a relational noun (such as relationship) and a “compound-like” expansion
(man-nature). In the first place, Barbu examines the grammatical relationship
between the noun and the expansion, but she also pays special attention to the
structure of the expansion itself. She shows that the construction under exami-
nation contains a relational valency noun and a complex complement that can
 Noun valency
be interpreted as a special coordination phrase that also implies reciprocity. Her

analysis is based on the data from a large Romanian corpus ZiareRom built from
non-annotated newspaper texts and is presented within the framework of Con-
struction Grammar.
Petya Osenova (Classifier noun phrases of the type N1N2 in Bulgarian) dis-
cusses noun phrases expressing measures, containers, and forms of grouping
in Bulgarian. Working within the framework of Head-driven Phrase Structure
Grammar, she shows that these “apposition-like” phrases in fact behave as head–
complement constructions. She proposes several semantic subtypes and evalua-
tion criteria useful for identifying them. Her research is based on data from the
Bulgarian National Reference Corpus 〈www.webclark.org〉, developed by the Bul-
TreeBank Group, of which she is a constitutive member. The author used the online
version of a large corpus, comprising more than 400 million tokens of tokenized
but non-annotated texts as well as the syntactically annotated part of the corpus –
the BulTreeBank, which comprises 216,000 tokens (about 15,000 sentences).
Rossella Resi (Noun phrasal complements vs. adjuncts) is concerned with
the syntactic differences between restrictive and non-restrictive relative clauses
in German. She argues that restrictive clauses function as modifiers exhibiting
a complement relationship with their head, unlike non-restrictive clauses, which
function as adjuncts. These properties suggest a close analogy with central and
peripheral subordinate clauses depending on a verb (the distinction introduced
by Haegeman), in that the former type of relative, noun-dependent clauses is cen-
tral, the latter peripheral. The author adopts the framework of Generative Gram-
mar and uses the data collected from native speakers who were asked to evaluate
the grammaticality of restrictive and non-restrictive relative clauses in a specific
context.
Olga Spevak (Noun valency in Latin) addresses several issues concerning
noun valency in Latin, such as some historical considerations, predictability of
the number and type of arguments used with different nouns, nominalization
of verbal notions in Early Latin, and the construction of the so-called dominant
participle. Latin provides interesting evidence for argument marking at a noun
phrase level: arguments are normally encoded in the genitive case (the adnominal
case par excellence); other morphological cases as well as prepositional phrases are
restricted to specific semantic categories of nouns. This work is based on several
case studies made with the help of the (non-annotated) electronic database Library
of Latin Texts.
The editor would like to thank all the contributors for participating in this
project. Special acknowledgments go to Prof. Eva Hajičová and all the colleagues
who willingly accepted the request for review and participated in the peer-
reviewing process. Their expertise contributed not only to the improvements of
Editor’s foreword 
individual articles but also to the fruitful discussion of the topic of noun valency.
Many thanks go to Peter T. Daniels for his valuable help with proofreading and his
critical remarks, and also to Vandana Bajaj, Matthew Barros, Nick Danis, Natalie
DelBusso, Ryan Denzer-King, and Jeremy Perkins for proofreading the chapters
in this volume.
Olga Spevak
References
Alexiadou Artemis & Rathert Monika (eds). 2010. The Syntax of Nominalizations Across Lan-
guages and Frameworks [Interface Explorations 23]. Berlin: De Gruyter Mouton.
van Durme, Karen (ed.). 1997. The Valency of Nouns [Odense Working Papers in language and
communication 15]. Odense: Odense University Press.
Gilbert, Glenn G. 1994. Valency and valency grammar. In The Encyclopedia of Language
and Linguistics, Vol. 9, Ronald E. Asher & Joy M.Y. Simpson (eds), 4878–4886. Oxford:
Pergamon Press.
Grimshaw, Jane. 1990. Argument Structure. Cambridge MA: The MIT Press.
de Groot, A. Willem. 1949. Structurele Syntaxis. Den Haag: Servire.
Helbig, Gerhard & Schenkel, Wolfgang. 1969. Wörterbuch zur Valenz und Distribution deutscher
Verben. Leipzig: VEB Bibliographisches Institut.
Mackenzie, J. Lachlan. 1997. Nouns are avalent – and nominalizations too. In The Valency of
Nouns [Odense Working Papers in Language and Communication 15], Karen van Durme
(ed.), 89–118. Odense: Odense University Press.
Panevová, Jarmila. 1974. On verbal frames in Functional Generative Description, Part I. The
Prague Bulletin of Mathematical Linguistics 22: 3–40.
Panevová, Jarmila. 1975. On verbal frames in Functional Generative Description, Part II. The
Rathert, Monika & Alexiadou Artemis (eds). 2010. The Semantics of Nominalizations Across
Languages and Frameworks [Interface Explorations 22]. Berlin: De Gruyter Mouton. DOI:
10.1515/9783110226546
Sgall, Petr, Hajičová, Eva & Panevová, Jarmila. 1986. The Meaning of the Sentence in its Semantic
and Pragmatic Aspects. Dordrecht & Prague: Reidel & Academia.
Sommerfeldt, Karl Ernst & Schreiber, Herbert. 1996. Wörterbuch der Valenz etymologisch
verwandter Wörter: Verben, Adjektive, Substantive. Tübingen: Niemeyer. DOI: 10.1515
/9783110918878
Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Klincksieck.
Contributors
Ana-Maria Barbu is a Senior Researcher at the Institute of Linguistics of the

Romanian Academy, in Bucharest. Her papers – some of them published in
Springer journals – mainly approach topics of Romanian morphology, syntax
and semantics from both theoretical and computational perspectives. Her current
research focuses on argument realization and subcategorization frames.
Elisa Bekaert is Research and Teaching Assistant of Romance linguistics at Ghent

University. Her interests are mainly in the syntax–semantics interface, and cur-
rently she is preparing a Ph.D. dissertation in which she focuses on the importance
of empirical analyses within the domain of deverbal nominalizations.
Věra Dvořák received her M.Phil. in theoretical linguistics at the University of

Tromsø in Norway, with a thesis analyzing the argument structure of Czech event
nominals. She is currently a Ph.D. candidate in linguistics at Rutgers University in
New Jersey; her dissertation under the supervision of Prof. Mark Baker focuses
on phonetically null objects in generic and episodic contexts. She has published
several articles on topics like nominalizations, ditransitives, dative case, and voic-
ing assimilation in Czech.
Renata Enghels is Associate Professor of Hispanic linguistics at Ghent University.

Her interests are mainly in the interplay between syntax and the mental construal
of verbal and nominal semantics. She has published articles in Linguistics, Revue
de Linguistique Romane, and Revue Romane among others and is author of the
book Les modalités de perception visuelle et auditive: différences conceptuelles et
répercussions sémantico-syntaxiques en espagnol et en français (Niemeyer, 2007).
Veronika Kolářová is Senior Research Associate at the Institute of Formal and

Applied Linguistics, Faculty of Mathematics and Physics, Charles University in
Prague. Her research interests lie in theoretical dependency syntax and corpus lin-
guistics, with special focus on interfaces between syntax, morphology, and the lex-
icon in the domain of noun valency. She participates in building of the annotated
corpus of Czech, the Prague Dependency Treebank. She has published articles in
Slavia and The Prague Bulletin of Mathematical Linguistics and is author of a book
on the valency of Czech deverbal nouns (Karolinum, 2010), written in Czech.
 Noun valency
Petya Osenova is associate professor of Morphology and Syntax in the Division of

Bulgarian Language, Faculty of Slavic Languages, Sofia University “St. Kl. Ohrid-
ski”. Her interests are in the areas of Formal and Corpus Linguistics as well as in
Grammar Writing. She is the author of a book on Bulgarian noun phrases (2009)
and co-author of a book on the formal grammar of Bulgarian (2007), both books
in Bulgarian. Petya Osenova is also the author of the Bulgarian language sec-
tion: “The Languages of the New EU Member States”, Revue Belge de Philologie et
d’Historie (2011). In 2010 she was a Fulbright Fellow at Stanford University (USA).
Jarmila Panevová is a Professor of General and Computational Linguistics at the

Institute of Formal and Applied Linguistics, Faculty of Mathematics and Phys-
ics, Charles University in Prague. Her main research interest is the grammar of
contemporary Czech, especially syntax, and, more broadly, basic issues of general
linguistics. She is a co-author of The Meaning of the Sentence in its Semantic and
Pragmatic Aspects (D. Reidel, Dordrecht 1986) and the author Forms and Func-
tions in the Structure of the Czech Sentence (Academia, Prague 1980), written in
Czech. She is a leading figure and an excellent and popular teacher in the field
of valency studies, focusing on verbs and nouns. Her ideas concerning valency
were formulated as early as the 1970s in the article “On verbal frames in Func-
tional Generative Description” (The Prague Bulletin of Mathematical Linguistics,
1974 and 1975). She is one of the authors of the conceptual scheme of the Prague
Dependency Treebank and continues to serve as an advisor to the team building
updated versions of this annotated corpus of Czech.
Rossella Resi is a Ph.D. student at the University of Verona (Italy) working primar-
ily on Germanic syntax. At the moment she is working on her dissertation con-
cerning word order phenomena in German at the interface between syntax and
informational structure. She wrote “The position of restrictive clauses in German”
(Lingue e Linguaggio, 2011).
Olga Spevak is Assistant Professor of Classics at the University of Toulouse 2. Her

interests are primarily in the areas of Latin syntax, semantics, and pragmatics.
Her current research focuses on the structure of Latin noun phrases. Her books
include The Noun Phrase in Classical Latin Prose (Brill, 2014) and Constituent
Order in Classical Latin Prose (John Benjamins, 2010).
chapter 1
Contribution of valency to the analysis

of language
Jarmila Panevová
Charles University, Prague
Faculty of Mathematics and Physics
The main criteria used for the valency of verbs in the framework of Functional
Generative Description (FGD) are introduced, leading to the recognition of
three classes of verbal modifications: inner participants (IPs), obligatory free
modifications (FMs), and quasivalency modifiers (QMs). The sources of surface
deletions of valency modifications are presented. For an analysis of noun valency,
word formation is taken into consideration. The criteria proposed for the
valency of verbs are applied in order to determine the valency of deverbal nouns.
Conversion of the verbal valency frame into the valency frame of the noun is
accompanied by formal changes in the morphemic form of inner participants.
The behavior of specific noun modifiers is studied with regard to their position in
the corresponding valency frame (VF).
1. Why valency?
At present, the topic of valency is one of the most important and attractive issues
in grammatical description. It appears in different frameworks and is covered by
different terminology – case grammar, theta-roles, verb and its arguments, actants
and circonstants, complements and modifiers, (inner) participants and (free) modi-
fications, etc. The starting point for valency studies was the valency of verbs, which
stimulated the “verbocentric” approach to syntax. During recent decades, how-
ever, the topic of valency has attracted the attention of researchers working in
lexicology and lexicography. Several valency dictionaries for different languages
have been published, Helbig and Schenkel (1969) being the first. This fact reflects
another aspect of valency studies. The description of valency belongs to both syn-
tax and the lexicon, which clearly reflects the interplay of the two modules of lan-
guage description, the lexicon and the grammar.
 Jarmila Panevová
An extension of the study of valency to the other parts of speech is again a

feature which emphasizes its contribution to language description. The diversity of
approaches presented in the present volume is also unsurprising.
It is then quite evident that contemporary language description could hardly
be possible without a notion of valency, whatever the terminology or framework.
The importance of data on valency for the teaching of languages is unquestionable.
For the rule-based systems of natural language processing, the existence of valency
dictionaries, and of rules specifying how the valency members are realized in the
texts, is an inseparable part of the automatic parsing of input sentences, as well as
for text generation. Violation of the valency requirements of the head of a con-
struction is a marker of ungrammaticality. Within automatic language processing,
such a situation usually leads to failure of the analysis; during generation, however,
the need to observe the valency requirements of the head is one of the guarantees
of the grammaticality of the generated output.
Though here we stress the attractiveness and importance of the topic of
valency, we must admit that presentations of valency phenomena differ greatly
depending on the terminology and explanations used.
Therefore, before presenting the empirical issues connected with noun
valency, we briefly describe our framework and the terminology used in the
description that follows. Our contribution is based on the framework of Func-
tional Generative Description (FGD) that was proposed for Czech in the 1960s
(Sgall 1967) and developed over the following decades, including the theory of
valency (Panevová 1974 and 1975, Sgall et al. 1986). Within this framework,
valency theory was applied to the valency of verbs and was later broadened to the
valency of nouns (see Piťha 1981, Panevová 2000 and Kolářová 2010) and adjec-
tives (Panevová 2008).
2. What is valency about?
Our approach is first demonstrated via the criteria used for valency of verbs, and
then these criteria are applied to nouns.
Modifications (complementations) of verbs are classified according to two cri-
teria testing their compatibility with their verbal heads:1
1. The repertoire of modifiers is based on traditional Czech syntactic handbooks (e.g.
Šmilauer 1947; Grepl & Karlík 1986; Daneš et al. 1987) and on empirical studies within FGD
(new types of modifiers were introduced in FGD by Panevová 2003; Mikulová et al. 2005;
Panevová & Mikulová 2012).
Chapter 1. Contribution of valency to the analysis of language 
i. A modifier may be connected with only a limited number of heads, which

must be listed;2
ii. A modifier may be connected with any verb.3
Modifiers satisfying criterion (i) (see Panevová 1974: 11f.) can occur with a given
head only once, while modifiers satisfying criterion (ii) can be repeated with a
single governor (see also the notion “repeatable” in Mel’chuk 1988: 143).
The verbal modifications which satisfy criterion (i) are called inner partici-
pants (IP), for example dům ‘house’ in stavět dům ‘to build a house.’ The other
class of verbal modifications, satisfying (ii), is called free modification (FM), for
example ‘in the train’ in usnul ve vlaku ‘he fell asleep in the train’, rychle ‘quickly’ in
běžet rychle ‘to run quickly’. The compatibility of a particular verb with every type
of modifier from the list was tested.4 On the basis of these tests, the verbal modifier
enters either the class of inner participants (a) or the class of free modifications (b):
a. Actor (ACT), Patient (PAT), Addressee (ADDR), Effect (EFF), Origin (ORIG);
b. other modifiers (location, direction, temporal, manner, cause, regard, etc.).
A “middle” class of modifiers, called quasivalency (QM), was added later (see
Panevová 2003; Lopatková & Panevová 2006). This concept covers cases such as
Obstacle modification (OBST), e.g. Jan zakopl o stůl ‘John stumbled over the table’
or Difference modification (DIFF), e.g. Inflace se zvýšila o několik procent ‘The
inflation has increased by several percents.’ The QMs share some of the features
typical of an IP (they occur with a limited class of verbs, their forms are governed
by their respective heads, they are not repeatable), but some of their features are
shared with the class of FMs – compared to IPs (such as ACT, PAT), which are
semantically heterogeneous, QMs have specific semantics.
Thus, the valency of an item can be defined as a set of required dependents
for a given item, with a desired dependency function – and for some of them, e.g.
for the class of IPs and QMs, with a prescribed form. It is reflected in the lexicon
in the shape of a valency frame (VF). In the valency dictionaries compiled for
FGD and its application for corpus annotation (Lopatková et. al. 2008 and Urešová
2011), the VF contains all the IPs (marked for obligatoriness or optionality and for
2. The Actor modifier (looking like an exception from this criterion) has a special position.
3. Some exceptions to this general statement exist; they are influenced by the lexical incom-
patibility of the head with its modifier.
4. A full list of inner participants and free modifications, used in the classical version of
FGD, is given in Sgall et al. (1986: 198); an enriched list is given in Mikulová et al. (2005: 425f.),
where the term ‘functor’ is used.
 Jarmila Panevová
their morphemic form(s)) and all obligatory FMs. For ambiguous and polysemic
lexical units, a VF is specified for each meaning they have. The VF can be empty
for specific verbs, e.g. for pršet ‘to rain.’
3. The deletion/absence problem
So far, our assumptions have been based on the valency members (VMs) present
on the surface level. The situation is complicated, however, by the existence of the
surface deletion (absence) of a VM. The sources of absence of a member of a VF
are the following:5
i. An IP is part of a VF, but it is marked as optional (opt). ADDR and ORIG with
the verb koupit ‘to buy,’ PAT and ADDR with the verb chlubit se ‘to boast,’ and
PAT with the verb mluvit ‘to talk’ are examples of optional participants (see
the illustrations of VFs in (1), and the applications of the VF in concrete utter-
ances in (2)–(3)). In these cases the IP is not deleted – patient in (2), addressee
and origin in (3) – but, as its VF allows, is simply not used.
(1) mluvit ‘to talk’ – ACT(NOM), PATopt (o + LOC) ‘about’, ADDR (s + INS)
‘with’
koupit ‘to buy’ – ACT(NOM), PAT(ACC), ADDRopt (DAT), ORIGopt
(od + GEN) ‘from’
(2) Jan často mluví se svými prarodiči.
John-nom often talk-prs with his-ins grandparents-ins
‘John often talks with his grandparents.’
(3) Jan včera koupil nové auto.
John-nom yesterday buy-pst new car-acc
‘John bought a new car yesterday.’
ii. The VM is generalized. The notion of generalization of ACT occurs in tra-

ditional syntactic approaches under different terms. In Generative Grammar
it is called “arbitrary” (PROarb). The semantic interpretation of a generalized
actor refers prototypically to the group of persons typical/customary for a
given action/state. In Czech, the typical form for generalization of the actor
(ACT) is a reflexive verbal form (psát (se) ‘to be written;’ cf. French s’écrire)
with absence of a slot for an ACT, see (4).
5. Deletions connected with text structure (textual anaphora and ellipsis) are not considered
here.
Chapter 1. Contribution of valency to the analysis of language 
(4) Na začátku věty se píše velké písmeno.

At beginning sentence refl write-3sg-prs capital letter-nom
‘A capital letter is written at the beginning of a sentence.’
However, the notion of generalization is applicable for the other valency members
as well. See general PAT and general ADDR in (5) with the verb having obligatory
PAT and ADDR in its verbal frame:
(5) Zuzana prodává v supermarketu.
Sue-nom sell-3sg-prs in supermarket
‘Sue sells in the supermarket.’
The annotated corpus (Prague Dependency Treebank, PDT 2.0),6 for the manual
annotation of which the FGD framework was used, makes it possible to compare
the number of overtly expressed IPs with the number of their generalization. The
result of this comparison is given for participants of verbs in Appendix 1, for noun
IPs in Appendix 2.
iii. In Panevová (1974: 17f.), a test for determining the obligatoriness of VMs was
proposed, and it still appears to be a useful tool in doubtful cases, with surface
absence of VM expected in the text. Example (6) shows an application of this
test (called a dialogue test) to an omitted obligatory FM in a dialogue between
two speakers.
(6) a. Přátelé už přišli.

Friend-nom.pl already come-3pl.pst
‘My friends have already arrived.’
b. Kam?
‘Where?’
c. *Nevím.
*‘I don’t know.’
Under normal circumstances, the question in (6b) cannot be answered by the

speaker of (6a) with (6c). The unacceptability of such a dialogue demonstrates that
the slot for direction modification (DIR) is semantically obligatorily with the verb
přijít ‘to come’ (though deletable on the surface). The marker of deletability must
be reflected in the VF in the valency lexicon. However, empirical studies of this
issue still continue for both the verbs and nouns.
6. PDT 2.0, see 〈http://ufal.mff.cuni.cz/pdt2.0〉 contains 49,431 sentences with 833,195
tokens.
 Jarmila Panevová
4. Labelling of inner participants
The interpretation of the labels used for IPs in our framework is influenced by
Tesnière’s (1959) approach. If a verb has a single IP (argument), it is an ACT(or), as
in otec pracuje ‘father is working’, as well as in otec spí ‘father is sleeping’ and komín
kouří ‘the chimney smokes’. The two-argument verbs have ACT(or) and PAT(ient)
in chlapec chytil rybu ‘a boy caught a fish’ as well as in muž kope jámu ‘a man digs
a hole’ and chlapec viděl lva ‘a boy saw a lion’. In these examples the semantically
based theories would assign different labels to ACT in the first three examples and
to PAT in the last three. With verbs with three (or more) arguments, the semantics
of the participants is taken into consideration: after the ACT and PAT are deter-
mined, the third and following slots are candidates for ADDR(ressee), EFF(ect), or
ORIG(in). The surface requirements of the given item (its “government”) are also
a useful tool for this labelling. In the example řečník oslovil publikum ‘a speaker
addressed the hearers’, semantically oriented theories would be labeled the 2nd
argument (‘hearers’) an ADDR, while FGD considers it PAT. We speak here about
a (cognitive) shift (in this case the shift of ADDR into the position of PAT).7
5. C
onclusions on verbal valency as a source for the examination
of valency of nouns
1. As stated in Section 2 above, verbal modifiers (they number approximately

50) are classified into 3 groups: inner participants (IPs), free modifications
(FMs), and quasivalency modifications (QMs).
2. Any modifier specified for a particular verb as its IP, obligatory FM, and QM
has a slot in the corresponding valency frame (VF) in the lexicon.
3. No verbal VF can contain a slot for PAT without having a slot for ACT; no
verbal VF can contain a slot for ADDR, EFF, or ORIG without having slots
for ACT and PAT.
6. Valency of nouns
The criterion that is most reliable for the nature of the valency of the verb modi-
fier (and the deletion of which causes ungrammaticality, as in *Jan navštívil *‘John
visited’) fails with nouns. Omission of a noun modifier occurs much more often
7. The arguments for the shifting are given in Panevová (1974) in comparison with T
esnière’s
and Fillmore’s approach.
Chapter 1. Contribution of valency to the analysis of language 
than with verbs (for comparison of expressed and generalized participants see
Appendixes 1 and 2). Any noun with an obligatory valency slot can occur with-
out such a modifier and still retain grammaticality (e.g. skupina ‘group’, část ‘part’,
tucet ‘dozen’). On the other hand, while verbs with empty VFs are rare (e. g. prší ‘it
rains’), there are many nouns with no valency slots in their VFs. This holds espe-
cially for primary, non-derived nouns.
The derivation of nouns plays an important role in the study of their valency.
Therefore we separate the examination of the valency of deverbal (and possibly
deadjectival) nouns (Section 6.1) from that of primary nouns (Section 6.2).
6.1 Valency of deverbal nouns

There are two classes of deverbal nouns (NV):
i. nouns derived by syntactic derivation, such as ošetření ‘treatment’ (see

Kuryłowicz 1936, and applied here in Kolářová’s contribution, this volume).
The syntactic function of the source verb changes, while the lexical meaning
is not touched (6.1.1);
ii. nouns derived by lexical derivation, such as actor nouns like učitel ‘teacher’,
where the lexical meaning as well as the syntactic function are changed (6.1.2).
6.1.1 Deverbal nouns derived by syntactic derivation

Nouns derived by syntactic derivation inherit the VF slots of their source verbs.
Though the morphemic forms of the specific participants often change in com-
parison with the forms of the IPs of corresponding verbs, the assumption of
inheritance of VF is in principle valid. Formal changes of the morphemic forms
are partially regular, partially irregular, sometimes even unexpected. The lat-
ter cases are the topic of Kolářová’s contribution to this volume. The transfor-
mation of VFs with two-argument verbs into VFs of nouns is illustrated by
Examples (7)–(11).
(7) a. Lékař ošetřil pacienta.
Doctor-nom treat-pst patient-acc
‘The doctor treated a patient.’
b. ošetření pacienta lékařem
treatment-nom patient-gen doctor-ins
lékařovo ošetření pacienta
doctor-adj.poss treatment-nom patient-gen
pacientovo ošetření lékařem
patient-adj.poss treatment-nom doctor-ins
 Jarmila Panevová
pacientovo ošetření od lékaře

patient-adj.poss treatment-nom from doctor-gen
*ošetření pacienta lékaře
treatment-nom patient-gen doctor-gen
*pacientovo ošetření lékaře
patient-adj.poss treatment-nom doctor-gen
‘treatment of a patient by the doctor’
c.VF of V: ACT(NOM), PAT(ACC) →
NV: (a) ACT (INS / ADJ-POSS / od + GEN), PAT(GEN);
(b) ACT (INS / od + GEN/*GEN), PAT(ADJ-POSS).
In the nominalized form of the verb (NV), both participants, ACT and PAT, can be
expressed by the genitive form (see (7c)), but if these two IPs co-occur, the genitive
form is strongly preferred for the position of PAT. The GEN form is also excluded
for the ACT if the PAT is expressed by a possessive adjective (see (7b)). PAT is
expressed by a possessive adjective, which is an adjective derived from a noun, as
in lékař-ovo ‘doctor’s’ or pacient-ovo ‘patient’s’ in several variants of (7b).
The noun VF derived from the verb odjet ‘to leave’ is given in the right side of
(8a), and its application is illustrated by (8b) and (8c).
(8) a. VF: V ACT(NOM), DIR18→ NV ACT(GEN) DIR1

b. Můj kolega odjel ze Stockholmu.
my colleague-nom leave-pst from Stockholm
‘My colleague left Stockholm.’
c. odjezd mého kolegy ze Stockholmu
leaving my colleague-gen from Stockholm-gen
‘my colleague’s departure from Stockholm’
The possessive adjective (kolegův) is excluded in (8c) by the syntactic rules of

Czech whereby a derivation of the possessive adjective is allowed only with a single
human noun (cf. 7b), while in Example (8c) ACT is expressed by a noun phrase.
The single genitive form in these deverbal noun patterns conveys the well-
known ambiguity of the type střílení lovců ‘the shooting of the hunters.’ Karlík and
Nübler (1998) and other authors explain this ambiguity as resulting from different
sources. In these cases the ACT label correspond to nominalization of the active
sentence (‘the hunters shoot’), while PAT in these constructions has as its source a
passive sentence (‘the hunters were shot’).
8. DIR1 is used for the meaning of direction from where?

Chapter 1. Contribution of valency to the analysis of language 
The verb hodit ‘to throw’ with the VF ACT(NOM), PAT(INS) is transformed
into a nominal construction with the VF NV: ACT(INS / GEN / ADJ-POSS / od +
GEN), PAT(INS), see (9a) and (9b).
(9) a. Sportovec hodil diskem.
sportsman-nom threw-pst discus-ins
‘A sportsman throw a discus.’
b. hod diskem od sportovce
throw-nom discus-ins from sportsmen-gen
sportovcův hod diskem
sportsman-adj.poss throw-nom disc-ins
‘the sportsman’s throwing of the discus’
As a result of the regular transformation between VFs of verb and noun, the
instrumental form for ACT is expected here. However, pragmatic conditions (the
euphony of two instrumental cases) lead ACT to prefer the other two possible
forms (GEN or possessive adjective).
The verbs with PAT in the genitive, such as dosáhnout, V: ACT(NOM),
PAT(GEN), ‘to achieve’ preserve this form in the NV frame, but they exclude the
GEN form for the ACT position; see (10a) and (10b).
(10) a. Učitel dosáhl zamýšleného cíle.
teacher-nom achieve-pst intended target-gen
‘The teacher has achieved the intended target.’
b. učitelovo dosažení zamýšleného cíle
teacher-adj.poss achievement-nom intended-gen target-gen
*dosažení učitele zamýšleného cíle
*achievement-nom teacher-gen intended-gen target-gen
‘the teacher’s achievement of the intended target’
In transformations of verb VFs, with the PAT in a prepositional case, the form
of the PAT is usually preserved with NV, while for the ACT, all possible forms
are available. See (11a) and (11b); the variations of (11b) with GEN are not given
explicitly here.
(11) a. Překladatel odkázal na původní verzi článku.
translator-nom refer-pst to original version-acc article-gen
‘A translator referred to the original version of the article.’
b. překladatelův odkaz na původní verzi článku
translator-adj.poss reference to original version-acc article-gen
‘the translator’s reference to the original version of the article’
 Jarmila Panevová
With verbs having more than two arguments, some combinations of IPs block
certain forms. Verbs without a passive voice block the instrumental form for the
ACT of NV, see (12a) and (12b).9
(12) a. Nepřátelé bojují s protivníky o moc.
enemy-nom.pl fight-prs with rival-ins.pl for power
‘The enemies fight with their rivals for power.’
b. boj nepřátel s protivníky o moc
fight-nom enemy-gen.pl with rival-ins.pl for power
‘the fight of enemies with their rivals for power’
Two variants of morphemic expressions of noun valency modifications are con-

nected with the alternation between ADJ-POSS and GEN for the IPs; if the form of
GEN occupies the ACT slot, this form cannot be used for the ADDR, though the
source verb fills the ADDR slot with ACC (which is regularly converted into GEN
with NV), see (13a) and (13b).
(13) a. Místní rozhlas informoval občany o
local radio-nom inform-pst citizen-acc.pl about
hrozbě povodní.
danger-loc flood-gen
‘A local radio informed the citizens about the danger of floods.’
b. informace občanů o nebezpečí povodní
information-nom citizen-gen.pl about danger-loc flood-gen
místním rozhlasem
local radio-ins
‘the informing of citizens about the danger of floods by the local radio’
Two genitives are allowed with NVs whose source verbs have the GEN form for PAT.
One of the genitive forms is preserved for the IP that is also expressed by the GEN in
the source verb; the other IP is expressed by an accusative, as with the verb zbavit ‘to
get rid of ’ with the VF: ACT(NOM), PAT(GEN), ADDR(ACC). The corresponding
NV has the VF: ACT(INS), PAT(GEN), ADDR(GEN), see (14a) and (14b).
(14) a. Novela zákona zbavila
amendment-nom law-gen relieve-pst
rodiče odpovědnosti.
parents-acc responsibility-gen
‘Parents were relieved of responsibility by an amendment of a law.’
9. The possessive adjective is blocked as well, but the source of blocking is different (plural
form of the noun).
Chapter 1. Contribution of valency to the analysis of language 
b. zbavení rodičů odpovědnosti

relief-nom parents-gen responsibility-gen
novelou zákona
novel-ins zákona law-gen
‘relief of parental responsibility by an amendment of a law’
An instrumental form for an ACT is blocked in the transformation of the verbal

VF that has the PAT in INS, as in vyhrožovat ‘to threaten’ with the VF: ACT(NOM),
PAT(INS), ADDR(DAT), see (15a) and (15b).
(15) a. Věřitelé vyhrožují podnikateli exekucí.

lender-nom.pl threaten-prs entrepreneur-dat distraint-ins
‘Lenders threatened the entrepreneur that they would seize his
property.’
b. výhrůžka exekucí podnikateli od věřitelů
threat-nom distress-ins entrepreneur-dat from lender-gen.pl
‘the threat to the entrepreneur of distress from (his) lenders’
6.1.2 Deverbal nouns derived by lexical derivation

The second group of deverbal nouns is represented by nouns derived by suf-
fixes that contribute a specific semantic feature to the noun. Many Czech nouns
belong to this class, but a great number also straddle the boundary between syn-
tactic and lexical derivation. The analysis of the latter is the topic of Kolářová’s
contribution to this volume. We present here only selected representatives of
specific semantic classes illustrating the approach based on processes of word
formation.
Nouns derived by an agentive suffix such as -tel, -eč, or -ce essentially absorb
the ACT position, which is not – nor can it be – expressed by a separate lexical
unit. Nouns such as uči-tel ‘teacher’, žada-tel ‘applicant’, and zástup-ce ‘representa-
tive’ have a PAT in their VFs as an obligatory or optional slot, while other agentive
nouns as kup-ec ‘merchant’, kov-ář ‘black smith’, saz-eč ‘typesetter’, and naklada-tel
‘editor’ are not combined with PAT. Examples with an incorporated ACT (some of
them combined with a PAT) are given in (16); in the last three examples the PAT
is obligatory.
(16) soudní znalec / znalec písma
court-adj expert expert handwriting-gen
‘expert witness’/‘handwriting expert’
stánkový prodavač / prodavač vysavačů
‘stalls-adj seller salesman vacuum cleaners-gen
‘street vendor’/‘vacuum cleaner salesman’
 Jarmila Panevová
brusič / brusič drahých kamenů

cutter cutter gemstones-gen
‘cutter’/‘gemstone cutter’
zástupce ministerstva
representative ministry-gen
‘representative of the ministry’
předkladatel zákona
proposer law-gen
‘proposer of the law’
vyznavač heavymetalu
follower heavy metal-gen
‘follower of heavy metal (music)’
In the following examples, it is the PAT which is absorbed by the word-formation

process: dárek ‘gift’ (an object gifted is a gift), see (17), výdaj ‘expense’ (of money),
výpis ‘(bank) statement’, zápis ‘record’ (of data). Such absorbed PATs do not appear
on the surface at all.
(17) dárek mladší dcery mamince
gift younger daughter-gen mother-dat
dárek mamince od mladší dcery
gift mother-dat from younger daughter-gen
‘the gift from the younger daughter to her mother’
In some cases, the other modifiers are absorbed in derived nouns (names of places,
means, etc. see Kolářová 2010: 46f.), but the incorporated modifier does not have
a valency feature, so they are not considered here.
6.2 Primary nouns

Modifications specific to nouns were studied by Piťha (1981). According to the
criteria used for verb valency (see Section 2 above) the modifier Material (often
called partitive; MAT) was introduced as an inner participant, obligatory with
some nouns (i), optional with others (ii):
i. část (pozemku) ‘part (of the land)’, konec (filmu) ‘end (of the movie)’, polovina
‘half ’, člen ‘member’, skupina ‘group’, odstavec ‘section’;
ii. sklenice (vína) ‘glass (of wine)’, talíř ‘plate’, šálek ‘cup’, košík ‘basket’.
Applying the “dialogue” test (see above Section 3) for the names of parts, groups,
and similar notions in (i), we conclude that the speaker using them properly must
have knowledge of the whole.
Chapter 1. Contribution of valency to the analysis of language 
Class (ii) includes the names of containers and similar nouns. With this class
the modifier MAT is not obligatory, if these nouns are presented as the names of
objects that can be counted, manipulated; or if they are used as measures (contain-
ers), they have MAT in their frames as an optional IP.
The modification APP(urtenance), specific to nouns, was tested as to its
valency character as well. As it is compatible with any noun, it belongs to the class
of FMs; however, with some nouns with relational meaning, APP is semantically
obligatory, as in bratr (mého otce) ‘(my father’s) brother’, předseda ‘chairman’, vlast-
nost ‘quality’. We can test the obligatoriness again by the dialogue test. Speaking
about the election of a chairman, the speaker must know what he/she will chair.
The modifier ID(entity) belongs among the optional IPs of the listed head
nouns. It can be expressed by a nominative, as in hotel Zvon ‘the hotel Bell’, loď
Queen Mary ‘the ship Queen Mary’, or by a genitive: pojem čas-u ‘notion of time’.
6.3 Linguistic meaning or cognitive content?

Names of artifacts need special attention and perhaps special treatment as to their
valency behavior. On the one hand, they express results (effects, in our terminol-
ogy) of human activities. On the other, they denote some countable, relocatable
objects (as do the names of containers). However, this is not a strong argument for
regarding them as ambiguous, though their different usage influences their valency
behavior. As the names of artifacts, they notionally have ACTs, and most of them
have PATs. As to their formal expression of PATs, the names of artifacts constitute
two classes:. In class (i) the PAT is optional, expressed by the prepositional case o +
LOC ‘about + LOC’, as with povídka ‘story’, báseň ‘poem’, dopis ‘letter’, píseň ‘song’.
In class (ii), the optional PAT is expressed by GEN or by ADJ-POSS, as with portrét
‘portrait’, fotografie ‘photo’, obraz ‘painting’, rukopis ‘manuscript’, and perhaps all
names of buildings, parts of furniture, etc., understood as artifacts. ACTs in both
classes are expressed by the same forms. Their VFs may look like (18):
(18) VF of (i): ACT (GEN / ADJ-POS / od + GEN)opt, PAT (o + LOC)opt;

VF of (ii): ACT (GEN / ADJ-POS / od + GEN)opt, PAT (GEN / ADJ-
POS)opt.
The identity of the forms expressing the two participants is a source of many
semantic interpretations. Moreover, the same form of GEN can also be used with
the same heads for the free modification of APP. For example, the genitive in
román L. N. Tolstého ‘novel by Leo Tolstoy’ is interpreted as an ACT, but in portrét
L. N. Tolstého ‘portrait of Leo Tolstoy’ it is an optional PAT and in dopisy mojí
maminky ‘letters of my mother’ the genitive most probably represents an APP. The
interpretation is here based on our knowledge of the world, the situation, or the
context. See (19)–(22).
 Jarmila Panevová
(19) román L. N. Tolstého vs. portrét L. N. Tolstého

novel-nom L.N. Tolstoy-gen vs. portrait-nom L.N.Tolstoy-gen
‘a novel by Leo Tolstoy vs. a portrait of Leo Tolstoy’
(20) a. Picassovy portréty G. Braqua
Picasso-adj.poss portraits-nom G. Braque-gen
‘Picasso’s portraits of G. Braque/Picasso’s portraits from/by G. Braque’
b. Braquovy portréty P. Picassa
Braque-adj.poss portraits-nom P. Picasso-gen
‘Braque’s portraits of P. Picasso/Braque’s portraits from/by P. Picasso’
(21) a. Corbusierova lenoška od Druota
Corbusier-adj.poss armchair-nom from Druot-gen
‘Corbusier’s armchair from/by Druot’
b. Druotova lenoška od Corbusiera
Druot-adj.poss armchair-nom from Corbusier-gen
‘Druot’s armchair from/by Corbusier’
(22) a. dědečkovy dopisy od maminky
grandfather-adj.poss letters-nom from mother-gen
‘grandfather’s letters from my mother/grandfather’s letters written by
my mother’
b. maminčiny dopisy od dědečka
mother-adj.poss letters-nom from grandfather-gen
‘mother’s letters from grandfather/mother’s letters written by
grandfather’
The correct interpretation of Tolstoy’s role in (19) is possible only if we know that
he was a writer and that he is an ACT rather than a PAT in the first noun phrase
in (19), while the PAT role is more probable for the second noun phrase in (19).
For the interpretation of (20), even knowledge of the world will not help identify
the proper meaning as to who is the ACTor and who is the PATient in both (a)
and (b) (since both participants were painters and friends). The situation is com-
plicated by the meaning of Possessor (APP) of the participating forms (ADJ-POSS
and GEN), which we did not reflect in the description of the forms in (19)–(22).
Similarly, knowledge of the famous architect (known not only for his buildings,
but also for his furniture and paintings) and the lesser-known antique dealer is of
no help for the relevant interpretation of (21). Possession plays a role here. A fam-
ily story is perhaps behind the content of (22), but it is difficult to come up with an
appropriate interpretation.
To sum up our attempt to answer the question formulated as the title of
this section we can conclude that both ADJ-POSS and a noun in GEN in the
Chapter 1. Contribution of valency to the analysis of language 
c onstructions studied in this section are vague rather than ambiguous, so that
they are all marked with the relatively vague label APP. A detailed analysis of the
form of the prepositional phrase od + GEN ‘from’ occurring in (21) and (22) is still
lacking. In some patterns, this form fits the ACT slot (e.g. in (22a)–(22b)), but the
meaning of origin (ORIG) is available as well. The ORIG assignment with nouns
denoting human beings offers a more open interpretation than ACT: an ORIG
could be understood as an ACT as well as a mediatory person who is actively par-
ticipating in the situation described, but is not its real actor.
6.4 Conclusions on noun valency

To conclude our discussion, the following statements can be made:
1. The same list of modifications used for verbs is applicable for nouns. It is
enriched by three special noun modifications participating in their valency
frames (MAT, APP, ID).
2. The “dialogue” test appears to be a useful tool for determining the semantic
obligatoriness of nouns as well as of verbs.
3. The principles of shifting do not seem to be applicable or useful for nouns.
(There are nouns having PAT in their VF even though ACT is absent.)
7. Future perspectives for studies of noun valency
Though in the last decades noun valency has become an attractive domain of the
linguistic study of particular languages as well as of general linguistic studies,
some issues remain open:
i. To compile and publish noun valency dictionaries for languages that do not
yet have any, which appears to be a realistic task.
ii. To specify a boundary between noun valency frames that could be trans-
formed by a set of simple rules from those inherited from their respective
sources. By means of such a transformation, many noun VFs can be achieved
without great effort.
iii. Though determining the conditions for the surface deletion of obligatory
valency members belongs to the domain of grammatical studies, it is also a
great challenge for text studies.
The list of issues given here is certainly not exhaustive for the future develop-
ment of noun valency studies. However, these issues appear to be both topical and
interesting.
 Jarmila Panevová
Acknowledgments
The research reported in this paper was supported by the Czech Science Foun-
dation project P406/10/0875 and used the language resources developed by the
LINDAT-Clarin project of the Ministry of Education of the Czech Republic
LM2010013.
References
Daneš, František, Grepl, Miroslav & Hlavsa, Zdeněk (eds). 1987. Mluvnice češtiny 3: Skladba
(Czech grammar 3: Syntax). Prague: Academia.
Grepl, Miroslav & Karlík, Petr. 1986. Skladba spisovné češtiny (Syntax of literary Czech). Prague:
SPN.
Helbig, Gerhard & Schenkel, Wolfgang. 1969. Wörterbuch zur Valenz und Distribution deutcher
Verben. Leipzig: VEB Bibliographische Institut.
Karlík, Petr & Nübler, Norbert. 1998. Poznámky k nominalizaci v češtině (Notes on nominalisa-
tion in Czech). Slovo a slovesnost 59: 105–112.
Kolářová, Veronika. 2010. Valence deverbativních substantiv v češtině (na materiálu substantiv
s dativní valencí) (Valency of deverbal nouns in Czech: On the basis of nouns with dative
valency). Prague: Karolinum.
Kuryłowicz, Jerzy. 1936. Dérivation lexicale et dérivation syntaxique. Bulletin de la Société lin-
guistique de Paris 37: 79–92.
Lopatková, Markéta & Panevová, Jarmila. 2006. Recent development of the theory of valency in
the light of Prague Dependency Treebank. In Insight into Slovak and Czech Corpus Linguis-
tics, Mária Šimková (ed.), 83–92. Bratislava: Veda.
Lopatková, Markéta, Žabokrtský, Zdeněk & Kettnerová, Václava. 2008. Valenční slovník českých
sloves (Valency lexicon of Czech verbs). Prague: Karolinum. 〈http://ufal.mff.cuni.cz/
vallex/〉
Mel’chuk, Igor A. 1988. Dependency Syntax: Theory and Practice. Albany NY: New York State
University Press.
Mikulová, Marie et al. 2005. Anotace na tektogramatické rovině Pražského závislostního kor-
pusu. Anotátorská příručka (Annotation on the Tectogrammatical Level in the Prague
Dependency Treebank. Annotation manual). Technical Report TR-2005-28. Prague:
ÚFAL MFF UK.
Panevová, Jarmila. 2000. Poznámky k valenci podstatných jmen (Remarks on valency of nouns).
In Čeština – univerzália a specifika 2, Zdeňka Hladká & Petr Karlík (eds), 173-180. Brno:
Masarykova univerzita.
Panevová, Jarmila. 2003. Some issues of syntax and semantics of verbal modifications. In Pro-
ceedings MTT 2003, First International Conference on Meaning-Text Theory, 139–146. Paris:
École Normale Supérieure.
Chapter 1. Contribution of valency to the analysis of language 
Panevová, Jarmila. 2008. Povaha stupňování adjektiv: K “nesrovnávacímu” užití stupňovaných

forem (Degrees of comparison of adjectives: On their non-comparative meaning). In Iugi
Observatione... Jubilejný zborník na počesť L’ubomíra Ďuroviča, Slavomír Ondrejovič (ed.),
149–156. Bratislava: Veda.
Panevová, Jarmila & Mikulová, Marie. 2012. Assimetrii mezhdu hlubinnym i poverxnostnym
predstavleniem predlozhenija (na primere dvux tipov obstojatel’stv v cheshskom jazyke)
(Asymmetries between deep and surface representation of sentence: Case of two types of
adverbials in Czech). In Smysly, teksty i drugie zachvatyvajushchie sjuzhety. Sbornik statej v
chest’ 80-letija I. A. Mel’chuka, Jurij D. Apresjan, Igor’ M. Boguslavskij et al. (eds), 486–499.
Moscow: Jazyki slavjanskoj kul’tury.
Piťha, Petr. 1981. On the case frames of nouns. In Prague Studies in Mathematical Linguistics,
Vol. 7, 215–224. Prague: Academia. DOI: 10.1075/llsee.9.16pit
Sgall, Petr. 1967. Generativní popis jazyka a česká deklinace (A generative description of lan-
guage and the Czech declension). Prague: Academia.
Sgall, Petr, Hajičová, Eva & Panevová, Jarmila. 1986. The Meaning of the Sentence in Its Semantic
Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Klincksieck.
Urešová, Zdeňka. 2011. Valence sloves v Pražském závislostním korpusu (Verb valency in Prague
Dependency Treebank) [Studies in Computational and Theoretical Linguistics]. Prague:
Institute of Formal and Applied Linguistics.
Appendix 1. Inner participants of verbs in PDT 2.0
IP’s total Participants expressed Participants

on the surface generalized
ACT(or) 87,111 74,280 7,523 9%

PAT(ient) 68,185 65,211 2,711 4%
ADDR(essee) 10,150 5,211 4,649 46%
EFF(ect) 7,208 6,980 199 3%
ORIG(in) 849 766 61 7%
Appendix 2. Inner participants of nouns (with functions

corresponding the IP of verbs) in PDT 2.0
IP’s total Participants expressed Participants

on the surface generalized
ACT(or) 20,786 9,689 8,345 40%

PAT(ient) 25,705 22,011 3,248 13%
ADDR(essee) 2,172 963 764 35%
EFF(ect) 340 284 30 9%
ORIG(in) 248 240 4 2%
chapter 2
Special valency behavior of Czech

deverbal nouns*
Veronika Kolářová
Charles University, Prague
Faculty of Mathematics and Physics
This chapter provides an in-depth analysis of the valency properties of Czech

deverbal nouns. It focuses on the forms of complementation they take. These can
be typical, related to those of the source verbs, or special, without any relationship
to them. We present an overview of these special shifts in valency. Special forms
of participants have an impact on the syntactic behavior of the noun and its
meaning. We argue that it is not always a plain shift in meaning but sometimes
only a slight meaning nuance. Such nouns with special forms of participants
require creating a new valency frame; they represent a separate category on the
boundary between syntactic and lexical derivation.
1. Introduction
1.1 Aims and objectives

Czech deverbal nouns can exhibit multiple forms of valency complementation
(participants); for example, lov ‘a hunt’: lov velryb ‘hunt of whales’ with a Patient
encoded in the genitive which represents a typical form of adnominal comple-
mentation, inherited from the source verb, and lov na velryby with a prepositional
phrase (‘hunt for whales’, i.e. ‘whales hunt’), which is an atypical, special form of
complementation, with no relationship to the form of the corresponding com-
plementation of the source verb. This study aims at a description of the special
valency behavior of Czech deverbal nouns, focusing on the special forms of the
participants and their influence on the meaning of the nouns involved. From a
* I would like to thank Olga Spevak, the editor of the present publication, and two anony-
mous reviewers for many stimulating remarks and comments.
 Veronika Kolářová
theoretical point of view, such a description is important for better understand-

ing the relationship between verbal and nominal valency. Moreover, as a practical
application of a theory, the description of the special valency behavior of nouns is
also of great importance to the lexicographic treatment of noun valency.
The crucial issue addressed in this chapter is whether two different forms of a
participant represent mere variants that can be captured in one valency frame, or
whether there are reasons for splitting them into two different valency frames. In
case two different valency frames are recognized, the question arises as to whether
these two different valency frames correspond to two different meanings of the noun.
We try to show that a special form of a participant does not automatically
imply a change in the meaning of a deverbal noun. On the other hand, we argue
that a participant in a special form cannot co-occur with the same set of forms of
other participants from one valency frame as typical forms. Such difference in the
syntactic behavior of a deverbal noun has a certain impact on its meaning, even
if it is only a slight nuance. Consequently, we propose creating a special valency
frame in the valency lexicon.
We suggest that nouns with a participant in a special form that have meanings
similar to those of nouns with participants in a typical form but that exhibit dif-
ferent syntactic behavior should be classified as a transitional category, namely as
nouns on the boundary between syntactic and lexical derivation.
1.2 Methods
Several syntactic tests are used, making it possible to describe the differences not
only between the forms of a participant but also between the syntactic behavior
of a noun modified by a participant in a typical form and the syntactic behavior
of the same noun modified by that participant in a special form. Our hypothesis
is that a participant in a special form cannot combine with the same set of forms
of other participants from the valency frame as the same participant in a typical
form. The tests that were used are:
–– possibility of expressing an agent (Actor) in the prepositionless instrumental

case;
–– combination with adjectives that specify the content of a noun with respect
to time and iteration (adjectives such as ‘frequent’, ‘repeated’, ‘constant’, ‘occa-
sional’, and ‘everyday’);
–– combination with adverbs such as předem ‘in advance’.
When describing the valency behavior of Czech deverbal nouns, we take advan-
tage of our experience in the annotation of nominal valency in the Prague Depen-
dency Treebank (PDT), cf. Mikulová et al. (2006), and in its valency lexicon,
Chapter 2. Special valency behavior of Czech deverbal nouns 
which is called PDT-Vallex (Hajič et al. 2003). The detailed analysis and the exam-
ples of special valency behavior presented in this chapter are based on the data
from five morphologically annotated subcorpora of the Czech National Corpus
(CNC, Český národní korpus), namely SYN2000, SYN2005, SYN2006PUB, SYN-
2009PUB, and SYN2010. For the analysis, syntactically and semantically compact
groups of nouns (ca. 140 lemmas) were selected; the selection of nouns, the search
methods, and the frequency data concerning the particular observed phenomena
are described in Kolářová (2006) and Kolářová (2010).
1.3 General framework of Functional Generative Description

Our approach to deverbal nouns is based on the theory of valency (especially the
valency of verbs) developed in the framework of Functional Generative Descrip-
tion (FGD) by Panevová (1974 and 1975) and Sgall, Hajičová & Panevová (1986).
The valency frames of verbs, which are presumably stored in the (mental)
lexicon, are reflected in the tectogrammatical representation of sentences in FGD.
The following types of complementation (i.e. the individual dependency relations)
are able to fill the individual slots of the valency frames of verbs:
a. inner participants or arguments that can be obligatory or optional:1 Actor

(ACT), Patient (PAT), Addressee (ADDR), Effect (EFF), Origin (ORIG);
b. obligatory free modifications or adjuncts, especially those with the meanings
of direction (e.g. přijet někam.DIR32 ‘to arrive somewhere’), location (e.g.
přebývat někde.LOC ‘to dwell somewhere’), and manner (e.g. chovat se dobře.
MANN ‘to behave well’).
Within the concept of nominal valency in the FGD framework (Piťha 1984 and
Panevová 2002), the meaning of a given noun is the most important factor in
determining its valency frames. According to Kuryłowicz (1936), two basic types
of word-formation, lexical derivation and syntactic derivation, can be distin-
guished. In syntactic derivation, only the syntactic function of the derived word
differs from that of the source word; for example, uplynout ‘to expire’ – uplynutí
‘expiration’, ostrý ‘sharp’ – ostrost ‘sharpness’, dobrý ‘good’ – dobře ‘well’. In lexi-
cal derivation, not only the syntactic function but also the lexical meaning of the
derived word differs: concerning deverbal nouns, these do not denote an action or
1. This point is discussed in more detail in Panevová, this volume. For definitions of argu-
ments see Mikulová et al. (2006: 460–479).
2. The following labels (so-called functors) are used for particular sub-meanings of direc-
tion: DIR3 (where), DIR2 (which way), DIR1 (from where).
 Veronika Kolářová
a state but for example a physical entity such as a thing (ukazovat ‘to point’ – uka-
zovátko ‘pointer’) or a person (učit ‘to teach’– učitel ‘teacher’). This distinction is
useful for the description of the valency of deverbal nouns. From this perspective,
our approach to the valency of deverbal nouns focuses on three groups of nouns
that represent the most dominant stages in the process of substantivization:
i. Nouns derived from verbs by syntactic derivation. Their meaning is parallel

to the meaning of their source verbs in that they denote an action3 (vyrábění/
vyrobení ‘manufacturing’ // výroba ‘production’) or a state (vyskytování se/
vyskytnutí se ‘occurring’ // výskyt ‘occurrence’); the forms of their participants
are typical.
ii. Nouns on the boundary between syntactic and lexical derivation. The mean-
ing of these nouns is slightly different from the meaning of action or state
nouns; however, the nouns belonging to this category, such as pochvala ‘praise’,
are still abstract nouns; the forms of their participants can be either typical or
special.
iii. Nouns derived by lexical derivation. This category includes nouns whose
meaning is unmistakably different from the meaning of action or state nouns;
it is represented by names of physical entities related to actions (semantically
concrete nouns) such as actor nouns (učitel ‘teacher’), nouns denoting a thing
(dodávka ‘van’, otvírák ‘opener’) or nouns denoting a place (stoupání ‘slope’,
východ ‘exit’, čekárna ‘waiting room’); forms of their participants can be either
typical or special.
There are two basic types of Czech deverbal nouns that in one of their meanings
can sometimes denote an action or a state and so can belong to group (i), namely
nouns derived by syntactic derivation: nouns derived from verbs by productive
means, by suffixes -(e)ní/tí, as in honění ‘hunting’ or hubnutí ‘losing weight’ (cf.
Dvořák, this volume); and nouns derived from verbs by non-productive means
or by zero suffix, such as honba ‘hunt’, hon ‘hunt’. These two types of nouns are at
the center of attention in this study since they can often exhibit both typical and
special valency behavior.
1.4 Treatment of deverbal nouns in Czech valency lexicons

There are two valency dictionaries that in addition to the valency of verbs and adjec-
tives also capture the valency of Czech nouns, namely a printed valency d ictionary
3. We use the term “nouns denoting an action” in a broader sense, for all non-stative nouns,
also subsuming accomplishments and achievements (Dowty 1979).
Chapter 2. Special valency behavior of Czech deverbal nouns 
prepared by Svozilová, Prouzová and Jirsová (2005) and an electronic valency dic-
tionary built during a tectogrammatical annotation of the Prague Dependency
Treebank (PDT), called PDT-Vallex (Hajič et al. 2003; Urešová 2011a).
The lexicon compiled by Svozilová, Prouzová and Jirsová (2005) presents a
traditional way of capturing noun valency by giving only examples of particular
complementation, regardless of possible combinations with other types of com-
plementation expressed by various forms, such as lov na medvěda ‘hunt for bear’,
i.e. ‘bear hunt’, lov ryb ‘hunt of fish’, i.e. ‘fishing’ (Svozilová, Prouzová & Jirsová
2005: 130).
In contrast, in the PDT-Vallex, the core valency information is encoded in
valency frames in which the possible combinations of complementation expressed
by various forms are taken into account. The PDT-Vallex valency entry contains
a lexeme (represented in PDT by the “tectogrammatical lemma”) and its valency
frame(s). As for verbs, the authors suppose that one valency frame corresponds
prototypically to one meaning (sense) of a verb.4 Although the PDT-Vallex does
not explicitly work with the term lexical unit, a meaning of a word (i.e. a verb, a
noun, or an adjective) with its particular valency frame corresponds to a lexical
unit, understood roughly as ‘a given word in a given sense’ (Cruse 1986).
Concerning nouns, the PDT-Vallex contains 3727 entries. So far, special
attention has been paid first to capturing the valency properties of nouns derived
from verbs by productive means, such as the noun balení ‘pack(ing)’, and second
to nouns occurring as nominal components in “support verb” constructions such
as the noun nabídka ‘offer’ in učinit nabídku ‘to make an offer’. The delimitation
of boundaries between particular meanings of a noun is one of the most diffi-
cult tasks in nominal valency annotation and consequently in the treatment of the
valency properties of the nouns in the valency lexicon. For example, in the PDT-
Vallex the noun (lexeme) balení ‘pack(ing)’ is represented by three valency frames
corresponding to three meanings of the noun, see (1)–(3). Different meanings can
sometimes be distinguished by different types or forms of complementation. In
(1), we encounter the semantic roles of ACT(or) and PAT(ient), optionally also
EFF(ect); in (2), MAT(erial). The valency frame in (3) is empty.
(1) balení1 ‘the process of packing’: ACT(GEN, INS, POSS) PAT(GEN, POSS)
EFFopt (na ‘on’ + ACC, …)
balení dárků.PAT rodiči.ACT ‘packing gifts by parents’
4. For further details and more complicated cases see Urešová (2011b: 39). For the treat-
ment of the relationship between a valency frame and a meaning of a verb in another valency
lexicon of Czech verbs, the so-called VALLEX, see Lopatková, Žabokrtský & Kettnerová
(2008: 16–17).
 Veronika Kolářová
(2) balení2 ‘a container’: MAT(GEN)

dárkové balení vína.MAT ‘a gift pack of wine’
(3) balení3 ‘design’: an EMPTY valency frame
kniha v brožurkovém balení ‘a book in a paperback binding’
1.5 Outline
This chapter, providing an in-depth analysis of the special valency behavior of
Czech deverbal nouns, is organized as follows. At the beginning we give an over-
view of primary as well as secondary general tendencies that determine the filling
of particular positions in the valency frames of deverbal nouns (Section 2). Then
we provide a brief description of the typical valency behavior of Czech dever-
bal nouns (Section 3). After that we deal with a manifestation of special valency
behavior (Section 4), focusing on special forms of participants (Section 4.1). Dif-
ferent meanings of deverbal nouns are analyzed (Section 5), and the impact of
special shifts on the meaning of a noun is studied (Section 6). Subsequently, we
provide an overview of nouns situated on the boundary between syntactic and
lexical derivation (Section 7). Finally, we summarize the factors influencing the
manifestation of special valency behavior of Czech deverbal nouns (Section 8).
2. General principles determining the surface forms of participants
Before discussing the valency behavior of Czech deverbal nouns, it is worth men-
tioning some general points. Czech is a highly inflectional language. Following
Karlík (2000: 183), a distinction between structural cases (nominative and accu-
sative) and non-structural cases (genitive, dative, locative, and instrumental) is
useful for a description of verbal valency. A similar distinction turns out to be
important in the nominal domain as well.
The primary general principle (Karlík 2000: 184) is as follows: within the pro-
cess of nominalization, the forms of verbal structural cases change whereas the
non-structural ones stay the same. This primary general principle explains the
typical shifts in the surface forms of participants and makes it possible to describe
the valency behavior of most Czech deverbal nouns.
Secondary general principles involve various special shifts in the surface forms
of participants; we formulate these principles as follows:
1. A verbal structural or non-structural case changes to a semantically more dis-

tinct (or less ambiguous) non-structural adnominal case, especially a non-
prepositional dative and a prepositional phrase.
Chapter 2. Special valency behavior of Czech deverbal nouns 
2. Verbal non-structural case changes to the standard form of adnominal modi-

fication, a non-prepositional genitive, or a possessive adjective/pronoun.
3. In general, the number of possible expressions of nominal modification is
higher than the number of corresponding verbal ones. Nominal comple-
mentation referring to an individual or an object expressed primarily by a
noun typically combines a non-prepositional case and a prepositional phrase.
Nominal complementation referring to a proposition can combine up to four
forms, namely a content clause (CC), an infinitive (INF), a non-prepositional
genitive (GEN), and a prepositional phrase (PP), although the corresponding
participants of their source verbs allow only three of them; for further details
see Kolářová (2010: 63–74).
3. T
ypical valency behavior and typical shifts in the surface forms
of participants
The valency behavior referred to as typical can be observed with the nouns derived
by syntactic derivation (group (i) in Section 1.3). This group includes mainly pro-
ductively derived nouns, but sometimes non-productively derived nouns as well.
The meaning of these nouns (or one of its meanings, if there are several) is parallel
to the meaning of their source verbs in that they denote an action or a state, on a
par with their source verb.
When determining the valency frames of deverbal nouns denoting an action
or a state, the same complementation as with verbs is envisaged. These nouns are
expected to inherit all participants that are present in the valency frame of their
source verbs, including the “verbal” character of the participants such as Actor,
Patient, and Addressee (unlike the nouns with special nominal valency comple-
mentation described in Section 4.4).
However, the forms of the participants in a noun phrase undergo some regular
shifts that can be described in terms of rules. These rules have been discussed by
many authors, e.g. Šmilauer (1966: 172), Novotný (1980), Panevová (2002: 30–32),
and Kolářová (2010: 54–56). The rules presented here follow the overview for-
mulated by Karlík & Nübler (1998: 107–111), with additional remarks by other
authors.
Rule (A): A valency slot using the nominative in an active or passive verbal
construction changes within the process of nominalization to a valency slot using
the genitive, see (4)–(6).
(4) (náš) Petr píše → psaní (našeho) Petra
(our) Peter-nom writes writing-nom (our-gen) Peter-gen
‘(our) Peter writes → writing of (our) Peter’
 Veronika Kolářová
(5) psi štěkají → štěkání / štěkot psů

dogs-nom bark barking/bark dog-gen.pl
‘dogs bark → barking/bark of dogs’
(6) dopis je psán5 → psaní dopisu
letter-nom is written writing-nom letter-gen
‘the letter is being written → writing of the letter’
Rule (B): Complementation expressed in a verbal construction by a noun (phrase)

or a content clause that is neither a subject in the nominative nor an object in
the non-prepositional accusative does not change its form in nominalization, see
(7)–(10).
(7) žák naslouchá učiteli

pupil-nom listens-prs teacher-dat
‘The pupil listens to the teacher.’
→ naslouchání žáka učiteli
listening pupil-gen teacher-dat.sg
‘listening of the pupil to the teacher’
(8) touží po tom, aby vyhrál / vyhrát / po výhře
long-prs.3sg for it to-conj win-pst win-inf for victory-loc
‘He desires to win / he longs for victory.’
→ touha po tom, aby vyhrál / vyhrát / po výhře
desire for it to-conj win-pst win-inf for victory-loc
‘desire to win/for victory’
(9) vnikl do pokoje → vniknutí do pokoje
enter-pst.3sg into room entering into room
‘He entered the room → entering the room’
(10) Pacient byl ošetřen lékařem / od lékaře.
patient was treated doctor-inst / from doctor-gen
‘The patient was treated by a doctor.’
→ ošetření pacienta lékařem / od lékaře
treating patient-gen doctor-inst from doctor-gen
‘treating of the patient by a doctor’
5. The ACC (dopis) in an active verbal construction psát dopis ‘to write a letter’ corresponds
to the GEN (dopisu) in the nominalized construction psaní dopisu ‘writing of the letter’, which
can be understood as the typical shift ACC → GEN.
Chapter 2. Special valency behavior of Czech deverbal nouns 
Rule (C): In the nominalization of a verbal construction where one valency slot
has a nominative and the other a non-prepositional accusative, only one of the
participants can change to a non-prepositional genitive, see (11).6
(11) Náš Petr píše dopis.
‘Our Peter writes a letter.’
→ psaní našeho Petra / psaní dopisu
writing-nom our Peter-gen.sg writing-nom letter-gen.sg
‘writing of our Peter/writing of the letter’
→ *psaní našeho Petra dopisu
writing-nom our-gen.sg Peter-gen.sg letter-gen.sg
‘writing of our Peter of the letter’
In Czech, a nominal construction with two participants in the non-prepositional

genitive is possible only when one of them is also in the genitive in the source
verbal construction, as in (12); according to rule (B), such a valency slot does not
change its form.
(12) Zbavil ženu starostí.
relieve-pst.3sg woman-acc worry-gen.pl
‘He relieved the woman of worries.’
→ zbavení ženy starostí
relieving woman-gen worry-gen.pl
‘relieving the woman of worries’
Rule (D): All the valency slots of a source verbal construction may be deleted in the
corresponding nominal construction; this means that they need not be expressed
on the surface at all (see Panevová, this volume). However, most nouns produc-
tively derived from perfective transitive verbs need to have an overt complement if
none is known from the previous discourse (see Dvořák, this volume).
Rule (E): Adverbs convert to adjectives, as in (13). Sometimes adverbs can be
kept, especially when an adjective cannot be derived from the adverb, see (14).
(13) píše rychle → rychlé psaní
he writes quickly → quick writing
6. The second participant (not changing into a genitive) either takes another typical form,
e.g. psaní dopisu naším Petrem ‘writing of the letter by our Peter’, or is deleted. However, a com-
bination of PAT expressed by a possessive adjective or pronoun and ACT in the postnominal
genitive is not possible, e.g. *pacientovo.PAT ošetření lékaře.ACT ‘patient’s treatment of the
doctor’, see Panevová, this volume. The second participant of nouns that allow specific shifts
in the surface forms of their participants (Section 4.1) can also take a specific form, e.g. obdiv
diváků k hercům ‘admiration of the audience for actors’.
 Veronika Kolářová
(14) útočí zvenčí → útok zvenčí

he attacks from_the_outside attack from_the_outside
Rule (F): Subject to certain conditions, an adnominal genitive can alternate with
a possessive adjective (ADJPOSS) or a possessive pronoun (PRONPOSS), as in (15).
(15) psaní Petra / Petrovo psaní / jeho psaní
writing Peter-gen.sg Peter-adj.poss writing his-pron.poss writing
‘writing of Peter/Peter’s writing/his writing’
In this chapter, the formal changes reflecting the above rules from the perspective
of active verbal constructions are called “typical shifts in the surface forms of par-
ticipants”; these typical shifts are schematically captured in Figure 1.
As the genitive form can be the result of different shifts (i.e. both NOM →
GEN and ACC → GEN, and also the correspondence GEN → GEN), in some
nominal constructions syntactic ambiguity can occur, e.g. střílení vojáků ‘shooting
of the soldiers’, where the genitive can indeed correspond either to the Actor or to
the Patient.
Surface forms of participants of verbs
NOM GEN DAT ACC LOC INS PP INF CC ADV
POSS GEN DAT LOC INS PP INF CC ADV ADJ
Surface forms of participants of nouns
Figure 1. Typical shifts in surface forms of participants
4. Special valency behavior
Every manifestation of valency behavior different from the typical one presented
in Section 3 is treated here as a special valency behavior.7 The manifestation of
special valency behavior concerns two basic issues: changes in the meaning of a
noun, discussed in Sections 6 and 7, and properties of valency complementation
7. However, in certain cases the specific behavior is “typical” for some groups or types of
nouns; for example, the specifically nominal complementation in partitive genitives is typical
for nouns that denote a container (Section 4.4).
Chapter 2. Special valency behavior of Czech deverbal nouns 
of a noun. The latter involves three phenomena: special forms of valency comple-
mentation (Section 4.1), reduction of the number of slots in the valency frame of
a noun (Section 4.3), and change of the character of valency complementation
(Section 4.4). The issue of special word order within a nominalized structure (cf.
Kolářová 2010: 83) is left out in this chapter. Various factors influencing the mani-
festation of special valency behavior are summarized in Section 8.
4.1 Special shifts in surface forms of participants

The form of a participant does not always undergo typical shifts (Section 3), but
some special shifts can be recognized alongside the typical ones. The term “special
shift” covers the situation when the form of an adnominal participant differs from
the form of the corresponding verbal participant and, at the same time, the new
form does not correspond to any of the typical shifts.
Special shifts frequently occur with nouns derived from verbs by n on-productive
means; however, they can also occasionally occur with nouns derived by productive
means. Special forms sometimes coexist with typical forms (Section 4.2) but the
following overview describes only the special forms. The impact of special shifts on
the meaning of the noun is discussed in Section 6.
A full overview of special shifts in the surface forms of participants is given by
Kolářová (2010: 60–73).8 The special shifts can be divided into several categories.
First, we distinguish special shifts of participants that modify action nouns and
nouns on the boundary between syntactic and lexical derivation from those that
modify nouns with an incorporated participant.
Participants of action-denoting nouns and nouns on the boundary between
syntactic and lexical derivation undergo special shifts that can be classified accord-
ing to the participant’s meaning, which determines its primary form: the partici-
pant can refer to an individual or an object, expressed primarily by a noun; or to a
proposition, expressed primarily by a content clause.
Special shifts of participants expressed primarily by a noun are exemplified
in (16)–(21). These participants take the special form of a prepositional phrase,
dative case, or genitive case.
– ACC → PP
(16) obdivovat herce → obdiv k hercům
to admire actor-acc.pl admiration to actor-dat.pl
‘to admire actors → admiration for actors’
8. We provide lists of nouns that can be modified by a participant in the form corresponding
to a specific shift and document them by examples that occur in the subcorpora of CNC;
selected specific shifts are also supported with statistical data.
 Veronika Kolářová
– ACC → DAT
(17) upozornit cestující → upozornění cestujícím
to warn passenger-acc.pl warning passenger-dat.pl
‘to warn passengers → warning addressed to passengers’
– GEN → DAT
(18) otázat se kamaráda → otázka kamarádovi
to ask friend-gen.sg question friend-dat.sg
‘to ask the friend → question addressed to the friend’
– INS → GEN
(19) kývnout hlavou → (naznačit) kývnutím hlavy
to nod head-ins.sg (to hint by) nod-ins.sg head-gen.sg
‘to nod the head → (to hint by) nodding of the head’
– GEN / DAT / INS → PP
(20) křivdit komu → křivda na kom
to do wrong sb-dat wrongdoing at sb-loc
‘to do wrong to sb → wrongdoing at sb’
(21) opovrhovat kým → opovržení nad kým

to despise sb-ins contempt over sb-ins
‘to hold sb in contempt → contempt for sb’
The special shifts of participants expressed primarily by a noun are schematically

captured in Figure 2 by solid arrows, whereas typical shifts are marked by dashed
arrows.
NOM GEN DAT ACC LOC INS PP INF CC ADV
POSS GEN DAT LOC INS PP INF CC ADV ADJ
Figure 2. Special shifts of participants expressed primarily by a noun
Special shifts of a participant expressed primarily by a content clause are

given in (22)–(25). The participant takes the special form of a prepositional
phrase, g enitive case, infinitive, or a content clause. We assume the participant is
Chapter 2. Special valency behavior of Czech deverbal nouns 
a participant in a propositional character (PPC)9 even if it is expressed by a noun

or a prepositional phrase.
– PPC → PPC(PP)
(22) Nabídl Janovi, aby spolupracoval / spolupracovat /
offered Jan-dat to-conj cooperate-pst cooperate-inf
spolupráci. → nabídka (Janovi), aby spolupracoval /
cooperation-acc offer (Jan-dat) to-conj cooperate-pst
spolupracovat / spolupráce / na spolupráci /
cooperate-inf cooperation-gen on cooperation-acc
ke spolupráci
to cooperation-dat
‘He offered Jan to cooperate/cooperation. → offer to cooperate/
of cooperation to Jan’
– PPC → PPC(GEN)
(23) Navykl si, že kouří / kouřit / na kouření. → návyk,
got_used refl that smokes smoke-inf to smoking-acc habit
že kouří / kouřit / kouření / na kouření
that smokes smoke-inf smoking-gen to smoking-acc
‘He got used to smoking. → the habit of smoking’
– PPC → PPC(INF)
(24) Ne-nadál se, že se dožije…/ dožití…
neg-hope-pst refl that refl lives_until living_until-gen
→ naděje, že se dožije.../ dožít se.../
hope that refl lives_until live_until-inf refl
dožití…/ na dožití…
living_until-gen for living_until-acc
‘He did not expect that he would live until… → the hope that he would live
until…/to live until…’
9. This participant is in most cases marked by the functor PAT but sometimes it can be
marked by the functor EFF. The functor EFF is used when a noun modified by PPC is derived
from a verb of communication that undergoes splitting of the “theme” (what is being talked
about) and “dictum” (what is being said about it); the characteristic feature of these construc-
tions lies in the fact that one of their participants – participant “message” – occupies two
valency slots (the “theme” is marked by the functor PAT, the “dictum” is marked by the functor
EFF, e.g. vyprávění o babičce.PAT, že byla.EFF pracovitá ‘talking about the grandmother that she
was hardworking’, see Kettnerová 2009).
 Veronika Kolářová
– PPC → PPC(CC)
(25) Odvážil se skočit / skoku / ke skoku. → odvaha,
dare-pst refl jump-inf jump-gen to jump-dat courage
aby skočil / skočit / skoku / ke skoku
to-conj jump-pst jump-inf jump-gen to jump-dat
‘He dared to jump. → courage to jump’
We assume that the verbal counterpart of a special adnominal form is PPC as

a whole, with all its surface forms. This assumption is schematically captured in
Figure 3 by the framed box containing all potential verbal forms of PPC. Typical
shifts are marked by dashed arrows. Figure 3 captures only the most frequent spe-
cial shift of PPC, i.e. the shift PPC → PPC(PP); for the diagrams of other special
shifts of PPC see Kolářová (2010: 68–69).
Verbal surface forms of PPC
NOM / GEN / DAT / ACC / INS / Ø INF CC
POSS GEN DAT INS PP INF CC
Adnominal surface forms of PPC
Figure 3. Special shift PPC → PPC(PP)
Nouns with an incorporated participant, for example, actor nouns (cf.

Section 4.3, paragraph ii), are derived by lexical derivation and represent a special
category. Since the “genitive” position left by the incorporated participant is free,
another participant takes either this form or the alternative form of a possessive
adjective or a possessive pronoun. These nouns are exemplified in (26)–(28).
– DAT → GEN / ADJPOSS/PRONPOSS

(26) dluží bance → dlužník banky
owes-prs bank-dat debtor bank-gen
‘he owes to a bank → debtor of a bank’
(27) (po)radí prezidentovi → poradce prezidenta /
advises-prs president-dat adviser president-gen
Chapter 2. Special valency behavior of Czech deverbal nouns 
prezidentův poradce / jeho poradce

president-adj.poss adviser his adviser
‘he advises the president → the adviser of a president/president’s adviser/
his adviser’
– PP → GEN / ADJPOSS/PRONPOSS
(28) spolupracuje s Petrem → spolupracovník Petra /
collaborates-prs with Peter collaborator Peter-gen
Petrův / jeho spolupracovník
Peter-adj.poss pron.poss collaborator
‘he collaborates with Peter → collaborator of Peter/Peter’s collaborator/
his collaborator’
The special shifts of participants that modify nouns with an incorporated partici-
pant are schematically captured in Figure 4 by solid arrows, whereas typical shifts
are marked by dashed arrows.
NOM GEN DAT ACC LOC INS PP
POSS GEN DAT LOC INS PP
Figure 4. Special shifts: nouns with an incorporated participant
4.2 Coexistence of typical and special forms of one participant

Nominal participants can take various grammatically correct forms correspond-
ing to both typical and special shifts. However, the two types of shift are not used
to the same extent. The coexistence of the typical and special forms of one par-
ticipant with respect to their frequency in the CNC subcorpora is captured in
Table 1. In situation A, the two forms have similar frequency, e.g. varování řidičů.
ADDR ‘warning of drivers’ versus varování řidičům.ADDR ‘warning addressed to
drivers’. In situations B and C, one of the forms is preferred over the other, which
is less f requent or rare though still possible. In situation B, the special form is less
 Veronika Kolářová
frequent: lov velryb ‘hunt of whales’ versus less common lov na velryby ‘hunt for
whales’, see (75)–(76). In situation C, the typical form is less frequent or rare, e.g.
Patient in the genitive with the nouns like dotyk ‘touch’ and obdiv ‘admiration’:
dotyk míče.PAT ‘touch of the ball’ versus more common dotyk s míčem ‘touch/
being in contact with the ball’, see (80)–(82); obdiv interiéru ‘admiration of the
decoration of the room’, but obdiv ?herců.PAT ‘admiration of actors’ versus more
common obdiv k hercům ‘admiration toward actors’. Interestingly, there are also
isolated cases (situation D) where the form corresponding to a typical shift is
ungrammatical and the only attested form corresponds to special shifts, e.g. spo-
lupracovník Petra ‘collaborator of Peter’/*spolupracovník s Petrem ‘collaborator
with Peter’, see (28); otázka kamarádovi ‘question addressed to the friend’/otázka
na kamaráda ‘question at the friend’/*otázka kamaráda.ADDR ‘question of the
friend’, see (86); or obava z následků ‘fear from consequences’/*obava následků
‘fear of consequences’, see (95). The opposite situation E, in which only the forms
corresponding to typical shifts are grammatical, is very frequent; for example,
Patients modifying nouns of exchange preserve the typical postnominal genitive
form and do not normally allow a special form, cf. vydávání/výdej jídla ‘handing
out/distribution of meals’.
It should be noted that it is not always possible to combine the form cor-
responding to a special shift with certain forms of other participants (for further
details see Sections 6.2 and 7).
Table 1. Coexistence of typical and special forms of one participant

Situation Grammatical form Ungrammatical form
Preferred/frequent Possible/less
frequent or rare
A Typical form and – –

special form have
similar frequency
B Typical form Special form –
C Special form Typical form –
D Special form – Typical form
E Typical form – Special form
4.3 Reduction of the number of slots in the nominal valency frame

Two types of reduction of the number of slots in the valency frame of a noun can
be distinguished: (i) simple (pure) reduction, (ii) incorporation of a participant.
Chapter 2. Special valency behavior of Czech deverbal nouns 
(i) Simple (pure) reduction. The first case is the situation when a participant
simply disappears from the valency frame. For example, the non-productively
derived noun důvtip ‘ingenuity’ does not allow modification by Patient in any
form, unlike its source (reflexive) verb dovtípit se ‘to infer’ and its productively
derived counterpart dovtípení se ‘inferring’, see (29).
(29) a. dovtípit se něčeho.PAT
infer-inf refl sth-gen.sg
‘to infer something’
b. dovtípení se něčeho.PAT
inferring-nom.sg refl sth-gen.sg
‘inferring of something’
c. *důvtip něčeho.PAT
ingenuity-nom.sg sth-nom.sg
‘ingenuity of something’
In the case of some polyvalent nouns undergoing changes in their meaning, the
participants that are defined on the basis of their semantic rather than their syn-
tactic properties (esp. ADDR and EFF) tend to disappear from the valency frame
of the noun. In contrast, Patient which is also defined syntactically, is often kept.
In (30) the noun návrh ‘suggestion/proposal’ is modified by all three participants
inherited from the source verb of communication (PAT of this noun of communi-
cation is the participant having a propositional character, so it can be expressed by
a content clause (CC)). However, when the noun návrh ‘proposal’ instead denotes
a piece of text, e.g. a draft or a bill, ADDR tends to disappear, as in (31).
(30) návrh vlády.ACT soudu.ADDR, aby… CC.PAT
suggestion government-gen.sg court-dat.sg to-conj
‘suggestion made by the government addressed to the court to...’
(31) návrh zákona.PAT ke schválení
proposal-nom.sg law-gen.sg for approval
‘proposal of a law for approval’
(ii) Incorporation of a participant. Sometimes the semantics of a certain partici-
pant (esp. ACT or PAT) which is part of the valency frame of the source verb is
incorporated into the meaning of the derived noun. As a result, in the valency
frames of nouns with an incorporated participant, the incorporated valency slot is
eliminated (absorbed) and cannot be expressed (see Panevová, this volume). For
example, actor nouns incorporate agents, compare (32) with (33).
(32) Někdo.ACT učí matematiku.PAT
somebody-nom.sg teach-3sg.prs mathematics-acc.sg
‘Somebody teaches mathematics.’
 Veronika Kolářová
(33) učitel matematiky.PAT

teacher-nom.sg mathematics-gen.sg
‘mathematics teacher’
According to our hypothesis, any type of reduction of the number of slots implies
a change in the meaning of the noun (see also Karlík & Nübler 1998: 106).
4.4 Specifically nominal valency complementation

Some deverbal nouns undergo further changes in their meaning, and then the
nouns can behave similarly to the non-deverbal nouns in that the character of
their valency complementation changes from inherited “verbal” participants to
specifically nominal complementation.
Two types of specifically nominal valency complementation, Partitive and
Appurtenance, were elaborated by Piťha (1984: 236). Partitive (or MAT(erial)) is
an inner participant modifying nouns that denote a container, e.g. skupina lidí.
MAT ‘a group of people’. APP(urtenance) is a free modification expressing mere
relation to a person or a thing, e.g. organizace neslyšících.APP ‘organization of the
hearing-impaired’, or expressing possession, e.g. Petrovo.APP auto ‘Peter’s car’;
with some nouns this modification is obligatory and belongs to their valency
frame, e.g. otec dítěte.APP ‘father of the child’.
These special types of complementation, elaborated primarily for non-deverbal
nouns, can also be found with deverbal nouns undergoing changes in their mean-
ing. For example, in the case of nouns that in addition to an action or a state can
denote a container, an inner verbal participant, usually PAT, changes to the inner
nominal participant MAT. In (34) the imperfective noun balení ‘wrapping’ denotes
an action but in (35) the same noun denotes a container and takes a MAT.
(34) balení másla.PAT do obalu brigádníky.ACT
wrapping butter-gen.sg into cover-gen.sg volunteer-ins.pl
‘wrapping of butter into a wrap by volunteers’
(35) jedno balení másla.MAT
one-nom.sg package-nom.sg butter-gen.sg
‘one package of butter’
In addition, modification in the special form of the prepositional phrase o ‘about’
+ LOC with the meaning “concerning/about something” (marked for the time
being by the functor PAT in the PDT and the PDT-Vallex) can indicate a seman-
tic shift. This modification is probably used by analogy to the so-called “theme”
which modifies nouns derived from some verbs of communication, e.g. vyprávění
o té době ‘talking about that period’ (cf. Section 4.1, Footnote 9). However, this
modification can sometimes be used with nouns that are derived from verbs that
do not have such a modification in their valency frame, as in vzkaz o telefonátu
Chapter 2. Special valency behavior of Czech deverbal nouns 
‘message about a phone call’. We assume that the presence of this modification
signals a subtle change in the meaning of the noun (an action → an abstract result
of an action), see (64)–(65) and Kolářová (2010: 154–155).
5. Different meanings of a deverbal noun
There is a close relationship between changes in the meaning of a noun and the
special properties of its valency complementation. The meaning of nouns mani-
festing special valency behavior (i.e. special forms of valency complementation,
reduction of the number of slots in the valency frame of a noun, or a change in
valency complementation character, see Section 4) is expected to be different from
the meaning of the verbs these nouns are derived from; the nouns are expected at
least not to denote an action or a state as their source verbs do.
The differences in meaning of deverbal nouns are typically described in the lit-
erature as the difference between event and result nominals, see Grimshaw (1991),
Apresjan (1995: 193–203), Alexiadou (2001), Melloni (2011). Sketching out the
argument structure of Czech nominals, Procházková (2006: 16) uses the labels
“eventive” vs. “resultative” for various interpretations of primarily event-denoting
nouns, in contrast to pure “result” nouns with only a resultative interpretation
(they can denote only physical entities). In order to establish which sense(s) is/are
activated in a given context, Ježek & Melloni (2011) pay attention to the selectional
properties of the adjectival and verbal collocates of action nominals.
However, careful examination of concrete instances shows that the issue of
“different meanings” of deverbal nouns is a complex one. Therefore, we suggest
to not limit it to various figurative senses (cf. zásah cíle ‘hit of the target’ vs. zásah
právnímu vědomí občanů ‘a jolt to the legal conscience of citizens’) or to clearly rec-
ognizable differences as in the case of deverbal nouns referring to a physical entity
(semantically concrete nouns such as nouns denoting a person, place or thing,
for example, vybavení ‘equipment’). We also want to pay attention to the semantic
nuances such as “an abstract result of an action”, which is more difficult to capture.
The issue is how to recognize manifestations of different meanings. The types
of derivational suffixes of the nouns in question (Section 5.1) and their syntactic
behavior (Section 5.2) seem to be important factors in resolving this issue.
5.1 Derivational suffixes

When a deverbal noun is derived by a “meaningful” or a semantically specific suf-
fix, its meaning usually does not pose any problem. This is true especially for actor
nouns (e.g. with the suffix -tel), nouns denoting a place of an action/activity (e.g.
suffix -árna), and nouns denoting a tool (e.g. suffix -dlo).
 Veronika Kolářová
However, two basic types of Czech nouns derived from verbs by non-specific
suffixes-that is, the nouns derived from verbs by productive means (e.g. balení
‘wrapping’) and by non-productive means or the zero suffix (e.g. dražba ‘auction’,
výskyt ‘occurrence’)-very often have several meanings: sometimes, they denote an
action or a state as their source verbs do; sometimes they denote physical entities
related to actions, such as a person (vedení ‘management’, rada ‘council’), a thing
(pohoštění ‘refreshments’, stavba ‘building’, dodávka ‘van’), or a place (stoupání ‘slope’,
východ ‘exit’). Sometimes, it is not clear whether there is really a shift in meaning.
This is the case especially for nouns derived from verbs by non-productive means; for
example, non-productively derived nouns of communication (e.g. prosba ‘request’,
návrh ‘suggestion’) seem to denote an action (speaking) in one of their meanings,
compare (30) with (31) above. However, when they are compared to their produc-
tively derived counterparts (prošení ‘asking’, navrhování/navržení ‘suggesting’), it
seems as if their action-like meaning has taken on a subtle resultative nuance.
Moreover, a noun that obviously has different meanings (e.g. vybavení1
‘furnishing’/‘equipping’, vybavení2 ‘furnishings’/‘equipment’) can appear in a con-
text that does not provide sufficient information for disambiguation, as in (36).
Identification of the actual meaning of nouns occurring in annotated databases
is often complicated by surface deletions of participants of the noun (rule (D) in
Section 3).
(36) a. Do vybavení1 pokoje investoval velkou částku.
in furnishing room-gen invest-3sg.pst large amount-acc
‘He invested a large amount of money in furnishing the room.’
b. Do vybavení2 pokoje investoval
in furnishings room-gen invest-3sg.pst
velkou částku.
large amount-acc
‘He invested a large amount of money in the room furnishings.’
The relation of the derivational suffixes of Czech deverbal nouns to the types of
nominal derivation in Kuryłowicz’s sense is illustrated in Table 2. We propose
that nouns derived by specific suffixes belong only to lexical derivation, whereas
nouns derived by non-specific suffixes can belong to all three groups – syntactic,
lexical, and the boundary between the two. In other words, a noun derived by
non-specific suffixes (productive or non-productive) is supposed to be derived by
syntactic derivation in one of its meanings; in another of its meanings, it can be
derived by lexical derivation or situated on the boundary between syntactic and
lexical derivation.
Chapter 2. Special valency behavior of Czech deverbal nouns 
Table 2. Czech deverbal nouns: derivational suffixes vs. type of derivation
Type of Syntactic derivation Boundary Lexical derivation
derivation in between
Kuryłowicz’s syntactic
sense and lexical
derivation
Abstract nouns Abstract Esp. concrete nouns

nouns
Derivational
suffixes Actions States People Things Places
Non- Productive – vedení vyskytování examples in vedení – první stoupání

specific means podniku se Section 7 ‘management’ vydání ‘slope’
někým ‘occurring’ knihy
‘managing ‘the first
of the edition of
company the book’
by sb’ – jedno
– vydání balení
knihy másla
někým ‘one
‘publishing package of
of the book butter’
by sb’
– balení
dárků
někým
‘wrapping
of gifts
by sb’
Non- – dodávky výskyt ropy examples in rada – dodávka východ
productive zboží ‘occurrence Section 7 ‘council’ ‘van’ ‘exit’
means zákazníkům of oil’ – 5 výskytů
or firmou v korpusu
the zero ‘deliveries ‘5
suffix of goods to occurrences
customers in the corpus’
by the
company’
– východ
slunce
‘sunrise’
Specific E.g. -árna, – – – učitel otvírák čekárna
-ák, -tel ‘teacher’ ‘opener’ ‘waiting
room’
When differences in the meanings of the noun in question are expected but
not clearly expressed, we suggest that its particular meaning (and consequently the
type of its derivation) can be inferred from various aspects of its valency behavior;
see examples in Sections 5.2, 6.2, and 7.
 Veronika Kolářová
5.2 D
ifferences between nouns derived by productive
and non-productive means
When it comes to understanding of Czech deverbal nouns derived by productive or
non-productive means, different approaches are available, applied within different
theoretical frameworks; and even the term “derivation” is used in different ways.10
The differences between productively and non-productively derived nouns
are described in detail in Karlík (2002: 20–21), among other sources. As for the
meaning and valency properties of these nouns, the following phenomena are dis-
cussed most frequently:
When nouns derived by productive means denote an action, they take over
the characteristic “verbal” properties (Veselovská 2001: 18–23):11
–– they can be modified by an agent expressed by prepositionless instrumental;
–– they are sensitive to aspect;
–– they combine with the reflexive particle se/si;12
–– they allow adjectives that specify the content of the noun with respect to
time and iteration (časté přepisování ‘frequent rewriting’) as well as adjectives
that signal the presence of an agent (úmyslné přepisování ‘intentional/wilful
rewriting’);
–– they do not allow adjectives that apply to concrete, physical objects (*špinavé
přepisování ‘dirty rewriting’).
On the other hand, productively as well as non-productively derived nouns share
some properties, especially the possibility of the following modifications:
–– modification by demonstratives;
–– modification by content clauses;
–– modification by possessives or genitives,13 which can be interpreted as sub-
jects or objects of the related verbal form.
10. Veselovská (2001) and Karlík (2002: 14) follow a Chomskyan lexicalist hypothesis
(Chomsky 1972) and claim that in Czech only nouns derived by productive means are syn-
tactically derived, whereas nouns derived by non-productive means are derived in the lexicon
module. However, some Czech scholars (Sgall 1967: 70; Panevová 1980: 142; Novotný 1980;
Dokulil 1982: 261; Kolářová 2010: 43) assume that some non-productively derived nouns are
also derived by syntactic derivation, e.g. zpěv ptáků ‘singing of birds’ (action), příchod někoho
někam ‘arrival of somebody somewhere’ (action), výskyt ropy ‘occurrence of oil’ (state).
11. See also Karlík (2002), Panevová (1980: 142–143) and Křížková (1968: 132–143).
12. The occurrences found in our corpora show that the reflexive particle is often preserved
but it can be omitted, see (79).
13. Conditions on expressing an agent in the genitive form are not investigated here.
Chapter 2. Special valency behavior of Czech deverbal nouns 
The tests proposed by Grimshaw (1991) for distinguishing complex event nomi-
nals, event nominals and result nominals, which were elaborated for English
nominals, should also be mentioned. The tests were effectively applied to Czech
nouns by Procházková (2006: 17–24), who states that Czech eventive nouns fit into
Grimshaw’s classification according to some but not all of the diagnostics. Never-
theless, the tests are still very useful for the comparison of syntactic properties of
productively and non-productively derived nouns, see (37)–(38).
(37) a. čtení knihy hodinu
reading-nom book-gen.sg hour-acc.sg
‘reading of the book for an hour’
b. ??četba knihy hodinu
reading-nom book-gen.sg hour-acc.sg
‘reading of the book for an hour’
(38) a. utěšování pacientů doktorem
consoling-nom patient-gen.pl doctor-ins.sg
‘consoling of patients by a doctor’
b. ??útěcha pacientů doktorem
consolation-nom patient-gen.pl doctor-ins.sg
‘consolation of patients by a doctor’
Two points need to be taken into consideration in this comparison. First, when we
measure the frequency of noun occurrences in the corpora, we find that tokens of
productively derived nouns are much less frequent than tokens of non-productively
derived ones. Sometimes, there is not even one occurrence of a productively
derived noun while its non-productively derived counterpart is amply represented.
Table 3 presents the frequencies found in the CNC subcorpus SYN2010.
The second point is that non-productively derived nouns that are derived
from transitive verbs can sometimes be modified by an agent (ACT) in preposi-
tionless INS, which is typical of syntactic derivations denoting an action, e.g. the
noun prodej ‘sale’ in (39) and the noun obliba ‘liking/popularity’ in (40). We thus
hypothesize that these nouns have an action as one of their meanings. It is a mat-
ter of debate whether such non-productively derived nouns (e.g. obliba ‘liking/
popularity’) take over the function of their barely used productively derived coun-
terpart, even though this counterpart is theoretically possible.
(39) prodej aut.PAT zákazníkům.ADDR firmou.ACT
sale-nom car-gen.pl customer-dat.pl company-ins.sg
‘sale of cars to customers by a company’
(40) obliba prezidenta Putina.PAT ruskou veřejností.ACT
liking president Putin-gen.sg Russian public-ins.sg
‘popularity of President Putin among the Russian community’
 Veronika Kolářová
Table 3. Comparison of frequencies of productively and non-productively derived nouns

Nouns derived by productive means Nouns derived by
non-productive means
Imperfective nouns Perfective nouns
Lemma Frequency Lemma Frequency Lemma Frequency
létání 612 letění 0 let 5661

‘flying’ ‘flying’ ‘flight’
oblibování si 0 oblíbení si 0 obliba 2507
‘getting to like’ ‘getting to like’ ‘liking /
popularity’
odvažování se 0 odvážení se 0 odvaha 4658
‘daring’ ‘daring’ ‘courage’
tázání se 94 otázání se 0 otázka 41875
‘asking’ ‘asking’ ‘question’
přicházení 11 přijití 0 příchod 5043
‘coming’ ‘coming’ ‘arrival’
6. Special shifts and their impact on the meaning of nouns
The main issue discussed in this section is whether a special shift in valency
implies change in the meaning of the noun. Jirsová (1966: 74) claims that the spe-
cial form of an adnominal participant is not coincidental: it is always associated
with a change in the noun’s meaning and represents one of the manifestations of
the new semantic nuance the noun receives.
There is no doubt that special shifts may be found with nouns undergoing
plain semantic shifts (Section 6.1). However, our question is whether a special
shift always implies change in meaning. The problem is that the difference in the
meaning of a noun modified by a participant in a special form is not always clearly
recognizable when compared with the same noun modified by that participant
in a typical form, as is demonstrated in Sections 6.2 and 7. The consequences of
special shifts for valency frames are discussed in Section 6.2.
6.1 Nouns undergoing plain semantic shifts

Sometimes, special forms of participants accompany nouns used in a figurative
sense, such as the various figurative senses of nouns expressing “negative touch”,
see (41) where the noun zásah modified by Patient in the dative, PAT(DAT), cor-
responding to the special shift, means ‘jolt’14 rather than ‘hit’; the latter m
eaning
14. The source verb itself can be used in the same figurative sense but in that case it does not
lead to the shift in the form of the verbal Patient.
Chapter 2. Special valency behavior of Czech deverbal nouns 
is exemplified in (42), where the noun zásah ‘hit’ is modified by PAT(GEN), cor-
responding to the typical shift. Using PAT(GEN) with the noun zásah ‘jolt’ is
ungrammatical, as is using PAT(DAT) with the noun zásah ‘hit’, see (43) and (44).
(41) zásah právnímu vědomí.PAT občanů
jolt-nom legal conscience-dat.sg national-gen.pl
‘a jolt to the legal conscience of citizens’
(42) zásah cíle.PAT
hit-nom target-gen.sg
‘hit of the target’
(43) *zásah právního vědomí.PAT občanů
jolt-nom legal conscience-gen.sg national-gen.pl
‘a jolt of the legal conscience of citizens’
(44) *zásah cíli.PAT
hit-nom target-dat.sg
‘hit to the target’
The fact that a noun allows participants in special forms is related to the category
of verbal aspect: in the case of nouns derived by productive means, special shifts
are mostly allowed with nouns derived from perfective verbs, but not with their
imperfective counterparts. It may be related to the fact that perfective nouns have
a greater tendency to lose an action meaning than the imperfective nouns. Some-
times, in addition to actions they can denote mental states or dispositions.15 Such a
semantic shift (i.e. action → mental state or disposition) can be accompanied by a
special form of a participant. In (45) the perfective noun politování ‘pitying’ denotes
an action and is modified by PAT(GEN) corresponding to a typical shift, unlike in
(46), where the noun politování ‘regret’ denotes the mental state and is modified by
PAT expressed by PP (nad ‘about’ + INS), corresponding to a special shift.16
(45) politování obětí.PAT útoku
pitying-nom victim-gen.pl attack-gen.sg
‘pitying of victims of the attack’
(46) vyjádřit politování nad oběťmi.PAT
express-inf regret-acc.sg about victim-ins.pl
‘to express regret about victims’
15. Křížková (1968: 143) claims that the perfective nouns often express states as the results of
an action, e.g. konstatovat s uspokojením ‘to state with satisfaction’.
16. We suppose that the nouns that represent a nominal part of a support verb construction,
such as politování ‘regret’ in vyjádřit politování ‘to express regret’ (46), have their own nominal
valency (Cinková & Kolářová 2006); we do not consider it the valency of the whole support
verb construction.
 Veronika Kolářová
When the semantic shift (change in the meaning) is indisputable, we consider the
nouns to be lexical derivates.
6.2 Special shifts and valency frames

In this section the assumption that one valency frame corresponds to one mean-
ing (sense) of a noun is discussed, with the focus on nouns with special forms
of participants. The general issue is whether two different forms of a participant
(two typical forms, two special forms or one typical and one special form, i.e. the
case focused on here) represent mere variants that can be captured in one valency
frame, or whether there are reasons for treating them within two different valency
frames (Kolářová 2010: 174). Subsequently, the question arises whether a new
valency frame created due to a special form of a participant corresponds to a dif-
ferent meaning of the noun at issue.
The indisputable reasons for creating a new valency frame are the following:
–– a clear change in the meaning of a noun (Sections 5 and 6.1);

–– reduction of the number of valency slots (Section 4.3);
–– a change of the character of valency complementation to exclusively nominal
(Section 4.4).
However, there are cases of nouns with a special form of a participant that keep
all participants inherited from the source verb; the participants do not change to
specifically nominal ones. Moreover, the difference between the noun’s meaning
and the verb’s meaning is not clear (Section 7). Especially when not all the par-
ticipants are expressed,17 the meaning of the noun modified by a participant in a
special form seems to be the same as the meaning of the noun modified by that
participant in a typical form, as in (47) and (48).
(47) americká podpora Evropy.PAT
American support Europe-gen.sg
‘American support of Europe’
(48) americká podpora Evropě.PAT
American support Europe-dat.sg
‘American support to Europe’
We propose that a further reason for creating a new valency frame is the different
syntactic behavior of a noun with a participant in a special form, when compared
17. Relational adjectives, such as American in (47) and (48), are not considered participants;
for discussion of this issue, see Kolářová (2010: 80–81).
Chapter 2. Special valency behavior of Czech deverbal nouns 
with the syntactic behavior of the same noun with the participant in a typical form.
We suggest that if a participant in a special form cannot combine with the same
set of forms of other participants from one valency frame as typical forms, a new
valency frame should be created in the lexicon. Moreover, we suppose that when
such a differentiating syntactic behavior is detected, it should also be understood
as a manifestation of a difference in meaning, even if it is only a slight nuance.
One of the syntactic tests we use is the possibility of combining an Actor
expressed by the prepositionless instrumental case with a participant expressed
either by the typical form or by the special form. Other tests are presented in
Section 7.
The noun podpora ‘support’, for example, can combine the typical (genitive)
form of Patient with an Actor expressed in the instrumental when it denotes an
action, see (49), unlike the same noun when it is action-like, however slightly dif-
ferent meaning. Such a noun does not typically allow the combination of Patient
in the special (dative) form and Actor in INS, see (51).18 A noun such as podpora
‘support’ should therefore be described in the valency dictionary with the two
valency frames illustrated in (50) and (52).
(49) podpora Evropy.PAT Amerikou.ACT
support-nom.sg Europe-gen.sg America-ins.sg
‘support of Europe by America’
(50) podpora1 ‘support’:
ACT(GEN, POSS, INS) PAT(GEN, POSS)
(51) ??podpora Evropě.PAT Amerikou.ACT
support-nom.sg Europe-dat.sg America-ins.sg
‘support to Europe by America’
(52) podpora2 ‘support’: ACT(GEN, POSS) PAT(DAT)
In contrast, sometimes different forms of complementation are considered to be

mere variants because they can be combined with other participants to the same
extent. This is the case of the noun vyučování ‘teaching’ which can combine two
forms of Patient, the genitive and the dative, both classified as typical, with Actor
expressed by a prepositionless instrumental, as in (53) and (54). Thus the two
18. Surprisingly, three examples of such allegedly ungrammatical construction occur in

the CNC subcorpora, e.g. vzpomínky na výraznou podporu vznikajícímu izraelskému státu
Československem v letech 1947–1949 (SYN2006PUB) ‘memories of substantial support to the
rising State of Israel by Czechoslovakia in years 1947–1949’. It’s true that this example is ac-
ceptable; however its shorter variant, e.g. podpora Izraeli Československem ‘support to Israel by
Czechoslovakia’, is not acceptable for the author of this chapter.
 Veronika Kolářová
ifferent forms of Patient can be captured in one valency frame as variants, as

d
shown in (55).19
(53) vyučování latiny.PAT učitelem.ACT
teaching-nom.sg Latin-gen.sg teacher-ins.sg
‘teaching of Latin by a teacher’
(54) vyučování latině.PAT učitelem.ACT
teaching-nom.sg Latin-dat.sg teacher-ins.sg
‘teaching of Latin by a teacher’
(55) vyučování1 ‘teaching’: ACT(GEN, POSS, INS) PAT(GEN, POSS, DAT)
ADDR(GEN, POSS)
We believe that this theoretical proposal for treating nouns with special forms
of participants in the valency lexicon makes it possible to properly capture their
valency properties. However, we are aware that such a fine-grained approach can
cause indecision when particular valency frames are being assigned to the nouns
that occur in the annotated databases, especially when not all the participants of
a noun are expressed. However, the problem with unexpressed participants also
holds for nouns that do not allow special forms of participants, as shown in (36).
7. O
verview of nouns on the boundary between syntactic
and lexical derivation
This section is devoted to nouns that represent a transitional category, namely

nouns situated on the boundary between syntactic and lexical derivation. This cat-
egory of nouns was envisaged by Panevová (2002), but a detailed description of the
valency behavior of nouns belonging to this category has not been provided yet.
In this section we show that these nouns – mostly non-productively, but also
a few productively derived – exhibit truly “transitional” semantic and valency
properties. They are thus not only rightly situated on the boundary between
syntactic and lexical derivation, but they also require special lexicographic treat-
ment. These nouns allow participants in a special form but they neither have
a plain action meaning nor exhibit a clear shift in meaning (such as toward a
figurative meaning). They do not denote physical entities either. As for their
valency behavior, it is not typical; the nouns do not behave completely as syn-
tactic derivates, but they take over some properties typical of syntactic derivates.
19. For systematic combinatory restrictions, see Section 3 and Panevová, this volume.
Chapter 2. Special valency behavior of Czech deverbal nouns 
At the same time, these nouns do not exhibit valency properties typical of plain
lexical derivates either.
We want to examine here how a special form behaves in combination with
various forms of modification implying an action reading, using the same tests as
in Sections 5.2 and 6.2. We compare (a) nominal constructions in which nouns
with typical “verbal” properties that obviously denote an action are modified
by participants in forms corresponding to typical shifts, with (b) nominal con-
structions in which the same nouns are modified by complementation implying
an action reading and, in addition, by a participant in a form corresponding to
a special shift (Section 4.1). We propose the following parameters as the main
criteria for identifying barely recognizable semantic nuances: Nouns denoting an
action are expected to allow modification by ACT(INS) and adjectives that spec-
ify the content of the noun with respect to time and iteration (adjectives such as
‘frequent’, ‘repeated’, ‘constant’, ‘occasional’, or ‘everyday’). Possibly, they can also
admit adverbs such as předem ‘in advance’. In contrast, some nouns (esp. nouns
of communication) that tend to exhibit a resultative meaning can be identified
by the possibility of modification expressed by a special PP (o ‘about’ + LOC)
‘concerning/about something’ which is not present in the valency frame of their
source verbs (Section 4.4). Additionally, nouns denoting a physical entity related
to an action can be modified by adjectives that apply to concrete, physical objects
(e.g. česky psaná varování zlodějům ‘warnings to thieves written in Czech’). Nouns
derived by non-productive means are also compared with their productively
derived counterparts. For this pilot study, nouns representing the most frequent
special shifts were chosen.
7.1 Typical shift ACC → GEN vs. special shifts ACC → DAT or ACC → PP
Nouns modified by a participant whose form undergoes the special shift ACC →
DAT (or ACC → PP) are illustrated by two productively derived nouns of com-
munication: the perfective noun upozornění ‘warning’ and the both perfective and
imperfective noun varování ‘warning’; and one non-productively derived noun of
evaluation, the noun pochvala ‘praise’.
Example (56) illustrates the perfective noun upozornění ‘warning’ modified
by participants ACT and ADDR, the form of which corresponds to typical shifts.
This noun can also be modified by ADDR in the forms corresponding to special
shifts, DAT and PP, see (57). However, a combination of these two special forms of
ADDR with ACT(INS) is not possible, see (58)–(59).
(56) upozornění myslivců.ADDR Dr. Novákem.ACT
warning-nom hunter-gen.pl Dr. Novak-ins.sg
‘warning of hunters by Dr. Novak’
 Veronika Kolářová
(57) upozornění myslivcům.ADDR / pro myslivce.ADDR

warning-nom hunter-dat.pl for hunter-acc.pl
‘warning addressed to hunters’
(58) *upozornění myslivcům.ADDR Dr. Novákem.ACT
warning-nom hunter-dat.pl Dr. Novak-ins.sg
‘warning addressed to hunters by Dr. Novak’
(59) *upozornění pro myslivce.ADDR Dr. Novákem.ACT
warning-nom for hunter-acc.pl Dr. Novak-ins.sg
‘warning addressed to hunters by Dr. Novak’
The noun upozornění ‘warning’ allows modification by the adverb předem ‘in
advance’, but only the combination with ADDR(GEN) is then acceptable, see (60).
Constructions in which the adverb is combined with ADDR in the forms corre-
sponding to special shifts are ungrammatical, as shown in (61) and (62). On the
other hand, the adjectives opakovaný ‘repeated’ and opětovný ‘repeated’ are allowed
in combination with the special form of ADDR, i.e. with ADDR(DAT), see (63).
(60) upozornění poslanců.ADDR předem, že...CC.PAT
warning-nom.sg deputy-gen.pl in advance that
‘warning of deputies in advance that...’
(61) *upozornění poslancům.ADDR předem, že...CC.PAT
warning-nom.sg deputy-dat.pl in advance that
‘warning addressed to deputies in advance that...’
(62) *upozornění pro poslance.ADDR předem, že...CC.PAT
warning-nom.sg for deputy-acc.pl in advance that
‘warning addressed to deputies in advance that...’
(63) po opětovném upozornění zmocněnci.ADDR
after repeated-loc.sg warning-loc representative-dat.sg
‘after repeated warning addressed to the representative’
The situation changes when the noun upozornění ‘warning’ is modified by a spe-
cial PP (o ‘about’ + LOC) ‘concerning/about something’ (marked by functor PAT).
The presence of this modification possibly signals a subtle change in meaning (an
action → an abstract result of an action). A combination of this modification and
ADDR(DAT) or ADDR(PP) is possible, as in (64) and (65), but a combination
with ADDR(GEN) is marked, see (66).
(64) upozornění řidičům.ADDR o výměně.PAT
warning-nom.sg driver-dat.pl about changing-loc.sg
řidičských průkazů
driving licence-gen.pl
‘warning addressed to drivers concerning changing of driver’s licenses’
Chapter 2. Special valency behavior of Czech deverbal nouns 
(65) upozornění pro řidiče.ADDR o výměně.PAT…

warning-nom.sg for driver-acc.pl about changing-loc.sg
‘warning addressed to drivers concerning changing…’
(66) ??upozornění řidičů.ADDR o výměně.PAT...
warning-nom.sg driver-gen.pl about changing-loc.sg
‘warning of drivers concerning changing…’
Noteworthily, the special modification by ADDR(DAT) or ADDR(PP) is only

acceptable with the perfective noun upozornění ‘warning’ but it cannot appear
with its imperfective counterpart, *upozorňování řidičům ‘warning addressed to
drivers’.
Another factor leading to the use of the special modification by ADDR(DAT)
or ADDR(PP) instead of the typical modification by ADDR(GEN) could be the
tendency to avoid the syntactic ambiguity between Actor and Addressee interpre-
tation of a postnominal genitive. This tendency is especially characteristic of nouns
of communication (both participants, Actor as well as Addressee, are animate).
Another noun of communication, the productively derived both perfective and
imperfective noun varování ‘warning’, behaves in a similar way. It does not allow
a combination of special modification by ADDR(DAT) together with ACT(INS),
but a combination of ADDR(DAT) and adjectives with the meaning ‘constant’ or
‘repeated’ is possible. Moreover, it allows a combination of ADDR(DAT) and the
adjective ‘intentional’, see (67). In contrast, when denoting a physical entity, the
noun varování ‘warning’ with ADDR(DAT) can also be modified by adjectives
implying resultative or “physical” interpretation of the noun, as in (68).
(67) úmyslná varování plagiátorovi.ADDR, že...CC.PAT
intentional warning-nom.pl plagiarist-dat.sg that
‘intentional warnings addressed to the plagiarist that…’
(68) česky psaná varování zlodějům.ADDR
in_Czech written warning-nom.pl thief-dat.pl
‘warnings addressed to thieves written in Czech’
Pochvala ‘praise’ is a noun derived by non-productive means that allows special

modification in contrast to its counterpart pochválení ‘praising’, a perfective noun
derived by productive means. Example (69) illustrates the noun pochválení ‘prais-
ing’ modified by the participants ACT and PAT in the forms corresponding to
typical shifts. Pochvala ‘praise’ can be modified by PAT(GEN) which can be both
animate (pochvala ministra ‘praise of minister’) and inanimate (pochvala kravaty
‘praise of necktie’), see (70). We determined that the noun pochvala ‘praise’ modi-
fied by PAT(GEN) in combination with ACT(INS), as in (71), does not appear
in the subcorpora of CNC. The noun pochvala ‘praise’ can be also modified by
PAT(DAT), corresponding to the special shift ACC → DAT, but the PAT must
 Veronika Kolářová
probably be animate in such a case, see (72). The construction in (73), where
PAT(DAT) is combined with ACT(INS), is ungrammatical. The dative form of
PAT can be influenced by the valency of the verbal part of support verb construc-
tions in which the noun pochvala ‘praise’ occurs, e.g. udělit pochvalu + DAT ‘lit. to
give praise to’.
(69) pochválení ministra.PAT prezidentem.ACT
praising-nom.sg minister-gen.sg president-ins.sg
‘praising of the minister by the President’
(70) pochvala ministra.PAT / kravaty.PAT
praise-nom.sg minister-gen.sg necktie-gen.sg
‘praise of the minister/of the necktie’
(71) ?pochvala ministra.PAT prezidentem.ACT
praise-nom.sg minister-gen.sg president-ins.sg
‘praise of the minister by the president’
(72) pochvala ministrovi.PAT / *kravatě.PAT
praise-nom.sg minister-dat.sg necktie-dat.sg
‘praise addressed to the minister/*to the necktie’
(73) *pochvala ministrovi.PAT prezidentem.ACT
praise-nom.sg minister-dat.sg president-ins.sg
‘praise addressed to the minister by the president’
7.2 Typical shift ACC → GEN vs. special shift ACC → PP

Example (74) shows the imperfective noun lovení ‘hunting’ derived by produc-
tive means with ACT and PAT forms corresponding to typical shifts. Its non-
productively derived counterpart lov ‘a hunt’ has the same valency structure in
(75). In addition, this noun can be modified by PAT in PP (na ‘on’ + ACC), corre-
sponding to the special shift ACC → PP, see (76), in which case it can be modified
by the adjective ‘occasional’, see (77). However, a combination of PAT(PP) with
ACT(INS) is ungrammatical for lov ‘a hunt’, see (78).
(74) lovení velryb.PAT rybáři.ACT
hunting-nom.sg whale-gen.pl fisherman-ins.pl
‘hunting of whales by fishermen’
(75) lov velryb.PAT rybáři.ACT
hunt-nom.sg whale-gen.pl fisherman-ins.pl
‘hunt of whales by fishermen’
(76) lov rybářů.ACT na velryby.PAT
hunt-nom.sg fisherman-gen.pl on whale-acc.pl
‘hunt of fishermen for whales’
Chapter 2. Special valency behavior of Czech deverbal nouns 
(77) občasný lov na velryby.PAT

occasional hunt-nom.sg on whale-acc.pl
‘occasional hunt for whales’
(78) *lov na velryby.PAT rybáři.ACT
hunt-nom.sg on whale-acc.pl fisherman-ins.pl
‘hunt for whales by fishermen’
7.3 Typical correspondence GEN → GEN vs. special shift GEN → PP

Examples (79) and (80) illustrate the productively derived perfective noun dotk-
nutí (se) ‘touching’ and the non-productively derived noun dotyk ‘touch’ modified
by PAT(GEN) and ACT(INS). This kind of modification (rare with dotyk) sug-
gests their reading as action nouns. The possibility of combining with ACT(INS)
is interesting in this case because the source verb dotknout-IPFV se /dotýkat-PFV
se ‘to touch’ is an intransitive inherently reflexive verb that can only be “passiv-
ized” in its figurative sense; see (81), where dotčen means ‘offended’. The noun
dotyk ‘touch’ can also be modified by PAT in PP (s ‘with’ + INS), corresponding to
the special shift GEN → PP, and at the same time by the adjectives ‘constant’ and
‘repeated’ as in (82) and (83). However, a combination of PAT(PP) with ACT(INS)
is ungrammatical, see (84).
(79) dotknutí (se) míče.PAT hráčem.ACT

touching-nom.sg refl ball-gen.sg player-ins.sg
‘touching of the ball by the player’
(80) dotyk míče.PAT hráčem.ACT

touch-nom.sg ball-gen.sg player-ins.sg
‘touch of the ball by the player’
(81) Petr byl dotčen jednáním toho člověka.

Peter was offended action-ins.sg that-gen man-gen.sg
‘Peter was offended by the action of the man.’
(82) pravidla nařizující neustálý dotyk se zemí.PAT

rule-nom.pl ordering continual touch with ground-ins.sg
‘rules ordering continual touch with ground’
(83) opakovaný dotyk s látkou.PAT

repeated touch-nom.sg with cloth-ins.sg
‘repeated touch with a cloth’
(84) *dotyk s míčem.PAT hráčem.ACT
touch-nom.sg with ball-ins.sg player-ins.sg
‘touch/being in contact with the ball by a player’
 Veronika Kolářová
ypical correspondence GEN → GEN vs. special shifts GEN → DAT

7.4 T
or GEN → PP
Another set of examples showing differences between productively and non-
productively derived deverbal nouns is given in (85)–(89). Construction (85)
illustrates the reflexive perfective noun derived by productive means otázání se
‘asking’, modified by the participant ADDR in the typical, genitive form.20 The
construction with ADDR(GEN) is ungrammatical with the corresponding non-
productively derived noun otázka ‘question’, see (86); instead, the Addressee com-
monly takes DAT or PP (na ‘at’ + ACC) which correspond to special shifts, see
(87). The dative form of PAT could be influenced by the tendency to avoid ambi-
guity between ACT(GEN) and ADDR(GEN), which is characteristic of nouns of
communication, and by the valency of the verbal part of the support verb con-
structions with the noun otázka ‘question’, e.g. dát otázku + DAT ‘to ask a question’,
lit. ‘to give a question to sb’. ACT(INS) does not appear at all with the noun otázka
‘question’. Adjectives meaning ‘repeated’ or ‘frequent’ can combine with the noun
otázka ‘question’ when modified by ADDR in DAT or PP, see (88) and (89). The
last two constructions could represent a condensation of a construction involv-
ing an adjectival participle, e.g. častá otázka ‘frequent question’ can correspond to
často kladená otázka ‘frequently asked question’.
(85) otázání se dvou sedících mužů.ADDR

asking-nom.sg refl two-gen sitting-gen.pl man-gen.pl
‘asking of two sitting men’
(86) *otázka dvou sedících mužů.ADDR

question-nom.sg two-gen sitting-gen.pl man-gen.pl
‘question of two sitting men’
(87) dnešní otázka někomu.ADDR /

today’s question-nom sb-dat.sg
na někoho.ADDR, zda …CC.PAT
at sb-acc.sg whether
‘today’s question addressed to somebody whether …’
(88) básníkova.ACT opakovaná otázka lidem.ADDR

poet-adj.poss repeated question-nom people-dat.pl
‘poet’s repeated question addressed to the people’
20. However, such a construction is rare (only one example occurs in all of the CNC sub-
corpora).
Chapter 2. Special valency behavior of Czech deverbal nouns 
(89) nejčastější otázka prvnímu kosmonautovi.ADDR

most frequent question-nom first-dat.sg cosmonaut-dat.sg
‘the most frequent question addressed to the first cosmonaut’
7.5 Special shift PPC → PPC(PP)

In order to illustrate nominal modification by a participant having a propositional
character (PPC) and undergoing a special shift, we chose one productively derived
noun of communication, the perfective noun doporučení ‘recommending’, and one
non-productively derived noun denoting a mental state or disposition, the noun
obava ‘fear’.
The propositional participant of the perfective noun doporučení ‘recommend-
ing’ undergoes the special shift PPC (CC / INF / ACC) → PPC (CC / INF / GEN /
PP). Examples (90) and (91) illustrate this noun when modified by participants in
the forms corresponding to typical shifts. Besides the forms of PPC corresponding
to the typical shifts, i.e. CC, INF, and GEN, there is one more possible form of PAT,
namely PP (k ‘to’ + DAT), which corresponds to a special shift. Adjectives mean-
ing ‘frequent’, ‘constant’, or ‘repeated’ do not occur when the noun is modified by
PAT(PP); however, the modification by the adjective ‘lasting’ in (92), interestingly,
implies the progressive meaning of the noun. The combination of PAT(PP) and
ACT(INS) is ungrammatical, see (93).
(90) doporučení produktu.PAT firmou.ACT
recommending product-gen.sg company-ins.sg
‘recommending of the product by the company’
(91) doporučení volby.PAT rodinou.ACT
recommending election-gen.sg family-ins.sg
‘recommending who to vote by the family’
(92) již druhý týden trvající doporučení
already second week lasting recommendation
k nákupu.PAT akcií
to purchase-dat.sg share-gen.pl
‘recommendation lasting for two weeks to purchase shares’
(93) *doporučení k volbě.PAT rodinou.ACT
recommending to election-dat.sg family-ins.sg
‘recommendation who to vote by the family’
The propositional participant of the non-productively derived noun obava ‘fear’

undergoes the special shift PPC (CC / (INF) / GEN) → PPC (CC / (INF) / PP).
Example (95) shows that the noun obava ‘fear’ does not manifest the typical cor-
respondence of the adnominal and verbal GEN; this correspondence can be found
 Veronika Kolářová
only with the productively derived reflexive imperfective noun obávání se ‘fearing’
that allows the modification by PAT in the typical, genitive form, see (94).21 The
noun obava ‘fear’ allows several other forms of PAT: besides the typical forms CC
and INF it also allows PP (z ‘of ’ + GEN), which is a special shift. ACT(INS) does
not appear at all with this noun. However, the adjective ‘everyday’ is possible with
the noun obava ‘fear’ modified by PAT(PP), see (96).
(94) obávání se něčeho.PAT
fearing-nom.sg refl sth-gen.sg
‘being afraid of something’
(95) *obava něčeho.PAT
fear-nom.sg sth-gen.sg
‘fear of something’
(96) každodenní obava z něčeho.PAT
everyday fear-nom.sg of sth-gen.sg
‘everyday fear of something’
7.6 Discussion of results

When evaluating the examples of nouns derived from action verbs (upozornění-
PFV ‘warning’, varování-PFV/IPFV ‘warning’, pochvala ‘praise’, lov ‘hunt’, dotyk
‘touch’, otázka ‘question’, doporučení-PFV ‘recommending/recommendation’), we
must conclude that there is no example in which a noun modified by a partici-
pant in a special form satisfies all the criteria proposed for the identification of
an action meaning; in other words, none of them is an indisputable example of a
noun with an action meaning.
Contrary to expectation, nouns derived from action verbs as well as the noun
obava ‘fear’, which is derived from a psychological verb, typically allow adjectives
such as ‘frequent’, ‘repeated’, ‘constant’, ‘occasional’, or ‘everyday’.22 Moreover, these
nouns can be modified by a participant in the special form and by all other par-
ticipants from their valency frame at the same time (we saw that the form of ACT
cannot be INS; however, GEN or ADJPOSS/PRONPOSS are possible expressions of
ACT). The participants keep their “verbal” character and do not change to the
specifically nominal valency complementations MAT or APP (Section 4.4). We
also observed that the typical forms are sometimes rare or even ungrammatical.
21. However, such a construction does not occur in the CNC subcorpora that we used.
22. We must realize, though, that these adjectives can be used as a vague condensation of
more difficult constructions, accompanied by the deletion of some parts of the construction;
see the discussion of častá otázka ‘frequent question’ in Section 7.4.
Chapter 2. Special valency behavior of Czech deverbal nouns 
The meaning of the abstract nouns with special forms of participants is slightly
different from that of action nouns.
Since the characteristics described do not completely satisfy the criteria pro-
posed for the identification of an action meaning nor do they match the proper-
ties typical of a plain result(ative) meaning (a plain semantic shift or a physical
interpretation, reduction of the number of slots in the valency frame, or change
of the character of a valency complementation, see Sections 4.3 and 4.4), we can
confirm the hypothesis that these nouns should be regarded as representatives of a
transitional (and heterogeneous) category situated on the boundary between syn-
tactic and lexical derivation. Their semantic interpretation in most cases could be
specified as an abstract result of an action. This interpretation was distinguished
from a “concrete object” interpretation when a noun denotes a physical entity, for
example, česky psaná varování zlodějům ‘warnings to thieves written in Czech’.
We suppose that special forms of the participants themselves cannot serve as
a reason to regard the given deverbal nouns as the results of lexical derivation. We
assume that some “irregular” nominal forms occasionally replace other “more reg-
ular” forms in the language system; we determined two such situations (which can
also combine in one construction): (i) a non-productively derived noun is used
instead of a productively derived one (the productively derived noun is very rare
or not used at all, see examples in Table 3), or (ii) a special form of a participant
is used instead of a typical one (the typical form is rare or even ungrammatical,
see situations C and D in Table 1 and examples (86) and (95)). Such a substitution
entails syntactic behavior that, for grammatical reasons, does not correspond to
the typical syntactic behavior characteristic of the syntactic derivation (Section 3).
However, such nouns seem to have a similar function in a sentence as the out-
comes of regular syntactic derivation.
Regardless of the difficulties with identifying the exact meaning of these
nouns, we suggest creating a new valency frame in the valency lexicon for them
(Section 6.2) if a special form of one of their participants cannot combine with the
same set of forms of other participants from a valency frame that has already been
described, as the typical form of that participant. The prototypical example of a
form with which special forms of participants cannot combine is ACT(INS), and
we rely on this fact in our tests that manifest different syntactic behavior of nouns
with special shifts.
8. S
ummary of factors influencing manifestation
of special valency behavior
This study has revealed various factors influencing the use of the special forms
of complementation. These factors affect not only the nouns on the boundary
 Veronika Kolářová
between syntactic and lexical derivation, but also the nouns derived by lexical
derivation. They can be classified as follows:
a. Semantic properties of a noun: the semantic group it belongs to (such as nouns

of communication, e.g. otázka ‘question’; nouns denoting positive or negative
mental states or dispositions, compare naděje spolupráce ‘hope of cooperation’
in (24) with *obava následků ‘fear of consequences’ in (95)); plain semantic
shifts that deverbal nouns have undergone (e.g. the shift to a figurative sense
of nouns expressing “negative touch”, see zásah ‘jolt’ in (41) and zásah ‘hit’ in
(42); the shift action → mental state or disposition, see politování ‘pitying’ in
(45) and politování ‘regret’ in (46)), etc.
b. Properties of valency complementation, such as animacy vs. inanimacy,
compare the forms of Patient modifying the noun pochvala ‘praise’ in (70)
and (72).
c. Grammatical properties of the source verb, especially perfective vs. imper-
fective aspect, e.g. upozornění-PFV řidičům ‘warning addressed to drivers’/
*upozorňování-IPFV řidičům ‘warning addressed to drivers’ (Section 7).
d. Use of the noun within support verb constructions such that the valency of the
verbal component of the support verb construction can influence the valency
of its nominal component, compare pochválit někoho ‘praise somebody’ with
udělit pochvalu někomu ‘to give praise to somebody’ and the dative form of
Patient modifying the noun pochvala ‘praise’ in (72).
e. The tendency to avoid syntactic ambiguity, for example, the tendency to dif-
ferentiate Addressee in the genitive, modifying some nouns of communica-
tion, from Actor in the genitive; as a result the noun employs a special form
for the Addressee (Section 7).
f. Some other factors such as various types of analogy. For example, the dative
form of Addressee when it represents a special shift, as in combination with
some nouns of communication, could be used by analogy with the most typi-
cal and the most frequent form of Addressee in general, which is the dative
form.
9. Conclusion
We have given a complex survey of the typical as well as special valency behavior
of Czech deverbal nouns. We have focused on special shifts in the surface forms of
participants and their impact on the meaning of the noun. After considering the
syntactic and semantic properties of the nouns modified by a participant under-
going a special shift in its form, we delimited a group of nouns situated on the
Chapter 2. Special valency behavior of Czech deverbal nouns 
boundary between syntactic and lexical derivation which allow these shifts. We
have taken into account not only the possible combinations of particular forms of
complementation (i.e. combinations of the forms corresponding to special shifts
and the forms of other modification implying action reading) but also the function
of the noun in the language related to its meaning. We suppose that the nouns on
the boundary between syntactic and lexical derivation have a similar role as the
fully syntactic derivations because they can occasionally replace them.
Acknowledgments
The research reported in the chapter was supported by the Czech Science Founda-
tion under the projects P406/10/0875 and P406/12/P190.
References
Alexiadou, Artemis. 2001. Functional Structure in Nominals. Nominalization and Ergativity [Lin-
guistik Aktuell/Linguistics Today 42]. Amsterdam: John Benjamins. DOI: 10.1075/la.42
Apresjan, Jurij D. 1995. Leksičeskaja semantika. Sinonimičeskije sredstva jazyka (Lexical seman-
tics. Synonymic means of the language). Moscow: Vostočnaja literatura, RAN.
Cinková, Silvie & Kolářová, Veronika. 2006. Nouns as components of support verb construc-
tions in the Prague Dependency Treebank. In Insight into Slovak and Czech Corpus Linguis-
tics, Mária Šimková (ed.), 113–139. Bratislava: Veda.
Chomsky, Noam. 1972. Remarks on nominalization. In Studies on Semantics in Generative
Grammar, 11–61. The Hague: Mouton.
Cruse, David Alan. 1986. Lexical Semantics. Cambridge: CUP.
Český národní korpus (Czech National Corpus) – SYN2000. Ústav Českého národního korpusu
FF UK, Prague. 〈http://www.korpus.cz〉 (2000).
Český národní korpus – SYN2005. Ústav Českého národního korpusu FF UK, Prague. 〈http://
www.korpus.cz〉 (2005).
Český národní korpus – SYN2006PUB. Ústav Českého národního korpusu FF UK, Prague.
〈http://www.korpus.cz〉 (2006).
Český národní korpus – SYN2009PUB. Ústav Českého národního korpusu FF UK, Prague.
〈http://www.korpus.cz〉 (2009).
Český národní korpus – SYN2010. Ústav Českého národního korpusu FF UK, Prague. 〈http://
www.korpus.cz〉 (2010).
Dokulil, Miloš. 1982. K otázce slovnědruhových převodů a přechodů, zvl. transpozice (On
transfers and transitions among parts of speech: The case of transposition). Slovo a sloves-
nost 43: 257–271.
Dowty, David R. 1979. Word Meaning and Montague Grammar: The Semantics of Verbs and
Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel. DOI: 10.1007
/978-94-009-9473-7
 Veronika Kolářová
Hajič, Jan, Panevová, Jarmilla, Urešová, Zdeňka, Bémová, Alevtina, Kolářová, Veronika & Pajas,
Petr. 2003. PDT-Vallex: Creating a large-coverage valency lexicon for Treebank annotation.
In Proceedings of The Second Workshop on Treebanks and Linguistic Theories, 57–68. Vaxjo:
Vaxjo University Press.
Ježek, Elisabetta & Melloni, Chiara. 2011. Nominals, polysemy and co-predication. Journal of
Cognitive Science 12(1): 1–31.
Jirsová, Anna. 1966. Vazby u dějových podstatných jmen označujících duševní projevy (Valency
of non-productively derived nouns denoting mental states or dispositions). Naše řeč
49: 73–81.
Karlík, Petr. 2000. Valence substantiv v modifikované valenční teorii. (Valency of nouns in a
modified valency theory) In Čeština – Univerzália a specifika 4, Zdeňka Hladká & Petr
Karlík (eds), 181–192. Brno: Masarykova univerzita.
Karlík, Petr. 2002. Ještě jednou k českým deverbálním substantivům (Once more on Czech
deverbal nouns). In Čeština – Univerzália a specifika 4. Zdeňka Hladká & Petr Karlík (eds),
13–23. Prague: Lidové noviny.
Karlík, Petr & Nübler, Norbert. 1998. Poznámky k nominalizaci v češtině (Notes on nominaliza-
tion in Czech). Slovo a slovesnost 59: 105–112.
Kettnerová, Václava. 2009. Konstrukce s rozpadem tématu a dikta v češtině (Constructions with
the splitting of the theme and dictum in Czech). Slovo a slovesnost 70: 163–174.
Kolářová, Veronika. 2006. Valency of deverbal nouns in Czech. The Prague Bulletin of Math-
ematical Linguistics 86: 5–19.
Kolářová, Veronika. 2010. Valence deverbativních substantiv v češtině (na materiálu substantiv s
dativní valencí) (Valency of deverbal nouns in Czech, with a special regard to nouns with
dative valency). Prague: Karolinum.
Křížková, Helena. 1968. Substantiva s dějovým významem v ruštině a v češtině (Nouns with
action meaning in Russian and Czech). In Kapitoly ze srovnávací mluvnice ruské a české III,
O ruském slovese, 81–152. Prague: Academia.
Kuryłowicz, Jerzy. 1936. Dérivation lexicale et dérivation syntaxique. Bulletin de la Société lin-
guistique de Paris 37: 79–92.
Lopatková, Markéta, Žabokrtský, Zdeněk & Kettnerová, Václava. 2008. Valenční slovník českých
sloves (Valency lexicon of Czech verbs). Prague: Karolinum. 〈http://ufal.mff.cuni.cz/vallex/〉
Melloni, Chiara. 2011. Event and Result Nominals. A Morpho-semantic Approach. Bern: Peter
Lang.
Mikulová, Marie et al. 2006. Annotation on the Tectogrammatical Level in the Prague Dependency
Treebank. Annotation manual. Technical Report TR-2006–30. Prague: ÚFAL MFF UK.
Novotný, Jiří. 1980. Valence dějových substantiv v češtině (Valency of non-productively derived
nouns in Czech) [Sborník pedagogické fakulty v Ústí nad Labem]. Prague: SPN.
Panevová, Jarmila. 1980. Formy a funkce ve stavbě české věty (Forms and functions in the struc-
ture of Czech sentences). Prague: Academia.
Panevová, Jarmila. 2002. K valenci substantiv (s ohledem na jejich derivaci) (On noun valency
(with respect to their derivation)). Zbornik Matice srpske za slavistiku 61: 29–36.
Chapter 2. Special valency behavior of Czech deverbal nouns 
PDT-Vallex: Czech Valency lexicon linked to treebanks. 〈http://ufal.mff.cuni.cz/PDT-Vallex/〉

Piťha, Petr. 1984. Case frames of nouns. In Contributions to Functional Syntax, Semantics, and
Language Comprehension [Linguistic and Literary Studies in Eastern Europe 16], Petr Sgall
(ed.), 225–238. Amsterdam: John Benjamins.
Procházková, Věra. 2006. Argument Structure of Czech Event Nominals. MPhil thesis, Univer-
sity of Tromsø.
Sgall, Petr. 1967. Generativní popis jazyka a česká deklinace (A generative description of lan-
guage and the Czech declension). Prague: Academia.
Sgall, Petr, Hajičová, Eva & Panevová, Jarmila. 1986. The Meaning of the Sentence in its Semantic
Svozilová, Naďa, Prouzová, Hana & Jirsová, Anna. 2005. Slovník slovesných, substantivních a
adjektivních vazeb a spojení (Dictionary of complementation and phrases of verbs, nouns
and adjectives). Prague: Academia.
Šmilauer, Vladimír. 1966. Novočeská skladba (Syntax of modern Czech). Praha: SPN.
Urešová, Zdeňka. 2011a. Valenční slovník Pražského závislostního korpusu PDT-Vallex (Valency
lexicon of the Prague Dependency Treebank PDT-Vallex) [Studies in Computational and
Theoretical Linguistics]. Prague: Institute of Formal and Applied Linguistics. 〈http://ufal.
mff.cuni.cz/PDT-Vallex/〉
Urešová, Zdeňka. 2011b. Valence sloves v Pražském závislostním korpusu (Valency of verbs in the
Prague Dependency Treebank) [Studies in Computational and Theoretical Linguistics].
Prague: Institute of Formal and Applied Linguistics.
Veselovská, Ludmila. 2001. K analýze českých deverbálních substantiv (On the analysis of Czech
deverbal nouns). In Čeština – univerzália a specifika 3, Zdeňka Hladká & Petr Karlík (eds),
11–27. Brno: MU.
chapter 3
Nominalizations of Spanish perception verbs

at the syntax–semantics interface
Elisa Bekaert & Renata Enghels

Ghent University
Most studies on nominalization focus on nouns derived from action verbs and
pay little attention to other semantic types. This article aims to fill this gap by
studying a group of perception nominalizations in Spanish. It is well known that
the semantic distinctions that cross the field of perception (corresponding to the
perception modalities and agentivity of the perceiver) influence the syntax of the
corresponding verbs, and especially their complementing pattern. Yet, it has not
been studied to what extent these oppositions have an impact on the argument
structure of perception nominalizations. Moreover, as opposed to previous
analyses, which mainly focus on theoretical assumptions of the nominalization
process, this study is corpus-based and provides a quantitative as well as a
qualitative analysis.
1. Introduction
It has frequently been stated that deverbal nominalizations should be consid-

ered as reclassifications of their corresponding verbal clauses (Heyvaert 2003),
and thus that there is a systematic link between a nominalization and its argu-
ments, on the one hand, and between clauses and their verbal head, on the other
(Hopper & Thompson 1985; Givón 2001; Heyvaert 2003; Melloni 2007).1 The lit-
erature on nominalizations has generally focused on prototypical instances, i.e.
deverbal nominalizations derived from verbs with an agentive subject (e.g. con-
struction, translation, description). These nomina actionis have traditionally been
divided into two types, event and result nominals, which are said to influence the
. As defined by Fradin (2012: 129), a nominalization fulfills two criteria: (i) it is discursively
equivalent to a verbal predicate, and (ii) it presents the distributional and semantic properties
of a noun.
 Elisa Bekaert & Renata Enghels
orpho-syntactic and denotative properties of the nominal (Grimshaw 1990;

m
Alexiadou 2001; Melloni 2007). Consider the following examples in Spanish:2
(1) la traducción de una carta (por (parte de) Emilio)
the translation of a letter (by (part of) Emilio)
‘the translation of a letter (by Emilio)’
(2) una/la traducción (de Emilio) (de una carta)

a/ the translation (of Emilio) (of a letter)
‘Emilio’s translation of a letter’
The first type (1) is an event nominal and behaves like a passive nominal. It is
therefore traditionally said to carry an overt patient (de una carta), while the
agent, introduced by por (parte de) ‘by’, can be omitted. Event nominals also admit
the same aspectual modifiers as their verbal counterparts, indicating that they pre-
serve the aspectual properties of the base verb. This is illustrated in (3).3
(3) a. El bombardeo destruyó la ciudad en solo dos días.
‘The bombing destroyed the city in only two days.’
b. La destrucción total de la ciudad en solo
the destruction total of the city in only
dos días espanta a todos.
two days shock.prs.3sg to all
‘Everyone is appalled that the city has been completely destroyed in
only two days.’
The second type of nomina actionis, result nominals (2), refers to the outcome of
the event denoted by the base verb. As a result, they can omit both the agent and
the patient, and thus display a merely nominal character. They are said to appear
with the entire range of determiners (4a), can be used in the plural form (4b), and
they carry optional modifiers (4c-d):4
(4) a. la / una / aquella / alguna traducción de la Eneida
‘the / a / that / some translation of the Aeneid’
b. las traducciones de la Eneida
‘the translations of the Aeneid’
. Glosses will be provided only in those cases where the Spanish and English equivalents
present a different syntactic structure. When there is a direct word-to-word correspondence,
simple alignment is introduced.
. Examples are based on Grimshaw (1990: 58).
. Examples based on Picallo (1999: 377, 382).
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
c. la traducción de la Eneida fue publicada por Alianza

‘the translation of the Aeneid was published by Alianza’
d. La traducción fue publicada por Alianza.
‘The translation was published by Alianza.’
However, more recently, many authors have pointed out the need to refine this
classification.
In the first place, in recent literature, the narrow interpretation of the result
category has been put into question, as it is such a heterogeneous class (Osswald
2005; Melloni 2007, 2010; Scott 2010). In this paper, we will therefore follow the
classification provided by Melloni (2007: 11), who adopts the term referential
nominal. This category contains the following semantic subsets: product/result
(construction), means (connection), psychological stimulus (attraction), path (pro-
longation), agentive-collective (administration), and locative (entry).
Next, as has been mentioned above, most studies have focused on nominaliza-
tions derived from action verbs and have paid little attention to other semantic
types. This clearly reflects on the interpretation of the notion of event nominal.
According to Melloni (2007: 12), this category should equally comprise action
nominals and state nominals, since both types usually preserve the argument struc-
ture of the verbs they are derived from. So, more recently, state nominalizations
and nominalizations derived from verbs with an experiencer subject have been
analyzed in more detail (Meinschaefer 2003; Giammatteo, Albano & Ghio 2005;
Fábregas & Marín 2012; Fábregas, Marín & McNally 2012). Since the grammatical
properties of these two types have been proven to be different both from the cat-
egory of event nominals as from that of referential nominals, they deserve a proper
treatment (Huyghe & Marín 2007; Melloni 2007, Barque, Huyghe, Jugnet & Marín
2009; Fábregas & Marín 2012). The first characteristic of states is that, as opposed
to events, they denote non-dynamic situations, by which they cannot be the subject
of a predicate like tener lugar ‘take place’ (5a). In this property, they behave like
result nouns. Second, states have a temporal extension, just like event nominaliza-
tions. Therefore, both events and states can be modified by a time expression, such
as the adjective constante ‘constant’ (5b) (Fábregas & Marín 2012: 36).
(5) a. *La preocupación de Juan por la economía
the preoccupation of John by the economy
tuvo lugar el verano pasado.5
have.pst.3sg place the summer passed
*‘John’s preoccupation with the economy took place last summer.’
. Examples based on Fábregas & Marín (2012: 36)

 Elisa Bekaert & Renata Enghels
b. la preocupación constante de Juan por la economía

the preoccupation constant of John by the economy
‘John’s constant preoccupation with the economy’
Throughout, one particular category of deverbal nominalizations has remained

fairly neglected in the literature. While many authors have studied the syntactic
and semantic behavior of perception verbs (cf. Sweetser 1990; Ibarretxe-Antuñano
1999; Enghels 2007 among others), surprisingly their derived nominalizations
have not yet been analyzed in depth.6 Hence, the main objective of this article is
to examine the linguistic behavior of perception nominalizations (PN). The group
comprises the following cases that are derived from visual perception verbs: visión
‘vision’, vista ‘sight’, and mirada ‘look’. For auditory perception verbs the follow-
ing cases are relevant: audición ‘hearing, audition’, oído ‘ear, hearing’ and escucha
‘listening, eavesdropping’.7 More particularly, this paper aims to answer these two
research questions: (1) to what extent do PNs show the same semantic and syn-
tactic complexity as compared to their verbal bases; and (2) how do PNs fit in the
above mentioned classification of event vs. referential vs. state nominal, and to
what extent does this classification influence their syntactic behavior.
The outline of this paper is as follows. First, we will motivate our selection
of perception nouns, and pay special attention to the properties of their verbal
bases (Section 2). Next, we will analyze the polysemy of the PNs, emphasizing the
impact of the semantics of their suffixes (Section 3). It will be demonstrated that
the PNs can be defined as near-synonyms that, however, seem to have semanti-
cally specialized. Finally, in Section 4, we will examine the syntax of our selected
PNs, systematically establishing a correlation with their semantic properties on
the one hand, and the cognitive-semantic features of their corresponding verbs, on
the other hand. The main area that will be studied is the behavior of postnominal
prepositional phrases, including de-phrases and other obliques.
. The analysis of English perception nominals by Gisborne (1993) is the only exception that
we are aware of.
. Note that we did not include the nominalized infinitives el ver ‘the seeing’, el mirar ‘the
looking’, el oír ‘the hearing’, and el escuchar ‘the listening’, which require a separate study (De
Miguel 1996 ; Rodríguez Españeira 2004). A third nominalization derived from ver is vistazo
‘glance’, which has not been included in this study because it is morphologically derived from
vista, and not directly from ver. It should also be noted that oído etymologically derives from
oír (first attestation between 1220–50) (Corominas 1970), whereas audición (auditio, audi-
tionis) is a ‘cultismo’, namely a Latin word introduced in the 14th century. Vista and visión
come from Latin (videre) and were first attested in the 12th–13th centuries. Mirada was first
attested around 1495, and derives from the Spanish form mirar; escucha was first attested at
the end of the 13th century. All these etymological differences will not have any impact on
our study, though.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
2. Syntax and semantics of perception verbs
In order to better understand the nature of perception nominalizations, it is nec-

essary to first provide a general characterization of perception verbs (PVs) and
briefly describe their argument structure. Perception verbs are a class of verbs that
share a number of linguistic properties. Nevertheless, in most languages, their
semantic nature varies according to two parameters, namely the modality of per-
ception and the degree of agentivity of the perceiver.
As to the former property, the semantics of each verb changes according to the
modality of perception: visual, auditory, gustative, tactile, and olfactory. As to the
latter, dependent on the relationship between the subject perceiver and the stimulus
object, in Spanish one can distinguish between the involuntary PVs ver ‘to see’ and
oír ‘to hear’ and the voluntary PVs mirar ‘to look at’ and escuchar ‘to listen to’. Vol-
untary perception consists of an activity controlled by the perceiver, whereas invol-
untary perception is a mental process experienced by this participant (Kirsner &
Thompson 1976). Therefore, the subject of voluntary PVs has been said to be close
to a prototypical agent, whereas involuntary PVs would be performed by an expe-
riencer subject (Gisborne 1993; Taylor 1995). This semantic distinction has been
shown to have its impact on the syntax of the corresponding verbs.8
Syntactically, perception verbs are transitive verbs. However, the complex
nature of perception triggers a wide range of possible complement types, namely
an NP (6a) – with the subtype NP + relative clause (6b) –, NP + gerund (6c), an
infinitive complement (6d), and a that-clause (6e).
(6) a. Veo a Juan.
see.prs.1sg prep John
‘I see John.’
b. Veo a Juan que toca la guitarra.
see.prs.1sg prep John who play.prs.3sg the guitar
‘I see John who plays the guitar.’
c. Veo a Juan tocando la guitarra.
see.prs.1sg prep John play.prog the guitar
‘I see John playing the guitar.’
d. Veo a Juan tocar la guitarra.
see.prs.1sg prep John play.inf the guitar
‘I see John play the guitar.’
. In addition to the discussion on the thematic role of the subjects of perception verbs,
the nature of their objects has also been the subject of debate. These have been defined as the
product of the perception event or the cause of this event (see Enghels 2007: 62–66).
 Elisa Bekaert & Renata Enghels
e. Veo que Juan tiene razón.

see.prs.1sg that John have.prs.3sg reason
‘I see that John is right.’
As for involuntary PVs, they appear in a wide range of syntactic constructions, in

accordance with their semantic productivity. Moreover, as has been extensively
described in the literature (for instance by Miller & Lowrey 2003), the meaning
of PVs, mainly visual PVs, depends on the complement they are combined with.
Most complement types – in particular infinitives, gerunds and relative clauses –
coincide with a direct physical perception reading, whereas others – mainly subor-
dinated that-clauses – trigger an indirect perception or cognitive reading.9 Given
that voluntary perception is said to always be direct perception, mirar ‘to look at’
and escuchar ‘to listen to’ cannot be complemented by a that-clause. On the other
hand, since voluntary perception involves a process of directing attention toward a
stimulus, its corresponding verbs, and in particular the visual ones, are easily com-
bined with directional prepositional phrases introduced by a ‘to, at’, hacia ‘toward’,
para ‘to, at’, etc. (Hanegreefs 2006):
(7) Mira a–l / hacia el sol.
look.prs.3sg to-det toward the sun
‘He looks at the sun.’
The next section examines whether the different PNs reflect the semantic opposi-
tions that cross the field of perception. It also investigates to what extent visión,
vista, mirada, on the one hand, and audición, oído, escucha, on the other hand,
have suffered a process of semantic specialization, or whether they are to be
defined as near-synonyms.
3. Perception nominalizations: A semantically heterogeneous class
3.1 Corpus study: The data

As opposed to previous studies, which have mainly focused on theoretical
assumptions of the nominalization process (Grimshaw 1990; Alexiadou 2001;
Picallo 1999; Fábregas & Marín 2012), our approach is corpus-based and provides
a quantitative as well as a qualitative analysis. To begin with, Table 1 presents the
. During a process of direct perception, the perceiver maintains a physical relationship
with the external stimuli that directly provide him with information concerning his environ-
ment. During an act of indirect perception, however, the perceiver obtains these data through
deductive reasoning (Enghels 2009: 786).
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
token frequencies10 of each of the six above-mentioned forms in the Corpus de

Referencia del Español Actual (CREA) between 2003 and 2004.11
Table 1. Frequencies PNs 2003–2004 (CREA)12

absolute /100.000
vista 869 17,029779

visión 562 11,013505
mirada 326 6,3886168
oído 128 2,508414
audición 36 0,7054914
escucha 24 0,4703276
Note that the quantitative data confirm the superiority of visual perception
compared to other modalities of perception, as has been observed in regard to per-
ception verbs (Viberg 1984; Sweetser 1990 among others). In order to get a clear
view of the semantic and syntactic behavior of these PNs, 1200 examples (200 for
each PN) were retrieved from CREA, half of which were drawn from the litera-
ture and the other half from the press (1996–2004).13 PNs can occur in singular
(cf. (1)) or plural form. Plurality has been associated with a referential interpreta-
tion (see Grimshaw 1990; Picallo 1999; Alexiadou 2001 among others), as well as
with a bounded event reading (Picallo 1999; Bisetto & Melloni 2005; Knittel 2011).
As these uses deserve a proper treatment, they will not be taken into account in the
remainder of this study.
A first step toward a better comprehension of the linguistic functioning of
perception nominalizations in Spanish consists of a thorough analysis of their
. It thus counts both singular and plural forms, in the different text types.
. Real Academia Española: Online database CREA: Corpus de referencia del español actual
〈http://www.rae.es〉.
. The CREA corpus (2003–2004) counts 5.102.826 words. In order to get a clear view of
the occurrence of each PN, the relative frequency is calculated on a total of 100.000 words.
. Idiomatic expressions had to be eliminated from the corpus since they could distort our
results. According to definitions proposed by the Real Academia Española (2009: 53–54) and
Piera & Varela (1999: 4403), the following types were distinguished: (a) nominal idioms: punto
de vista ‘point of view’, umbral de audición ‘threshold of hearing’, infección de oído ‘ear infec-
tion’, etc. (b) verbal idioms: perder de vista ‘to lose sight of ’, hacer oídos sordos ‘to turn a deaf
ear’, etc. (c) prepositional idioms: a la vista de ‘in view of ’, con vista(s) a ‘with a view to’, etc.;
and (d) adverbial idioms: a vista de pájaro ‘from a bird’s eye view’, a la escucha ‘listening’, etc.
 Elisa Bekaert & Renata Enghels
polysemy. However, the study of deverbal nominalizations must also include an

analysis of the suffixes added to the verbal bases (Díaz Hormiga 2005; Amador
Rodríguez 2009). As will be shown in the next section, the different suffixes -ión,
-do/-da/-ta and zero affixation add their own semantic and aspectual values when
involved in the process of nominalization of perception verbs.
3.2 The semantics of suffixation

Following the principle of iconicity, an extra formal marking implies an additional
semantic value (Haiman 1980). We therefore assume that all suffixes add a certain
semantic content to their base, albeit a weak one (Melloni 2007: 74).
The suffix -ión, which appears in visión and audición, is generally defined as
a transpositional suffix, which principally acts as a category changer (Bisetto &
Melloni 2005; Fábregas 2010). Fábregas & Marín (2012) for instance, believe that
this suffix does not contain any aspectual information in itself and hence does not
change the aspectual properties of the verbal base. However, the suffix -ión does
present some specific features. It can convey concrete meanings to the nominal,
often through metonymical transpositions (e.g. construcción ‘construction’, tra-
ducción ‘translation’) (Pharies 2002; Melloni 2007; Amador Rodríguez 2009).14 It
thus constitutes a very flexible suffix which can add polysemy to its corresponding
nominals.
The suffixes -do/-da and their allomorphs, which occur in mirada, oído and
vista, are not mere transpositional nominalizers (like -ión or -miento), as they
derive from the past participial form of the corresponding verb and therefore
display Aktionsart properties (Melloni 2007: 108).15 Their semantics is, in fact,
strongly related to the perfective value of the past participle (Amador Rodríguez
2009), by which these suffixes are said to trigger the referential reading. Further-
more, several authors assume that the suffix -(a)da can require a path interpreta-
tion and is therefore agglutinated to verbs of inherent direction to form both event
and referential nominals (e.g. llegada ‘arrival’, salida ‘exit’) (Melloni 2007; Fábregas
2010; Fábregas, Marín & McNally 2012).
Finally, zero suffixation or conversion implies that the derivation is not real-
ized through the addition of a suffix, but through a categorical change of the base
(e.g. escuchar > la escucha) (Amador Rodríguez 2009). As opposed to Fábregas
and Marín (2012) we include zero suffixation within the possibilities for deriving
. For a more detailed definition of the notion of ‘metonymical transposition’, cf. infra
Note 19.
. The allomorphs arise as a consequence of the existence of irregular participles in Spanish
(e.g. dicho, escrito, respuesta, vista).
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
deverbal nominalizations, but, consistent with the iconicity principle, we presume

that it does not add any additional information to the verbal base. In other terms,
it merely acts as a category-changing process.
3.3 Polysemy of perception nominalizations

A detailed analysis of the corpus shows that the PNs under scrutiny constitute
a heterogeneous class. More particularly, all six PNs are semantically ambiguous
and most of them, except for oído and escucha, can denote both an event and a
referential object, as will be illustrated below. However, in accordance with the
semantics of their verbal bases, only visual PNs can refer to a perception event
and to a cognitive process. Table 2 displays the frequency of all three categories
for each PN.16
Table 2. Semantics of PNs17

visión vista mirada audición oído escucha
#17 % # % # % # % # % # %
Event (perception) 25 14,8 3 9,7 – – 95 69,8 – – 69 100

Event (cognition) 113 66,9 – – 55 35,1 – – – – – –
Referential 31 18,3 28 90,3 102 64,9 41 30,1 95 100 – –
Total 169 100 31 100 157 100 136 100 95 100 69 100
To begin with, within the domain of visual perception, visión seems to be the
nominal that displays the richest polysemy, in accordance with the semantics of
its base and its suffix. It can indicate a process of direct physical visual perception,
whether it be real (8a) or not (8b) – this example concerns St. Francis’ vision – but
it can also receive a more referential interpretation when it denotes the faculty of
perception (8c) or a general overview (8d). However, it is most frequently used to
refer to a cognitive process (8e), which recalls the polysemy of its verbal base (e.g.
Sweetser 1990).18
. Note that the total number of instances is inferior to 200, because of some semantically
ambiguous cases and the large number of idiomatic expressions.
. I.e. absolute frequency.
. Only the relevant portion of the examples is provided; the references allow the reader to
retrieve the larger context.
 Elisa Bekaert & Renata Enghels
(8) a. ante la visión de un niño mutilado

‘in front of the sight of a child mutilated
por las bombas
by the bombs’ (CREA: La Razón, 09/04/2003)
b. melodías seguidas por la visión de un ángel
‘melodies followed by the vision of an angel’
(CREA: Press, 2003)
c. dos causas importantes de pérdida de visión
two causes important of loss of sight
‘two important causes of loss of eyesight’ (CREA: El País, 13/05/2003)
d. ofrecer una amplia visión retrospectiva de
offer.inf a wide vision retrospective of
lo mejor y más vanguardista de
the best and most avant-garde of
la cinematografía española de los años sesenta
the cinematography Spanish of the years sixty
‘to offer a broad retrospective overview of the best and the most
avant-garde of the Spanish cinematography in the sixties’
(CREA: El Cultural, 21/11/2003)
e. su visión conservadora de la religión
its vision conservative of the religion
‘its conservative view on religion’ (CREA: El Mundo, 10/11/2004)
The PN vista can also receive an event interpretation – albeit mainly when it is
used in the context of a lawsuit (9a). Yet, in the overwhelming majority of the
cases, it refers to a referential object. This referential reading includes the interpre-
tation of the faculty of perception (9b), or a panorama (9c).
(9) a. un segundo aplazamiento de la vista de su apelación
a second delay of the sight of her appeal
‘a second delay of (the treatment of) her appeal’
(CREA: El País, 04/06/2003)
b. La vista de-l pájaro cazador, cuando más
the sight of-det bird hunter, when more
alto subía, se deslumbra, se
high rise.pst.3sg refl dazzle.prs.3sg refl
ciega, por el exceso de luz.
blind.prs.3sg by the excess of light
‘The eyesight of the bird of prey, when rising, was impaired by the
excess of light.’ (CREA: Press, 2003)
c. esa vista desde Plaza Nueva de-l
‘that view from Plaza Nueva of-det
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
Sacromonte vestido de blanco

Sacromonte decorated in white’ (CREA: Ideal Digital, 11/01/2003)
Finally, mirada has specialized in the referential meaning when referring to the
look or gaze as instrument of perception (10a). However, as opposed to the seman-
tics of the verb it derives from, mirada is also frequently used to indicate a cogni-
tive process (10b):
(10) a. con el gesto cordial y la viva mirada penetrante
with the expression friendly and the lively look penetrating
‘with the cordial expression and the intense penetrating look’
(CREA: La Razón Digital, 16/12/2003)
b. Menos abundante, pero no menos
‘Less abundant, yet not less
instructiva, fue su mirada sobre el arte.
instructive, was his view on det art.’
(CREA: El Cultural, 02/01/2003)
The auditory PNs seem to be less polysemous and can more easily be situated
within one particular semantic category. This is certainly the case for oído, which
in all its occurrences is referential and either refers to the faculty of perception
(11a) or, through a process of metonymical transposition, to the instruments of
auditory perception, namely the ears (11b).19 Escucha on the other hand, always
denotes an event, and more particularly it can either refer to the process of listen-
ing (12a) or a true act of eavesdropping (12b):
(11) a. sobre todo gracias a–l oído y a la vision
above all thanks to-det hearing and to the visión
‘especially thanks to his senses of hearing and sight’
(CREA: El País, 30/01/2003)
b. susurrándo-le melodías tanto en su oído
whisper.prog-ben melodies both in her ear
como en su vientre
and in her belly
‘whispering melodies both in her ear and her belly’
(CREA: Revista Natural, 03/2003)
(12) a. una escucha a lo que puede ser pensado
a listening to what can be thought
. Within the category of referential nominals, a particular subset has been distinguished,
namely when the interpretation of the nominal refers to a concrete object due to a process that
has been defined by Bisetto and Melloni (2005: 400) as ‘metonymic transposition of meaning’:
La traducción está en la mesa. ‘The translation is on the table’.
 Elisa Bekaert & Renata Enghels
o sentido desde el interior

or felt from the interior
‘a listening to what can be thought or felt from the interior’
(CREA: Press, 2004)20
b. una probable y no suficientemente probada escucha
a probable and not sufficiently proven listening
a la novieta de turno de su hijo
at the girlfriend on duty of his son
‘a probable and not sufficiently proven act of eavesdropping at the
present girlfriend of his son’ (CREA: El Mundo, 15/10/1996)
As was the case in the domain of visual perception, the nominal derived by -ión is
the most polysemous one. Audición denotes a process of direct physical auditory
perception, whether it be real (13a) or divine (13b), but has also become semanti-
cally specialized to denote a concert (13c), an audition (13d) or an interrogation
(13e), which are also events. On the contrary, its limited referential uses denote the
faculty of perception (13f):
(13) a. caen en éxtasis a–l hacer la
fall.prs.3pl into ecstasy prep-det do.inf the
oración mental o durante la audición de-l canto
prayer mental or during the hearing of-det hymn
‘become ecstatic while saying a mental prayer or while hearing
the hymn’ (CREA: Press, 2003)
b. les provoca alucinaciones y audición
ben cause.prs.3sg hallucinations and hearing
de voces extrañas
of voices weird
‘causes hallucinations and hearing of weird voices’ (CREA: Press, 1997)
c. ofrecerán una audición en el colegio
offer.ft.3pl an audition in the college
‘will give a concert at the college’ (CREA: El Diario Vasco, 13/03/2001)
d. para acceder a esta escuela era necesario
to access at that school be.pst.3sg necessary
. Note that the PN is introduced by an indefinite article. In the literature (Grimshaw 1990),
this property has been defined as an indication of the referential nature of the nominaliza-
tion. However, we believe that – just as was the case for plural formation – in this context the
indefinite article introduces a bounded event reading.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
superar una dura audición

endure.inf a hard audition
‘in order to be admitted to this school, a rigorous audition needed
to be endured’ (CREA: Val, C. del, Nacho Duato. Por vos muero, 1998)
e. Udre había sido criticada en su
‘Udre had been criticized during her
audición ante una comisión de
interrogation in front of a commission of
la Eurocámara.
the European Parliament.’
(CREA: ABC, 03/11/2004)
f. en cuanto a las personas con dificultad
in regard to the persons with difficulty
de visión, habla y audición
of vision, speech and audition
‘with regard to persons with sight, speech and hearing difficulties’
(CREA: Press, 2004)
3.4 First conclusions

From this survey it becomes clear that, within each field defined by the modality
of perception, the PNs appear to be semantically specialized. Moreover, both the
semantics of the base verb and of the suffix have an impact on the polysemy of the
perception noun. The profile of each PN can be summarized as follows:
visión: [visual, involuntary, – aspectual suffix]
vista: [visual, involuntary, + perfective suffix]
mirada: [visual, voluntary, + perfective suffix]
audición: [auditory, involuntary, – aspectual suffix]
oído: [auditory, involuntary, + perfective suffix]
escucha: [auditory, voluntary, zero affixation]
First, the perfective suffixes -ta/-da/-do often – but not exclusively – trigger a ref-
erential reading. This is certainly the case with oído (‘ear’) and vista (‘gaze’), and
to a lesser extent also with mirada. As has already been stated, the suffix -(a)da
particularly appears with verbs that require a path interpretation to form both
event and referential nominals. Indeed, the semantics of mirar has been defined as
‘directing his view toward’, emphasizing the aspect of movement and orientation
(Hanegreefs 2006). As has been claimed in the literature, -ión seems to be the most
flexible suffix that allows the PN to refer to different aspects of the base verb: next
to a referential interpretation, it can indicate a direct perception event, as well as
 Elisa Bekaert & Renata Enghels
a cognitive process (with visual perception). Indeed, the polysemy of visual PNs
is more complex than that of auditory PNs. More particularly, only visual PNs,
in accordance with the semantics of their base verbs, allow the extension toward
the domain of cognition and require a distinction to be made between event uses
referring to physical perception on the one hand, and cognitive processes on the
other. Audición also refers to an act of direct auditory perception, besides its ref-
erential use. Finally, escucha is the only PN that lacks a referential use and always
receives a (perception) event interpretation.
Consequently, it could be said that, as to the semantics of the PNs, the oppo-
sition between voluntary vs. involuntary perception seems to be subordinated
to the dichotomy between visual and auditory perception. Visual PNs most fre-
quently extend to cognitive event readings, apart from referential uses, whereas
auditory PNs – except for oído, which has gone through a lexicalization p rocess –
most frequently receive an event reading and extend toward the domain of com-
munication. This opposition recalls a previously stated opposition between both
modalities of perception. Indeed, apart from the fact that the two main percep-
tion modalities share a number of characteristics (Sweetser 1990: 38–39), visual
and auditory perception are two very different ways of acquiring information
about the external world (Enghels 2009). In the first place, for visual percep-
tion the experiencer subject has more control over the perception process than
that of auditory perception, which is less agentive. Secondly, the visual percep-
tion of an entity merely follows from its presence in the field of view, whereas
auditory perception is a necessary consequence of the effect of a stimulus being
present. Finally, and most importantly, what is seen is always located in space
and thus displays an objectival character. However, we do not hear an object
directly but merely the sounds it produces. Sound is therefore located in time,
not space, which explains the predominantly eventive nature of auditory percep-
tion (Miller & Johnson-Laird 1976).
The next sections will examine how the different PNs syntactically realize the
semantic participants of the events they denote and to what extent they have a
surface syntax comparable to the verbs they are derived from.
4. Realization of the argument structure of perception nominalizations
4.1 Introduction
In our corpus, PNs have four types of syntactic constructions at their disposal to
realize the perceiver (i.e. First Argument, FA) and the stimulus of perception (i.e.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
Second Argument, SA):21 (1) prepositional phrases headed by de ‘of ’ (14a), (2)
prepositional phrases headed by prepositions other than de, such as por (parte de)
‘by’, a ‘to’, hacia ‘toward’, sobre ‘on, at’, etc. (14b), (3) possessive determiners such as
su ‘his/her’ (14c) and (4) relational adjectives (Picallo 1999) (14d).22
(14) a. Dicha visión de-l personaje iba
that view of-det character go.pst.3sg
destinada a un “lector pío”.
addressed to a reader pious
‘That view on the character was meant for a “pious reader.”’
(CREA: Press, 2003)
b. una simple mirada a una página de-l “Zohar” en hebreo
a simple look at a page of-det Zohar in Hebrew
‘a simple look at a page of the “Zohar” in Hebrew’
(CREA: La Razón, 01/12/2004)
c. Días después sería su primera audición in Bilbao.
days after be.cond.3sg his first audition en Bilbao
‘Several days later, he would give his first concert in Bilbao.’
(CREA: ABC Cultural, 06/09/1996)
d. Lo achacaron a la histórica mala vista
do attribute.pst.3pl to the historical bad sight
de los germanos comparada con la siempre
of the Germans compared with the ever
excelente vista gala.
excellent sight Gallic
‘[They] attributed it to the historical bad eyesight of the Germans
compared with the ever excellent Gallic eyesight.’
(CREA: Sabadell, M.A., El hombre que calumnió a los monos, 2003)
. We follow the terminology of Gisborne (1993) among others. The objective of this article
is not to determine what the syntactic status of these complements is, as stated by M
einschaefer
(2003: 234): “A crucial difference between verbs and derived nominalizations is that for verbs,
syntactic realization of semantic participants is obligatory, while for nominalizations (as well
as for non-derived nominals) it is optional. Therefore, it is controversial whether the syntactic
complements of nominalizations corresponding to arguments of the base verbs should be
termed arguments too, or whether they should rather be classified as adjuncts (…).” Hence, in
order to refer to the syntactic realizations of semantic participants, we will alternately use the
terms ‘complement’ and ‘argument’.
. A relational adjective refers to a set of features. For instance, marítimo ‘maritime’ denotes
the entire range of properties that define the substantive mar ‘sea’ (Demonte 1999: 137).
 Elisa Bekaert & Renata Enghels
As is shown in Table 3, the six PNs under scrutiny seem to realize their argument
structure quite differently.23
Table 3. Syntactic realization of PN-arguments24

# % # % # % # % # % # %
[+FA] [–SA] 23 12,9 20 69 88 75,2 11 25 25 100 2 8,3

[+FA (de)] [–SA] 6 [26,1] 6 [30] 45 [51,1] 4 [36,4] 11 [44] – –
[+FA (other prep)] – – – – – – – – – – 1 [50]
[–SA]
[+FA (poss)] [–SA] 16 [69,6] 13 [65] 42 [47,7] 6 [54,5] 12 [48] 1 [50]
[+FA (rel adj)] [–SA] 1 [4,3] 1 [5] 1 [1,1] 1 [9,1] 2 [8] – –
[–FA] [+SA] 114 64 9 31 22 18,8 33 75 – – 22 91,7
[–FA] [+SA (de)] 109 [95,6] 9 [100] 1 [4,5] 27 [81,8] – – 19 [86,4]
[–FA] [+SA 5 [4,4] – – 21 [95,5] – – – – 3 [13,6]
(other prep)]
[–FA] [+SA (poss)] – – – – – – 3 [9,1] – – – –
[–FA] [+SA (rel adj)] – – – – – – 3 [9,1] – – – –
[+FA] [+SA] 41 23 – – 7 6 – – – – – –
[+FA (de)] [+SA (de)] 1 [2,4] – – – – – – – – – –
[+FA (de)] 4 [9,8] – – 3 [42,9] – – – – – –
[+ SA (other prep)]
[+FA (poss)] 31 [75,6] – – – – – – – – – –
[+SA (de)]
[+FA (poss)] 3 [7,3] 4 [57,1]
[+SA (other prep)]
[+FA (rel adj)] 2 [4,9] – – – – – – – – – –
[+SA (de)]
Total 178 100 29 100 117 100 44 100 25 100 24 100
. Note that the examples without any trace of argument realization have not been included
in the table, which explains why the total number of examples differs from the one presented
in Table 2.
. In this table, the grey rows indicate the argument combinations, namely [+FA][-SA],
[-FA][+SA] or [+FA][+SA], whereas the white zones show the frequency of the several syn-
tactic constructions within these types. The composition of Table 5 will be similar.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
First, referential nominals such as oído (15a) and vista (15b) often realize their
FA through a possessive. This possessive replaces the FA and SA which otherwise
could be introduced by the preposition de ‘of ’:
(15) a. Para vivir así, susurró él en su
to live.inf like this whisper.pst.3sg he in her
oído, más vale no morirse.
ear, more be_worth.prs.3sg not die.inf
‘To live like this, he whispered in her ear, it is best not to die.’
(CREA: Pérez-Reverte, A., La Reina del Sur, 2002)
b. Mi mala vista me jugó más de una mala pasada.
my bad sight ben play.pst.3sg more than one bad trick
‘My bad eyesight has played more than one trick on me.’
(CREA: El País, 22/03/2003)
Secondly, referential nominals do not allow for their SA being introduced by a

relational adjective (Picallo 1999: 385), which explains their absence with oído,
vista and mirada.
In what follows, only the distribution of prepositional phrases will be consid-
ered. Given that our aim is to describe the syntactic realization of the participants
by a complement, examples where none of the participants were expressed have
also been excluded (16):25
(16) versos divinos cantados por un ente sobrenatural
verses divine sing.ptcp by a being supernatural
en medio de una visión
in middle of a vision
‘divine verses sung by a supernatural being in a vision’ (CREA: Press, 2003)
Indeed, as stated by Heyvaert (2008: 248–249), “there is no automatic transfer

from clause to NP as far as obligatory participants are concerned”, which also
seems to be the case for PNs given the high frequency of zero realization in the
nominal domain.
4.2 Results
Table 4 shows the results of classifying our examples according to the presence of
one or both complements, when expressed by a prepositional phrase.
. These methodological restrictions reduce our corpus considerably and hence leaves us
with a sample of 275 relevant occurrences of the abovementioned PNs.
 Elisa Bekaert & Renata Enghels
Table 4. Frequency of prepositional PN-arguments

# % # % # % # % # % # %
[+ FA] [– SA] 6 4,8 6 40 45 63,8 4 12,9 11 100 1 4,3

[–FA] [+ SA] 114 91,1 9 60 22 31,9 27 87,1 – – 22 95,7
[+ FA] [+ SA] 5 4,1 – – 3 4,3 – – – – – –
Total 125 100 15 100 70 100 31 100 11 100 23 100
As opposed to what could be observed for the PVs, it immediately becomes

clear that the expression of both arguments with PNs is very rare. Only eight cases
were observed, all of which were situated within the domain of visual perception
(cf. infra Eg. (23)). The data also point toward a dichotomy between, on the one
hand, nominalizations that clearly tend to syntactically realize the argument refer-
ring to the perceiver (17a), namely mirada, oído and to some extent also vista, and,
on the other hand, PNs that combine with arguments referring to the stimulus of
perception (17b), namely visión, audición, and escucha.
(17) a. bajo la mirada de-l internauta
under the look of-det netter
‘in front of the view of the netter’
(CREA: El País, 30/01/2003)
b. La maternidad te cambia la visión
the motherhood ben change.prs.3sg the view
de las cosas y el estándar de valores.
of the things and the standard of values
‘Motherhood changes your view on things and the standard of values.’
(CREA: El País, 14/06/2004)
This varied behavior of the PNs in the expression of their complements raises
a number of interesting questions. First, as opposed to what has been observed
for PVs, the opposition between voluntary and involuntary perception does not
seem to be as important for the syntactic behavior of PNs; mirada and escucha for
instance, are both voluntary PNs but display a very different syntactic behavior.
Hence, the question arises which of the PNs’ semantic properties actually have an
impact on their argument realization.
The second interesting issue that follows from the differences in how PNs
behave involves the distinction between first argument and second argument real-
ization. This opposition recalls the difference between referential nominals, on
the one hand, and event and state nominals, on the other. Indeed, it is generally
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
accepted that events (the translation of a letter by Emilio) and states (John’s pre-
occupation with the economy) are naturally in opposition to referential nominals
(a construction, that translation), as they usually preserve the argument structure
of the base verb (Melloni 2007; Fábregas & Marín 2012). Indeed, traditionally,
event nominals are said to carry an obligatory SA (Grimshaw 1990; Picallo 1999;
Alexiadou 2001). This claim, albeit an uncompromising one, assumes that there
is a systematic link between the presence of the SA and the event reading.26 On
the contrary, referential nominals are often said to absorb the internal argument
of the corresponding verb, by which they only allow the expression of the FA. This
FA is usually interpreted as a possessor (Alexiadou 2001; Bisetto & Melloni 2005;
Gutiérrez Ordóñez 2005; Amador Rodríguez 2009). Although many authors, as
well as our own results, have pointed to counter-examples, referential nominals
are thus generally assumed to prefer the expression of the FA.27 In other words,
these data confirm that visión, audición and escucha mainly receive an event read-
ing, while mirada, oído and, to some extent vista, which presents an intermediate
status, favor the expression of the FA and thus syntactically behave like referential
nominals. However, in what follows, we will demonstrate that within the category
of event readings, perception events and cognitive events also present some salient
differences, which put into question the classification of the cognitive event type.
To this end Table 5 provides a more detailed account of the prepositions that
introduce the postnominal complements of PNs.
Table 5. Syntactic realization of prepositional PN-arguments28

# % # % # % # % # % # %
[+FA] [– SA] 6 4,8 6 40 45 64,3 4 12,9 11 100 1 4,3

[+FA (de)] [–SA] 6 [100] 6 [100] 45 [100] 4 [100] 11 [100] – –
[+FA (por (parte – – – – – – – – – – 1 [100]
de))] [–SA]
[–FA] [+ SA] 114 91,2 9 60 22 31,4 27 87,1 – – 22 95,7
[–FA] [+ SA (de)] 109 [95,6] 9 [100] 1 [4,5] 27 [100] – – 19 [86,4]
(Continued)
. The statement has been challenged by emphasizing the influence of pragmatic and dis-
course factors on argument realization (Heyvaert 2008 ; Melloni 2007 ; Sleeman & Brito 2010).
. A clear counter-example is the following sentence: The translation of this text is full of
mistakes. (Bisetto & Melloni 2005: 400)
. Cf. Table 3 (Note 24).
 Elisa Bekaert & Renata Enghels
Table 5. (Continued)

# % # % # % # % # % # %
[–FA] [+ SA 5 [4,4] – – 21 [95,5] – – – – 3 [13,6]
(other prep)]
[+FA] [+ SA]* 5 4 – – 3 4,3 – – – – – –
[+FA (de)] 1** [20] – – – – – – – – – –
[+ SA (de)]
[+FA (de)] [+ SA 4 [80] – – 3 [100] – – – – – –
(other prep)]
Total 125 100 15 100 70 100 31 100 11 100 23 100
*We did not include [+FA (por)] as it was not attested in our data.
**This example constitutes an exceptional case where both arguments are coordinated through y ‘and’:
[…] la visión particularista de los Estados Unidos y de lo que podía ofrecer a un mundo que ya se
estremecía ante los albores de la democracia. ‘[…] the particularistic vision of the United States and of
what it could offer to a world that was already staggering at the beginning of democracy.’ (CREA: ABC
Cultural, 29/11/2003)
When we consider the syntactic realization of the FA, we observe that it is

expressed almost exclusively by de-phrases ‘of ’. The preposition por (parte de) ‘by’
typically introduces agents and is therefore related to an event reading, so that
both state and referential nominalizations reject the use of this preposition. More
particularly, since only pure agents can be introduced by por (parte de) ‘by’, state
nominals merely allow the expression of a non-volitional causer of the state by de
‘of ’ (el aburrimiento de Juan) (Melloni 2007; Fábregas & Marín 2012). Moreover,
referential nominals combine with genitive possessors, which are expressed by a
possessive determiner or the preposition de ‘of ’ (Picallo 1999; Amador Rodríguez
2009). It is thus significant that in our sample the only occurrence of por (parte de)
‘by’ was attested by escucha, which always denotes a volitional perception event
(cf. Table 2):
(18) para impedir cualquier escucha por parte de
to avoid any listening by part of
personas no autorizadas
persons non-authorized
‘to avoid any eavesdropping by non-authorized persons’
(CREA: García de Enterría, E.; Tizzano, A.; Alonso García, R.,
Código de la Unión Europea, 1996)
The expression of the SA offers a more complex pattern. In our corpus, the SA is
generally expressed through a de-phrase ‘of ’, except for mirada, which can be used
with a wide range of prepositions, such as a ‘to, at’, hacia ‘toward’, sobre ‘on, at’, etc.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
This syntactic variation can be explained by two factors. First, as has been stated
above (cf. Section 2), voluntary perception (mirada, escucha), and in particular
visual perception (mirada), involves a process of directing attention toward a stim-
ulus. Indeed, in our corpus, mirada (59,1%) (19), and to a minor extent escucha
(9,1%) (12a;b), allow for the locative directional prepositions a ‘to, at’ and hacia
‘toward’ to introduce the SA, just as their corresponding PVs.
(19) a. la mirada a los valores esenciales
the look at the values essential
‘the look at the essential values’ (CREA: Press, 2003)
b. la mirada hacia el usuario único
the look toward the user single
‘the look at the particular user’ (CREA: Royo, J., Diseño digital, 2004)’
The second factor to explain the syntactic variation of the SA involves the use of
certain prepositions and their relationship with the perception vs. cognitive event
reading. As has already been shown, PVs differ from PNs in that the opposition
between voluntary and involuntary perception is crucial for the former, while the
modality of perception seems to be more determinative within the field of the lat-
ter. This is confirmed by the syntactic analysis of our corpus in which the use of
de ‘of ’ coincides with both a cognitive event (20a) and a perception event reading
(20b). However, other prepositions and in particular sobre ‘on, at’, clearly trigger
a cognitive event reading only of visual perception, in both voluntary (21a) and
involuntary (21b) PNs.
(20) a. la visión pragmática de la relación educación-trabajo
the view pragmatic of the relationship education work
‘the pragmatic view of the relationship between education and work’
(CREA: Comunidad Escolar, 03/12/2003)
b. ha posibilitado la visión de la respuesta en pantalla
make_possible.pst.3sg the view of the answer on screen
‘made possible to view the answer on the screen’
(CREA: Rodríguez Calafat, D., Informática avanzada
al alcance de todos, 2004)
(21) a. desarrollar una visión crítica sobre los
develop a view critical on the
acontecimientos cotidianos
events daily
‘to develop a critical view on the daily events’
(CREA: Arqueoweb, 01/05/2004)
b. Este film constituye la insólita mirada
this movie constitutes the unusual look
 Elisa Bekaert & Renata Enghels
sobre la represión emocional y sexual

on the repression emotional and sexual
de un grupo de monjas en un monasterio tibetano.
of a group of nuns in a convent Tibetan
‘This movie provides an unusual view on the emotional and sexual
repression of a group of nuns in a Tibetan convent.’
(CREA: Fotogramas, 11/2003)
What is more, the presence of an abstract SA must be considered as a clear indica-

tion of the cognitive event reading of the PN. Table 6 demonstrates that [+SA (de)]
allows for a cognitive reading, as these SAs are frequently abstract (20a). More-
over, with visión, which clearly prefers this cognitive reading (66,9%) (cf. Table 2),
the use of a preposition other than de, such as hacia ‘toward’ (22a) or respecto a
‘about’ (22b), is restricted to abstract SAs:
(22) a. una visión directa, humanísima y llena de
a view direct, most human and full of
comprensión hacia el oficio más viejo de-l mundo
comprehension toward the profession most old of-det world
‘a direct, most humane view full of comprehension toward the world’s
oldest profession’
(CREA: La Razón, 09/04/2003)
b. la nueva visión respecto a–l uso de
the new view respect to-det use of
la tecnología
the technology
‘the new view on the use of technology’
(CREA: Geofocus, 2003)
Table 6. Semantic nature of the SA29

visión mirada
# % # %
[–FA] [+ SA] 114 95,8 22 88

[–FA] [+ SA (de)] 109 [95,6] 1 [4,5]
SA [+ conc] 44 [40,4] 1 [100]
(Continued)
. Just like Tables 3 and 5, Table 6 contains subdivisions. Whereas the dark grey rows show
the argument combinations, the light grey rows indicate the frequency of the different prepo-
sitions within these combinations and the white zones give an idea of the concrete or abstract
character of the SA.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
Table 6. (Continued)
visión mirada
# % # %
SA [+ abstr] 65 [59,6] – –
[–FA] [+ SA (other prep)] 5 [4,4] 21 [95,5]
SA [+ conc] – – 11 [52,4]
SA [+abstr] 5 [100] 10 [47,6]
[+FA] [+ SA] 5 4,2 3 12
[+FA (de)] [+ SA (de)] 1 [20] – –
SA [+conc] – – – –
SA [+abstr] 1 [100] – –
[+FA (de)] [+ SA (other 4 [80] 3 [100]
prep)]
SA [+conc] – – – –
SA [+abstr] 4 [100] 3 [100]
Total 119 100 25 100
To summarize, two factors elicit the use of a preposition other than de ‘of ’ to
introduce the SA: (a) the directionality of voluntary perception in general, and
(b) the cognitive event reading of visual perception. As both features converge in
mirada, this PN allows for the widest range of different prepositions.
Finally, in our corpus, the syntactic realization of both arguments is limited
to visual PNs, visión and mirada, and in all instances provokes a cognitive event
reading (23), which is emphasized by the absence of concrete SAs (cf. Table 6).
(23) a. la “visión
positiva” de Mohamed VI respecto
the view positive of Mohamed VI respect
a “los múltiples desafíos que perturban
to the numerous challenges that disturb.prs.3sg
la buena marcha de la integración africana”
the good course of the integration African
‘Mohamed VI’s “positive view” on “the various challenges that interfere
with a fluent African integration”’ (CREA: El Mundo, 10/11/2004)
b. 24 Hour Party People, la mirada más canalla
24 Hour Party People, the look most nasty
y documentalista de Michael Winterbottom
and documentary of Michael Winterbottom
sobre la escena musical de Manchester
 Elisa Bekaert & Renata Enghels
on the scene musical of Manchester

durante las décadas de los 70 y 80.
during the decades of the 70 and 80
‘24-Hour Party People, Michael Winterbottom’s nastiest and most
documentary look on the Manchester music scene during the seventies
and the eighties.’ (CREA: El Cultural, 02/01/2003)
5. Conclusions
To sum up, a number of interesting conclusions can be drawn from exploring the
similarities and differences between perception verbs and their related nominals.
Just like the verbs they are derived from, the different PNs cannot form a single
unified class since they vary considerably in their semantic and syntactic behavior.
In the first part of this study, it was shown that the different subsets coinciding
with the modality of perception, namely visión, vista, mirada and audición, oído,
escucha present semantic specialization. More particularly, a distinction has been
made between more eventive nominalizations – visión, audición and escucha –
and nominalizations tending toward referential interpretations – vista, mirada,
oído. This dichotomy was shown to be largely predictable from the semantics of
the suffixes that the nominalizations carry. In this semantic characterization, the
opposition between voluntary and involuntary perception turned out to be less
decisive than the distinction between visual and auditory perception: except for
oído, auditory PNs clearly tend toward direct perception event readings (often
with communicative meaning), whereas visual PNs frequently trigger the cogni-
tive event reading. This semantic extension coincides with what has been observed
for their corresponding verbs.
In the second part of this study, it was shown that, as has been argued for
other types of nominalization, the semantics of PNs leads to differences in their
morpho-syntactic behavior. A study of the behavior of postnominal preposi-
tional phrases has demonstrated that the six PNs display clear preferences as
to the realization of their arguments: the referential nominals oído, mirada and
vista most frequently combine with the syntactic expression of the First Argu-
ment or perceiver, whereas the event nominals visión, audición and escucha in
most cases trigger the expression of the Second Argument or object of percep-
tion. A more detailed analysis revealed that in the case of visual PNs the cogni-
tive reading is to some extent related to abstract SAs and the use of particular
prepositions, which parallels the use of the that-clause after PVs. Moreover, this
reading exhibits some similarities with state nominals, such as the impossibility
to use the preposition por (parte de) ‘by’ as FA marker. This hybrid character of
PNs, behaving both as event nominals and state nominals, recalls the difficult
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
aspectual classification of the verbs they derive from and will constitute the topic
of future research.
References
Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization and Ergativity [Lin-
Amador Rodríguez, Luis Alexis. 2009. La derivación nominal en español: nombres de agente,
instrumento, lugar y acción. Frankfurt: Peter Lang.
Barque, Lucie, Huyghe, Richard, Jugnet, Anne & Marín, Rafael. 2009. Two types of deverbal
activity nouns in French. In 5th International Conference on Generative Approaches to the
Lexicon, 169–175. Pisa: Istituto di Linguistica Computazionale.
Bisetto, Antonietta & Melloni, Chiara. 2005. Result nominals: A lexical-semantic investigation.
In On-line Proceedings of the Fifth Mediterranean Morphology Meeting (MMM5), Geert
Booij, Luca Ducceschi, Bernard Fradin, Emiliano Guevara, Angela Ralli & Sergio Scalise
(eds), 393–412. 〈http://mmm.lingue.unibo.it/mmm-proc/MMM5/MMM5-Proceedings_
full.pdf〉 (September 2012).
Corominas, Joan. 1970. Diccionario crítico etimológico de la lengua castellana. Bern: Francke.
De Miguel, Elena. 1996. Nominal infinitives in Spanish: An aspectual constraint. Canadian
Journal of Linguistics 41: 29–53.
Demonte, Violeta. 1999. El adjetivo: Clases y usos. La posición del adjetivo en el sintagma nomi-
nal. In Gramática descriptiva de la lengua española, Ignacio Bosque & Violeta Demonte
(eds), 129–215. Madrid: Espasa.
Díaz Hormiga, María Tadea. 2005. Entre el léxico y la sintaxis: a propósito de la denominada
sintaxis interna de las formaciones derivadas. In Algunos problemas específicos de la descrip-
ción sintáctico-semántica, Juan Cuartero Otal & Gerd Wotjak (eds), 77–95. Berlin: Frank &
Timme.
Enghels, Renata. 2007. Les modalités de perception visuelle et auditive: différences conceptuelles
et répercussions sémantico-syntaxiques en espagnol et en français. Tübingen: Max Niemeyer.
DOI: 10.1515/9783110944884
Enghels, Renata. 2009. The syntactic position of the perceived participant as indicator of the
internal structure of the Spanish and French infinitival complement. Linguistics 47(3):
759–791. DOI: 10.1515/LING.2009.025
Fábregas, Antonio. 2010. A syntactic account of affix rivalry in Spanish nominalizations. In The
Syntax of Nominalizations across Languages and Frameworks, Artemis Alexiadou & Monika
Rathert (eds), 67–92. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110245875.67
Fábregas, Antonio & Marín, Rafael. 2012. The role of Aktionsart in deverbal nouns: State
nominalizations across languages. Journal of Linguistics 48: 35–70. DOI: 10.1017/
S0022226711000351
Fábregas, Antonio, Marín, Rafael & McNally, Louise. 2012. From psych verbs to nouns. In Telic-
ity, Change and State: A Cross-categorical View of Event Structure, Violeta Demonte & Louise
McNally (eds), 162–185. Oxford: OUP. DOI: 10.1093/acprof:oso/9780199693498.003.0007
Fradin, Bernard. 2012. Les nominalisations et la lecture ‘moyen’. Lexique 20: 129‒156.
Giammatteo, Mabel, Albano, Hilda & Ghio, Adalberto. 2005. Clases de predicados y nomi-
nalización. In Algunos problemas específicos de la descripción sintáctico-semántica, Juan
Cuartero Otal & Gerd Wotjak (eds), 35–48. Berlin: Frank und Timme.
 Elisa Bekaert & Renata Enghels
Gisborne, Nikolas. 1993. Nominalizations of perception verbs. UCL Working Papers in Linguis-
tics 5: 23–44.
Givón, Talmy. 2001. Syntax: An Introduction, Vol. 2, 1–37. Amsterdam: John John Benjamins.
Gutiérrez Ordóñez, Salvador. 2005. Diátesis no verbal. In Algunos problemas específicos de la
descripción sintáctico-semántica, Juan Cuartero Otal & Gerd Wotjak (eds), 17–33. Berlin:
Frank & Timme.
Haiman, John. 1980. The iconicity of grammar: Isomorphism and motivation. Language 56:
515–540. DOI: 10.2307/414448
Hanegreefs, Hilde. 2006. La construcción preposicional con mirar: Un análisis sintáctico-
semántico. Boletín de Lingüística 18(25): 22–65.
Heyvaert, Liesbet. 2003. A Cognitive-functional Approach to Nominalization in English. Berlin:
Mouton de Gruyter.
Heyvaert, Liesbet. 2008. The periphrastic realization of participants in nominalizations.
Semantic and discourse constraints. In Asymmetric Events [Converging Evidence in Lan-
guage and Communication Researach 11], Barbara Lewandowska-Tomaszczyk (ed.),
245–259. Amsterdam: John Benjamins.
Hopper, Paul J. & Thompson, Sandra A. 1985. The iconicity of the universal categories ‘noun’
and ‘verb’. In Iconicity in Syntax [Typological Studies in Language 6], John Haiman (ed.),
Huyghe, Richard & Marín, Rafael. 2007. L’héritage aspectuel des noms déverbaux en français et
en espagnol. Faits de Langues 30: 265–274.
Ibarretxe-Antuñano, Iraide. 1999. Polysemy and Metaphor in Perception Verbs: A Cross-
linguistic Study. Ph.D. dissertation, University of Edinburgh.
Kirsner, Robert S. & Thompson, Sandra A. 1976. The role of pragmatic inference in semantics:
A study of sensory verb complements in English. Glossa 10(2): 200–240.
Knittel, Marie Laurence. 2011. French event nominals and number inflection. Recherches lin-
guistiques de Vincennes 40: 127–148.
Meinschaefer, Judith. 2003. Nominalizations of French psychological verbs. In Selected Papers
from Going Romance [Current Issues in Linguistic Theory 245], Josep Quer, Jan Schro-
ten, Mauro Scorretti, Petra Sleeman & Els Verheugd (eds), 231–246. Amsterdam: John
Benjamins.
Melloni, Chiara. 2007. Polysemy in Word Formation: The Case of Deverbal Nominals. Ph.D.
dissertation, University of Verona.
Melloni, Chiara. 2010. Action nominals inside: Lexical-semantic issues. In The Semantics of
Nominalizations Across Languages and Frameworks, Monika Rathert & Artemis Alexiadou
(eds), 141–168. Berlin: Mouton de Gruyter. DOI: 10.1515/9783110226546.141
Miller, George A. & Johnson-Laird, Philip N. 1976. Language and Perception. Cambridge: CUP.
Miller, Philip & Lowrey, Brian. 2003. La complémentation directe et indirecte des verbes de per-
ception en anglais. In Actes des journées scientifiques 2000/2001. La préposition dans la rection
des verbes, Vol. 19, Jean Pauchard (ed.), 115–135. Reims: Presses Universitaires de Reims.
Osswald, Rainer. 2005. On result nominalization in German. In Proceedings of Sinn und Bedeu-
tung 9, Emar Maier, Corien Bary, and Janneke Huitink (eds), 256–270. 〈http://pi7.fernuni-
hagen.de/osswald/papers/sub04.pdf〉 (September 2012).
Pharies, David A. 2002. Diccionario etimológico de los sufijos españoles y de otros elementos fina-
les. Madrid: Gredos.
Chapter 3. Nominalizations of Spanish perception verbs at the syntax–semantics interface 
Picallo, M. Carme. 1999. La estructura del sintagma nominal: las nominalizaciones y otros sus-
tantivos con complementos argumentales. In Gramática descriptiva de la lengua española,
Ignacio Bosque and Violeta Demonte (eds), 363–393. Madrid: Espasa.
Piera, Carlos & Varela, Soledad. 1999. Relaciones entre morfología y sintaxis. In Gramática
descriptiva de la lengua española, Ignacio Bosque & Violeta Demonte (eds), 4367–4422.
Madrid: Espasa Calpe.
Real Academia Española. 2009. Nueva gramática de la lengua española. Madrid: Espasa Libros.
Real Academia Española. Online database CREA.Corpus de referencia del español actual. 〈http://
corpus.rae.es/creanet.html〉 (September 2012).
Rodríguez Espiñeira, María José (2004). Nominalizaciones con infinitivo. In Lecciones de sin-
taxis española. Santiago de Compostela: Universidad de Santiago de Compostela. 79–110.
Scott, Alan K. 2010. Accounting for the semantic extension of derived action nouns. Journal of
Linguistics 46: 711–734. DOI: 10.1017/S0022226710000034
Sleeman, Petra & Brito, Ana Maria. 2010. Nominalization, event, aspect and argument structure:
A syntactic approach. In Argument Structure from a Crosslinguistic Perspective [Linguistik
Aktuell/Linguistics Today 158], Maia Duguine, Susana Huidobro & Nerea M adariaga
(eds), 113–129. Amsterdam: John Benjamins.
Sweetser, Eve. 1990. From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Seman-
tic Structure. Cambridge: CUP. DOI: 10.1017/CBO9780511620904
Taylor, John R. 1995. Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Claren-
don Press.
Viberg, Åke. 1984. The verbs of perception: A typological study. In Explanations for Language
Universals, Brian, Butterworth (ed.), 123–162. Berlin: Mouton de Gruyter.
chapter 4
Case assignment, aspect, and (non-)expression

of patients
A study of the internal structure
of Czech verbal nouns
Věra Dvořák
Rutgers University
After reviewing various surface realizations of agents, patients and goals in

Czech nominalizations, I present a syntactic analysis which straightforwardly
accounts for the case form of these arguments, based on the well-known idea in
the literature that nouns can share with verbs a substantive part of the extended
verbal projection. Moreover, both imperfective verbs and nouns can combine
with null existentially interpreted patients while neither perfective verbs nor
perfective nouns allow them. I explain this as the interaction of the properties of
verbal Aspect/Quantity category and the missing number projection of implicit
patients. Finally, I show that only nominals (regardless of their aspectual value)
but not verbs can combine with null patients referring to an entity from the
previous discourse.
1. Introduction
Nouns derived from verbs preserve to a certain extent the argument structure of
verbs (Chomsky 1970; Abney 1987; Grimshaw 1990, and many others). Among
these deverbal nouns, names of events are the most faithful to the original valency –
as opposed to names of agents, instruments, places, or results. One of the reasons
is that they combine with both an agent as well as various internal arguments, as
in the classical example in (1), dating back to Chomsky (1970).
(1) the enemy’s destruction of the city
On the other hand, the amount of syntactic structure shared by verbs and nouns
is a matter of a long-standing debate in linguistics. This article contributes to this
 Věra Dvořák
debate by arguing that certain event nouns have not only the same valency frames
as corresponding verbs but also the same aspectual properties affecting the behav-
ior of their internal arguments in the same way in both verbal as well as nominal
types of structures (see Borer 1999 or Alexiadou 2001 for a similar conclusion).
I focus on one type of Czech deverbal event-naming nouns, those ending in
-ní/tí. They are sometimes called “verbal nouns” because they are the most “verb-
like” of all Czech deverbal nominals. These nouns also represent the most pro-
ductive class of deverbal nouns in Czech because they can be derived from all
verbal stems except those of a few modal verbs. In English, they have the closest
parallel in the so-called “ing-of ” gerundial nouns such as Peter’s smoking of cigars,
cf. Abney (1987). The prenominal position, which is in English occupied by the
Saxon genitive (the possessive expression formed by adding an apostrophic ’s to a
noun), is in Czech occupied by the so-called possessive adjective, formed from a
noun by adding a suffix -ovo (for masculine and neuter nouns in the role of pos-
sessors) or -ino (for feminine nouns as possessors). An example is given in (2),
with a possessive adjective nepřítelovo ‘enemy’s’ derived from the noun nepřítel
‘enemy’. The English postnominal prepositional of-phrase formally corresponds to
a noun in non-prepositional genitive morphological case in Czech.
(2) nepřítel-ovo zničení měst-a
enemy-poss.nom.sg destroying.nom.sg town-gen.sg
‘the enemy’s destruction of the town’
2. Nominal valency frames and their surface realization
In this article I look only at verbal nouns that combine with at least two arguments,
because they are the most relevant for the analysis I am going to present. (How-
ever, my analysis could be extended to monovalent verbal nouns as well.) Nominal
valency frames of different types of verbal nouns have been already discussed mul-
tiple times in a more general way in Czech literature, esp. by Novotný (1980), Piťha
(1992), Karlík (2000 and 2002), Panevová (2000), Veselovská (2001), Procházková
(2006), and Kolářová (2010).
2.1 (Mono)transitive nominalizations

If both agent and patient are overtly expressed with verbal nouns, the patient is
assigned postnominal genitive case and the agent is assigned postnominal instru-
mental case (a parallel to an English by-phrase). I use the terms ‘agent’ and ‘patient’
in the sense of theta-roles, i.e. as generalizations over a part of the semantics of pred-
icates that is syntactically relevant. One theta-role can thus subsume several specific
thematic relations, which are nevertheless all associated with one structural position.
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
(3) přemlouvání stařenk-y podvodn-ým prodavač-em

persuading.nom.sg grandma-gen.sg fraudulent-ins.sg seller-ins.sg
‘the persuading of an old lady by a fraudulent seller’
Singular animate nouns without any modifier, both agents and patients, can also
take the form of prenominal possessive adjectives. For example, prodavačovo
‘seller’s’ in (4) functions as an agent, and stařenčino ‘old lady’s’ in (5) functions
as a patient. The construction in (5) has a question mark because it is slightly
degraded with respect to the more preferred construction with the equivalent
meaning in (4).
(4) prodavač-ovo přemlouvání stařenk-y
seller-poss persuading.nom.sg grandma-gen.sg
‘a seller’s persuading of an old lady’
(5) ?stařenč-ino přemlouvání prodavač-em
grandma-poss persuading.nom.sg seller-ins.sg
‘an old lady’s persuading by a seller’
Agents can appear in the postnominal genitive phrase only if the patient is not
present overtly. The resulting construction is then ambiguous between an agent
and a patient interpretation of the genitive phrase. The same ambiguity is attested
with the prenominal possessive adjective if no postnominal genitive phrase is
present.
(6) a. přemlouvání podvodn-ého prodavač-e
persuading.nom.sg fraudulent-gen.sg seller-gen.sg
‘a fraudulent seller’s persuading’ (‘of/by a seller’)
b. prodavač-ovo přemlouvání
seller-poss persuading.nom.sg
‘a seller’s persuading’ (‘of a seller/by a seller’)
Multiple postnominal genitives are not allowed unless one of them is inherent
case associated with the underlying verb; see Panevová (2000) and this volume for
examples.
(7) *přemlouvání stařenk-y prodavač-e
persuading.nom.sg grandma-gen.sg seller-gen.sg
‘the persuading of an old lady of a seller’
2.2 Ditransitive nominalizations

Dvořák (2010) shows that there are two types of ditransitives in Czech:
1. Dat>Acc ditransitives which have the goal-like argument (corresponding to

the indirect object) above the patient (corresponding to the direct object). They
 Věra Dvořák
involve the projection of the applicative phrase vApplP (Marantz 1993) assign-
ing dative case in its specifier and associated with a beneficiary/recipient theta-
role. On account of this dative-assigning position, these ditransitives belong to
“high dative verbs”; see example (8a) with a verb darovat ‘to give’.
2. Acc>Dat ditransitives which have the direct object above the indirect one
and which should be analyzed as involving the conflated null preposition
(P0) assigning oblique case and associated with a path theta-role, cf. the simi-
lar analysis of McFadden (2004) for German. These ditransitives belong to
“low dative verbs” and they are represented by the verb přizpůsobit ‘to adjust’
in (8b).
(8) a. Učitel daroval Mari-i knih-u.
teacher.nom gave Mary-dat book-acc
‘The teacher gave Mary a book.’
b. Učitel přizpůsobil cvik-y student-ům.
teacher.nom adjusted exercises-acc students-dat
‘The teacher adjusted the exercises to students’ needs.’
On the surface, both word orders (an accusative argument preceding a dative one
as well as a dative argument preceding an accusative one) are allowed for both
types of verbs, but Dvořák (2010) provides multiple arguments showing that only
one of these orders is basic for each class of ditransitive verbs.
We just saw in 2.1 that a patient argument in accusative always changes its
form to genitive case (or a possessive adjective) in monotransitive nominaliza-
tions. When ditransitives are nominalized, a patient also always bears postnomi-
nal genitive (or a prenominal possessive adjective form), as expected, while a goal
retains its dative case. Interestingly, the patient and the goal arguments appear in
the Gen>Dat order for both Acc>Dat as well as Dat>Acc verbs undergoing nomi-
nalization. If we expected the plain parallelism between the verbal and the nomi-
nal structure, the word order under nominalization would be as expected for low
dative verbs but the opposite from the expected one for high dative verbs.
(9) Dat>Acc ditransitive nominalized
a. Darování knih-y Mari-i (se učitel-i nevyplatilo).
giving.nom book-gen Mary-dat refl teacher-dat not_paid_off
‘Giving a book to Mary (didn’t pay off to the teacher).’
b. ??Darování Mari-i knih-y (se
giving.nom Mary-dat book-gen refl
učitel-i nevyplatilo).
teacher-dat not_paid_off
‘Giving Mary a book (didn’t pay off to the teacher).’
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
(10) Acc>Dat ditransitive nominalized

a. Přizpůsobení cvik-ů student-ům (bylo marné).
adjusting.nom exercises-gen students-dat was useless
‘Adjusting the exercises to students’ needs (was useless).’
b. *Přizpůsobení student-ům cvik-ů (bylo marné).
adjusting.nom students-dat exercises-gen was useless
It was noted by an anonymous reviewer that the requirement for the genitive to
precede the dative in nominalizations can be subsumed under a more general
requirement that the genitive always has to precede all other postnominal argu-
ments and adjuncts. See the contrast between (3) where the agent in the instru-
mental follows the patient in the genitive and the following example where the
agent in the instrumental precedes the patient in the genitive.
(11) *přemlouvání prodavač-em stařenk-y
persuading.nom.sg seller-ins.sg grandma-gen.sg
‘the persuading by a seller of an old lady’
2.3 Syntactic realization of participants

When analyzing the syntactic structure of verbal nouns, I advocate the mor-
phosyntactic approach based on the assumption in Generative Grammar that a
sequence of functional categories is superimposed above the lexical entry (see esp.
the literature within the Distributed Morphology framework). Nominals derived
from verbs have a partially common functional sequence that is “closed-off ” by a
lexical-category morpheme at a certain point (Fu et al. 2001). Concretely, Czech
verbal nouns have a nominalizing head (spelled-out as a suffix -í) attached to the
extended verbal projection (the term due to Grimshaw 1991). This verbal projec-
tion contains at least a (big) verb phrase (VP), spelled out as a verbal root,1 a little
verb phrase (vP), spelled-out as various stem suffixes, see Jablońska (2007), and
a voice phrase (VoiceP), spelled-out as a “passivizing” morpheme -n/t-. See the
schematic tree diagram for -ní/tí nominalizations in the Figure 1 below, followed
by a brief discussion of how various types of participants are introduced syntacti-
cally and how they receive case. A more detailed discussion of this syntactic imple-
mentation can be found in Dvořák (2011).
. What I label as “VP” is nowadays often labeled as “√P” instead, in order to capture the
intuition that the root itself is uncategorized and it is the little v that functions as a verbalizer,
as proposed in Marantz (1997 and 2007). I stick to a more traditional way of labeling here.
 Věra Dvořák
DP
D nP
n[Gen] VoiceP
-í DPi[Gen]
Voice[-active] (vApp1P)
-n/t- (ti) (vApp1P)
(DP[Dat])
(vApp1[Dat]) vP
v VP
stem ti
suffix
V (PP)
verb (PØ[Dat]) (DP[Dat])
Figure 1. Schematic partial tree for -ní/tí nominalizations
2.3.1 Patients
In my analysis I employ the tripartite verbal phrase, introduced in Pylkkänen
(2002), consisting of a VoiceP, a vP and a lexical projection. The [–active] Voice
present in nominalizations is defective in the sense that it has no case-marking
capacity. As a result, the Determiner phrase (DP) merged in the internal argument
position of the specifier of a verbal phrase (Spec,VP) gets the patient theta-role
(thus conforming to UTAH (Baker 1997)) but not the canonical object case-
marking. The patient-expressing DP is then case-marked later: in the spirit of Case
Filter (Chomsky 1980: 25), it undergoes raising that leads to the DP’s local rela-
tionship with the genitive-valuing nominal head.
2.3.2 Agents
The unavailability of accusative case-marking is connected with the inability of
a verb to project an agent theta-role (Burzio 1986). The agent, normally pro-
jected in Spec,VoiceP (Hale & Keyser 1993; Chomsky 1995; Kratzer 1996) thus
becomes free to reappear in an adjunct instrumental DP, cf. Roeper & van Hout
(1999: 187). The deficiency of Voice in nominals is confirmed by the presence of
an -n/t- morpheme which is in Czech found also in passive structures (but see
Procházková 2006: 64–67 for discussion of the different scope of this suffix in pas-
sives as opposed to nominalizations).
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
2.3.3 Other arguments

If such categories like dative-assigning vAppl or P0 are present in the nominalized
verbal structure, it explains not only the preserved dative morphological marking
but also that dative DPs are associated with the same theta-roles as they were in
the purely verbal structure.
2.3.4 Details of implementation

Under the assumption that the derivation proceeds by phases (Chomsky 2000,
2001) whereby Spell-Out applies to the complements of phase heads, the patient
DP merged within a VP has to move to the edge of a VoiceP phase right before vP
(or vApplP) is sent to Spell-Out. Otherwise the derivation won’t converge due to
the unchecked/unassigned case feature on the object DP. This movement complies
with the Last Resort condition (Chomsky 1995: 256) because it is a necessary step to
eliminate unchecked case features on a DP that could not be eliminated otherwise.
This DP movement explains why direct objects have to precede indirect
objects in both types of nominalized ditransitives, as shown in (9b) and (10b).
Dative DPs, on the other hand, stay in situ because their theta-marking and their
(inherent) case-assignment are simultaneous. The same would be true for all other
inherently case-marked DPs as well as for prepositional phrases which appear in
the valency frames of nominalized verbs.
3. Aspectual characteristics of verbal nouns
3.1 Aspectual affixes

Czech verbal nouns can combine with the same set of aspectual affixes as
verbs, namely with perfectivization prefixes such as pře- or do-, and secondary-
imperfectivization suffixes such as -va-.
(12) a. malova-l pře-malova-l
paint.ipfv-pst pref-paint.pfv-pst
‘(he) was painting’ ‘(he) repainted’
pře-malová-va-l
pref-paint-ipfv-pst
‘(he) was repainting’
b. malová-ní pře-malová-ní
paint.ipfv-ing.nom pref-paint.pfv-ing.nom
‘painting’ ‘repainting’
pře-malová-vá-ní
pref-paint-ipfv-ing.nom
‘process of repainting’
 Věra Dvořák
(13) a. malova-l do-malova-l

paint.ipfv-pst pref-paint.pfv-pst
‘(he) was painting’ ‘(he) finished painting’
do-malová-va-l
pref-paint-ipfv-pst
‘(he) was finishing painting’
b. malová-ní do-malová-ní
paint.ipfv-ing.nom pref-paint.pfv-ing.nom
‘painting’ ‘finishing painting’
do-malová-vá-ní
pref-paint-ipfv-ing.nom
‘process of finishing painting’
It has been agreed in the literature that there are two types of prefixes in Slavic:
lexical (or VP-internal) such as na- in nanést ‘to dash something somewhere’ or
po- in pomalovat ‘to cover something with painting’, and superlexical (or VP-
external) such as cumulative na- in napéct ‘to bake a lot of ’ or distributive po- as in
poutrácet ‘to spend bit by bit’; see Smith (1991) for the introduction of this distinc-
tion. Many researchers working on Slavic languages assume that lexical prefixes
are generated within a VP under Res (the head of a Resultative Phrase) or a paral-
lel category that introduces a result state subevent (Svenonius 2004; Ramchand
2004; Romanova 2004, among others). Since these prefixes change the event type,
they constitute so-called inner aspect (telicity). Superlexical prefixes, on the other
hand, are generated outside of VP, much like adverbs, and they constitute what is
called outer aspect (AspP); see Gehrke (2008), who explicitly argues that Czech
superlexical prefixes function as adverbial modifiers. Secondary imperfectiviza-
tion suffixes (-va-) are typically analyzed as a specific instantiation of the Asp head
itself (Ramchand 2004: 355).
The purpose of this section is not to embrace the extremely complicated area of
aspectual affixes in Slavic but to point out that if we want to capture fully the paral-
lels between verbs and nominals, including those exemplified in (12) and (13), we
have to assume that such verbal syntactic projections like ResP or AspP are present
within the nominals as well, as depicted in Figure 2 below. Moreover, we can expect
that the aspect- and telicity-related projections will further interact with the pro-
jections introducing arguments that were discussed in 2.3. Section 4.3 is devoted
to one example of such interaction, namely the interaction between VP and AspP.
3.2 Aspectual adverbials

Other support for the presence of an aspectual node in the tree in Figure 2 comes
from aspect-sensitive adverbials. Perfective verbs (those describing telic events)
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
DP
D nP
n[Gen] VoiceP
-í DPi[Gen]
Voice[-active] AspQP
-(e)n/t- ti
AspQ[±PF] vP
v VP
stem ti
suffix
V (ResP)
verb (Res) (DP)
Figure 2. Schematic tree for -ní/tí nominalizations (dative-assigning heads omitted)
combine only with terminative adverbials like “in an hour”2 while imperfective
verbs (verbs with a primary atelic interpretation) combine with durative adverbi-
als like “for an hour” (see Dowty 1979 for the introduction of this test in English).
In Czech, imperfective verbs can describe telic events as well but only in habitual
contexts. As expected, in that case they can combine with terminative adverbials
(Dočekal & Kučerová 2010). Importantly, perfective (14b) and imperfective (15b)
nominals fully follow the same pattern.
(14) a. na-malova-t jeden portrét *hodin-u/za hodin-u
pref-paint.pfv-inf one portrait.acc hour-acc/in hour-acc
‘to paint a portrait for an hour/in an hour’
b. na-malová-ní jednoho
pref-paint.pfv-ing.nom one
portrét-u *hodin-u/za hodin-u
portrait-gen hour-acc/in hour-acc
‘painting a portrait for an hour/in an hour’
. “In time x” adverbials are ambiguous between the meaning where the whole event takes
time x and the meaning where the event described by the VP starts after time x has passed.
Here I consider only the former, event-measuring interpretation.
 Věra Dvořák
(15) a. malova-t jeden portrét hodin-u/za hodin-u

paint.ipfv-inf one portrait.acc hour-acc/in hour-acc
‘to be painting a portrait for an hour/in an hour (hab.)’
b. malová-ní jednoho portrét-u hodin-u/za hodin-u
paint.ipfv-ing.nom one portrait-gen hour-acc/in hour acc
‘painting a portrait for an hour/in an hour (habitually)’
Notice that Czech -ní/tí nominals are very different from English -ing nominals
in this respect. It is known that English nominalizations neutralize the difference
between durative and terminative adverbials, as the full grammaticality of the
gloss in (14b) confirms.
4. Aspect-related patient obligatoriness
4.1 Introducing the contrast between perfectives and imperfectives

Czech verbal nouns take not only the same aspectual affixes and aspectual modifiers
as verbs but they also seem to impose the same requirements on the expression of
patients. Procházková (2006) shows that the patient of nouns derived from imper-
fective verbal stems does not have to be expressed overtly but can be just “implicitly
satisfied” by being existentially quantified over (Zucchi 1989: 186). In contrast, nouns
derived from transitive perfective stems become typically ungrammatical unless the
slot for the patient gets filled by some overt DP. See the parallelism between the verbal
structures in (16a) and (17a), and the nominal structures in (16b) and (17b). The
underscore in the following examples stands for the phonetically null patient.
(16) a. Včera večer jsme kreslili__ / *nakreslili__.
yesterday evening aux.1pl drew.ipfv drew.pfv
‘We were drawing/drew yesterday in the evening.’
b. Celý večer nám zabralo kreslení__ / *nakreslení__.
whole evening us took drawing.ipfv drawing.pfv
‘Drawing took us the whole evening.’
(17) a. V neděli jsme lovili__/*ulovili__ v lese.
on Sunday aux.1pl hunted.ipfv/pfv in forest
‘On Sunday, we were hunting/hunted in the forest.’
b. Tát-ovi udělalo radost lovení__/*ulovení__ v lese.
dad-dat made joy hunting.ipfv/pfv in forest
‘Hunting in the forest made my dad happy.’
Even though the observation made in Procházková (2006) is correct, the matter is
not as simple as it might seem. The contrast between perfective and imperfective
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
nominals holds only in all-new contexts (typically represented by “out-of-the-

blue” contexts). I leave the discussion of the cases where nominals do not behave
like verbs when it comes to non-expressing patients for the Section 5.1.
Note on resultativity
Before we look at the aspect-driven contrast in (16) and (17), I would like to point
out that all of the generalizations presented so far hold for verbal nouns which
describe events. The eventive interpretation is the primary interpretation of -ní/tí
nominals though they can often have the derived, resultative interpretation. The
valency frames of such resultative nouns are more varied (see esp. Kolářová 2010
and this volume) and the aspect-related properties described above do not hold for
them either. See the contrast between the eventive (18b) and the resultative (18c)
noun oznámení, derived from the verbal stem oznámi- ‘announce’. The patient is
only omissible with the resultative noun in (18c).
(18) a. Komise konečně oznámila *(výsledk-y).
committee finally announced.pfv results-acc
‘The committee finally announced the results.’
b. Oznámení *(výsledk-ů) za tak krátkou chvíli
announcing.pfv results-gen in such short while
všem vadilo.
all.dat embarrassed
‘Nobody liked the announcement of the results in such a short time.’
c. Na nástěnce viselo oznámení (změn-y
on board hang announcing.pfv change-gen
sídl-a společnost-i).
seat-gen company-gen
‘There was an announcement (of the change of the company’s
headquarters) on the notice board.’
Note on perfective verbs

The generalization that perfective transitive verbs need an overt object does not
hold for a few lexicalized cases, which are well-known in the literature (Daneš
1971; Panevová 1974, among others). The most prominent of these are the verbs
denoting various chores in (19) and the verbs denoting committing something
wrong in (20a) and (20b). That this is not a systematic exception is shown by two
facts: omitting an overt object is allowed neither for all the house works nor for all
the wrong-doings. Moreover, near synonyms behave differently when it comes to
allowing null objects, cf. the contrast between the verbs zabít ‘kill’ versus zavraždit
‘murder’ in (20a), or zradit ‘betray’ versus podvést ‘deceive’ in (20b).
 Věra Dvořák
(19) Karel vytřel__, vysál__, uklidil__, zametl__,

Charles wiped.pfv vacuumed.pfv cleaned.pfv swept.pfv
vypral__, nakoupil__, ustlal__, uvařil__.
laundered.pfv shopped.pfv made_the_bed.pfv cooked.pfv
‘Charles wiped the floor, vacuumed, cleaned up, swept the floor, did the
laundry, did the shopping, made the bed, and cooked dinner.’
(20) a. Karel zabil__. / ??Karel zavraždil__.
Charles killed.pfv Charles murdered.pfv
‘Charles killed.’ ‘Charles murdered.’
b. Karel zradil__. / ??Karel podvedl__.
Charles betrayed.pfv Charles deceived.pfv

‘Charles betrayed.’ ‘Charles deceived.’
None of the null patients combining with perfective verbs in (19) gets interpreted
existentially, i.e. as a null “something”. The empty position is rather interpreted as
a prototypical object that the given housework applies to, and which can be further
contextually determined. For example, object-less uvařit ‘to cook (completely)’
doesn’t mean just to cook something but to make a full meal, i.e. to make dinner/
supper, depending on the situation. Similarly, the perfectives with the grammatical
null object in (20a) and (20b) entail that there was one particular person or group
of people who were killed/betrayed by Charles, not just some unquantified peo-
ple. Most relevantly for the line of analysis presented here, both verbs and nouns
behave equally when it comes to allowing these sorts of null patients in combina-
tion with perfective stems.
(21) Vyprání a uklizení zabralo Karl-ovi celý den.

laundering.pfv and cleaning.pfv took Charles-dat whole day
‘Doing the laundry and cleaning up took Charles the whole day.’
4.2 Existentially interpreted null patients

There are numerous proposals in the literature addressing how to analyze the non-
overt implicit arguments, ranging from analyses that give them a full syntactic
status to analyses that treat them as only conceptual elements with no linguistic
impact whatsoever; see Bhatt & Pancheva (2006) for a nice summary of the field.
In Dvořák (2013), I argue that implicit patients of the sorts exemplified in (16) and
(17) (i.e. those that systematically combine with imperfectives) should be analyzed
on a par with indefinitely interpreted bare plurals such as knihy ‘books’, and bare
mass nouns such as sůl ‘salt’ because they behave identically under various syn-
tactico-semantic tests: (i) they do not express quantity, (ii) they allow only a weak
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
(low-scope) indefinite reading, (iii) neither indefinite bare plurals or mass terms,
nor indefinite implicit patients can combine with perfective transitive verbs.
Bare plural and mass terms are standardly treated as existentially quantified
elements that do not undergo quantifier raising but stay in-situ where they are
existentially closed over (Diesing 1992). As a result, these “nonquantificational
indefinites”, as Diesing and Jelinek (1995) call them, can get only a weak (non-
specific) reading unless they overtly move out of VP. Benedicto (1997) further
refines this proposal by showing that the nuclear scope of the existential closure
corresponds to the c-commanding domain of the verb (after its movements),
rather than to the VP itself as in the original proposal by Diesing. I suggest that
this analysis of bare plurals and mass nouns should be extended to existentially
interpreted null patients and I employ the framework of Borer (2005a, b) to do so.
In Borer’s theory, both DPs and VPs have an articulated internal functional
structure where each projection is headed by a categorially labeled open value 〈e〉
that needs to be assigned range by some functional item, i.e. needs to be bound by
the appropriate functional operator.
DP TP
Det DP
〈e〉d #P T AspQP
Quantifier DP
(Subj-of-quant)
〈e〉# CIP 〈e〉# VP
Classifier …
〈e〉DIV NP
Figure 3. Structure of a decomposed DP and a decomposed TP (Borer 2005a and b)
Since going into the details of Borer’s general framework would take us too far
from the topic of this study, I will present just the part which is directly involved in
accounting for the difference between perfectives and imperfectives introduced in
4.1. In this framework, count interpretation is the property of syntax so all noun
denotations are mass (they are of type 〈e,t〉). As non-quantity structures, neither
bare plurals nor mass nouns project a category of number/quantity associated
with an open quantity value 〈e〉# (and only plurals project a category of classi-
fier associated with 〈e〉DIV where DIV stands for “divided”). In order to capture
 Věra Dvořák
the parallelism between overt non-quantity nouns like mass nouns and non-overt
non-quantity nouns like implicit patients, we have to assume that all of the follow-
ing three nominal structures are possible:
(22) a. bare plural noun: [DP 〈e〉d [ClP 〈e〉DIV [NP √noun ]]]
b. bare mass noun: [DP 〈e〉d [NP √noun ]]
c. null implicit patient: [DP 〈e〉d [NP pro〈e,t〉 ]]
Borer (2000a) further proposes that when the low-scope existential closure à la
Diesing applies, it assigns a range to an open value 〈e〉d, i.e. it binds the logical
variable 〈e〉d, but it is not capable of binding 〈e〉#. The result of combining a non-
quantity noun with a verb could be schematically captured as follows (subscripts
stand for the binding relation).
(23) a. existentially quantified overt noun:
∃i [V √verb] [DP 〈ei〉d [NP √noun ]]
b. existentially quantified non-overt noun:
∃i [V √verb] [DP 〈ei〉d [NP pro〈e,t〉 ]]
This proposal was originally formulated for verbs, but if we assume that verbal
nouns contain a full syntactic projection of the verbal head that introduces the
patient argument, as suggested in 2.3.1 (as well as several other higher verbal
heads connected by verb-movement), we can readily account for the existence of
indefinite null patients within nominals, without a need to introduce any addi-
tional theoretical machinery.
4.3 Incompatibility of null patients with perfective verbs and nouns

Once we understand the structure and the interpretation of existentially quanti-
fied null patients, we can proceed to analyzing the incompatibility of these patients
with perfective verbs and perfective nominals exemplified in (16) through (18b).
The strategy I am defending in this study is that whatever principle guides this
incompatibility in the case of verbs should also guide it for nominals, precisely
because in both cases it is tied to a purely verbal category of aspect (the presence
of which in the structure of verbal nouns was already justified in Section 3). In the
rest of this section I will sketch an analysis based on this logic.
I propose (Dvořák 2013) that it is the perfective verb’s need for a quantity
patient DP that excludes implicit patients (but also indefinite bare plural and mass
nouns) as licit direct objects of transitive perfective verbs. Borer (2005b) makes a
proposal along the same lines for English telic predicates even though she talks
only about bare plurals and mass terms, not about non-overt patients. I rely on
her framework once again when formally capturing the same incompatibility in
Czech.
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
Recall that according to Borer, not only nominals but also verbs project a cat-
egory of quantity which is in the case of verbs associated with an aspectual phrase
(labeled as AspQP whereby Q stands for quantity). The open value 〈e〉# heading
such aspectual projection is in need of range assignment, but in contrast to nomi-
nals, which can value their 〈e〉# by means of various quantifiers, perfective verbs do
not have any such direct range assigner. However, the verbal 〈e〉# can still be valued
indirectly, through specifier-head agreement. This happens when there is a quan-
tity DP in Spec,AspQ and the particular quantity value of this DP (labeled as Qi in
Figure 4) gets assigned not only to a nominal number [#P 〈e〉# ] but also gets copied
onto the verbal “number” [AspQP 〈e〉#]. As a consequence, only the combinations of
quantity DPs and transitive perfective verbs are well-formed.
AspqP
DP〈Qi〉
〈ei〉# VP
〈e〉d #P
…
Qi
〈ei〉# (CIP)
(〈e〉DIV) NP
Figure 4. Perfective verbs: indirect assignment of range to an open verbal-quantity value
When it comes to imperfective verbs, there are at least two theoretical pos-
sibilities: They either do not project AspQP associated with 〈e〉# at all, or they
do project it but in contrast to perfective verbs, they have an internal 〈e〉#-range
assigner. While the former path is taken by Borer (2005b), the latter one seems
more plausible to the author of this study, given that imperfectivity is in Slavic
languages associated with a specific progressive-like interpretation (e.g. Altshuler
2010), and the aspectual node should be the locus of this interpretation. One of
the morphology-based arguments for the latter approach is the existence of a mor-
phological reflex of imperfectivization, namely the suffix -va-, which is the most
productive means of deriving imperfective stems in Czech, cf. (12) and (13). On
the other hand, there is no unique perfectivizing affix (see Filip 2003 for an argu-
ment against the view of prefixes as perfectivity markers). The imperfectivizing
suffix -va- and its counterparts can thus fulfill the role of functional items assign-
ing the range to a verbal quantity value 〈e〉# directly, which in turn means that
 Věra Dvořák
imperfective verbs do not need an external range-assigner in the form of a quan-

tity DP in Spec,AspQ.
AspQP
DP
〈PROG〉
〈ePROG〉# VP
v
……
Figure 5. Imperfective verbs: direct assignment of range to an open verbal-quantity value
I hope I have demonstrated that if we acknowledge the presence of such verbal

projections like AspQP in the nominalized structure, and we further accept that
the head of this projection can interact with the argument introduced lower in the
structure in the same way as it does in the purely verbal structure, we can explain
an important parallel between verbs and nouns regarding the non-expression of
their objects. Since providing a full-fledged analysis of the interaction between
phonological nullity of patients and imperfectivity of verbs is not the task of this
study, precisely because there is nothing specifically nominal about this inter-
action, I will stop at this point in order to look at some interesting differences
between verbs and argument-taking nouns.
5. Other types of null patients
5.1 Contextually dependent null patients

In the previous section we discussed cases where patients with a meaning roughly
corresponding to “something” did not have to be expressed overtly with nominals
but could be just tacitly present. We found out that the same type of silent patient
can occur with verbs, and that nominal and verbal null patients can be analyzed
uniformly. In addition to such existentially interpreted null patients, Czech verbal
nouns can combine with null patients referring to a contextually salient entity, i.e.
they behave like true overt pronouns. Interestingly, this is something that verbs do
not allow.
(24) Context: Učitelka dala Karlíkovi za úkol nakreslit přes víkend sněhuláka.
‘The teacher gave Charlie the task to draw a snowman over the weekend.’
a. Karlík #kreslil__/ *nakreslil__ v neděli odpoledne.
Charlie drew.ipfv drew.pfv on Sunday afternoon
‘Charlie was drawing/drew on Sunday afternoon.’
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
b. Karlík nechal kreslení__ / nakreslení__ na neděli odpoledne.

Charlie left drawing.ipfv drawing.pfv for Sunday afternoon
‘Charlie left drawing (it) for Sunday afternoon.’
The possibility of taking a contextually-dependent null patient has no relation to

aspect: it is attested for both perfective and imperfective nominals, and it is not
attested for either perfective or imperfective verbs.
The expected form of the sentence in (24a) is with an overt pronoun ho ‘him’,
referring to the snowman. The number sign (#) in front of the verb kreslil in (24a)
marks that the sentence with the imperfective verb is not ungrammatical but only
semantically awkward because it can have an alternative meaning, involving a
patient’s implicit existential quantification. As we already discussed in relation to
(16a) and (17a), the sentence “Karlík kreslil v neděli odpoledne” could be roughly
paraphrased as “Charlie was drawing something on Sunday afternoon”. Since we
know from the context that he was supposed to draw a snowman, not just “some-
thing”, most speakers deduce that this “something” was presumably a snowman.
However, this meaning has the character of an implicature and as such it can be
canceled (in the same context as given in (24)).
(25) Karlík kreslil___ v neděli odpoledne, ale ne sněhulák-a.
Charlie drew.ipfv on Sunday afternoon but not snowman-acc
‘Charlie was drawing something on Sunday afternoon, but not the
snowman.’
On the other hand, the null patient of the perfective nominal nakreslení in (24b) has
to refer to the snowman mentioned in the previous sentence, therefore its context-
dependent interpretation cannot be canceled (it is a part of the presupposition).
(26) *Karlík nechal nakreslení___, ale ne sněhulák-a,
Charlie left drawing.pfv but not snowman-gen
na neděli odpoledne.
for Sunday afternoon
‘Charlie left the drawing but not (the drawing) of the snowman for Sunday
afternoon.’
The null patient of the imperfective nominal kreslení in (24b) allows both inter-
pretations: the existential one which we saw already in (16b) and (17b), or the
contextually dependent one. The latter, contextual one is most probably going to
be picked because the existential interpretation would be pragmatically odd in the
given context.
In (27) I provide one more set of examples showing the contrast between verbs
and nouns when it comes to allowing a null contextually dependent patient. This
example shows that the antecedent does not have to be a syntactically u niform
 Věra Dvořák
category. In this case, it is the subject of the preceding clause that serves as an ante-
cedent for the null object in the following clause.
(27) Context: Soutěž Supertalent měla letos dva moderátory.
‘The Supertalent contest had two hosts this year.’
a. Na začátku uváděla___ / *uvedla___
at beginning introduced.ipfv introduced.pfv
švédsk-á moderátork-a.
Swedish-nom host-nom
‘A Swedish host was introducing/introduced at the beginning.’
b. Počáteční uvádění___ / uvedení___
initial introducing.ipfv introducing.pfv
(švédsk-ou moderátork-ou) bylo trochu těžkopádné.
Swedish-ins host-ins was bit clumsy
‘The initial presentation (by a Swedish host) was a bit clumsy.’
While I gave a systematic account of various properties of verbal nouns that were
presented in the previous sections, I do not have a full-fledged explanation for the
contrast presented here, i.e. why nominals allow a discourse anaphoric null patient
but verbs don’t. I only suggest a possible way to go here.
Landau (2010) distinguishes two types of implicit arguments: strong implicit
arguments (which he labels pro) consisting of a D-feature (“D” stands for “Deter-
miner”) and a set of phi-features (person, number, gender), and weak implicit
arguments consisting only of phi-features. He also notes that the presence of a
case-feature requires the presence of a D-feature because case is a property of DPs
(Landau 2010: 380).
There is an important difference between verbs and nominals when it comes
to the presence of an unchecked case feature: projections of transitive verbs con-
tain an accusative (Acc) feature if there is an active Voice introducing an external
argument (or an active little v which is the head with the same function as Voice in
many proposals). Nominalized structures, on the other hand, contain a passive-like
Voice, cf. 2.3.1 and 2.3.2, so they do not have an unchecked Acc feature. Moreover,
genitive-assignment is optional in transitive nominals, as opposed to the obligatory
assignment of accusative case in transitive verbs (see Dvořák 2011 for more details
on structuralness versus inherentness of postnominal genitive). It follows that only
active verbal structures but not nominalizations based on transitive stems require
a full DP that would bear a direct object theta-role and structural case associated
with it. Nominals can combine also with non-D arguments because they do not
need to check their case against an unchecked case on a DP.
I already proposed that existentially quantified null patients have a D-layer,
cf. (22c), so they would count as strong implicit arguments for Landau. In Borer’s
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
framework, it is precisely the variable associated with the Determiner category that
gets bound by the existential quantifier, as shown in (23b). On the other hand, it is
plausible that discourse anaphoric null patients, exemplified in (24b) and (27b), do
not project the D-layer at all. Their incompatibility with active verbs would then
follow from the fact that they cannot satisfy the case feature on Voice/v. I leave the
verification of this hypothesis for further research.
5.2 [+HUM] null patients

So far, we have been focusing on non-overt patients that correspond to inani-
mate objects. The last set of data that I present here shows that verbal nouns can
also combine with non-overt patients denoting humans. The behavior of these
patients is different from that of non-human null patients in one important aspect:
they seem to be able to combine with perfective nominals even in out-of-the-blue
contexts, i.e. without the possibility of referring to an entity from the previous
discourse.
(28) a. Na vánočních trzích šidili__ / *ošidili__.
at Christmas markets cheated.ipfv cheated.pfv
‘They were cheating/cheated at the Christmas market.’
b. Na vánočních trzích došlo k šizení__/ošizení__.
at Christmas markets went to cheating.ipfv/pfv
‘At the Christmas market, cheating occurred.’
(29) a. Během obřadu kněz rychle křtil__ / *pokřtil__.
during ceremony priest.nom quickly baptized.ipfv baptized.pfv
‘During the ceremony, the priest was also quickly baptizing/also
quickly baptized.’
b. Obřad byl spojený s rychlým
ceremony was connected with quick
křtěním__ / pokřtěním__.
baptizing.ipfv baptizing.pfv
‘The ceremony was connected with quick baptizing.’
The sentences (28a) and (29a) show that perfective verbs taking [+HUM] direct
objects cannot combine with a null implicit patient in episodic contexts,3 on a
par with non-[+HUM]-object taking verbs, cf. (16a) and (17a). In contrast, verbal
nouns derived from the same stems are not ungrammatical if combined with a
null patient, regardless of their aspect; see (28b) and (29b) on one side, and (16b)
. Perfective verbs with human objects behave differently in generic sentences, but I will not
discuss those here at all.
 Věra Dvořák
and (17b) on the other. The perfective nominal is of particular interest here: Why
does the perfectively marked noun allow combination with a null patient when the
perfective verb derived from the same stem does not?
A detailed analysis of this interesting contrast goes far beyond the scope of
this paper so I make just a short related note here. The compatibility of a null
human patient and a perfective nominal in (28b) and (29b) would follow if: (1)
these patients expressed the category of nominal quantity, which allows them to
assign range to the open quantity value of perfective stems (see Figure 3), (2) they
were similar to contextually-dependent implicit patients discussed in 5.1 in that
they would not have the Determiner category so they could not be assigned case.
Consequently, they could not appear in structures with active Voice, which are
characterized by obligatory case assignment to their internal arguments.
6. Conclusion
This article had a threefold aim: (1) to give an overview of certain assumptions
about the syntactic structure of verbal nouns, especially the assumptions related to
the case forms of various members of nominal valency frames (Sections 2 and 3),
(2) to analyze one type of non-expressed patient that can be present in nominal
structures as well as verbal structures (Section 4), and (3) to examine whether
there are any other types of non-expressed patients (Section 5). The logical link
between the first two aims is that we can systematically account for all sorts of
verb – noun parallels without a need to extend our theoretical apparatus if we
assume that verbs and nouns share a substantial part of their functional structure.
When it comes to the first aim, the Czech data confirmed that nouns can
share with verbs all projections where internal arguments are syntactically intro-
duced: VP, PP, vApplP, but also other phrases in the extended verbal projection,
namely vP, AspQP, VoiceP. However, nouns differ from (active) verbs in project-
ing passive-like Voice which is deprived of case-marking capacity. Nouns provide
their own structural slot, licensed by the nominalizing suffix (of category n) where
both patient and agent DPs can get genitive-case marked.
In the second step, it has been shown that there are at least two types of non-
overt arguments that combine with verbal nouns: existentially quantified indefinite
patients and contextually dependent patients. While the former ones are parallel
to non-overt indefinite verbal arguments in not combining with perfective verbs,
the latter ones are attested only in nominal structures. Figure 6 summarizes which
types of null patients can appear in which types of structures.
The behavior of existential non-overt patients in nouns, which mirrors the
behavior of the same types of patients in verbs, was accounted for by the presence
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
∃-quantified indefinite patient Context-dependent patient
Verbs Perfective  
Imperfective  
Nouns Perfective  
Imperfective  
Figure 6. Typology of implicit non-human patients in Czech
of an aspectual quantificational head in their structures and its need for a quantity
object in its specifier. The requirement associated with perfective aspect is not met
in the case of mass-like null patients.
On the other hand, I related the non-existence of contextually interpreted null
patients in active verbal structures with the structural element missing in nomi-
nals: accusative-assigning Voice. Finally, in 5.2 I discussed what seems like yet a
third type of null arguments, human implicit patients.
Even though I made several important findings throughout the paper, con-
tributing to a better understanding of the internal structure of verb-like nouns, the
last section revealed that there still remains a lot more work to be done in this area.
The classical research rule saying that the more questions get answered, the more
new questions arise got confirmed once again.
References
Abney, Steven P. 1987. The English Noun Phrase in its Sentential Aspects. Ph.D. dissertation,
MIT.
Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization and Ergativity [Lin-
Altshuler, Daniel G. 2010. Temporal Interpretation in Narrative Discourse and Event Internal
Reference. Ph.D. dissertation, Rutgers University.
Baker, Mark. 1997. Thematic roles and syntactic structure. In Elements of Grammar, Liliane
Haegeman (ed.), 73–117. Dordrecht: Kluwer. DOI: 10.1007/978-94-011-5420-8_2
Benedicto, Elena. 1997. The Syntax and Semantics of Non-canonical NP Positions. Ph.D. dis-
sertation, University of Amherst.
Bhatt, Rajesh & Pancheva, Roumyana. 2006. Implicit arguments. In Blackwell Companion to
Syntax, Vol. 2, Martin Everaert & Henk van Riemsdijk (eds), 558–588. Oxford: Blackwell.
Borer, Hagit. 1999. The form, the forming, and the formation of nominals. Ms, University of
Southern California.
Borer, Hagit. 2005a. Structuring Sense: An Exo-skeletal Trilogy, Vol. 1: In Name Only. Oxford:
OUP.
Borer, Hagit. 2005b. Structuring Sense: An Exo-skeletal trilogy, Vol. 2: The Normal Course of
Events. Oxford: OUP. DOI: 10.1093/acprof:oso/9780199263929.001.0001
 Věra Dvořák
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. DOI: 10.1007/978-94-009-4522-7

Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English Transformational
Grammar, Roderick A. Jacobs & Peter S. Rosenbaum (eds), 184–221. Waltham MA:
Ginn & Company.
Chomsky, Noam. 1980. On binging. Linguistic Inquiry 11: 1–46.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by Step: Essays on Mini-
malist Syntax in Honor of Howard Lasnik, John Martin, David Michaels & Jane Uriagereka
(eds), 89–155. Cambridge MA: The MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael
Kenstowicz (ed.), 1–53. Cambridge MA: The MIT Press.
Daneš, František. 1971. Větné členy obligatorní, potenciální a fakultativní (Obligatory, potential,
and facultative sentential constituents). In Miscellanea Linguistica, Miroslav Komárek (ed.),
131–138. Ostrava: Profil.
Diesing, Molly. 1992. Indefinites. Cambridge MA: The MIT Press.
Diesing, Molly & Jelinek, Eloise. 1995. Distributing arguments. Natural Language Semantics
3(2): 123–176.
Dočekal, Mojmír & Kučerová, Ivona. 2010. Aspectual presuppositions in Slavic and Romance.
In Formal Approaches to Slavic Linguistics 18: The Ithaca Meeting 2009, 125–139. Ann
Arbor MI: Michigan Slavic Publications.
Dowty, David R. 1979. Word Meaning and Montague Grammar: The Semantics of Verbs and
Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel. DOI: 10.1007
/978-94-009-9473-7
Dvořák, Věra. 2010. On the syntax of ditransitive verbs in Czech. In Formal Approaches to
Slavic Linguistics 18: The Ithaca Meeting 2009, 161–177. Ann Arbor MI: Michigan Slavic
Publications.
Dvořák, Věra. 2011. Inherent case and locality requirement: Evidence from ditransitives and
their nominalizations. In U. Penn Working Papers in Linguistics 17(1): Proceedings of PLC
34, 95–104. Philadelphia PA: Penn Linguistics Club.
Dvořák, Věra. 2013. When silent ‘something’ and silent ‘someone’ behave like mass nouns. Talk
presented at RULing VIII, Rutgers University. 〈http://vera.indus3.net/docs/RULingVIII.pdf〉
Filip, Hana. 2003. Prefixes and the delimitation of events. Journal of Slavic Linguistics 11(1):
55–101.
Fu, Jingqui, Roeper, Thomas & Borer, Hagit. 2001. The VP within process nominals: Evidence from
adverbs and the VP anaphor do – so. Natural Language and Linguistic Theory 19: 549–582.
DOI: 10.1023/A:1010654105760
Gehrke, Berit. 2008. Goals and sources are aspectually equal: Evidence from Czech and Russian
prefixes. Lingua 118: 1664–1689. DOI: 10.1016/j.lingua.2007.05.011
Grimshaw, Jane. 1991. Extended projection. Ms, Brandeis University.
Hale, Kenneth & Keyser, Samuel J. 1993. On argument structure and lexical expression of syntac-
tic relations. In The View from Building 20: Essays in Linguistics in Honor of Sylvian Brom-
berger, Kenneth Hale & Samuel J. Keyser (eds), 53–109. Cambridge MA: The MIT Press.
Jablońska, Patrycja. 2007. Radical Decomposition and Argument Structure. Ph.D. dissertation,
University of Tromsø.
Karlík, Petr. 2000. Valence substantiv v modifikované valenční teorii (Valency of nouns in a
modified valency theory). In Čeština – univerzália a specifika 2, Zdena Hladká & Petr
Karlík (eds), 181–192. Brno: Masarykova Univerzita.
Chapter 4. Case assignment, aspect, and (non-)expression of patients 
Karlík, Petr. 2002. Ještě jednou k českým deverbálním substantivům (Once more on Czech
deverbal nouns). In Čeština – univerzália a specifika 4, Zdena Hladká & Petr Karlík (eds),
13–23. Praha: Lidové noviny.
Kolářová, Veronika. 2010. Valence deverbativních substantiv v češtině (na materiálu substantiv s
dativní valencí) (Valency of deverbal nouns in Czech, with a special regard to nouns with
dative valency). Praha: Karolinum.
Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase Structure and
the Lexicon, Johan Rooryck & Laurie Zaring (eds), 109–137. Dordrecht: Kluwer. DOI:
10.1007/978-94-015-8617-7_5
Landau, Idan. 2010. The explicit syntax of implicit arguments. Linguistic Inquiry 41(3): 357–388.
DOI: 10.1162/LING_a_00001
Marantz, Alec. 1993. Implications of asymmetries in double object construction. In Theoretical
Aspects of Bantu Grammar, Sam Mchombo (ed.), 113–150. Stanford CA: CSLI.
Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy
of your own lexicon. In U. Penn Working Papers in Linguistics 4(2): Proceedings of PLC 21,
Alexis Dimitriadis (ed.), 201–225. Philadelphia PA: Penn Linguistics Club.
Marantz, Alec. 2007. Phases and Words. In Phases in the Theory of Grammar, Sook-Hee Choe.
(ed.), 196–222. Seoul: Dong In.
McFadden, Thomas. 2004. The Position of Morphological Case in the Derivation: A Study on
the Syntax–Morphology Interface. Ph.D. dissertation, University of Pennsylvania.
Novotný, Jiří. 1980. Valence dějových substantiv v češtině (Valency of event nouns in Czech)
[Sborník pedagogické fakulty v Ústí nad Labem]. Prague: SPN.
Panevová, Jarmila. 1974. On verbal frames in Functional Generative Description. Prague Bul-
letin of Mathematical Linguistics 22, 3–40.
Panevová, Jarmila. 2000. Poznámky k valenci podstatných jmen (Notes on the valency of
nouns). In Čeština – univerzália a specifika 2, Zdena Hladká & Petr Karlík (eds), 173–180.
Brno: Masarykova Univerzita.
Piťha, Petr. 1992. Posesívní vztah v češtině (Possessive Relation in Czech). Prague: Aved.
Procházková, Věra. 2006. Argument Structure of Czech Event Nominals. M.Phil. thesis, Uni-
versity of Tromsø.
Pylkkänen, Liina. 2002. Introducing Arguments. Ph.D. dissertation, MIT.
Ramchand, Gillian. 2004. Time and the event: The semantics of Russian prefixes. Nordlyd.
Tromsø University Working Papers on Language and Linguistics 32(2): 323–361.
Roeper, Thomas & van Hout, Angeliek. 1999. The impact of nominalization on passive, -able
and middle: Burzio’s generalization and feature-movement in the lexicon. In MITWPL 35:
Papers from the UPenn/MIT Roundtable on the Lexicon, Liina Pylkkänen, Angeliek van
Hout & Heidi Harley (eds), 185–211. Cambridge MA: MIT.
Romanova, Eugenia. 2004. Superlexical versus lexical prefixes. Nordlyd. Tromsø University
Working Papers on Language and Linguistics 32(2): 255–278.
Smith, Carlota S. 1991. The Parameter of Aspect. Dordrecht: Kluwer. DOI: 10.1007/978-94-015
-7911-7
Svenonius, Peter. 2004. Slavic prefixes inside and outside VP. Nordlyd. Tromsø University Work-
ing papers on Language and Linguistics 32(2): 205–253.
Veselovská, Ludmila. 2001. K analýze českých deverbálních substantiv (On the analysis of Czech
deverbal nouns). In Čeština – univerzália a specifika 3, Zdena Hladká & Petr Karlík (eds),
11–27. Brno: Masarykova Univerzita.
Zucchi, Alessandro. 1989. The Language of Propositions and Events: Issues in the Syntax and
Semantics of Nominalization. Ph.D. dissertation, University of Amherst.
chapter 5
A data-driven analysis of the structure type

‘man–nature relationship’ in Romanian*
Ana-Maria Barbu
Romanian Academy, Bucharest
The Romanian tri-nominal juxtaposition structure relație om – natură

‘man-nature relationship’ is cross-linguistically widespread and typical of a series
of relational nouns, such as agreement, interaction, and mixture, which can have a
“compound” expansion (Canada–U.S. agreement, parent–child interaction,
air–water mixture, etc.). Our analysis is twofold: we first examine the grammatical
relationship between relație ‘relationship’ and om – natură ‘man–nature’, and
second the construction om – natură. On the basis of data from a large Romanian
newspaper corpus, we show that the “compound” construction om – natură is
in fact a free phrase; we call it a Relational Coordination Construction (RCC). It
usually embodies valency complements of a relational noun, and it semantically
implies reciprocity. The analysis adopts a non-transformational, data-driven
perspective within the Construction Grammar framework.
1. Introduction
This chapter approaches the general problem of noun valency in Romanian by

concentrating on one specific issue. Noun valency is often analyzed analogously to
verb valency, especially in the case of deverbal nouns (see Panevová, this volume).1
* I express my gratitude to Olga Spevak for her numerous comments and suggestions. My
thanks also go to the two anonymous reviewers of this paper. All remaining errors are mine.
. In Romanian, usually, through nominalization the subject of a verb becomes the valency
element in the genitive, inflectionally expressed, of the corresponding deverbal noun. Alter-
natively it becomes a de către-PP ‘by’-PP. The direct object becomes the genitive or the de-PP
‘of ’-PP complement, while the indirect and oblique objects are kept as such: Ion taie pâinea cu
cuțitul ‘John cuts the bread with the knife’ > tăierea pâinii de către Ion cu cuțitul ‘the cutting of
the bread by John with the knife’.
 Ana-Maria Barbu
Nevertheless, a series of nouns – relational nouns – have also been identified

as nouns with valency that require complements, some of them without being
deverbal.
The literature devoted to this topic has focused on only a small group of non-
deverbal relational nouns. These are nouns expressing inter-human relationships
(kinship, friendship, neighborhood), such as father, brother, friend, and neighbor.
All the other relational nouns have been virtually ignored, although obviously
they are more numerous. Among this latter group are the typical relation(ship)
and also alliance, antagonism, asymmetry, etc.; that is, a series of nouns expressing
a symmetrical (or reciprocal) relationship between at least two elements. Such a
relational noun always denotes a state of fact (not a class of individuals).2 In this
chapter, we argue that the elements involved in the relationship expressed by such
a noun represent its valency structure.
In Romanian, the relationship between two nominals is usually expressed
by the genitive case: lucrarea studentului (work-art student-art.gen, ‘the stu-
dent’s work’), or by a prepositional phrase: ușă de casă (door of house, ‘door of
a house’). Much less frequently, nominal dependency can also be expressed by
juxtaposition (also called apposition), when a noun in the nominative is simply
attached to a head noun: sintagmă cu centru pronume (phrase with head pro-
noun, ‘a phrase with a pronoun head’). This is the case of the structure discussed
here. The dependent elements are nouns or noun phrases in the nominative
juxtaposed with one another. Further, they are appositively attached to a noun
head, and the structure thus obtained is a phrase with three members N1 N2 N3
(a tri-nominal structure). A typical example in this respect is relaţie om – natură
‘man–nature relationship’, where relaţie = N1 om = N2 natură = N3 (man = N2
nature = N3 relationship = N1).
This type of structure appears extensively in newspapers and scientific litera-
ture. It is preferred for its concision. In our research, we mainly used the Internet
and the corpus ZiareRom, a corpus of written newspapers.3 It is worth pointing
out that there are several ways of spelling the construction N2 N3: with a slash, a
hyphen, a dash, or a blank space. Here are some examples of the structure N1 N2
N3 extracted from this corpus:
. This feature distinguishes a relational noun such as father from those like relationship;
compare the entity X is a father with *the entity X is a relationship. The relational noun father
can also be used as a sortal noun, but relationship cannot. For the distinction sortal vs. rela-
tional nouns, see Partee & Borschev (2012: 446).
. For a detailed description of the corpus, see 〈www.lingv.ro/Resurse lingvistice〉.
Chapter 5. Man–nature relationship 
(1) a. Tratăm adesea cuprins-ul relaţi-ei om –

treat.1pl often scope-art.acc relationship-art.gen man-nom.sg
natură la Literatură.
nature-nom.sg at literature-acc.sg
‘We often treat the scope of the man–nature relationship in Literature
class.’ (Jurn. 15/04/11)
b. Trebuie să țină cont de un nou raport
must-3sg to take-sbvj.3sg account of a new ratio-acc
adâncime / lățime.
depth-nom breadth-nom
‘It must take into account a new depth/breadth ratio.’ (EvZ 23/07/07)
c. Combinați-a alcool – medicament-e
combination-art.nom alcohol-nom drug-nom.pl
antidepresiv-e este foarte des întâlnită
antidepressive-nom.pl is very often encountered
‘The alcohol–antidepressive drugs combination is encountered very
often.’ (EvZ 03/10/07)
In previous studies of relational nouns, few references to the above nominal con-
struction are found. The dominant point of view treats the sequence of arguments
N2 N3 (for instance man–nature in (1a)) as a compound noun. As for the syntactic
function of this sequence, we have found references in Noailly (to appear), where
the compound noun N2 N3 is considered the adjunct of N1. Likewise, on a merely
semantic basis, Olsen (2001 and 2004) treats the sequence N2 N3 as a complex
argument whose parts stand in the ‘between’ relation to the head N1.
In this chapter, we first argue that N2 and N3 form not a compound noun
but a phrase of coordinated elements of a special type, called here Relational
Coordination Construction (hereafter RCC). In this construction, components
N2 and N3 preserve their morphological, syntactic and semantic individual-
ity. In Section 3, arguments are supplied that N2 and N3 should be considered
complements of N1. Using data in the corpus, in Section 4 some possible
realizations of the arguments of relational nouns (that is, subcategorization
frames) are given; RCC is also characterized in detail. Upon deeper examination
it appears that RCC, as an independent phrase, is, in most cases, a subcategoriza-
tion frame for relational nouns, but in some cases it can also be an adjunct. In
Section 5, a formal representation of RCCs is proposed within the framework
of Construction Grammar, which, in our opinion, best reflects the properties
of the corpus data. Finally, in the last section we put forward some conclusions
about the tri-nominal juxtapositional structure, which we consider a specific
realization of the relational noun arguments.
 Ana-Maria Barbu
2. Relational Coordination Construction (RCC)
The first step in approaching a structure headed by a relational noun (for

example, relaţie ‘relationship’ in the example relaţia om – natură ‘the man–nature
relationship’ is to question the status of the sequence N2 N3. Noailly (to appear)
analyses the French examples un conflit parents-enfant ‘a parents–child c onflict’,
le malentendu mère-fille ‘mother–daughter misunderstanding’, l’alternative
privatisation-étatisation ‘privatization–nationalization alternative’, la cohésion
verbe-nom ‘verb–noun cohesion’ as follows: “Les formes du type parents-enfants
sont des mots composés [emphasis added] du même type que histoire-géographie,
dénotant une entité plurielle, faite de plusieurs parties.”4 Olsen (2004: 27) in turn
considers N2 and N3 to be “copulatives occurring as front forms of more complex
compound stems.” In other words, the whole structure, for instance man–nature
relationship, is a compound in which man and nature are word parts contributing
to the meaning of the whole.
In the absence of any argument, however, it is very difficult to accept that
parents-enfants (in un conflit parents-enfants) and om – natură (in relația om –
natură) are lexical (or morphological) compounds.5 In what follows, we bring into
discussion some arguments which prove that the sequence N2 N3 is not a com-
pound but a free construction (or phrase) without any lexical or morphological
cohesion. This position is endorsed by tests elaborated by Bouvier (2000). His tests
are precisely designed to identify the properties that help to distinguish lexical
compounds from phrases.
According to Bouvier, a defining property of a lexical compound is the fact
that it represents a new and unique concept obtained through lexicalization
(Bouvier 2000: 170). From this point of view, the construction om – natură ‘man-
nature’ does not represent a lexical compound not only because it does not express
a new and unique concept, but because it cannot actually be assigned any concept
at all, and moreover because no set of semantic features capable of defining an
entity – even plurality, according to Noailly – might be associated with it. Thus,
unlike constructions such as istorie-geografie ‘history-geography’, which is really
a plurality, whose parts are connected by means of the hypernym scientific disci-
pline, the parts of the construction om – natură ‘man–nature’ are not conceptually
connected. In the domain of semantic ontology, there would be no room for such
. “Forms such as parents-child are compound words of the same type as history-geography,
which denotes a plurality formed out of several elements.”
. See Mathieu-Colas (1995) and Gushchina (2008), who also believe, without explicit argu-
ments, that such constructions are not compounds.
Chapter 5. Man–nature relationship 
a construction. The immediate consequence of this fact is that if one wished to

introduce the construction in a dictionary, like any other lexical compound, no
definition of its meaning would be possible. In short, this construction violates
Olsen’s Principle of Ontological Coherence, according to which a complex concept
as the denotation of a morphological object should pick out a coherent individual
from one of the domains of individuals (Olsen 2004: 19).
On the other hand, as Bouvier (2000: 170) notes, a lexical compound does not
preserve its meaning if one of its elements is replaced with a synonym. This is not
the case with the constructions under scrutiny here. In the example (relația) om
– natură ‘the man–nature (relationship)’, om ‘man’ may be replaced by the phrase
ființă umană ‘human being’, the result being (relația) ființă umană – natură ‘the
human being–nature (relationship)’ with no consequent change in meaning. This
shows again that what we are dealing with here is not a lexical compound.
Finally, the type of construction analyzed here differs from lexical compounds
in morphology. Thus, unlike lexical units, in the case of om – natură ‘man–nature’
it is impossible to determine the grammatical gender. In Romanian, om is mascu-
line and natură is feminine and neither of them imposes its gender on the whole
sequence. At the same time, the parts of the construction may have independent
inflection (with slight modification of meaning):
(2) (relați-a) părinț-i – copil / părinț-i –
(relationship-art.nom) parent-nom.pl child-nom.sg parent-nom.pl
copi-i / părint-e – copi-i /
child-nom.pl parent-nom.sg child-nom.sg
‘the relationship between parent(s) and child(ren)’
Moreover, the elements of the construction may appear in inverted order in the
same context (3).
(3) Am urmărit una, televizată, despre relația
aux watched-1sg.pst one televised about relationship
jurnaliști-scriitori și scriitori-jurnaliști.6
journalists-writers and writers-journalists
‘I watched a TV debate about reciprocal relationships between writers and
journalists.’ (Ziua 14/06/07)
All of this shows a certain morphological independence which is not characteristic

of the elements of a lexical compound.
. In this example, the relationship is considered to be oriented from journalists toward
writers and from writers toward journalists, respectively.
 Ana-Maria Barbu
Further evidence that N2 N3 (in a relational structure N1 N2 N3) is a phrase

comes from the following facts:
1. N2 and N3 have distinct referents. Proper names (of persons, political parties,
organizations, etc.), which are very frequent in such a construction, illustrate
this property very well: relația NATO – Uniunea Europeană ‘the NATO–
European Union relationship’ (Ziua 01/02/07).
2. N2 and N3 may have their own dependents (4). Bouvier (2000: 167) uses
this feature, too, as evidence for rejecting the idea that they form a lexical
compound.
(4) a. cercetarea experimentală a interacțiunii roată cu pneu –
investigation experimental of interaction wheel with tire
cale nedeformabilă de rulare
track undeformable of traffic
‘the experimental investigation of the interaction between wheel with
tire and undeformable traffic track’ (Internet, 20.06.2012)
b. relaţia drept-urile om-ului – mass media –
relation right-art.nom.pl human-art.gen.sg mass media
societatea civilă
society-art.nom civil
‘the human rights – mass media – the civil society relationship’
(Euractiv-ro 05/04/07)
Note that sequences are not restricted to two elements N2 and N3 but can contain
three or more nouns, as in (4b).
Having accepted that the sequence N2 N3 does not represent a lexical com-
pound but a free phrase, we must still determine what kind of phrase the construc-
tion actually is.
Coordination as a type of relationship between the members of a (bi-nomi-
nal) “compound” is widespread in the literature. For instance, Noailly (1990: 65)
identifies a logical relationship of coordination between two nouns: when they
have equal importance in the structure which they form by juxtaposition. Exam-
ples clearly illustrating this relationship, in Noailly’s opinion, are l’ambivalence
violence-indolence ‘violence–indolence ambivalence’ or la cohésion verbe-nom
‘verb–noun cohesion’: that is, the complex structure N1 (N2 + N3) where the
coordinate nouns N2 + N3 have the same head, the relational noun itself ( Noailly
1990: 84). The same type of coordination structure is mentioned in Barbaud (1971,
apud Gushchina (2008)), Bauer (2008), and Olsen (2001 and 2004).
However, Olsen (2001: 299) adds that the characterization of N2 N3 as
members of a coordination construction is not enough, even if coordination is
Chapter 5. Man–nature relationship 
nderstood as being copulative – that is, a construction creating a plurality. Olsen

u
correctly observes that conceiving of this construction as a simple operation of
plural formation does not capture the correct semantic representation of the whole
structure (i.e. N1 (N2 + N3)). Besides, this characterization actually fails to explain
the ill-formedness of the structures obtained just by expressing the coordinating
conjunction: *relația om și natură ‘the man and nature relationship’, *ambivalența
violență și indolență, ‘the violence and indolence ambivalence’, or *coeziunea verb
și nume ‘the verb and noun cohesion’.
What is essential in the analysis of these coordination-based constructions
are the facts that, as Noailly noticed, they have the same noun as their head, and
also that this noun is relational in its nature. A relationship obligatorily involves
at least two elements – that is, a plurality. The plurality may be expressed in two
different ways:
1. by morphological means – mainly, the plural number:

(5) relația dintre oam-eni / *om
relationship between man-pl man-sg
‘the relationship between men / *man’
2. by syntactic means:
a. conjunctive copular coordination: relația dintre om *(și natură) ‘the
relationship between man *(and nature)’;
b. preposition phrase headed by associative preposition: relația omului *(cu
natura) ‘the relationship of man *(with nature)’.
At the same time, these examples show that relational nouns involve not
merely plurality, but also reciprocity.7 Reciprocity is expressed by the reciprocal
prepositions dintre (with the variant între) ‘between’ and cu ‘with’ that are gram-
matically and semantically selected by relational nouns. One may thus conclude
that p lurality and reciprocity are two semantic features simultaneously imposed
by relational nouns. As a consequence, in the tri-nominal structure discussed here
(e.g. relația om – natură, ‘the man–nature relationship’), the construction N2 N3,
in the absence of any preposition, has to incorporate both of these features, plural-
ity and reciprocity. Therefore this construction represents a coordination phrase
which is not merely copulative but also involves a reciprocal (or relational) status
of its conjuncts. We term this phrase a Relational Coordination Construction.
. Reciprocity is meant to define a relation between equipotent members.

 Ana-Maria Barbu
In Section 3 we provide a detailed characterization of this construction and

the tri-nominal structure discussed here using, evidence from the corpus.
The next step of the analysis is to establish the syntactic function of N2 and N3
in the triadic structure headed by N1.
3. N2 and N3 as complements of N1
As already mentioned, Noailly (to appear) puts forward the view that the sequence
N2 N3 is a compound word with the syntactic function of adjunct of N1. The com-
pound makes implicit reference to a relationship expressed by a preposition. This
structure (N1 N2 N3) patterns like binominal French noun phrases such as un pro-
blème cheveux or un programme enfants, where the prepositions expressing the
relationship between the two nouns are avec ‘with’: un problème avec les cheveux ‘a
problem with the hair’ and pour ‘for’: un programme pour les enfants ‘a program for
the children’, respectively. In the same vein, in a structure such as la cohesion verbe-
nom ‘verb–noun cohesion’, the compound verbe-nom ‘verb–noun’ establishes a
relationship with cohesion which may be glossed by the preposition entre ‘between’:
la cohesion entre un verbe et un nom ‘the cohesion between a verb and a noun’.
We showed in the previous section that the sequence N2 N3 (like verb–noun
in the above example) is not a lexical compound. We now present evidence that
these nominals are not adjuncts but, in most cases, complements of N1, and that
they occupy argument positions. Additionally, we try to prove that the sequence
N2 N3 represents an alternative argument realization to the between-PP. The facts
that support our interpretation are the following:
1. The arguments of a relational noun must be known.

2. The complements of a verb become the complements of the corresponding
relational deverbal noun.
3. The arguments of a relational noun may have multiple realizations.
4. The arguments of a relational noun may be involved in the phenomenon of
argument-sharing.
3.1 The mandatory realization of the arguments

From a semantic point of view, relational nouns are predicates with two places
or more: relationship′(x, y), mix′(x, y), combination′(x, y), conflict′(x, y),
dialog′(x, y), etc. Knowing the arguments is compulsory.8 For example, (6) is
. Olsen (2001: 299) refers to these arguments as “semantic requirements of certain types of
relational heads”, while Bauer (2008: 12) calls them co-participants.
Chapter 5. Man–nature relationship 
bad both semantically and pragmatically, if the universe of the discourse does
not make available the elements which the fair relationship refers to:
(6) #S-a stabilit o relație echitabilă.
pass-aux establish-3sg.pst a relationship fair
‘A fair relationship has been established.’
Example (6) becomes acceptable if one adds an RCC as in (7):

(7) S-a stabilit o relație om – natură echitabilă.
pass-aux establish-3sg.pst a relationship man nature fair
‘A fair man–nature relationship has been established.’
However, we may note that even if the arguments are not overtly expressed, they
must be recoverable from the (discourse or knowledge of the world) context:
(8) Omul a înțeles că trebuie să protejeze natura.
man aux understand-3sg.pst that must sbvj-protect-3sg nature
Astfel, s–a stabilit o relație echitabilă.
thus pass-aux establish-3sg.pst a relationship fair
‘Man understood that he must protect nature. Thus, a fair relationship has
been established.’
This situation is examined in Herbst (1999), who foregrounds the so-called contex-
tually optional complements necessary to the syntactic realization of the arguments
that are obligatorily recoverable from the context. Herbst makes a distinction
between optional complements, whose omission yields an indefinite meaning – as
in the example John is reading, equivalent to John is reading something – and con-
textually optional complements, whose omission is allowed only if the complement
in question can be retrieved from the context, as in the example John objects / is
objecting which is meaningful if it is already known what John is objecting to.
Relational nouns, therefore, align with predicates that have contextually
optional complements.
3.2 The complements of derived relational nouns

Another argument in support of the fact that N2 and N3 are complements of the
relational noun N1 is offered by relational nouns which have in their lexical family
verbs with uncontroversial complements. For example, Guțu-Romalo (2005: 416)
says that prepositional complements (or oblique objects) are verb phrase com-
ponents which have semantic roles. Verbs selecting an obligatory prepositional
phrase are, for instance, “symmetric” verbs (those having the semantic feature [+
Reciprocal]), such as a se asemăna cu ‘to resemble with’, a colabora cu ‘to collabo-
rate with’, a comunica cu ‘to communicate with’, a rivaliza cu ‘to rival with’, etc. as
exemplified in Guțu-Romalo (2005: 416). So in (9a) Romania and (cu) Bulgaria
 Ana-Maria Barbu
are the arguments (expressed as the subject and the oblique object, respectively)
of the verb colaborează ‘collaborates’. If the verb is nominalized, the corresponding
noun has the same arguments, Romania and (cu) Bulgaria, expressed by a genitive
and the same oblique object (9b). These arguments remain the arguments of the
nominal head even if they are expressed by an RCC, as in (9c).
(9) a. România colaborează cu Bulgaria în turism.

Romania collaborates with Bulgaria in tourism.
b. Există o colaborare a Români-ei cu Bulgaria
exists a collaboration of Romania-gen with Bulgaria-acc
în turism.
in tourism
‘Collaboration of Romania with Bulgaria in tourism exists.’
c. Există o colaborare România – Bulgaria în turism.
exists a collaboration România-nom Bulgaria-nom in tourism
‘Romania–Bulgaria collaboration in tourism exists.’
Example (10) is similar, where the noun similaritate ‘similarity’ is derived from the
adjective similar ‘similar’ and inherits its complements:
(10) a. Om-ul este similar cu animal-ele.

human-art.nom.sg is similar to animal-art.acc.pl
‘Humans are similar to animals.’
b. Similaritate-a om-ului cu animal-ele
similarity-art.nom human-art.gen.sg to animal-art.acc.pl
este evidentă.
is obvious
‘The similarity of humans to animals is obvious.’
c. Similaritate-a om – animal
similarity-art.nom human-nom.sg animal-nom.sg
este evidentă.
is obvious.
‘The human–animal similarity is obvious.’
We may therefore extrapolate and say that relational nouns marked [+ Reciprocal],
such as relație ‘relationship’, armonie ‘harmony’, and simetrie ‘symmetry’, require
the same types of complements, even if they do not come from verbs or adjectives.
In essence, then, from the point of view of the present discussion, these predicates
are of the same nature as colaborare ‘collaboration’ or similaritate ‘similarity’. That
is, accepting that the construction Romania–Bulgaria expresses the complements
Chapter 5. Man–nature relationship 
of collaboration in (9c) triggers the fact that the same construction expresses the
complements of relationship in a similar example, such as A Romania–Bulgaria
relationship in tourism exists, as well.
3.3 Multiple argument realizations

Levin (1993) and Meyers et al. (1996) observe that argument phrases participating
in valency alternations are usually complements, not adjuncts. In (11), the argu-
ments of the verb a colabora ‘to collaborate’ illustrate two different realizations: as
multiple subjects in (11a), and as a subject and prepositional object in (11b):
(11) a. Statele Unite și Rusia colaborează.
the United States and Russia collaborate
b. Statele Unite colaborează cu Rusia.
the United States collaborates with Russia
Multiple argument realizations also appear in the case of relational nouns, both
derived (as already reflected in (9) and (10) above) and underived, as may be seen
in (12) for the noun relație ‘relationship’:
(12) a. Relați-a Stat-elor Unit-e cu Rusia
relationship-art.nom State-art.gen.pl United-pl with Russia-acc
este stabilă.
is stable
‘The relationship of the United States with Russia is stable.’
b. Relația dintre Stat-ele Unit-e și Rusia
relationship between State-art.acc.pl United-pl and Russia-acc
este stabilă.
is stable
‘The relationship between the United States and Russia is stable.’
c. Relația Stat-ele Unit-e – Rusia
relationship State-art.nom.pl United-pl Russia-nom
este stabilă.
is stable
‘The United States–Russia relationship is stable.’
This example displays three different subcategorization frames for the noun rela-
tionship: genitive complement and with-PP (12a); between-PP (12b); RCC (12c).
These frames are common to the majority of relational nouns.
It is important to highlight that RCC represents just another argument realiza-
tion, equivalent to between-PP, and not derived from it by preposition omission.
 Ana-Maria Barbu
Olsen (2004: 28) and Noailly (to appear) make a connection between the sequence
N2 N3 (that we call RCC) and between-PP, by rendering RCC as between-PP more
explicitly (e.g. lawyer–client relationship > relationship between a lawyer and a cli-
ent). Mathieu-Colas (1995: 163) in turn interprets RCC as an ellipsis and proposes
more possible paraphrases.
In the following, we show that RCC is not a result of the ellipsis of the
between-PP.
First, the hypothesis of the omission of the preposition dintre ‘between’ is not
supported by morphological data: in Romanian, the presence of the preposition
requires the accusative (13a), whereas RCC requires nominative forms of the argu-
ment nouns (13b) and excludes accusative ones (13c). Note that in a true ellipsis,
the morphological case is preserved (13d).
(13) a. relația dintre mine și tine

relationship between me-acc and you-acc
‘the relationship between me and you’
b. relația eu – tu
relationship I-nom you-nom
‘the I – you relationship’
c. *relația mine – tine
relationship me-acc you-acc
‘the me – you relationship’
d. Ion a dat o floare Mariei, iar Radu __ Danei.
John gave a flower Mary-dat and Radu Dana-dat
‘John gave a flower to Mary, and Radu to Dana.’
The omission hypothesis, therefore, has to explain why a simple deletion involves
systematic case modification of the complements.
Second, an explanation is needed of why the construction in (12c) comes
from (12b) through preposition deletion, and not from (12a). And finally, there
is no answer to the legitimate question of why the preposition dintre ‘between’
may not be elided in the structure with a locative dintre: pomul dintre casă și
garaj ‘the tree between the house and garage’ / *pomul casă – garaj ‘*the house–
garage tree’.
All these problems disappear if it is admitted that there is a certain type of
phrase, identified here as a Relational Coordination Construction, which is not
reducible to another “more basic” one. This phrase is composed of nominative
NPs (see (13b)) that represent an alternative realization of the arguments of a rela-
tional noun.
Chapter 5. Man–nature relationship 
3.4 Argument sharing

Another reason for treating N2 N3 as a construction of complements of N1 and not
as an adjunct is the fact that the arguments of the relational noun may be inherited
by its governor.
In general, a constituent may have arguments which are shared with the head
of the projection it belongs to. For example, in (14a) the subject of the verb a citi
‘to read’ is also the subject of its governor, the verb a încerca ‘to try’. This is the
phenomenon of control, which is marked by coindexation (indicated by i in the
examples below). In (14b) the verb a putea ‘can, may’ shares the direct object with
the verb complement a contacta ‘to contact’. This time, we are dealing with a phe-
nomenon of raising.
(14) a. Ioni încearcă [ _i să citească].
John tries to read
‘John tries to read.’
b. Ion îli putea [contacta _i]
John him.acc could contact
‘John could contact him.’
Relational nouns too can exhibit cases of shared arguments, but this sharing is not
mandatory as in cases of control and raising. In (15), for example, the comple-
ments of the noun relație ‘relationship’ are at the same time the multiple subjects
of its governor, the verb a stabili ‘to establish’:
(15) [Omul și natura]i stabilesc o armonioasă [relație _i].
man and nature establish a harmonious relationship
What proves the existence of shared arguments is the ill-formedness of (16). (16)
does not specify with whom (or what) nature establishes a harmonious relation-
ship. At the same time, it ought to be noted that it is not the verb a stabili ‘to
establish’ that imposes multiple subjects in (16), because this verb may also have
a subject in the singular. So, the multiple subjects are in fact required by the noun
relație. And this means that the structure is ill-formed because the valency of the
noun relație is not saturated, even if the matrix-verb valency is saturated with the
subject and the direct object.
(16) *Natura stabilește o armonioasă relație.
nature establishes a harmonious relationship
For a better understanding of the manner in which argument structures of rela-

tional nouns and, especially, RCCs are expressed, we have developed a detailed
 Ana-Maria Barbu
analysis based on data in the ZiareRom corpus (we also resorted to Internet lin-
guistic samples). The results of the analysis are given below.
4. Further corpus data
4.1 Argument realization structures

Given that our corpus (ZiareRom) is large (over 80 million words), the analysis had
to use both machine and manual processing. First, we built an inventory of rela-
tional nouns found in the corpus. In order to achieve this goal, we first extracted
all the nouns followed by prepositional phrases headed by the preposition dintre
‘between’-PP. We then kept in our inventory only the nouns with a between-PP
expressing reciprocity and we identified 224 relational nouns.9 We further checked
how many of them occurred in at least one context with RCC. We thus obtained a
list of 173 nouns, given in the Appendix.
In what follows, our analysis concentrates on the context of the noun relație
‘relationship’, which is neutral from a semantic point of view and well represented
in the corpus (13,358 occurrences). We focus on the argument realizations of this
noun and in particular on the argument realization through RCC. The results are
summarized in Table 1.
Table 1. Argument realization structures of the noun relație ‘relationship’

Subcategorization cu-PP dintre-PP RCC Others
frame (with-PP) (between-PP) (eg. man-nature)
No. of occurrences 6,015 2,146 529 4,668

Rate 45% 16% 4% 35%
As Table 1 shows, the noun relationship has several subcategorization frames.

By far the most frequent structure is that involving cu-PP ‘with’-PP. A typical
example is: relația mea cu familia ‘my relationship with my family’. Another is
given in (17).
(17) Un medic chiar are o relație cu pacient-ul.
a physician really has a relationship with patient-art
. It is worth noting that in Romanian, as in English, the preposition dintre ‘between’ also
denotes a position (as in numărul dintre doi și patru ‘the number between two and four’) or
partition (as in alegerea dintre Dumnezeu și Diavol ‘the choice between God and the Devil’).
Chapter 5. Man–nature relationship 
The structure in (17), where one argument of the predicate relationship is shared
with the subject argument of its governor, probably satisfies discourse require-
ments the best and therefore is the most frequent.
The examples in (18) illustrate the subcategorization frame with dintre-PP
‘between’-PP:
(18) a. relația dintre cele două state
relationship between the two states
b. relațiile dintre Tokyo și Phenian
relations between Tokyo and Pyongyang
Under the “Others” heading we list the cases displaying a plural possessive adjec-
tive, their relationship; a genitive plural noun phrase, the relationship of the two
states; a compound “group” adjective, American–Russian relations; an inter-
derived adjective, inter-human relations; or simply the cases where the noun rela-
tionship lacks arguments in the same sentence, because they are recoverable from
the larger context.
It is worth mentioning that the relational nouns grouped by Olsen (2004: 299)
under the heading Collection Type, such as combination and mixture, in our cor-
pus also select a between-PP. This fact shows that they are ordinary relational
nouns with reciprocal meaning. However, their meanings express a relationship
from which new entities result, so that the elements participating in the relation
become parts of the resulting entity. Due to this fact, collective relational nouns
can select a de-PP ‘of’-PP expressing the content of the entity (19).
(19) a. o combinație între alb și negru / de alb și negru
a combination between white and black / of white and black
b. amestec-ul dintre sare și apă / de sare și apă
mixture-art between salt and water / of salt and water
Note that with nouns like dialogue, interaction, relationship, etc., in which par-
ticipants keep their individuality, the of-PP realization is not allowed: *dialogul
de mamă și fiică ‘*the dialogue of mother and daughter’, *interacțiunea de asfalt și
pneu ‘*the interaction of asphalt and tire’, or *relația de SUA și Rusia ‘*the relation-
ship of the USA and Russia’.
In what follows, we pay special attention to evidence from the corpus regard-
ing the RCC.
The elements of the RCC are of two major types: proper nouns (simple or
compound), relația Cotroceni – Palatul Victoria ‘the Cotroceni–Victoria Palace
relationship’, and bare nouns (that is, nouns without determiner or dependents)
relația antrenor – jucător ‘the coach–player relationship’. However, there are also
cases where RCC elements are nouns with dependents and determiners. In this
 Ana-Maria Barbu
situation, the nouns are always generic: relația autor – locul faptei ‘the author–
crime scene relationship’ (see also (4) above).
It is worth mentioning that the determiner allowed to occur in RCC seems
to be exclusively the definite article (-ul ‘the’). This fact strengthens the generic
interpretation of this type of argument realization, unlike the other realization
types, such as those involving between-PP or with-PP, which are used precisely for
specific individuals. This fact could also explain the relatively low percentage (4%,
Table 1) of the use of RCC.
Despite these quite severe restrictions in use – proper nouns, bare nouns, defi-
nite article – RCC allows users’ creativity sometimes to be manifest.10
4.2 On RCC independence

In the following, we attempt to answer the question whether RCC is to some extent
independent by looking for evidence in our corpus. This question arises from
Olsen’s statement that the constituents N2 N3 saturate the argument position of
the head at the morphological level (2004: 28).11 In other words, the structure N1
N2 N3 forms a complex morphological compound.
The following data in the corpus cast doubt on the adequacy of treating tri-
nominal structure as a morphological unit.
1. The head noun may have an adjunct placed between it and the RCC: o relație
armonioasă PNL-PSD (‘a harmonious relationship between the PNL and the
PSD parties’) (7plus 03/04/07).
2. There are coordinate RCCs, as we have seen in (3), and also coordinate heads:
comunicarea și cooperarea cluburi-FRF-AJF (‘the communication and coop-
eration between the FRF, AJF, and clubs’).
3. The appositive status of RCC can be explicitly emphasized:
(20) o relație de tip protector – protejat
a relationship of type protector – protégé
. See for example “Combinația sacul cu bani – Steaua – cruci mari făcute cu dreapta de
față cu ziariștii – câțiva consilieri pricepuți – charisma celui care știe cum se mulge o oaie pare,
în anul 2007, formula ideală pentru un prezidențiabil cert” (EvZ 30/03/07). ‘The combination
bag full of money – Steaua football team – the making of the sign of cross in the presence of the
press – some skilled advisors – charisma of the one who knows how to milk a sheep seems in
2007 the best device for a successful candidate for the presidency.’
. Bernard Fradin also considers that the whole structure N1 N2 N3 forms a construction
and that N2 N3 cannot be used independently (p.c.).
Chapter 5. Man–nature relationship 
All these examples show that the head and the RCC in a tri-nominal structure
can independently undergo syntactic operations like adjunction, coordination or
appositive expansion. This fact invalidates the assumption that the structure is a
morphological unit.
Special attention should be paid to example (20). It suggests that RCC is not
always a valency complement: it can also be the adjunct of the relational noun.
This hypothesis is supported by examples like (21).
(21) Avem o relație jucător - antrenor.
have.1pl a relationship player coach
‘We have a player–coach relationship.’
It is obvious that the elements participating in the relationship are referred to by

we, which is inferable from the verb avem. In this case, the construction player–
coach is an adjunct showing the kind of relationship. The adjunct status becomes
more obvious when the noun relationship is modified by a de-PP ‘of ’-PP, as in (22).
(22) a. Eu am avut cu el o relație de patron – manager.
I had with him a relationship of owner manager
‘I had a relationship of owner–manager with him.’
b. Politicienii au o relație de iubire-ură cu presa.
politicians have a relationship of love hate with press
‘The politicians have a relationship of love–hate with the press.’
The PP headed by de ‘of ’ in (22) seems to specify the type of relationship (or its
property), not the arguments of the relationship.12 In this respect, the above struc-
tures follow the pattern “I had a relationship of love/of cooperation/of mutual
respect … with him”.
In this situation the of-PP is an adjunct. At the same time, the constructions
patron-manager ‘owner–manager’ in (22a) and iubire-ură ‘love–hate’ in (22b) may
be considered RCCs, too. The former denotes the relationship between an owner
and a manager, as the following paraphrase shows: “I had a relationship with him
like the one between an owner and a manager”. The latter is special because its
meaning is not “The politicians have a relationship with the press like the one
between love and hate” but “The politicians have, with the press, a relationship
alternating or combining love and hate”. Thus, the RCC love–hate proves to have
its proper relational meaning, that of alternation or amalgamation.
We also recognize the possibility of associating an RCC with non-relational
nouns as an argument that RCC can independently have a relational meaning – thus
. The arguments are I and him in (22a), and the politicians and the press in (22b).
 Ana-Maria Barbu
expressing an underspecified type of relation. Mathieu-Colas (1995: 165) notes that

in these cases the articulation between the elements remains implied.
(23) și dus-ul la film e o activitate mamă– fiică

also go-ptcp-art to movie is an activity mother daughter
‘going to movie is also a mother-daughter activity’ (Cancan 17/04/11)
In (23), activity is not a relational noun, yet the speakers do understand that it
refers to an activity that involves the interaction between mother and daughter.
With respect to this fact, Olsen (2001: 300) says that, in examples such as mind–
body problem, cost–benefit analysis, Clinton–Lewinsky saga, the between relation
seems to be a conceptual inference induced by the meaning of the head (i.e. prob-
lem, analysis, saga). We, on the contrary, sustain the opposite idea, namely that the
conceptual inference of a relation is induced by the RCC itself.
From a syntactic point of view, the relative independence of RCC is proved by
the operation of topicalization, as in (24a) (a headline) and (24b).
(24) a. SUA – Rusia: pași de întâmpinare,

dar nu cu
USA – Russia: steps toward negotiations but not at
orice preț
any price (Finan. 10/03/09)
b. La ANRE, între 18 persoane există relații de rudenie, un
at ANRE between 18 persons exist relations of kinship a
tip de “famiglia” din energie: soț-soție,
type of famiglia in energy husband-wife
frate-soră, mamă-fiică.
brother-sister mother-daughter
‘At ANRE, there are 18 persons between whom kinship relationships

exist, a type of “famiglia” in the energy domain: husband–wife, brother–
sister, mother–daughter.’ (EvZ 06/06/2007)
On the basis of a similar example in French (Royal-Sarkozy: projet contre projet

‘Royal–Sarkozy: project versus project’), Gushchina (2008: 2514) holds that the
link between the two components (Royal and Sarkozy) expresses two semantic
nuances: coordination and complementation induced by the preposition contre
‘versus’. Just as this chapter does, she holds that this type of example seems to
represent the free juxtaposition of two independent NPs.
The independent meaning of RCC is sometimes hard to obtain if an appropri-
ate context is not provided. The constructions USA–Russia, mother–daughter, and
Greece–Russia: 1–0, are easy to understand because political, familial, and sports
relationships are common. Conversely, the construction man–fence is weird, so
Chapter 5. Man–nature relationship 
long as it is not placed in a context that specifies the relationship between a man
and a fence, such as in The man–fence contact triggers the house alarm, or Man–
fence: a dangerous contact (if the fence is electric).13 Besides, being a phrase, RCC
creates the necessary syntactic opportunities for expressions like “A star–mercy
relationship doesn’t exist”, where two completely different concepts (e.g. star and
mercy) are connected into a meaningful utterance.
In sum, assuming the independence of RCC should not imply that an RCC
can occur anywhere in a sentence. It obeys syntactic rules like any other phrase.
What we did was merely to point out that RCC exhibits some features supporting
its relative independence, namely:
a. the availability of syntactic operations like coordination, adjunct insertion,

and explicit classifying apposition;14
b. a proper meaning allowing RCC to be attached to nouns other than relational
nouns;
c. the ability to be topicalized;
d. the capacity to fulfill different grammatical functions such as complement or
adjunct of a relational noun, as well as complement of the preposition de ‘of’.
In Figure 1 we offer a schematic syntactic representation of the tri-nominal struc-

ture discussed above. The head is a relational noun, and the complements in an
RCC are expressed by bare nouns.
N′
Head Complements
RCC
N1
relație N2 N3
‘relationship’ antrenor jucător
‘coach’ ‘player’
Figure 1. Tri-nominal syntactic structure
. Note that a copulative coordination phrase, whose independence is uncontroversial, such
as man and fence is as weird as RCC, if a context is not provided.
. For the meaning of the term classifying apposition, see Bosque & Demonte (1999: 4779).
In short, a classifing apposition is one allowing the insertion of a word like type, style, class,
etc.: una pintura Renacimiento = una pintura (estilo) Renacimiento.
 Ana-Maria Barbu
In what follows, a formal representation of the structure under scrutiny here

is given. It is expressed within the framework of Construction Grammar, which, at
Fradin’s (2009) suggestion, best reflects the data from the corpus.
5. A Construction Grammar representation
5.1 Advantages of the constructionist approach

Construction Grammar (CG) is a usage-based grammar model. Its goal is to
account for all the constructions in a given language, regardless of the so-called
‘core–periphery’ distinction. In this respect, even if RCCs appear ‘marginal’
because of their relatively low frequency, e.g. 4% in Table 1, CG is able to account
for their existence in a simple yet principled way.
At the same time, CG provides the best solution to the oscillations in treating
these tri-nominal structures in the syntax or morphology or the interface between
them. Aiming for a satisfactory explanation of the data encountered in the corpus,
we argued in this paper for the free phrase nature of RCC and the whole structure
that contains it. Nevertheless, CG imposes no a priori choice in this respect (see
also the comments on this issue in Fried & Östman 2004: 17). In other words, con-
structions are representational and analytical devices appropriate for approaching
expressions belonging to any type of linguistic level: lexicon, morphology, syntax,
semantics, pragmatics, or discourse.
Another attractive aspect is that CG is surface-oriented. Unlike transfor-
mational grammar, which obtains structures from more basic ones through
derivational processes, CG approaches ‘unpredictable’ structures directly. For
example, instead of proposing a sophisticated succession of rules which could
derive RCC from a between-PP, CG is able to describe RCC by means of a specific
construction.
One of the main sections of the analysis of the tri-nominal structure focused
on showing that RCC is just one of several possible realizations of the relational-
noun arguments. This concerns Linking Theory, which occupies an important
place within the constructionist framework. From this point of view, it ought to
be noted that our approach follows that of Fried and Östman (2004: Chapter 2),
which differs from that of Goldberg (1995). An important difference between
the two versions is that Fried and Östman’s analysis allows the complements to
successively attach to the predicate that requires them, through ‘linking con-
structions’. Goldberg’s, on the other hand, avoids succession and treats both
the complements and their predicate in only one construction. As shown in
Chapter 5. Man–nature relationship 
S ection 4.2., we consider the representation of RCCs as constructions indepen-

dent of the relational noun to be more suitable. This representation allows RCC
to occur as a construction attached to a (non-)relational noun or, alternatively,
to occur freely.
Technically speaking, a construction is a form–meaning pair learned as such.
Any linguistic pattern, then, is a construction, as long as a given aspect regarding
its form or function cannot be predicted on the basis of its components or on the
basis of other already inventoried constructions. At the same time, though, if the
patterns are sufficiently frequent, they are themselves considered constructions,
even if they are fully predictable.
Constructions combine freely to form actual expressions (technically termed
constructs), insofar as they do not conflict with each other. The tool for construction
representation is the formal device of feature structures. The principle that governs
what set of feature structures is allowed to form a construction is the operation
of unification (see Goldberg 1995, 2006; Fried & Östman 2004). An example of
how a constructional device works is given in the following section, together with
the formal representation of the tri-nominal structure relație om – natură ‘man–
nature relationship’.
5.2 Formal representation of tri-nominal structure

In the constructional representation of the tri-nominal structure, we concentrate
on specific constructions. Only the relevant features of these constructions are
mentioned. In order to represent the autonomy of RCCs, we propose a general
construction, depicted in Figure 2. The following elements are relevant in it.
Semantic information (encoded as sem) is represented as an unspecified relation-
ship (R) between two or more equipotent items (for more than two items, we use
Kleene’s sign ‘+’).15 The items are tagged (#1, #2, etc). Multiple use of the same tag
is the mark for coindexing (that is, sharing the same element). Each tag denotes
an item displaying certain syntactic features (encoded as syn), namely the noun
category and the nominative case (cat n, case nom). The semantic information of
the items is left unspecified (…). The construction formed from these elements
is general enough to be a valency of a relational noun or an adjunct in other
environments.
. The conjunction of predicates R & EQUIPOTENT actually notes a relational coordination.
 Ana-Maria Barbu
RCC
sem R(#1 x, #2 y, …) & EQUIPOTENT (x, y,…)
+
#1 #2
syn [cat n, case nom] syn [cat n, case nom]
sem […] sem […]
Figure 2. Relational coordination construction
If an RCC is used as a valency frame (not an adjunct), it combines with the

linking construction Relational Noun Valency (RNV, see Figure 3, where val
stands for valency frame). RNV stipulates that a relational noun realizes its argu-
ments by means of complements with the grammatical function of object (gf obj).
The connection between the two constructions is achieved through the indication
Inherit RCC. This indication specifies the subcategorization frame of the relational
noun. In the present case, it specifies that it is an RCC (not a between-PP or with-
PP) that is appropriate to the relational noun.
RNV
Inherit RCC
syn [cat n]
sem R(#1 x, #2 y, ...)
val {#1[gf obj], #2[gf obj], …}
Figure 3. Relational noun valency construction
Figure 4 depicts the construction which corresponds to the noun relație

‘relationship’. This is a lexical construction (marked by the feature lxm, lexeme),
indicating that this noun is a predicate with two or more arguments.16 Argument
realization can be achieved through RCC (because the construction inherits
RNV, which inherits RCC). The representation also specifies the type of relation-
ship expressed by the noun (i.e. relationship′). This type will later override the
unspecified relation type of RCC.
. Note that there are some predicates that restrict the number of arguments. For instance,
two arguments define a pair relationship (pair′(x,y)), whereas three arguments define a trio
relationship (trio′(x, y, z)).
Chapter 5. Man–nature relationship 
RELAŢIE
Inherit RNV
syn [cat n]
sem relationship′ (#1 x, #2 y, …)
val {#1[…], #2 […], …}
lxm relaţie
Figure 4. Lexical construction of the noun relație
Before describing the construct that corresponds to the phrase relație om –
natură ‘man–nature relationship’, we also have to specify the device which
licenses expansions of the noun. This is the general instantiation construction NP
(Figure 5). The NP construction may get the desired constructs ‘projected’ by a
relational noun. The semantics of the NP construction incorporates the semantics
of its parts.17 This is expressed by the symbols ↓↑1 and ↓↑j. It is worth mention-
ing that the NP construction also allows the combination of a non-relational noun
with an RCC, as in the example mother–daughter activity. In such cases, RCC con-
tributes the semantics of the whole nominal phrase with an unspecified relation-
ship, whose members are equipotent.
NP
syn [head #3 [ ]]
sem ↓1 [ ], ↓j [ ]+
+
syn [head #3[cat n]] sem ↑j […]
sem ↑1 […]
Figure 5. Noun phrase construction
We are now in a position to supply the representation of the whole construct

relație om – natura ‘man–nature relationship’. This is done in Figure 6.
. As the reader can see, the semantic representations are not elaborated here due to their
complexity, which would require a specific study.
 Ana-Maria Barbu
syn [head #3 [cat n]]

sem ↓4 ↓5 relationship′ (#1 x, #2 y) & EQUIPOTENT (x, y)
syn [head #3[cat n] sem ↑5 R(#1 x, #2 y) & EQUIPOTENT (x, y)

sem ↑4 relationship′ (#1 x, #2 y)
#1 #2
val {#1 [gf obj], #2 [gf obj]} syn [cat n, case nom] syn [cat n, case nom]
lxm relație sem […] sem […]
lxm om lxm natură
Figure 6. The construct relație om – natură
The right-hand box incorporates the lexeme boxes (i.e. lexical construc-
tions) om and natură into RCC, by indicating that they are the arguments of
an unspecified equipotent relation and are expressed, at a syntactic level, by
nouns in the nominative. The left-hand box displays the superposition of sev-
eral constructions. The lexical construction relație describes the noun relație as
the semantic predicate relationship′ with (two or more) arguments that are
valency elements as well. This construction combines with the RNV, which adds
the information that the valency elements are objects (not obliques). Further,
the result combines with the RCC formed earlier, which instantiates the valency
elements (by also limiting them to two). The semantic predicate relationship′
overrides the unspecified relation R of the RCC (see the upper level) and takes
over its arguments by coindexings #1 and #2. Finally, all combine with the NP
construction, which imposes relație as the head of a noun phrase and combines
the semantic information of the NP components in a complex way only sketched
here (by ↓↑4 and ↓↑5).
6. Conclusion
This chapter has discussed a problem regarding the valency complements of sym-
metric relational nouns. We have shown that these types of nouns may express
their arguments by means of a juxtapositional construction of two nominals. We
termed this construction Relational Coordination Construction (RCC). The head
noun along with its RCC forms a tri-nominal structure of the type relație om –
natură ‘man–nature relationship’. It ought to be emphasized that RCC is only spe-
cific to nouns. Verbs cannot use this option of expressing their arguments (not
even those verbs belonging to the same lexical family as the relational deverbal
noun).
Chapter 5. Man–nature relationship 
Examination of the corpus data reveals, on the one hand, that within an RCC
nominals may be gather together which are morphologically, syntactically, or
semantically independent of one another. It is true, however, that bare nouns and
proper names are the most frequent. The RCC meaning is generic. Whenever it
occurs independently of a relational noun, RCC expresses an underspecified rela-
tionship between equipotent members.
On the other hand, the corpus shows that there are also some (few) situations
where RCCs are not valency complements of a relational noun, but adjuncts. In
this latter case RCCs make explicit the type of relationship of the noun. They thus
work as classifying appositions or prepositional adjuncts.
In the last section we sketched a formal representation within the construc-
tionist framework, which essentially accounts for all these data. We illustrated the
way the analysis works by modeling the structure relație om – natură.
References
Barbaud, Philippe. 1971. L’ambiguïté structurale du composé binominal. Cahiers de linguistique

1: 71–116.
Bauer, Laurie. 2008. Dvandvas. Word Structure 1(1): 1–18. DOI: 10.3366/E1750124508000044
Bosque, Ignacio & Demonte, Violeta. 1999. Gramática descriptiva de la lengua española
(Descriptive grammar of Spanish language). Madrid: Real Academia Española, ESPASA.
Bouvier, Yves-Ferdinand. 2000. Définir les composés par opposition aux syntagmes. GG@G
(Generative Grammar in Geneva) 1: 165–187. 〈http://www.unige.ch/lettres/linge/syntaxe/
journal/pdf_volume_one/article6_bouvier.pdf〉
Fried, Miriam & Östman, Jan-Ola. 2004. Construction Grammar in a Cross-Language Perspective
[Constructional Approaches to Language 2]. Amsterdam: John Benjamins. DOI: 10.1075/
cal.2
Fradin, Bernard. 2009. IE, Romance: French. In Oxford Handbook on Compounding, Rochelle
Lieber & Pavol Štekauer (eds), 417–435. Oxford: OUP.
Goldberg, Adele. 1995. Constructions. A Construction Grammar Approach to Argument Struc-
ture. Chicago IL: The University of Chicago Press.
Goldberg, Adele. 2006. Constructions at Work. The Nature of Generalization in Language.
Oxford: OUP.
Gushchina, Olga. 2008. Les constructions directes “substantif + substantif ”. Le cas du nom pro-
pre. In Actes du CMLF 2008 – Congrès Mondial de Linguistique Française, Jacques Durand,
Benoît Habert & Bernard Laks (eds), 2505–2518. Paris: Institut de Linguistique Française.
Guțu-Romalo, Valeria. 2005. Gramatica limbii române, Vol. II: Enunțul (Grammar of Romanian
Language, Vol. II: The Sentence). Bucharest: Editura Academiei Române.
Herbst, Thomas. 1999. English valency structures – a first sketch. Erfurt Electronic Studies in
English 6. 〈http://webdoc.gwdg.de/edoc/ia/eese/rahmen22.html〉
Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. Chicago
IL: The University of Chicago Press.
 Ana-Maria Barbu
Mathieu-Colas, Michel. 1995. Syntaxe du trait d’union: Structures complexes. Linguisticae Inves-
tigationes 19(1): 153–171. DOI: 10.1075/li.19.1.10mat
Meyers, Adam, Macleod, Catherine & Grishman, Ralph. 1996. Standardization of the comple-
ment adjunct distinction. In Euralex ‘96. Proceedings, Martin Gellerstam, Jerker Järborg,
Sven-Göran Malmgren, Kerstin Norén, Lena Rogström & Catarina Röjder Papmehl (eds),
141–150. Gothenburg: Gothenburg University, Department of Swedish.
Noailly, Michèle. 1990. Le substantif épithète. Paris: Presses Universitaires de France.
Noailly, Michèle. To appear. Les séquences nom-nom. In La grande grammaire du français, Anne
Abeillé, Danièle Godard & Anne Delaveau (eds). Arles: Actes Sud.
Olsen, Susan. 2001. Copulative compounds: A closer look at the interface between syntax
and morphology. In Yearbook of Morphology 2000, Geert Booij & Jaap van Marle (eds),
279–320. Dordrecht: Kluwer. DOI: 10.1007/978-94-017-3724-1_11
Olsen, Susan. 2004. Coordination in morphology and syntax. The case of copulative com-
pounds. In The Composition of Meaning. From Lexeme to Discourse [Current Issues in Lin-
guistic Theory 255], Alice ter Meulen & Werner Abraham (eds), 17–37. Amsterdam: John
Benjamins.
Partee, Barbara & Borschev, Vladimir. 2012. Sortal, relational, and functional interpretations of
nouns and Russian container constructions. Journal of Semantics, 29: 445–486. DOI: 10.1093
/jos/ffs009
ZiareRom. 〈http://www.lingv.ro〉
Appendix
The following symmetric relational nouns autostradă ‘highway’

(selecting between-PP) occur in at least one balanță ‘balance’
context with RCC, on the web. bătălie ‘battle’
acord ‘accord’ blocaj ‘freeze’
adversitate ‘adversity’ căsătorie ‘marriage’
afacere ‘affair’ căsnicie ‘matrimony’
alianță ‘alliance’ ceartă ‘controversy’
alternanță ‘alternation’ ciocnire ‘encounter’
amestec ‘mixture’ coabitare ‘cohabitation’
amiciție ‘friendship’ coaliție ‘coalition’
analogie ‘analogy’ colaborare ‘collaboration’
antagonism ‘antagonism’ combinație ‘combination’
antiteză ‘antithesis’ comparație ‘comparison’
apropiere ‘approach’ compatibilitate ‘compatibility’
aranjament ‘arrangement’ competiție ‘competition’
armistițiu ‘armistice’ complicitate ‘complicity’
armonie ‘harmony’ comunicare ‘communication’
asemănare ‘resemblance’ comuniune ‘communion’
asimetrie ‘asymmetry’ concordanță ‘concordance’
asociere ‘association’ concubinaj ‘cohabitation’
atracție ‘attraction’ concurență ‘competition’
Chapter 5. Man–nature relationship 
conexiune ‘connexion’ duel ‘duel’

conflict ‘conflict’ dușmănie ‘enmity’
confluență ‘junction’ echilibru ‘equilibrium’
confruntare ‘confronting’ egalitate ‘equality’
conlucrare ‘cooperation’ finală ‘final game’
contact ‘contact’ fricțiune ‘friction’
contract ‘contract’ frontieră ‘frontier’
contradicție ‘contradiction’ fuziune ‘fusion’
contrast ‘contrast’ gâlceavă ‘quarrel’
convenție ‘convention’ grup ‘group’
conviețuire ‘cohabitation’ identitate ‘identity’
convorbire ‘talk’ idilă ‘idyll’
cooperare ‘cooperation’ impact ‘impact’
corelare ‘correlation’ incompatibilitate ‘incompatibility’
corespondență ‘correspondence’ inegalitate ‘inequality’
culoar ‘corridor’ interacțiune ‘interaction’
cumetrie ‘godfathership’ interconectare ‘interconnection’
cuplu ‘couple’ interconexiune ‘interconnection’
decalaj ‘discrepancy’ interdependență ‘interdependence’
demarcaţie ‘demarcation’ interfață ‘interface’
deosebire ‘difference’ interferență ‘interference’
departajare ‘deciding between’ interoperabilitate ‘interoperability’
dependență ‘dependence’ intersecție ‘intersection’
derby ‘derby’ interval ‘interval’
desincronizare ‘desynchronisation’ iubire ‘love’
dezacord ‘disagreement’ îmbinare ‘joining’
dezbatere ‘debate’ încăierare ‘skirmish’
dezechilibru ‘imbalance’ încredere ‘confidence’
dialog ‘dialogue’ înfrățire ‘fraternity’
diferend ‘argument’ înfruntare ‘facing’
diferență ‘difference’ întâlnire ‘meeting’
discrepanță ‘discrepancy’ întrevedere ‘meeting’
discriminare ‘discrimination’ înțelegere ‘agreement’
discuție ‘discussion’ joc ‘game’
disociere ‘dissociation’ legătură ‘link’
disonanță ‘dissonance’ limită ‘limit’
disproporție ‘disproportion’ litigiu ‘dispute’
dispută ‘dispute’ luptă ‘fight’
distanţă ‘distance’ mariaj ‘marriage’
distincție ‘distinction’ meci ‘match’
divergență ‘divergence’ negociere ‘negotiation’
divorț ‘divorce’ opoziție ‘opposition’
dragoste ‘love’ ostilitate ‘hostility’
drum ‘way’ pact ‘pact’
dualism ‘dualism’ pereche ‘pair’
duet ‘duet’ paralelism ‘parallelism’
 Ana-Maria Barbu
parteneriat ‘partnership’ simbioză ‘symbiosis’

partidă ‘contest’ similitudine ‘similitude’
perioadă ‘period’ simultaneitate ‘concomitance’
polemică ‘polemic’ sincronizare ‘synchronisation’
prietenie ‘friendship’ solidaritate ‘solidarity’
protocol ‘protocol’ supercupă ‘super-cup’
racord ‘junction’ summit ‘summit’
raport ‘rapport’,‘ratio’ tandem ‘tandem’
răfuială ‘scuffle’ tensiune ‘tension’
război ‘war’ transfer ‘transfer’
reconciliere ‘reconciliation’ tranzacție ‘transaction’
relație ‘relation(ship)’ tranziţie ‘transition’
relaționare ‘relation(ship)’ traseu ‘tract’
reuniune ‘reunion’ tratat ‘treaty’
rivalitate ‘rivalry’ trecere ‘transition’
ruptură ‘rupture’ trio ‘trio’
rută ‘route’ troc ‘exchange’
scandal ‘scandal’ tronson ‘portion of line’
schimb ‘exchange’ unificare ‘unification’
schismă ‘schism’ unitate ‘unity’
separație ‘separation’ violență ‘violence’
chapter 6
Classifier noun phrases of the type N1N2

in Bulgarian
Petya Osenova
Sofia University
This chapter discusses classifier* noun phrases of the type Noun1 Noun2
(N1N2) in Bulgarian. The data analysis and the language-specific properties
described show that the apposition-like relation between the two nouns is in fact
a government relation. For that reason, within the framework of Head-driven
Phrase Structure Grammar (HPSG) I consider classifier noun phrases to be
phrases of type head–complement, where the first noun (N1) is the syntactic and
semantic head, while the second (N2) is an argument, which, however, might
be optional on the syntactic level. Three semantic subtypes are presented and
discussed: measure – substance, container – contained, and form of grouping –
grouped entities. The fine-grained distinctions among them are also made explicit
within the ideas of qualia structure and the generative lexicon.
1. Introduction
Many languages allow nominal phrases of the type Noun1 Noun2 (N1N2) or
Noun Phrase1 Noun Phrase2 (NP1 NP2). They can exhibit two relations between
the first (N1) and the second (N2) nouns: either apposition or government. Hence,
* The categorization of nouns in a cross-linguistic context is far from a trivial task. Usually
‘classifiers’ are viewed as an intermediate state of noun classification systems – positioned
between lexical noun terms and morphosyntactic noun classes (see further in Grinevald
2002). Although, when following these considerations, Bulgarian is not considered a ‘classifier
language’, I believe that it has classifier-like constructions. These constructions are function-
ally close to the true classifiers, typical of languages like Chinese. Like Lehrer (1986) I assume
that the notion of ‘classifier’ subsumes the notions of measures, containers, and forms of
groupings (collectives).
 Petya Osenova
these relations are language-dependent. By ‘apposition’ I understand a grammati-

cal relation in which two nouns are placed next to each other and one modifies the
other. By ‘government’ I mean a grammatical relation, in which the head subcat-
egorizes for its dependent.
In contrast to other Slavic languages, Bulgarian is mainly an analytic language.
Historically, its development (mainly lexicon and grammar) has been influenced
by the Greek and Turkish languages. Concerning the nominal system, Bulgarian
has lost its case declension. The weakening of synthetism (including cases) dates
back to the 12th–14th centuries AD. Only the vocative still exists, but only some
nouns have a distinct vocative form. Instead, a reduced set of case forms is pre-
served in some types of pronouns. Bulgarian has a definite article, which is realized
as a postposed making part of the word. However, syntactically it is considered a
phrasal affix within an NP.
Bulgarian noun phrases, in particular, can show four internal syntactic rela-
tions, three of which are presented below:1
1. agreement in gender and number (Adjective Noun),

2. prepositional linkage (N preposition N),
3. juxtaposition (NN). Semantically, the third, apposition-like type can be divided
into two subgroups: (3a) true appositions and (3b) classifier expressions.
(1) Agreement: A N
xubav-a žena
pretty-f.sg woman-f.sg
‘a pretty woman’
(2) Prepositional linkage: N p N
žena ot Sofia
woman-sg from Sofia
‘a woman from Sofia’
(3) Juxtaposition: N N
a. čičo Ivan
uncle-m.sg Ivan-m.sg
‘uncle Ivan’
b. čaša voda
glass-sg water-sg
‘a glass of water’
. The fourth syntactic relation – government – if considered in a broader sense (i.e. not only
as a case assignment) is typical of deverbal and relational nouns.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
I consider examples like (3a) to be appositions, in which one of the nouns modi-
fies the other. There are arguments in favor of analyzing N1 as the head and argu-
ments in favor of selecting N2 as the head. However, this issue is not the focus of
this text.
My aim is to show that although type (3b) displays an appositional ordering,
that is, it lacks overt case marking, which is lost in Modern Bulgarian, and some-
times it also lacks number agreement – in fact it exhibits government relations.
Furthermore, the nouns involved are transitive nouns with an obligatory argu-
ment. And, finally, at the syntactic–semantic interface these classifier nouns are
placed closer to relational nouns. They are also viewed as supporting the work of
quantifiers.
In traditional Bulgarian grammars, both phrases (3a and 3b) are viewed as
appositions, and discussion mainly concerns the problem of which noun is the
head and which is the modifier. In Vakarelijska (2011: 48), these constructions
are considered head–modifier phrases, in which N1 is the syntactic head and N2
is the modifier. Such phrases are called partitive–category constructions. As has
recently been proposed for other languages, such as Russian, German, and Dutch
(see Corbett et al. 1993; Teubert 2003; Trawinski 2000; Van Eynde 2006), I assume
that in these phrases the first noun functions as head. However, in contrast to
the predominant analyses, in which these classifier nouns are viewed as types of
numerals and thus – selecting their nominal heads, I argue that N1 subcategorizes
for N2. The proposed analysis in this text follows the HPSG theory (Pollard & Sag
1994). In the HPSG literature the ‘mutual-selection’ feature is preferred when NP-
internal agreement is considered. This means that there is a mechanism through
which the head selects its dependent as well as the dependent selects its head at
the same time. The mechanism differentiates between specifiers and modifiers:
“The factor which distinguishes specifiers from modifiers is lexical selection: while
specifiers are selected by their head, modifiers are not” (Van Eynde 2006: 159).
This is thus a good alternative analysis to the one I propose here. However, I con-
sider N1 nouns to be closer to relational and deverbal nouns than to determiners
and quantifiers. This means that I view N1 nouns as subcategorizing heads rather
than modified heads or dependents. In this way, I pursue a unified analysis of
Bulgarian argument-taking nouns.
The semantics of nominal constructions with classifiers has been explored in
many theoretical frameworks and for a number of languages. The problem of their
typology remains complex and non-homogeneous. Borer (2005), among others,
considers lexical expressions of quantity to be ‘stuff dividers’, which seems appli-
cable to quantity nouns, too. Pit’ha (1981: 219) calls such nouns ‘quantifiers in a
broad sense’ and says that nouns of the type ‘group’ have ‘frames with obligatory
actants’.
 Petya Osenova
In this chapter, as mentioned above, the phenomenon is analyzed in the

constraint-based framework of Head-driven Phrase Structure Grammar, in which
I incorporate Pustejovsky’s ideas of the generative lexicon (Pustejovsky 1998). The
incorporation adds to the semantic nature of the classifiers in their exophoric and
endophoric usages.
The chapter is structured as follows: Section 2 presents the typology of Bul-
garian classifier noun phrases. It also provides some language-specific properties,
whose selection properties differ with respect to the complement/adjunct distinc-
tion2 Section 3 describes the proposed analysis, according to which classifier NPs
are considered head–complement phrases. Additionally, the specific behavior of
each subtype is outlined. Finally, the semantic contribution of the head is dis-
cussed. The last section offers conclusions.
2. Typology and language-specific properties
I use the term ‘classifier’ as a hypernym for all the specific types listed below. Need-
less to say, the content of this notion is very different from that used extensively for
languages like Chinese, where nouns and quantifiers are systematically mediated
by ‘classifiers’. However, as that sense also exists idiosyncratically in languages like
English (one pair of shoes), I adopt it for clarity when presenting the typology in
Bulgarian.
The N1N2 classifier expressions can be subdivided into three groups: (i) mea-
sure – substance (4), (ii) container – contained (5), and (iii) form of grouping –
grouped entities (6). As the names of the types suggest, in type (i) the first noun
indicates measure; in type (ii) it indicates a container; and in type (iii), a grouping
formation.
(4) measure – substance

a. litâr mljak-o
liter-sg milk-sg
‘a liter of milk’
b. kilogram jabâlk-i
kilo-sg apples-pl
‘a kilo of apples’
. Note that in HPSG both roles, modifier and adjunct, are called adjuncts.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
(5) container – contained

a. čaš-a vod-a
glass-sg water-sg
‘aglass of water’
b. košnic-a domat-i
basket-sg tomatoes-pl
‘a basket of tomatoes’
(6) form of grouping – grouped entities
a. oblak prax
cloud-sg dust-sg
‘a cloud of dust’
b. tâlp-a xora
crowd-sg people-pl
‘a crowd of people’
For the describing properties of these constructions, I use several linguistic cri-
teria: selectional restrictions (2.1), paraphrases with a prepositional phrase (2.2),
comparison with synthetic Slavic languages (2.3), subject–predicate agreement
vs. NP-internal agreement (2.4), pronominal coreference substitution (2.5), and
obligatoriness of the noun within the NP (2.6). Then I give some additional local
properties by type. The rationale behind outlining these properties is to show that
the phenomenon is complex at the syntax–semantic interface, and also to provide
evidence that the proposed head–complement analysis is appropriate for Bulgarian.
My hypotheses have been verified using data from the Bulgarian National
Reference Corpus.3
2.1 Selectional restrictions

The second noun (N2) in the classifier expressions is an argument required by
the valency frame of the first noun (N1), which is the classifier. On the syntactic
level, such arguments are called the complements of the head. All head nouns
(N1) require their dependents (N2) to be determiner-less and quantifier-less. The
presence of indefinites, determiners, or quantifiers renders the phrases ungram-
matical, as illustrated in (7).
(7) a. litâr *tova mljako
liter this milk
b. litâr *mljako-to
liter milk-art
. 〈www.webclark.org〉
 Petya Osenova
c. chasha *edno kafe

cup one coffee
d. grupa *njakakvi studenti
group some students
It can be seen that all three types of N1 introduced above require a noun in either
the singular (mljako, kafe), or the plural (studenti), depending on the seman-
tic characteristics of N1, that is, whether it requires a non-discrete or a discrete
entity. In this respect classifier nouns differ from numerals, since numerals usually
require discrete entities to combine with.
This property shows that N1 imposes grammatical restrictions on N2. Thus, it
is the head, which selects for its complement.
2.2 Paraphrase with a prepositional phrase

Classifier expressions compete with prepositional phrases, as in (8). Appositives
do not allow this kind of alternative marking. They usually compete with pred-
icative constructions that express attributive equality. Compare the appositional
construction uncle John with the predicated paraphrase John is my uncle.
(8) measure – substance
litâr ot mljako-to
liter-sg from milk-sg.art
The type presented in (8) shows the strongest partitive relation ‘something from/of
something’ of the three types. Here are examples of the other two:
(9) container – contained
čaša s voda
glass with water
‘a glass filled with water’
(10) form of grouping – grouped entities
tâlpa ot/s xora
crowd from/with people
‘a crowd of people’
As is evident, in type (9) a non-partitive preposition is used, while in type (10) a

partitive preposition (ot ‘of, from’) alternates with an additive one (s ‘with’).4 No
definiteness marker on N2 is required as it is in type (8).
. The preposition s has a phonetically conditioned allomorph sâs, which is used when the
next word begins with s- or z-. Note that the additive preposition is not used very frequently,
but it does occur.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
Although it is assumed that NPs with prepositional phrases ‘compete’ with

the above mentioned NPs of type N1N2, it must be noted that the former phrases
are rarer, in general, than the latter to express exactly the same meaning. Thus,
using nouns with prepositional phrases is evoked by a special context, such as the
need to stress the partitive relation (as in (8)), the content (as in (9)), or the form
(as in (10)). For example, the statistics of the word group in NPs of the type N1N2
or NP PP in the corpus showed the following ratio: of 783 relevant occurrences,
560 are of the NN type while 223 are of the NP with a PP type that have the same
meaning.
This criterion shows that the obligatory argument can be expressed in alterna-
tive ways, which is good evidence for considering N1 the head subcategorizing for
its arguments.
2.3 Comparison with other Slavic languages

In similar N1N2 phrases, the second noun (N2) in synthetic Slavic languages
has overt genitive case marking. Since modern Bulgarian has lost its declension
system, such contrastive approaches are justified and helpful for interpreting the
grammatical relations, which are overtly non-marked. Here are two examples
comparing Russian with Czech:
(11) Russian: litr molok-a
Czech: litr mlék-a
liter milk-gen
(12) Russian: stakan vod-y
Czech: sklenice vod-y
glass water-gen
‘a glass of water’
Since these modern Slavic languages are closer to Old Bulgarian5 in their gram-
matical systems, it can be assumed that in contemporary Bulgarian N1N2 classifier
phrases, the relation between the two nouns is government rather than apposition.
This property shows that N2 is morphologically marked in other Slavic
languages as a dependant. Thus, N1 is the head of the construction, subcategoriz-
ing for N2.
Apart from the morphosyntactic properties listed in 2.1, 2.2, and 2.3 above,
there are also other properties typical of classifier expressions that, however, also
display non-local features (2.4, 2.5, and 2.6 below), and thus are conditioned by
discourse.
. According to our tradition, I use Old Bulgarian as a synonym to Old Church Slavonic.
 Petya Osenova
2.4 Subject-predicate agreement vs. NP-internal agreement

In subject position, the agreement of a classifier NP allows for alternative patterns:
the predicate agrees either with the head (dojde), or with the dependent (dojdoxa).
For example:
(13) Grupa student-i dojd-e/dojd-oxa.
group-sg students-pl come-aor.3sg/aor.3pl
‘A group of students came.’
In contrast to Romanian,6 where the use of a modifier/determiner/quantifier other
than the indefinite article itself (a group) would require only singular agreement,
in Bulgarian there is no such restriction. The type form of grouping – grouped
entities seems to be the most unrestrictive type with respect to alternative agree-
ment behavior. The other two types, however, measure – substance and container –
contained, seem to be more dependent on specific conditions in the sentence, such
as word order and the semantics of the predicate.
Thus, this property does not favor any of the nouns as head or dependent. This
fact is not as peculiar as it seems at first sight, since similar variance of agreement
can be observed in coordinated NPs in subject position. If the coordinated NPs
are semantically close, then the agreement with the predicate might be singular.
The agreement within NPs of all the N1N2 types considered here shows
another tendency. Preposed extensions (such as adjectives) agree only with N1,
while postposed modifiers (such as relatives) might agree with either N1 or N2
depending on the semantics. This fact supports the idea that N1 is the head of the
N1N2 construction, since it takes the local morphosyntactic agreement, while N2
is its governed complement.
2.5 Pronominal coreference substitution

Either the first noun (N1) or the second (N2) can be pronominalized in semanti-
cally unspecified contexts. This is shown in (14) with the type container – contained,
where ja pronominalizes čaša or go pronominalizes kafe:
(14) Vze-x čaša kafe. Iz-pi-x ja/go.
take-1sg.aor cup-f.sg coffee-n.sg pfv-drink-1sg.aor her-f.sg/it-n.sg
‘I took a cup of coffee. I drank it up.’
The same observation holds for the other two types: measure – substance and form
of grouping – grouped entities. Like the subject–predicate criterion discussed above,
this one does not distinguish any of the nouns as head or dependent.
. I owe this observation regarding Romanian to Ana-Maria Barbu.

Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
The observations in 2.4 and 2.5 comply with Corbett’s agreement hierarchy
(Corbett 2003), in which from left to right syntactic agreement weakens, while
semantic agreement becomes stronger: attributive < predicative < relative pronoun
< personal pronoun. In the case of Bulgarian classifier NPs of the type N1N2, the
attributive use of the modifier is always syntactically bound, while agreement with
predicates, relatives, and anaphoric pronouns is semantically justified, thus allow-
ing alternative patterns. Recall that N1 has no restrictions on its attributive agree-
ment (it might agree locally with determiners, quantifiers, adjectives, etc.), while
N2 exhibits some restrictions on its local syntactic agreement (it cannot agree with
determiners or quantifiers).
To sum up, local attributive agreement in the N1N2 nominal phrase, which is
completely morphosyntactic, has as its locus N1, thus supporting my hypothesis
that N1 is the syntactic head, subcategorizing over N2.
2.6 Obligatoriness of the nouns within an NP

This criterion might be related to the phenomenon of ellipsis, and it also depends
on the lexical semantics of the predicate. In this respect I support Zwicky’s opinion,
which is cited in (McGlashan 1993: 205): “The notion of obligatoriness, however,
must be approached with caution, since the absence of a category can indicate
either optional or elliptical status.” The obligatoriness criterion is also closely
related to the issue of the reference status of this NP type. In my opinion, the mea-
sure – substance type as well as the grouping type have only one referent, while the
container – contained type has two specific referents. This assumption is justified
by the fact that in the first two types either of the two nouns can be omitted within
the discourse, while in the last case the omission can also be evoked by the lexical
meaning of the predicate. An example is (15): omission of butilka ‘bottle’ would
produce an ungrammatical sentence, because the lexical semantics of the verb ‘to
break’ requires an overt expression of some container. However, such occurrences
are rare in general. In Bulgarian, this meaning would be normally expressed by
other means (NPs with PPs, as discussed above; without the complement; or by a
whole sentence). Additionally, it seems that concrete words denoting containers
have different combinatorial powers. For example, the word bottle is more natural
in such expressions than the word cup/glass.
(15) Sčupix butilka vino.
break-1sg.aor bottle wine
‘I broke a bottle of wine.’
The omission of the container is allowed with another type of verb, such as ‘to give’
(to order, to drink, etc.), which evokes metonymical uses of the liquid taking on
the role of the container, as in (16).
 Petya Osenova
(16) Dadoxa mu [čaša] aromatno kafe.

give-3sg.aor him-dat [cup] fragrant coffee
‘They gave him a cup of fragrant coffee.’
This property helps in distinguishing semantically among the three subtypes rather
than in assigning the appropriate phrase type (head–complement or head–adjunct).
2.7 Other local properties

The three types of NP under examination show some further specific properties.
First, the measure – substance type can take count as well as mass nouns. Hence,
a restriction of number applies to them in that count nouns appear in the plural
and mass nouns in the singular. Additionally, apart from the numerals or other
quantifiers, which stand in anteposition, only a reduced number of pre-modifiers
of N1 can be used. For example, it can be modified by ‘whole’ or ‘additional’ (17a)
but not by a qualifying adjective such as ‘beautiful’ (17b).
(17) a. Izpih cjal / dopâlnitelen litâr voda.

drink-1sg.aor whole additional liter water
‘I drank up the whole/an additional liter of water.’
b. *Izpih xubav litâr voda.
drink-1sg.aor beautiful liter water
‘I drank up a beautiful liter of water.’
Apart from the DP (Determiner Phrase)/QP (Quantifier Phrase) analysis, where

the determiner or the quantifier is the head, in the linguistic literature there is also
an alternative approach, according to which some combinations can be regarded
as complex specifiers that modify their head. This is the view adopted, for example,
in HPSG (Pollard & Sag 1994: 360). Following this perspective, classifier nouns,
which are countable, might form a complex specifier with the preceding numeral,
as in (18).
(18) dva litr-a mljako

two liter-pl milk
‘two liters of milk’
However, in our analysis we prefer to regard classifier expressions as heads. In

accordance with this, combinations of a numeral plus a classifier like two liters
might be treated as complex heads (not specifiers), which then subcategorize for
N2 (milk). This issue is not further elaborated or pursued as a solution in this
chapter, since in the singular and in some plurals the presence of a quantifier is
optional in Bulgarian.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
In the PP paraphrase, this type strongly favors the partitive preposition ot

‘from/of ’. But in such cases it assumes the entire quantity of N2. For that reason N2
within the PP paraphrase can only be definite (marked either by a definite article
or by a demonstrative pronoun: dva litra ot mljako-to; two liters of milk-art). This
criterion also shows that N2 is an argument of N1, since the PP is a paraphrase of
the N1N2 type. Compare the example where the noun ‘milk’ in square brackets is
presupposed as a default argument, which remains unexpressed: dva litra [mljako]
ot mljako-to (two liters [milk] of milk-art).
Second, the container – contained type also takes count as well as mass nouns.
The PP paraphrase prefers the preposition s ‘with’. If the whole quantity is indicated
by N2 via a definite article, then N1 also has to be definite. Thus, some kind of
definiteness agreement is imposed by N2. For example, a phrase with an indefinite
N1 (19a) is ungrammatical, while one with definiteness agreement (19b) is gram-
matical. This fact seems to be a counterargument to my claim that N1 imposes
various restrictions on N2. However, the role of the category of definiteness in
these phrases needs more elaboration, which remains beyond this chapter’s scope.
(19) a. *čaša s voda-ta
glass with water-art
‘a glass of the water’
b. čaša-ta s voda-ta
glass-art with water-art
‘the glass with the water’
Note that the expressions become grammatical when the partitive preposition ot
‘from/of ’ is used with a definite N2. This situation is shown in (20). In such cases
this type becomes closer to the type measure – substance.
(20) Izpix čaša ot vino-to.
pfv-drink-1sg.aor glass from wine-art
‘I drank up a glass of the wine.’
Third, the form of grouping – grouped entities type can also take count as well as
mass nouns, depending on the entity that is formed (e.g. grupa studenti ‘a group
of students’, but oblak prax ‘a cloud of dust’). The PP paraphrase can take the par-
titive preposition ot ‘from/of ’, s ‘with’, or na ‘of ’. The situation with imposed defi-
nite agreement is similar to the previous type. For example, the phrase without
definiteness agreement (21a) is questionable, but the phrase with agreement is
grammatical.
(21) a. ?grupa sas/ot/na studenti-te
group with/of students-art
‘a group of the students’
 Petya Osenova
b. grupa-ta sas/ot/na studenti-te

group-art with/of students-art
‘the group of the students’
Note that when some syntactic definiteness is added, these constraints might be
loosened. For example, the addition of a relative clause that modifies the noun
studentite (‘the students’) makes the sentence grammatical (22).
(22) Pokanix grupa ot studenti-te, koito uchat tuk.

invite-1sg.aor group of students-art who study-3pl.prs here
‘I invited a group of the students that are studying here.’
3. The proposed analysis
Our analysis is based on the distinction between the argument structure and
valency lists, which is defined in the theory of HPSG (Manning & Sag 1995). The
argument structure (ARG-ST) encodes the potential obligatory participants in
the situation as an ordered list. This list follows the obliqueness hierarchy. It is
used for modeling binding phenomena. The valency lists encode the projections
of ARG-ST, which are syntactically realized. These lists include subject, comple-
ments, and specifiers.7 Thus the mappings between the two levels of representa-
tion might be:
– The argument exists in ARG-ST and it is also realized in Valency list(s).

– The argument exists in ARG-ST but it is not realized in Valency list(s).
An example of the first case is (23).
(23) Izvikax grupa-ta studenti.

pfv-call-1sg.aor group-art students
‘I called the group of students.’
N1 (‘group’) subcategorizes for N2 (‘students’). An example of the second case is

the same sentence (24), but without the word studenti (‘students’).
(24) Izvikax grupa-ta.

pfv-call-1sg-aor group-art
‘I called the group.’
. Specifiers are not considered here. They are mentioned just for completeness.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
In this case N1 still requires N2 semantically, but N2 is omitted, since it is deriv-

able through the wider context. As mentioned above, the measure and grouping
types can omit their arguments within the given discourse, while omission in the
container type also depends on the lexical semantics of the predicate. Figure 1 is
a simplified lexical specification for these nouns. Note that the example is for the
grouping type, but it also holds for the other two types.
word
PHON < group >
HEAD noun
SYNSEM|LOC|CAT
VAL|COMP < >
ARG-ST < NP >
Figure 1. The lexical specification of all the classifier noun types
The specification means that for the noun group in the syntactic-semantic
domain (SYNSEM) and in the local environment (LOC), the syntactic character-
istics in the category domain (CAT) are head (HEAD) and valency (VAL). The
value of the feature HEAD is noun, while the empty value of the complement
(COMP) within VAL indicates the optionality of the argument. At the same time,
the argument is encoded in the argument structure (ARG-ST) as a NP, which is
governed.
This approach is combined with the qualia structure approach presented in
Pustejovsky (1998). The qualia structure, together with the argument structure,
event structure, and inheritance structure, constitutes a mechanism of lexical
decomposition, which underlies the notion of a generative lexicon. According to
Pustejovsky (1998), each word can have the following qualia: Formal, Telic, Consti-
tutive, Agentive. The Formal feature encodes the differentia specifica of the concept
designated by the word. This refers to the properties that differentiate one con-
cept from other concepts. The Telic feature presents the purpose (use, function).
The Constitutive feature refers to the parts of an entity. The Agentive characteristic
presents the origin. Pustejovsky also introduces four types of arguments. These
are: ARG0 (the ontological label of the word), T-ARG (true argument, i.e. the one
that has to be realized syntactically), D-ARG (default argument, i.e. the one that
might not be realized syntactically), and S-ARG (argument in shadow, i.e. the one
that is realized syntactically only under special conditions, since it is part of the
lexical meaning of the word).
 Petya Osenova
Figure 2 gives an example for the relational noun brother from (Pustejovsky
1998).
brother
ARG x : human
ARG-ST D-ARG y : human
CONST male (x)
QUALIA FORMAL brother_of (x,y)
Figure 2. The relational noun brother
The schema says that brother itself has the ontological value (ARG) human. It
has a default argument (D-ARG) whose value is also human. The Qualia encode
the differentia specifica, which is the value Formal: brother_of, and the more spe-
cific ontological restriction, the value Constitutive: male.
Osenova (2009: 104) shows that these qualia can be ordered in a hierarchy
with respect to the relational properties of the various relational nouns. The pro-
posed hierarchy is as follows: Formal > Constitutive > Telic > Agentive. The idea
is that the ‘relational property’ of the nouns diminishes from left to right. In the
present analysis, only Formal and Constitutive features are used, which means that
classifier nouns have a high degree of relational properties.
Classifier N1N2 types are argument-having, and thus complement-taking
nouns. I place them closer to the relational nouns and view them as supporting
the work of quantifiers, since at the same time they presuppose another entity and
also facilitate its quantification. The representations of measure and grouping types
are as follows:
[QUALIA|FORMAL measure_of (x, y)]
[QUALIA|FORMAL group_of (x, y)]
As can be seen, both have values with two arguments for the feature Formal. These
values are relations. In the first case it is measure_of, while in the second case it is
group_of.
The representation of the container type is as follows:
FORMAL container_of (x, y)]
QUALIA
CONSTITUTE container (x)]
Note that only the container type representation has a value for the Constitu-
tive qualia. In this way the fact that N1 refers to an object in the world is reflected,
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
in contrast to the other two types, in which N1 is an abstract noun (being measure,
form).
Let us consider the integrated view of HPSG and qualia specifications for the
three subtypes.
word
PHON <litâr>
noun
HEAD NUMBER [1]sg
CAT GENDER [2]m
VAL|COMP <>
SYNSEM|LOC
nom_obj
PERSON 3
CONT INDEX [x] NUMBER [1]
GENDER [2]
QUALIA|FORMAL measure_of (x,y)
ARG-ST|D-ARG <[y]NP>
Figure 3. The integrated specification for measure type with example word ‘liter’
Figure 3 presents more detailed information than Figure 1, which concen-

trates on syntax. In the semantic part (CONT) there are two semantic features:
INDEX and QUALIA. The syntactic and semantic features NUMBER and GEN-
DER have structure-shared values. This word has a default argument – D-ARG
(following Pustejovsky), since it is semantically required, but optional at the
syntactic level. The argument is structure-shared with the second argument of
the measure_of relation within the features QUALIA|FORMAL. The variable is
named y. The variable x (the ontological label of ‘liter’) is structure-shared with
INDEX. Thus, more fine-grained semantic information is added into the lexical
specification.
In Figure 4 the lexical specification is given for the form-of-grouping type. As
can be seen, there is no difference from Figure 3 except for the presence of the rel-
evant relation group_of. Thus, all other differences would be found in the semantic
selectional requirements of the head over its complement. In this way, the head in
Figure 3 will take only nominals that can be measured, while the head in Figure 4
will take nominals that can be grouped.
 Petya Osenova
word
PHON <grupa>
noun
HEAD NUMBER [1] sg
CAT GENDER [2] f
VAL|COMP <>
SYNSEM|LOC
nom_obj
PERSON 3
GENDER [2]
QUALIA|FORMAL group_of (x,y)
Figure 4. The integrated specification for the form-of-grouping type with example word ‘group’
word
PHON <čaša>
noun
HEAD NUMBER [1]sg
CAT GENDER [2]f
VAL|COMP <>
SYNSEM|LOC
nom_obj
PERSON 3
GENDER [2]
FORMAL container_of (x,y)
QUALIA CONST container (x)
Figure 5. The integrated specification for the container type with example word ‘cup/glass’
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
The container type specification in Figure 5 follows the same representation as

above. However, the QUALIA feature has the additional feature CONST(itutive)
with the ontological variable x as its value. In this way, as noted above, the dif-
ference remains between the abstract index and the specific index of the nouns.
Again, the idea is that selectional properties over and by this noun are made
explicit and thus the proper semantic combinatorics are ensured.
The idea of viewing these nouns so close to relational nouns is neither new
nor revolutionary. However, it is not typical, either. If we consider only the pro-
totypical relational nouns (such as kinship or part–whole), then such a decision
would sound strange. But in NomBank (Meyers 2007), for example, the nouns
assembly and type are considered relational (among 16 types of relational nouns):
an assembly of shareholders and a different type of filter. (Barker & Dowty 1993)
also include abstract nouns in the group of relational nouns. Hence I support the
broader understanding of what belongs to the relational nouns group.
Now let us discuss the head issue. Since the HPSG theory has been adopted for
the present analysis, phrases of the type N1N2 are considered head–complement
phrases, in which N1 is the syntactic head and N2 is the complement. In HPSG,
in head–complement phrases the syntactic head is also a semantic head. Thus, the
contribution of the semantics of N2 is incorporated into that of the head noun.
The split head analysis, in which one daughter of the phrase might be con-
sidered the syntactic head while another might be viewed as the semantic head,
would be appropriate for the head–adjunct phrase, where the adjunct (modifier)
is the semantic head. This means that in those phrases the syntactic and semantic
heads are different, while in phrases of the type head–complement the syntactic
head is also the semantic head of the phrase per se. Here the question arises: what
is the grammatical role of the paraphrasal PPs? Are they also complements, bear-
ing in mind that some of the PPs can be considered ‘partitives’? There are two pos-
sible solutions. One is to keep the type N1N2 as a head–complement relation, while
the NpN paraphrase is to be viewed as a head–adjunct relation. This is one of the
possibilities, which follows Butt et al. (1999). It suggests a trade-off between verbal
and nominal syntactic domains. Thus Butt et al. consider all the dependents in the
nominal domain to be adjuncts. This choice has the advantage that it reduces the
ambiguity of the prepositions being part of complement versus adjunct phrases.
On the other hand, the disadvantage is that two semantically identical phrases
(N1N2 and N1pN2) receive different syntactic analyses. The second possible solu-
tion is to assume that the PPs under consideration are also complements that alter-
nate with N1N2 types. Giving preference to the semantic identity, I would support
the second option, namely, considering NPs and PPs uniformly to be comple-
ments. However, this issue needs more elaboration and argument-strengthening
in the future.
 Petya Osenova
Semantically, the three N1N2 subtypes that have been considered are not
completely identical. The measure and container types facilitate the quantifiers’
job, although the container type has a specific referential index in contrast to the
measure type’s abstract referential index. The grouping type is closer to relational
nouns, but also to appositions, since the ‘equality’ test turns out to be valid for this
type only. Compare the typical apposition in (25) and the grouping phrase in (26),
both used predicatively.
(25) Teresa e majka.
Teresa is mother
‘Teresa is a mother.’
(26) Student-i-te sa grupa.
student-pl.art are group
‘The students are a group.’
The apposition constructions might show some agreement markers of genre, for
example, both nouns majka Teresa ‘Mother Teresa’ are feminine, but no other
grammatical restrictions are observed, no PP-paraphrasing is possible.
4. Conclusions
This chapter aims to contribute, first, to the discussion of the properties that would
be useful for identifying nominal phrases of the classifier type N1N2 in Bulgarian as
head–argument constructions, and hence as head–complement rather than head–
modifier (head–adjunct) phrases. The criteria discussed give enough evidence for
considering N1 the head. The semantic treatment of N2 as argument as well as its
syntactic analysis as complement rely mainly on semantic grounds, which are pro-
vided by Pustejovsky’s generative lexicon architecture and the s yntactic–semantic
interface apparatus behind the HPSG theory.
Second, the chapter provides evidence that the classifier ‘apposition-like’ NPs
in Bulgarian show hidden government relations. These findings are important
for distinguishing between constructions that exhibit the same syntax but are not
identical from a semantic point of view.
The evidence presented here shows that: the three subtypes that have been
distinguished share some common properties but also exhibit their own specific
features – especially in the lexical semantics area; that all the types can be placed
very close to the set of relational nouns and can be viewed as quantifier facilita-
tors, with a different degree of similarity; and that the PP-paraphrases need to be
the object of further research as to whether they should be viewed as adjuncts or
complements to N1.
Chapter 6. Classifier noun phrases of the type N1N2 in Bulgarian 
References
Barker, Charles & Dowty, David. 1993. Nominal thematic proto-roles. 〈http://www.ling.ohio-
state.edu/~dowty〉
Borer, Hagit. 2005. Structuring Sense, Vol. 1: In Name Only. Oxford: OUP.
Butt, Miriam, Holloway King, Tracy, Niño, María-Eugenia & Segond, Frédérique. 1999. A Gram-
mar Writer’s Cookbook. Stanford CA: CSLI.
Corbett, Greville G., Fraser, Norman M. & McGlashan, Scott. 1993. Heads in Grammatical
Theory. Cambridge: CUP. DOI: 10.1017/CBO9780511659454
Corbett, Greville G. 2003. Agreement: Terms and boundaries. In The Role of Agreement in Natu-
ral Language. Proceedings of the 2001 Texas Linguistic Society Conference, William E. Griffin
(ed.), 109–122. Austin TX: University of Texas.
Grinevald, Colette. 2002. Making sense of nominal classification systems. In New Reflections
on Grammaticalization [Typological Studies in Language 49], Ilse Wischer & Gabriele
Diewald (eds), 259–275. Amsterdam: John Benjamins.
Lehrer, Adrienne. 1986. English classifier constructions. Lingua 68: 109–148. DOI: 10.1016/0024-
3841(86)90001-X
Manning, Christopher D. & Sag, Ivan A. 1995. Dissociation between argument structure and
grammatical relations. In Lexical and Constructional Aspects of Linguistic Explanation, Gert
Webelhuth, Jean-Pierre Koenig & Andreas Kathol (eds), 63–78. Stanford CA: CSLI.
McGlashan, Scott. 1993. Heads and lexical semantics. In Heads in Grammatical Theory, Greville
G. Corbett, Norman Fraser & Scott McGlashan (eds), 204–230, Cambridge: CUP. DOI:
10.1017/CBO9780511659454.010
Meyers, Adam. 2007. Annotation Guidelines for NomBank. Noun Argument Structure for
PropBank, 〈http://nlp.cs.nyu.edu/meyers/nombank/nombank-specs-2007.pdf〉
Osenova Petya. 2009. Imennite frazi v bâlgarskija ezik. (The nominal phrases in Bulgarian).
Sofia: Eto.
Piťha, Petr. 1981. On the case of frames of nouns. In Prague Studies in Mathematical Linguistics,
Vol. 7, 215–224. Prague: Academia.
Pollard, Carl & Sag, Ivan A. 1994. Head-driven Phrase Structure Grammar. Stanford CA: CSLI.
Pustejovsky, James. 1998. The Generative Lexicon. Cambridge MA: The MIT Press.
Teubert, Wolfgang. 2003. Die Valenz nicht-verbaler Wortarten: Das Substantiv. In Depen-
denz und Valenz. Ein internationales Handbuch der zeitgenössischen Forschung, Vilmos
Ágel, Ludwig M. Eichinger, Hans Werner Eroms, Peter Hellwig, Hans Jürgen Heringer &
Henning Lobin (eds), 820–835. Berlin: Walter de Gruyter.
Trawinski, Beata. 2000. Die Struktur der Deutschen Nominalphrase: Die HPSG-analyse im
b8-fragment. In Aspekte eines HPSG-Fragments des Deutschen [Arbeitspapiere des SFB
340, Bericht Nr. 156], Jesse Tseng (ed.), 1–37. Stuttgart: University of Stuttgart/University
of Tübingen.
Vakareliyska, Cynthia M. 2011. The new English [N [N]] construction in the Slavic languages,
and why the Baltic languages don’t have it. Slavistica Vilnensis: Kalbotyra 56(2): 45–52.
Van Eynde, Frank. 2006. NP-internal agreement and the structure of the noun phrase. Journal of
Linguistics 42: 139–186. DOI: 10.1017/S0022226705003713
chapter 7
Noun phrasal complements vs. adjuncts
Rossella Resi
University of Verona
This paper draws a parallel between verbs selecting phrasal complements or

adjuncts and nouns followed by Relative clauses (RCs), referring in particular
to German. According to Meinunger (2000) a restrictive relative clause (RRC)
attached to a noun and a sentential complement of a factive verb are both sister
complement phrases (CPs) of lexical heads. This paper aims at providing evidence
of the fact that a RRC and its head are linked by a head-complement-relationship
while a non-restrictive relative clause (NRRC) can be seen as an adjunction
to its head. NRRCs merge with maximal projection late in the derivation of
the associated clause rather than being selected by a lexical head during the
construction of the matrix clause. This paper presents evidence for this difference,
such as the parallelism with Haegeman (2002–2008)’s dichotomy between central
and peripheral clauses.
1. Introduction
The concept of valency is traditionally connected to verbs and refers to their inter-
nal property of selecting any argument controlled by the verb predicate itself. This
also includes the possibility of not selecting any arguments. Verb valency is often
related to the traditional notion of verb transitivity, although do not correspond
entirely as valency may also involve the external argument whereas transitivity
does not. In this paper, the traditional concept of verb valency (in specific cases
referred to as transitivity) is taken a bit further and extended to the nominal
domain. The aim of this work is to explore specific syntactic verbal and nominal
(in particular phrasal) structures, which share similar properties, to provide evi-
dence for a parallelism between verb and noun arguments, and finally to find a
pattern of generalization for the concept of valency for verbs and nouns as far as
phrasal arguments are concerned. The syntactic structures that support the idea of
a parallelism between verb and noun valency are phrasal arguments for verbs and
 Rossella Resi
relative clauses for nouns. The aim of this paper is to draw an analogy on the basis
of the German language.
2. Noun and verb valency
One of the most explicit formal proposals of the parallelism between nouns and
verbs selecting or not selecting phrasal arguments is Meinunger’s (2000: 206) work
on topic effects over extraction and extraposition. In an endeavour to explain how
different types of syntactic movements are all restricted by the blocking effect of
topics, he draws “a parallel between this finding and the behavior of argument sen-
tences of factive predicates, which turn out to be of the same category (= topics)”
(Meinunger 2000: 179). Following these assumptions, and in order to provide
more accurate evidence on similarity between leftward movements like wh-
extraction, topic movement out of noun phrases and rightward movements like
relative clause extraposition, he makes some claims about the position of relative
clauses with respect to their nominal antecedent. Relative clauses in this respect
are then compared with phrasal factive complements. Accordingly, relative clauses
and nouns share the same index, just as a verb shares the same index with its argu-
ment. It is the lexical head that provides its phrasal argument with index, which
can be both referential, or not. If the index is referential, the argument is said to
carry a referential theta-role and counts as a true participant in the event; if it is
not referential, the argument does not participate directly in the event and it is
considered a “quasi-argument or a non-referential expression” (Meinunger 2000:
206). The starting point of my analysis was the assumption made at this point by
Meinunger (2000), and specifically the following:
A noun that is identified by a restrictive relative clause assigns a referential
index to it in the same way as, for example, a verb of saying marks its sentential
complement with argument index, since in both cases we are dealing with sister
CPs of lexical heads (Meinunger 2000: 206).
N′ V′
N0 CPi V0 CPi
Figure 1. Sister CPs of lexical heads
Meinunger’s idea accounts for the fact that in the above mentioned construc-
tion, the CP is linked to the head by a very close head-argument relationship
and that the base position of a restrictive relative clause or of a factive phrasal
Chapter 7. Noun phrasal complements vs. adjuncts 
c omplement is the sister of the lexical head it refers to (we will see in 3.2 that the
lexical head N0 is not exactly the lexical head we want the restrictive relative clause
to be linked to). The aim of this work is to support this analogy with further evi-
dence and to show that there is a hitherto undiscussed difference with respect to
whether the CP of the structure given in Figure 1 is a restrictive or non-restrictive
relative clause. The difference in the syntactic derivation of the two sentences will
not invalidate the original parallelism between nominal and verb valency but, on
the contrary, we will propose that the hierarchical location for non-restrictive rela-
tive clauses is similar to that of a verb adjunction.
Since the distinction between these two types of relative clauses is crucial
to my argument, the following section will focus on some relevant syntactic and
semantic properties of the two types of relative clauses, which will be useful for
later argumentation.
3. Restrictive and non-restrictive relative clauses
3.1 Semantics
Contrary to English and Italian, German relative clauses do not have any overt
syntactic elements that are unambiguously and graphically able to distinguish
between two types of clauses. The relative pronouns are always der, die, das (with
declination), and in none of the cases can they be omitted. This is why in our
German examples an “appositive” adverb such as übrigens has been added inside
the non-restrictive clause: it forces the parenthetical reading without destroying
the minimal pair.
The semantic difference between restrictive and non-restrictive relative
clauses is immediately evident in the following minimal pair of clauses (1a–b).
(1) a. Julia kennt weinige Frauen, die sehr gut kochen können ≠
Julia knows few women that very good cook can
Julia kennt wenige Frauen
Julia knows few women
‘Julia knows few women that know how to cook very well. ≠ Julia
knows few women.’
b. Julia kennt wenige Frauen, die (übrigens) gut kochen
Julia knows few women, who (by the way) good cook
können = Julia kennt wenige Frauen
can Julia knows few women
‘Julia knows few women, who (by the way) know how to cook. = Julia
knows few women.’
 Rossella Resi
A restrictive relative clause narrows down the field of reference of the anteced-
ent and provides relevant information to limit, restrict or unequivocally identify
the noun it modifies. It is therefore a real post-nominal modifier of its anteced-
ent because by removing the relative clause, the underlying meaning of the main
sentence changes. If we look at the definitions of modifiers proposed in the litera-
ture we see that no doubts arise about the modifying nature of restrictive relative
clauses. According to the theoretical definition, a modifier is restrictive if the set
of objects denoted by a modified head is properly contained in the denotation of
the head alone. From a procedural point of view, a restrictive modifier contrib-
utes to identifying the referent of a complex term expression of the form ‘Deter-
miner Head’ (Alexiadou 2001; Umbach 2006). There is little consensus about the
discourse-related status of restrictive modifiers. They may be an active part of the
presupposition but they can also represent new information (Fabricius-Hansen
2009).
According to these definitions, sentence (1b) does not seem to have the same
status. In fact the denotation of the head and the head itself are equal and the
relative clause provides additional information about some already identified dis-
course referent. From a discourse-related point of view non-restrictive relative
clauses consist of new information that can also often be stated as a separate asser-
tion without changing the meaning.
Non-restrictive antecedents are not identified by non-restrictive relative
clauses but by the context or by the noun itself. The relative clause provides sup-
plemental information which does not limit the meaning domain of the noun it
modifies.
By removing the relative clause, the truth-value of the main clause does not
change and the relative clause turns out to be only a supplement to the basic mean-
ing of the sentence.
Based on this assumption, it seems that non-restrictive relative clauses cannot
be modifiers of the antecedent, and consequently, following Meinunger's analysis
they should not be able to profit from a co-indexing relationship with a lexical
head.
3.2 The antecedent

Based on the assumptions made so far about the semantic difference between
restrictive and non-restrictive relative clauses, we can deduce which nouns can
exclusively introduce one type or the other. Non-restrictive relative clauses relativ-
ize already defined nouns (such as proper nouns, but rarely pronouns), which are
well identified and limited (2a). Proper nouns in the sense of (2a), with respect to
substantives in general, are in a way “intransitive” and cannot therefore have any
Chapter 7. Noun phrasal complements vs. adjuncts 
arguments. The syntactic characterisation of proper names as intransitive deter-

miners is motivated by the fact that they can form DPs by themselves, they can be
modified, and are incompatible with determiners (Egg 2007). Again, we see that a
tree like (1a) is not suitable for non-restrictive relative clauses because the lexical
head must be at least transitive to project an argument.
(2) Peter, der übrigens auf die Party kommt, ist ein Freund
Peter who by the way to the party comes is a friend
von meinem Bruder.
of my brother
‘Peter, who by the way comes to the party, is a friend of my brother.’
A non-restrictive relative clause is also sensitive to the kind of determiner that

precedes the antecedent noun, especially when it is a quantifier that makes the fol-
lowing noun unidentifiable. This is the case with a distributive quantifier like jeder,
which extends the meaning of a specific reference to a broader undefined unit of
references (3a) or a negative determiner like kein (3b).
(3) a. *Jeder Professor, der übrigens eingeladen wurde, hat
every professor, who by the way invited was, has
an der Konferenz teilgenommen.
to the conference attended.
‘Every professor, who was invited, attended the conference.’
b. *Keine Frau, die (übrigens) Julia kennt, kann
no women, who (by the way) Julia knows can
gut kochen.
good cook
‘No women, who (by the way) Julia knows, knows how to cook.’
We see from (3a) that the possibility of attaching a non-restrictive relative clause
depends not only on the nature of the noun it refers to but also on the kind of
determiner or quantifier merged with it. This could be because a relative clause
merges after the entire DP has been built and that the determiner of a non-restric-
tive head is therefore part of what we call the relative clause’s head; otherwise an
external determiner could not interfere in the grammaticality of a relative clause.
Semantically, only a complete DP can be understandable on its own, and this is
what a non-restrictive relative clause like (1b) requires to attach to and none the
less be an independent speech act, conveying new information.
If the point of attachment of a non-restrictive relative clause is a maximal pro-
jection like DP (or NP), the relative clause is definitely a non-argument because
there is no transitive lexical head, as in (1a), which can project it. Most of the
 Rossella Resi
analysis that put restrictive and non-restrictive relative clauses in different syntac-
tic positions choose a lower position for restrictive relative clauses and a position
that is higher for non-restrictive ones (Ziv & Cole 1974; Emonds 1979; Fabb 1989)
even if the configuration proposed at different stages of the theory differs consis-
tently. Sections 3.2 – 3.4 will deal with the actual syntactic derivation of relative
clauses.
We then need to consider that a restrictive relative clause can only have nomi-
nal antecedents, while a non-restrictive relative clause can also take entire CPs
as antecedent (4). This means that restrictive relative clauses are limited in this
manner because they are selected through a single lexical head which they are
complement to. Non-restrictive relative clauses, on the other hand, are not ruled
by sisterhood relationship and since they do not profit from co-indexing relation-
ship, they may take other possible maximal projections.
(4) Er wohnt in Berlin, worüber er auch froh ist.

he lives in Berlin about this he also happy is
‘He lives in Berlin, which makes him very happy.’
What forces the two types of relative clauses to appear in different positions?
Semantically, the fact that just one of the two is a modifier is sufficient to predict
that there is a different structural relation; their relation with the antecedent and
with determiner is the empirical evidence that supports the semantic intuition.
A difference in the way relative clauses relate with the determiner of their
antecedents concerns the possibility of leaving a floating quantifier of the ante-
cedent inside the relative clause (Bianchi 2000: 46). German quantifiers like alle
(8a–b) or beide, which introduce a definite DP, can be stranded when the noun
they refer to moves to the left.
(5) a. Alle Studenten haben die Prüfung bestanden

all students have the exam passed
b. [DPDie Studenten] haben [alle tDP] die Prüfung bestanden.
the students have all the exam passed
‘All the students have passed the exam.’
If the DP becomes antecedent of a restrictive relative clause, the quantifier strand-

ing is no longer possible (6b):
(6) a. Alle Studenten, die die Prüfung bestanden haben, wollen mit
all students that the exam passed have want with
dem Professor sprechen.
the professor speak
‘All the students that passed the exam want to speak with the professor.’
Chapter 7. Noun phrasal complements vs. adjuncts 
b. *Studenten, die alle die Prüfung bestanden haben, wollen mit

students that all the exam passed have want with
dem Professor sprechen.
the professor speak
Non-restrictive relative clauses, instead, allow for a stranded quantifier because

the head is a full DP (5b). Moreover, a stranded quantifier forces the non-restric-
tive reading of the relative clause, as the minimal pair (7a–b) shows.
(7) a. Die ausländischen Studenten, die alle im Wohnheim wohnen,

the foreign students who all in the dorm live
müssen pünktlich zu Hause sein.
must on time at home be
b. Die ausländischen Studenten, die im Wohnheim
the foreign students that in the dorm
wohnen, müssen pünktlich zu Hause sein.
live must on time at home be
‘The foreign students, who live in the dorm, must be home on time.’
This is possible only if we consider that the determiner of the antecedent of a

restrictive relative clause is the only part of the antecedent that does not directly
belong to the clause. According to Cinque (2008: 6), the noun of an Italian restric-
tive relative clause does not show any definiteness and intersects its value with
external material (which is D). Here we will attempt to see if this can be extended
to German relative clauses as well. If the definite article precedes the noun in a
declarative sentence (8a), the non-specific reading we have with an indefinite arti-
cle (8b) is of course excluded. The indefinite article enables both the specific and
non-specific reading. In (8b), the speaker may refer either to a specific actor he/she
has in mind, or to any actor in general.
(8) a. Ich weiss, dass der bekannte Schauspieler auf die Party
I know that the famous actor at the party
kommen wird
come will
b. Ich weiss, dass ein bekannter Schauspieler auf die Party

I know that a famous actor at the party
kommen wird.
come will
‘I know that the/a famous actor will come to the party.’
 Rossella Resi
c. Der bekannte Schauspieler, der auf die Party kommen

the famous actor that at the party come
wird, wird bestimmt sehr elegant aussehen.
will will of course very elegant look
‘The famous actor that will come to the party will be of course very
elegant.’
It is interesting to see that the definite article acquires a non-specific reading if a

restrictive relative clause follows (8c). According to Cinque (2008: 7), the non-
specific reading of the relative clause’s head (which has no article) is not lost.1 The
Head Raising Analysis proposed by Kayne (1994) was a brilliantly innovative intu-
ition, even though, as proposed in Resi (2011), it should not have been generalized
to include all kinds of relative structures. Only in restrictive relative clauses is the
noun the internal head, which raises to the specCP without determiner. It is a bare
noun that merges with an external determiner once the movement has already
taken place and the CP is completely merged (Kayne 1994: 86). This is the reason
why a restrictive relative clause is not affected by the kind of determiner, which
precedes the noun that has to be restricted by the relative clause. The relationship
between das and Buch in an isolated nominal phrase (9a) and the relationship
between das and Buch when a relative clause follows (9b) has a completely differ-
ent syntactic derivation.
(9) a. [N0+ restrictive relative clause] =
[buch + das ich t gelesen habe]
book that I read aux.1sg
b. [DP [N0+ restrictive relative clause]] =
[das [Buch, das ich gelesen habe]]
the book that I read aux.1sg
‘the book that I have read’
On the other hand: nouns introducing a non-restrictive relative clause form a

constituent with the preceding determiner analogously to a non-relativized DP/
NP. They are grammatical both on their own and when they are relativized, and
their meaning does not change in either case. This is not so with restrictive rela-
tive clauses where the head noun, modified by a restrictive relative clause, cannot
be the only direct complement of the determiner as it is generated inside the CP
1. He even suggests that a silent indefinite article raises between the determiner and the
noun: Der [ein] bekannte Schauspieler, der auf die Party kommen wird, wird bestimmt sehr
elegant aussehen (Cinque 2008: 7).
Chapter 7. Noun phrasal complements vs. adjuncts 
of the relative clause. I suggest that the entire CP (including the internal head) is
complement of the lexical head D0 and that the determiner is the external element
selecting the restrictive sentence (Figure 2).
DP DP
D DP CP
D0 CP
Figure 2. Syntactic derivation of RRCs and NRRCs
It is worth considering if a restrictive relative clause also behaves like a com-

plement of the D0 lexical head it refers to. And if this is really the case, what about
non-restrictive relative clauses? What is the exact hierarchical location for non-
restrictive relative clauses?
3.3 Complementation and adjunction

Following Williams’ (1980) idea, Meinunger (2000) suggests that co-indexing
between two elements necessarily requires a c-command relationship, and that
CP is sister of the lexical head D0 or V0. In fact, everything inside the CP is
c-commanded by the external lexical head which is a V0 in the case of phrasal
complements of verbs or, as we have proposed, a D0 in the case of relative clauses.
If we assume this, we would expect, following our analysis, that everything inside
a restrictive relative clause is in the domain of the matrix clause. On the other
hand, a non-restrictive relative clause is next to the head DP but, as syntactic tests
will show, neither is c-commanded by anything inside the matrix clause nor is
c-commanded by the DP itself. There is evidence that non-restrictive relative
clauses merge much later in the derivation of the sentence than restrictive relative
clauses, and that they function as adjunction to the maximal projection of DP/NP.
The first piece of evidence is the fact that the external head is a maximal pro-
jection and adjunction can in fact only add to maximal projections. We see that
the CP of a non-restrictive relative clause behaves like phrasal adjunctions with
respect to some syntactic phenomena. We assumed from the beginning that a
restrictive relative clause and its head are linked by a head-argument-relationship
similar to that of verbs selecting their arguments. We will try now to provide evi-
dence of both syntactic derivation and see if the complement/adjunction distinc-
tion suits sentences in the nominal domain as well as in the verbal domain.
The idea that adjuncts to verbs can be merged late in the derivation has been
used extensively and is not particularly controversial. The reason is that adjuncts
 Rossella Resi
are exempted from the theta-criterion, which is the condition that forces comple-
ments to be inserted in the derivation as early as the predicates they receive a
theta–role from. So the key criterion to distinguish arguments from adjuncts in the
verbal domain is that only arguments obey the theta–criterion and we suggest that
this difference can be implemented in the restrictive/non-restrictive dichotomy.
First of all I will show that variable binding, licensing of polarity items, and the
isle condition correspond for restrictive relative clauses to those of a complement
rather than an adjunction. A VP-internal position like the indirect complement,
jedem in (10a) must be able to c-command the internal head and consequently the
pronoun er within the relative clause. In fact in (10a) the pronominal variable er is
regularly bound with the quantifier of the matrix clause. If the subordinate clause
were an adjunction, this would not be possible, as in (10b). The non-restrictive
relative clause in (10c) behaves exactly like an adverbial clause in adjunct position.
Examples are taken from Haider (1997: 130).
(10) a. Hast du jedemi die Details genannt, an denen eri
have you everyone the details told in that he
interessiert war?
interested was
‘Have you sent the details to anyone who is interested?’
b. *Hast du jedemi die Details genannt, nachdem eri die
have you everyone the details told, after which he the
Situation erklärt hat?
situation explained has?
c. *Hast du jedemi die Details genannt, an denen
have you everyone the details told in which
eri übrigens interessiert war?
he by the way interested was?
A second piece of evidence in favour of the presence of a c-command relation

between the head and the material internal to the restrictive relative clause (which
proves the fact that a restrictive relative clause is a complement of its external
head D0), derives from the behavior of polarity items. These are licensed from the
external head material, as is usually the case for heads and the material internal
to its complement. If the antecedent of a restrictive relative clause has a negative
determiner (11a, c), this should be able to license a polarity item inside its comple-
ment CP (the restrictive relative clause).
(11) a. Kein, Schuler der in den Ferien jemals gelernt hat,...
no pupil that on holiday never learned has
‘No students that studied during the holiday…’
Chapter 7. Noun phrasal complements vs. adjuncts 
b. *Der Schuler, der in den Ferien jemals gelernt hat,…

the pupil that on holiday never learned has
c. Keine Frau, die Julia sonderlich gut kennt, kann
no woman, that Julia particularly good knows can
gut kochen.
good cook
‘No woman that Julia knows particularly well how to cook.’
d. *Die Frau, die Julia sonderlich gut kennt, kann
the woman that Julia particularly good knows can
gut kochen.
good cook
Our expectations on the c-command relationship between the external D0 and the
material internal to CP were right. Sonderlich (11c) and jemals (11a) are allowed
only if the head is a negative determiner because they belong to its c-command
domain. The result of this analysis is not surprising if we consider our suggestion;
a complement is always in the domain of the head from which it has been selected
exactly like the CP-internal material is in the domain of the D0 which selects it.
3.4 Central and peripheral clause

So far we have distinguished two types of relative clauses, according to their index-
ing, antecedent and of course relationship with the matrix clause; additionally, we
have proposed a configuration where these two subordinate clauses, depending
semantically on noun phrases, actually can be aligned syntactically with classi-
cal subordinate clauses depending on verbs. In particular, we have suggested that
restrictive relative clauses are complements and non-restrictive relative clauses are
adjunctions, the first being attached to a lexical head D0 which projects its argu-
ment and the latter being merged to a maximal projection DP. This analogy with
verb transitivity is supported by another syntactic consideration with respect to
the moment of merge of the two structures. As complement, a restrictive relative
clause should merge inside the matrix clause at the time when its head projects it,
while a non-restrictive relative clause should merge later in the derivation when
the entire matrix clause has already been projected.
According to Haegeman’s (2002, 2004a–b, 2006, 2010) analysis, subordinate
clauses depending on verbs in general do not have the same syntactic derivations
and do not behave homogeneously with respect to the matrix clause. Two main
subtypes can be syntactically distinguished with respect to their degree of subor-
dination to the associated clauses, the timing of merge and the consequent internal
structure, which may or may not convey independence from the speech act of the
 Rossella Resi
sentence. In particular, central clauses are merged with the associated clause early
in the derivation of the sentence. Specifically, they are merged before IP is com-
pleted. Peripheral clauses are adjoined after the associated CP has been projected.
Central clauses are part of the speech of the matrix clauses while peripheral clauses
have their own illocutory force. In the latter case, we have two different speech
acts. As a consequence of their external syntactic properties, these two types of
subordinate clause also differ in their internal structure and, in particular, in the
complexity of the CP domain.
Haegeman (2004b) first tests the dichotomy of central and peripheral clauses
on English conditional clauses, and secondly on other types of subordination,
including complement clauses. As far as complement clauses are concerned,
Thurmair (1989: 74ff), Haegeman (2002: 159f) and Meinunger (2000: 206ff) make
a distinction between complement clauses depending on verba dicendi and the
other types of complement clauses. They suggest that properties of central clauses
only occur in the first type of complement clauses, which Meinunger calls factive
complement clauses and represents with the structure (1a) that was provided at
the beginning of the present article. Adverbial subordinate clauses may also be
central or peripheral according to various criteria. What is more, German subor-
dinate clauses (complement and adverbial clauses) do not behave homogeneously
and display the same differences for most of the relevant criteria proposed by
Haegeman (Coniglio 2011).
In this chapter, the criteria for external and internal syntax proposed by
Haegeman (2000, 2004b, 2010) for English and by Coniglio (2011) for German
subordinate clauses will be briefly presented and applied, where possible, to
German restrictive and non-restrictive relative clauses. From a semantic point
of view, the analogy between peripheral and non-restrictive relative clauses is
straightforward. Non-restrictive relative clauses and matrix clauses constitute two
different illocutory speech acts. This type of relative clause provides background
information for the main assertion but it is are not relevant for processing the
proposition expressed in the matrix clause like it is for central clauses. Restrictive
relative clauses, on the other hand, contribute to the proposition expressed in
the associated clause and are necessary. The restrictive relative clause helps iden-
tify the nominal group of the matrix clause. The non-restrictive relative clause
describes a context but is not essential. Does this semantic analogy have a syntac-
tic counterpart?
3.4.1 Differences in external syntax

As far as external syntax is concerned, Haegeman (2000, 2004b, 2010) lists syn-
tactic properties on the basis of English examples that account for the difference
between central and peripheral clauses. As we said before, a central clause merges
Chapter 7. Noun phrasal complements vs. adjuncts 
earlier in the derivation of the matrix clause while peripheral clauses merge later.
Consequently, the time of the central clauses depends on the time of the matrix
clauses (12a), while the tense of the associated clauses does not affect the time line
of the peripheral clause (12b) at all. Examples are taken from Haegeman (2002:
123–4).
(12) a. If your back-supporting muscles tire (future), you

will be at increased risk of lower-back pain.
b. If we are short of teachers […], why don’t we send
our children to Germany to be educated? […]
This aspect is not directly evident in relative clauses because they depend on
nominal elements of the associated clause, which do not have the time property.
However, it seems that relative clauses do bear covert time properties, which is
evident when we consider prenominal participial construction. Only a restrictive
relative clause can be converted into a prenominal participial construction with an
implicit verb (13a–b).
(13) a. Der Mann, der an Krebs gestorben war, war noch

the man that from cancer died was was still
sehr jung.
very young
‘The man that died of cancer was very young’
b. Der an Krebs gestorbene Mann war noch sehr jung.
The from cancer died man was still very young
c. Paul, der an Krebs gestorben war, war noch sehr jung.
Paul that from cancer died was, was still very young
d. *Der an Krebs gestorbene Paul war noch sehr jung.
thee from cancer died Paul was still very young
This is also evident in Italian, where only restrictive relative clauses can have an
infinitive verb. (14) can have only a restrictive reading.
(14) le cose da mangiare (= le cose che devono/possono essere mangiate)

things to eat (= things that must/can be eaten)
If central clauses are said to be within the scope of temporal operators in the asso-
ciated clauses we can suppose that the reason that only restrictive relative clauses
can be placed prenominally without overtly expressing the finiteness of the verb is
because their time depends on the time of the matrix clause, and is therefore not
necessary. Since non-restrictive relative clauses have their own time reference, this
 Rossella Resi
must always be expressed as in peripheral clauses, where tense interpretation is

independent (Haegeman 2004b, 2010).
Moreover, central clauses are said to be within the scope of operators in the
matrix clause because they are c-commanded by material internal to the matrix
clause. We have seen in 3.3 that the c-command relationship is suitable for restrictive
relative clauses, but not for non-restrictive relative clauses, which, on the contrary,
are completely outside the scope of the operators. Polarity items (Examples 11) and
binding operators (Examples 10) only have an effect on c-commanded material.
Another syntactic phenomenon, which should account for this difference, is
licensing of parasitic gaps. In English, parasitic gaps are licensed only with central
clauses because the gaps need to be c-commanded by the matrix clause ( Haegeman
2002, 2004a, 2010). Unlike English, the parasitic gap in German may never appear
in a complement clause, even in a factive one, as we would expect if we extend
Haegeman’s analysis on the German language. It has even been suggested that par-
asitic gaps may not exist in the German language and that instead, the construc-
tion that appears to resemble a parasitic gap construction is actually an instance of
forward deletion (Fanselow 2001). Sabel (1996) also suggests that what look like
parasitic gaps in German are actually ‘pseudo-parasitic gaps’ in the sense of Postal
(1994). Whether or not parasitic gaps exist in German, it is interesting to note that
against any expectation, they are not licensed in factive complement clauses.
(15) a. *Weri hat Kevin ti überredet [CP dass wir ei anrufen sollen]?
who has Kevin convinced that we call should
‘Who convinced Kevin, that we should call?’
b. *Wemi hast du ti gesagt [CP dass wir am Sonntag ei
who have you said that we on Sunday
treffen würden]
meet would
‘Who did you say we would meet on Sunday?’
If the analogy we have drawn so far is correct, we would expect that restrictive
relative clauses in Standard German, being central clauses, do not license parasitic
gaps either. If relative clauses behave like central clauses in German, they should in
fact also respect the syntactic behavior of central clauses with respect to the possi-
bility of licensing parasitic gaps. Examples (16a) and (16b) meet our expectations.
(16) a. *Dies ist ein Manni den Leute [die ei treffen] ti
this is a man that people that meet
echt mogen.
truly like
‘This is a man, that people really like.’
Chapter 7. Noun phrasal complements vs. adjuncts 
b. *Welches Autoi mogen Leute [die ei fahren] ti?

Which auto like people that drive
‘Which car do people like?’
While this analysis is based on Standard German, an examination of more than

one dialect of German would be beyond the scope of this work; it is, how-
ever, worth noticing that the parasitic gap construction appears in certain simi-
lar contexts in Bavarian. The following data on Bavarian are taken from Felix
(1985: 6), where (17a) is grammatical in the Bavarian variation but ungram-
matical in Standard German (17b).
(17) a. Das ist eine Fraui, diei wenn ei etwas verspricht hält ei
that is a woman who when something promises keeps
es auch.
it also
‘She is a woman that keeps the promises she makes.’
b. *Das ist eine Fraui diei wenn ei etwas verspricht hält ei es auch.
For the sake of parallelism between central and restrictive relative clauses, we
expect that a parasitic gap is licensed in the types of central clauses described for
English by Haegeman (2002 and following works). For example, according to
Bayer (1984), conditional central clauses in Bavarian license parasitic gaps (18).
(18) Den, wann i ei derwisch, derschlog i ei.
him if I catch slay I
‘I slay him if I catch him.’
3.4.2 Differences in internal syntax

As far as internal syntax of central and peripheral clauses is concerned, Haege-
man lists other criteria, which influence how far the two different speech acts
(associated clause and subordinate clause) are independent and, as a result of that,
how complete their internal CP-domain is. According to Haegeman, peripheral
clauses, like root clauses, have a much more complete left periphery than central
clauses and they are therefore much more independent with respect to the matrix
clause than central clauses are. The latter lack projections in the CP domain, which
encode illocutory force and therefore the possibility of anchoring the utterance
to the speaker’s opinion. This is evident from the fact that peripheral clauses may
contain elements of epistemic modality or elements, like modal particles, which
relate the utterance to the speaker.
These speaker-oriented elements must depend on the presence or absence of a
projection with these properties. ForceP encodes illocutory force, which is a prop-
erty of root like clauses, independent speech time and epistemic modality. As far
 Rossella Resi
as speech time is concerned, central clauses do not need this projection because
their time depends on the time of the matrix clause. They are allowed to have a
non-finite verb and occur in prenominal position as participial construction. To
describe this difference, Haegeman adopts Rizzi’s (1997, 2001, 2004) well-known
theories on the fine structure of the CP. Central clauses present a reduced CP com-
pared to that of peripheral clauses.
Clauses displaying a reduced structure not only cannot license fronted argu-
ments and focused elements (Haegeman 2002, 2004a, 2006, 2010), but they can-
not contain modal particles either, since they cannot licence root phenomena at
all. Modal particles can only occur in those contexts that, according to Haegeman
(2002, 2004a, 2006, 2010), display a full left periphery and thus root proper-
ties. They are banned from non-root contexts since the latter do not constitute
independent speech acts, and also from central clauses. We therefore expect that
they cannot occur in restrictive relative clauses (19a-b). Examples are taken from
Zimmermann (2004: 32).
(19) a. *Eine Kollegin, die ja in Syracuse wohnt, wird kommen.

A collegue that ptcl in Siracuse lives, will come.
b. *Die Firma sucht einen Angestellten, der ja immer
The company looks for an employee that ptcl always
pünktlich ist
on time is.
c. Peter, der ja nichts verpassen will, ist immer dabei.
Peter, who ptcl nothing miss want is always present
‘Peter, who does not want to miss anything is always present.’
Non-restrictive relative clauses, which are far more independent than restrictive
relative clauses, have no problem with modal particles and epistemic modality.
The use of modal particles can even disambiguate the reading of German relative
clauses and influence the grammaticality of the proposed sentence. (20b) can only
have a non-restrictive reading because of the presence of a projection that hosts
the modal particle ja.
(20) a. Autos, die laut sind, sollten mit einer

autos that loud are should with a
geschlossenen Motorkapsel versehen werden.
closed motorcapsule provided be
‘Loud cars should be provided with a motorcapsule.’
Restrictive reading:
Chapter 7. Noun phrasal complements vs. adjuncts 
Nur diejenigen Autos, die laut sind…

only those autos that loud are
Non restrictive reading:
Alle Autos, die übrigens laut sind…
all autos which by the way loud are
b. Autos, die ja laut sind, sollten mit einer geschlossenen
autos that ptcl loud are, should with a closed
Motorkapsel Motorkapsel versehen werden.
motorcapsule motorcapsule provided be
Non restrictive reading:
Alle Autos, die übrigens laut sind…
all autos which by the way loud are
On the one hand we have root-clause-like embedded clauses with a full structure,
which are also endowed with illocutory force; on the other hand there are embed-
ded clauses displaying a reduced CP domain without illocutory force, which
depend on the matrix clause as far as the anchoring of force to the speaker is con-
cerned. They do not have a ForceP projection on their own (Figure 3).
Central clauses Sub ModP FinP
Peripheral clauses Sub ForceP TopP FocusP ModP FinP
Main Clauses ForceP TopP FocusP ModP FinP
Figure 3. Internal syntax
3.4.3 Implementation of Haegeman’s analysis

Due to the previous analysis we can see that there are syntactic phenomena that
account for an implementation of Haegeman’s distinction to relative clauses both
for external and internal syntax. The time dependency and the possibility of some
elements of scoping over the relative clause account for a different time in merging
with the associated clause. Availability of epistemic modality, licencing of modal
particles or, in general, the lack of a position that guarantees anchoring to the
speaker and which is usually projected in root clauses, together account for a dif-
ferent internal structure. Restrictive relative clauses have a shorter left periphery
than non-restrictive relative clauses, which, like root clauses, are provided with a
 Rossella Resi
full structure. It is by no means accidental that non-restrictive relative clauses can

be transformed into two different clauses where the second one refers anaphori-
cally to the first one through a pronoun (21b). This is not possible for restrictive
relative clauses (21d).
(21) a. Peter ist ein Freund von meinem Bruder, der übrigens auf
Peter is a friend of my brother, that by the way to
die Party kommt.
the party comes.
‘Peter is a friend of my brother, who by the way comes to the party.’
b. Peter ist ein Freund von meinem Bruder. Er kommt
Peter is a friend of my brother he comes
auf die Party.
to the party.
c. An der Konferenz hat jeder Professor teilgenommen,
at the conference has every professor taken part,
der eingeladen wurde.
that invited was.
‘Every invited professor came to the conference.’
d. *Jeder Professor hat an der Konferenz teilgenommen.
Every professor has at the conference taken part.
Er wurde eingeladen.
He was invited.
What is interesting for our analysis is that relative clauses have two different syn-
tactic structures. Relative clauses, like central clauses, are embedded in the matrix
clause and, as we have seen, they are selected as complement of a lexical head
(22a). The restrictive relative clause is integrated in the main clause. It is merged
within the IP of the matrix clause and the nominal head belongs to the relative
clause while the external determiner governs over the entire clause as if it were its
complement (Resi 2011). Non-restrictive relative clauses are adjunctions to maxi-
mal projections NP/DP and they are merged later in the derivation when the IP
has already derived, most probably as adjunctions (22b).
(22) a. Das [CP[NPBild], [das [Peter tNP gesehen hat]]]
the picture that Peter seen has
‘The picture that Peter has seen.’
b. [DP Mein Vater], [CP<mein Vateri> den du ti (übrigens)
my father my father that you by the way
schon kennengelernt hast]
already known has
‘My father, who you, by the way, already know.’
Chapter 7. Noun phrasal complements vs. adjuncts 
The non-restrictive relative clause is integrated in the matrix clause, but it is merged
or adjoined to the CP of the associated clause in a structure which resembles coor-
dination rather than subordination. The external head, which is the entire NP or
DP, is not part of a syntactic chain.
I therefore propose that restrictive relative clauses, like central clauses, are
merged at an earlier point of the derivation than non-restrictive or periph-
eral clauses. Restrictive relative clauses are merged within the matrix clause;
non-restrictive relative clauses are adjoined once the entire antecedent is fully
projected.
4. Conclusion
We have seen that some syntactic differences between restrictive and non-restric-
tive relative clauses can derive (a) from the fact that the former is a modifier and
the latter is not, (b) that one is c-commanded by the matrix clause and the other
is not, (c) that one is the complement of a functional head while the other has
a full maximal projection as antecedent, and (d) that one is central while the
other is peripheral. All these properties account for a close analogy with the way
that subordinate clauses depend on verbs. In the case of central clauses, comple-
ment clauses and restrictive relative clauses, we can apply the original concept of
valency as they are considered arguments selected by a lexical head. A lexical ver-
bal head projects a phrasal argument as complement while a determiner projects
a restrictive relative clause as complement. Proper names, which are already dis-
cursively identified, are not allowed to have complements. The clause is, instead,
like an adjunct sentence, which provides a background to the main assertion
and conveys additional information that is not directly relevant. Adjunction, like
peripheral clauses and non-restrictive relative clauses, can also be asserted sepa-
rately from the matrix clause like an independent clause, displaying discourse
anaphora.
It is generally assumed that complements always precede adjuncts in German
(23a–b) because adjunctions are merged higher than complements, resulting in
a linear order where complements always precede adjuncts (Plaztack 2000: 265).
(23) a. der Freund von Peter mit blauen Augen

the friend of Peter with blue eyes
b. *der Freund mit blauen Augen von Peter
the friend with blue eyes of Peter
We would therefore expect a CP adjunct to follow a CP complement and an oppo-

site linear order to be ungrammatical.
 Rossella Resi
(24) a. Der Mann der uns gestern zum Mittagessen Eingeladen hat,
The man that us yesterday for lunch invited has,
und übrigens 35 Jahre alt ist, kommt aus Australien.
and by the way 35 years old is, comes from Australia.
‘The man that invited us yesterday for lunch and, by the way, is 35 years
old, comes from Australia.’
b. *Der Mann der übrigens 35 Jahre alt ist, und uns gestern
The man who by the way 35 years old is, and us yesterday
zum Mittagessen eingeladen hat, kommt aus Australien.
for lunch invited has, comes from Australia.
As in (24a), a restrictive relative clause precedes a non-restrictive relative clause

while the opposite order (24b) is ungrammatical, which confirms our hypothesis.
References
Alexiadou, Artemis. 2001. Adjective syntax and noun raising: Word order asymmetries in the
DP as the result of adjective distribution. Studia Linguistica 55: 217–248. DOI: 10.1111
/1467-9582.00080
Bayer, Josef. 1984. COMP in Bavarian syntax. The Linguistic Review 3: 209–274. DOI: 10.1515/
tlir.1984.3.3.209
Bianchi, Valentina. 2000. Some issues in the syntax of relative determiners. In The Syntax of
Relative Clauses [Linguistik Aktuell/Linguistics Today 32], Artemis Alexiadou, Paul Law,
André Meinunger & Chris Wilder (eds), 53–81. Amsterdam: John Benjamins.
Cinque, Guglielmo. 2008. More on the indefinite character of the head of restrictive relatives.
Rivista di grammatica generativa 33: 3–24.
Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln: ihre Distribution und Lizen-
zierung in Haupt- und Nebensätzen [Studia Grammatica 73]. Berlin: Akademie-Verl. DOI:
10.1524/9783050053578
Egg, Markus. 2007. The syntax and semantics of relative clause modiﬁcation. In Proceedings of
the Sixteenth Computational Linguistics in the Netherlands, Khal’il Simanan, Maarten de
Rijke, Remko Scha & Rob van Son (eds), 49–56. Amsterdam: University of Amsterdam.
Emonds, Joseph. 1979. Appositive relatives have no properties. Linguistic Inquiry 22: 1–25.
Fabb, Nigel. 1989. The difference between English restrictive and nonrestrictive relative clauses.
Journal of Linguistics 26: 57–78. DOI: 10.1017/S0022226700014420
Fabricius-Hansen, Cathrine. 2009. Überlegungen zur pränominalen Nicht-Restriktivität. In
Koordination und Subordination im Deutschen [Linguistische Berichte, Sonderheft 16],
Veronika Ehrich, Christian Fortmann, Ingo Reich & Marga Reis (eds), 89–112. Hamburg:
Buske.
Fanselow, Gisbert. 2001. Features, theta-roles, and free constituent order, Linguistic Inquiry 32:
405‒436. DOI: 10.1162/002438901750372513
Chapter 7. Noun phrasal complements vs. adjuncts 
Felix, Sascha. 1985. Parasitic gaps in German. In Erklärende Syntax des Deutschen, Werner
Abraham (ed.), 173–200. Tübingen: Narr.
Haegeman, Liliane. 2002. Anchoring to speaker, adverbial clauses and the structure of CP.
Georgetown University Working Papers in Theoretical Linguistics 2: 117–180.
Haegeman, Liliane. 2004a. Topicalization, CLLD and the left periphery. In Proceedings of the
Dislocated Elements Workshop vol. 1 [ZAS Papers in Linguistics 35], Benjamin Shaer,
Werner Frey & Claudia Maienborn (eds), 157–192. Berlin: ZAS.
Haegeman, Liliane. 2004b. The syntax of adverbial clauses and its consequences for t opicalization.
In Current Studies in Comparative Romance Linguistics [Antwerp Papers in Linguistics
107], Martine Coene, Gretel De Cuyper, Yves D’Hulst (eds), 61–90. Antwerp: University
of Antwerp.
Haegeman, Liliane. 2006. Conditionals, factives and the left periphery. Lingua 116: 1651–1669.
DOI: 10.1016/j.lingua.2005.03.014
Haegeman, Liliane. 2010. The internal syntax of adverbial clauses. Lingua 120: 628-648. DOI:
10.1016/j.lingua.2008.07.007
Haider, Hubert. 1997. Extraposition. In Rightward Movement [Linguistik Aktuell/Linguistics
Today 17], Dorothee Beerman, David LeBlanc & Henk van Riemsdijk (eds), 115–151.
Amsterdam: John Benjamins.
Kayne, Richard S. 1994. The Antisymmetry of Syntax [Linguistic inquiry Monographs 25].
Cambridge MA: The MIT Press.
Meinunger, André. 2000. Syntactic Aspects of Topic and Comment [Lingustik Aktuell/Linguistics
Today 38]. Amsterdam: John Benjamins. DOI: 10.1075/la.38
Platzack, Christer. 2000. A complement of n° account of restrictive and non-restrictive
relatives: The case of Swedish. In The Syntax of Relative Clauses [Linguistik Aktuell/Lin-
guistics Today 32], Artemis Alexiadou, Paul Law, André Meinunger & Chris Wilder (eds),
Postal, Paul M. 1994. Parasitic and pseudoparasitic gaps. Linguistic Inquiry 25(1): 63–117.
Resi, Rossella. 2011. The position of relative clauses in German. Lingue e Linguaggio 1: 87–118.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar. A Handbook
of Generative Grammar [Kluwer International Handbooks of Linguistics 1], Liliane Haege-
man (ed.), 281–337. Dordrecht: Kluwer.
Rizzi, Luigi. 2001. On the position Int(errogative) in the left periphery of the clause. In Cur-
rent Studies in Italian Syntax. Essays Offered to Lorenzo Renzi [North-Holland Linguistic
Series: Linguistic Variations 59], Guglielmo Cinque & Giampaolo Salvi (eds), 287–296.
Amsterdam: Elsevier.
Rizzi, Luigi. 2004. Locality and left periphery. In Structures and Beyond [The Cartography of
Syntactic Structures 3], Adriana Belletti (ed.), 223–251. Oxford: OUP.
Sabel, Joachim. 1996. Asymmetries partial wh-movement. In Papers on Wh-Scope Marking
[Arbeitspapiere des Sonderforschungsbereichs 340, 76], Uli Lutz & Gereon Müller (eds),
289–315. Stuttgart: University of Stuttgart.
Thurmair, Maria. 1989. Modalpartikeln in ihre Kombination [Linguistische Arbeiten 223].
Tübingen: Max Niemeyer. DOI: 10.1515/9783111354569
Umbach, Carla. 2006. Non-restrictive modification and backgrounding. In Proceedings of the
Ninth Symposium on Logic and Language, Beáta Gyuris László Kálmán, Chris Piñón &
Károly Varasdi (eds), 152–159, Budapest: Hungarian Academy of Sciences.
 Rossella Resi
Williams, Edwin. 1980. Predication. Linguistic Inquiry 11: 203–238.

Zimmermann, Malte. 2004. Discourse particles in the left periphery. In Proceedings of the
Dislocated Elements Workshop, Vol. 2 [ZAS Papers in Linguistics 35], Benjamin Shaer,
Werner Frey & Claudia Maienborn (eds), 254–566, Berlin: ZAS.
Ziv, Yael & Cole, Paul. 1974. Relative extraposition and the scope of definite description in
Hebrew and English. In Papers from the Tenth Regional Meeting of the Chicago Linguistic
Society, April 19-21, Michael W. La Galy, Robert A. Fox & Anthony Bruck (eds), 772–786.
Chicago IL: CLS.
chapter 8
Noun valency in Latin
Olga Spevak
University of Toulouse 2
This article has two objectives. The first is to present an account of valency nouns
in Latin. Lyons’ typology (1977) envisaging three orders of entities is useful
for predicting the number and type of complements used with various nouns.
Expansions of all the categories are distinguished: concrete entities, relational
nouns, agent nouns, verbal nouns, and nouns expressing qualities. Furthermore,
Latin shows interesting phenomena closely related to noun valency, namely
nominalization of verbal notions in Early Latin and the construction of the
dominant participle. The second objective is to examine argument marking at the
noun phrase level. The genitive is the “adnominal” case par excellence; other cases
(the dative, accusative, and ablative) as well as prepositional phrases are atypical
noun complements in Latin, and furthermore they are often restricted to specific
categories of nouns.
1. Aims and objectives
The aim of this contribution is to present an overview of several issues related

to noun valency, including the typology of valency nouns and a brief charac-
teristic of each category (Section 2), argument marking at the noun phrase
level (Section 3), and semantic shifts between verbs and nouns (Section 4). I
use examples from Latin, a language with case marking.1 The Latin case system
shows an interesting functional differentiation. In general – disregarding spe-
cial uses – the genitive marks adnominal complements; the dative is used for
recipient/addressee, beneficiary, and experiencer; the accusative marks on the
1. This article is based on a previous investigation of a series of individual words and their
behavior (Spevak 2014) with the help of the electronic database the Library of Latin Texts
(LLT). Digital corpora available for Latin lack specific annotations, thus making it possible to
proceed to a detailed corpus analysis.
 Olga Spevak
one hand direct objects, and on the other, extension in space and time; the abla-
tive is used for various circumstantial expressions. Furthermore, Latin provides
testimony of ancient “verbal” constructions of verbal nouns and their specific
use for the resumption of a state of affairs, which is presented in Section 2.2.1.
Participialization of state of affairs is a counterpart of verbal nouns, which repre-
sent nominalizations of states of affairs; the dominant participle construction as
a competing means of verbal nouns is the topic of Section 2.2.2. Several observa-
tions concerning Ancient Greek and Roman grammatical theory are presented
in an appendix (Epilogue).
2. The typology of valency nouns
Derivation usually serves as a criterion for the description of valency nouns (see
Panevová, this volume). I myself approach valency nouns using the concept of
orders of entities established by Lyons (1977: 442‒447) and combine it with
valency. Lyons distinguishes between spatial entities (first-order), temporal enti-
ties (second-order), and propositional content (third-order). Examples are given
in Table 1. Note that relational nouns (father) and agent nouns (judge), which
belong to the first-order entities, as well as various verbal nouns (arrival) and
abstract nouns (opinion), are valency nouns requiring an expansion. Although
the criterion of valency does not directly match the orders of entities, it is useful
to consider both parameters because their combination shows that (i) first-order
entities are zero-valent, except relational nouns and agent nouns; (ii) second-order
entities are mono- and bivalent, except nouns with a merely temporal meaning
(day, for example); and (iii) there are no zero-valent nouns among third-order
Table 1. Three orders of entities

First order Spatio-temporal entities: zero-valent:
persons, animals, and objects that boy, soldier, horse, book, ship
can be situated in space collective nouns: family
mass nouns: water
mono- and bivalent
relational nouns: father, son
agent nouns: judge, actor
Second order temporal entities: events, processes, zero-valent: day, hour
and states of affairs that can be located mono- and bivalent:
in time arrival, enjoyment, pain, sedition, love
Third order non-spatial and non-temporal entities: mono- and bivalent:
propositional content opinion, judgment, memory, suspicion
Chapter 8. Noun valency in Latin 
entities. Additionally, as is shown below, some nouns can belong to more than one
order of entity.
Furthermore, there are other, monovalent nouns that seem to form a special
category: great number, crowd, part, amphora, jar when they are used as expres-
sions of measure. They are called “containers” (see Panevová, this volume).
Distinguishing orders of entities makes it easier to understand the types of
complement that nouns take: first-order entities mostly combine with expres-
sions of possession, second- and third-order entities take expressions of agent or
patient, and third-order entities admit complements with the form of a clause. In
sum, valency nouns thus fall into the following categories:
1. first-order relational nouns (father);

2. first-order agent nouns (judge);
3. second-order nouns denoting various states of affairs (verbal nouns) and
abstract qualities;
4. third-order nouns (verbal nouns);
5. “containers” including nominal quantifiers (part, amphora, great number),
“classifiers” (type, sort), and similar expressions.
2.1 First-order nouns

Zero-valent nouns are sufficient on their own from the semantic point of view. For
example, book, horse, house, and letter do not need any complement in order to be
used in an utterance. Therefore, a genitive complement marking the entity repre-
sented, for example signum Minervae ‘a statue of Minerva’ in (1) or a genitive of
the author in (2), are optional, because they are not implied by the semantic value
of signum (Pinkster 1990/1995: § 6.6; cf. Panevová, this volume). Expressions of
ownership (alienable possession), which represent “prototypical possession” in the
sense that the possessed object (possessum) is a property to which the possessor
has a legal right (Heine 1997: 34), also function as satellites (3). Expressions com-
peting with the possessive genitive, i.e. possessive pronouns and adjectives derived
from proper names (4), have the same status.2
(1) Minervae signum ex ebore pulcherrimum
Minerva-gen statue-nom from ivory-abl very nice-nom
‘the very nice ivory statue of Minerva’ (Cic. Brut. 257)3
2. However, possessive genitives are not interchangeable with the personal name adjectives
in Latin, unlike in Slavic languages (Corbett 1995). For the situation in Latin, see Baldi & Nuti
(2010: 356).
3. Abbreviations of Latin authors and their works follow the Oxford Latin Dictionary.
 Olga Spevak
(2) signa Praxiteli

statues-nom.pl Praxiteles-gen
‘Praxiteles’ statues’
(3) domus Pomponii
house-nom Pomponius-gen
‘Pomponius’ house’
(4) domus mea / Rabiriana
house-nom my-poss.nom Rabirius- adj.poss.nom
‘my/Rabirius’ house’
Genitives expressing matter, origin, price, or content (5) are not required by the
sense of their governing nouns. The same holds for an adjectival modifier such as
pulcherrimum ‘very nice’ and a prepositional phrase expressing material (ex ebore
‘ivory’) in (1).
(5) liber legum
book-nom laws-gen.pl
‘book of laws’
The optional character of these complements is manifested at the syntactic level.

Complements not required by valency are more loosely related to their govern-
ing nouns than valency complements are (Bolkestein 1989: 15). In Latin, this
explains the difference between caedes Clodi ‘the murder of Clodius’ (objective
genitive, argument), where Clodi cannot be used predicatively (6a), and the pos-
sessive genitive (satellite) in domus Clodi ‘Clodius’ house’, where Clodi can be used
predicatively (6b) and, furthermore, competes with the dative Clodio (cf. Dressler
1970: 27).
(6) a. caedes Clodi
murder-nom Clodius-gen
‘the murder of Clodius’
*caedes est Clodi
murder-nom is-3sg.prs Clodius-gen
*‘the murder is of Clodius’
b. domus Clodi
house-nom Clodius-gen
‘Clodius’ house’
domus est Clodi / Clodio
house-nom is-3sg.prs Clodius-gen Clodius-dat
‘the/this house belongs to Clodius’
Among the first-order entities, relational nouns and agent nouns are monovalent
or bivalent: the complements they take are required by their semantic value. In this
Chapter 8. Noun valency in Latin 
Section, I use the abbreviations N1 for the first argument (the agent), and N2 for
the second argument (mainly, the patient).
Kinship nouns and nouns referring to body parts and (some) personal
properties – they represent relational nouns4 expressing inalienable possession –
encode the possessor as a genitive complement or a possessive pronoun. Two
examples with kinship nouns are given in (7).
(7) a. uxor Cleomeni Syracusani
wife-nom Cleomenes-gen Syracusan-gen
‘the wife of Cleomenes the Syracusan’ (Cic. Ver. 5.82)
b. filia mea
daughter-nom my-poss.nom
‘my daughter’
The valency frame of relational nouns can be represented as:

uxor + POSSESSION gen. N1/possessive pronoun.
The syntactic form of complementation is thus the same as for the expressions of
ownership mentioned above. In other words, there is no formal difference between
alienable and inalienable possession in Latin.5 However, a distinction between a
closer and a more distant possession relationship is manifested at a different level,
that of expression and non-expression of the possessor. In Latin, closer a pos-
session relationship is normally related to the subject of the sentence (or possi-
bly, to the speaker) and remains unexpressed. Explicit expression of the possessor
either is required for avoiding ambiguity between two possessors or else entails an
emphatic or contrastive interpretation.6
Agent nouns are monovalent or bivalent. Bivalent agent nouns take genitive
complements with the semantic function of patient. For example, the genitive
complement factorum et scriptorum meorum ‘of my doings and writings’ of the
agent noun laudator ‘encomiast’ in (8a) corresponds to the clausal expression
given in (8b), where this element functions as direct object. Instead of the genitive,
a possessive pronoun may be used for encoding the patient, such as noster ‘our’ in
amator noster ‘our admirer’. As pointed out by Panevová (this volume), the pecu-
liarity of agent nouns is the fact that they incorporate the semantic role of agent,
4. Certain linguists use the term “relational” in a larger sense for “bivalent nouns” in contrast
with “absolute” (i. e. zero-valent) nouns, especially Seiler (1983: 11) and Lehmann (1985: 72).
5. See Baldi & Nuti (2010: 347) and Chappell & McGregor (1996: 3–4).
6. For further details, see Lehmann (2005) and Spevak (2010: 251).
 Olga Spevak
N1. As a consequence of this, the agent does not appear in the valency frame that
can be formulated as follows:
laudator + PATIENT gen. N2/possessive pronoun.
(8) a. Tu..., laudator et factorum et

you-nom encomiast-nom and doings-gen.pl and
scriptorium meorum.
writings-gen.pl my-gen.pl
‘You..., the encomiast of my doings and writings.’ (Cic. Att. 7.1.4)
b. Atticus facta et scripta
Atticus-nom doings-acc.pl and writings-acc.pl
mea laudat.
my-acc.pl praises-3sg.prs
‘Atticus praises my doings and writings.’
2.2 Second-order nouns

Second-order entities (temporal entities) include a number of monovalent and
bivalent nouns: on the one hand, verbal nouns of various types; on the other,
abstract nouns expressing qualities. Nouns derived from verbs of movement
behave in a specific way and are treated separately here. Before presenting an over-
view of second-order nouns in Latin, I discuss two points that are important for
understanding the construction of verbal nouns in Latin: their function in Early
Latin, and the use of the dominant participle.
2.2.1 Verbal nouns in Early Latin

Verbal nouns are usually regarded as nominalizations of verbal notions that
denote states of affairs.7 Latin provides interesting evidence of this phenomenon.
In Early Latin, verbal nouns are typically used for nominalizing of states of affairs,
“as a Wiederaufnahme (‘resumption’) of a preceding, formally finite, verbal form”
(Rosén 1981: 17).8 This situation is shown in (9). The speaker, Periplectomenus,
is describing what the slave Palaestrio is doing while thinking up a plan for fur-
ther action. By using the noun aedificatio ‘action of building’ accompanied by the
7. For Latin, see especially Dressler (1970: 28), Rosén (1981: 71) and Fugier (1983: 247). See
also in this volume Enghels & Bekaert and Kolářová.
8. The resumptive function of verbal nouns (“Wiederaufnahme”) was first identified by
Porzig and developed in the 1938 dissertation by Seitz. Porzig himself published a monograph
on this topic in 1942.
Chapter 8. Noun valency in Latin 
anaphoric pronoun illaec, the speaker refers to the state of affairs that is taking
place: aedificat ‘he is building’.
(9) Ecce autem aedificat: columnam

look-interj but is building-3.prs pillar-acc
mento suffigit suo. Apage, non
chin-abl fix beneath-3.prs his-abl away with neg
placet profecto mi illaec aedificatio.
like-3.prs assuredly me-dat that-nom building-nom
‘But look, he is building something: he is supporting his chin with a pillar.
Away with that, I sure don’t like this way of building of his.’ (Pl. Mil. 209)
This aspect of Latin verbal nouns explains the use of the accusative complement
(hanc) referring to a young girl in (10a) instead of the adnominal genitive (huius
‘her’); furthermore, the ablative (digito) marks the instrument (satellite). The con-
struction of the verbal noun tactio is the same as that of its cognate verb tango ‘to
touch’ (10b).9
(10) a. Quid tibi hanc digito tactio est?
what you-dat this-acc finger-abl touching-nom is-3.prs
‘What right have you to touch this girl here with your finger?’
(Pl. Poen. 1308)
b. Quid hanc digito tangis?
what this-acc finger-abl touch-2.prs
‘Why do you touch her with your finger?’
In Early Latin, “verbal” construction of verbal nouns is found alongside genitive

marking. This property is lost around 100 BC (Rosén 1983: 201) together with the
decline of verbal nouns functioning as nominalizations. In Classical Latin, verbal
nouns with such a nominalizing function can still be found, but their construction
with the genitive is regular. Not only do verbal nouns henceforth take “nominal”
marking, the genitive, but a number of them also disappear from the vocabulary
(e.g. statio ‘action of standing’, aditio ‘approaching’, expectatio ‘awaiting’) or are
found with a specialized meaning (Rosén 1983: 76), as in (11). Although there is a
gradual increase in various abstract nouns from Classical Latin onward, they are
not productive for nominalizing of state of affairs; other means, especially ger-
unds, gerundives, and participles, are used instead.
9. For the construction of action nouns with an accusative in other ancient languages, see
Panagl (2006: 52).
 Olga Spevak
(11) a. quaestio ‘action of seeking’ (Early Latin, Pl. Trin. 1012) vs.
‘investigation’ (Classical Latin, Cic. Caec. 29)
b. curatio ‘action of taking charge’ (Early Latin, Pl. Poen. 354) vs.
‘(medical) treatment’ (Classical Latin, Cic. Inv. 1.6)
Additionally, some verbal nouns with “verbal” marking are found in periphrastic
constructions in Early Latin: with the verb sum ‘to be’, for example, in expectatione
esse (+ accusative) ‘to be in expectation’ (Pl. St. 283) or in (10), which indicate that
a state of affairs is in progress, and with support verbs (Rosén 1981: 131‒5). For
example, mentionem facio ‘to make mention’ is found not only with the regular
genitive but also with an accusative (Pl. Pers. 283), the case used for object mark-
ing with the verb memoro ‘to mention’ (Rosén 1981: 143).
2.2.2 Dominant participle

Verbal nouns compete with other means, such as supines, gerunds, gerundives,
and infinitives (Rosén 1981: 21). Another competitor is the so-called ab urbe con-
dita construction (Heick 1936) or “dominant participle”. It consists of participi-
alization of the predicate and expresses a factual process (Bolkestein 1981 and
Pinkster 1990/1995: § 7.4.7). In a sense, it represents a counterpart of nominaliza-
tion of the predicate, a phenomenon that yields verbal nouns.
A good example of a dominant participle is occisus dictator Caesar in (12). As
subject complement, the inanimate noun facinus ‘crime’ is used because its referent
is inanimate: the murder of Caesar, and not Caesar himself (Longrée 1995: 175).
(12) Occisus dictator Caesar pulcherrimum
killed-pst.ptc dictator-nom Caesar-nom beautiful-adj.nom
facinus videretur.
crime-nom seemed-3.pass.subjv.impf
‘The murder of Caesar, the Dictator, seemed a most beautiful crime.’
(Tac. An. 1.8.6)
The construction ante Carthaginem deletam ‘before the destruction of Carthage’
(lit. ‘before Carthage destroyed’) (13a) might, theoretically, interchange with the
verbal noun deletio ‘destruction’ (13b). However, deletio is not used in this way in
Classical Latin (see ThLL, s.v.). It seems likely that this happens because the verbal
noun deletio is not suitable for expressing the achievement of the process of deleo
‘to destroy’. The job is done by the past perfect participle, deleta ‘destroyed’. Addi-
tionally, there is another competitor for expressing a virtual process: the gerundive
delendus ‘to be destroyed’ (13c).
(13) a. ante Carthaginem deletam
before Carthage-acc destroyed-pst.ptc.acc
‘before the destruction of Carthage’ (Sal. Jug. 41.2)
Chapter 8. Noun valency in Latin 
b. ante Carthaginis deletionem

before Carthage-gen destruction-acc
c. ad Carthaginem delendam
to Carthage-acc destroyed-grdv.acc
Another illuminating example is (14): a dominant participle construction (de

proelio facto) is coordinated with the verbal noun oppugnatio ‘besieging’ accompa-
nied by a genitive complement (Cirtae). The verbal noun oppugnatio means ‘action
of besieging’, and not its accomplishment.
(14) Romae de proelio facto et oppugnatione
at Rome about battle-abl done-pst.ptc-abl and siege-abl
Cirtae audiebatur.
Cirta-gen heard-3.pass.impf
‘A report was spread at Rome of the battle which had been fought, and of
the siege of Cirta.’ (Sal. Jug. 22.1)
In sum, the participialization of the predicate – the dominant participle construc-

tion – does not seem to be a true equivalent of nominalization. Verbal nouns such
as oppugnatio ‘besieging’ do not seem to have a perfective meaning. This explains
the almost complete absence of the use of nouns such as deletio ‘destruction’ or
conditio ‘foundation’ in Classical Latin. They emerge in Late Latin.
2.2.3 Verbal nouns expressing actions and states

The category of bivalent verbal nouns is represented by nouns expressing actions or
states that are derived from verbs or associated with a state of affairs. Among Latin
derived nouns there are, for example: amor ‘love’ (derived from amo ‘to love’), fides
‘faith’ (fido ‘to trust’), timor ‘fear’ (timeo ‘to fear’), existimatio ‘opinion’ (existimo ‘to
think’), clamor ‘cry’ (clamo ‘to cry’). Nouns related to bivalent verbs are expanded
by arguments N1 and N2; nouns associated with monovalent verbs only encode N1.
However, arguments of verbal nouns can remain unexpressed, maybe more fre-
quently than in the case of arguments of verbs (see Panevová, this volume).
Consider desperatio ‘despair’ (15a), which is derived from the bivalent transi-
tive verb despero. From the semantic point of view, its complements represent the
agent (N1 omnium) and the patient (N2 salutis) of the action and correspond to the
clausal expression given in (15b) (cf. Pinkster 1990/1995: § 6.6). The agent is an
animate entity encoded in the so-called subjective genitive (omnium); the patient
refers to an inanimate entity and takes the form of an objective genitive (salutis).
The obligatory character of the complements expanding verbal nouns such as des-
peratio is proved by the fact that they cannot be used predicatively (Bolkestein
1989: 13) in the clausal expression given in (15c).
 Olga Spevak
(15) a. in desperatione omnium salutis

in despair-abl all-gen.pl safety-gen.sg
‘in the despair of safety experienced by everybody’ (Caes. Civ. 1.5.3)
b. Omnes salutem desperant.
all-nom.pl safety-acc despair-3.pl.prs
‘Everyone despairs of safety.’
c. ???
Desperatio omnium / nostra est.
despair-nom all-gen.pl our-nom.poss is-3.prs
‘Despair is of everyone/our.’
The valency frame of this verbal noun is:

desperatio + AGENT gen. N1; PATIENT gen. N2.
2.2.4 Nouns expressing the result of an action or a process

Among verbal nouns, it is necessary to distinguish nouns expressing the result of
an action or a process as a separate category. The peculiarity of these nouns con-
sists of the fact that they can express both an (ongoing) action and the result of an
action. Compare the following examples with oratio ‘speech’, which is related to oro
‘to speak (as orator)’:
(16) a. Sed nihil te interpellabo: continentem

but neg you-acc interrupt-1.fut continuous-acc
orationem audire malo.
speech-acc hear-inf prefer-1.prs
‘But I shall not interrupt you: I prefer to hear a continuous speech.’
(Cic. Tusc. 1.16)
b. oratio Ciceronis / eius
speech-nom Cicero-gen his-gen
‘Cicero’s/his oration’ (cf. Cic. Phil. 1.2)
(17) a. Orationes me duas postulas.

speeches-acc.pl me-acc two-acc ask-2.prs
‘You ask me for two speeches.’ (Cic. Att. 2.7.1)
b. orationes meae / Ciceronis
speeches-nom.pl my-nom.pl Cicero-gen
‘my/Cicero’s speeches’ (cf. Cic. Att. 2.1.3)
Oratio in (16a) has a temporal meaning and denotes ‘a speech’ as ‘an action of speak-
ing’ (second-order entity) but oratio in (17a) is the result of the action, a material-
ized speech, in our case, in writing. The semantic difference between these two uses
is that a genitive complement or a possessive pronoun oratio Ciceronis/oratio eius
Chapter 8. Noun valency in Latin 
‘Cicero’s speech/his speech’ – inferrable from the context and thus not expressed in
(16a) – mark the agent (N1) of the action. They correspond to Cicero orat ‘Cicero
speaks’ at the clause level and function as arguments. In the second instance, the
complement orationes Ciceronis/meae ‘Cicero’s/my speeches’ marks the author and
represents a satellite. Furthermore, oratio as a materialized, countable object joins
the first-order nouns such as liber ‘book’, and can take the plural number and a
prepositional phrase with de expressing content.
In Latin, this category is represented by such words as actio ‘action’ (ago ‘to
act’), especially a legal process and its materialization, ‘a plea’; consilium ‘delibera-
tion’ (consulo ‘to deliberate’) and its result, ‘a plan’; iudicium ‘legal proceedings’
(iudico ‘to judge’) and ‘a decision, a verdict’; commeatus ‘passage’ (commeo ‘to go
and come’) and ‘supplies’; rogatio ‘request’ (rogo ‘to ask’) and ‘a bill’.
Additionally, there are nouns stemming from verbs that have completely lost
any relationship with them and, consequently, any valency. These are, for example:
remedium ‘medicine’ (re, medior ‘to heal’); a signum ‘mark; statue’ is the result of
seco ‘to cut’; or exercitus ‘army’ is what has been trained (exerceo). They do not
have temporal meaning and belong to first-order entities.
2.2.5 Verbal nouns of movement

Verbal nouns derived from verbs of movement, for example profectio ‘departure’
(proficiscor ‘to set out’), reditus ‘return’ (redeo ‘to come back’), or aditus ‘access’
(adeo ‘to approach’), constitute a special category. Their second argument N2
expresses direction (argument), origin (argument with nouns that semantically
involve an idea of ‘return’), or path (satellite). It is never encoded in the adnominal
genitive: the second argument always retains the construction of the basis verb
and takes the form of a directional accusative or a prepositional phrase with in or
ad + accusative for expressions of direction, such as in Hispaniam in (18). Expres-
sions of the origin appear as prepositional phrases with a, ex + ablative, or as sim-
ple ablatives (19). The first argument N1 is encoded as a genitive (eius in 18) or a
possessive pronoun; it can also remain unexpressed when the agent of the action
is inferrable from the context, as is Antonius in (19).
(18) Quis de C. Cethego atque eius in Hispaniam
who about G. Cethegus-abl and his-gen into Spain-acc
profectione... cogitat?
expedition-abl thinks-3.prs
‘Who even thinks of Gaius Cethegus and his expedition into Spain...?’
(Cic. Sulla 70)
(19) Qui vero Narbone reditus!
what indeed Narbonne-abl return-nom
‘And what a return from Narbonne!’ (Cic. Phil. 2.76)
 Olga Spevak
Due to their specific encoding of the second argument N2, which reflects the
semantic function it fulfils, the construction of verbal nouns of movement receives
no further discussion here.
2.2.6 Abstract nouns expressing qualities

Second-order entities include various expressions of human qualities and feelings,
mostly derived from adjectives, such as pietas ‘dutiful conduct’ (derived from pius
‘dutiful’), dignitas ‘dignity’ (dignus ‘worthy’), ineptiae ‘absurdities’ (ineptus ‘silly’).
Since these expressions apply to human beings, their expansions, encoded as geni-
tives with a generic referent such as hominum in (20) or a specific referent, are
arguments. Such abstract nouns allow absolute use.
(20) praestans deorum natura hominum
exalted-adj gods-gen.pl nature-nom men-gen.pl
pietate coleretur
respect-abl worship-3.pass.sbjv.impf
‘the exalted nature of the gods would receive men’s pious worship’
(Cic. N. D. 1.45)
2.3 Third-order nouns

Third-order entities have the character of propositional content that can be true
or false, can be asserted or denied, remembered or forgotten (Lyons 1977: 442;
Rijkhoff 2002: 19). They are related to some mental or intellectual activity and
include verbal nouns, such as opinio ‘opinion’ (opinor ‘to think’), iudicium ‘judg-
ment’ (iudico ‘to judge’), nouns derived from adjectives: memoria ‘memory’
(memor ‘mindful’), and underived nouns, such as laus ‘praise’, spes ‘hope’.
Nouns belonging to third-order entities are for the most part bivalent and
they are distinguished by the richness of their valency frames (cf. Panevová 2002).
Apart from subjective and objective genitives, they admit complements such as
gerundives, gerunds, or indirect questions, because their semantic value implies
content. (21) illustrates complements found with the noun spes ‘hope’ ((21d) and
(21e) are support verb constructions).
(21) spes ‘hope’ + gen. N1 (Catilinae ‘Catilina’s’)/+ gen. N2 (pacis ‘of peace’)
Syntactic variants of N2:
a. Accusative + infinitive clause
magna in spe sum {mihi nihil
great-abl in hope-abl I am me-dat no
temporis prorogatum iri}
time-gen postponed-ptc.prf will be-fut.inf
‘I am very hopeful that there will be no postponement’ (Cic. Att. 6.2.6)
Chapter 8. Noun valency in Latin 
b. Gerundive:
spes {libertatis recuperandae}
hope-nom liberty-gen recovering-gen.gerdv
‘hope of recovering liberty’ (Cic. Agr. 1.17)
c. Prepositional phrase with de:
mea {de tua erga me benevolentia} spes
my-poss of your-abl towards me-acc goodwill-abl hope-nom
‘my hope of your kindly disposition toward me’ (Cic. Fam. 13.29.8)
d. Prepositional phrase with in:
in avaritia nobilitatis et pecunia sua
in avarice-abl nobility-gen and money-abl their-poss.abl
spem habere
hope-acc have-inf
‘having hopes in the avarice of the nobility and in his own wealth’
(Sal. Jug. 13.5)
e. Complement clause:
Quae te ratio in istam spem induxit {ut
what you-acc reason-nom in this hope-acc led-3.prf that
eos tibi fideles putares
these-acc.pl you-dat faithful-acc.pl consider-2.impf.sbjv
fore, quos pecunia corrupisses}?
be-fut.inf who-acc.pl money-abl corrupted-2.pqpf.sbjv
‘What reason led you to entertain the thought that men you had cor-
rupted with money would be faithful to you?’ (Cic. Off. 2.53)
It is worth mentioning the case of a primarily first-order noun, locus ‘place’, which
extends its meaning to ‘right, privilege’ and thus joins the third-order entities. As a
consequence of this, it can take a gerundive clause as complement (22).
(22) antiquiorem in senatu {sententiae
higher in order-acc in senate-abl opinion-gen
dicendae} locum
delivering-gen.gerdv right-acc
‘precedence in delivering my opinion in the senate’ (Cic. Ver. 5.36)
2.4 Nominal quantifiers and measure expressions (“containers”)

Nominal quantifiers and various measure expressions are monovalent. They apply
to various entities to indicate their small or large quantity: multitudo ‘quantity’,
magnus numerus ‘great amount’, grex ‘troop’, caterva ‘crowd’. The entity modified is
usually encoded in the genitive (23).
 Olga Spevak
(23) a. aquae sextarius

water-gen pint-nom
‘a pint of water’ (Cic. Off. 2.56)
b. magnus numerus militum
great-adj number-nom soldiers-gen.pl
‘great number of soldiers’
Nouns indicating a portion of an entity, such as pars ‘a part’ or dimidium ‘a half ’,

combine with a genitive or a partitive expression in the form of a de or ex prepo-
sitional phrase (‘from’) (24). However, these expressions are not synonymous
(Pinkster fc., chapter 11).
(24) ex illa pecunia magna pars

from that money-abl great-adj part-nom
‘a considerable part of that sum of money’ (Cic. Div. Caec. 57)
3. Argument marking at the noun phrase level
3.1 The adnominal genitive

In Latin, the genitive is the means par excellence for marking nominal complements.
Furthermore, this case is almost entirely restricted to them; verbs rarely combine
with the genitive. It serves to encode various arguments that at the sentence level
take different forms and stand in the nominative, accusative, dative, genitive, or
ablative (Pinkster 1990/1995: § 6.6). We have seen above the verbal noun desperatio
‘despair’ derived from the bivalent verb despero ‘to despair’, which takes the second
argument in the accusative, and the treatment of the agent and patient of the action
involved at the noun phrase level (15). Arguments in the dative used with verbs
such as fido ‘to trust’, invideo ‘to envy’, and servio ‘to serve’ are, at the noun phrase
level, also encoded in the genitive. This is the regular construction of such nouns
as fides ‘faith’, invidia ‘envy’, servitus ‘servitude’. The genitive is also found with
oblivio ‘oblivion’ (from obliviscor + genitive) and usus ‘use’ (from utor + ablative),
see examples in (25). Argument marking is thus simplified at the noun phrase level
with respect to the clause level, where distinctions such as agent/patient/recipient
are clearly distinguished through the use of different case endings.
(25) a. fides huius defensionis
faith-nom his-gen defense-gen
‘(my) good faith for his defense’ (Cic. Clu. 118)
huius defensioni fido
his-gen defense-dat trust-1.prs
‘I trust his defense.’
Chapter 8. Noun valency in Latin 
b. in oblivionem totius negotii

in oblivion-acc whole-gen affair-gen
‘into oblivion of the whole affair’ (Cic. Ver. 4.79)
totius negotii obliviscuntur
whole affair-gen forgot-3.pl.prs
‘They forgot the whole affair.’
c. usus navium
usage-nom ships-gen.pl
‘usage of ships’ (Caes. Gal. 3.14.7)
navibus utuntur
ships-dat.pl use-3.pl.prs
‘They use ships.’
Other means of encoding second arguments at the noun phrase level are infrequent
in Latin. The dative is sometimes found with verbal nouns such as o btemperatio
‘obedience’ (26), responsio ‘answer’, or plausus ‘applause’, corresponding to the
argument marking of their source verbs; it is rare with nouns other than verbal.10
(26) Iustitia est obtemperatio scriptis legibus
justice-nom is-3.prs obedience-nom written laws-dat.pl
institutisque
populorum.

customs-dat.pl=and people-gen.pl
‘Justice is conformity to written laws and national customs.’ (Cic. Leg. 1.42)
The accusative is not used for encoding patients in Classical Latin (cf. Section 2.2.1
above). Nor does the ablative serve to encode the semantic patient of nouns. There
are only a few attestations of optional complements in the ablative for expressing
time, manner, or instrument (27).
(27) An exercitus nostri interitus ferro
q army-gen our-gen destruction-nom sword-abl
fame frigore pestilentia?
famine-abl cold-abl pestilence-abl
‘Or the destruction of our army by sword, famine, cold, and pestilence?’
(Cic. Pis. 40)
Prepositional phrases, which unlike the genitive make the semantic relationship
between two entities explicit, can be used for argument marking in Latin only
10. See Kühner & Stegmann (1914: I.317) and Rosén (1981: 96–100) for Early Latin. There
are only a few instances of dative marking of the beneficiary, for example pabulum bubus
‘forage for cattle’ (Cato Agr. 27).
 Olga Spevak
with nouns belonging to very specific semantic fields. These are nouns implying
interactivity or sharing, for example bellum cum Iugurtha ‘the war with Jugurtha’
(Cic. Man. 60), expressions of content such as nuntius de ‘message concerning
something’ (28), and nouns expressing affections, which are dealt with in the next
Section.
(28) Nullus umquam de Sulla nuntius
no-nom ever about Sulla-abl message-nom to
ad me (pervenit).
me arrived-3.prf
‘No message about Sulla came ever to me.’ (Cic. Sul. 14)
3.2 Case marking vs. prepositional phrases

Genitive marking of nominal arguments of verbal nouns has as a consequence the
fact that the semantic function of the constituent involved remains unspecified;
that is, a genitive argument allows the interpretation as a subjective as well as an
objective genitive (cf. metus hostium ‘fear of the enemies’, quoted and discussed
below (41)). Such an ambiguity arises especially when the noun in the genitive
refers to an animate, human entity; an inanimate entity in the genitive is almost
always interpreted as patient. Furthermore, both genitives may co-occur in one
noun phrase, as in (15a) quoted above.11
Ambiguity due to effacement of the agent/patient distinction at the noun
phrase level can be avoided by the formal differentiation of one of the comple-
ments, especially the objective genitive. Such a substitution is frequent in Czech.
As Kolářová (this volume, Section 4.1) shows, the patient takes either a case
other than the genitive or a prepositional phrase. The genitive marking (*herců)
of the patient would be ungrammatical in a noun phrase with both participants
expressed (29).
(29) obdiv diváků *herců / k hercům
admiration-nom audience-gen.pl actors-gen.pl to actors-dat.pl
‘admiration of the audience toward the actors’
Something similar happens in Latin but only in the case of one specific semantic
group of nouns: those expressing emotions, such as love, hatred, fear, flattery, anger,
praise (Torrego 1991). The patient, which has an animate, human referent – or a
11. Noun phrases with one subjective and one objective genitive are listed in Latin gram-
mars, see Kühner & Stegmann (1914 I: 416); cf. also Devine & Stephens (2006: 316). According
to Rosén (1981: 78), there are no sure instances in Early Latin.
Chapter 8. Noun valency in Latin 
referent associated with human beings like patria ‘fatherland’ – can take a prep-
ositional phrase instead of the objective genitive, as in (30a). The corresponding
expression at the clause level is indicated in (30b). The use of the prepositional
phrase in patriam makes explicit whom the action is oriented toward and focuses
on the person. The valency frame of amor can be established as follows:12
amor + AGENT gen. N1; PATIENT gen./in ‘for’/inter ‘among’/erga ‘toward’ N2.
(30) a. posteaquam L. Flacci amor
after L. Flaccus-gen love-nom
in patriam perspectus esset
for fatherland-acc perceive-3.pass.pqpf.sbjv
‘after Lucius Flaccus’ love for our country has been clearly seen’
(Cic. Flac. 2)
b. L. Flaccus patriam amat.
L. Flaccus-nom fatherland-acc loves-3.prs
‘Lucius Flaccus loves his fatherland.’
Let us now look at the distribution of the complements used with the nouns amor
‘love’ and odium ‘hatred’ in Cicero’s speeches (Table 2) to see the extent of such
substitution in Latin. Data were collected with the help of Merguet (1877–1884).
In the majority of cases, the nouns under examination have only one argument;
noun phrases with two explicit arguments are in the minority (row 5: subjective
genitive + prepositional phrase).
Table 2. Arguments used with amor and odium
in Cicero’s speeches
Syntactic form of the complement amor odium
objective genitive animate 3 9

objective genitive inanimate 7 11
subjective genitive 2 16
in-prepositional phrase 11 8
subj. gen. + prep. phrase 2 7
other prepositional phrase 2 1
possessive pronoun (subject) 3 5
Total 30 57
12. Prepositions do not seem to be used in Early Latin (cf. ThLL, s.v. amor 1969.66). For
interchange of genitives with prepositional phrases, see Torrego (1989) and Nutting (1932,
esp. 268–279).
 Olga Spevak
These data show, first, a competition between objective and subjective geni-
tives (20 vs. 16) in the case of odium; the other participant is understood from
the context. Although inanimate referents are likely to be interpreted as objec-
tive genitives, there are still 9 animate referents functioning as patients. Second,
prepositional phrases do not really resolve the problem of ambiguity in the case of
odium (9 instances of objective genitives competing with 8 prepositional phrases).
In the case of amor, prepositional phrases are used more frequently than objective
genitives with animate referents (11 vs. 3). However, there does not seem to be a
tendency to replace the objective genitive with a prepositional phrase. The prepo-
sitional phrase is likely a more expressive means for encoding patients than the
objective genitive. On the other hand, there is no instance of the combination of a
subjective and an objective genitive in this sample; in the case of co-occurrence of
both expressions, the agent is encoded as a genitive, the patient as a prepositional
phrase.
Additionally, it is worth pointing out that the encoding of patients as prepo-
sitional phrases is peculiar to noun phrases in Latin. Expansions with in ‘for’ of
amor (30a) function as arguments at the noun phrase level and, furthermore, they
do not stem from the valency frame of the verb amo (*amo in *‘to love for’). When
such a prepositional complement is found with other verbs, especially verbs of
saying, it functions as a satellite; for example, carmen in eum scribo ‘to write a
poem in someone’s honor’ (cf. Cic. de Orat. 2.352). This point must be stressed.
The preposition erga ‘toward’ is typical of noun phrases (31); it is not used with
verbs at all.
(31) quae voluntas erga Caesarem totius provinciae

what feeling-nom toward Caesar-acc whole-gen province-gen
‘what was the feeling of the whole province toward Caesar’
(Caes. Civ. 2.17.1)
3.3 The syntactic form of noun arguments

We have seen that some nouns can develop a construction independent from the
construction of the source verbs. This point invites a closer look at the other nouns
denoting actions or states.
Among these, there are valency nouns that do not have a verbal origin. Words
such as the following belong to this category: spes ‘hope’ (→ spero ‘to hope’),13 cura
13. Arrows indicate the direction of the derivation: x → y means y is derived from x.
Chapter 8. Noun valency in Latin 
‘care’ (→ curo ‘to care’), iniuria ‘injustice’ (←negative in + ius ‘justice’, ‘law’; no
cognate verb), gratia ‘favor’ (← gratus ‘grateful’; no cognate verb), or ius iurandum
‘oath’ (a binding formula to be sworn; ius ‘justice, law’ → iuro ‘to swear’). Their
valency frames are various and sometimes very rich. It is tempting, obviously, to
interpret the genitive used with ius iurandum + gen. N1 in (32a) as a subjective
genitive. At the clause level, (32b) can be envisaged.
(32) a. [cum habeas]

integerrimi municipi ius iurandum
most respectable-gen town-gen oath-acc
‘[when you have] the oath of a town of the highest integrity’(Cic. Arch. 8)
b. integerrimum municipium iuravit /
most respectable-nom town-nom swore-3.prf
ius iurandum dedit
oath-acc gave-3.prf
‘a town of the highest integrity swore an oath/took an oath’
Such an analysis seems fully justified: we can assume that these nouns, without
having a verbal origin, have joined the category of verbal nouns due to analogy:
spes enters the group of nouns and verbs of thinking and willing; ius iurandum,
that of nouns and verbs of speaking. If we admit that semantic analogies, which
link underived nouns with semantically cognate groups – cf. Pinkster (fc. chap. 12)
on transcategorial parallelism –, are at work, we can more easily understand the
richness of valency frames of such nouns, for example of spes, illustrated above in
(21), as well as the fact that a word like ius iurandum ‘oath’ can take a complement
clause with ut + subjunctive as expansion (33) – as does the verb dico ‘to say’ for
expressing volitive content.
(33) ius iurandum poscere, {ut}, quod esse ex usu

oath-acc ask-inf that what be-inf to advantage-abl
Galliae intellexissent, {communi
Gaul-gen judge-3.pl.pqpf.sbjv common-abl
consilio administrarent}
consent-abl execute-3.pl.impf.sbjv
‘to ask for their oath that they would by common consent execute whatever
they judge to be for the advantage of Gaul’ (Caes. Gal. 5.6.6)
The lack of a cognate verb is sometimes compensated for by “support verb” con-
structions (Ger. Funktionsverbgefüge) involving a semantically weak verb such as
facio ‘to make’, habeo ‘to have’, ago ‘to do’, gero ‘to bear’, or adfero ‘to cause’ and a
 Olga Spevak
verbal noun or a noun related to a verb.14 The noun is the bearer of the mean-
ing, and furthermore it often imposes the syntactic form of the expansion – for
example, the prepositional phrase with cum (34); there is no *gero cum ‘to bear
with’. The contribution of the verb is to actualize the process (gesserint). A counter-
example is gratias ago + dat. ‘to thank sb’ where the dative can be reduced neither
to the noun gratias nor to the verb ago (Happ 1976: 454).
(34) Rhodii qui... bellum illud superius
Rhodians-nom who-nom war-acc that previous-acc
cum Mithridate rege gesserint
with Mithridates-abl king-abl waged-3.pl.prf.sbjv
‘The Rhodians who... carried on the first war against Mithridates’
(Cic. Ver. 2.159)
The formation of such periphrastic constructions results, on the one hand, from
the fact that combination of a verbal noun and a verb with a weak semantic value
makes it possible to explicitly express aspectual nuances (Flobert 1996) – impetum
facio ‘to make an attack’ (Cic. Mil. 29) marks execution of impetus ‘attack’ – and
to express causativity: spem adfero ‘to arouse hope’ (Cic. Amic. 68) (Hoffmann
1996: 204). On the other hand, support verb constructions sometimes compensate
for the absence of a verb as such; for example, cognationem habeo cum ‘to have
affinity with’ or auctoritatem habeo apud ‘to have influence with’ have no matching
semantic expressions in the verbal domain.
Several nouns involved in support verb constructions have the same form of
expansion as their cognate verbs, such as pactio cum ‘a compact with’ – paciscor
cum ‘to arrange an agreement with’; coniectura de ‘a conjecture about’ – conicio de
‘to conjecture about’; verbal nouns of movement retain the construction of their
cognate verbs as well. However, there are valency nouns that do not have morpho-
logically cognate verbs and thus cannot “copy” a verbal construction. The ques-
tion to be asked is where the syntactic form of their complements comes from.
For example, verbum ‘word, discourse’ in the construction verba facio cum (lit. ‘to
make words with’) ‘to talk with’ (Roesch 2001) seems to be explained as analogous
to loquor cum ‘to talk with’; bellum ‘war’ in bellum gero cum ‘to wage war with’
models pugno cum ‘to fight with’. The construction of auctoritas ‘authority, influ-
ence’ (derived from the agent noun auctor ‘who authorizes’) in auctoritatem habeo
apud ‘to have influence with’ could be explained as analogous to valeo apud ‘to
have influence on somebody’.
14. Support verbs are used in all periods of Latin; see Rosén (1981: 139 and 190), Flobert
(1996) and Hoffmann (1996).
Chapter 8. Noun valency in Latin 
Here again, analogy with expressions belonging to the same semantic field is
obvious. Several examples are presented in Table 3.
Table 3. Analogous constructions

Semantic field verbs nouns analogical construction
verbs of speaking dico ‘to say’ ius iurandum ‘oath’ ius iurandum +
expressions of scribo ‘to write’ spes ‘hope’ AcI/+ ut-clause ‘oath that’
– factual content: respondeo ‘to answer’ spes + AcI /+ ut-clause
accusative + infinitive ‘hope that’
– volitive content:
ut-clause + subj.
concern scribo in ‘to write amor ‘love’ amor in/erga ‘love for’
“A is oriented toward B” in favor of odium ‘hatred’ odium in/erga ‘hatred for’
in ‘for’, erga ‘toward’ sb./against sb’
interactivity paciscor cum + abl. verbum ‘word’ bellum cum ‘war with’
“A has to do with B” ‘to arrange an bellum ‘war’ verba cum ‘words with’
cum ‘with’ agreement with’
loquor cum ‘to
talk with’
3.4 Verbal nouns derived from trivalent verbs

Languages seem to behave in different ways as to the number of complements
expressed and their syntactic form. Whereas Latin is rather reluctant to express
both the patient and the recipient/addressee with nouns derived from trivalent
verbs, such expressions are common in Czech (35a) and German (Happ 1976: 177).
The patient appears in the genitive, but the recipient (kamarádovi) retains the
dative, which is in the valency frame of its cognate verb, vrátit ‘to return’, in (35b)
(see Kolářová 2006 and this volume).
(35) a. vrácení knihy kamarádovi
returning-nom book-gen friend-dat
‘returning of the book to his friend’
b. vrátil knihu kamarádovi
returned-3.prf book-acc friend-dat
‘he returned the book to his friend’
In Latin, trivalent verbs such as dono aliquid alicui ‘to give something to somebody’,
which usually present a great formal variety of arguments ( Pinkster 1985: 170‒2),
constitute a category that is less easily subject to nominalization. Subjective (36a)
and objective (36b) genitives accompanying nouns such as donatio ‘donation’,
responsum ‘answer’, absolutio ‘acquittal’, privatio ‘privation’ or rogatio ‘request, bill’
are found.
 Olga Spevak
(36) a. responsum haruspicum

answer-nom soothsayers-gen.pl
‘the answer of the soothsayers’ (Cic. Har. 9)
b. de donatione regnorum
on donation-abl kingdoms-gen.pl
‘on the grants of foreign kingdoms’ (Cic. Fam. 1.9.7)
There are also several examples of verbal nouns with the third argument alone,
such as liberatio culpae ‘a release from all guilt’ (Cic. Lig. 1). However, it is excep-
tional to find instances of verbal nouns derived from a trivalent verb with more
than one complement expressed together in one noun phrase as in (37) with roga-
tiones ‘bills’.
(37) ut etiam Catonis rogationibus de Milone et
that also Cato-gen bills-dat.pl about Milo and
Lentulo resistamus
Lentulus-abl stand against-1.pl.prs.sbjv
‘to make a stand against Cato’s bills concerning Milo and Lentulus’
(Cic. Q. fr. 2.3.4)
It is not usual to encode the recipient/addressee of an action in the genitive at the

noun phrase level (Pinkster 1990/1995: § 6.6). Additionally, it rarely appears in the
dative or in a prepositional phrase, such as the ad-phrase with translatio in (38a),
which is derived from transfero ‘to transfer’. The expression of origin (a-phrase)
functions as a satellite. This noun phrase corresponds to the clausal expression
given in (38b).
(38) a. Quare L. Sullae C. Caesaris pecuniarum
therefore L. Sulla-gen G. Caesar-gen money-gen.pl
translatio a iustis dominis ad alienos
transference-nom from lawful-abl owners-abl to others-acc
non debet liberalis videri.
neg ought-3.prs liberal-nom seen-pass.inf
‘Consequently, the transference of money by Lucius Sulla and Gaius
Caesar from its lawful owners to others ought not to be seen as liberal.’
(Cic. Off. 1.43)
b. L. Sulla C. Caesar pecunias a
L. Sulla-nom G. Caesar-nom money-acc.pl from
iustis dominis ad alienos transtulerunt.
lawful owners-abl to others-acc.pl transfer-3.pl.pf
‘Lucius Sulla and Gaius Caesar transferred money from its lawful
owners to others.’
Chapter 8. Noun valency in Latin 
The number and syntactic form of allowed complements can be due to the degree
of grammaticalization of verbal nouns and/or to the phenomenon of reduced
valency frame that concerns especially verbal nouns derived from trivalent nouns.
Latin, at least, avoids the expression of both a second and a third argument together
in one noun phrase.
4. Semantic shifts between verbs and nouns
The valency reduction that accompanies nominalization often goes together with
changes in the meaning of the noun, the verb, or both. There are pairs that more
or less maintain their semantic properties as well as their construction in paral-
lel, such as paciscor ‘to arrange an agreement’ and pactio ‘agreement’; but other
expressions underwent change. First, there is a tendency to restrict the meaning
of the noun: the meaning of the verbal noun coniectura ‘conjecture’ is only one
among multiple meanings of conicio ‘to throw together, to put, to dispatch’. Factio
‘manner of doing’ and facio ‘to do’ became completely separated: factio denotes
a social group, especially a ‘faction’. Second, the couple oratio – oro illustrates a
change in meaning on the part of the verb. Oratio ‘action of speaking, speech’ is
derived from oro in the meaning ‘to speak (as orator)’; however, the verb special-
ized its meaning into ‘to pray’, a semantic feature that does not affect the verbal
noun. Although detailed research in this domain is necessary to determine the
extent to which such semantic shifts took place, it does not seem likely that a Latin
verb always has available a matching verbal noun such as ‘to do’/‘doing’, that is, a
productive derivation of nouns denoting states of affairs.
5. Conclusions
Complements required by the valency of a noun are more closely related to their
head noun than optional complements are. Valency nouns whose the semantic
value requires complementation fall into several categories: relational nouns,
verbal nouns or nouns associated with states of affairs, and quantifying and clas-
sifying expressions (“containers”). These categories are distinguished by specific
semantic properties as well as by the way they encode noun complements. The
concept of three orders of entities, although further investigation is necessary to
discuss in detail sub-categories of the second and third orders, makes it easier to
predict the complementation of a valency noun.
In Latin, a language with case marking, expressions of alienable and inalien-
able possession are not formally distinguished: both are encoded as genitives.
 Olga Spevak
“Containers” also require a genitive complement. The category of verbal nouns

and nouns related to states of affairs is the most complex one. A further subdivi-
sion is necessary to describe the behavior of (i) nouns related to verbs of move-
ment; (ii) nouns expressing actions and states; (iii) nouns expressing results of an
action or a process; and (iv) nouns expressing propositional content.
The question of noun valency is closely related to argument marking at the
noun phrase level. In Latin, arguments of nouns (agent and patient) are typically
encoded as genitives; prepositional phrases are confined to specific semantic cat-
egories (expressions of interactivity, content, and affection). Latin is reluctant to
express a recipient/addressee at the noun phrase level; this can be due to a weak
degree of grammaticalization or to the productivity of verbal nouns in Latin. A
point that requires more detailed investigation is the extension of valency frames
resulting from analogy with semantically related words and, conversely, reduction
of valency frames in the case of nouns expressing the result of an action. Addition-
ally, support verb constructions as well as periphrases with the verb ‘to be’ call for
special attention, because such constructions often compensate for the lack of a
verb and their complementation is imposed by the nominal component.
Epilogue: “Noun valency” in ancient grammatical theory
When dealing with valency, it is worth adding several points that were consid-
ered in ancient Greek and Roman grammatical theory, which was based on the
concept of parts of speech. First, the ancients envisaged “completeness” of utter-
ances. How much and in what detail this concept was used is difficult to evalu-
ate, especially because we have only fragmentary knowledge of Stoic doctrine and
because of the loss of the third part of Varro’s treatise On the Latin language.15
Nevertheless, the Greek grammarian Apollonius Dyscolus (2nd century AD), fol-
lowed by the grammarian Priscian (6th century AD), worked with the concept
of “complete” and “incomplete” utterances, which are now described in terms of
valency, omissibility, and obligatory and optional complements.16 Second, ancient
15. Cf. a fragment from book 14 quoted by Gel. 16.8.

16. See, for example, Priscian’s (GL Keil III 116.9) demonstration of omissibility of the com-
plements in the sentence: idem homo lapsus heu hodie concidit ‘unfortunately, the same man
having fallen down has died today’. In this sentence, which contains a representative of each
part of speech (in the sense of ancient theory), the only non-omissible constituents are homo
concidit ‘(the) man has died’: Si tollas nomen aut verbum, imperfecta fit oratio; sin autem cetera
subtrahas omnia, non necesse est orationem deficere. ‘If you omit the noun or the verb, the sen-
tence will be incomplete; if you omit the rest, the sentence will not necessarily lack anything.’
Chapter 8. Noun valency in Latin 
grammarians separated several categories of nouns (see Luhtala 2005: 75), such as

relational nouns (ad aliquid), which cover the construction pater meus/Ciceronis
‘my/Cicero’s father’ (39), and verbal nouns (verbalia), which may take inflected
complements in the genitive (40). 18
(39) ‘There are nouns used with respect to something (ad aliquid). They cannot
be understood without relation with another noun, for example, pater
‘father’. We cannot say pater unless he has a child.’
(Serv. GL Keil IV 430.19)17
(40) ‘Transitive verbal nouns in -or and -rix, which are formed from verbs
denoting an action, combine with the genitive, for example, amo illum
‘I love him’, amator ‘lover’ and amatrix illius ‘his (female) lover’. They are
derived from verbs, for example, doctor ‘teacher’, lector ‘reader’.’
(Prisc. gramm. GL Keil III 215.23)18
Third, an interesting testimony is that of the so-called objective and subjective

genitives, reported by Aulus Gellius (2nd century AD) in a Section devoted to
words that are used with two opposite meanings, both active and passive. He dis-
cusses in detail the verbal noun metus ‘fear’ (41a).
(41) a. ‘Metus also and iniuria, and some other words of that kind, may be
used in this double sense; for metus hostium ‘fear of the enemies’ is
a correct expression both when the enemies fear and when they are
feared (cum timent hostes et cum timentur). Thus Sallust in the first
book of his History speaks of ‘the fear of Pompey’ (metus Pompei), not
implying that Pompey was afraid, which is the more common meaning,
but that he was feared.’ (Gel. 9.12.13‒15)
b. metus hostium
fear-nom enemies-gen.pl
‘fear of the enemies’
Priscian’s argument is based on Apollonius Dyscolus (Synt. 1.14). The concept of “complete
utterance” (λεκτὸν αὐτοτελές [lekton autoteles], oratio perfecta) and “incomplete utterance”
(λεκτὸν ἐλλιπές [lekton ellipes], oratio imperfecta)” goes back to Stoic doctrine (see Ildefonse
1997: 146).
17. Sunt quae ad aliquid dicuntur, quae penitus non possunt sine alterius coniunctione intellegi,
ut pater: non enim patrem possumus dicere, nisi filium habeat.
. Verbalia quoque in -or desinentia vel -rix transitiva, quae a verbis actum aliquem signifi-
cantibus fiunt, genetivo iunguntur, ut amo illum, amator et amatrix illius. Sunt facta de verbo, ut
doctor, lector.
 Olga Spevak
hostes timent
enemies-nom.pl fear-3.pl.prs
‘the enemies fear’
hostes timentur
enemies-nom.pl fear-3pl.prs.pass
‘the enemies are feared’
Gellius establishes a parallel between the noun phrase metus hostium ‘fear of
the enemy’ and the clauses hostes timent ‘the enemies fear’ and hostes timentur
‘the enemies are feared’, which is its passive counterpart (41b). From the syntac-
tic point of view, the genitive hostium can correspond to the subject (agent) or
the object (patient) of the related verb, timeo ‘to fear’. In modern literature from
Benveniste ([1962] 1966: 146‒7) onward (cf. also Kuryłowicz ([1949] 1960: 145),
noun phrases involving a verbal noun and a genitive complement are regarded as
“transpositions” of direct objects and subjects, e.g. tolerantia frigoris ‘endurance of
cold’ going back to tolerare frigus ‘to endure cold’, and adventus consulis ‘arrival of
(the) consul’ to consul advenit ‘(the) consul arrives’. Benveniste labels such geni-
tives “genitives of transposition” (génitifs de transposition), but they are commonly
called objective and subjective genitives.19 They are arguments of the noun as are
direct objects and subjects.
References
Baldi, Philip & Nuti, Andrea. 2010. Possession. In New Perspectives on Historical Latin Syntax,
Vol. 3: Constituent Syntax: Quantification, Numerals, Possession, Anaphora, Philip Baldi &
Pierluigi Cuzzolin (eds), 239–387. Berlin: De Gruyter Mouton.
Benveniste, Émile. 1966[1962]. Pour l’analyse des fonctions casuelles: Le génitif latin. In Prob-
lèmes de linguistique générale, Vol. 1, 140–148. Paris: Gallimard.
Bolkestein, A. Machtelt. 1981. Factitivity as a condition for an optional expression rule in Latin:
The ‘ab urbe condita’ construction and its underlying representation. In Predication and
Expression in Functional Grammar, Machtelt Bolkestein, Caspar de Groot & J. Lachlan
MacKenzie (eds), 206–233. New York NY: Academic Press.
Bolkestein, A. Machtelt. 1989. Parameters in the expression of embedded predications in
Latin. In Subordination and Other Topics in Latin. Proceedings of the Third colloquium on
Latin Linguistics (Bologna, 1–5 April 1985) [Studies in Language Companion Series 17],
Gualtiero Calboli (ed.), 3–35. Amsterdam: John Benjamins.
19. For the term objective/subjective genitive, see Rosén (1978). Laurentius Valla was the
first to establish the relationship between the genitive and the subject or object of a corre-
sponding clause; it is termed genitivus obiecti/subiecti by Vossius. Genitivus obiectivus/subiec-
tivus is current in grammars in 19th century.
Chapter 8. Noun valency in Latin 
Chappell, Hilary & McGregor, William. 1996. Prolegomena to a theory of inalienability. In The
Grammar of Inalienability: A Typological Perspective on Body Part Terms and the Part-whole
Relation, Hilary Chappell & William McGregor (eds), 3–30. Berlin: Mouton de Gruyter.
DOI: 10.1515/9783110822137.3
Corbett, Greville G. 1995. Slavonic’s closest approach to Suffixaufnahme: The possessive adjective.
In Double Case: Agreement by Suffixaufnahme, Frans Plank (ed.), 265–282. Oxford: OUP.
Devine, Andrew M. & Stephens, Laurence D. 2006. Latin Word Order. Structured Meaning and
Information. Oxford: OUP. DOI: 10.1093/acprof:oso/9780195181685.001.0001
Dressler, Wolfgang. 1970. Comment décrire la syntaxe des cas en latin? Revue de Philologie 44:
25–36.
Flobert, Pierre. 1996. Les verbes supports en latin. In Acten des VIII. internationalen Kollo-
quiums zur lateinischen Linguistik, Alfred Bammesberger & Friedrich Heberlein (eds),
193–199. Heidelberg: Carl Winter.
Fugier, Huguette. 1983. Le syntagme nominal en latin classique. In Aufstieg und Niedergang der
römischen Welt 2, 29, 1, Wolfgang Haase (ed.), 212‒269. Berlin: Walter de Gruyter.
GL: Keil, Henricus. 1855–1880. Grammatici Latini. Leipzig: Teubner.
Happ, Heinz. 1976. Grundfragen einer Dependenz-Grammatik des Lateinischen. Göttingen:
Vandenhoeck & Ruprecht.
Heick, Otto William. 1936. The Ab Urbe condita Construction in Latin. Lincoln NB: Lincoln
University Press.
Heine, Bernd. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge:
CUP. DOI: 10.1017/CBO9780511581908
Hoffmann, Roland. 1996. Funktionsverbgefüge im Lateinischen. In Acten des VIII. internation-
alen Kolloquiums zur lateinischen Linguistik, Alfred Bammesberger & Friedrich Heberlein
(eds), 200–212. Heidelberg: Carl Winter.
Ildefonse, Frédérique. 1997. La naissance de la grammaire dans l’Antiquité grecque. Paris: Vrin.
Kolářová, Veronika. 2006. Valency of deverbal nouns in Czech. The Prague Bulletin of Math-
ematical Linguistics 86: 5–19.
Kühner, Raphael & Stegmann, Carl. 1914. Ausführliche Grammatik der lateinischen Sprache 2,
Satzlehre, 2 Vols. Hannover: Hahn.
Kuryłowicz, Jerzy. 1949[1960]. Le problème du classement des cas. In Esquisses linguistiques,
131–150. Wrocław: Akademia Nauk.
Lehmann, Christian. 1985. On grammatical relationality. Folia linguistica 19: 67–109. DOI:
10.1515/flin.1985.19.1-2.67
Lehmann, Christian. 2005. Sur l’évolution du pronom possessif. In Latin et langues romanes.
Études de linguistique offertes à József Herman à l’occasion de son 80e anniversaire, Sandor
Kiss, Luca Mondin & Giampaolo Salvi (eds), 37–46. Tubingen: Niemeyer. DOI: 10.1515
/9783110944532.37
Longrée, Dominique. 1995. Du fonctionnement syntaxique de la construction ab urbe condita
chez Tacite. In De usu. Études de syntaxe latine offertes en hommage à Marius Lavency,
Dominique Longrée (ed.), 175–188. Louvain-la-Neuve: Peeters.
LLT: Library of Latin Texts, Series A. 〈www.brepolis.net〉.
Luhtala, Anneli. 2005. Grammar and Philosophy in Late Antiquity: A Study of Priscian’s Sources
[Studies in the History of the Language Sciences 107]. Amsterdam: John Benjamins. DOI:
10.1075/sihols.107
Lyons, John. 1977. Semantics, 2 Vols. Cambridge: CUP.
 Olga Spevak
Merguet, Hugo. 1877–1884. Lexikon zu den Reden des Cicero, 4 Vols. Jena: Fischer.
Nutting, Herbert C. 1932. On the adnominal genitive in Latin. University of California Publica-
tions in Classical Philology 10(10): 245–308.
Panagl, Oswald. 2006. Zur verbalen Konstruktion deverbativer Nomina. In Word Classes and
Related Topics in Ancient Greek, Emilio Crespo, Jesús de la Villa & Antonio R. Revuelta
(ed.), 91–104, Louvain-la-Neuve: Peeters.
Panevová, Jarmila. 2002. K valenci substantiv s ohledem na jejich derivaci (On noun valency
with respect to their derivation). Zbornik Matice Srpske za Slavistiku 61: 29–36.
Pinkster, Harm. 1985. Latin cases and valency grammar. Some problems. In Syntaxe et latin.
Actes du IIe Congrès international de linguistique latine (Aix-en-Provence, 28–31 mars
1983), Christian Touratier (ed.), 163‒189. Aix-en-Provence: Presses Universitaires.
Pinkster, Harm. 1990/1995. Latin Syntax and Semantics. London: Routledge. Revised and
enhanced edition Sintaxis y semántica del latín. Madrid: Ed. Clásicas, 〈http://perseus.uchi-
cago.edu/cgi-bin/philologic/ navigate.pl?NewPerseusMonographs.19〉 (September 2012).
Pinkster, Harm. Forthcoming. The Oxford Latin Syntax.
Porzig, Walter. 1942. Die Namen für Satzinhalte im Griechischen und im Indogermanischen.
Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft 10. Berlin: Walter
de Gruyter. DOI: 10.1515/9783111679358
Rijkhoff, Jan. 2002. The Noun Phrase. Oxford: OUP. DOI: 10.1093/acprof:oso/9780198237822
.001.0001
Roesch, Sophie. 2001. Les emplois de verbum et sermo dans les expressions à verbe support
verba facere, verba habere et sermonem habere. In De lingua Latina novae quaestiones. Actes
du Xe Colloque international de linguistique latine, Paris-Sèvres, 19–23 avril 1999, Claude
Moussy (ed.), 859–874. Louvain: Peeters
Rosén, Hannah. 1978. The emergence of a syntactic notion: Genitive of the object and geni-
tive of the subject in continental Renaissance grammars of Latin. Folia linguistica 12(3–4):
267–283.
Rosén, Hannah. 1981. Studies in the Syntax of the Verbal Noun in Early Latin. Munich: Fink.
Rosén, Hannah. 1983. The mechanisms of Latin nominalization and conceptualization in a his-
torical view. In Aufstieg und Niedergang der römischen Welt 2, 29, 1, Wolfgang Haase (ed.),
179–211. Berlin: Walter de Gruyter.
Seiler, Hansjakob. 1983. Possession as an Operational Dimension of Language. Tübingen: Gunter
Narr.
Seitz, Johannes. 1938. Über die Verwendung der Abstrakta in den Dialogen Gregors des Grossen.
Borna: R. Noske.
Spevak, Olga. 2010. Constituent Order in Classical Latin Prose [Studies in Language Companion
Series 117]. Amsterdam: John Benjamins. DOI: 10.1075/slcs.117
Spevak, Olga. 2014. The Noun Phrase in Classical Latin Prose [Amsterdam Studies in Classical
Philology 21]. Amsterdam: Brill. DOI: 10.1163/9789004265684
ThLL: Thesaurus Linguae Latinae. 1900–. Leipzig: Teubner
Torrego, M. Esperanza. 1989. Caracterización funcional de los sintagmas preposicionales en
latín: pro + abl., contra, adversus, in + ac. In Actas del VII Congreso español de estudios clási-
cos, 609–616. Madrid: Universidad Complutense.
Torrego, M. Esperanza. 1991. The genitive with verbal nouns in Latin: A functional analysis.
In New Studies in Latin Linguistics. Selected Papers from the 4th International Colloquium
on Latin Linguistics, Cambridge, April 1987 [Studies in Language Companion Series 21],
Robert Coleman (ed.), 281–294. Amsterdam: John Benjamins.
Index
A external/internal argument marking the addressee/

ablative 184, 189, 193, 196–197 79, 89–90, 94, 106, 108, 161 recipient 56, 92–93, 183,
absorption (of an argument) first argument 74, 78, 84, 196, 202–204
11–12, 35, 79 see also 187, 193 compound adjective 127
incorporation of a verb 1, 6, 10, 21, 92, compound noun 115–118, 120,
accusative 94, 109, 124, 183, 197 122–123, 127, 161, 170 127–128
as argument of a verb 10, of a noun 61, 63, 75n21, container 13, 28n7, 36,
27, 92, 106, 193, 196 78, 83–84, 90–93, 102, 144–146, 148–149, 151,
with verbal nouns 189, 190 104, 115, 120, 161, 179, 153–154, 157–158, 185, 195,
actor 3–4, 6, 14–15, 20–21, 23, 191, 193–194, 199, 200, 205–206
25, 28, 45, 49, 56 see also 204–206, 208 content clause see complement
agent second argument 75, 78, 84, clause
addressee 3–4, 6, 21, 25, 49, 52, 155, 187, 193–194, 196–197
56, 183, 203–204, 206 third argument 204–205 D
adjunct 21, 93–94, 115, 120, 123, with a classifier 143, 145, 147, determiner 62, 94, 106–108,
125, 128–129, 131, 133–134, 151–155, 158, 127–128, 143, 145, 148–150,
137, 144, 157–158, 163, aspect 40, 43, 56, 73, 96, 102, 164–171, 178–179
169–171, 178–179 105, 107, 109, 202 see also derivation
adjunction see adjunct telic, atelic, perfective, lexical 7, 11, 20–22, 24, 29,
agent 20, 35, 40–41, 62, 65, imperfective 32, 38, 46, 55–57
80, 89–91, 93–94, 108, aspectual adverbials 96 syntactic 7, 11, 20–22,
185, 187–188, 191, 193, 196, aspectual affixes 68, 73, 95–96, 24–25, 29, 38, 41,
198, 200, 206, 208 see also 98 46, 55–57
actor aspectual modifiers 62, 98, 103 dialogue test 5, 12–13, 15
agreement 103, 142–143, 145, aspectual properties 90, 99
148–149, 151, 158 atelic 97 E
ambiguity effect 3, 6, 13, 21, 23
semantic 13, 15, 24, 38, 69, B
97n2, 176, 287 beneficiary 92, 183, 197n10 F
syntactic 8, 28, 49, 52, 56, 91, free modification 1, 3, 6, 13,
198, 200 C 21, 36
antecedent 162, 164–167, central/peripheral clause functor 3n4, 21n2, 31n9,
170–171, 179 171–179 36, 48
apposition 114, 141–143, 147, 158 classifier 101, 143–145, 185
appositive (expression, classifying apposition 131, 137 G
status) 128–129, 146 complement 77–79, 114–115, genitive
argument see also complement 120–123, 125, 129, 131–132, adnominal 25, 27–29, 30–32,
argument structure 37, 134, 136–137, 144–146, 148, 114, 183, 189–190, 192–194,
63, 65, 74, 76, 79, 89, 125, 153–158, 162–179 196, 207
152–153 complement clause 25–26, 29, as verb argument 9, 27, 196
arguments of relational 30, 35, 40, 172, 174, 179, 201 in other expressions 13, 186
nouns 115, 120–121, 132 adnominal 29, 95, 197 marking possession 13, 80,
argument realization 78, marking possession 186 185–187, 205
83–84, 120, 123–126, 128, marking the patient 42, 45, marking the addressee/
134 50, 52, 56 recipient 49, 52, 56, 204
 Index
marking the agent 8, 40, 56, measure (expressions) 13, 144, 206, 208 see also null
108, 113n1, 193, 206, 208 146, 148–151, 153–155, 158, patients
see also subjective genitive 185, 195 perfective meaning, value 68,
marking the patient 8, 19, 191
34, 40, 45, 54, 90–93, 106, N perfective noun 43, 47, 49,
108, 113n1, 187, 203, 206, nominative 13, 24–27, 114, 124, 51–53, 97, 102–103, 105,
208 see also objective 133, 136, 196 107–108
genitive nouns (nominals, perfective suffix 73
noun phrases with two nominalizations) perfective verb 27, 43, 96,
genitives 10, 27, 91, 198, action nouns 29, 51, 55, 98–103, 105, 107–108
200 189n9 plural number (with
objective genitive 186, 191, agent (actor) nouns 7, 11, deverbal nouns) 10n9,
194, 196, 198–199, 200, 22, 32, 35, 184–187, 202 62, 67, 72n20, 100–102,
203, 207–208 denoting actions (nouns 119, 193
subjective genitive 191, 194, with action meaning) plurality 116, 119, 127
198–201, 203, 207–208 21–22, 25, 29, 36–38, polysemy (polysemic) 4, 64,
with measure 40–41, 43, 45–48, 51, 68–69, 71–74
expressions 147, 195, 206 54–57, 61–63, 191–193, possessive
with relational nouns 200, 206 adjective 8–9, 10n9,
122–123, 127 denoting a result 13, 37, 25, 27n6, 28, 32, 40,
gerund 65–66, 90, 189–190, 43n15, 48–49, 55, 192–193, 90–92
194–195 206 determiner 75, 80, 77, 127
government 141–143, 147, 158 denoting a state 22, 25, pronoun 25, 27n6, 28, 32,
grouping (forms of) 144–146, 36–38, 43, 53, 56, 114, 191, 185, 187–188, 192–193
148–149, 151, 153–155, 158 200, 206 prepositional phrase
event nouns 37, 41, 61–64, adnominal 19, 25, 29–31, 36,
H 68, 73, 78–79, 84, 89–90, 64, 66, 75, 77, 84, 114, 126,
head-complement phrase 99 145–147, 186, 193, 196–198,
144–145, 150, 157–158 nouns with eventive 202, 204, 206
meaning (interpretation) as verb argument 95, 121,
I 37, 67–72, 74, 79–84, 99 123, 193
imperfective noun 36, 43, 47, result nouns 37, 41, 61–63, marking the patient 9, 13,
49–50, 54, 97, 105 89, 99 198–200
imperfective verb 97–98, state nouns 63–64, 78, 80, proper name 118, 137, 127–128,
100–101, 103–105 84 164–165, 179
incorporation (of participants null patients 98, 100–102,
into a valency frame) 104–109 Q
11–12, 29, 32–35, 187 see quantifier 101, 103, 107,
also absorption O 143–145, 148–150, 154, 158,
infinitive 25, 30, 64n7, 65–66, origin 3–4, 6, 15, 21 165–167, 170, 185, 195
173, 190
inner participant 1, 3, 6, 12, P R
17, 21, 36 parasitic gap 174–175 recipient 92, 183, 196, 203–204,
instrumental 9–11, 20, 40, 45, participle 52, 68, 184, 188–191 206
90, 93–94 passive (voice, sentence) 8, reciproc(ity) 114, 119, 121–122,
10, 25, 94, 106, 108, 126–127
J 207–208 reduction of slots (in the
juxtaposition 114–115, 118, 130, passive nominal 62 nominal valency frame)
136, 142 patient 3–4, 6, 14, 19, 21, 23, 25, 34–36, 44, 55, 205–206
28, 34–35, 42, 45–46, 56, referential
M 62, 90–112, 185, 187–188, nominal 63–64, 69, 70,
material 12, 23, 36 191–192, 196–200, 203, 71n19, 72–73, 77–80, 84
Index 
interpretation 67–69, 71, support verb construction 23, V

73–74, 84 43n16, 50, 52, 56, 190, 194, valency complementation 19,
relational noun 114–129, 201–202, 206 25, 29, 36–37, 44, 54–56
130–138 T valency frame
relative clause 65–66, 152, telic 96–97, 102 of nouns 15, 20, 23–25, 29,
162–180 theta-role 1, 90, 92, 94–95, 106, 36, 42, 44, 45–47, 54–55,
162, 170 90, 95, 99, 108, 134, 145,
S time expressions (with verbal 187–188, 192, 194, 199,
satellite 185–186, 189, 193, 200, nouns) 20, 40, 47, 51, 63, 200–201, 203, 205–206
204 97, 197 of verbs 3, 6, 21, 23

Noun Valency (PDFDrive) PDF

Uploaded by

Copyright:

Available Formats

Noun Valency (PDFDrive) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Noun Valency (PDFDrive) PDF

Uploaded by

Copyright:

Available Formats

Noun Valency

Studies in Language Companion Series (SLCS)

John Benjamins Publishing Company

the American National Standard for Information Sciences – Permanence

Library of Congress Cataloging-in-Publication Data

Noun Valency / Edited by Olga Spevak.

© 2014 – John Benjamins B.V.

For interlinear glosses The Leipzig Glossing Rules were used.

Phrase S­tructure Grammar (HPSG), the Minimalist Program within Genera-

be interpreted as a special coordination phrase that also implies reciprocity. Her

Ana-Maria Barbu is a Senior Researcher at the Institute of Linguistics of the

Elisa Bekaert is Research and Teaching Assistant of Romance linguistics at Ghent

Věra Dvořák received her M.Phil. in theoretical linguistics at the University of

Renata Enghels is Associate Professor of Hispanic linguistics at Ghent University.

Veronika Kolářová is Senior Research Associate at the Institute of Formal and

Petya Osenova is associate professor of Morphology and Syntax in the Division of

Jarmila Panevová is a Professor of General and Computational Linguistics at the

Olga Spevak is Assistant Professor of Classics at the University of Toulouse 2. Her

Contribution of valency to the analysis

1. Why valency?

An extension of the study of valency to the other parts of speech is again a

2. What is valency about?

i. A modifier may be connected with only a limited number of heads, which

3. The deletion/absence problem

ii. The VM is generalized. The notion of generalization of ACT occurs in tra-

(4) Na začátku věty se píše velké písmeno.

(6) a. Přátelé už přišli.

Under normal circumstances, the question in (6b) cannot be answered by the

4. Labelling of inner participants

1. As stated in Section 2 above, verbal modifiers (they number approximately

6. Valency of nouns

6.1 Valency of deverbal nouns

i. nouns derived by syntactic derivation, such as ošetření ‘treatment’ (see

6.1.1 Deverbal nouns derived by syntactic derivation

pacientovo ošetření od lékaře

(8) a. VF: V ACT(NOM), DIR18→ NV ACT(GEN) DIR1

The possessive adjective (kolegův) is excluded in (8c) by the syntactic rules of

8. DIR1 is used for the meaning of direction from where?

Two variants of morphemic expressions of noun valency modifications are con-

b. zbavení rodičů odpovědnosti

An instrumental form for an ACT is blocked in the transformation of the verbal

(15) a. Věřitelé vyhrožují podnikateli exekucí.

6.1.2 Deverbal nouns derived by lexical derivation

brusič / brusič drahých kamenů

In the following examples, it is the PAT which is absorbed by the word-formation

6.2 Primary nouns

6.3 Linguistic meaning or cognitive content?

(18) VF of (i): ACT (GEN / ADJ-POS / od + GEN)opt, PAT (o + LOC)opt;

(19) román L. N. Tolstého vs. portrét L. N. Tolstého

6.4 Conclusions on noun valency

7. Future perspectives for studies of noun valency

Panevová, Jarmila. 2008. Povaha stupňování adjektiv: K “nesrovnávacímu” užití stupňovaných

Appendix 1. Inner participants of verbs in PDT 2.0

IP’s total Participants expressed Participants

ACT(or) 87,111 74,280 7,523 9%

Appendix 2. Inner participants of nouns (with functions

IP’s total Participants expressed Participants

ACT(or) 20,786 9,689 8,345 40%

Special valency behavior of Czech

the American National Standard for Information Sciences – Permanence

Phrase Structure Grammar (HPSG), the Minimalist Program within Genera-

a participant in a propositional character (PPC)9 even if it is expressed by a noun

ifferent forms of Patient can be captured in one valency frame as variants, as