Вы находитесь на странице: 1из 227


Studies in Natural Language and Linguistic Theory


Managing Editors
Liliane Haegeman, University ofGeneva
Joan Maling, Brandeis University
James McCloskey, University ofCalifornia. Santa Cruz

Editorial Board
Carol Georgopoulos, University ofUtah
Guglielmo Cinque, University ofVenice
Jane Grimshaw. Rutgers University
Michael Kenstowicz, Massachusetts Institute ofTechnology
Hilda Koopman, University ofCalifornia. Los Angeles
Howard Lasnik, University ofConnecticut at Storrs
Alec Marantz. Massachusens Institute ofTechnology
John J. McCarthy, University ofMassachusens. Amherst
lan Roberts, University ofWales. Bangor

TM titles published in this series are listed at the end ofthis volume.
Department 0/ Linguistics,
Tel Aviv University,lsrael


The Syntax of DPs

Library of Congress Cataloging-in-Publication Data
SIlonI, Tat.
Noun phrases and no.lnallzatlons : tha syntax of OPs I by Tal
p. e., -- fStudles In natural languaga and llngulstle theory
; v, 401
Based on the author's thesIs Idoetorall--Unloerslty of Oaneva,
Ineludes blbllographleal reFerenees fp, 1 and Index,
ISBN 978-90-481-4866-0 ISBN 978-94-015-8863-8 (eBook)
DOI 10.1007/978-94-015-8863-8
1. Ora ••• r, Co.parattop. .nd general--Noun phrase. 2. Gra •• ar,
Co.par.t'.e .nd general--No.tnals. 3. Gra ••• r, Co.para'lve and
gener.I--Syntax, I. Tltle. 11. Sertes.
P27 1.857 1997
415--de21 97-15607

ISBN 978-90-481-4866-0

Printed on acid-free paper

All Rights Reserved

© 1997 Springer Science+Business Media Dordrecht
Originally published by K.luwer Academic Publishers in 1997
Softcover reprint ofthe hardcover 1st edition 1997
No part of the material protected by this copyright notice may be reproduced or
utilized in any fonn or by any means. eJectronic or mechanical.
incJuding photocopying. recording or by any infonnation storage and
retrieval system, without written permission from the copyright owner.

Preface ......................................... ix

Chapter 1: Theoretical issues . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.0. Preliminaries ................................. 1
1.1. Nominalizations and DPs ........................ 2
1.1.1. Background ............................. 2 Nominalizations ...................... 2 Functional structure ................... 5
1.1.2. Outline ................................ 7
1.2. Some basic theoretical assumptions ................ 10
1.2.1. Levels of representation .................... 11
1.2.2. The economy guideline and the checking technique .. 12
1.2.3. Clause structure and Case ................... 13

Chapter 2: Noun raising and genitival relations . . . . . . . . . . . . . 19

2.0. Background ................................ 19
2.1. Construct states versus free states ................. 21
2.1.1. The article constraint ...................... 22
2.1.2. The position of modifying adjectives ........... 23
2.1.3. The genitive noun phrase ................... 25
2.2. Free states ................................. 26
2.2.1. Binding phenomena ....................... 28
2.2.2. The structure ............................ 30
2.3. Construct states .............................. 33
2.3.1. Setting the stage ......................... 35
2.3.2. Structural genitive Case .................... 40
2.3.3. The definiteness agreement .................. 45
2.3.4. Construct states and seI phrases .............. 47
2.3.5. Modifying adjectives and adjacency ............ 51
2.4. Clitic and clitic doubling constructions .............. 54
2.4.1. Data ................................. 54
2.4.2. Generation site .......................... 55
2.4.3. Analysis ............................... 57
2.5. Summary .................................. 59

Chapter 3: Event nominals . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.0. Introduction ................................ 65
3. 1. Event versus result nominals ..................... 67
3.1.1. Grimshaw's diagnostics . . . . . . . . . . . . . . . . . . . . . 68
3.1.2. More diagnostics ......................... 72
3.1.3. Syntactic derivation: the VP-analysis ............ 74
3.2. Adverbial PPs - not adverbs ..................... 75
3.3. Two types of accusative Case .................... 78
3.4. Inherent accusative ............................ 84
3.4.1. 'et insertion ............................. 84
3.4.2. The relevance of AgroP ..................... 86
3.4.3. Idiosyncratic information .................... 87
3.5. Subjectless event nominals ...................... 89
3.5.1. Against a passive analysis ................... 89
3.5.2. The implicit Agent ........................ 91
3.6. Concluding remarks ........................... 95
Appendix: Concrete nouns .......................... 96

Chapter 4: Semi-relatives and reduced relatives ............ 109

4.0. Introduction ................................ 109
4.1. Hebrew semi-relatives ......................... 111
4.1.1. Regular relatives versus semi-relatives .......... 111
4.1.2. The structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.1.3. The relative ha- versus the article ............. 123
4.2. On the parallelism between CP and DP ............. 126
4.3. French reduced relatives ....................... 129
4.3.1 Internal structure . . . . . . . . . . . . . . . . . . . . . . . . . 13 2 Past participie reduced relatives .......... 134 Present participle reduced relatives ........ 135 Tense ............................ 137
4.3.2. Against PRO............................ 140
4.4. Unified analysis ............................. 144
4.4.1. Selectional properties of D . . . . . . . . . . . . . . . . . . 144
4.4.2. The ban against wh-elements ................ 147
4.4.3. The variable ............................ 148
4.5. Concluding remarks .......................... 150

Chapter 5: Verbal and nominal ge runds .................. 159

5.0. Introduction ................................ 159
5.1. Verbal gerunds: properties ...................... 160
5.1.1. Verbal characteristics ..................... 160

5.1.2. Gerund clauses versus infinitivals ............. 163

5.2. Sentential approaches to verbal gerunds ............. 165
5.2.1. The P-CP analysis ........................ 166
5.2.2. The CP analysis ......................... 168
5.3. Nominalization .............................. 172
5.3.1. Against NP ............................. 172
5.3.2. Tense ................................ 176
5.4. DP-zation .................................. 177
5.4.1. Construct states ......................... 178
5.4.2. Internal structure ......................... 180
5.4.3. The nature of 0 ......................... 183
5.4.4. More on gerund clauses and noun phrases ....... 184
5.4.5. Gerund clauses and infinitivals ............... 188
5.5. Nominal gerunds ............................ 190
5.5.1. Properties ............................. 190
5.5.2. Lexical nominalization .................... 193
5.6. Conclusion ................................ 196

Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .... 201

References ..................................... 203

Index of names .................................. 213

Index of subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215


As is clear from its title, this book deals with two main topics. First, it
explores various aspects of the syntax of noun phrases. Second, it
suggests a systematic investigation of the categorial nature and syntactic
characteristics of nominalizations, which can be formed prior to lexical
insertion or in the syntactic component.
The syntactic nature of structures showing a categorially ambiguous
behavior is not always easy to determine, as the distinction between
syntactic phenomena that merely echo lexical information and those that
result from an actual syntactic operation is sometimes blurred due to
miscellaneous factors. In fact, the tension between the lexicon and the
syntactic component is natural under any approach assuming aseparate
lexical component. This tension, which, of course, is fed by theoretical
developments and empirical discoveries, is methodologically healthy as
it encourages the theory to reexamine the division of labor between its
components. The present study sheds some light on this issue as it
arises in the domain of nominalizations.
The book concentrates mainly on empirical data taken from (Modern)
Hebrew. The nominal system (in the broad sense) that characterizes
Semitic languages in general and Hebrew in particular is rich and
intriguing. It shows rather unique properties, whose investigation has
significant consequences for universal issues such as the characteristics
of syntactic nominalization or the relationship between functional and
lexical categories. It is thus not a mere coincidence that led me, a
native speaker of Hebrew, to explore the wonders of nominal
expressions. As is by now a common practice in syntactic research -
which aims to deepen our understanding of what is a possible variation
between related gramm ars and between language families, and what
remains constant across languages - this study adopts a comparative
methodology. The book often compares characteristics of Hebrew
grammar to properties shown by other grammars, Semitic or non-
Semitic, drawing conclusions of theoretical interest.
As its subtitle indicates, the book adopts the claim that syntactic
nominalizations share with noun phrases (whether simple nouns or
outputs of lexical nominalization) the same external layer, the functional


projection DP. It argues that the discrepancies between noun phrases (in
particular, event nominals) and syntactic nominalizations follow from
the fact that noun phrases have an internal nominal structure, while
syntactic nominalizations entail a verbal projection that is allowed to be
dominated by the functional projection DP because it lacks temporal
specifications. If my proposals are on the right track, then the nominal
properties of syntactic nominalizations never stern from a syntactically
present NP.
The first part of the book is devoted to noun phrases. It examines
different sorts of genitival constructions, analyzes their structure and
justifies a lexicalist approach to deverbal event nominals. The second
part of the book examines instances of syntactic nominalization, defines
the context required for propositions to be nominalized in syntax and
discusses some (a)symmetries between nominalized propositions and
their sentential counterparts.
The book has grown out of my doctoral dissertation completed at the
University of Geneva in March 1994. While working on the material in
the book, I have had fruitful discussions and exchanges with many
linguists and colleagues. It would hardly be possible to acknowledge all
those who have contributed to my work in one way or another. I would
nonetheless like to thank again the many people acknowledged in my
dissertation, in particular, Luigi Rizzi and Hagit Borer, who have been
of central influence from the beginning, both through their comments
and suggestions, and through the example set by their own research.
Among those who have read earlier vers ions of the manuscript or
various components thereof and extensively eommented are Adriana
Belletti, Guglielmo Cinque, Mare-Ariel Friedemann, Liliane Haegeman,
Tanya Reinhart, Ur Shlonsky, and an anonymous SNLLT Reviewer. The
material in the book has benefitted from presentations at the University
of Leiden, the Hebrew University of Jerusalem, MIT, CUNY, Ben
Gurion University of the Negev, the University of Niee, Bar-Han
University and the University of Geneva. Audienees at these plaees
made helpful suggestions. Those attending my classes at Tel Aviv
University also provided valuable discussions of mueh of the material
in the book. FinaUy, I would like to thank aU those who supplied data,
judgments, observations regarding Hebrew, Freneh and other languages,
and in partieular Aminadav Dykman and Mare-Ariel Friedemann for
their endless patience.

Tal Siloni
Tel Aviv



The ultimate goal of linguistic inquiry in the particular tradition known

as generative gramm ar is to understand the nature of the language
faculty. The central assumption is that humans are endowed from birth
with a system predisposed to the acquisition of a natural language (e.g.
Chomsky 1965, 1975). This assumption suggests a way to account for
the rapidity and apparent ease with which children acquire the
remarkable complexities of languages, without systematic instruction,
on the basis of incomplete data, and with no negative evidence.
The innate component of the human mind yields a particular language
through interaction with a particular linguistic experience. The idealized
model of language acquisition takes the initial state of the language
faculty to be a function mapping linguistic experience into a natural
language. The theory of the initial state of the language faculty, prior
to any exposure to linguistic data, is called universal grammar (UG).
UG determines the class of possible languages. The theory of the steady
state, that is the state of the language faculty of a person who knows a
particular language, is often called grammar.
UG must reconcile two seemingly conflicting requirements: it must
be predetermined enough to explain the process of native language
acquisition, and at the same time sufficiently flexible to allow the
diversity of natural languages. In the beginning of the eighties, certain
ideas regarding UG crystallized into the principles and parameters
approach (e.g. Chomsky 1981). Under this approach, UG consists of
certain invariable principles that hold of any natural language, and
finitely valued parameters to be set through the particular linguistic
experience of the learner. UG supplies a format of principles and
parameters; a particular instantiation of this format constitutes a specific
language. The parameters provide UG with flexibility and account for
the diversity of languages, apart from Saussurean arbitrariness (the
phonological encoding of concepts in the lexicon). Often, clusters of
different properties distinguishing two or more languages can be
reduced to a single difference, to a distinct setting of one single
parameter (for instance, see Rizzi 1982, 1986a for discussion of the

Null Subject parameter and related characteristics). Linguistic variation

is constrained by the principles and parameters of VG, which suggests
a way to explain the considerable rapidity and ease of native language
acquisition. The parametric range may be restricted to functional
elements and general properties of the lexicon (Borer 1984, Chomsky
1991, 1993).
Many specific variants of the approach have been developed and
explored in recent years. The current diversity of notions within the
generative approach necessitates a short introduction of the specific
path adopted here (section 1.2). Prior to this introduction, however, I
briefly discuss the goals of the research (section 1.1).


1.1.1. Background Nominalizations. It is weil known that verbs and their

corresponding deverbal nouns appear to share some basic semantic
properties. Thus, for example, the noun examination in (la) appears to
bear the same semantic relation to the noun phrases Dan arid the
papers, as the verb examined does in (lb). However, while the
appearance of those noun phrases is optional with examination (2a), it
is obligatory with examined (2b): .

(1) a. Dan's examination of the papers

b. Dan examined the papers.

(2) a. the examination

b. • Examined.

The issue of the relationship between verbs and their corresponding

deverbal nouns already enjoyed an important position among inquiries
in the earliest works in generative grammar. Lees (1960) had deverbal
nouns generated as clauses and mapped onto a noun phrase structure by
aseries of nominalization transformations. This derived the fact that the
contexts in which a verb and its derived noun appear are closely
related. The differences between the two categories were accounted for
by ordering certain roles after the nominalization transformations.
Within the framework of the theory available at that time, there was,

in fact, no alternative way to express the similar properties of verbs and

their related nouns. Chomsky's Syntactic Structures (1957) lacked a
lexical component in the current sense, and could not formulate the
affinities between verbs and nouns in lexical terms. These affinities had
to be handled by the syntactic-transformational component. With the
introduction of aseparate lexicon (Chomsky 1965), it became possible
to express the relationship between verbs and deverbal nouns via lexical
representations, without assuming that deverbal nouns entail a syntactic
transformation of the source verb. The restricted productivity that
characterizes the formation of deverbal nouns, certain idiosyncrasies
they show, their nominal behavior, among other things, led Chomsky
(1970) to take a lexica/ist position with respect to deverbal nouns. In
rough terms, this means that entries like verbs and deverbal nouns share
their lexical representations as far as their thematic properties are
During tl1e eighties, variants of the lexicalist approach to deverbal
nouns have been dominant. Linguists have generally agreed that
deverbal nouns are inserted in the syntactic component as nouns, and
have been concerned with the extent and character of similarities and
differences nouns and verbs show with respect to argument structure
and 6-theory (Cinque 1980, 1981, Milner 1982, Anderson 1983-84,
Kayne 1984, Safir 1987, Zubizarreta 1987, among others). It has often
been asserted that nouns, contrary to verbs, take arguments only
optionally (see, for example, Higginbotham 1983, Dowty 1989).
In a consequential study ofthe nominal system, Grimshaw (1990) has
established clear diagnostics to distinguish between two types of nouns
that are often homophonous: event nouns, which express an event (or
a process), and result nouns, which name the output of the event or an
entity related to it. This disambiguation enables Grimshaw to show that
event nouns obligatorily have an argument structure as part of their
lexical representation; they assign specific 6-roles, just Iike verbs. The
lexical representation of result nominals, which do not express an event,
does not specify an argument structure; result nouns do not take real
arguments, which bear specific 6-roles, but rather a kind of semantic
participants that are more loosely associated with them.
For example, Grimshaw shows that certain modifiers, like frequent,
can modify a (singular) noun only when it expresses an event. They can
thus serve to diagnose eventhood. Once a noun is disambiguated, it
becomes clear that a noun without arguments cannot have an event
interpretation. In (3a) the referent of Dan is somehow associated with
the referent of construction, which is a concrete entity. Dan can be the
owner, the caretaker, the admirer, or the creator of this entity. Thus,

Dan is not a real argument of construction as it does not bear a specific

6-role; rather it has some claim of possession over the concrete entity.
When the modifier frequent is added, construction is forced to have an
event interpretation, and consequently it assigns specific 6-roles (3b),
just like its corresponding verb (3c). Hence, Dan must be interpreted as
the agent of construction, and the appearance of the constructed
element becomes obligatory:

(3) a. Dan's construction impressed uso

b. Dan's frequent construction *(ofsailing boats) impressed uso

c. Dan constructs *(sailing boats).

In short, lexical entries that denote an event (whether verbs or nouns)

have an argument structure. The clear split between event and result
nominals undoubtedly highlights the common properties verbs and event
nouns share. Certain important asymmetries between verbs and deverbal
nouns are in fact to be associated only with result nominals. This has
paved the way for the revival of the syntactic approach to event
The modern syntactic approach takes the presence of an event
reading and an argument structure to be a lexical property of verbs, not
nouns (e.g. Borer in progress). It inserts deverbal nouns as verbs that
raise to incorporate with a nominal head in the course of the syntactic
derivation, as schematized in (4). Event nouns thus have an event
reading and an argument structure because they contain a verbal
projection in syntax:

(4) NP




Apriori, the syntactic approach simplifies the lexicon, dispensing

with the need to lexically represent the nominal ambiguity
(event/result). It does so at the cost of greater complexity of the
syntactic component, which must allow structures of the type in (4).
The lexicalist approach, in contrast, simplifies the syntactic component
at the price of a richer lexicon. As noted by Chomsky (1970), there do
not seem to be general considerations that settle the matter; deciding
between the two approaches is mainly an empirical problem.
Alongside event nominals, languages also exhibit gerundive forms
such as the English gerund given in (5). On a par with event nominals,
English gerunds of the type in (5) have the distribution of noun phrases
and take a genitive subject (5a). Unlike event nominals, however, they
are formed fairly freely, their semantic interpretation is straightforward
with regard to the source verb, and their internal structure is not
nominal (e.g. they cannot be modified by adjectives, nor realize their
article (5b-c», but rather verbal (e.g., they can take an accusative
argument (5a»:

(5) a. John's constructing sailing boats impressed uso

b. *John's rapid constructing sailing boats impressed uso

c. * The constructing sailing boats impressed uso
Any study of nominalizations has to take up the challenge of
accounting for the discrepancies and similarities between event
nominals and their gerundive counterparts. If there are good empirical
reasons to believe that both event nominals and ge runds are derived
from the base verb in the syntactic component, the more verbal nature
of gerunds is apriori unexpected and requires an explanation. In
contrast, if it can be shown that event nominals are the output of a
process of lexical nominalization and gerunds are the product of
syntactic nominalization (as suggested in Chomsky's Remarks on
Nominalization 1970), the discrepancies between the two
nominalizations follow rather straightforwardly. Functional structure. A more recent issue in the investigation

of noun phrases concerns the functional structure they entail.
Traditionally, noun phrases were naturally taken to be the maximal
projections ofN, as depicted below (Jackendoff 1977, among others):

(6) NP

Determiner N'


The theoretieal developments in the eighties alongside the growing

interest in the nominal system have ealled this strueture into question.
More speeifieally, the extension of the X-bar schema to the sentential
funetional elements (Chomsky 1986b) and the inereasing understanding
of the workings of head movement (Chomsky 1986b, Baker 1988) have
led linguists to elaborate a more artieulated syntaetie representation for
the noun phrase.
Arguing that the funetional nominal material, too, should fit into the
X-bar schema, Abney (1987) has hypothesized that noun phrases, like
elauses, are headed by a funetional element. He has proposed that noun
phrases are the maximal projeetions of D, the base position of artieles
(see also Szabolesi 1983-84, Fukui and Speas 1986):

(7) DP



This proposal has reeeived strong empirieal support by aseries of

studies arguing that the head noun overtly raises to D in Semitie (Ritter
1987, 1988, Mohammad 1988, Ouhalla 1988, Fassi Fehri 1989, Hazout
1990, Siloni 1990b, 1991a) and Seandinavian languages (Delsing 1988,
Taraldsen 1990). Longobardi (1994) has shown that instanees of noun
raising to D are also likely to oeeur overtly in Romanee and eovertly in
English and German. Subsequent studies of Semitie, Romanee and
Germanie languages have suggested that the strueture of noun phrases
is even more artieulated and ineludes additional infleetional strueture
between DP and NP (Ritter 1991, Valois 1991, Cinque 1993, Bernstein
1993, Fassi Fehri 1993, Penner and Sehönenberger 1993, among others).
Alongside the aeeumulating empirieal evidenee, several studies have
developed a prineipled explanation of why the nominal expression is the

maximal projection of D (Szabolcsi 1987 ,1989, Stowe11 1989, 1991,

Longobardi 1994). Somewhat simplified, their insight is that D is the
element that converts the nominal expression into a referential phrase,
which consequently is able to serve as an argument. In this respect, it
can be argued that D paralleis the complementizer of sentential
complements: each turns its complement (NP and IP, respectively) into
an expression that is able to appear in an argument position, that is, to
bear a 6-role (Szabolcsi 1987 ,1989).
While the discussion of nominalizations is particularly concerned
with their categorial structure, recent investigations of the structure of
noun phrases are interested in defining their functional structure and its
workings. These related issues, which I informally phrase below, form
the grounds for this research on noun phrases and nominalizations,
which concentrates on a variety of constructions in (Modern) Hebrew,
often comparing them to parallel constructions in Semitic and non-
Semitic languages:

A. The categorial syntactic structure of event nominals and other


B. The functional structure of DPs and the different facets of D.

In the subsequent section I outline the main claims advanced in this


1.1.2. Outline

The first part of this work (chapters 2-3) has two major goals:
examining the functional structure of noun phrases, and justifying a
lexicalist approach to event nominals. I supply empirical evidence that
event nominals are purely nominal and do not contain a syntactically
projected VP. Yet they share with the corresponding verbs their
argument structure (Grimshaw 1990), which they equally map onto a
hierarchical syntactic structure (Giorgi and Longobardi 1991). I present
the empirical arguments Hebrew offers in favor of viewing noun
phrases as maximal projections of D. Further, I motivate the projection
of an agreement phrase between NP and DP in noun phrases involving
the so-called construct state. This allows a straightforward account of
the distinct syntactic behavior of three types of genitival constructions
in Hebrew: the construct state, the free state, and the clitic doubling

The second part (chapters 4-5) is devoted to cases of "mixed"

structures, DPs containing a verbal projection (the structures show
typical verbal characteristics). The question immediately arises as to
what it is that enables D to take a verbal complement instead of its
standard nominal complement. Observing that verbal projections
embedded under D are non-tensed propositions, I suggest that their
untensed nature is the crucial factor that makes them legitimate
complements of D. This is what they share with NPs and this is what
renders them compatible with D. According to my proposal, embedding
by D of a verbal projection is what syntactic nominalization is in
essence. Syntactic nominalization does not involve a syntactic
transformation incorporating a verb into a noun. Rather, it entails a DP
dominating a verbal projection that does not contain tense
specifications. Thus, English gerunds, Hebrew gerunds, or Italian
nominalized infinitives may all be cases of syntactic nominalization. I
suggest a detailed analysis of Hebrew gerunds along these lines.
Inspired by the analogy between articles and complementizers
outlined by Szabolcsi (1987, 1989), I characterize D as the equivalent
of C with regard to non-tensed phrases. Both C and D turn the
expression they head into a referential argument, wh ich is consequently
able to bear a 6-role. But while C heads tensed propositions, D
introduces non-tensed phrases. Following Stowell (1982), I assume that
the CP level is obligatory in tensed clauses (whether finite clauses or
infinitivals). Stowell entertains the idea that the tense operator has to
raise to C (COMP) at LF to take scope over its clausal operand (see En y
1987 for a detailed discussion of the Anchoring Conditions, which link
tense to C). Now, if C must be associated with a tense operator,
whereas D cannot do so, it becomes clear why C must introduce tensed
clauses, while D is the "complementizer" of non-tensed expressions,
whether noun phrases or gerund clauses.
The investigation of participial relatives strongly reinforces this
functional parallelism between CP and DP. In Hebrew (or Standard
Arabic), these non-tensed relative clauses surface headed by D. In other
languages (e.g. French), they do not manifest any overt element of this
type. Various considerations, however, suggest that they do contain a
covert D. This strengthens the claim that verbal projections can be
embedded under D only if they are not tensed. To the extent that
syntactic nominalization means the occurrence of a verbal constituent
as a component of DP (say, DP-zation of a VP), participial relatives
constitute an additional instance of this syntactic phenomenon.
Moreover, the occurrence of D as the head of participial relatives
suggests that D can not only introduce non-tensed argumental phrases,

but also non-tensed modifying phrases. This brings to light a novel facet
of D and extends the functional analogy between complementizers and
articles: C as weil as 0 can head arguments as weil as modifiers.
Let me summarize the main arguments of the following chapters in
rough lines.

(a) Event nouns basically share the same argument structure with the
corresponding verbs (see Grimshaw 1990), and map it, like verbs,
onto hierarchical syntactic structures (Giorgi and Longobardi
1991) (chapter 2).

(b) Given the hierarchical structure of noun phrases and the order of
constituents they exhibit, it must be concluded that overt noun
raising is obligatory in Hebrew. I suggest that 0 is the landing
site of the raised noun, thus supplying support for the claim that
noun phrases are the maximal projections of 0 (Abney 1987)
(chapter 2).

(c) Genitival relations in Hebrew can be expressed via the construct

state, the free state, or a clitic doubling construction. The
construct state avails itself of structural Case, the free state
involves inherent Case assigned via the Case marker sei ('of'),
and the clitic doubling configuration has recourse to both Case
assignment mechanisms. If structural Case is always the
realization of Spec-AgrO relation (Chomsky 1991, 1993),
construct states and clitic doubling configurations entail an
agreement projection. The syntactic properties of all three
genitival constructions fall out (chapter 2).

(d) Hebrew event nominals show some arguably verbal properties:

they can take accusative arguments and be modified by adverbs.
This seems to justify a syntactic approach to event nominals
(Hazout 1990, 1995, Borer in progress). I show that the verbal
properties of Hebrew event nominals are only apparent: the
accusative Case of event nominals is an inherent Case assigned
by a Case marker, and the adverbs that can modify them are all
adverbial PPs and not genuine adverbs. There are no empirical
reasons to believe that Hebrew event nominals in particular, and
event nominals in general, contain a verbal projection. On the
contrary, a lexicalist approach can better handle the data. I
suggest that syntactic incorporation of V into N is not a process
allowed by UG (chapter 3).

(e) There is a functional analogy between articles and

complementizers. D is the equivalent of C in non-tensed phrases.
D, just like C, renders its complement a referential expression,
which is able to bear a 6-role (Szabolcsi 1987, 1989). Moreover,
D, on a par with C, can introduce modifying phrases. While C
heads tensed sentential structures, D heads nominal expressions
as weil as non-tensed verbal projections such as participial
relative clauses. I offer a detailed analysis of participial relatives
in Hebrew (where D is overt) and French (where I suggest it is
covert) along the above lines (chapter 4).

(f) Syntactic nominalization always entails a DP (not an NP)

dominating a verbal projection that lacks tense specifications.
Analyzed along these lines, the particular behavior of Hebrew
gerunds receives a straightforward account, as do the similarities
and distinctions between them and event nominals (chapter 5).

Before turning to the study itself, I briefly set the theoretical

framework. The current diversity of approaches with respect to central
notions in generative grammar necessitates a short introduction of the
specific path adopted in this work. The presentation does not intend to
offer a comprehensive discussion, but rather to draw the basic
theoretical assumptions. For detailed discussions, the reader is referred
to the references cited throughout the presentation. Notions directly
relevant to the study will be explained in the pertinent chapters.


This study adopts a minimalist approach to Iinguistic theory and is

embedded in the minima/ist program, as put forward in Chomsky (1993,
1995:chapter 3). As its titte suggests, the minimalist program is a
research design (not a worked-out theory), which assumes minimalism,
or simplicity, as a central criterion. Continuing the theoretical trend in
generative grammar to move from specific grammatical rules to simple
general principles that interact to produce linguistic expressions, the
program suggests a severe tightening of the Iinguistic apparatus
developed in recent years. Like earlier versions of generative grammar,
the approach assumes that there is a component of the human brain
dedicated to language, and that the language faculty has a cognitive
system that interacts with the performance systems by means of levels
of representations. Additional assumptions are subject to critical

The notion of simplicity is essential to the program not only as a
forceful working hypothesis but also as a theory internal notion. Like
any rational inquiry, generative grarnmar has always adopted simplicity
as a theoretical criterion, a natural procedure to reach explanatory
adequacy. And, indeed, over the years it has been repeatedly shown that
overlapping principles were simply wrong formulations. As a theory
internal notion, simplicity is argued to be an essential characteristic of
the computational system of the language faculty. It is instantiated in
the form of economy principles selecting among derivations.
Ongoing research in the minimalist program has already gone through
somewhat distinct stages. The newly developed linguistic mechanisms
are constantly subject to investigation and reconsideration (see, in
particular, Chomsky 1995:chapter 4). As will become clear in what
follows, this book adopts the original minimalist design (Chomsky
1993, 1995:chapter 3) as a working framework.

1.2.1. Levels 0/ representation

In descendants of the extended standard theory (EST, as developed by
Chomsky 1973, 1975, 1976), each linguistic expression has been taken
to be a sequence of representations at several levels: D-structure, S-
structure, Phonetic Form (PF), and Logieal Form (LF) (Chomsky and
Lasnik 1977, Chomsky 1981). These levels constitute the computational
system of the language, which is fed by aseparate lexicon. D-structure
is said to be a pure representation of thematic relations; the linking
level between the lexicon and the computational system, formed by an
"all-at-once" insertion of lexieal items. Mapping to the following level,
S-structure, takes place through the application of a general rule, Move-
IX, which displaces an element leaving a trace in the original position.
S-structure branches to LF and PF independently. LF is the interface
level with the conceptual-intentional faculties of the brain. PF is the
interface level with the articulatory-perceptual faculties. S-structure is
related to each of the three other levels simultaneously. The properties
of each level and the conditions it has to satisfy are specified (or
pararnetrized) by UG.
While LF and PF are conceptually necessary interface levels, D-
structure and S-structure have only theory internal motivation. It is a
subtle question whether or not they are indispensable (see, for instance,
Baker 1988, Chomsky 1991). A minimalist approach would take the
interface levels to be the only levels of representation, trying to account

for the considerable empirical consequences of the additional levels in

some other way. This is the position taken by Chomsky's Minimalist
Program for Linguistic Theory (1993). I adopt this particularly simple
approach, showing that at least the empirical issues discussed in this
book do not necessitate having recourse to conditions that apply at the
additional levels of representation.
The minimalist model is derivational and works as follows (for
detailed discussion, see Chomsky 1993). The lexicon specifies a set of
items with their phonetic, semantic and syntactic idiosyncratic
properties. The computational system, which is constrained by economy
principles (see below), uses these items to derive linguistic express ions.
A derivation converges at PF if its PF representation is legitimate, and
crashes if it is not. Likewise, a derivation converges at LF if its LF
representation is legitimate, and crashes if it is not. The computational
system selects elements from the lexicon and projects them in parallel
into X-bar structures. Two types of operations are possible: a binary
operation, which forms a single phrase marker from two distinct phrase
markers, and a singulary operation, Move-cx, which applies within one
phrase marker, leaving a trace in the original position. There are no
radically empty positions; positions are created only to be filled. At any
point, the operation of speil-out can apply, switching to PF. By then the
derivation must have been merged into one phrase marker, or else the
derivation crashes at PF. After speil-out, (covert) computation may
continue until an LF representation is generated, but there is no more
access to the lexicon.

1.2.2. The economy guideline and the checking technique

As already mentioned, notions of economy are fundamental to the

minimalist model. Derivations and representations are forced to be
economical in a sense to be discussed shortly.
Entries have inflectional features in the lexicon as an intrinsic
property. A lexical item is in~rted with its inflectional features (Case,
agreement, tense or others), which must be checked against the features
of the corresponding inflectional head by LF. A derivation containing
unchecked features will crash at LF.
The minimalist program suggests deriving overt movement by means
of the checking technique (Chomsky 1993). The features on the
inflectional head disappear once checking has taken place. Initially they
can be strong or weak. Strong features must be checked prior to speIl-
out as they are not legitimate objects at PF. Movement executed in

order to check weak features cannot take place before LF, due to
Procrastinate, an economy principle that states that movements should
be delayed as long as possible. Thus, for example, the differenee
between English and French concerning verb raising is due to the
distinct force of the features to be cheeked on the relevant inflectional
head. In French, they are strong and consequently movement is overt.
In English, in contrast, they are weak, which forces covert movement.
Once an element has ehecked all its features, it can no more move
because movement, according to the economy guideline, is a Last
Resort operation (Chomsky 1986a, 1991, 1993). An element ean move
only to check its features; a legitimate element cannot move further.
The checking technique dispenses the theory with the need to assume
in English-type languages (where the infleeted verb does not overtly
raise) an unusual proeess of (inflectional) affix-Iowering. The verb has
its inflectional features from the outset, and their ehecking takes plaee
at LF. The theory need not assurne lowering.
Derivations must contain the shortest possible links (Shortest
Movement Condition), in essenee, a derivational reformulation of the
basic insight of the system proposed in Rizzi's Relativized Minimality
(1990). At the same time, they must also have the smallest number of
steps. To reconeile these two apparently contradictory requirements,
Chomsky (1993) defines the operation forming a chain (for example, an
A-chain, whose tail is a 6-position and whose head is a Case position)
as one single step. In terms of ehains, derivations can have the fewest
steps and the shortest links.
Full Interpretation, which requires that every element receive an
appropriate interpretation at the interface levels, may be the only
economy principle applying to representations (Chomsky 1986a, 1993).
Note that some version of 6-theory must be assumed at LF (for
example, the principles of 6-discharge suggested by Higginbotham
1985). The system itself rules out raising to 6-positions (Chomsky
1993), thereby deriving the major consequences of the 6-criterion,
which requires a biuniqueness relation between an argument (or an A-
chain) and a Case position (Chomsky 1981).

1.2.3. Clause structure and Case

The computational system projects structures constrained by X-bar

theory (Jackendoff 1977). For the purposes of this work, it is not really
important whether the principles constructing X-bar structures are
primitives, or can be derived from more basic principles (see Chomsky

1994 and Kayne 1994 for specific suggestions). An X-bar structure

consists of projections of heads. These projections are configurationally
uniform across categories, lexical as weil as functional (Chomsky
1986b). The basic X-bar structure assumed here is a two-level
configuration restricted by binary branching (Kayne 1984) and
composed of a maximal projection (XP) containing a specifier and an
intermediate projection (XI). XI contains the head of the projection (XO),
whose sister is the complement. The linear order may be subject to
parametric variation (e.g. Rizzi 1987) or universal (as argued by Kayne

(8) XP

Spec XI

A Compl

The head-complement relation is typically associated with thematic

relations. The head-head relation (resulting from head adjunction) and
the Spec-head relation (resulting from XP-substitution, which creates
the specifier position) are the core configurations for checking of
inflectional morphology.
Two types of features characterize inflectional heads: XO-features and
XP-features. Thus, in the clausal system, an inflectional head bears V-
features, which are checked in a head-head relation with the raised
verb, and DP-features (Case features and agreement features), wh ich are
checked in a Spec-head relation after raising of the appropriate DP. DP-
positions in a local relation with lexical heads or with heads checking
lexical features (say V-features) are the traditional A-positions.
Chomsky (1986a) draws a distinction between structural Case and
inherent Case. Inherent Case is assigned by IX to DP only if IX 8-marks
DP, while structural Case imposes no such thematic requirement. This
means that accusative and nominative are structural Cases as they are
not thematically related. Oblique Case assigned by prepositions or
dative are inherent Cases, as their assigners also 8-mark the assignee.
Inherent Case is assigned in situ under sisterhood, while structural Case
has often been argued to be available either und er government or in a
Spec-head configuration (Koopman and Sportiche 1991, Roberts 1993).
As will beeome elear presently, this disjunetive formulation may be
dispensed with.

Elaborating proposals by Pollock (1989) and Belletti (1990),

Chomsky (1991, 1993) proposes the following basic sentential structure,
which splits the inflectional structure into subject agreement phrase
(AgrsP)' tense phrase (TP), and object agreement phrase (AgroP):

(9) CP


Agro VP

He further suggests unifying the conditions necessary for structural Case

assignment (or checking) to take place. Structural Case can be
conceived as the manifestation of a Spec-AgrO relation. The basic
assumption is that there is a symmetry between the subject and the
object inflectional systems. The object raises to SpecAgroP and checks
accusative Case with the complex head [Agr V+Agr]. And the subject
raises to SpecAgrsP and checks nominative Case with the complex head
[Agr T+Agr].
AgroP' where structural accusative Case is checked, has commonly
been identitied with the participle agreement projection (see Belletti
1990, Chomsky 1991, Kayne 1993, among others). A participle
agreement projection was first proposed by Kayne (1985, 1989a) in
order to account for participle agreement with the (overtly) raised direct
object in French or Italian.
Friedemann and Siloni (1993) provide considerable evidence against
this identification. First, they point out the fact that in French (and
ltalian) participle agreement is obligatory in passive sentences and with
unaccusative verbs, although there is no accusative argument in the

(10) a. Cette porte sera ouvert-*(e) par Johnny.

this door(FMSG) will+be opened-FMSG by Johnny

b. Comelia est arrive-*(e).

Cornelia is arrived-FMSG

In addition, they mention cases where participle agreement is triggered

with a non-accusative element, although an accusative argument appears
in the sentence:

(11) Maria si e comprat-a un libro.

Maria to+herself is bought-FMSG a book(MSSG)

Moreover, they observe that in Hebrew, participles always agree with

the nominative argument, whether an accusative argument is present or
not in the sentence:

(12) a. hem hayu kotv-im 'agadot ba-yad.

they(MsPL) were writing-MsPL legends(FMpL) in+the-hand
'They were writing legends by hand'

b. hem hayu magi'-im ba-zman.

they(MsPL) were coming-MsPL in+the-time
'They were coming on time'

As accusative Case checking does not coincide with participle

agreement, the two phenomena are not likely to take place within the
same projection. If so, there are two distinct agreement projections:
AgroP' where accusative Case is checked, and AgrpP, where participle
agreement takes place. This raises the question as to their relative
positioning. Friedemann and Siloni (1993) show that in complex tenses
AgroP is generated in distinct locations in French-type languages and
Hebrew-type languages. While in French AgroP is associated with the
VP of the auxiliary (the highest auxiliary) (13a), in Hebrew it must
immediately dominate the participial VP (13b):

(13) a. [AgrsP hp [AgroP [VPaux V [AgrpP [vPp OPs [VI Vp OPo ]]]]]]]
b. [AgrsP hp [VPaux V [AgrpP [AgroP [vPp OPs [VI Vp OPo ]]]]]]]

This structural difference results from two factors. First, participles

appearing in complex tenses in the two types of languages are of

distinct nature. While the Hebrew form can assign accusative Case by
itself (see Siloni 1995), the French form cannot do so (see Hoekstra
1984, Belletti 1990). In structural terms, this means that in Hebrew (but
not in French) the participial VP Iicenses AgroP' Second, French, but
not Hebrew, has a have-type auxiliary, which is the auxiliary that
restores the accusative Case capacity of an otherwise passive participle,
like mange ('eaten') (Hoekstra 1984). In structural terms, this means
that in French (but not in Hebrew), the VP of the auxiliary can Iicense
As Friedemann and Siloni (1993) show, this proposal allows a
straightforward explanation of participle agreement phenomena in the
two languages. Moreover, it correctly predicts the possible main
constituent ordering the two languages show and them only. The
discussion of impoverished (non-tensed) clausal structures embedded
under D, participial clauses (chapter 4) and Hebrew gerund clauses
(chapter 5) supplies further support in favor of the sentential structures
in (13).
Before concluding, a word on the status of AgrPs is in order. The
function of AgrPs is to provide a structural configuration in which
features are checked. Overt movements of the subject and/or the object
(depending on the language) are taken to show that feature checking
takes place outside the base positions, arguably in the corresponding
AgrP. In his recent work, Chomsky (1995:chapter 4) suggests moving
from an Agr-based system to a system with no AgrPs, in which
checking takes place in additional specifiers. This move means that the
features of the different AgrOs should be added to the relevant heads
(say T or V). Thus, for example, in the revised system the object would
check its Case with V, in an (additional) outer specifier of VP and not
in SpecAgroP' and the subject would check its Case with T, in a
specifler of TP and not in SpecAgrsP' This move is natural in a
minimalist model, because AgrPs, whose motivation is theory internal,
are not indispensable. In the present study, I employ an Agr-based
system to draw detailed structures more as an expository device than for
theoretical reasons. I do not think that this choice of working hypothesis
has any crucial consequences; much of the discussion can be easily
reconstructed with structures containing multiple specifiers.
In sum, the minimalist design is simple and restrictive. The empirical
burden is consequently rather considerable. Many empirical domains
demand a thorough examination. In this study I try to shed some light
on the domain of noun phrases and nominalizations.



Traditionally the noun phrase has been structurally represented as an

NP, the maximal projection of N, with the determiner in its specifier
position (Jackendoff 1977, among others):

(1) NP



The extension of the X-bar schema to functional elements (Chomsky

1986b) and the accumulating evidence for head movement (Baker 1988,
among others) have called this representation into question. On the
basis of various empirical considerations, linguists have proposed that
the noun overtly raises in Semitic (Ritter 1987, 1988, Mohammad 1988,
Ouhalla 1988, Fassi Fehri 1989, Hazout 1990, Siloni 1990b, 1991a,
Cohen 1992) and Scandinavian languages (Delsing 1988, Taraldsen
1990). The syntactic behavior of articles in both families of languages
has led scholars to identify the landing site of the raised noun as 0, the
base position of articles. This has supplied strong support in favor of
Abney's (1987) claim that noun phrases are the maximal projections of

(2) OP




Alongside the empirical motivation, several studies have developed

a principled explanation of why the nominal expression is the maximal
projection ofD (Szabolcsi 1987, 1989, Stowelll989, 1991, Longobardi
1994). Somewhat simplified, their insight is that D is the element that
provides the nominal expression with reference, which is necessary in
order for the noun phrase to be able to function as an argument.
Szabolcsi (1987, 1989) notes that in this respect, D paralleis the
complementizer of sentential complements, as both render their
complement (NP and IP, respectively) an expression that is able to bear
a 8-role. Siloni (1990a, 1995) further advances the functional analogy
between CP and DP, thereby arguing that D should head the noun
phrase in the same way that C heads the clause (see chapter 4 for
This chapter reviews the basic empirical evidence offered by Hebrew
in favor of noun raising, hence in favor of the structure depicted in (2).
Relying specifically on Siloni (1990b, 1991a), I elaborate a somewhat
different generalized noun raising analysis of two distinct genitiv al
constructions known as the construct state and the free state.
First, I show that Hebrew noun phrases have a hierarchical internal
structure. I then argue that given their hierarchical structure on the one
hand, and the order of constituents they exhibit on the other hand, it
must be concluded that in Hebrew overt noun raising always applies. If
so, there must exist an appropriate landing site to host the raised noun.
I argue that this host is the head position D, thus supplying support for
the structure given in (2).
Second, I present evidence that the Case of the construct state and
that of the free state are of different nature and merit a distinct
structural treatment. As the Case of the construct state behaves like
structural Case, Iassume it is checked in an agreement projection
occurring between DP and NP whenever a construct state is involved.
The first section of this chapter presents the free state and the
construct state and discusses some structural distinctions between the
two. In the second section I analyze the internal structure of free states,
arguing on the basis of binding phenomena that they must have a
hierarchical structure. This leads me to conclude that overt noun raising
must apply. The third section is devoted to the analysis of construct
states. First I argue that noun raising must be generaIized. I then throw
light upon the Case of the construct state and analyze the peculiar
properties the construction shows. Secti on four examines noun phrases
involving a clitic doubling configuration, arguing that they involve both
the Case assignment mechanism of the construct state and that utilized
in the free state.


In Hebrew, arguments of the noun (as weil as adjuncts) can never

appear prenominally. A genitival relation between a head noun and a
noun phrase can be indicated in two distinct ways, as illustrated below.
While in (3a) genitive Case is realized via sei ('of'), in (3b) sei does
not appear and (abstract) genitive Case seems to be assigned by the
head noun itself: I

(3) a. ha-bayit sei ha-fis

the-house 0/ the-man
'the man's house'

b. beyt ha-fis
house .the-man
'The man's house'

The second construction (3b) is known in the literature of Semitic

languages as the construct state. When the head noun is in the construct
state, it loses stress (the main stress always falls on the genitive
member of the construct state) and it is therefore subject to
phonological rules which operate in non-stressed environments. Hence
the alternation between bayit in (3a) (henceforth the free state) and beyt
in the construct state (3b).2
The genitive complement of a head in the free state must surface in
a sei ('of') phrase (4a), whereas its construct state equivalent cannot do
so (4b):3

(4) a. ha-bayit *(sel) ha-fis

the-house (oj) the-man
'the man's house'

b. beyt (*sel) ha-fis

house (oj) the-man
'the man's house'

Deverbal nouns also appear in the two constructions: the free state
(5a) and the construct state (5b):

(5) a. ha-hofa'a §el ha-saxkan

the-appearance 0/ the-actor
'the actor's appearance'

b. hofa'at ha-saxkan
appearance the-actor
'the actor's appearance'

There have been different attempts to define a systematic semantic

difference, between the two constructions. Thus, regarding concrete
(non-deverbal) nouns, Rosen (1957), adapted by Doron (1989), claims
that the complement is connected to the head of the construct state with
some kind of inalienable possession, while its free state counterpart is
not in such a relation with the head noun. Berman (1978), on the other
hand, reports that this observation does not seem to accord with the
intuitions of the native informants she interrogated. As my intuitions do
not reflect any such systematic difference either, the two constructions
will be treated here as essentially synonymous. 4
There are, however, some important syntactic distinctions between
free states and construct states, which are worthy of discussion. The
distinctions concem the article, the position of modifying adjectives,
and the obligatoriness/optionality of a genitive complement. 5

2.1.1. The artic/e constraint

Hebrew exhibits only adefinite article, which is aprefix and does not
inflect (ha-). Its indefinite counterpart is arguably phonetically null.
Note that the head noun of say (3b) is interpreted as definite, just like
the head noun of (3a), as is clear from the glosses, although the definite
article does not accompany its head. In fact, the article can never be
attached to the head of a construct state; it results in ungrammaticality
The [±definite] value of the head of the construct state is determined
by that of its complement. Adefinite complement renders the head
definite and an indefinite complement renders it indefinite, as shown,
for instance, by the behavior of the accusative marker 'ei. This marker
appears exclusively with definite objects, as illustrated in (7a-b). It can
therefore serve as a reliable test distinguishing between definites and
indefinites. Thus, when the head of the construct state receives
accusative Case, it is obligatorily accompanied by 'et (hence definite),
if its complement is definite (7c). If the complement is indefinite, 'et

cannot appear (that is, the head noun is indefinite) (7d):

(6) (*ha-)sifrey ha-me§orerim

(the-)books the-poets
'the poets' books'

(7) a. hu kone 'et ha-sfarim bezol.

he buys ACC the-books cheaply
'He buys the books cheaply'

b. hu kone (*'et) sfarim bezol.

he buys (ACC) books cheaply
'He buys books cheaply'

c. hu kone 'et sifrey ha-me§orerim ha-ce'irim.

he buys ACC books the-poets the-young
'He buys the young poets' books'

d. hu kone (*'et) sifrey me§orerim ce'irim.

he buys (ACC) books poets young
'He buys young poets' books'

Moreover, if the complement of a construct state becomes the head

of another construct state, its article can no more appear. In astring of
two (or more) construct states, only the right most noun can carry the
article (8). The definiteness of the others depends on that of this last

(8) gag (*ha-)beyt ha-'i§

roof (the-)house the-man
'the roof of the house of the man'

2.1.2. The position of modifying adjectives

Adjectives in Hebrew follow the noun they modify and agree with it in
number, gender and definiteness, as ilIustrated in (9a-b):

(9) a. ha-bayit ha-gadol

the-house(MSSG) the-big(MSSG)
'the big house'

b. biktot gdolot
hut(FMPL) big(FMPL)
'big huts'

In the free state, the adjective must always immediately follow the noun
it modifies. A pp cannot intervene between a noun and its modifying
adjective (compare (IOa) with (lOb». Likewise, the sei ('of') phrase in
the free state cannot intervene between the two (lOc), rather it must
follow the adjective (IOd):

(l0) a. • ha-bayit 'im ha-'aruba ha-yafe

the-house with the-chimney the-beautiful
'the beautiful house with the chimney'

b. ha-bayit ha-yafe 'im ha-'aruba

the-house the-beautiful with the-chimney
'the beautiful house with the chimney'

c. • ha-bayit §el ha-'i§a ha-gadol

the-house 0/ the-woman the-big
'the woman's big house'

d. ha-bayit ha-gadol §el ha-'i§a

the-house the-big 0/ the-woman
'the woman's big house'

In the construct state, in contrast, an adjective modifying the head noun

must appear at the end of the construct state complex (11b) and not
immediately following the head (lla):

(11) a. • beyt (ha-)gadol ha-'i§a

house (the-)big the-woman
'the woman's big house'

b. beyt ha-'ga ha-gadol

house the-woman the-big
'the woman's big house'

An adjective appearing at the end of the complex can refer to either the
head or the complement, providing that these two agree in number and
gender; recall that they obligatorily share the same [± definite] value.
Thus, (12) is ambiguous between the two interpretations given in (i)
and (ii):

(12) beyt ha-'g ha-gadol

house(MSSG) the-man(MSSG) the-big(MSSG)
i. 'the big man's house'
ii. 'the man's big house'

Notice that the adjective in (12) is definite whether it modi fies the
complement or the head. This by itself supplies another piece of
evidence that the head of the construct state must agree in definiteness
with its complement.

2.1.3. The genitive noun phrase

Normally, nouns can appear without a possessor. Indeed, in the free

state (13), the sei ('of') phrase is optional:

(13) ha-bayit (sei ha-'is)

the-house (0/ the-man)

Noticeably, in construct states the genitive noun phrase is obligatory

(14a). It must be phonetically realized, either as a full noun phrase
(14b) or as a pronominal clitic (14c) (discussion of clitics is deferred
until section 2.4):

(14) a. • beyt

b. beyt ha-'is
house the man
'the man's house'

c. beyt-o
'his house'

For completeness, note that in the free state a pronominal clitic surfaces
on sei ('of'):

(15) ha-bayit §el-o

the-house of-him
'his house'

To summarize, this section has presented two distinct genitival

constructions: the free state and the construct state. The latter shows
some peculiar properties which will be dealt with in section 2.3. I first
turn to the analysis of the internal structure of the free state.


In this section I present evidence that Hebrew noun phrases have a

hierarchical internal structure. I then show that, given their hierarchical
structure on the one hand, and the order of constituents they exhibit on
the other hand, it must be concluded that in Hebrew overt noun raising
must apply. If so, there must exist an appropriate lariding site higher
than NP to host the raised noun. I argue that this host is the head
position D, thus supporting the claim that noun phrases are the maximal
projections of D, as schematized in (2).
Consider deverbal nouns, specifically those expressing a process, an
event. Basically, they share the 6-grid of their associated verbs. 6 Thus,
for instance, in (16a) the deverbal noun harisa ('destruction') assigns
the same 6-roles as those assigned by the corresponding verb haras
('destroy') in (16b). In both constructions ha-'ir ('the city') is assigned
a Patient 6-role, while ha-cava ('the army') is assigned an Agent 6-role.
The Agent argument receives nominative Case in the verbal
construction and genitive Case in the nominal construction, whereas the
Patient argument receives accusative Case in both constructions (the
availability of accusative Case in Hebrew noun phrases is discussed in
the subsequent chapter):

(16) a. ha-harisa §el ha-cava 'et ha-'ir

the-destruction 0/ the-army ACC the-city
'the army's destruction of the city'

b. ha-cava haras 'et ha-'ir.

the-army destroyed ACC the-city
'The army destroyed the city'

Verbal O-grids are projected hierarchically onto syntactic structures.

Thus, for instance, the Agent is structurally more prominent than the
Theme or the Patient. In light of that, the question arises whether
nominal O-grids are projected as weIl onto hierarchical syntactic
structures. Obviously, a positive answer constitutes the null hypothesis
here, as it enables the theory to generalize principles across categories.
In order to shed light on the syntactic structure of noun phrases, I will
utilize syntactic processes that refer to structural information. Prior to
that, however, a note on word order is in place.
In (16a) the Agent immediately follows the noun. In fact, this is the
only possible word order. The Agent (in other terms, the subject) can
never be preceded by the Patient (the object) (17a). This is reminiscent
of the word order in Hebrew clauses attesting a postverbal subject; the
subject must immediately follow the verb (17b), and cannot be preceded
by the object (17c):

(17) a. * ha-harisa 'et ha-'ir §el ha-cava

the-destruction ACC the-city 0/ the-army
'the army's destruction of the city'

b. 'etmol haras ha-cava 'et ha-'ir.

yesterday destroyed the-army ACC the-city
'Yesterday the army destroyed the city'

c. • 'etmol haras 'et ha-'ir ha-cava.

yesterday destroyed ACC the-city the-army
'Yesterday the army destroyed the city'

This by itself already seems to reveal at least some structural similarity

between noun phrases and their verbal counterparts.

2.2.1. Binding phenomena

The data discussed so far do not offer any crucial evidence regarding
the internal syntactic structure of noun phrases. In principle, given the
X'-schema, three major possible structures can surface as (16a). Either
we have a flat tripartite structure (18a), or the Patient is more
prominent than the Agent (18b), or the Agent is more prominent than
the Patient (18c), as the null hypothesis would predict. 7 The third
structure, however, can surface as (16a) only if the noun undergoes
obligatory leftward movement prior to speIl-out, Agent lowering being
excluded as movement cannot take place to non c-commanding

(18) a. NP

Agent Patient

b. NP

N' Patient

A Agent

c. NP

Agent N'

A Patient
move a I
As far as the c-command relationship is concerned, the structure in
(ISa) predicts that both arguments ought to c-command each other. The
structures in (ISb-c), on the contrary, dictate an asymmetrical c-
command relationship between the two arguments (assuming the
original definition of Reinhart 1976): either only the Patient c-
commands the Agent (l8b), or vice versa, only the Agent c-commands
the Patient (l8c). As the theory of binding crucially utilizes the c-

command relationship, it offers a testing-field for the issue at stake, as

already pointed out by Giorgi and Longobardi (1991). Relying on their
work, I argue for the structure in (18c).
Principle A of the binding theory requires that anaphors be bound in
the domain of the closest subject. Before examining their distribution
within the noun phrase, it should be pointed out that neither sei ('of')
nor 'et (ACC) block c-command of the argument they introduce toward
another argument (as already pointed out by Borer 1984). This is
illustrated below. In (l9a) the anaphor is bound (hence c-commanded)
by the argument introduced by sei, and in (l9b) by the argument
introduced by 'et:

(19) a. ha-harisa ~el ha-cavaj 'et 'acmoj

the-destruction 0/ the-army ACC itself
'the army's destruction of itself'

b. hu her'a 'et ha-tinokj le-'acmoj ba-mar'a.

he showed ACC the-baby to-himself in+the-mirror
'He showed the baby to hirnself in the mirror'

As is already clear from (19a), the Patient can be an anaphor. If so,

it must be bound within the noun phrase. The only antecedent available
is the Agent, which must therefore c-command the Patient. Hence, the
options in (18) can al ready be narrowed down: (l8b) cannot be the
right structure, because the Agent would not c-command the Patient.
We are left with two options.
Principle C of the binding theory requires that R-expressions be free.
The flat structure (I8a) then erroneously predicts a Principle C violation
in cases like (l9a). However, it has been argued that when an anaphor
and its antecedent mutually c-command, no disjointness effect arises
(Borer 1984, Giorgi and Longobardi 1991). Nonetheless (ISa) must be
dismissed, because contrary to (t8c), it predicts that the Patient ought
to c-command the Agent. This prediction is wrong, as shown in (20):

(20) * ha-harisa ~el 'acmo 'et ha-cava

the-destruction 0/ itself ACC the-army

Thus, the Agent c-commands the Patient, but not vice versa. Put
differently, the Agent (the subject) must be structurally more prominent
than the Patient (the object). (l8c) then is the only structure compatible
with the data. 8
The same conclusion can be drawn on the basis of other syntactic

processes referring to c-command. One such process is binding of a

pronoun by a quantified noun phrase. The bound pronoun must be c-
commanded by the quantified noun phrase, or else it results in
ungrammaticality known as a weak crossover violation (see Reinhart
1976, 1983, Koopman and Sportiche 1982). As expected, then, the
Theme (the object) can function as the bound pronoun (21a), whereas
the Agent (the subject) cannot do so (21b), as it asymmetrically c-
commands the object:

(21) a. ha-te' ur §el kol ha-na§imj 'et ba'aley-henj

the-description 0/ all the-women ACC husbands-their
'all the women's description of their husbands'

b. *ha-te'ur §el ba'aley-henj 'et kol ha-na§imj

the-description 0/ husbands-their ACC all the-women

Summing up, in this section it was shown on the basis of binding

phenomena that the Agent must occupy a higher structural position than
the Theme or the Patient within the noun phrase. This singles out (18c)
as the only possible structure. This conclusion provides further support
for the Thematic Correspondence Hypothesis (Giorgi and Longobardi
1991), which states that verbs and their corresponding nouns identify
the same O-role as the role assigned to the subject position, available
outside V' in VPs (Fukui and Speas 1986, Kuroda 1988, Sportiche
1988) and outside N' in NPs; the external O-role in Williams' (1981)
terms. The other O-roles are assigned internally, within V' and N',
respectively (hence internal O-roles). This hypothesis suggests a
principled account of salient similarities between noun phrases and
clauses, thus dispensing with construction specific stipulations.
Moreover, it fits in with a restrictive view of grammatical processes that
allows reference to structural information only, and not to types of 0-
roles (see Belletti and Rizzi 1988).

2.2.2. The structure

For the sake of concreteness, consider again (16a), repeated here as


(22) ha-harisa seI ha-cava 'et ha-'ir

the-destruction 0/ the-army ACC the-city
'the army's destruction of the city'

The external argument, ha-cava ('the army'), is generated in the

specifier position of NP, and the internal argument, ha-'ir ('the city'),
is generated within N'. The surface word order must therefore be
derived via leftward movement of the noun. If so, there must be a head
position higher than NP to host the raised noun. 9 Obviously, if noun
phrases were projections of N (1), such a position would not be
available. On the other hand, if noun phrases are DPs (2), then D
emerges as a possible landing site. Given the affixal nature of the
Hebrew (definite) article, D constitutes a natural landing site, just like
functional heads in the verbal system:

(23) DP



DPext. N'

N DPint.
move a I
Recall now that Hebrew shows a definiteness agreement between the
noun and its modifying adjective (see (9) above) as weH as between the
head of the construct state and its complement (see (7c-d), (8) above).
This agreement phenomenon may suggest that definiteness in Hebrew
is a feature of nouns, which coincides with the prefixal nature of the
article (see Borer 1989). Under a checking theory, lexical entries are
inserted with their morphological features, wh ich must be checked with
an inflectional head by LF. If definiteness is a feature in Hebrew, it is
reasonable that the noun is inserted with its definiteness specification,
be it the definite article or its phonetically null indefinite counterpart.
This lexical feature must be checked with D in the course of the

derivation, just like the tense specification of a verb is checked with T

in the course of the derivation. Noun raising to D applies overtly in
Hebrew. This amounts to saying that the corresponding features D bears
and checks with the raised noun (its N-features) are strong in Hebrew
and must therefore be eliminated prior to speil-out (because they are not
legitimate objects at PF, as discussed in chapter 1).10,11
An example like (22) then involves the following representation:

(24) [~p [0 ha-harisajl [NP §el ha-cava [N' t j let ha-lir]]]

the-destruction 0/ the-army ACC the-city

The movement the noun undergoes is local, as it does not skip any head
position; thus it does not pose any problem for current approaches to
V(erb)S(ubject)O(bject) surface order (for instance (17b» has often
been analyzed along similar lines (Fassi Fehri 1989, Sproat 1985, Borer
1995, among others), as exemplified in (25):

(25) ... hp h harasj 1 [vp ha-cava [v' t j let ha-lir]]]

destroyed the-army ACC the-city

Under the present approach, the word order of both (24) and (25) can
be accounted for in a sim~le unified way, that is, via overt noun and
verb raising, respectively. 2
Consider now modifying adjectives. Recall that they normally appear
immediately to the right of the noun, as illustrated again in (26):

(26) ha-harisa ha-mehira §el ha-cava let ha-lir

the-destruction the-rapid 0/ the-army ACC the-city
'the army's rapid destruction of the city'

Consequently, they ought to be base-generated in a position left-

adjoined to NP, so that they will follow the noun after it raises to D, as
illustrated in (27):

(27) [~p [0 ha-harisaj 1[NP

ha-mehira [NP §el ha-cava [N' t j ...

the-destruction the-rapid 0/ the-army

This suggestion does not entail any special assumptions. On the
contrary, it is straightforward, if Chomsky (1986a) is correct in arguing
that maximal projections can adjoin exclusively to maximal projections.
Note that the fact that English manifests prenominal adjectives whereas

Hebrew manifests postnominal adjectives simply falls out, if in English

overt noun raising cannot apply while in Hebrew it must apply. If the
noun remained in its base position in Hebrew, we would have to
assurne that these two languages differed essentially with regard to the
positioning of modifying adjectives. 13
Finally, it should be pointed out that although noun raising was
motivated here on the basis of evidence offered by deverbal (event)
nouns, it must be generalized to concrete nouns as weIl. First, it is not
plausible to suppose that the force of the features D checks with the
noun is dependent on the deverbal/concrete nature of the noun. Second,
as modifying adjectives follow both deverbal and concrete nouns, a true
generalization concerning their positioning can be reached only if nouns
must overtly raise to D, whether they are deverbal or concrete, definite
or indefinite:

(28) a. [op [0 ha-bayit j] [NP ha-yafe [NP t j ]]]

the-house the-beautiful
'the beautiful house'

b. [op [0 bayit j ] [NP yafe [NP t j ]]]

house beautiful
'a beautiful house'

In conclusion, arguments based on binding phenomena justify a

configurational approach to Hebrew noun phrases, which captures in a
natural way important generalizations across categories. Word order can
then be accounted for solely in terms of overt noun raising, which,
obviously, implies that noun phrases cannot simply be projections of N.
In the next section, I turn to examine noun phrases involving a
construct state.


Turning our attention to nouns occurring in construct states, we would

obviously expect them, too, to reflect the insight expressed in the
Thematic Correspondence Hypothesis. This is in fact predicted by the
Uniformity of Theta Assignment Hypothesis (UTAH; Baker 1988). Put
in somewhat minimalist terms, UTAH requires that thematic paraphrases
have their arguments inserted in the same positions. Hence, the free
state (22) and its construct state paraphrase (29a) (in which the internal
argument receives accusative Case as in (22), and the external argument

is part of the construct state) should share the same basic structure. The
external argument should be generated in the specifier position of NP
and the internat argument inside N'. Consequently, surface word order
should be derived via noun raising (29b), as in the case of free states:

(29) a. harisat ha-cava 'et ha-'ir

destruction the-army ACC the-city
'the army's destruction of the city'

b. [oP [0 harisatj ] [NP ha-cava [N' t j 'et ha-'ir]]]

destruction the-army ACC the-city

If that is indeed correct, we predict that construct states ought to show

the same subject-object asymmetry manifested in free states. The
prediction is borne out: (30a) is grammatical because an anaphor is
Iicensed in the object position, while (30b) is ruled out because it is not
Iicensed in the subject position. Similarly, (31a) is grammatical because
a quantified noun phrase in the subject position can bind a pronoun in
the object position, but not vice versa (31 b): 14

(30) a. harisat ha-cava 'et 'acmo

destruction the-army ACC itself
'the army's destruction of itself'

b. • harisat 'acmo 'et ha-cava

destruction itself ACC the-army

(31) a. te'ur kol ha-na§imj 'et ba'aley-henj

description all the-women ACC husbands-their
'all the women's description of their husbands'

b. • te'ur ba'aley-henj 'et kol ha-na§imj

description husbands-their ACC all the-women

I now turn to a more complete analysis of the syntax of construct

states. Three major issues deserve attention: first, the type of Case
associated with the construct state (section 2.3.2); second, the
definiteness agreement between the components of the construct state
(section 2.3.3); and finally, the irregular location of modifying
adjectives (section 2.3.5). A detailed discussion of the distribution of

construct states and sei ('of') phrases is given in section 2.3.4. I start
by summarizing the groundwork that led me to the present analysis.

2.3.1. Setting the stage

In certain languages the noun agrees with its possessor in the same way
that the verb agrees with its subject. Hungarian is a language of this
type: the agreement between the noun and its possessor is by and large
identical to the agreement between the verb and its subject; possessors
are morphologically unmarked, like (nominative) subjects (32a-b) (see
Szabolcsi 1983-84, 1987, 1989). Yupik, a Central Alaskan Eskimo
language manifests the same phenomenon; the subject as weil as the
possessor show ergative Case (33a-b) (Abney 1987). In Turkish, too,
the noun agrees with its possessor; the agreement morpheme, however,
is distinct from the verbal agreement with the subject, and the possessor
shows genitive Case (34) (Komfilt 1984):

(32) a. Ate ir-od.

yOU(NOM) write-2SG
'You write'

b. a te titk-od
the YOU(NOM) secret-2SG
'your secret'

(33) a. angute-t kiputa-a-t.

'The men bought it'

b. angute-t kuiga-t
men-ERG river-SA
'the men's river'

(34) sen-in ei-in

you-GEN hand-2sG
'your hand'

In order to explain this parallelism between clauses and noun

phrases, Abney (1987) has proposed that the noun phrase is the
maximal projection of a functional I-like element, D. According to him,
the class of elements generated in D consists of determiners and
Agr(eement) features. Agr in D assigns Case to the possessor in SpecDP
(or in SpecNP) in the same way that Agr in I assigns Case to the
subject in SpeclP (or in SpecVP):

(35) a. IP

(subject) A I'


A!r I


b. DP

(possessor) A DI




This proposal has inspired aseries of studies of Semitic construct

states, starting with Ritter (1987, 1988). These studies have advanced
slightly different variants of the same basic idea that in construct states
the noun raises to D and incorporates with Agr. Once supported
morphologically, Agr assigns genitive Case under government to the
argument in the specifier position of NP (see Ritter 1987, 1988, Fassi
Fehri 1989, Siloni 1990b, 1991a, among others):IS

(36) DP


The position of modifying adjectives clearly poses a problem for an

analysis along these lines. If modifying adjectives are base-generated in
a position left-adjoined to NP (see section 2.2.2), we would predict that
they ought to appear in between the head of the construct state and its
genitive DP, while, in fact, they can never intervene between these two
(see (11) above). In Siloni (1990b, 1991a), I proposed that this could
stern from a requirement that the Case assigner and the assignee be
adjacent. An adjective intervening between the two blocks Case
assignment. To rescue the structure, then, the assignee itself must adjoin
to NP higher than the adjective, yielding the order "N DP AP ... ":

(37) DP




Indeed, strict adjacency has often been mentioned as a requirement

imposed on Case assignment under government (see Chomsky 1980,
Stowell 1981, Borer 1984, and for more recent discussions, Belletti
1990, McCloskey 1991, Rizzi 1991a, Haegeman 1992, Friedemann
1993-94). Notice that this proposal assumes that multiple adjunctions
and Case assignment to an adjoined position are possible mechanisms.
Altematively, Ritter (1991) has suggested that noun phrases contain
an additional functional category between DP and NP, NumP, whose
head bears the number specification of the noun. Following work by
Bat-EI (1986), Ritter argues that number marking is a syntactic process.
NumP provides an additional specifier to which the assignee (the DP
member of the construct state) can raise in order to be adjacent to its
Case assigner D. The trigger for the movement is identical in both (37)
and (38); the landing site, however, is different:

(38) OP




A N'


In short, assuming that in construct states Case assignment takes

place under government and that this process is constrained by an
adjacency requirement, one can technically account for the data.
However, in light of the growing tendency to split up functional heads
into their more basic components (starting with Pollock's 1989 split of
IP into AgrP and TP), the question arose as to whether the conflation
of determiners and Agr-features in one functional head position (D) was
indeed justified.
On the basis of examples like (32b), Szabolcsi (1987, 1989) has
argued that Hungarian noun phrases contain two functional elements:
an Agr-type element that is responsible for the Case of the possessor in
the same way that I is responsible for the Case of the subject, and
dominating it a complementizer-like element that accommodates the
article (D). This proposal syntactically separates the two distinct
functional elements, Agr-features and determiners, which are both
generated in 0 according to Abney (1987) and much subsequent
research. Abstracting away from Szabolcsi's precise execution, we

obtain the structure schematized below:

(39) DP

I 0'

A AgrP





It is nonetheless unclear to which extent this approach should be

generalized; after all , in many languages (say, French, Italian) noun
phrases do not involve any direct Case assignment to DPs, rather they
always require the intervention of a Case marker like of. It would be
somewhat dubious to stipulate in these cases, too, a functional element
like Szabolcsi's (1987, 1989) AgrP. As will be shown below, Hebrew
offers us some insight into the matter.

2.3.2. Structural genitive Case

An interesting piece of data given in (40) reveals an important

distinction between the behavior of the genitive DP of the construct
state and that of sei ('of'). (40a-b) contain the deverbal noun meci'a
(,finding'), which takes a small clause as its complement. While the
genitive complement of sei cannot be the subject of the small clause
(40a), the genitive DP of the construct state can do so (40b), just like
the accusative complement of the corresponding verb (40c):

(40) a. • ha-meci'a sei ha-ne'esam 'asem

the-finding 0/ the-accused guilty

b. meci'at ha-ne'esam 'asem

finding the-accused guilty

c. hu maca 'et ha-ne'esam 'asem.

he found ACC the-accused guilty

In (40c), although the DP ha-ne'e§am ('the accused') is not 6-marked

by the verb, but rather within the small clause, it can still receive
accusative Case (an exceptional Case marking configuration (ECM».
Analogously, the deverbal noun in (40a-b) does not 6-mark the DP ha-
ne'e§am ('the accused'). Notably, while it cannot receive Case via sei
('of'), it can receive Case via the construct state. The Case of the
construct state is available in ECM configurations, whereas the case
assigned via sei is not.
This is of course reminiscent of the distinction Chomsky (l986a)
draws between structural Case and inherent Case. Recall that inherent
Case is assigned by IX to DP only if IX 6-marks DP, while structural
Case imposes no such thematic requirement. According to Chomsky
(1986a), accusative and nominative are structural Cases; they are not
thematically dependent. Oblique Case assigned by prepositions, dative
and presumably genitive are inherent Cases, as their assigners also 6-
mark the assignee. However, if the differentiation between inherent
Case and structural Case is on the right track, it follows that the
genitive Case of the construct state is a structural Case, as it is not
contingent upon a thernatic relationship, while the genitive Case
assigned via sei is an inherent Case determined by the noun, as it can
only be granted to noun phrases that the noun 6-marks.
Importantly, the construct state seems to involve structural Case in
other Semitic languages, too. Thus, for instance, in Standard Arabic,
just like in Hebrew, the Case of the construct state is not thernatically
dependent; it is available in ECM configurations (see Fassi Fehri 1993):

(41) öann-u r-rajul-i zakiyy-an xata'-un.

belief-NOM the-man-GEN clever-ACC error-NOM
'Believing that the man is clever is an error'
(Fassi Fehri 1993)

It has commonly been argued that structural Case can be assigned either

und er government (accusative Case) or in a Spec-head relation

(nominative Case). To replace this disjunctive formulation of the
configuration allowing structural Case assignment, Chomsky (1991,
1993) suggests that structural Case is always the manifestation of a
Spec-Agr relation. In other words, structural Case is always checked in
SpecAgrP, prior to speil-out or following it, depending on the force of
the relevant features of the Agr-head.
Adopting this line of reasoning, it follows that noun phrases
involving a construct state must contain an AgrP, whose specifier serves
as the locus of Case checking for the genitive argument. More
generally, it follows that only in languages where the Case available
within the noun phrase is not thematically dependent, such an AgrP
must be present. A parallel AgrP in languages that do not manifest
structural Case within the noun phrase would perform no checking and
would therefore be inert or simply missing. Thus, as genitive Case in
English or French noun phrases is not available in ECM configurations,
it is an inherent Case, wh ich does not entail a Spec-Agr relation: 16

(42) a. * The belief of John to be the winner

b. *John's consideration intelligent
c. * La consideration de Jean intelligent
the consideration of Jean intelligent

As expected under the present analysis, in Hungarian (43) just like in

Hebrew (40b), the relevant Case (nominative, in Hungarian) is available
in ECM configurations, as it is a structural Case: 17

(43) a. Janos okos-nak tartas-a

Janos(NOM) smart(DAT) consideration-3sG
'considering Janos smart'

b. Janos j6 tanar-nak tartas-a

Janos(NoM) good teacher(DAT) consideration-3SG
'considering Janos a good teacher'

Let us now examine the derivation of noun phrases involving a

construct state. As these noun phrases involve structural genitive Case,
they must contain an AgrP (as in (39) adapted below):

(44) DP


D AgrgenP



I label the agreement projection AgrgenP, but this notation is only

mnemonic: it is an AgrP where structural Case is checked in the noun
phrase, be the Case genitive Case as in Hebrew, nominative Case as in
Hungarian, or ergative Case as in Yupik. Just like the Case of the
subject of the clause is determined by T and checked by Agr and the
Case of the direct object is determined by V and checked by Agr, the
Case of their nominal equivalent is determined by N and checked by
Agr. Recall now that a head in the construct state loses stress and
therefore undergoes phonetic changes (section 2.1). This may be
conceived as some phonetic reflex of the presence of Agrgen features on
the noun.
For concreteness, consider again (29a), repeated here as (45a). Its LF
representation and the relevant steps in its derivation are given in (45b)
(as noted earlier, discussion of the accusative Case that the direct object
(DP 0) can bear in Hebrew noun phrases is deferred until the next
chapter, where it is shown that it is not a structural Case and does not
involve an AgrP). The noun is inserted with certain cj>-features it must
check with Agrgen . Correspondingly, then, Agrgen has N-features it has
to check with the noun. As empirical evidence shows (see discussion
above), noun raising in Hebrew is overt. This is straightforward, if the
N-features of Agrgen are strong and hence must be eliminated prior to
spell-out. 18 As mentioned earlier, the genitive DP (the subject, DPs in

(45b» must precede an adjective modifying the head noun (11), which
is base generated in a position left-adjoined to NP (see (27), and section
2.3.5 for more discussion). It follows that the genitive DP undergoes
overt raising to SpecAgrgenP, where it checks genitive Case. Hence, the
DP-features of Agrgen are also strong. The resulting word order is "DP
N ...", which is never attested in Hebrew noun phrases. However, recall
that D also bears strong N-features in Hebrew (see section 2.2.2), which
requires further overt N-raising to D and results in the observed word
order (-+ indicates movement):

(45) a. harisat ha-cava 'et ha-'ir

destruction the-army ACC the-city
'the army's destruction of the city'

b. DP



A N'

N -+ Agrgen I DPs -+ SpecAgrgenP I Agrgen -+ D

The same ought to apply to concrete nouns. If the Case of the

construct state is a structural Case, it must involve an AgrP, whether
the head noun is concrete or deverbal. It is hardly plausible to argue
that construct states involving concrete nouns are derived differently,
since both construct states (namely of deverbal and concrete nouns)
show the same syntactic properties (as described in section 2.1).

When sei ('of') is inserted as in (46), DPs receives its inherent

genitive Case within the sei phrase and has no reason to raise further,
because movement is a Last Resort operation (it thus follows an
adjective modifying the head noun (26-27». Given that, either an inert
AgrgenP is generated, which does not check any features, or the
projection is missing all together. Apriori, the presence of an inert
AgrgenP does not seem to have any empirical consequences and seems
to me incongruous with a minimalist guideline; I therefore take it to be
absent (in the same way that it is missing in English or French noun

(46) ha-harisa seI ha-cava 'et ha-'ir

the-destruction 0/ the-army ACC the-city
'the army's destruction of the city'

The question now arises as to why a noun that checks cj>-features in

Agrgen cannot bear the article, and why there is a definiteness
agreement between the components of the construct state.

2.3.3. The definiteness agreement

Recall that the head of the construct state cannot realize its article, as
illustrated in (6) and repeated here:

(47) (*ha-)sifrey ha-mesorerim

(the-)books the-poets
'the poets' books'

Thus, in astring of two (or more) construct states, the article can
surface only on the right most noun (see (8) above). This strongly
suggests that the article of a head in the construct state can never occur.
This impossibility is even more salient in Standard Arabic, wh ich has
both adefinite and an indefinite article; the former is aprefix (48a)
while the latter is a suffix (48b). In the construct state, none of them
can appear with the head noun (48c-d):19

(48) a. 'al-bayt-u
'the house'

b. bayt-u-n
'a house'

c. bayt-u r-rajul-i
house-NOM the-man-GEN
'the man's house'

d. bayt-u rajul-i-n
house-NOM man-GEN-a
'a man's house'

Moreover, recall that there is an obligatory agreement in (in)definiteness

between the head of the construct state and its genitive DP. Evidence
that they must agree was presented above on the basis of the behavior
of the accusative marker 'ei (section 2.1.1) as weil as the behavior of
modifying adjectives (section 2.1.2). A genitival relation between
elements that do not share the same [±definite] value cannot be
expressed through the construct state (49a), rather sei ('of') must appear

(49) a. beyt ha-'is

house the-man
i. *'a house of the man'
ii. 'the man's house'

b. bayit sei ha-'is

house 0/ the-man
'a house of the man'

This means that a noun can either be inserted with Agrgen features,
which gives rise to a construct state, or with the article, out not with
both. However, even a noun inserted with Agrgen features has a
[±definite] value. This value must accord with the definiteness value of
the genitive DP. Why should that be so?
My proposal involves the following assumptions. In Hebrew (or
Arabic), a noun can realize its definiteness feature as part of its Agr~en
morpheme. It either does it this way or in the form of an article \as
customary). When the definiteness feature is part of Agr gen , it must
match the definiteness value of the genitive DP.

The (in)definiteness effect is derived as folIows. If a noun is inserted

with Agrgen features, its [±definite] feature is part of Agrllen , and the
article cannot appear. Agr en and the article are incompatible because
both include the [±definitef feature, which cannot have two realizations
on the same head (N). A noun endowed with Agrgen features raises to
Agrgen to check them. The features must match the features of the
genitive DP, or else the derivation would not converge; the definiteness
agreement between the members of the construct state folIows. Once
this checking has taken place, the complex head in Agrgen raises to D. 20
In contrast, if the noun is inserted with the article, it cannot bear Agrgen
features (again, Agrgen and the article are incompatible, as both contain
the [±definite] feature). Hence, AgrgenP is not projected, and the noun
raises immediately to D to check the relevant features.
The assumption that Agrgen includes the feature [±definite] is not
baseless, as nominal agreement features in Hebrew (and Arabic) contain
definiteness in addition to eil-features. Modifying adjective, for example,
must agree with the noun they modify in number, gender, and
definiteness (section 2.1.2). The novel component of the proposal is the
claim that within Agrgen , the noun realizes its own definiteness feature.
This means that while a singular noun can be in an agreement relation
with a plural genitive DP (that is, albeit their agreement relation, their
intrinsic eil-features do not have to match), a noun can be in an
agreement relation with a genitive DP only if it bears an identical
definiteness value. It follows that the head of the construct state cannot
bear the article and must agree with its genitive DP in definiteness,
because it is in an agreement relation with it. 21

2.3.4. Construct states and äel phrases

In this section, I extend the paradigm of data and briefly discuss the
distribution of structural and inherent genitive arguments. Up to now,
I have discussed two major cases: one in whieh the external argument
has struetural genitive Case through the eonstruet state (45) and the
other, where it has inherent genitive via sei ('of') (46). In both eases
the internal argument must be associated with aceusative Case, and
eannot have Case via another sei (trivially, it eannot have Case via
another eonstruet state, as only one AgrgenP and only one head noun are
available within one noun phrase):22

(50) a. • harisat ha-cava §el ha-'ir

destruction Ihe-army 0/ the-city
'the army's destruction of the city'

b. • ha-harisa §el ha-cava §el ha-'ir

the-destruction 0/ the-army 0/ the-city
'the army's destruction of the city'

When the external argument is not realized, the internal argument can
have genitive Case either through the construct state or via sel: 23

(51) a. harisat ha-migdal

destruction the-tower
'the destruction of the tower'

b. ha-harisa §el ha-mi gd al

the-destruction 0/ the-tower
'the destruction of the tower'

If the internal argument is the genitive argument, the external

argument can still surface in an 'al-yedey ('by') phrase (resulting in the
so-called nominal passive):

(52) a. harisat ha-migdal 'al-yedey ha-cava

destruction Ihe-Iower by the-army
'the destruction of the tower by the army'

b. ha-harisa §el ha-migdal 'al-yedey ha-cava

the-destruction 0/ the-tower by the-army
'the destruction of the tower by the army'

In contrast, the external argument cannot surface in a sei phrase:

(53) a. • harisat ha-migdal §el ha-cava

destruction the-tower 0/ the-army

b. • ha-harisa §el ha-migdal §el ha-cava

the-destruclion 0/ the-tower 0/ the-army

Having presented this paradigm, I start with the simple cases


involving one realized argument (51). If the noun is inserted with

Agrgen features (as in (51a), repeated in (54) below), checking must
take place in AgrltenP, or else the derivation would contain unchecked
features and theretore would crash. Hence, AgrgenP must be projected.
N raises to Agrgen' the internal argument (OPo ) to SpecAgrgenP, where
it checks genitive Case, and finally the complex head in Agrgen raises
to 0:

(54) a. harisat ha-migdal

destruction the-tower
'the destruction of the tower'

N .... Agrgen I OPo .... SpecAgrgenP I Agrgen .... 0

The internal argument in (54a) cannot receive Case via sel. .If it did, it
would have no reason to raise further (movement being a Last Resort
operation) and Agrgen could not check its OP-features, which would
result in a crashing derivation:

(55) * harisat ~el ha-migdal

destruction 0/ the-tower

According to the same reasoning, if the noun is not inserted with

Agr,en features, AgrgenP is absent (see discussion regarding (46» and
the tnternal argument must receive Case via sei:

(56) a. ha-harisa *(§el) ha-migdal

the-destruction 0/ the-tower

N .... O

As reported above, when the external argument receives genitive

Case, the internal argument can bear accusative Case but cannot surface
in a sei phrase (50). This may suggest that sei is not available in the
complement position, but only higher in the structure, in the specifier
position of NP. The fact that the internal argument can surface in a sei
phrase only when the external argument is not realized (5tb) further
supports this assumption, suggesting that in the absence of the external

argument, the internal argument can raise to the specifier position of

NP, where sei is available. On the same grounds, (S3b) is excluded: if
the internal argument surfaces in a sei phrase in the specifier position
of NP, the external cannot do so. It can however surface as an adjunct
in an 'al-yt!dey ('by') phrase, as shown in (S2b).
There is however no apparent reason why (S3a) is not acceptable.
Raising of the internal argument to SpecAgrgenP should be able to skip
SpecNP just like raising of its verbal counterpart to SpecAgroP is able
to skip SpecVP. The Agent then should be able to surface in a sei
phrase in SpecNP. This impossibility may suggest that object raising to
SpecAgrgenP must nonetheless pass through SpecNP, wh ich excludes
realization of the Agent in this position. But why should that be so? I
would like to suggest a more radical solution that is independently
motivated and accounts for the ungrammaticality of both (SO) and (53)
without the above assumptions.
It is implicit in some versions of Case theory and explicit in others
(e.g., Vergnaud 1985) that there is a biunique relation between Case
assigners and Case assignees. Let us assume with Cinque (1994) that
genitive Case assignment is restricted accordingly. More precisely, I
suggest that a head noun can assign genitive Case to one argument only
(henceforth, the Biunique Relation Requirement). There are, of course,
apparent counterexamples involving more than one instance of genitive
Case per noun phrase, to which I return in section 2.4. and in the
appendix to chapter 3.24 Now, if genitive Case assignment is restricted
by the Biunique Relation Requirement, it immediately derives the
impossibility of (SOb) and (S3b), which contain two sei phrases, thus
rendering the stipulation that sei is not available in the complement
position superfluous.
Suppose further that the restriction on genitive Case assignment does
not discriminate between inherent genitive (the genitive of seI) and
structural genitive (the genitive of the construct state). If so, then it also
blocks the same head noun from taking an inherent genitive argument
and a structural genitive argument simultaneously, thus ruling out (SOa)
as weH as (S3a), which contain both a construct state and a sei phrase.
If this reasoning is on the right track, then two distinct arguments of
one head noun cannot both bear genitive Case, whether the Case is
inherent or structural. As mentioned above, an account along these lines
enables us to dispense with both the stipulation that sei is not available
in the complement position and the stipulation that a DP ending up in
SpecAgrgenP must have moved through SpecNP.
Finally, recall that a noun in the construct state, namely, a noun
inserted with Agrgen features (say, hofa'at ('appearance') in (S7a», must

take a genitive OP (section 2.1.3). As is already clear, Agrgen must

check its OP-features, or else the derivation would crash due to
unchecked features. Note that the genitive OP must be phonetically
realized, which means that Agrgen cannot license a null element. 2S As
expected, a noun inserted without Agrgen features (57b) can appear with
no complement:

(57) a. • hofa'at

b. (ha-)hofa'a

I now turn to discuss the distribution of modifying adjectives and the

adjacency required between the members of the construct state.

2.3.5. Modifying adjectives and adjacency

As already noted above, it is straightforward that in construct states an

adjective modirying the head noun can never precede the genitive OP,
as the latter raises to SpecAgrgenP, thus preceding any element internal
to NP and, more pertinently, adjoined to it, like adjectives (58a). In
contrast, an argument receiving Case via seI ('of') does not raise out of
NP and therefore must follow a modifying adjective (58b):

(58) a. harisat ha-migdal ha-mehira

destruction the-tower the-rapid
'the rapid destruction of the tower'

b. ha-harisa ha-mehira sei ha-migdat

the-destruction the-rapid 0/ the-tower
'the rapid destruction of the tower'

In fact, judgments regarding the occurrence of an adjective modifying

the head of the construct state are slightly more subtle. When a noun
in the construct state has both its external and internat arguments
realized, the occurrence of a modifying adjective renders the structure
somewhat c1umsy and marginal (59a). Importantly, however, an
adjective intervening between the head noun and the genitive OP
renders the structure completely impossible (59b):

(59) a. ? harisat ha-cava ha-mehira let ha-migdal

destruction the-army the-rapid ACC the-tower
'the army's rapid destruction of the tower'

b. * harisat ha-mehira ha-cava let ha-migdal

destruction the-rapid the-army ACC the-tower
'the army's rapid destruction of the tower'

The status of (59a) is comparable to that of its "passive" counterpart

(60), which possibly sounds a bit better:

(60) ? harisat ha-migdal ha-mehira 'al-yedey ha-cava

destruction the-tower the-rapid by the-army
'the rapid destruction 01 the tower by the army'

Except for a certain heaviness, I do not see why (59a) and (60) should
be somewhat marginal. Still more interestingly, when the deverbal noun
has a small clause as its complement, the occurrence of an adjective
modifying the head noun renders the structure impossible. Again, the
same structure with an adjective intervening between the head noun and
its genitive DP is clearly worse, in fact, inconceivable:

(61) a. ?*meci'at ha-ne'e§am ha-mehira 'dem

jinding the-accused the-rapid guilty

b. * meci'at ha-mehira ha-ne'e§am 'dem

jinding the-rapid the-accused guilty

A parallel impossibility is found in clauses. In (62) the object is an

unstressed pronoun, which must overtly raise out of VP (arguably to
SpecAgroP; see Friedemann and Siloni 1993 for empirical evidence).
Nonetheless a VP-initial adverb cannot intervene between the pronoun
(the subject of the small clause) and its predicate:

(62) ?* hu maca toto bi-mehirut 'dem.

he lound him quickly guilty

The impossibility· of (61) could probably be explained on a par with

that of (62); for some reason (which I will not investigate here), such
an interference in the sequence of the small clause is impossible.
Note that I derive the adjacency requirement between the head noun

of the construct state and its genitive DP from their respective raising
to D and to speCAgryenP. Iassume that modifying adjectives in Hebrew
appear in a position eft-adjoined to NP. Hence, they cannot intervene
between the head of the construct state and its genitive DP, which both
occupy higher positions.
We have seen that the head of the construct state loses its main
stress, phonologically forming one unit with the genitive DP (section
2.1). Thus, the adjacency between the head of the construct and its
genitive DP resulting from syntactic processes seems to be further
required on PF grounds. While I will not explore the nature of this PF
requirement here, I would nonetheless like to discard the possibility that
construct states are units (words) created via syntactic incorporation.
The idea that construct states are words formed in the syntactic
component is due to Borer (1989). Elaborating this idea, Shlonsky
(1990) has suggested that construct states are words created via
syntactic incorporation into the head of the construct. This proposal is
hard to maintain, given the grammaticality of examples like (63), where
the genitive DP is a coordinate structure. The coordinate structure as a
whole cannot be claimed to incorporate with the head of the construct
state, nor can one member of the coordinate structure (or its head) be
extracted, as it would violate the Coordinate Structure Constraint:

(63) a. harisat [[ha-muze'on le-'omanut] ve- [ha-rexovot ha-

destruction the-museum to-art and-the-streets the-
smuxim 10]]
close to+it
'the destruction of the museum of art and the streets next
to it'

b. beyt [[ha-rabi mi-kiryat 'arba] ve- [ra'ayat-o ha-

house the-rabbi jrom-Kiryat Arba and-wiJe-his the-
'the house of the rabbi of Kiryat Arba and his honorable

Finally, Hebrew nouns can also realize their genitive argument as a

clitic. A word on such constructions is therefore in order. The closing
section examines the behavior of noun phrases involving clitics and
clitic doubling configurations in light of the analyses proposed in the


2.4.1. Data

Much research in generative grammar has focused on pronominal clitics

crosslinguistically and in various environments. Roughly speaking,
clitics have the defining property that they are phonologically dependent
on an adjacent host. They thus cannot occur in isolation nor be
coordinated. Nonetheless, they do not seem to form a homogenous class
(see Everett 1989 for a typology). This section deals with Hebrew
nominal clitics.
In Semitic languages, when a head noun in the construct state is in
a genitival relation with a pronominal element, it surfaces with a suffix
(whether the noun is deverbal (64a) or not (64b»:

(64) a. hofa'at-o
'his appearance'

b. beyt-a
'her house'

The suffix is in complementary distribution with a lexical DP in (65),

which suggests that it absorbs the structural genitive Case of the
construct state, thus leaving the lexical DP short of Case:

(65) * hofa'at-o ha-'is

appearance-his the-man

This behavior sets these suffixes apart from agreement markers, which
do not absorb Case, and classifies them with pronominal clitics. For
example, the Hungarian nominal agreement morpheme, which formally
resembles the suffixes in (64) and (65), does occur with a coreferring
noun phrase, wh ich, in fact, cannot appear in its absence:

(66) a Peter hang *(-ja)

the Peter voice (-3sG)
'Peter's voice'
(Szabolcsi 1989)

This way of viewing things gains further support by the fact that the
Hebrew suffix can cooccur with a lexical DP, only if the latter receives
Case via seI ('of') (67). The resulting configuration is the so-called
clitic doubling construction, where the clitic and the lexical DP must
corefer and share the same semantic role:

(67) hofa'at-o".,
~el ha-'g,
appearance-his 0/ the-man
'the man's appearance'

Moreover, if indeed the clitic absorbs Case, it is clear why a head noun
in the free state cannot host a clitic (68a); seI must be present to
provide the clitic with Case (68b):26

(68) a. * (ha)-hofa'a-'o

b. ha-hofa'a ~el-o
the-appearance of-him
'his appearance'

In order to analyze clitic constructions, it must be determined where

the clitic is generated.

2.4.2. Generation site

Syntactic analyses of clitics have developed in two main directions. One

view argues that clitics are inserted where the corresponding full noun
phrases are generated (see Kayne 1975, 1989b, Quicoli 1976, and
subsequent work). A second line of thought takes them to be inserted
in their surface position (Strozer 1976, Rivas 1977, Jaeggli 1982, Borer
1984, among others). If clitics are inserted where the corresponding full
noun phrase is generated, the doubled noun phrase in clitic doubling
constructions of the type in (67) ought to occupy a dislocated position
(as suggested by Hurtado 1984 with regard to clitic doubling in

Spanish). Doubling constructions, however, differ from dislocation

constructions in important respects, which argues against the dislocation
approach (as observed by Jaeggli 1986).
Dislocations are typically set off from the rest of the sentence with
an intonational break (69a), while doubled noun phrases are not (69b)
(# denotes an intonational break). Second, dislocations do not involve
a Case marker, while doubled noun phrases require it. As shown in
(69c), the addition of the Case marker sei ('of') to the dislocated
element results in ungrammaticality:

(69) a. dan # pniyat-o la-cibur 'orera ~a'aruriya.

Dan addressing-his to+the-public created scandal
'Dan, his addressing the public created a scandal'

b. pniyat-o ~el dan la-cibur 'orera ~a'aruriya.

addressing-his 0/ Dan to+the-public created scandal
'Dan's addressing the public created a scandal'

c. * ~el
dan # pniyat-o la-cibur 'orera ~a'aruriya.
of Dan addressing-his to+the-public created scandal

Moreover, a doubled noun phrase can be a quantified noun phrase

(70a), while dislocated elements cannot be quantified elements (70b)
(see Rizzi 1986b for some discussion of this restrietion):

(70) a. hakranat-o ~el kol seret nim~exa ~a'atayim.

screening-it 0/ every movie lasted two+hours
'The screening 0/ every movie las ted two hours'

b. * kol bamay, hu hikrin seret.

every director, he showed film

It is indeed on the basis of clitic doubling constructions that linguists

have advanced the claim that clitics are generated in their surface
position, and not where the corresponding full noun phrase is generated.
If this reasoning is correct, then Hebrew nominal clitics ought to be
generated (inserted) with their nominal host, and are therefore pre-
syntactic suffixes. As they exhaust the structural Case of the noun (the
Case of the construct state), I suggest that they realize its structural
Case features. The intuition behind this suggestion is that in the same
way that heads realize their tense or agreement features as inflectional

affixes, they can also materialize their Case features as affixes. 27

If this proposal is on the right track, other Case assigning heads
ought to be able to utilize this strategy, too. As expected, not only
nouns can appear with pronominal suffixes, but also verbs (71a),
prepositions (71 b), or quantifiers (71 c):

(71) a. hu roce li-r'ot-o.

he wants to-see-him
'He wants to see hirn'

b. bi~vil-o

c. kul-o

Let me then turn to analyze the occurrences of Hebrew nominal

clitics. I first analyze clitic doubling constructions, then simple clitic
constructions, and finally address the following closely related
questions: how come clitic doubling constructions can simultaneously
realize both structural and inherent genitive Case, in apparent
contradiction with the Biunique Relation Requirement imposed on
genitive Case assignment (section 2.3.4)1 And why must the clitic and
the doubled noun phrase be coindexed and share the same semantic role
(see (67))1

2.4.3. Analysis

If N realizes its structural Case features as a clitic, the latter ought to

undergo checking in Agrgen . Typically, structural Case is checked in a
Spec-head configuration. However, apriori, nothing enforces this
pairing. If the grammatical system allows materialization of structural
Case features as pronominal affixes, their checking ought to involve a
local head-head configuration, like checking of tense features, for
example. This means that Case checking may alternatively involve both
core confi~urations for inflectional morphology: Spec-head or head-head
relations. 2 The clitic exhausts the structural Case of the noun, which
explains why a full noun phrase cannot receive the Case of the
construct state (65). Thus, when a clitic cooccurs with a fuH noun
phrase in the relevant argument position, the latter must have re course
to inherent genitive Case available through sei ('of'):

(72) hofa'at-o • (§el) ha-'i§

appearance-his (of) the-man
'the man's appearance'

When the clitic is not doubled by a fuH noun phrase, the argument
position must be occupied by an empty category, pro, according to
current typologies of empty categories. If pro has to be licensed in a
Spec-head relation with a richly specified functional head (Chomsky
1993 based on work by Rizzi 1982, 1986a), SpecAgrgenP would be an
appropriate licensing site. Reasonably, after checking its Case and <1>-
features with the clitic, Agrs.en qualifies as a richly specified head that
can license pro in its specitler position.
Now, consider the coindexation pattern. The Hebrew nominal clitic
can be coindexed either with pro (73a) or with a phonetically realized
noun phrase in a sei ('of') phrase. However, if a sei phrase is present,
the clitic must be coindexed with the sei phrase (73b). If coreference
is impossible because the two elements do not share the same <1>-
features, it yields ungrammaticality (73c) (as already observed by Borer

(73) a. te'ur-oj pro j

'his description'

b. te'ur-oj/.j §el danj

description-his 0/ Dan
'dan's description'

c. • te'ur-a §el dan

description-her 0/ Dan
In other words, in the presence of a sei phrase, the clitic cannot be
coindexed with pro:

(74) • te'ur-o·I pro·I §el dan·J

description-his 0/ Dan
'his description of Dan'

I would like to suggest that this strict coindexation pattern follows from
the Biunique Relation Requirement discussed in section 2.3.4. Recall
that the requirement states that two arguments of one head noun cannot

each receive genitive Case (whether structural or inherent). In apparent

contradiction with this restriction, in clitic doubling constructions the
noun licenses both structural genitive Case (realized by the clitic) and
inherent genitive Case (realized by the sei phrase). However, in contrast
with the examples investigated in 2.3.4, clitic doubling constructions
involve two genitive elements that are not two distinct arguments, but
rather share the same semantic role. If the restriction on genitive Case
assignment concerns only distinct arguments of the noun, the puzzle is
solved. First, clitic doubling constructions are not counterexamples to
the restriction, as they do not entail two distinct arguments. And
second, the use of two genitive Cases obligatorily results in a clitic
doubling construction, because coindexation and thematic sharing are
aprerequisite for this use.
As expected, when the noun takes an accusative argument, the clitic
may be coindexed with pro:

(75) te'ur-oj proj let dan

description-his ACC Dan
'his description of Dan'

In sum, it is proposed that the strict coindexation pattern in nominal

clitic doubling constructions stems from the constrained genitive Case
capacities of nouns. I resume discussion of this matter when I disctiss
concrete nouns, which present the second type of apparent
counterexamples to the suggested restriction on genitive Case
assignment (see appendix to chapter 3).


The chapter has focused on genitival constructions in Hebrew,

examining various aspects of their syntax. In particular, it has shown
that given their configurational internal structure and the relative
ordering of their arguments, Hebrew noun phrases must always involve
overt noun raising. The landing site has been identified as the head D,
which means that the nominal expression cannot simply be the maximal
projection of N. Moreover, it has been shown that noun phrases
involving a construct state have a more articulated internal structure
than noun phrases involving a free state, as they contain an additional
functional category between NP and DP, AgrP, where structural genitive
Case is checked. In free states, inherent genitive case is realized in a sei
('of') phrase. Nominal clitics materialize the structural Case features of

the noun, and clitic doubling constructions exploit both kinds of

genitive Case: inherent and structural.


I Although the head noun in (3b) is not accompanied by the definite article, it is
interpreted as definite (see section 2.1.1).

2 The phonological alternation between a head noun in the construct state and its
free state counterpart is clearly due to the absence of stress on the former. Thus,
for instance, an initial vowel Ia! is reduced to a schwa when non-adjacent to stress;
a suffix It/, elided under stress, fails to delete. For phonological discussion of the
construct state, see McCarthy (1979).

3 When a head in the construct state takes a pronominal clitic as its complement,
a coreferring noun phrase can surface in a sei ('of') phrase, resulting in the so-
called clitic doubling construction:
(i) beyt-oj §el ha-'i~j
house-his 0/ the-man
'the man's house'
Clitic doubling constructions are dealt with in section 2.4.

4 Regarding deverbal nouns, Rosen (1957) observes that when the complement of
the construct state is the only realized complement within that noun phrase, it
generally refers to the internal argument; the sei ('of') phrase, according to hirn,
tends to designate the possessor. Rappaport and Doron (1990) derive this
difference from the fact that the relevant deverbal nouns have more readily event
reading when they appear in the construct state. As this matter is not relevant for
the arguments in this chapter, I will disregard it here. See next chapter for
discussion of event versus result nominals.

S Besides construct state nominals which are reasonably productive, Hebrew also
has compounds. Compounds consist of exactly two no uns wh ich result in
idiosyncratic meaning:
(i) beyt sefer
house book
(ii) ro§ 'ir
head city
In some respects they behave Iike construct states; for instance, the head noun in
both cases loses stress and undergoes the same phonological alternations. For

discussion of compounds, see Borer (1989); see also Agmon-Fruchtman (1982).

6 I put aside deverbal nouns expressing result, which have much freer relations
with their complements, as weil as known peculiarities of event nominals (like the
optionality of the subject), because the issue in this section is how nominal 6-grids
are mapped onto syntactic structures. For discussion of result versus event
nominals, see chapter 3.

7 The tripartite structure (18a) should apriori be excluded under a strictly binary
X'-theory. For reasons of completeness, I show that it must also be dismissed on
empirical grounds.

8 It has been argued that anaphors in noun phrases are used logophorically and are
not governed by grammatical conditions (Reinhart and Reuland 1993). However,
anaphors appearing with event nouns «19a), (20», unlike logophoric anaphors, do
not give rise to variations in judgements and seem to require a c-commanding
antecedent. Anaphors appearing with concrete nouns indeed seem to be logophoric
(see appendix to chapter 3). This may suggest that eventhood is relevant to the
definition of the domain where Principle A of the binding theory applies (as
suggested by Ben-Shalom and Weijler 1991).

9 If the noun phrase contains more structure than NP, one could argue that the
prominence of the Agent argument is obtained via covert raising. However, as will
become clear in what folIows, the Agent receives its Case within the sei ('of')
phrase, and has no reason to raise further (movement being a Last Resort
operation; see chapter 1).

10 A further step would be to generalize and argue that even in languages where
D constitutes an independent word, it bears features it must check with the noun,
which probably trigger (covert) noun raising to D (Chomsky 1993). Even if
definiteness in these languages is not a feature of the noun, but of a free standing
article, N may still need to check other features (say cI>-features) in D.

11 I abstract away from the possibility that there are additional functional
categories between NP and DP that bear other inflectional nominal specifications
such as number (Ritter 1991, Valois 1991, Cinque 1992, 1993, Bernstein 1993,
among others), as it has no real effect on the analysis defended here. Although the
existence of say NumberP would provide a landing site for the raised noun,
reasonably Number does not head the noun phrase, nor does it have the determiner
in its specifier position. See subsequent section for the claim that noun phrases
involving a ~onstruct state contain an AgrP that expresses an agreement relation
between the head of the construct state and its genitive DP.

12 A parallelism between the derivation of VSO surface order and that of NSO
surface order was first drawn by Ritter (1987) in the context of construct states
(discussed here in section 2.3); see also Fassi Fehri (1989).
I use IP instead of a more articulated sentential structure when details are
irrelevant (as in (25». In addition, Iassume here for the sake of simplicity that the

subject is in its base position. It may bowever be in an intermediate position

between SpecVP and SpecAgr.P, given a more articulated IP-structure.

13 This issue is somewbat more complicated in Romance, as both prenominal and

postnominal adjectives can occur in tbe same language. For discussion of the
matter, see Cinque (1992, 1993), Bernstein (1993), among others.

14 In fact, as will become clear in wbat folIows, the genitive argument of the
construct state surfaces in a position higher than SpecNP.

U Government may be defined as folIows: a governs XP if a and XP c-command

each other; and if a governs XP, then a governs the specifier and the head X of
XP (Chomsky 1986a).

16 In (42b) John is not affected by the head noun and cannot be realized in the
specifier position, according to the Affectedness Constraint, formulated by
Anderson (1979) as follows (but see also Tenny 1987, Giorgi and Longobardi
1991, Guasti 1992, among others):
(i) If a head noun does not express an action which "affects" (i.e.
modifies) the state of tbe object, tbe latter cannot be
For our purposes, it only matters that tbis realization of genitive Case like the
realization via 0/ is contingent upon thematic relationships, and therefore not a
structural Case.

17 Tbe examples in (43) are due to Julia Horvath (personal communication), who
notes tbat the deverbal noun seems to form a complex with the predicate,
analogously to the corresponding verb. This is not of our concern here.

IS Alternatively, Agr CD may bear weak N-features; in this case, strong N-features
of 0 (see section 2.~.2) will suffice to trigger overt noun raising to 0 through
Agrson '

19 Traditionally, the suffix -n is analyzed as an indefinite marker. But see Fassi

Fehri (1993) for the claim that it is an instantiation of a functional head,

20Being part of Agr,en' the [±definite] feature presumably need to undergo both
Spec-head checking lwith the genitive DP) and head-head checking in D.

21 Alternatively, it can be claimed that there is some basic incompatibility between

Agrllen features and an intrinsic definiteness feature. The hypothesis would be that
this combination of features is morpbologically impossible (along Iines proposed
by Fassi Fehri 1989 and Siloni 1990b, 199Ia). This means that a noun in the
construct state is inserted underspecified with regard to definiteness. Consequently,
an additional mechanism has to be assumed to explain the way the noun acquires

a [±definite] value. As its definiteness value is identical to that of the genitive

member of the construct, with which it is in an agreement relation, it could be
suggested that agreement configurations allow feature endowment between Spec
and head in addition to feature checking (the dynamic agreement mechanism
utilized by Rizzi 1991 a for different reasons).

22 The examples in (50) are grammatical under the reading that takes sei ha-'ir ('of
the city') to be the complement of ha-cava ('the army'). This is completely
irrelevant here.

23 When the external argument is not phonetically realized, the internal argument
cannot receive accusative Case. This impossibility is dealt with in detail in the
subsequent chapter (section 3.5).

24 In section 2.4 I discuss clitic doubling configurations, and in the appendix to

chapter 3 concrete nouns, which can involve more than one instance of genitive
Case. English noun phrases involving both aprenominal and a postnominal
genitive DP (i) constitute a further, salient counterexample:
(i) the army's destruction of the city
One possibility that immediately comes to mind would be that the prenominal
instance of genitive Case is not assigned by the noun itself (see Abney 1987,
Cinque 1994). Another line of reasoning could develop the claim that the
prenominal DP is not a real argument ofthe noun (a la Grimshaw 1990). I will not
discuss this type of examples any further here.

25 Agrgen is too rich to allow PRO, and too poor to license pro (see chapter 5 for
more evidence in favor of this claim). When Agr en hosts a pronominal clitic, it
can license pro (see section 2.4). Moreover, recal' that the head of the construct
state loses stress and the main stress falls on the genitive member of the construct,
which means that the phonetic realization of the latter mayaiso be obligatory for
PF reasons (see also section 2.3.5).

26 A nominal clitic can be doubled by a sei ('oe') phrase which itself realizes a
pronominal clitic; it has though an emphasizing nuance that doubling with a full
DP does not have:
(i) hofa'at-oj §el-oj
appearance-his of-him
'his appearance'

27 Borer (1984) has already proposed that clitics are speil-outs of the Case features
of their host. She analyzes Semitic and Romance object clitics alike. There is,
however, growing evidence that Romance object clitics form a distinct group.
Unlike Semitic clitics, they are morphologically c10se to determiners, mostly
proclitic, and create clusters. In some Romance languages, clitic constructions
exhibit a sensitivity to syntactic principles typical of movement. For example, it
is known that wh-movement can take place out of a subcategorized PP (i), hut not

out of an adjunet (ii), as it would violate Huang's Condition on Extraetion

Domains (see Huang 1982, Chomsky 1986b, Cinque 1990, among others).
Clitieization patterns on a par «iii), (iv» (thanks to Luigi Rizzi for the Italian
data; for more diagnosties of movement, see Couquaux 1981, van Riemsdijk and
Williams 1986, Kayne 1989b, Sportiehe 1992, among others»:
(i) L' uomo [a eui]j Maria si e messa [pp aeeanto eCj ]
the man to whom Maria herself was pul near
(ii) * L' uomo [a eui]j Maria e feliee [pp aeeanto eCj ]
the man to whom Maria is happy near
(iii) Maria gli si e messa [pp aeeanto eCj ]7
Maria himj herself was put near
(iv) *Maria gli e feliee [pp aeeanto eCj ]7
Maria himj is happy near
This has led linguists to argue that at least in some Romanee languages
elitieization involves movement. For a more thorough eomparison of Semitie and
Romanee elities, see Shlonsky (1994) and Siloni (l994a). Shlonsky (1994) analyzes
Semitie elities as Agr heads that are ineorporated with their host syntaetieally and
do not absorb Case. Shlonsky's analysis does not prediet the eomplementary
distribution of elities and fuH noun phrases iIIustrated in (65).

28 Belletti (1993) independently proposes that Case Cheeking ean take plaee in
a head-head eonfiguration in Romanee.



The similarities and differences between verbs and their corresponding

derived nouns have been a central issue of inquiry in the last decade.
Distinguishing between event nominals (also known in the literature as
process or action nominals) and result nominals (sometimes concrete or
simple nouns), linguists have investigated the behavior of nouns with
respect to argument structure and 8-theory (Anderson 1983-84, Lebeaux
1986, Grimshaw 1990, among others).
In an influential study of the nominal system, Grimshaw (1990)
establishes clear diagnostics to discriminate between the two types of
nominals, showing that only event nominals share their argument
structure with the corresponding verbs. Result nominals do not entail an
event and do not have an argument structure; therefore, they do not
have specific 8-roles to discharge. Lexical entries involving an event
(whether verbs or nouns) have an event structure and an argument
structure to satisfy.
To derive this systematic symmetry, Borer (in progress) suggests that
noun phrases expressing an event contain a fully projected VP. This
idea is not new; it dates back to the earliest works in generative
grammar (Lees 1960). It is the projected VP which is responsible for
the event reading and the appearance of argument structure. However,
if event nominals contain a verbal projection, it is unclear why they do
not show typical verbal properties, such as adverbial modification or
accusative complements.
In this context, Hebrew event nominals are extremely interesting, as
they seem to show verbal behavior. First, Hebrew event nominals can
assign accusative Case to their direct object: 1

(1) a. ha-harisa §el ha-cava let ha-'ir

the-destruction 0/ the-army ACC the-city
'the army's destruction of the city'


b. harisat ha-cava 'et ha-'ir

destruction the-army ACC the-city
'the army's destruction of the city'

Second, they can be modified by adverbs:

(2) a. harisat ha-cava 'et ha-' ir bi-mehirut

destruction the-army ACC the-city in-quickness(=quickly)
'the army's quick destruction of the city'

b. harisat ha-cava 'et ha-'ir be-'axzariyut

destruction the-army ACC the-city in-cruelty(=cruelly)
'the army's cruel destruction of the city'

This has led Hazout (1990, 1995) to argue that Hebrew event
nominals contain a verbal projection. Borer (in progress) has adopted
this approach for event nominals crosslinguistically.
Apriori, the syntactic derivation of event nominals simplifies the
lexicon, enriching the syntactic component. It does not require any
lexical information regarding event nominals, which are derived by
some extension of the syntactic apparatus. The alternative approach, the
lexicalist approach, which does not derive event nouns syntactically
(Chomsky 1970, Grimshaw 1990, among others), simplifies syntax at
the price of a richer lexicon. The choice between the two approaches
is primarily an empirical issue.
This chapter shows that there is no empirical reason to believe that
Hebrew event nominals contain a verbal projection, since their verbal
properties are only apparent. On the contrary, this assumption
encounters problems of over-generation. I argue in favor of a lexicalist
approach to event nominals, taking the eventlresult distinction to be part
of the lexical information adeverbal noun can have. If event nouns do
not entail syntactic nominalization, it is not surprising that their
formation is not systematically productive, that they do not manifest
genuine verbal properties, and that they do show some otherwise
unexpected idiosyncrasies.
The first section of the chapter establishes the partition of Hebrew
nouns into event and result nominals, on the basis of tests suggested by
Grimshaw (1990). In addition, the section shows that in Hebrew,
accusative complements and adverbial modification safely make the
event versus result distinction, as they occur with event nominals only.
Section 3.2 shows that event nominals are in fact not modifiable by real

adverbs, and section 3.3 argues that there are important reasons to
believe that the accusative Case occurring in noun phrases is not the
ordinary accusative of transitive verbs. Section 3.4 adds arguments to
the effect that accusative Case in nominal contexts is an inherent Case
whose occurrence cannot be systematically predicted. Finally, section
3.5 analyzes subjectless event nominals.


Nouns can either refer to an event (or a process) or denote a result, that
is, the output of an event or an element associated with it. The labels
"event" and "result" may not clearly reflect the range of relevant cases,
but their coverage will become clear as we proceed. The same noun is
often ambiguous between these two interpretations. Grimshaw (1990)
shows that only nouns referring to an event are associated with an
argument structure, similarly to their verbal counterparts, and must
therefore realize their (internat) arguments, to which they assign specific
6-roles. 2 The term "event" is used here in the broad sense: it is the
aspectual dimension associated with the argument structure. While all
verbs, whether verbs of change, action verbs, or statives, have this
dimension (an event-position, under Higginbotham's 1985 view) and
obligatorily take arguments, no uns are divided into argument taking
nouns and nouns with no arguments. A noun takes arguments if it refers
to an event, and hence has an event structure. Result nominals are
nouns that do not involve an event, and hence do not have an argument
structure to satisfy. Thus, only the noun in (3a-b) can have an event
reading; it can refer to the event of examining the documents. Its
homonym in (3c) cannot have this reading, as it realizes no arguments.
As will become clear in what folIows, this division among nouns is
supported by aseries of tests:

(3) a. bxinat ha-mismaxim hayta hexrexit.

examination the-documents was necessary
'The examination of the documents was necessary'

b. ha-bxina shel ha-mismaxim hayta hexrexit.

the-examination 0/ the-documents was necessary
'The examination of the documents was necessary'

c. ha-bxina hayta hexrexit.

the-examination was necessary
'The exam was necessary'

3.1.1. Grimshaw 's diagnostics

Grimshaw (1990) suggests various diagnostics for discriminating

between event and result nominals. Applied to Hebrew, her diagnostics
confirm the claim that only event nominals have an argument structure.
Certain frequency modifiers occur only with the event interpretation
of particular singular nouns. For example, the modifier tadir
('frequent') requires the event reading of bxina ('examination'), thus
distinguishing between the event nominal and its result homonym. Once
we are able to disambiguate, we see that a noun with no arguments
cannot have an event reading (compare (4a-b) with (4C»:3

(4) a. bxinat ha-mismaxim ha-tedira hayta hexrexit.

examination the-documents the-frequent was necessary
'The frequent examination of the documents was necessary'

b. ha-bxina ha-tedira seI ha-mismaxim hayta

the-examination the-frequent of the-documents was
'The frequent examination of the documents was necessary'

c. * ha-bxina ha-tedira hayta hexrexit.

the-examination the-frequent was necessary

Aspectual modifiers, such as tox sa'a ('in an hour'), are also used to
diagnose eventhood. Only no uns denoting an event have the aspectual
structure needed to license aspectual modifiers (the same modifiers as
those admitted by the corresponding verbs (Vendler 1967, Dowty 1979).
An argument taking noun has an event structure and admits aspectual
modifiers (Sa-b); its homonym with no arguments disallow them,
because it is a result noun (Sc):

(5) a. bxinat ha-mismaxirn tox ~a'a hir~ima 'oto.

examination the-docurnents in hour impressed hirn
'The examination of the docurnents in an hour impressed

b. ha-bxina ~el ha-mismaxim tox ~a'a hir~ima

the-examination 0/ the-documents in hour impressed
'The examination of the documents in an hour impressed

c. * ha-bxina tox ~a'a hir~irna 'oto.

the-examination in hour impressed him

Agent oriented adjectives, too, can serve to disambiguate nouns.

These adjectives can modify only nouns that have an event reading,
because they dernand an explicitlirnplicit Agent. If there is no event,
there is no argument structure and there cannot be a true Agent.
Therefore an Agent oriented adjective renders (6c) (but not (6a-b»
ungramrnatical, as a noun cannot have an event reading without
satisfying the associated argument structure:

(6) a. bxinat ha-mismaxirn ha-mexuvenet 'arxa

examination the-documents the-intentional las ted
'The intentional examination of the documents lasted two

b. ha-bxina ha-mexuvenet ~el ha-mismaxirn 'arxa

the-examination the-intentional 0/ the-documents lasted
'The intentional examination of the documents lasted two

c. * ha-bxina ha-mexuvenet 'arxa ~a'atayirn.

the-examination the-intentional las ted two+hours

Rationale (infinitival purpose) clauses, similarly to Agent oriented

adjectives, require an event reading. Williams (1985) and Lasnik (1988)
argue that these infinitivals involve control by an event, while Roeper
(1987) claims that they require control by an (implicit) argument. At
any rate, under either approach, an event reading is crucial. Hence, a
rationale clause cannot be added to the noun in (7c):

(7) a. bxinat ha-mismaxim ba-gvul kedey

examination the-documents in+the-border in+order
li-lkod mavrixey samim
to-capture smugglers drugs
'the examination of the documents on the border in order
to capture drug smugglers'

b. ha-bxina §el ha-mismaxim ba-gvul kedey

the-examination 0/ the-documents in+the-border in+order
li-lkod mavrixey sam im
to-capture smugglers drugs
'the examination of the documents on the border in order
to capture drug smugglers'

c. ... ha-bxina ba-gvul kedey li-lkod

the-examination in+the-border in+order to-capture
mavrixey sam im
smugglers drugs

Grimshaw (1990) further notes that while the English by phrase can
be used independently of the presence of an argument structure, in other
languages the distribution of by phrases is more restricted. Indeed, the
Hebrew 'al-yedey ('by') phrase is contingent upon the presence of an
argument structure. Hence, it can only appear with an argument taking
noun (compare (8a-b) to (8C»:4

(8) a. bxinat ha-mismaxim 'al-yedey ha-rdut

examination the-documents by the-authority
'the examination of the documents by the competent

b. ha-bxina §el ha-mismaxim 'al-yedey ha-rdut

the-examination 0/ the-documents by the-authority
'the examination of the documents by the competent

C. * ha-bxina 'al-yedey ha-rdut ha-musmexet

the-examination by. the-authority the-competent
'the examination by the competent authority'

Notice that when a result nominal appears with a genitive noun

phrase, the relationship between the two is rather vague; result nominals
do not have specific 8-roles to discharge, as already noted by Anderson
(1983-84). In (9a), for example, the referent of dan can be the owner,
the caretaker, the admirer or the creator of the reconstruction. This
referent has some claim of possession over the physical object which
is the referent of ha-sixzur ('the reconstruction'), but it is not in a
specific 8-relation with it. It is not its argument, rather it is a semantic
participant, a modifier (in Grimshaw's 1990 terms), an adjunct (in
Szabolcsi's 1992 terms). If an event reading is forced, for instance by
a rationale clause as in (9b), where dan is clearly the Agent of sixzur
('reconstruction'), the sentence is ungrammatical, unless the argument
structure of sixzur is satisfied: 5

(9) a. ha-§ixzur §el dan nizok ba-srefa.

the-reconstrur;tion 0/ Dan was+damaged in+the-fire
'Dan's reconstruction was damaged in the fire'

b. ha-§ixzur §el dan *('et ha-pe§a) kedey

the-reconstruction 0/ Dan (ACC the-crime) in+order
le-havin 'et ha-macav
to-understand ACC the-situation
'Dan's reconstruction ofthe crime in order to understand the

To summarize, result nominals do not involve an event and do not

take arguments. Their apparent arguments are semantic participants with
which they are not in clear 8-relations. Event nominals, in contrast,
have an argument structure to satisfy, similarly to verbs. In this context,

Hebrew event nominals become especially interesting. Their

resemblance to verbs is even more salient, as they can be modified by
adverbs and take accusative complements, unlike their English
counterparts. These characteristics of Hebrew event nominals are
discussed below.

3.1.2. More diagnostics

In Hebrew, a noun denoting an event, an argument taking noun, can be

modified by an adverb. Result nouns disallow adverbs, as shown by the
contrast between (lOa-b) and (lOc):

(10) a. xakirat ha-ne'eäam bi-mehirut

interrogation the-accused in-quiclcness(=quickly)
'the quick interrogation of the accused'

b. ha-xakira äel ha-ne'eäam bi-mehirut

the-interrogation 0/ the-accused in-quiclcness(=quickly)
'the quick interrogation of the accused'

c. * ha-xakira bi-mehirut
the-interrogation in-quickness(=quickly)

In the same vein, a noun like re'ayon ('interview'), which can only
denote a result, as shown by the fact that the addition of the modifier
tadir ('frequent') results in ungrammaticality (lla), cannot be modified
by an adverb (llb):

(11) a. * ha-re'ayon ha-tadir 'im ha-nasi

the-interview the-frequent with the-president
'the frequent interview with the president'

b. *ha-re'ayon 'im ha-nasi bi-mehirut

the-interview with the-president in-quiclcness(=quickly)
'the quick interview with the president'

Moreover, Hebrew nouns allow an accusative Theme. A noun taking

an accusative complement has an event reading, as shown by the fact
that it licenses a rationale clause (12a), the modifier tadir ('frequent')
(12b), an aspectual modifier (12c) an Agent oriented adjective (12d) or

an Agent oriented adverb (12e):6

(12) a. hafcacat ha-cava let ha-'ir kedey le-havri'ax

bombing the-army ACC the-city in+order to-drive+away
'the army's bombing of the city to drive away inhabitants'

b. ha-hafcaca ha-tedira §el ha-cava let ha-'ir

the-bombing the-frequent 0/ the-army ACC the-city
'the army's frequent bombing of the city'

c. hafcacat ha-cava let ha-'ir tox §a'a

bombing the-army ACC the-city in hour
'the army's bombing of the city in an hour'

d. ha-hafcaca ha-mexuvenet äel ha-cava let ha-'ir

the-bombing the-intentional 0/ the-army ACC the-city
'the army's intentional bombing of the city'

e. hafcacat ha-cava let ha-'ir

bombing the-army ACC the-city
'the army's malicious bombing of the city'

On a par with adverbs, accusative arguments are impossible with

nouns that cannot denote an event. As re'ayon ('interview') can only
name a result, (13) is ungrammatical:

(13) * ha-re'ayon äel dan let ha-nasi

the-interview 0/ Dan ACC the-president

Adverbial modification and accusative Case assignment safely set

apart Hebrew event and result nominals. 7 However, these are properties
typical of verbs. Hebrew event nominals then seem to have an
ambiguous nature. On the one hand, they are clearly nominal: they have
the distribution of noun phrases, a definiteness value, can be modified
by adjectives, and license sei ('of'), which is strictly limited to nominal
contexts. On the other hand, they show some verbal behavior, as just

discussed. The question arises as to whether they really have an

ambiguous categorial nature, which should be echoed in syntactic

3.1.3. Syntactic derivation: the VP-analysis

In order to derive the verbal properties of Hebrew event nominals;

Hazout (1990, 1995) and Borer (in progress) have proposed that the
syntactic representation of, these nominals contains a verbal projection.
That is, they are not represented in the lexicon as nouns, but inserted
in syntax as verbs that are nominalized in the course of the derivation
via V-raising to N. The structure and the incorporation of V into N are
schematized in (14):

(14) OP


o NP

I N'


The presence of a verbal projection in the structure accounts for the
verbal properties of the construction: adverbial modification and
accusative complements. The nominal part derives its nominal
properties. Although languages differ with regard to the verbal behavior
of their event nominals, Borer (in progress) suggests that event
nominals are derived in syntax across languages. Under her analysis,

eventhood and argument structure characterize verbs, not nouns.

Before shedding more light on the arguably verbal behavior of
Hebrew event nominals, I would like to note that the incorporation of
V into N suggested in (14) differs from other processes of incorporation
of two lexical heads described by Baker (1988). More specifically, it is
not simply the counterpart of noun incorporation with V. Noun
incorporation puts together two semantically autonomous units. For
example, the noun 'money' incorporates with the verb 'lose', resulting
in the complex 'money-Iose' in Onondaga, an Iroquoian language (see
Baker 1988). In contrast, the incorporation of V into N in (14) involves
a semantically "empty" noun, which inherits the semantic content of the
incorporated verb. It is in fact a nominalizing functional affix labeled
N, which serves to convert the verb into a noun, to modify its categorial
specification. Moreover, while noun incorporation does not entail a
particular verbal subcategorization frame (as noun phrases are typically
complements of verbs), V incorporation into N requires that the
nominal affix subcategorize for a VP, which is exceptional for nouns.
In addition, this subcategorization frame is limited in an unpredictable
fashion, as it is not the case that any verb can head a VP
subcategorized by N: it is well-known that the formation of deverbal
no uns is not systematically productive. The approach schematized in
(14) then increases the syntactic machinery. Its lexicon may be simpler,
as it does not have to represent the nominal ambiguity (event/result). Is
the approach advantageous on empirical grounds?
In this chapter I show that the VP-analysis of event nominals (14)
encounters serious difficulties, which cast doubts on its adequacy.
Moreover, I put forward the restrictive hypothesis that a process of
nominalization that modifies the categorial specification of the verb in
the way just described is not part of UG. I first show that Hebrew event
nominals cannot be modified by genuine adverbs, in spite of


As mentioned above, Hebrew event nominals seem to be modifiable by

adverbs, which is clearly a verbal property. It can be suggested that
adverbs in noun phrases are licensed by the event (see Davidson 1967
and Higginbotham 1985 for the claim that adverbs are predicated of
events). This explanation is plausible, as shown by the fact that adverbs
are never allowed when the noun is a result nominal. Thus, as already
shown above, the result nominal re'ayon ('interview'), wh ich cannot be

modified by the adjective tadir ('frequent') as this adjective requires an

event reading (ISa), cannot be modified by an adverb either (lSb):

(IS) a. • ha-re'ayon ha-tadir 'im ha-nasi

the-interview the-frequent with the-president

b. • ha-re'ayon 'im ha-nasi bi-mehirut

the-interview with the-president in-quickness(=quickly)

This explanation, however, is not adequate. If an event sufficed to

license adverbs, we would expect them also to be admitted with English
or French event nominals, contrary to the facts:

(16) a. • The army's destruction ofthe city quickly

b. • La fabrication de chaussures rapidement

the manu/acture 0/ shoes quickly

I would like to suggest that it is the particular categorial status of

Hebrew adverbs which enables them to modify event nominals. Most
adverbs in Hebrew are, in fact, adverbial PPs. The class of genuine
adverbs consists of a small number of fixed elements. As shown in
(17), a genuine adverb cannot modify an event nominal in Hebrew, in
the same way that it cannot modify its English or French counterpart:

(17) a. • harisat ha-'ir maher

destruction the-city quickly

b. • slilat ha-kvi§ le'at

paving the-road slowly

c. • clilat ha-yeled 'amukot

diving the-boy deeply

If event nominals contained a verbal projection, as suggested by the

VP-analysis, this bifurcation would be rather mysterious. But if event
nouns are purely nominal, as argued here, it becomes clear why only
adverbial PPs are allowed. Reasonably, while PPs can occur within
noun phrases, genuine adverbial phrases (AdvP) cannot do so. It follows
that event nominals admit adverbial PPs, because PPs are legitimate
nominal modifiers and the presence of an event licenses adverbial

Note that these adverbials are real PPs: their internal structure is not
frozen; it is syntactically present, as shown by the fact that their
nominal complement can be modified. In (18a) mehirut ('quickness') is
modified by raba ('great') and in (18b) nimus ('politeness') is modified
by me'ulac ('forced'):

(18) a. slilat ha-kvis bi-mehirut raba

paving the-road in-quickness great

b. 'axilat ha-bananot be-nimus me'ulac

eating the-bananas in-politeness /orced

As expected, adverbial PPs can also modify event nominals in other


(19) a. La fabrication de chaussures en bäte

the manu/acture 0/ shoes in haste

b. The army's destruction of the city in haste

In Hebrew the adverbial modification of event nominals is more striking

simply because adverbs are mostly adverbial PPs, and consequently the
distinction between authentic adverbs and adverbial PPs is somewhat
blurred. 8
Hazout (1990, 1995) observes that adjectives and adverbs (our
adverbial PPs) have a different distribution: adjectives must precede the
sei ('of') phrase (20a-b), while adverbial PPs can only follow it (20c-d).
According to Hazout, lower than the sei phrase, the projection is verbal;
hence, adverbial PPs are allowed, but not adjectives. Higher, the domain
is nominal and therefore adjectives are allowed, but not adverbial PPs:

(20) a. ha-harisa ha-'axzarit sei ha-batim

the-destruction the-cruel 0/ the-houses

b. • ha-harisa sei ha-batim ha-'axzarit

the-destruction 0/ the-houses the-cruel

c. ha-harisa sei ha-batim be-'axzariut

the-destruction 0/ the-houses in-cruelness

d. • ha-harisa be-'axzariut sei ha-batim

the-destruction in-cruelness 0/ the-houses

However, an examination of concrete nouns reveals that the described

distribution is typical of Hebrew noun phrases in general, regardless of
their eventlresult nature. That is, this distribution is independent of the
conjectured existence of a verbal projection. An adjective cannot follow
the sei ('of') phrase, while a pp has to follow it: 9

(21) a. ha-bayit ha-yafe §el dina

the-house the-beautiful 0/ Dina

b. • ha-bayit §el dina ha-yafe

the-house 0/ Dina the-beautiful

c. ha-bayit §el dan 'im ha-'aruba

the-house 0/ Dan with the-chimney

d. • ha-bayit 'im ha-'aruba §el dan

the-house with the-chimney 0/ Dan

To sum up, genuine adverbs cannot appear in Hebrew noun phrases

in the same way that they cannot appear in their English or French
counterparts. Adverbial PPs, on the other hand, can modify event
nominals. Syntactically, they are allowed in noun phrases as they are
PPs. Semantically, they are licensed by the presence of an event.
I now turn to the other arguably verbal characteristic of Hebrew
event nominals, namely, their apparent ability to assign accusative Case.
I show that the accusative Case appearing in noun phrases, unlike its
verbal equivalent, is not a structural Case.


Crosslinguistically, accusative Case is by and large a type of Case

available in verbal contexts. A notorious puzzle raised by Hebrew event
nominals is their ability to assign accusative Case to their direct object,
an anomalous behavior as far as nouns are concerned. In this section I
show that there are significant distinctions between the accusative Case
of transitive verbs and that found in nominal contexts. I present five
major distinctions between the two, which concern the accusative
particle ('et), exceptional Case marking configurations, accusative
pronouns, the relative ordering of the direct and indirect objects, and
subjectless noun phrases.
The most salient distinction between the accusative Case of transitive

verbs and that of nouns concerns the particle 'et. This particle appears
with definite objects only (see also chapter 2, section 2.1.1). Thus,
when a verb takes adefinite accusative complement, 'et must precede
the complement (22a); when it takes an indefinite accusative
complement, 'et cannot appear (22b). Interestingly, nouns can assign
accusative Case only in the presence of 'et (22c) (as first observed by
Borer 1984). As 'et is limited to definite objects, (22d) is

(22) a. ha-cava haras *('et) ha-'ir.

the-army destroyed (ACC) the-city
'The army destroyed the city'

b. ha-cava haras (*'et) 'ir 'axat.

the-army destroyed (ACC) eity one
'The army destroyed one city'

c. harisat ha-cava 'et ha-'ir

destruetion the-army ACC the-city
'the army's destruction of the city'

d. * harisat ha-cava ('et) 'ir 'axat.

destruetion the-army (ACC) eity one
'the army's destruction of one city'

As there is no reason to assume that a definiteness requirement is

imposed on objects of derived nouns (when they receive genitive Case
they can of course be indefinite), it seems that 'et plays a crucial role
in the assignment of this Case in nominal contexts. This distinction
between the accusative Case of verbs and that of event nominals is
completely unexpected under the VP-analysis, which attributes the
occurrence of accusative Case in nominal contexts to the presence of
a verbal head.IO,11
Second, nouns, unlike verbs, are unable to license accusative Case to
subjects of sm all clauses (e.g. in exceptional Case marking
environments (ECM», as shown by the contrast between (23a-b) and

(23) a. ha-more maca let ha-ti'un mdaxne'a.

the-teacher found ACC the-argument convincing
'The teacher found the argument convincing'

b. ha-me'amen ra'a let ha-yeled nofel.

the-coach saw ACC the-boy falling
'The coach saw the boy falling'

c. * meci'at ha-more let ha-ti'un mdaxne'a

jinding the-teacher ACC the-argument convincing

d. * re'iyat ha-me'amen let ha-yeled nofel

view the-coach ACC the-boy falling

Again, if accusative Case in nominal contexts were the Case of

transitive verbs, as suggested by the VP-analysis, this behavior would
be rather mysterious. As already discussed in the previous chapters,
Chomsky (1986a) draws a distinction between inherent Case and
structural Case. Inherent (but not structural) Case is thematicaHy
related, that is, it can be assigned by IX to DP only if IX 6-marks DP. If
so, then the fact that nouns cannot Iicense accusative Case in ECM
configurations, where the noun does not 6-mark the assignee, suggests
that their accusative is an instance of inherent Case. 13
The third distinction between the accusative Case of transitive verbs
and that of nouns mayaiso suggest that the latter is an inherent Case.
Unlike verbs, event nominals are unable to realize their object as an
accusative pronoun, as shown by the contrast between (24a) and (24b).
The ungrammaticality of (24c) shows that it is not simply the
intervenience of a fuH noun phrase between the noun and the accusative
pronoun that is responsible for the ungrammaticality of (24b):

(24) a. ha-cava haras 'oto.

the-army destroyed him
'The army destroyed hirn'

b. *harisat ha-cava 'oto

destruction the army him
'the army's destruction of hirn'

c. * harisat-am 'oto
destruction-their him
'their destruction of him'

Accusative pronouns, which are clearly definite direct objects, do not

allow 'et. As they are diachronically derived from the combination of
'et with suffixal pronouns, they arguably contain the accusative particle

(25) ha-cava haras (*'et) 'oto.

the-army destroyed (ACC) him

Assuming that the Case these accusative pronouns bear is intrinsically

structural (along lines proposed by Friedemann and Siloni 1993), one
can suggest that their inability to occur in nominal contexts follows
from the fact that the accusative Case of nouns is an inherent Case (I
come back to this issue in section 3.4).14
The fourth distinction between the accusative Case of transitive verbs
and that appearing in nominal contexts concems the order of
constituents. VPs permit both word orders: accusative-dative (26a) as
weIl as dative-accusative (26b) (see Belletti and Shlonsky 1995,
Rapoport 1987). In noun phrases, the accusative complement must
precede its dative counterpart «26c) versus (26d)). In other words, the
accusative argument must be the first complement following the
(genitive) subject:

(26) a. ha-bank hexzir 'et ha-kesef le-§ula.

the-bank refunded ACC the-money to-shula
'The bank refunded the money to Shula'

b. ha-bank hexzir le-§ula 'et ha-kesef.

the-bank refunded to-shula ACC the-money
'The bank refunded the money to Shula'

c. haxzarat ha-bank 'et ha-kesef le-§ula

refund the-bank ACC the-money to-shula
'the bank's refund of the money to Shula'

d. * haxzarat ha-bank le-äula 'et ha-kesef

re/und the-bank to-shula ACC the-money
'the bank's refund of the money to Shula'

This asymmetry, too, would be unexpected, if the syntactic

representation of event nominals involved a verbal projection and if a
verbal head determined the relevant accusative Case. But if a nominal
head determines this Case, the contrast between (26b) and (26d) is less
surprising and can be attributed to the fact that the examples do not
involve the same type of Case: verbal contexts license structural
accusative Case, while their nominal counterparts determine inherent
accusative Case (see below for some more discussion).
A final distinction concerns nominals which do not (phonetically)
realize the external argument. In the three examples of (27), only the
internal argument is realized. While the structure is grammatical when
it receives genitive Case via sei ('of') (27a) or through the construct
state (27b), it is excluded when it receives accusative Case (27c), as
already observed by Ritter (1987, 1988):

(27) a. ha-harisa ~el ha-'ir

the-destruction 0/ the-city
'the city's destruction'

b. harisat ha-'ir
destruction the-city
'the city's destruction'

c. * ha-harisa 'et ha-'ir

the-destruction ACC the-city
'the city's destruction'

This behavior is not predicted by the VP-analysis; verbs can take an

accusative argument regardless of the phonetic realization of their
external argument (28):

(28) a. hem harsu 'et ha-'ir.

they destroyed ACC the-city
'They destroyed the city'

b. harastem 'et ha-'ir.

destroyed(2pL) ACC the-city
'You destroyed the city'

To explain this distinction, Borer (in progress) argues that in noun

phrases, the external argument can be absent only when the verb
embedded in the deverbal noun has undergone passivization. In a
passive environment, accusative Case is not available. Thus, according
to her, the examples of (27) are passive constructions, hence the
unavailability of accusative Case. In section 3.5, I argue against the
passive analysis of subjectless event nominals, and explain why they
cannot take an accusative argument.
In sum, it is hardly controversial to assume that the accusative Case
of deverbal nouns is not an ordinary type of accusative. It differs from
the accusative of transitive verbs in several important respects, which
renders the VP-analysis significantly less interesting. First, it can be
assigned only through a Case marker, 'et. Moreover, it cannot be
assigned in exceptional Case marking environments (Le. to subjects of
small clauses) nor can it be assigned to pronouns. In addition, its
licensing conditions are particular: it can only be assigned when the
subject is (phonetically) realized and in a position immediately
following it (preceding a dative argument).
Suppose now that accusative Case in nominal contexts is indeed an
inherent Case (assigned by 'et), as suggested by the fact that it is
blocked in ECM constructions and by the unavailability of accusative
pronouns. Other instances of inherent accusative Case have been
discussed in the generative literature: in Italian (Belletti and Rizzi
1988), in German (van Riemsdijk 1983) and in Icelandic (Zaenen,
Maling and Thrainsson 1985, Sigurösson 1989, among others). By
definition, inherent Case is assigned in situ under sisterhood (Chomsky
1986a and subsequent work). This can explain the order of constituents
illustrated in (26). In nominal environments (26c-d), the order is strict,
as both the dative argument and its accusative counterpart receive
inherent Case in their base positions and have no reason to move
further (movement being a Last Resort operation) and no reasonable
landing site. Verbal contexts (26a-b), in contrast, allow permutation in
the order of complements, as they involve a structural accusative
argument and a different structure (for a specific proposal and
discussion of the permutations in verbal complement order, see Belletti
and Shlonsky 1995). The question still arises as to why subjectless
event nominals do not license inherent accusative. I first add further
evidente in favor of the claim that the particle 'et can license inherent

accusative Case, and then discuss the puzzle raised by subjectless noun


3.4.1. 'et insertion

It seems then that while 'et appearing in verbal contexts may be a

morphological realization of structural accusative Case, its
homophonous form in nominal contexts is a dummy Case assigner
which actually assigns the Case (an inherent Case). Thus, accusative
Case can be assigned in Hebrew noun phrases precisely because this
dummy Case marker is available. But it cannot be assigned in English
noun phrases for example, because no such device is available. 1S
Consequently, one may wonder whether 'et can be shown to be the
Case assigner of a Theme argument elsewhere. Some evidence in this
direction is offered by passive and unaccusative configurations in
colloquial Hebrew. As first observed by Shoshani (1980), when the
Theme argument is definite, 'et can sometimes be inserted: 16

(29)a. 10 nimsar li 'et ha-hoda'a ha-zot.

NEG was+transmitted to+me ACC the-message the-this
'This message was not transmitted to me'

b. noda li 'et ze 'etmol.

was+known to+me ACC this yesterday
'I found it out yesterday'

c. 10 haya katuv §am 'et ha-§a'a.

NEG was written there ACC the-hour
'The hour was not written there'

d. kara li kvar 'et ha-te'una ha-zot.

happened to+me already ACC the-accident the-this
'I already had this accident'

When the Theme argument is indefinite, the verb must agree with it,
which shows that the Case it bears is not accusative but rather
nominative. Thus, while in (29a), for example, 'et appears and the verb
does not agree with its Theme argument, in (30a) 'et is absent and the

verb agrees with its Theme. Lack of agreement results In

ungrammaticality (30b):

(30) a. 10 nimser-a li hoda'a.

NEG was+transmitted-FM to+me message(FM)
'A message was not transmitted to me'

b. * 10 nimsar li hoda'a.
NEG was+transmitted to+me message(FM)

In section 3.3, it was suggested that accusative pronouns bear

structural Case. We therefore predict that they should be unable to
appear in examples of the type in (29), where only inherent accusative
is available. This prediction is borne out, as shown by the
ungrammaticality of the examples in (31) (recall that accusative
pronouns do not allow the accusative particle 'et):17

(31) a. * 10 nimsar li 'ota.

NEG was+transmitted to+me her

b. * kvar kara 'ota le-dan.

already happened her ta-Dan

Let us now consider the behavior of foreign nouns, such as

konstrukcia ('construction'), which are not assimilated to the system
and do not have corresponding verbs. Such foreign nouns cannot take
an accusative argument (32a). At first glance, this argues in favor of the
VP-analysis; why would 'et be unable to assign accusative Case in noun
phrases whose head is a foreign word? However, as observed by Borer
(in progress), foreign nouns of this type do not denote an event but
rather a result. Thus, for instance, konstrukcia ('construction') cannot
be modified by an adverbial pp (32b), contrary to its corresponding
event nominal bniya ('construction') (32c):

(32) a. * ha-konstrukcia sei ha-siltonot 'et ha-'ir

the-construction 0/ the-authorities ACC the-city
'the authorities' construction of the city'

b. ha-konstrukcia seI ha-' ir (*bi-mehirut)

the-construction 0/ the-city (in-quickness)
'the (quick) construction of the city'

c. ha-bniya ~el ha-'ir bi-mehirut

the-construction 0/ the-city in-quickness
'the quick construction of the city'

As already discussed above, Grimshaw (1990) convincingly argues that

result nominals do not take arguments, but rather se1nantic participants
with which they are not in specific 6-relations. As accusative Case can
only be assigned to a Theme (or Patient) argument, one would not
expect it to be available in noun phrases expressing a result, as they are
not argument taking nouns. Thus, it is only natural that foreign words
like konstrukcia cannot take an accusative complement. 18

3.4.2. The relevance 0/ AgroP

Chomsky (1991, 1993) has argued that structural accusative Case is
checked within the functional category AgroP' If so, then this functional
category is relevant to the issues at stake here. More precisely, the
syntactic approach can take event nominals to contain a bare verbal
projection; in the absence of AgroP' structural accusative would not be
available. 19 While under the approach advanced in this work, the lack
of structural accusative Case follows from the lack of any projection
related to V (whether lexical (VP) or functional), under the bare VP
approach, it would hinge specifically upon the lack of Agrop. I think
that there are good reasons to prefer the former approach.
First, the bare VP-analysis cannot explain why event nominals do not
allow adverbs. In the presence of a fully projected VP, modification by
adverbs should be licensed. Suggesting that the unavailability of adverbs
is related to the lack of AgroP seems not only ad hoc, but also
empirically incorrect. For example, French reduced relatives containing
past participles do not license accusative Case (33a). They are clausal
segments which do not contain AgroP (nor higher sentential projections;
see chapter 4 for argumentation and extensive discussion of reduced
relatives). But they do allow genuine adverbs (33b):

(33) a. * L' homme peint les murs ecrit des poemes.

the man painted the walls writes poems

b. Les m~rs peints rapidement craignent Phiver.

the walls painted rapidly /ear the winter

Thus, the unavailability of authentic adverbs argues direct1y against the


presence of a (bare) verbal projection.

Second, since the accusative Case available in event nominals is not
the ordinary accusative of transitive verbs, the putative presence a
verbal projection does not automatically derive its availability. Thus, as
far as this Case is concemed, the VP-analysis has apriori no advantage
over the present analysis. Moreover, as will be shown below, Doron
(1989) supplies evidence that the occurrence of accusative Case in noun
phrases is in fact idiosyncratic. If so, at any rate it cannot simply follow
from the syntactic presence of a transitive verb, as already observed by
Rappaport and Doron (1990).

3.4.3. Idiosyncratic information

Take the transitive verbs hitbi'a and tibe'a, which both mean 'sink'.
Their corresponding deverbal nouns are hatba'a and tibu'a, respectively,
wh ich can both have an event reading, as shown, for instance, by the
fact that they can be modified by adverbial PPs:

(34) a. hatba'at ha-sirot bi-mehirut

sinking the-boats in-quickness
'the quick sinking of the boats'

b. tibu'a ha-sirot bi-mehirut

sinking the-boats in-quickness
'the quick sinking of the boats'

Interestingly, as observed by Doron (1989), only hatba'a allows

accusative Case assignment (35a-b), although both corresponding verbs
are transitive:

(35) a. hatba'at ha-cava let ha-sirot

sinking the-army ACC the-boats
'the army's sinking of the boats'

b. • tibu'a ha-cava let ha-sirot

sinking the-army ACC the-boats
'the army's sinking of the boats'

This is not an isolated case. Doron (1989) lists many such minimal
pairs. The following table repeats some of her examples:


Accusative: allowed Accusative: disallowed Meaning

beki'a biku'a splitting
hariga hereg killing
harisa heres destruction
mi'ux me'ixa squeezing
netixa nitu'ax operation

This split among nouns is not systematic, cannot be predicted, and tend
to vary among speakers. For example, although mi'ux ('squeezing') and
nitu'ax ('operation') belong to the same morphological pattern, the
former can take an accusative complement, while the latter cannot.
Similarly, although beki'a ('splitting') and me'ixa ('squeezing') share
the same morphological pattern, they diverge regarding their capacity
to take an accusative complement. 20
This idiosyncrasy is completely unexpected under the VP-analysis. A
priori, the VP-analysis predicts that any event nominal derived from a
transitive verb should have the ability to assign accusative Case. Now,
if the capacity of nouns to take an accusative complement is
idiosyncratic and does not automatically follow from the putative
syntactic presence of a transitive verb, it means that the specific
information concerning this Case must be lexically encoded.
If deverbal nouns are always formed pre-syntactically and if the
ability to take an accusative complement is a lexical property of the
relevant event nominal, its idiosyncratic character is not surprising, as
it is typical of lexical structure. In contrast, if event nominals were
derived syntactically, the ability to assign inherent accusative would
have to be a lexical property of the source verb. This amounts to saying
that a verb would have to bear different lexical specifications according
to whether it is to be inserted in a clausal context or embedded under
N. This is rather unusual for lexical information.
The VP-analysis intends to simplify the lexicon at the price of
allowing more articulated syntactic processes, such as the process that
incorporates a verb into a noun (see section 3.1.3). However, it seems
that it cannot do with a significantly simpler lexicon, as most of the
charge removed from nouns must consequently figure as lexical
information of the corresponding verbs (and an unusual sort of lexical
information as it refers to the syntactic context in which the verb is

So even in Hebrew, which at first glance seemed to favor syntactic

derivation of event nominals, the VP-analysis encounters serious
problems, and therefore cannot be maintained. I would like to advance
the restrictive hypothesis that the incorporation of a verb into a
semantically "empty" noun, wh ich serves to modify the categorial
specification of the former (in the sense discussed in section 3.1.3), is
not a syntactic mechanism available in UG. The subsequent chapters
suggest that there are cases of syntactic nominalization, which do not
entail modification of the categorial specification of the embedded verb.


3.5.1. Against a passive analysis

As mentioned above, Borer (in progress) argues that subjectless event

nominals are passive constructions, thereby deriving the unavailability
of accusative Case in examples like (27). There are, however, good
reasons to believe that the passive analysis of subjectless event
nominals cannot be maintained.
First, Hazout (1990, 1995) reveals important discrepancies between
verbal passives (37a) and their putative nominal counterparts (37b) or
(27). These discrepancies (which I discuss shortly) show that it cannot
be claimed that subjectless event nominals (the so-called nominal
passives) are syntactically derived from verbs that have undergone
syntactic passivization:

(37) a. ha-'ir nehersa 'al-yedey ha-cava.

the-city was+destroyed by the-army
'The city was destroyed by the army'

b. harisat ha-'ir 'al-yedey ha-cava

destruction the-city by the-army
'the city's destruction by the army'

On the one hand, there are deverbal nouns that appear in the putative
passive construction, although their source verb cannot passivize. Thus,
for instance, the verb calax ('crossed') cannot undergo passivization
(38a-b), but its corresponding noun clixa ('crossing') does appear in the
putative passive construction (38c-d):

(38) a. ha-cava calax let ha-te'ala.

the-army crossed ACC the-canal
'The arrny crossed the canal'

b. • ha-te'ala niclexa 'al-yedey ha-cava.

the-canal was+crossed by the-army
'The canal was crossed by the arrny'

c. clixat ha-cava let ha-te'ala

crossing the-army ACC the-canal
'the arrny's crossing of the canal'

d. clixat ha-te'ala 'al-yedey ha-cava

crossing the-canal by the army
'the canal's crossing by the arrny"

And on the other hand, there are verbs that take a prepositional
complement and can nevertheless passivize, but their related noun
cannot do so. Consider, for example, the verb hispi'a ('influenced'). As
noted by Berman (1978), it takes a prepositional complement and not
a direct object (39a), but can nonetheless undergo passivization; the
preposition disappears and its complement becomes the derived subject
(39b). Hazout observes that the corresponding deverbal noun does not
allow a similar operation (39c-d):

(39) a. dan hi~pi'a 'al dina.

Dan influenced on Dina

b. dina hu~pe'a 'al-yedey dan.

Dina was+influenced by Dan

c. haspa'at dan 'al dina

injluence Dan on Dina

d. • haspa'at dina 'al-yedey dan

influence Dina by Dan

Second, in passive sentences by phrase dropping is not constrained;

it can be dropped independently of its interpretation, as is iIIustrated in

(40) a. ha-yeled huka ('al-yedey 'axiv).

the-boy was+beaten (by brother-his)
'The boy was beaten (by his brother)'

b. ha-yeled huka§ ('al-yedey ha-naxa§).

the-boy was+bitten (by the-snake)
'The boy was bitten (by the snake)'

In nominal contexts, in contrast, the Agent can be implicit only if it is

[+human] «41a) versus (41b», which is typical of arbitrary (arb)
interpretation (see Cinque 1988, Rizzi 1986a, Szabolcsi 1992, 1994).
This is completely unexpected under any passivization analysis: 21

(41) a. haka'at ha-yeled ('al-yedey taxi-v) zi'aze'a 'otanu.

beating the-boy (by brother-his) shocked us
'The beating of the boy (by his brother) shocked us'

b. hakdat ha-yeled *('al-yedey ha-naxa§) zi'aze'a 'otanu.

biting the-boy (by the-snake) shocked us
'The biting of the boy (by the snake) shocked us'

Discarding the claim that subjectless event nominals are the output
of a passivization process, I turn to discuss the syntactic status of their
implicit arb Agent, and their inability to take an accusative complement

3.5.2. The implicit Agent

Rizzi (1986a) defines a rule assigning arb interpretation to a 6-role. The

rule can apply in either the lexicalor the syntactic component (in the
spirit of Borer 1984). The question arises as to whether the implicit arb
Agent in subjectless event nominals is syntactically realized as a null
element, or its 6-role is saturated in the lexicon, along lines proposed
by Szabolcsi (1992, 1994) for Hungarian noun phrases.
In Siloni (1990b, 1994a), I elaborated an argument in favor of its
syntactic realization based on data referring to the binding theory,
following Chomsky's (1986a) reasoning on the matter. The argument
proceeds as follows. Consider the distribution of anaphors and
pronouns, as stated by Principles A and B of the binding theory: while
anaphors must be bound in a local domain, pronouns must be free in

that domain, which is the domain of the closest subject. Consider now
the examples in (42):

(42) a. hUj nifga me-horadat-oj be-darga.

he was+offended from-lowering-his in-rank
'He was offended by his demotion'

b. • hUj nifga me-horadat 'acmoj be-darga.

he was+offended from-lowering himself in-rank

The local domain in (42) cannot be the whole clause; if it were, we

would expect (42a) to be excluded, as the pronoun would be bound in
the local domain, and (42b) to be grammatical, as the anaphor would
be bound in this domain. It follows that the local domain must be the
noun phrase containing the anaphor/pronoun. As such, it must contain
a c-commanding subject structurally realized as an empty category. The
only available candidate is the Agent of horada ('Iowering'), which
cannot be coindexed with the subject of the matrix verb, according to
a reasonable interpretation. The judgments in (42) immediately follow. 22
Under this approach then, the implicit subject of event nominals must
be structurally realized. 23
A parallel argument can be constructed regarding non-event nouns,
and in fact Chomsky's (1986a) original discussion deals with concrete
nouns such as story. Advancing the above reasoning, Chomsky argues
that the grammaticality of examples such as (43) follows form the
presence of a structurally realized empty category (PRO, according to
current typologies of empty categories) that functions as the subject of
stories. The matrix verb heard dictates disjointness in reference
between PRO and they, hence a coindexed pronoun can occur in the
noun phrase:

(43) TheYj heard PROj stories about themj.

Recall that we conc1uded with Grimshaw (1990) that non-event

nominals do not take arguments (section 3.1). The representation in (43)
then structurally realizes as PRO the possessor of stories, which is not
an argument but rather a semantic participant. Moreover, according to
the above reasoning, PRO in (43) is obligatory (although it is not areal
argument); in its absence, the binding domain of the pronoun them
would be the whole clause, in which it would be bound in violation of
Principle B. But the obligatory realization of semantic participants
cannot be generalized. For example, as noted by Chomsky (1986a) and

Williams (1985), a parallel PRO in (44) would also have to be distinct

in reference from the subject they, given the interpretation of the matrix
verb, wh ich would erroneously predict the ungrammaticality of the
example. Hence, PRO is impossible here:

(44) TheYj heard (*PROj ) stories about themselvesj.

Furthermore, Williams (1985), arguing against the PRO approach,

shows that PRO in nominal environments would not obey the same
control principles as its verbal counterpart. Indeed, while the Agent of
the infinitival verb fe-nakot ('to clean') in (45a) is obligatorily
controlled by the subject of the matrix clause, the Agent of the noun
nikuy ('cleaning') in (45b) can have an arb interpretation:

(45) a. hem hivtixu la le-nakot bi-tedirut 'et ha-§ati'ax.

they promised to+her to-clean frequently ACC the-carpet
'They promised her to clean the carpet frequently'

b. hem hivtixu la nikuy tadir §el

they promised to+her cleaning(NOuN) frequent of
'They promised her a frequent cleaning of the carpet'

So, the PRO approach can account for binding data of the type
discussed in (42-44), but it is not really straightforward, as just
discussed (see also note 22). The generative literature contains
alternative approaches. Williams (1985) suggests an account of such
binding phenomena that does not require the structural realization of the
external argument. A more recent and thorough account of such data
that does not have recourse to a structurally realized null subject is
given by Reinhart and Reuland (1993). I refer the reader to the different
works, and return to the original question that has led us to this
discussion: why are subjectless event nominals unable to realize an
accusative Theme? The answer to this question will supply some
independent evidence against the structural realization of the implicit
Apriori, two possible solutions come to mind. First, inherent
accusative in Hebrew noun phrases is contingent upon the discharging
of genitive Case (Case dependency). Second, it is contingent upon the
discharging of the external 6-role (thematic dependency). As observed

by Kayne (1984), the external 6-role of nominals can be assigned not

only to DPs, but also to a subset of adjectives, which Giorgi and
Longobardi (1991) label referential adjectives (typically, adjectives
expressing nationality). An event nominal involving a referential
adjective can help us decide whether the dependency at stake is
thematic or Case related, because a referential adjective realizes the
external 6-role without realizing genitive Case. If the dependency is
thematic, a referential adjective should suffice to license inherent
accusative Case. If it is a Case dependency, it should not. As shown by
the contrast between (46) and (47), the presence of a referential
adjective renders inherent accusative Case assignment possible: 24

(46) a. * ha-hafcaca 'etlevanon

the-bombing ACC Lebanon
'the bombing of Lebanon'

b. *ha-§ixzur 'et ha-xoma

the-reconstruction ACC the-wall
'the reconstruction of the wall'

(47) a. ha-hafcaca ha-yisre'elit 'et levanon

the-bombing the-Israeli ACC Lebanon
'the Israeli bombing of Lebanon'

b. ha-§ixzur ha-sini 'et ha-xoma

the-reconstruction the-Chinese ACC the-wall
'the Chinese reconstruction of the wall'

Genitive Case is not discharged in (47). If follows that the dependency

at stake cannot ,be a Case dependency. If so, then inherent accusative
Case is available in Hebrew noun phrases only if the external 6-role has
been discharged. This is reminiscent of Burzio's (1986) generalization,
which states that accusative Case is assigned to the object if and only
if a 6-role is assigned to the subject. Indeed, Ritter (1987, 1988) has
proposed that the descriptive observation emerging from examples like
(46) or (27c) should be traced back to Burzio's generalization. 25
Now, if the implicit Agent were structurally realized as a null
element in subjectless event nominals, their inability to license inherent
accusative Case would be rather mysterious, as the external 6-role
would have to be assigned to the phonetically null subject. Given that
and the preceding discussion, it seems advantageous to adopt the lexical

saturation approach to the implicit Agent of subjectless event nominals.

In other words, if the above reasoning is correct, the understood subject
of event nominals ought to be saturated in the lexicon prior to syntactic
insertion, as suggested by Szabolcsi (1992, 1994) on the basis of
Hungarian data.
Finally, note that a noun that takes an accusative Theme argument
cannot realize an 'al-yedey ('by') phrase (48). This is expected if 'al-
yedey phrases are not arguments, but rather adjuncts associated with an
Agent interpretation, as commonly assumed (see Grimshaw 1990 for a
recent discussion). 'Al-yedey phrases do not bear the external 6-role in
the same way that genitive arguments or referential adjectives do.
Hence, they do not suffice in order for an accusative Theme to be

(48) • ha-harisa let ha-'ir 'al-yedey ha-cava

the-destruction ACC the-city by the-army

In sum, subjectless event nominals are not the output of a

passivization process. They cannot realize an accusative Theme
argument because accusative Case assignment is contingent upon the
discharging of the external 6-role in concert with Burzio's


The question addressed in this chapter has been whether event

interpretation is a property of lexical heads - verbs or deverbal nouns
- or whether it is a verbal property, which amounts to saying that event
nominals contain a verbal projection in syntax. I have argued in favor
of a lexicalist approach on the basis of various considerations.
First, English or French event nominals do not show any verbal
behavior: they cannot assign accusative Case nor can they be modified
by adverbs. Second, Hebrew event nominals, which seem to allow
adverbial modification as weil as accusative Case assignment, cannot
assign structural accusative Case, nor are they modifiable by genuine
Thus, the VP-analysis of event nominals does not receive any
independent empirical support. On the contrary, the above data follow
straightforwardly if event nouns are purely nominal. Furthermore, the
VP-analysis encounters problems of over-generation. It erroneously
predicts that any verb ought to have a corresponding event nominal, and

that any Hebrew event nominal ought to be able to take an accusative

complement if it is derived from a transitive verb.
A lexicalist approach to event nominals allows deverbal nouns to be
specified in the lexicon as to their interpretation (ev~nt/result). If they
have an event interpretation, they are associated with an argument
structure; if they do not, they are result nominals, which do not take
real arguments (Grimshaw 1990). The approach does not expect event
nominals to show verbal properties nor does it predict their
systematically productive formation.
This chapter advances the restrictive hypothesis that a syntactic
process that incorporates V into a semantically "empty" N is not part
of UG. The following chapters define syntactic nominalization as
embedding of a non-tensed verbal projection by the functional head D.
As will be shown in the course of these chapters, when syntactic
nominalization applies, the verbal properties of the construction are
crystal clear, and its productivity is systematic.


Concrete nouns, whether non-deverbal nouns or deverbal result nouns,

designate entities in the world as opposed to events. As is already clear,
nouns that do not express an event do not have an argument structure
to satisfy. Concrete nouns do not entail an event and do not have an
argument structure. Therefore, they do not have specific 6-roles to
discharge. They option~ly take semantic participants, with which they
are in rather loose relations, modifiers (in Grimshaw's 1990 terms).
adjuncts (in Szabolcsi's 1992 terms). The lack of argument structure is
of course echoed in syntactic structure, as already discussed in the
course of the chapter.
Obviously, concrete nouns can occur with no complements:

(49) a. ha-tmuna hidima 'oto.

the-picture impressed him

b. ha-re'ayon 'inyen 'oto.

the-interview interested him

As their lexical representation does not specify a Theme (Patient) 0-

role. they do not license inherent accusative Case, which can only be
assigned to Themes (see seetion 3.1.2):

(50) a. ha-tmuna §el dan (*'et ha-praxim)

the-picture 0/ Dan (ACC the-flowers)

b. ha-re'ayon §el dan (*'et ha-nasi)

the-interview 0/ Dan (ACC the-president)

Moreover, as they do not involve an event structure, they cannot

realize an 'al-yedey ('by') phrase (51a-b), because 'al-yedey ('by')
phrases are contingent upon the presence of an event . In this regard,
they differ from their English counterparts (SIe) and resemble their
French equivalents (51d) (see section 3.1.1):

(51) a. tmuna (*'al-yedey dan)

picture (by Dan)

b. re'ayon (*'al-yedey dan)

interview (by Dan)

c. a picture (by Dan)

d. une peinture (*par Dan)

a picture (by Dan)

Concrete nouns, however, can take more than one genitive noun
phrase. Unfortunately, when more than one genitive DP is involved,
judgements become rather difficult and vary among speakers. I will
nonetheless try to draw the picture in rough lines.
A concrete noun in the construct state can also take a sei ('of')
phrase if the latter can be interpreted as having some claim of
possession over the whole construct. The examples in (52) are
grammatical, as the second genitive DP lends itself to this
interpretation. The examples in (53) are excluded, as the second DP
cannot be interpreted as the possessor of the construct:

(52) a. tmunat ha-praxim §el ha-yeled

picture the-flowers 0/ the-boy

b. targumey ha-odise'a §el ha-sifriya

translations the-Odyssey 0/ the-library

(53) a. * tmunat ha-yeled §el ha-praxim

picture the-boy 0/ the-flowers

b. • targumey ha-sifriya §el ha-'odise'a

translations the-library of the-Odyssey

Recall that this option is not available with event nouns. Reasonably,
events cannot be possessed:

(54) • harisat ha-'ir §el ha-cava

destruction the-city of the-army

Analogously, in French, for instance, concrete nouns avail themselves

of a possessor position that is not available to event nouns (a second de
('of') phrase). Thus, (55a) is grammatical because presentation
('presentation') is the written text that appears in the newspaper, whose
"possessor" is ce journaliste ('this Journalist'). (55b) is impossible as
presentation expresses the event in which the books were presented,
and events cannot be possessed. (55b) becomes grarnmatical if ce
journaliste surfaces as in a par ('by') phrase (55c) (of course the
reading that takes the journalist to be the writer or owner of the books
is perfectly acceptable in both (55a) and (55b), but totally irrelevant

(55) a. La presentation de livres de ce journaliste est toujours

the presentation of books of this journalist is always

b. • La presentation frequente de livres de ce journaliste

the presentation frequent of books of this journalist
m' impressionne.
me(cL) impresses

c. La presentation frequente de livres par ce journaliste

the presentation frequent of books by this journalist
m' impressionne.
me(cL) impresses

Turning back to Hebrew, multiple sei ('of') phrases are (fairly

clumsy but) acceptable with concrete nouns, and completely impossible
with event no uns, as the contrast between (56a-b) and (56c) shows;
again, only concrete nouns can be possessed:

(56) a. ? ha-tmuna §el ha-praxim §el ha-yeled

the-pieture 0/ the-flowers of the-boy

b. ? ha-targumim §el ha-odise'a §el ha-sifriya

the-translations 0/ the-Odyssey 0/ the-library

c. • ha-harisa §el ha-'ir §el ha-cava

the-destruetion 0/ the-city 0/ the-army

In fact, this is not surprising. As they can be possessed, concrete

nouns can take possessors. Reasonably, their possessors are generated
as adjuncts. Much like VP-adjuncts then, possessors are not constrained
with regard to Case the way real arguments are. Arguably, they have
re course to default genitive Case, which is not assigned by the head
noun. Their occurrence thus is not in contradiction with the Biunique
Relation Requierment that qualifies the relationship between Case
assigners and Case assignees. Recall that this requirement was argued
to be responsible for the fact that event nouns can realize only one
genitive argument (chapter 2, section 2.3.4). As expected then,
examples like (56) can even include an additional possessor in an extra
sei phrase (of course, the growing heaviness renders the structure more

(57) a. ? ha-tmuna §el ha-praxim §el ha-yeled §el beyt-ha-sefer

the-pieture 0/ the-flowers 0/ the-boy 0/ the-sehool

In a multiple sei phrase structure of the type in (57), the noun realizes
a Theme-Iike participant (ha-praxim 'tbe flowers'), an Agent-like
participant (ha-yeled 'the-boy'), and an owner (beyt ha-se/er 'the
school'). Following work by Giorgi and Longobardi (1991), Shlonsky
(1988) argues that concrete nouns are hierarchically structured. The
possessor (the owner) is structurally the most prominent element in the
noun phrase. If an Agent-like participant and a Theme-like participant
are present, the former will be higher in the structure than the latter.
Shlonsky's arguments are based on the behavior of anaphors and
pronouns bound by a quantified noun phrase, similarly to the
argumentation developed in chapter 2. I refer the reader to Shlonsky's
work, noting nonetheless that judgments regarding anaphors in concrete
nominals are not clear. Speakers I have consulted did not agree on the
simplest facts. This may suggest that anaphors in concrete nominals (but
not in event nominals, see chapter 2) are logophoric, and thus not
subject to Principle A of the binding theory. Examples involving

logophors are typically subject to variations in judgements (for

discussion, see Reinhart and Reuland 1993). The behavior of pronouns
bound by a quantified noun phrase seems to supply more solid evidence
in favor of the suggested prominence, although here, too, judgements
vary. I leave further discussion of this matter for future research.
Consider now clitic doubling configurations. In chapter 2 (section
2.4), it was observed that in the presence of a sei phrase, the clitic must
be coindexed with it, obligatorily resulting in a clitic doubling
configuration. The examples below further show that in case two sei
phrases are present, the clitic must be coindexed with the first phrase
(58a). Coindexation with the second sei phrase (58b) or with an empty
category in the relevant position (pro) (58c) is inconceivable:

(58) a. ? tmunat-aj sei ha-malkaj sei ha-muze'on

pieture-her 0/ the-queen 0/ the-museum

b. *tmunat-oj sei ha-malka sei ha-muze'onj

pieture-its 0/ the-queen 0/ the-museum

c. * tmunat-aj proj sei klee sei ha-muze'on

pieture-her 0/ Klee 0/ the-museum

This may indicate that there is a structural difference between the first
sei phrase and the phrase(s) following it in (58), (56a-b) or (57). What
could this difference be?
In chapter 2 (section 2.4.3), the obligatory coindexation between the
clitic and the sei phrase was derived from the Biunique Relation
Requirement on Case assignment. The clitic and the doubled noun
phrase ought to be coreferential and share the same semantic role in
order to be each associated with genitive Case. If this is indeed so, it
follows that the doubled noun phrase can never have recourse to default
Case, which is not constrained by the Biunique Relation Requirement
(see discussion above). Or else it would not be forced to be coindexed
with the clitic. This in turn me ans that a sei phrase can receive default
Case within concrete nouns only once a lexical noun phrase has
absorbed the genitive Case of the noun, which forces a doubling
construction in cases like (58a), due to the Biunique Relation
Requirement. I have no explanation as to why the availability of default
genitive Case ought to be conditioned in this way.
Finally, note that some Hebrew speakers accept permutation in the
order of the sei phrases in examples like (56a-b), resulting in (59a-b)
below. According to Shlonsky (1988) such examples are completely

grammatical (as weIl as (56a-b». Other speakers find them quite

marginal (worse than (56a-b», though better with a slight intonational
break between the two sei phrases:

(59) a. ? ha-tmuna §el ha-yeled §el ha-praxim

the-picture 0/ the-boy 0/ the-flowers

b. ?ha-targumim §el ha-sifriya §el ha-odise'a

the-translations 0/ the-library 0/ the-Odyssey

The marginality of these examples for some speakers and the fact that
an intonational break improves them suggest that they result from a
process of reordering, whose exact nature is n'ot directly relevant here.
Shlonsky (1988) notes that the structural hierarchy described earlier is
independent of the linear order of participants. This suggests that the
process of reordering operates at a level that does not affect syntactic
operations. It does not belong to core syntax, and presumably applies
following the branching point to PF (see Giorgi and Longobardi 1991
for the claim that parallel permutations in linear order within the Italian
noun phrase are the result of a stylistic rule). üf course, a parallel
permutation is impossible when the head noun is in the construct state,
as the members of the construct form a phonological unit (chapter 2,
section 2.3.5). Along similar lines, it mayaiso be possible to explain
why examples such as (58b) cannot be derived via linear reordering.
To summarize, concrete nouns do not have an event structure and an
argument structure to satisfy. This is echoed in syntactic structure: they
license neither inherent accusative Case nor an 'al-yedey ('by') phrase.
Concrete nouns can nonetheless appear with more than one genitive DP
(unlike event nominals), as they can take possessors, which avail
themselves of default genitive Case and therefore are not subject to the
Biunique Relation Requirement between Case assigners and Case


1 The subject can have genitive Case either via sei ('of') as in (Ia) or through the
construct state as in (Ib) (see discussion in chapter 2), As this is irrelevant to the
issues discussed here, I will alternate between the two options.

2 The external argument of event nouns does not obligatorily surface, According
to Grimshaw (1990), the external argument of nouns, Iike the external argument
of passives, is suppressed and constitutes an A-adjunct, which is Iicensed by the

argument structure, but is not 6-marked. According to others, its optionality

follows from independent factors, such as the absence of a tense operator in
nominals (see Clark 1990), or the fact that nouns need not have a predication
subject in the sense of Rothstein (1983) (see Giorgi and Longobardi 1991).
Szabolcsi (1992, 1994) argues that in Hungarian event nominals, the unexpressed
subject is PRO, possibly present only in the lexicon. In Siloni (I990b, 1994a),
evidence is presented in favor of the claim that the unexpressed subject of Hebrew
event nouns is syntactically realized as a null element. See section 3.5 for detailed
discussion of subjectIess event nominals.

3 Using the term "event nominal", I refer only to deverbal nouns (Grimshaw's
1990 complex event nominals). In concert with Lebeaux (1986), I think that nouns
Iike trip or event are not event nominals (what Grimshaw calls simple event
nominals) in the same way that the deverbal noun presentation is not an event
nominal in (i). When presentation does not take arguments it cannot express an
event; thus, for instance, it cannot be modified by the modifier frequent (ii), which
requires an event reading. Trip (Hi) and event (iv) do not take arguments and do
not have an event interpretation; analogously, then, they cannot be modified by the
modifier frequent:
(i) A presentation that lasts three hours cannot be good.
(ii) The (*frequent) presentation is desirable.
(iii) The (*frequent) trip tires him.
(iv) The (*frequent) event tires him.
Trip, event, or presentation in (i-ii) are simply concrete or result nouns referring
to entities with temporal extension; yet this does not render them event nominals.

4 Grimshaw (1990) mentions some additional differences between event and result
nominals. For example, event (but not result) nominals do not pluralize (i-ii) and
do not allow the indefinite article (which is not testable in Hebrew, as it has only
adefinite article):
(i) ha-harisot nir'u lemeraxok.
the-destruetions were+seen from+afar
'The destructions were seen from afar'
(ii) * harisot ha-'ir bi-mehirut
destruetions the-city in-quiekness(=quiekly)

S As pointed out by Adriana Belletti (personal communication), a semantic

participant of a result noun (say, sei dan ('of Dan') in (9a» bears inherent Case,
although it does not receive a specific 6-role from its head noun. This may seem
incongruous with the definition of inherent Case, which refers to 6-marking (as
discussed in the previous chapters). However, although this semantic participant
does not receive a strictly 6-interpretation, the range of semantic roles it may have
as a possessor is clearly determined by the head noun itself (and not by a different
element, as in exceptional Case marking configurations). As such, it qualifies as

an inherent Case assignee.

6 An noun taking an accusative complement cannot realize an 'al-yedey ('by')

phrase. The complementary distribution of accusative arguments and 'al-yedey
phrases is discussed in section 3.5.

7 It may be important to note explicitly that Hebrew event nouns can be derived
from stative verbs, which is coherent with our broad use of the term "event" (see
discussion above). Thus, the stative deverbal noun yedi'a ('knowledge') can denote
an event and therefore take an accusative Theme:
(i) yedi'at ha-talmid 'et ha-xomer
knowledge the-pupil Ace the-material

8 It seems that temporal adjuncts like 'etmol ('yesterday') can modify event
nominals in Hebrew (i), as weil as in English (ii) or French (iii):
(i) harisat ha-cava 'et ha-'ir 'etmol
destruction the-army Acc the-city yesterday
(ii) the army's destruction of the city yesterday
(iii) la destruction de la ville hier par I' armee
the destruction 01 the city yesterday by the army
This, however, is not surprising: these adjuncts are not adverbial phrases but rather
noun phrases known as bare noun phrase adverbs (see Larson 1985 for

9 Moreover, note that in French an adverbial pp can intervene between an event

noun and its de ('of') phrase:
(i) la fabrication en häte de chaussures
the manulacture in haste 01 shoes

10 Note that it cannot be argued that in (22d) accusative Case is not available due
to lack of adjacency between the putative embedded V and its object (while in
(22c) this adjacency is somehow rescued via 'et), as in Hebrew an accusative
complement does not have to be adjacent to the verb; a postverbal subject (i) as
weil as an adverb (ii) can intervene between the two:
(i) 'etmol haras ha-cava 'ir 'axat.
yesterday destroyed the-army one city
(ii) ha-cava haras 'etmol 'ir 'axat.
the-army destroyed yesterday one city

11 When the indefinite direct object is heavy, it can appear without the particle 'et:

(i) bikur sar ha-xakla'ut 'olim xadaäim ha-mitgorerim

visit minister the-agriculture comers new that-Iive
be-parvarey yeruäalayim
in-suburbs Jerusalem
'Ihe minister of agriculture's visit to new corners who live in the
suburbs of Jerusalem'
Ihis might suggest that the process of Heavy NP Shift can rescue a construction
otherwise ungrammatical because of lack of Case.
However, in this regard genitive Case behaves differently in Hebrew as weil as
in Romance. As iIIustrated below in Hebrew and French, a heavy noun phrase can
not do without the genitive Case marker:
(ii) tmunat ha-äki'a *(§el) ha-calam §e-'avodot-av
picture the-sunset (of) the-photographer that-works-his
mucagot ba-Iobi
are+displayed in+the-Iobby
'Ihe picture of the sunset by the photographer whose works are
displayed in the lobby'
(iii) la destruction par I' armee *(d') une ville qui etait connue pour
the destruction by the army (of) a city that was known for
ses nombreux musees
its numerous museums

12 Although the judgements in (23c-d) are subtle, most of the speakers I have
consulted find them unacceptable.

13 Interestingly, in the same way that as, comme. and ke- rescue the constructions
in (i-iii), which involve inherent genitive, ke- rescues the configuration in (iv),
which involves accusative Case. Arguably these examples do not involve
exceptional Case marking:
(i) John's consideration of Bill *(as) his mentor
(ii) la nomination de Paul *(comme) directeur
the nomination of Paul (as) director
(iii) ha-minuy §el dan *(ke-) menahel
the-nomination of Dan (as-) director
(iv) minuy-am 'et dan *(ke-) menahel
nomination-their ACC Dan (as-) director

14 Likewise, event nominals do not allow dative pronouns, unlike verbs (for
arguments that indirect objects are datives and not PPs, see Borer 1984). Ihis
shows that the occurrence of dative pronouns, too, is contingent upon the presence
of a verbal element:

(i) • haxzarat-o la 'et ha-kesef

refund-his to+her ACC the-money
As will become clear from the contrast between (29) and (31) in section 3.4.1, for
dative pronouns a verbal environment suffices, but not for accusative pronouns,
which actually require structural accusative Case (for discussion, see Friedemann
and Siloni 1993).
Ur Shlonsky (personal communication) points out that when the pronoun is
stressed, its occurrence in noun phrases does not result in ungrammaticality (ii).
This may suggest that strong (stressed) pronouns can bear inherent Case:
(ii) ? kri'at ha-mefaked rak 'OTO le-seder
calling the-commander only HIM to-order

IS Accusative Case is available in noun phrases in Standard Arabic. No dummy

Case marker is used, but it seems that a special device should be assumed anyhow,
as accusative Case in this language is assigned not only to the Theme argument but
to a larger set of elements Iike certain adverbs and predicates.

16 Apriori, the fact that inherent accusative Case can sometimes be assigned in
passive environments also argues against the passive analysis of (27c) (to be
discussed in seetion 3.5.1). If this Case is sometimes available in verbal passives,
why would it never be available in the putative nominal passives?

17 As is clear from (29b), the pronoun ze is not morphologically marked for Case,
can receive inherent accusative, and can therefore appear in the configurations in
(29) as weil as in noun phrases:
(i) havanat-am 'et ze
understanding-their ACC this

18 Foreign nouns which are assimilated to the system and have corresponding
verbs Can denote an event. Therefore, they are able to assign accusative Case (i-ii)
and are modifiable by adverbial PPs (i) and by the modifier tadir ('frequent') (ii).
As clarified in note (3), event nouns are always deverbal nouns:
(i) minpul ha-cava 'et xayal-av
manipulation the-army Ace soldiers-its
'the army's systematic manipulation of its soldiers'
(ii) ha-minpul ha-tadir ~el ha-cava 'et xayal-av
the-manipulation the-frequent of the-army ACC soldiers-its
'the army's frequent manipulation of its soldiers'

19 Borer (in progress) develops a suggestion along these Iines.


20 Quite systematically, it seems, nouns Iike heres ('destruction'), which belong

to the pattern CeCeC (e stands for a consonant) do not take accusative
complements. Borer (in progress) suggests that this pattern of deverbal nouns is
a passive pattern; it is the morphological output of the incorporation of a
passivized verb into a nominal head. Still, this does not explain the idiosyncratic
behavior of other patterns nIustrated in table (36). On the so-called passive
nominals, see section 3.05.

21 Data of the type in (41) also argue against Grimshaw's (1990) claim that the
external argument of event nominals is suppressed on a par with the external
argument of passives (see note 2). Grimshaw's unified analysis does not explain
the fact that the omission of the external argument is not equally constrained in
nominal environments and in passives, as already observed by Szabolcsi (1992,

22 As shown below, it is not the case that anaphors are independently ruled out in
this configuration:
(i) hu nimna me-'ahavat 'acmo.
he avoided from-loving himself
'He avoided loving himself'
Note that if the genitive DP of the construct state occupies a position outside NP
(as suggested in chapter 2), a problem of structural hierarchy arises: how can an
Agent in SpecNP bind an anaphor higher in the structure.

23 Recall that Agent oriented adverbials and rationale clauses are Iicensed in
Hebrew noun phrases that do not contain a phonetically realized Agent (see section
3.1 and the examples below). According to Roberts (1987), they must be Iicensed
by a structurally realized Agent. But see Williams (19805), Lasnik (1988), and
Grimshaw (1990) for the claim that they are Iicensed by the event:
(i) harisat ha-'ir be-zadon
destruction the city in-maliciousness(=maliciously)
(ii) hafcacat ha-'ir kedey le-havri'ax to§avim
bombing the city in+order to-drive+away inhabitants

24 This was pointed out to me by Reuven Harari. Some speakers also find (i),
which does not seem to include a referential adjective, somewhat better than (46),
but not as good as (47). I have no explanation for this slight improvement:
(i) ?? ha-kri'a ha-mehira 'et ha-sefer
the-reading the-rapid ACC the-book
'the rapid reading of the book'

2' This is inconsistent with Belletti and Rizzi's (1988) proposal that Burzio's
generalization applies to structural accusative only. Moreover, the availability of
inherent accusative Case with passives and unaccusatives (29) should be further

investigated in light of the conclusion drawn in the text. I leave it for future

26 According to Grimshaw (1990), however, subjects of event nominals as weil as

referential adjectives are not true arguments either. The present study, of course,
diverges from Grimshaw on this matter.



Several recent studies have proposed a principled explanation of why

Dis the head ofthe noun phrase (Szabolcsi 1987, 1989, Stowe II 1989,
1991, Longobardi 1994). Somewhat simplified, their insight is that D
determines the referential capacity of the nominal expression, which
consequently is able to act as an argument. In this respect, Szabolcsi
(1987, 1989) argues, D functions on a par with the complementizer of
sentential complements: each renders its complement (NP and IP,
respectively) an expression that is able to bear a 6-role. In this chapter
I discuss a novel facet of D, which supports the functional analogy
between articles and complementizers: I claim that D can act as the
relative complementizer of certain clausal structures.
In some languages, e.g. Hebrew, participial relatives surface headed
by a particle homophonous to the definite article. The particular
properties of these relatives and their crosslinguistic distribution lead
me to analyze this particle as a D that functions like a relative

(1) a. 'i§ ha-kore 'iton ba-rexov hu meragel.

man the-reading newspaper in+the-street is spy
'A man reading a newspaper in the street is a spy'

b. hu kvar ra'a 'et kol ha-sratim ha-mukranim

he already saw ACC all the-movies the-projected
'He has already seen all the movies shown in town'

In other language, e.g. English or French, participial relatives (the so-

called reduced relatives) do not manifest any overt element of this type,
as illustrated below in French and English paraphrases:


(2) a. Un homme lisant un journal dans la rue est un espion.

b. A man reading a newspaper in the street is a spy.

c. Il a deja vu tous les films projetes en ville.

d. He has already seen all the movies shown in town.

Various considerations, however, suggest that they, too, contain a D (a

phonetically empty D) that has the same properties as its Hebrew
Both types of participial relatives contain a verbal projection. But,
typically, DP is a functional projection that is associated with the
nominal system. What is it then which allows DP to contain a verbal
projection? As will become clear in the course of the chapter,
participial relatives are tenseless structures. This is what they share with
noun phrases and this is what renders them appropriate components of
DP. The claim that verbal projections are compatible with D only if
they do not contain a tense operator is further supported in the
subsequent chapter.
The proposal that D can act as a relative complementizer raises
several questions which are addressed in detail in this chapter. What is
the internal structure of the relatives that D heads? What determines the
choice between C and D as the relative complementizer? And what are
the properties of D that enable it to head a clausal structure that
functions as a modifier? The discussion often refers to regular relatives
on the one hand and to argurnental noun phrases on the other hand as
points of comparison, since the construction under investigation is
argued to be at the same time both a relative clause and a DP.
In the first section of this chapter I discuss Hebrew participial
relatives (1). I show that they are tenseless clauses headed by a D that
functions as a relative complementizer. Section 4.2 throws some light
on the different types of D, drawing an explanatory analogy between its
occurrences and the occurrences of C. In section 4.3 I discuss reduced
relatives (2), concentrating mainly on French. I explore their properties
and analyze their internal structure. Finally, section 4.4 offers a unified
analysis; it accounts for the peculiarities of participial relatives and
discusses their common nature.


Consider the synonymous pair in (3). (3a) contains a regular relative

clause, which is headed by the standard complementizer se-. (3b), on
the other hand, contains a relative clause headed by ha-, otherwise the
definite article (recall that Hebrew has only adefinite article, which
shows no inflection). For the sake of clarity, let us call relative clauses
of the type in (3b) semi-relatives (following Siloni 1990a):\

(3) a. hine ha-'is se-xosev rak 'al kesef.

here the-man that-think(BEYNONI} only about money
'Here is the man that thinks only about money'

b. hine ha-'is ha-xosev rak 'al kesef.

here the man the-think(BEYNONI} only about money
'Here is the man that thinks only about money'

The question arises as to whether ha- in semi-relatives is simply an

alternative complementizer to se- (,that'), which appears in C and heads
a regular relative clause. In what folIows, I discuss several salient
distinctions between regular and semi-relatives, showing that se- and
ha- do not have the same distribution. The characteristics of semi-
relatives lead me to suggest that ha- does not occupy C. Rather, it is a
D (just like its homophonous form, the definite article) which heads a
non-tensed clausal structure.

4.1.1. Regular relatives versus sem i-relatives

Consider first the verbal form: it seems identical in both relatives of

(3). However, this form, the so-called beynoni form (which agrees with
its subject in number and gender), is ambiguous between the present
tense and the participle. A reliable test distinguishing between the two
lies in their ability to alternate with past and future forms. When the
beynoni is a tensed form, as in (4a) (kone 'buys'), it can obviously
alternate with other tensed forms. When it is participial, it cannot, as
illustrated by its behavior in the complement of the perception verb in
(4b) (kone 'buying'). In regular relatives the verb need not be in the
beynoni, as shown by the grammaticality of (5a-b). This is natural since
regular relatives are finite clauses. In semi-relatives, on the other hand,
the beynoni is the only acceptable verbal form, as shown by the
ungrammaticality of (5c-d). This implies that semi-relatives are not

finite clauses, but rather participial clauses: 2

(4) a. hu konel kanal yikne lexem.

he buysl boughtl will+buy bread

b. 'ani ro'e 'oto konel *kanal *yikne lexem.

I see him buyingl boughtl will+buy bread

(5) a. hine ha-'i§ §e-xa§av rak 'al kesef.

here the-man that-thought only about money

b. hine ha-'i§ §e-yadov rak 'al kesef.

here the-man that-will+think only about money

c. * hine ha-'i§ ha-xa§av rak 'al kesef.

here the-man the-thought only. about money

d. * hine ha-'i§ ha-yax§ov rak 'al kesef.

here the-man the-will+think only about money

Moreover, semi-relatives are not internally specified for tense in the

same way that regular relatives are. Their understood tense is
determined externally by the context. Consider, for instance, the
contrast between the grammatical semi-relative (6a) and its deviant
regular relative counterpart (6b), when a past tense interpretation is

(6) a. 'ad ha-§ana §e'avra kol ha-klavim ha-no§xim 'et

until the-year last all the-dogs the-biting ACC
ba'aley-hem hayu mumatim.
owners-theirs were killed
'Untiliast year, all the dogs biting their owners were killed'

b. * 'ad ha-§ana §e'avra kol ha-klavim §e-nobim 'et

until the-year last all the-dogs that-bite ACC
ba'aley-hem hayu mumatim.
owners-theirs were killed

Second, it is weH known that Hebrew lacks a copula in present tense.

However, a copular pronoun-like element can optionally appear, as
shown in (7a) (Doron 1983, Rapoport 1987, Rothstein 1995). In relative

clauses of the type in (7b), this copular element becomes obligatory for
reasons which are not relevant here (see Doron 1983). Semi-relatives
do not allow the copular element and therefore cannot be used in this
context (7c):

(7) a. ha-yeled (hu) ~amen.

the-boy (is) fat
'The boy is fat'

b. yeladim ~e-*(hem) ~menim 10 mesaxakim kaduregel.

boys that-(are) fat NEG play football
'Boys that are fat do not play football'

c. * yeladim ha-(hem) ~menim 10 mesaxakim kaduregel.

boys the-(are) fat NEG play football

Third, semi-relatives, unlike regular relatives, can be negated neither

by the standard negation 10 «8a) versus (8b» nor by the present tense
special negation 'eyn (which agrees with the subject when it follows it)
«8c) versus (8d»:

(8) a. hine ha-'i~ ~e-Io xo~ev 'al kesef.

here the-man that-NEG thinks about money
'Here is the man that does not think about money'

b. * hine ha-'i~ ha-Io xo~ev 'al kesef.

here the-man the-NEG thinking about money

c. hine ha-'is se-'eyn-o xosev 'al kesef.

here the-man that-NEG-AGR thinks about money
'Here is the man that does not think about money'

d. • hine ha-'iS ha-'eyn-o xo~ev 'al kesef.

here the-man the-NEG-AGR thinking about money

Fourth, topicalization cannot take place in semi-relatives, it results in

ungrammaticality (9b), although it is admissible in regular relatives (9a)
(the optional appearance of the resumptive pronoun hu is mentioned for
completeness; (9b) is ungrammatical in any case):

(9) a. hine ha-'i§ §e-rak 'al kesef (hu) xo§ev.

here the-man that-only about money (he) thinks

b. • hine ha-'i§ ha-rak 'al kesef (hu) xo§ev.

here the-man the-only about money (he) thinking

A fifth distinction between regular and semi-relatives concerns wh-

elements. This is shown by the behavior of free-relatives (as other
Hebrew relatives can only contain null operators). Whereas se- free
relatives require a wh-element in the specifier position of CP (1 Oa-b),
ha- free relatives never allow wh-elements (IOc); they do allow,
however, free relatives of the type in (IOd):3

(10) a. mi §e-macbi'a ba'ad-am roce §alom.

who that-votes for-them wants peace
'Whoever votes for them wants peace'

b. • §e-macbi'a ba'ad-am roce §alom.

that-votes for-them wants peace

c. • mi ha-macbi'a ba'ad-am roce §alom.

who the-voting for-them wants peace

d. ha-macbi'a ba'ad-am roce §alom.

the-voting for-them wants peace
'Whoever votes for them wants peace'

Finally, a notable distinction between semi-relatives and regular

relatives has to do with the relativized element. While in general the
process of relativization does not discriminate between grammatical
subjects (1Ia) and other grammatical functions, say objects (1Ib), semi-
relatives relativize subjects only. This is illustrated by the contrast
between (llc) and (lId):

(11) a. hine ha-'i§ §e-ma'aric 'et sara.

here the-man that-admires ACC Sara
'Here is the man that admires Sara'

b. hine ha-'i§ §e-sara ma'ariea.

here the-man that-Sara admires
'Here is the man that Sara admires'

e. hine ha-'i§ ha-ma'arie 'et sara.

here the-man the-admiring ACC Sara
'Here is the man admiring Sara'

d. • hine ha-'i§ ha-sara ma'ariea.

here the-man the-Sara admiring

Tbe following ehart summarizes the above distinetions. The behavior

of sem i-relatives regarding the first four entries of the ehart will beeome
elear shortly. For methodological reasons I postpone diseussion of the
last two entries until seetion 4.4:

Regular Relatives Sem i-relatives
Verbal Form finite (+tense) participial (-tense)
Copular pronoun obligatory impossible
Negation possible impossible
Topiealization possible impossible
wh-elements (in FRs) obligatory impossible
Relativized element subject I objeet I ... subject only

Semi-relatives are not tensed elauses and do not allow negation, whieh
suggests that they eontain neither NegP nor TP. Further, if we assume
that elausal reduetion eannot skip an arbitrary layer but ean only
proceed by cutting off external layers (see Rizzi 1993), then the
structure of semi-relatives should not contain sentential layers higher
than TP. If the sentential functional projections are hierarchically
organized as follows (Belletti 1990): ..... AgrsP-(NegP)-TP ...", it beeomes
clear why topicalization and copular pronouns are impossible. First, in
Hebrew a topicalized XP is attached immediately to the left of AgrsP
(9a) (Borer 1984). And second, a copular pronoun is generally
conceived as a phonetic realization of Agrs features (Doron 1983,
Rapoport 1987). If the structure of semi-relatives does not include

layers from TP and up, it does not contain AgrsP' and therefore
disallows topicalization and copular pronouns.
Now, if clausal reduction indeed proceeds in this fashion, ha- in
semi-relatives cannot simply be an alternative complementizer to se-
(,that'), because C does not normally take a reduced clause as its
complement. Trying to maintain that ha- occurs in C would require
either explaining its irregular selection pattern, or accounting for the
peculiar constraints noted above, without assuming that the relevant
functional projections are missing. Both lines of investigation seem to
me ad hoc and deprived of explanatory interest. 4
Rather, I would like to propose that ha- of semi-relatives occupies D,
just like its homophonous form the definite article, and heads a
participial clause, wh ich does not contain the sentential functional
categories from TP and up (on the selectional properties of D, see
section 4.4.1). This proposal immediately derives the non-finite nature
of semi-relatives as weIl as the ban against copular sentences, negation
and topicalization. D of sem i-relatives functions however like a relative
complementizer and involves the type of structural configuration found
in relative clauses.
Importantly, the proposal that ha- is a complementizer-like D
correctly predicts the crosslinguistic distribution of semi-relatives. I
discuss this crosslinguistic prediction in section 4.2, where lexamine
the properties of D that allow it to act as a complementizer. As will
become clear in the course of that section, analyses taking ha- to be
some element distinct from D (say C) do not make this prediction and
cannot explain its validity. Prior to that however, I clarify various
aspects of the proposal: section 4.1.2 discusses the exact structure of
the participial clause and the structural configuration that links it to the
relative head, and section 4.1.3 provides evidence that D of sem i-
relatives indeed behaves in a way analogous to the relative
complementizer se- ('that').

4.1.2. The structure

Given the pseudo-verbal behavior shown by Hebrew event nominals

(see chapter 3 for extensive discussion), it may be worth showing that
the participial form, the beynoni that appears in sem i-relatives, shows
clear verbal properties and therefore entails a verbal projection.
When this form is transitive, its direct object bears structural
accusative Case. (13-16) display several aspects in wh ich this Case
behaves like the ordinary accusative of transitive verbs and not like the

inherent accusative of event nominals. The accusative Case of semi-

relatives can be assigned to subjects of small clauses (13a), and is
compatible with pronominal forms (14a). The accusative argument can
be indefinite and receive Case without the particle 'et (ISa). Moreover,
it can follow a dative argument, unlike the direct object of event
nominals (16a) (while the (a) examples contain semi-relatives, the (b-c)
examples contain verbs and event nominals, respectively):5,6

(13) a. student ha-moce 'et ha-ti'un mesaxne'a

student the-jinding ACC the-argument convincing
yarim 'et yad-o.
will+raise ACC hand-his
'A student finding the argument convincing will raise his

b. ha-student moce 'et ha-ti'un mesaxne'a.

the-student jinds ACC the-argument convincing
'The student finds the argument convincing'

c. * meci'at ha-student 'et ha-ti'un mesaxne'a

jinding(NouN) the-student ACC the-argument convincing

(14) a. student ha-mevin 'oto yarim 'et yad-o.

student the-understanding him will+raise ACC hand-his
'A student understanding hirn will raise his hand'

b. ha-student mevin 'oto.

the-student understands him
'The student understands hirn'

c. * havanat-am 'oto
understanding(NOUN)-their him.

(15) a. ha-soxer ha-kone sfarim higi'a.

the-merchant the-buying books arrived
'The merchant buying books arrived'

b. ha-soxer kone sfarim.

the-merchant buys books
'The merchant buys books'

c. • kniyat ha-soxer sfarim

buying(NouN) the-merchant books

(16) a. ze hu ha-bank ha-maxzir le-§ula let ha-kesef.

this is the-bank the-refunding to-Shula ACC the-money
'rhis is the bank refunding the money to Shula'

b. ha-bank maxzir le-§ula let ha-kesef.

the-bank refunds to-Shula Ace the-money
'The bank refunds the money to Shula'

C. • haxzarat ha-bank le-§ula let ha-kesef

refund the-bank to-Shula ACC the-money

Likewise, the beynoni in semi-relatives is modifiable by genuine

adverbs (17a), just like other verbal forms (17b). As shown in chapter
3, event nominals, which do not contain a verbal projection, cannot be
modified by genuine adverbs (17c):

(17) a. 'ezraxim ha-nohagim maher mesaknim let xayey-hem.

citizens the-driving quickly risk ACC life-their
'Citizens driving quickly risk their life'

b. hem nohagim maher.

they drive quickly

c. • nehigat ha-traktor maher

driving the-tractor quickly

Semi-relatives then clearly contain a verbal projection. If structural

accusative Case is checked in AgroP' they must also contain an Agro
projection. Their verbal form displays participle agreement. If participle
agreement takes place in an agreement projection (Belletti 1990, Kayne
1989a), this projection must be distinct from AgroP' as shown by
Friedemann and Siloni (1993) (see chapter 1, seetion 1.2.3). In Hebrew,
this split is straightforward, as participle agreement is triggered by the

grammatical subject, while accusative Case is checked with the object.

If participle agreement indeed involves syntactic checking, it means that
the subject of sem i-relatives must be syntactically present. This receives
independent empirical support.
First, the (active) participle determines structural accusative Case (see
discussion above), hence it must assign its external 9-role to a null
element, in conformity with Burzio's Generalization (1986). Second,
anaphors are licensed in semi-relatives, which points to the same
direction (according to Principle A of the binding theory):

(18) hine 'iä ha-makir let 'acmo.

here man the-knowing ACC himself
'Here is a man knowing himself'

Third, subject floating quantifiers can occur in semi-relatives (19):

(19) ha-yeladim ha-xoävim kulam 'al kesef

the-children the-thinking al/ about money
'the children all thinking about money'

If Sportiche (1988) and Shlonsky (1990) are right in arguing that the
quantifier is stranded by the raised subject in its basic position, the
occurrence of floating quantifiers in sem i-relatives clearly shows that
they have a structurally realized subject. 7
If so, then semi-relatives contain both an AgroP' where accusative
Case is checked, and a participle agreement projection (AgrpP), where
agreement with the subject is checked. Friedemann and Siloni (1993)
show that in Hebrew Agr P must immediately <lominate AgroP. They
also motivate the choice of weak DP-features on Agro' which explains
why a fuH direct object surfaces in its base position, but strong features
on Agrp ' which forces movement of the verbal element to Agrp (through
Agro) and movement of the subject to SpecAgrpP prior to speil-out
(here nOthing crucial hinges on that). Recall that semi-relatives lack the
higher sentential functional categories. The semi-relative in (20a) then
involves the structure in (20b) by speil-out:

(20) a. ha-na'ar ha-Iomed 'ivrit higi'a.

the-guy the-studying Hebrew arrived
'The guy studying Hebrew arrived'

b. DP

A V'


Let me now consider what kind of an empty category instantiates the

subject and what the structural configuration is between the semi-
relative clause and its head noun.
Both regular relatives and sem i-relatives seem to occupy the same
structural position within the noun phrase whose head they modify. As
shown below, they must follow a sei ('of') phrase (21a-b) (as weIl as
other PPs), while a modifying adjective must precede it (21c). Both
types of relatives constitute the right-most element within the noun
phrase. Arguably, they are adjoined to NP (see Partee 1976 for semantic
motivation of the claim that relative clauses are lower than D):8

(21) a. ha-ben (*ha-gar be-roma) §el dina (ha-gar

the-son (the-Iiving in-Rome) 0/ Dina (the-Iiving
'Dina's son living in Rome'

b. ha-ben (*§e-gar be-roma) §el dina (§e-gar

the-son (that-Iives in-Rome) 0/ Dina (that-Iives
'Dina's son that lives in Rome'

c. ha-ben (ha-yafe) seI dina (*ha-yafe)

the-son (the-beautiful) 0/ Dina (the-beautiful)
'Dina's beautiful son'

I would like to suggest that regular and semi-relatives also involve the
same empty category. In both cases it is a variable bound by a null
operator, which occupies SpecCP in regular relatives and SpecDP in
semi-relatives. Technical problems regarding the formal licensing of the
variable in semi-relatives are postponed until section 4.4 and discussed
in a broader context, once the analysis has been further supported
crosslinguistically. Regular relatives (22a) and semi-relatives (22b) then
involve the same configuration, as schematized in (22c) and (22d),

(22) a. ha-'i!! OPj !!e- t j xo!!ev'al kesef

the-man that- thinks about money

b. ha-'is OPj ha- t j xosev 'al kesef

the-man the- thinking about money

.. .
c. .. ..


t·I Agr'

.. .
d. .. .


A D'


If this suggestion is on the right track. there is an interesting analogy
between the two functional categories CP and DP. D. it is suggested.
functions in (22d) just like C in (22c). It is able to head a relative
clause. that is, to act as a relative complementizer. Since relative
complementizers are, by hypothesis, licensed by being in a Spec-head
configuration with a non-interrogative operator, SpecDP must be
occupied by such an operator in semi-relatives (22d) on a par with
SpecCP in regular relatives (22c).9 If so, then ha- of semi-relatives
(henceforth the relative ha-) should behave more like a relative
complementizer than like its homophonous form. the article. This
prediction is borne out by empirical evidence. as shown below.

4.1.3. The relative ha- versus the article

Reeall that Hebrew has only adefinite artiele, whieh does not infleet.
As the artiele appears with both nouns and adjeetives, a note on the
latter is in order (see also ehapter 2). Adjeetives in Hebrew follow the
noun they modify and agree with it in number, gender and: definiteness.
The relevant facts are illustrated below:

(23) a. ha-'i§ ha-yafe

the-man the-beautiful
'the beautiful man'

b. 'i§ yafe
man beautiful
'a beautiful man'

c. • 'i§ ha-yafe
man the-beautiful

Thus, when the noun is definite, an artiele must surfaee not only with
the noun but also with its modifying adjective (23a). When the noun is
indefinite, the adjective cannot bear an article either (23b-c). Precisely
in this respeet, the relative ha- behaves differently, sinee it is entirely
independent of the (in)definite nature of the noun. (24a) is well-formed,
though the head noun is indefinite (eompare with (23e». In this regard,
the relative ha- patterns just Iike se- (24b): their appearance is not
eontingent upon the definiteness of the noun they modify. On the
contrary, (24c) is ungrammatical precisely because some relative marker
must be present, be it ha- or se-, in order to provide the proper Spec-
head configuration with the relative operator (Hebrew lacks a null
relative complementizer):

(24) a. hine 'i§ ha-xo§ev rak 'al kesef.

here man the-thinlcing only about money
'Here is a man thinking only about money'

b. hine 'i§ §e-xo§ev rak 'al kesef.

here man that-thinks only about money
'Here is a man that thinks only about money'

c. * hine 'i§xo§ev rak 'al kesef.

here man thinks/thinking only about money

Additionally, the relative ha- differs from the definite article since
the latter, but not the former, affects the interpretation of the quantifier
kol. Consider first the effect of the definite article: when the noun is
definite, kol has a collective interpretation, meaning 'all' (25a); when
it is indefinite, kol has a distributive interpretation, meaning 'each'
(25b). Shlonsky (1990) has proposed that the collective kol ('all') is a
functional head Q, which takes adefinite complement, while the
distributive kol ('each') is a Q whose complement is not definite:

(25) a. [QP kol [DP[+defj ha-sefer]]

all the-book

b. [QP kol [DP[-defj sefer ]]

every book

In contrast to the definite article, the relative ha- allows in principle

for the two interpretations, naturally depending on the context. Both
examples of (26) involve a bare quantifier and a semi-relative that
modifies the empty complement of the quantifier. IO In (26a) I choose
the convenient context to force the collective interpretation ('all'), and
in (26b) the distributive interpretation ('each'):

(26) a. kol ha-munax 'al ha-§ulxan §okel yaxad

all the-put on the-table weighs together
20 kilo.
20 kilograms
'All the things put on the table weigh together 20

b. kol ha-macbi'a ba'ad-am xo§ev §e-hu patriyot.

each the-voting for-them thinks that-he patriot
'Everyone voting for them thinks that he is a patriot'

Modulo the appearance of a wh-word in SpecCP (see (10», the same

facts hold when se- ('that') is used, that is, both interpretations are
available (27a-b):

(27) a. kol ma §e-munax 'al ha-§ulxan §okel yaxad

all what that-is+put on the-table weighs together
20 kilo.
20 lcilograms
'All the things that are put on the table weigh together 20

b. kol mi §e-macbi'a ba'ad-am xo§ev §e-hu patriyot.

each who that-votes for-them thinks that-he patriot
'Everyone that votes for them thinks that he is a patriot'

Assuming the distinct selectional properties proposed by Shlonsky

(1990) to distinguish between kol ('all') and kol ('each'), (26a) and
(27a) ought to have the structure represented in (28a,c), respectively,
where the empty complement of Q is forced by the context to be
interpreted as [+definite]. (26b) and (27b), in contrast, ought to have
the structure in (28b,d), respectively, where the empty complement of
Q is interpreted as [--definite] according to the context. The relative ha-,
just like the complementizer se- (,than, does not affect the
interpretation of the quantifier, since it does not head the complement
of Q, rather it heads the restrictive modifier of that complement, be the
latter specified [+det] (28a) or [--det] (28b):

(28) a. [QP kol [oP[+deij ec [oP ha- [AgrpP munax ... ]]] ...
all the- put

b. [QP kol [OP[-dcfl ec [~p ha- [AgrpP macbi'a... ]]] ...

each the- voting

c. [QP kol [OP[+dcij ec [cp ma §e- [AgrsP munax... ]]] ...

all what that- is+put

d. [QP kol [oP[-defl ec [cp mi §e- [AgrsP macbi'a ... ]]] ...
each who that- votes

In conclusion, although ha- of sem i-relatives occupies D, it behaves

more like a relative complementizer than like its homophonous form,
the article. This supports the hypothesis that sem i-relatives involve the
structural configuration typical of relative clauses.


Let me now further develop the analogy outlined above between the
two functional categories CP and DP. Consider first CP. Rizzi (1990)
introduces a feature system which provides a partial specification of the
different kinds of C. Following his proposal, I would like to utilize the
feature [±modifier] (henceforth [±mod]) to characterize two different
occurrences of C that are relevant for the present discussion. 11 C
headirig a relative clause is C[+modJ' since it heads a CP that is a
modifier. C(-mod) then must head a clause which is not a modifier, but
is a sententlai complement, an argumental CP:

(29) a. the thing [cp OPj [q+mod) that] you saw t j ]

b. I know [cp [q-mod) that] you saw it]

(adapted from Rizzi 1990)

(29a) contains a CP which is a relative clause, its head is therefore

marked [+mod]. (29b), on the other hand, involves a 9-marked CP,
hence its head is specified [-mod]. Discussing the functioning of C,
Szabolcsi (1987, 1989) argues that it turns the proposition it heads into
an expression that can act as an argument. Note that this is the function
performed by C[-mod) in our terms.
Turning our attention back to D, let us review its occurrences in
Hebrew. Primarily, it heads the noun phrase. In that case, as observed
by Stowell (1989, 1991) and Longobardi (1994), it determines the
referential capacity of the nominal expression, thus allowing it to occur
in an argument position. As such, it acts with respect to NP on a par
with C[-mod) with regard to AgrsP: both allow the phrase they head to
appear in an argument position. This parallelism has led Szabolcsi (op.
cit.) to label D "the complementizer of the noun phrase". Adopting the
above terminology, D in that occurrence is D[-mod) by analogy with
C[-mod)' as both head 9-marked expressions.
Recall now that D in Hebrew also occurs with modifying adjectives
as an agreement reflex with the noun they modify (23). Here, D does
not head an argumental expression; rather, it heads a modifying
expression. In that respect it paralleis C[+mod), since both head
modifying phrases, and it can therefore be classified as D[+mod) by
analogy. Evidently, languages vary on whether their modifying
adjectives realize such a D. For example, whereas Standard Arabic,
Gulf Arabic and Classical Greek resemble Hebrew on that score (30),
English, French and Italian contrast with it, as they do not allow

adjeetival Ds (31):

(30) a. 'ar-rajulu t-tawiilu

the-man the-tall
'the tall man'
(Standard Arabic)

b. I-bint I-matiina
the-girl the-Iat
'the fat girl'
(Gutf Arabic)

c. to biblion to kalon
the seroll the beautiful
'the beautiful scroll'
(Classical Greek)

(31) a. the (*the) fat man


b. I'homme (*Ie) gros

the man (the) lat

c. Ia ragazza (*Ia) ricca

the girl (the) rieh

Finally, consider the third occurrence of D in Hebrew: D that heads

semi-relatives. This D, the relative D, is c1early D[+mod), just Iike
C[+mod)' as both head relative clauses. Crucially, then, [+mod] is a
feature that the relative D shares with the adjectival D, that is, the one
appearing with modifying adjectives. Though the latter is dependent on
the definite nature of the noun it modifies (mirroring agreement with it),
whereas the former appears regardless of the (in)definite nature of the
noun Oust Iike other relative complementizers), both head modifying
phrases, thus bearing the feature [+mod]:

(32) a. ha-'i§ [DP [D[+mod] ha-] [AP §amen]]

the-man the- lat

b. ha-'i~ [OP OPj [O[+mod1 ha-] [AgrpP t j xo~ev 'al

the-man the- thinking about

The fact that only in semi-relatives is the occurrence of ha- obligatory

independently of the (in)definite nature of the modified noun follows
from the fact that only semi-relatives involve a null operator, which
must be licensed in the appropriate Spec-head configuration. APs do not
contain an operator-variable chain; they are basic predicate constituents
and therefore do not demand an empty category in order to serve as
modifiers (see Rothstein 1991).12
If this classification is on the right track, we expect a certain
crosslinguistic coherence in the manifestation of O[+modJ' That is, all
other things being equal, we predict some consistency in the appearance
of adjectival Os and semi-relatives. Importantly, this prediction is borne
out. Standard Arabic, Gulf Arabic and Classical Greek, which exhibit
adjectival Os (30) also permit sem i-relatives (33), precisely as does
Hebrew. English, French and Italian, on the other hand, do not manifest
adjectival Os (31) and do not allow semi-relatives (34):13

(33) a. 'ar-rajulu l-qaadimu yadan

the-man the-arriving tomorrow
'the man arriving tomorrow'
(Standard Arabic)

b. r-rayyaal l-yaay baacir

the-man the-arriving tomorrow
'the man arriving tomorrow'
(Gulf Arabic)

c. se ten neousan es pedon kara

you the /owering to ground head
'you, who lower your head to the ground'
(Classical Greek)

(34) a. The man (*the) arriving from Paris


b. L' homme (*1') arrivant de Paris

the man (the) arriving from Paris

c. L' uomo (*1') accusato dalla polizia

the man (the) accused by+the police

To conclude, the crosslinguistic correlation between the distribution

of adjectival Os and semi-relatives receives an elegant explanation
under the hypothesis that the relative ha- occupies O. Both phenomena
appear in the same grammatical systems as they are associated with one
single property of the same head (0). Analyses positing the relative ha-
in a different structural position do not predict this crosslinguistic
correlation and cannot explain it.


As shown above, English, French and Italian do not allow sem i-

relatives. However, they do exhibit participial clauses functioning as
restrictive modifiers, just like semi-relatives. These clauses, the so-
called reduced relatives, occupy a position within the noun phrase on
a par with other modifiers. It is important to notice that, though they
are homophonous with participial adjuncts (illustrated below),
semantically they function like relative clauses. As such, they are
semantically independent of the event expressed by the matrix verb to
the extent that noun phrase internal modifiers are. For the sake of
concreteness, let us go through some examples.
Consider first (35b) and (35c) in comparison with (35a). Although
both (35b) and (35c) contain the participial clause vetue de noir
('dressed in black'), in (35b) the participle vetue can be modified by the
adverb generalement ('generally'), whereas in (35c) this modification
results in semantic oddity. This difference follows from the fact that in
(35b) the participial clause appears within the noun phrase, hence the
event it expresses need not coincide with that of the main clause, as is
the case with relative clauses (35a). In (35c), in contrast, this clause is
a sentence-final adjunct, which is attached outside the noun phrase;
hence, it is under the scope of tense and must be simultaneous to the
event expressed by the matrix verb. The participial clause in (35c)
cannot be interpreted as a reduced relative, since the latter cannot be
extraposed (see Williams 1975):

(35) a. La fille qui est generalement vStue de noir

the girl that is generally dressed in black
est arrivee.

b. La fille generalement vStue de noir est arrivee.

the girl generally dressed in black arrived

c. La fille est arrivee (·generalement) vStue de noir.

the girl arrived (generally) dressed in black

AIthough sentence-initial adjuncts seem to be somewhat less strict

regarding simultaneity, they must nevertheless be in a certain semantic
relationship with the matrix clause, which may be best characterized in
terms of causality. Thus, while (36a) and (36b) are well-formed, since
the (reduced) relative only restricts the class of girls and does not have
to be in any particular semantic relationship with the matrix clause,
(36c) is odd because the fact that the girl arrives today in Geneva
cannot be a reason for her being born in Rome (and compare with

(36) a. La fille qui arrive aujourd'hui a Geneve est nee

the girl that arrives today in Geneva was born
a Rome.
in Rome

b. La fille arrivant aujourd'hui a Geneve est nee

the girl a"iving today in Geneva was born
a Rome.
in Rome

c. • Arrivant aujourd'hui a Geneve, la fille est nee

arriving today in Geneva, the girl was born
a Rome.
in Rome

d. Arrivant aujourd'hui a Geneve, la fille pourra

arrlvJng today in Geneva, the girl will be able
voir le film.
to see the movie

Thus, (35b) and (36b) contain participial clauses which function

semantically like relative clauses. Note that only these participial
clauses surface in Hebrew headed by D (37a); adjuncts can never
surface this way, because they are not modifying clauses (37b):14

(37) a. ha-yalda ha-Iove§et borim higi'a.

the-girl the-wearing black arrived
'The girl wearing black arrived'

b. ha-yalda higi'a (*ha-)Iove§et §xorim.

the-girl arrived (the-)wearing black
'The girl arrived wearing black'

Noticeably, reduced relatives share with semi-relatives their most

salient properties. First, their verbal form can be only participial.
Second, they permit relativization of grammatical subjects only;
relativization of other grammatical functions results in
ungrammaticality, as illustrated by the contrast between (38a) and

(38) a. L' homme lisant le journal est un espion.

the man reading the newspaper is a spy

b. * Le journal I' homme lisant est interessant.

the newspaper the man reading is interesting

Third, they do not allow any material normally occurring within CP.
(39a) is ungrammatical because of the appearance of the
complementizer qui (compare with (39b», and the unacceptability of
(39c) is due to the occurrence of a wh-element (compare with (39d»:lS

(39) a. * L' homme qui lisant le journal est un espion.

the man that reading the newspaper is a spy

b. L' homme qui Iit le journal est un espion.

the man that reads the newspaper is a spy

c. * Qui respectant ses amis ne les trompe pas.

who respecting his fr;ends NEG(CL) them(cL) cheat not

d. Qui respecte ses amis ne les trompe pas.

who respects his friends NEG(CL) them(CL) cheat not

I suggest that reduced relatives in essence share the same structural

configuration with semi-relatives, whence their common properties.
Before elaborating this proposal, however, I explore the internal
structure of reduced relatives.

4.3.1. Internal structure

French has both a present participle and a past participle. Both can
occur in reduced relatives, as shown below (40). While the past
participle agrees in number and gender with its grammatical subject
(40a), the present participle shows no agreement at all (40b):16

(40) a. Les voitures repeintes par le garagiste semblent neuves.

the cars repainted by the garagist seem new

b. Les actrices repeignant(·es) leurs voitures sont rares.

the actresses repainting their cars are rare

Moreover, whereas a present participle reduced relative is well-formed

regardless of the argument structure of the verb (41), its past participle
counterpart can only be a passive (42a) or an unaccusative (42b) (its
subject can only be the deep object); hence the ungrammaticality of
(42c) (an unergative verb) and (42d) (a transitive verb):

(41) a. Le policier arrStant toujours des etudiants a ete

the policeman arresting always students was

b. L' etudiant arrivant a Geneve est sud-americain.

the student a"iving in Geneva is South-American

c. L' etudiant telephonant chaque jour a sa mere

the student calling every day to his mother
est parti sans payer.
left without paying

d. L' etudiant connaissant Marie est mon voisin.

the student knowing Marie is my neighbor

(42) a. L' etudiant arrete par la police est mon cousin.

the student arrested by the police is my cousin

b. L' etudiant arrive a Geneve est sud-americain.

the student arrived in Geneva is South-American

c. • L' etudiant telephone a sa mere est parti sans.

the student calling to his mother left without

d. • L' etudiant connu Marie est mon voisin.

the student known Marie is my neighbor

I assume, along lines proposed by Hoekstra (1984), that the past

participle form by itself can neither assign its external 9-role nor realize
an accusative argument. The unacceptability of (42c) and (42d)
immediately folIows. Note that this participle can appear in active
sentences using the auxiliary avoir ('have'). It is arguably avoir which
restores the thematic capacities of the verb and its ability to assign
accusative Case (Hoekstra 1984, Belletti 1990):

(43) L' etudiant a telephone a sa mere.

the student has called to his mother

According to Friedemann and Siloni (1993), French past participles

cannot realize accusative arguments as they do not bear Agro features
and are not associated with AgroP' It is the auxiliary avoir ('have')
which bears the relevant features in complex tenses (AgroP immediately
dominates VPavoir). Further, past participles cannot realize their
external argument by themselves as the latter must undergo thematic
checking with avoir (see the quoted reference). This suggests that past
participle reduced relatives contain an impoverished functional
structure. As will be shown below, diverse empirical evidence supports
this conclusion. I first consider past participle reduced relatives, and
then turn to show that contrary to them, their present participle
counterparts have a sentential structure.
134 CHAPTER 4 Past participle reduced relatives. If there are no specific mies

of adverb movement in VG, the respective ordering of verbal forms and
VP-initial adverbs provides a diagnostic for the positioning of the
former (see Emonds 1978, Pollock 1989). Applied to past participles,
the test shows that they can stay within VP, as they follow VP-initial
adverbs in (44):

(44) a. L' etudiante toujours etonnee par vos reponses

the student always astonished by your replies
est jeune.
is young

b. *L' etudiante etonnee toujours par vos reponses

the student astonished always by your replies
est jeune.
is young

More importantly, past participle reduced relatives do not allow

sentential negation (ne pas). The past participle can neither precede the
negation adverb pas (4Sa) (which is evident if it does not raise (44»
nor follow it (4Sb) (ne, due to its clitic nature, can never follow the

(45) a. * L' etudiant n' arrete pas par la police est mon
the student NEG(CL) arrested not by the police is my

b. * L' etudiant ne pas arrete par la police est mon

the student NEG(CL) not arrested by the police is my

In addition, past participle reduced relatives disallow auxiliaries:

(46) a. * L' etudiant ete arrete par la police est un

the student bePastP arrested by the police is a

b. * L' etudiant ete arrive en retard est un etranger.

the student bepastP arrived late is astranger

The unavailability of auxiliaries and sentential negation seems to

indicate that these reduced relatives contain bare participial phrases,
wh ich do not involve any hierarchically higher projection. In concert
with current approaches (see discussion in section 4.1.2), I assume the
past participial phrase is an AgrpP, whose head contains participle
agreement features.
Finally, it is a known characteristic of French object clitics that they
appear only in sentential structures. If past participle reduced relatives
involve AgrpPs and contain no additional sentential structure (neither
AgroP nor any higher functional projection), they ought to disallow
clitics. This prediction is indeed borne out, strengthening our claim:

(47) a. * L' etudiant y arrive est mon frere.

the student there(cL) arrived is my brother

b. *L' etudiant en rentre est mon frere.

the student from+it(cL) returned is my brother

c. * Le livre lui donne est interessant.

the book to+him(cL) given is interesting

In short, past participle reduced relatives contain an AgrpP and no

further sentential structure. I now turn to examine present participle
reduced relatives, which differ considerably from their past participle
equivalents. Present participle reduced relatives. As is already clear from

the examples in (41), present participle reduced relatives can realize
their external 6-role and take accusative arguments. If accusative Case
is checked in AgroP' present participial clauses contain this projection.
In this regard, they resemble Hebrew semi-relatives (section 4.1.2).
However, as will become clear below, they contain more functional
First, unlike past participles, present participles obligatorily raise to
some higher functional head; they always precede VP-initial adverbs

(48) a. * L' etudiant toujours arrivant en retard est mon frere.

the student always arriving laie is my brolher

b. L' etudiant arrivant toujours en retard est mon frere.

the student arriving always late is my brother

Likewise, they must precede object quantifiers. As observed by Kayne

(1975) (see also Belletti 1990 and Friedemann 1993-94), these
quantifiers have to move out of their base position prior to speil-out
(unless they are stressed). Their landing site is lower than negation (49).
Present participles raise higher than object quantifiers (50):19

(49) Ces etudiants n' ont pas tout compris.

these students NEG(CL) have not everything understood

(50) a. Les etudiants comprenant tout peuvent partir.

the students understanding everything can leave

b. * Lesetudiants tout comprenant peuvent partir.

the students everything understanding can leave

Let us then try to define the landing site hosting the present participle.
Unlike their past participle counterparts, present participle reduced
relatives allow sentential negation (ne pas) (51a). Notice that the
participle must precede the negation adverb pas, as shown by the
ungrammaticality of (51b). If the sentential functional projections are
hierarchically organized as follows (Belletti 1990): " ... AgrsP-(Negp)-
TP ... ", then (51) indicates that the participle has raised higher than
NegP, arguably to Agrs :20

(51) a. Le policier n' arretant pas les etudiants a ete

the policeman NEG(CL) arresting not the students was

b. *Le policier ne pas arretant les etudiants a ete

the policeman NEG(CL) not arresting the students was

If present participles indeed involve a sentential structure, we expect

them to permit clitics and auxiliaries. This prediction turns out to be
correct, as shown in (52) and (53), respectively:21

(52) a. L' etudiant y arrivant est mon frere.

the student there(cL} arriving is my brother

b. L' etudiant en achetant toujours est mon frere.

the student of+it(CL} buying always is my brother

c. L' etudiant le faisant le mieux est mon frere.

the student it(CL} doing the best is my brother

(53) a. Les etudiants ayant manifeste hier sont invites

the students having manifested yesterday are invited
au poste de police.
to the police station

b. Les filles s'etant arr8tees en chemin sont mes soeurs.

the girls being stopped on the way are my sisters

The following chart summarizes the major distinctions between past

participle and present participle reduced relatives; put together, they
indicate that the former involve an AgrpP, while the latter entail an


Present participle Past participle

Position of adverbs following preceding
Sentential negation possible (following) impossible
Auxiliaries possible impossible
Clitics possible impossible Tense. Plausibly, the present participle morphology (the suffix

-ant) entails checking in Agrs; this triggers raising of the verbal form
up to Agrs .22 The fact that only present participial relatives have a
sentential structure raises the question of whether they contrast with
past participial relatives regarding the presence of tense. Interestingly,
both participial relatives can appear in all three tenses (given the
appropriate context), as shown in the following paradigms:

(55) a. Tous les etudiants commemorant les evenements

all the students commemorating the events
ont ete tues.
were lcilled

b. Tous les etudiants commemorant les evenements sont

all the students commemorating the events are
dans la rue.
in the street

c. Tous les etudiants commemorant demain les

all the students commemorating tomorrow the
evenements seront fiches par la police.
events will be put on files by the police

(56) a. Tous les etudiants admis a I' universite

all the students admitted to the university
I'annee passee font preuve de mauvaise volonte.
last year show bad will

b. Tous les etudiants admis a I' universite sont

all the students admitted to the university are
imm6diatement classifi6s.
immediately classified

c. Tous les etudiants admis a I' universite

all the students admitted to the university
I' annee prochaine seront soumis au nouveau reglement.
next year will be subject to the new regulation

This paradigm suggests that, contrary to what one could expect, both
kinds of participial relatives have no internally determined tense and
therefore their understood tense is determined extemally by the context.
Compare, for example, (55a) and (56c) with the entirely unacceptable
regular relatives, (57a) and (57b), respectively:

(57) a. • Tous les etudiants qui commemorent les evenements

all the students that commemorate the events
ont ete tues.
were killed

b. • Tous les etudiants qui ont ete admis a I' universite

all the students that were admitted to the university
I'annee prochaine seront soumis au nouveau reglement.
next year will be subject to the new regulation

Thus, the stru<!tural difference betwet'n present and past participle

reduced relatives does not seem to havt" <my correlate as far as tense in
concemed. Neither kind of participle is restricted to a particular tense,
in contrast with finite verbs and ak~n to Hebrew participles (I
nonetheless keep to the traditional terminology). What the participles do
seem to specify is the aspect of the event. Past participles express
perfectivity: the event has already taken place (it may, however, turn
into astate). Present participles, on the other hand, signal imperfectivity
or rather continuity, progression. This explains the impossibility of
(58a), in which the present participle is forced to a perfective
interpretation, and (58b), where the past participle can be interpreted as
perfective neither with respect to the moment of the utterance nor with
regard to the tense of the matrix:

(58) a. • Tous les etudiants commemorant les evenements sont

all the students commemorating the events are
deja morts.
already dead

b. • L' etudiant arrive demain est mon trere.

the student arrived tomorrow is my brother

(58a) becomes acceptable when the participle is rendered perfective as

in (59a). Notice that the verbal complex in (59a) does not parallel the
passe compose, as shown by the fact that it can be used to indicate an
event in the future (59b):

(59) a. Tous les etudiants ayant commemore les evenements

all the students having commemorated the events
sont dejä morts.
are already dead

b. Tous les etudiants ayant passe les examens ä la fin

all the students having passed the exams in the end
de I' annee seront admis comme doctorants.
0/ the year will be admitted as doctorants

I now turn to discuss the considerable opacity reduced relatives show

with regard to extraction. This leads me to suggest that, in essence, they
share the same structural configuration with regular and semi-relatives.
Two additional arguments point to the same direction.

4.3.2. Against PRO

Following Burzio (1981), Chomsky (1981) and Stowell (1981), Iassume

that the grammatical subject of the participle in reduced relatives must
be syntactically present, as simple tests affirm. Just Iike semi-relatives
(see arguments in section 4.1.2), reduced relatives allow accusative
objects and anaphors, and can contain subject floating quantifiers:

(60) a. L' homme s' examinant toujours dans le

the man himself(cL) examining always in the
miroir est malade.
mirror is siek

b. J' ai vu dix hommes regardant tous la television.

I have seen ten men watehing all the television

Again, one needs to ask which empty category best characterizes this
phonetically null element. In the references cited above, it has been
claimed that PRO is the most natural candidate. In what folIows, I
argue that the subject of reduced relatives cannot be PRO, rather it is
a variable bound by a null operator, as in other relative clauses.
Empirical evidence based, in particular, on subjacency effects supports
this view. Consider the following paradigrn: (61a) and (62a) involve an
infinitival clause, (61b) and (62b) a regular relative, and (61c) and
(62c) a reduced relative:

(61) a. Jean est le seul a parler regulierement aux

Jean is the only (one) to talk regularly to the
old people

b. Jean est le seul qui parle regulierement aux

Jean is the only (one) that talks regu/ar/y to the
old peop/e

c. Jean est le seul parlant regulierement aux

Jean is the only (one) talking regularly to the
old people

(62) a. Jean est le dernier a revenir content de ce

Jean is the last (one) to return happy from this

b. Jean est le dernier qui est revenu content de ce

Jean is the last (one) that returned happy /rom this

c. Jean est le dernier revenu content de ce

Jean is the last (one) returned happy /rom this

While extraction out of the infinitival clause is possible, as shown by

the grammaticality of (63a) and (64a), extraction out of the relative
clause gives rise to subjacency effects, as illustrated in (63b) and (64b).
Interestingly, when extraction takes place out of (both past and present)
reduced relatives, a subjacency violation results as weil «63c) and
(64c». Reduced relatives constitute islands for extraction on a par with
regular relatives. Tense cannot be the pertinent factor, since reduced
relatives, unlike regular relatives, are not tensed clauses (see discussion
above). I suggest that the crucial distinction between the infinitival
clause and the relative clause is that the latter, but not the former,
contains a relative operator in SpecCP, which renders the CP an island
for extraction. Given this reasoning, if PRO were the grammatical
subject of reduced relatives, we would not expect them to form islands.
On the other hand, if reduced relatives involve the same structural
configuration as regular relatives (a variable bound by an operator), we
expect them to constitute islands just like regular relatives:

(63) a. [A qui]j Jean est-il le seul a [cp [PRO

to whom Jean is-he the only (one)
parler regulierement tj ]]?
to talk regularly

b. * [A quilj Jean est-i1 le seul [Cp OPj qui [ t j

to whom Jean is-he the only (one) that
parle regulierement tj ]]
talks re~/arly

c. * [A quilj Jean est-ille seul [ OPj [ tj

to whom Jean is-he the only (one)
parlant regulierement tj 11
talking regularly

(64) a. [D' oU]j Jean est-ille demier a [cp [ PRO

/rom where Jean is-he the last (one)
revenir content tj ]]
to return hal'l'Y

b. * [D' OU]j Jean est-ille demier [cp OPj qui [ t j

/rom where Jean is-he the last (one) that
est revenu content tj ]]
returned hal'l'Y

c. *[D' oulj Jean est-ille demier [ OPj [ tj

/rom where Jean is-he the last (one)
revenu content tj ]]
returned hal'l'Y

If this is indeed correct, reduced relatives ought to contain a CP-like

functional category that accommodates the null operator, forming an
island for extraction. Importantly, parallel facts hold in Italian. Below
is the corresponding ltalian paradigm:23

(65) a. ? [Da dove]j Oianni era l'unico a [cp [ PRO

/rom where Gianni was the only (one)
tomar contento Ijl1
to return hal'l'Y

b. * [Da dove]j Oianni era l'unico [cp OPj che [ t j

/rom where Gianni was the only (one) that
tomava contento tj 11
returned hal'l'Y

c. • [Da dove]j Gianni era l'unico [OPj [ tj

/rom where Gianni was the only (one)
tomato contento tj ]]
returned happy

Moreover, ItaHan seems to suggest another piece of evidence against

PRO and in favor of the operator-variable approach. Consider the
participial construction (the so-called absolute construction) given in
(66a) below. The construction is not passive: the past participie
conosciuta agrees with its direct object me, to which it assigns
accusative Case, and assigns its external 9-role to PRO, as shown by
Belletti (1990). Crucial to the ensuing argument is the fact that it is a
control construction. Notice now that the corresponding reduced relative
is completely ungrammatical (66b). If reduced relatives involved a PRO
subject, there would be no obvious way to derive this bifurcation. Why
could the same participial clause not act as a reduced relative?
However, if the subject of reduced relatives is not PRO, but rather a
bound variable, and if a variable is not licensed in the subject position
of this absolute construction, the contrast between (66a) and (66b)
automatically follows (for discussion of the licensing conditions
imposed on variables see section 4.4.3):

(66) a. PRO Conosciuta me, hai cominciato ad apprezzare

known me, (you) started liking
il mare.
the seaside
(Belletti 1990)

b. • L' uomo conosciuta Maria eintelligente.

the man known Maria is intelligent

Finally, note that reduced relatives cannot be rendered appositive. When

a proper name is involved (as in (67», the participial clause must be
interpreted as an adjunct. Recall that participial adjuncts must be in a
certain semantic relationship with the matrix clause: simultaneity (as
shown in (35c» or causality (as in (36d) and (67a». In (67b) such a
relationship does not hold, which excludes the adjunct interpretation;
the sentence is ungrammatical, since the reduced relative reading is not

(67) a. Jean, arrivant aujourd'hui a Geneve, pourra

Jean, arriving today in Geneva, will be able
voir le film.
to see the movie

b. • Jean, arrivant aujourd'hui a Geneve, est ne a Rome.

Jean, arriving today in Geneva, was born in Rome

If both adjuncts and reduced relatives had PRO as their subject, the fact
that the occurrence of a proper name eliminates the reduced relative
reading only would be rather puzzling. On the contrary, if adjuncts and
reduced relatives do not share the same structural configuration, it
constitutes a reasonable starting point for an account. It is weil known
that appositive relatives show some particular characteristics. Thus, in
English for instance, while object restrictive relatives may have a
phonetically empty CP (The man I saw), their appositive counterparts
must contain a wh-element in SpecCP (John, ·(whom) I saw). This may
be related to the fact that appositive relatives are attached higher in the
structure than restrictive relatives (for a detailed discussion, see Emonds
1979). Suppose now that for some reason reduced relatives cannot be
attached higher in the structure; it follows that they cannot function as
appositive relatives. Note that this kind of explanation would not be
possible if reduced relatives shared the same structural configuration
with adjuncts, as the latter are attached higher in the structure.24
In sum, various considerations suggest that the grammatical subject
of reduced relatives is not PRO, but rather a variable bound by a null
operator. An appropriate functional category must therefore be present
to host the relative operator.


4.4.1. Selectional properties 0/ D

The previous section concluded that reduced relatives are headed by a
functional cat~gory that accommodates the null operator. The question
arises as to whether that functional category is CP, like in regular
relatives, or DP, as in semi-relatives. Before deciding the issue, let us
summarize the main characteristics of reduced relatives in comparison
with the properties of regular and semi-relatives:

Reduced relatives Semi- Regular
Present Past relatives relatives

Clausal compl. AgrsP AgrpP AgrpP AgrsP

Tense +
Wh-word (in FRs) impossible impossible impossible obligatory
Relativized EI. subject subject subject subj/obj/...

I first investigate the relevance of tense for the type of functional

category used in relative clauses. Sections 4.4.2 and 4.4.3 are devoted
to discussion of the two last entries of the table. The account of the
'only subject' constraint is based on a suggestion regarding the
licensing conditions imposed on variables.
Stowell (1982) argues for a correlation between the existence of a CP
level and the presence of a tense operator in the clause. 2s Elaborating
on that, I would like to suggest that what determines the choice between
CP and DP is the presence or absence of a tense operator. Relative
clauses that have a uniform internally determined tense are headed by
C; thus, the complementizer of regular relatives is C. In contrast,
relative clauses that have no internally determined tense are headed by
D. Reduced relatives, on a par with semi-relatives, do not have a fixed
time frame; it is therefore D which functions as their relative
complementizer, be its complement AgrpP or AgrsP (see chart (68) and
note 27). What discriminates then between semi-relatives and reduced
relatives is simply the fact that in the former, the complementizer-like
D (D[+mod]) is phonetically realized, while in the latter it is phonetically
null, as illustrated below:26

(69) a. Un homme [DP OPj [D ][AgrsP t j lisantun journal

a man reading a newspaper
dans la rue]] ...
in the street

b. 'i§ [OP OPj [0 ha- ][AgrpP t j kore 'iton .

man the- reading newspaper
ba-rexov ]] ...
'A man reading a news paper in the street. ..'

c. Les films [OP 0Pi [0 ][Agrpp t j projetes en ville ]] ...

the movies shown in town

d. ha-sratim [OP OPj [0 ha- ][AgrpP t j mukranim

the movies the- shown
ba-'ir ]] ...
'The movies shown in town .. .'

While C is associated with tensed clauses, D is associated with

tenseless clauses, that is, participial clauses (infinitival clauses, as
shown by Stowe11 1982, contain a tense operator and a CP level).
Typically, DP is a functional projection that is related to the nominal
system; NP (or an agreement projection related to it; see chapter 2) is
the standard complement of D. What is it then which allows DP to
contain a verbal projection in the case of participial relatives? The
common denominator between noun phrases and participial clauses is
their non-tensed nature. Suppose the set of legitimate complements of
D includes non-tensed phrases. If follows that participles, which are
verbal elements that do not bear tense features, constitute appropriate
complements of D, while tensed verbs do not. 27
Moreover, this is certainly not the only case in which DP appears to
contain a verbal projection. Presumably the same selection pattern is
found in English verbal gerunds (see Abney 1987 and the many
references cited there) and in the so-called Italian nominalized
infinitives, e.g. il leggere questo libro ('the to read this book') (see
Salvi 1983, Zucchi 1989, Bottari 1990, Siloni 1994b, among others).
All these cases may be treated along similar lines, namely, by assuming
that the relevant verbal forms (participles, Italian nominalized infinitives
and English gerunds) do not bear tense features. In the subsequent
chapter, I discuss Hebrew gerunds, claiming that they constitute another
manifestation of the same selection pattern. 28
The present approach expects reduced relatives to pattern with semi-
relatives, as they share the same structural configuration. I now turn to
explain their common behavior.

4.4.2. The ban against wh-elements

Given the absence of CP, the fact that wh-elements can appear neither
in sem i-relatives nor in reduced relatives (see chart 68) may seem
trivial, as SpecCP is not available. However, the question still arises as
to why they cannot appear in the specifier position of DP. Suppose
now, along lines proposed by Rizzi (1990) for certain relative
complementizers, that D[+mod] is inherently specified [-wh]. Participial
relatives containing wh-elements can then be ruled out as involving
inconsistent feature specifications of specifier and head, as shown

(70) a. * [op mi[+wh] lOt-wh] ha- ] macbi'a ba'ad-am ] roce

who the- voting for-them wants

b. * [oP qui[+wh] [O[-wh] ] respectant ses amis] ne

who respecting his friends NEG(CL)
les trompe pas.
them(cL) cheat not

Suppose further, following Rizzi, that null operators need not be

specified with respect to the feature [±wh]. This then is what enables
them to appear in participial relatives. 29
Why should D[+mod] be inherently marked [-wh]? To answer this, let
me briefly consider what it means to be marked [+wh]. C[+wh]' that is,
C whose specifier position contains a wh-element, typically implies that
its maximal projection, the clause, is an interrogative (71a) or a relative
c1ause (71b):

(71) a. [cp [oP Who]j [C[+wh] ] did John meet t j ] ?

b. the man [cP [oP who]j [C[+wh] ] John met t j ] ...

What does D[+wh] indicate regarding its maximal projection? To be more

precise, what is the nature of DPs containing a wh-element in their
specifier position, as is the case, for example, in (72) (see Hendrick
1990)7 Unlike a wh-element in SpecCP (wh ich turns the CP into an
interrogative or a relative c1ause), a wh-element in SpecDP turns the DP
into an operator. In (72) the DP how big a hause, whose head is

specified [+wh] (its specifier contains the wh-phrase how big) functions
as an operator, which moves to a scope position in the course of the

(72) [cP [oP How big [O[+wh] a] house]j [c ] did John see t j ] ?

Reasonably, a maximal projection cannot be simultaneously an operator

and a relative clause (notice that in (71 b) it is not the same projection
that fulfills the two functions; the DP who is the operator, while the CP
who John met is the relative clause). 1fthat is so, it follows that D[+mod]
can never be marked [+wh], as the same DP cannot serve
simultaneously as both an operator and a relative cIause. Simply, then,
as far as 0 is concemed, [+wh] and [+mod] are incompatible features,
which predicts that wh-elements in principle can never appear in
participial relatives.

4.4.3. The variable

This section examines the licensing of the variable in the subject

position of participial relatives. The approach adopted supplies a
straightforward account of the fact that these relatives relativize subjects
only (see chart 68). After a brief digression summarizing the long
controversy regarding the licensing conditions imposed on variables, I
sketch a proposal which relies on previous analyses and is of
considerable interest as far as participial relatives are concemed. I then
turn to the 'only subject' restriction.
Argurnental variables have generally been assumed to require Case
(Chomsky 1981, 1986a). As far as participial relatives are concemed,
this is problematic, since participles do not seem to be able to assign
Case to, their subject. However, there have been some important
attempts in the literature to eliminate or modify the Case requirement
imposed on variables, for theoretical and empirical reasons.
Theoretically, if arguments must be assigned Case in order to be visible
for 9-marking (according to the Visibility Principle, which states that
a chain must contain a Case position in order to be visible for 9-
marking (Chomsky 1986a», PRO and lexical expletive,s constitute
unclear exceptions. The former because it should be vfsible for 9-
marking, but it is not assigned Case, and the latter because it demands
Case although it is not assigned any 9-role. Empirically, several
configurations have been mentioned in which lexical noun phrases are
illicit due to lack of Case, whereas variables are entirely legitimate. 30

Consequently, two main approaches have been advanced. On the one

hand, it has been argued that only and all lexical noun phrases require
Case (Borer 1981, May 1981, Bouchard 1984); hence, only when the
operator is a phonetically realized element should Case be assigned to
the chain. On the other hand, it has been claimed that the Case
requirement imposed on variables should be somewhat loosened (see
Epstein 1987, Shlonsky 1987, and Siloni 1991b for specific
suggestions). I will outline here a proposal in the spirit of the latter
If some version of the Visibility Principle is to be taken seriously,
lexical expletives and PRO constitute two anomalies, which should be
accounted for. Recent work has argued that these cases do fall under
the visibility approach. Consider expletives first. Chomsky (1986a,
1993) has suggested that the noun phrase associated with the expletive
(the associate) raises to replace (adjoin to) the expletive at LF.
Following this proposal, Chomsky and Lasnik (1993) have argued that
the associate raised to the expletive position at LF would be Caseless
and therefore illicit, if the expletive were in a Caseless position. The
requirement that expletives have Case is thus explained in a principled
fashion. Next, consider PRO. Chomsky and Lasnik (1993) propose that
PRO, which they regard as a minimal noun phrase argument lacking
independent phonetic and referential properties, can bear null Case.
This Case is the realization of a Spec-head relation with the head
[Agrs+T] when it lacks tense and agreement specifications. The
assumption is that this inflectional head in infinitivals does not contain
tense nor agreement specifications. Note that this is not in contradiction
with Stowell's (1982) claim (adopted in section 4.4.1) that infinitivals
do contain a tense operator; simply, this operator has no inherent tense
reference (specification). Likewise, it is not the case that PRO does not
have cj)-features, but rather that it does not check them in a Spec-head
configuration, but through control or via arbitrary interpretation. If this
approach is on the right track, PRO is no more Caseless.
Interestingly, this approach to PRO paves the way toward a better
understanding of the distribution of variables. I would like to suggest
that variables, insofar as they share with PRO the lack of phonetic
realization, can also do with a sort of null Case - a Case determined by
lack of tense specifications; however, unlike PRO, they must have their
cj)-features checked in a local domain, more precisely, in a Spec-head
configuration. Hence, unlike PRO, they are usually not licensed in the
subject position of infinitivals, where cj)-features are not specified:

(73) • the man OPj that you tried t j to win


Let us now turn our attention back to participial relatives. Given the
proposal advanced above, variables can occupy the subject position of
participial clauses, since they are licensed in a Spec-head configuration
with an inflectional head (an Agr-head) that lacks tense features, but
contains ct»-features. 31 Importantly, if this is correct, the fact that
participial relatives can relativize grammatical subjects only (see table
68) is automatically derived. Relativizing another grammatical function
would entail lexical realization of the su~ect, which, in turn, will be
impossible due to lack of standard Case. 3
Now, if the ungrammaticality of non-subject participial relatives is
simply due to lack of standard Case, it is expected that if a language
has a special mechanism which enables it to assign Case to a lexical
noun phrase in that position, it should manifest non-subject participial
relatives. This prediction seems to be correct. Standard Arabic is an
example of a language of this type. As in this language, the subject of
the participle has recourse to nominative Case (the specific mechanism
enabling this is not of direct relevance here), non-subject participial
relatives are possible:

(74) marartu bi-r-rajuli l-maqtuulati 'ummu-hu.

met(I) at-the-man the-killed mother(NoM}-his
'I met the man whose mother is killed'

To sum up, 1 have argued that the Case requirement imposed on

variables should be restated: variables can do with null Case if they can
check their ct»-features in a Spec-head configuration. This makes them
appropriate subjects of participial relatives and explains why in general
the latter relativize subjects only. The link between the availability of
standard Case and the type of relativized element gets further support
by Standard Arabic data. The data show that non-subject participial
relatives are not apriori excluded, rather, they are generally impossible
for Case reasons.


The basic empirical observation of this chapter has been that in some
languages, for instance Hebrew, participial relatives surface headed by
0, whereas in other languages, e.g. French or English, phonetic',
evidence does not suggest its presence.
Despite the extensive research recently devoted to the functional
category DP, its occurrence in participial relatives has passed rather

unnoticed. I have proposed that participial relatives involve an operator-

variable chain, the scope position of the operator being the specifier of
DP. This elucidates the unexpected appearance of DP in this
environment, and requires its presence regardless of its phonetic
The chapter draws a functional parallelism between the two
categories CP and DP, and claims that the presence versus absence of
tense is the crucial factor determining the choice between the two.
Whereas C is the complementizer of tensed clauses, D qualifies as the
complementizer of non-tensed phrases. Participial clauses are not
tensed; this is what they have in common with noun phrases, and this
is what makes them appropriate complements of D. The subsequent
chapter further corroborates this claim, investigating the case of Hebrew
Finally, a rather old debate in linguistic literature regarding the
relevance of Case for the licensing of variables has received some
attention. It has been proposed that variables can receive a Case
different from the one assigned to lexical noun phrases. This explains
their occurrence in certain structural configurations that do not allow
lexical noun phrases, like the subject position of participial relatives.


1 For discussion of the beynoni form, see section 4.1.1.

2 The question whether the beynoni form is really ambiguous, or always

participial, appearing with a zero auxiliary in present tense (for a proposal along
these lines, see Shlonsky to appear), is irrelevant for the issue in question. In both
cases, the fact that it cannot alternate with past and future forms shows that the
clause is not finite.

3 Following Oroos and van Riemsdijk (1979) and Borer (1984), I assume that in
free relatives, the wh-element is in the CP-Iayer, and does not constitute the head
of the relative.

4 In the same vein, proposing that ha- is a special realization of T or Agrs does
not seem advantageous. First, it is not independently motivated. Second, although
it may suggest an explanation for the ban against tensed forms or copular
pronouns, it would faH to explain either why negation is excluded (if indeed T
licenses sentential negation (Laka 1990 and Zanuttini 1990», or why topicalization
is impossible (if indeed topics are adjoined to Agr.P in Hebrew).

S The beynoni form is often used as a noun or an adjective. However, when it

appears as a noun, it cannot take an accusative complement (ii), unlike its verbal
equivalent in semi-relatives (i):
(i) studentim ha-konim sfarim kol §avu'a...
students the-buying books every week
'Students buying books every week ... '
(ii) konim (*sfarim) nixnesu la-xanut.
customers (books) entered to+the-store
And when it occurs as an adjective it cannot take an 'al-yedey ('by') phrase (iv),
unlike the beynoni in semi-relatives (iii) (the beynoni form in (iv) is clearly an
adjective as verbal forms cannot cooccur with the copular pronoun in present
(iii) ha-xeder ha-mesudar 'al-yedey dan ...
the-room the-arranged by Dan
'The room arranged by Dan ... '
(iv) ha-xeder hu mesudar (* 'al-yedey dan).
the-room is arranged (by Dan)
Moreover, the beynoni form in sem i-relatives has necessarily the meaning of the
cqrresponding verbal form, while the related adjective may undergo semantic drift.
Consider, for instance, the difference between (v) and (vi) (the corresponding
active verbal form means 'internalize'):
(v) rega§ot ha-mufnamim le-'itim txufot hem nose ha-kenes.
sentiments the-internalized often are subject the-congress
'Sentiments that are often internalized are the subject of the congress'
(vi) yeled mufnam
boy introverted
'an introverted boy'

6 In addition, the beynoni in semi-relatives can realize its accusative object as a

suffixal clitic, just like other verbal forms (see chapter 2, section 2.4.2):
(i) yadid ha-makir-eni §anim rabot...
[riend the-knowing-me years numerous
'A friend knowing me for numerous years .. .'

7 Any analysis positing the relative head itself (say 'ezraxim in (i» as the
grammatical subject of the semi-relative c1ause should be immediately discarded.
Clearly, the structure in (i) does not properly reflect 9-assignment: the matrix verb
rocim ('want') does not assign its external S-role to DP t , whose head is ha-
('the'), but rather to DP2 , 'ezraxim ('citizens'), of which ha-macbi'im ba-'adam
('the-voting for them ') is a restrictive modifier:

(i) [AgrsP [OPI [OP2 'ezraxim] [0 ha-] macbi'im ba-'adaln) [Agrs' rocim
citizens the- voting for-them want
§alom ]]
'Citizens voting for them want peace'

In addition, if semi-relatives had the structure represented in (i), we would not

expect them to become appositive relatives when their relative head is a proper

(ii) dan, ha-magi'a hayom le-geneva, nolad be-roma.

Dan, the-arriving today to-Geneva, was+born in-Rome
'Dan, who arrives today in Geneva, was born in Rome'

8 Right-adjunction is reinterpreted in terms of leftward movement under Kayne's

(1994) approach to phrase structure. This is irrelevant here.

9 On different grounds, Tellier (1991) also argues that the specifier position of DP
can host a null operator; her analysis concerns cases like (i):
(i) Un gar\!on dont la maigreur t [vp disait [op Op les longues
A boy whose thinness told the long
souffrances e]]
See also Hendrick (1990), who claims that in cases like (ii) wh-movement applies
to SpecDP:
(ii) [oP How tall a man] did Jane see?

10 In what folIows, I will briefly sketch the structure of these particular free
relatives. Further questions concerning their structure will not be pursued here.

11 Rizzi (1990) does not use the term "modifier", but "predicative". I have
changed the term to prevent confusion, since the cases at stake are cases of
modification and not of predication in general.

12 More precisely, it should also be explained why ha- cannot appear with
modifying adjectives when the modified noun is indefinite. This distinction
between adjectives and semi-relatives may be related to the fact that the former are
structurally closer to the head they modify than the latter (see (21».

13 Gulf Arabic examples (30b), (33b) are from Qafisheh (1977); Classical Greek
(33c) is Antigone 441. Thanks to Ur Shlonsky for helping me with the Standard
Arabic data.

14 Adjectives behave the same way; only modifying adjectives surface headed by
(i) ha-yalda ha-xiveret higi'a.
the-girl the-pale arrived
'The pale girl arrived'
(ii) ha-yalda higi'a (·ha-)xiveret.
the-girl arrived (the-)pale
'The girl arrived pale'

IS Only in free relatives can French subject relatives realize a wh-element.

At first glance, qui in (39d) may appear ambiguous between the wh-element and
the complementizer. However, the contrast between (i) and (ii) clearly shows that
qui here is a wh-element, because the complementizer qui cannot occur when the
object is relativized (iii); que is then the appropriate complementizer (iv) (compare
with (ii»:
(i) ? Qui chacun admire n' est pas un homme de valeur.
Who everyone admires NEG(CL) is not a man 01 value
(ii) • Que chacun admire n' est pas un homme de valeur.
That everyone admires NEG(CL) is not a man 01 value
(iii) • L' acteur qui chacun admire n' est pas un homme de
The actor that everyone admires NEG(CL) is not a man 01
(iv) L' acteur que chacun admire n' est pas un homme de
The actor that everyone admires NEG(CL) is not a man 01

16 The same form can also appear as an adjective. However, when the form is
adjectival, it does show number and gender agreement. The form appearing in
reduced relatives is clearly verbal.

17 Past participle reduced relatives can be negated by the constituent negation pas,
as illustrated in (i). As is often the case when constituent negation is involved, the
sentence is more natural when the negation is emphatic:
(i) L' etudiant pas arrate par la police est mon cousin.
The student not arrested by the police is my cousin
Hebrew semi-relatives do not allow constituent negation. I put aside this particular
difference between Hebrew and French.

18 It has already been shown by Pollock (1989) that French present participles in
sentence-initial modifiers pattern with finite verbs as far as verb raising is

19 Past participle reduced relatives cannot realize object quantifiers, as their verbal
from is either unaccusative or in the passive diathesis.

20 Sentential structures are not negated by constituent negation; hence, as expected,

present participial relatives do not permit constituent negation via pas:

(i) • Le policier pas arretant les etudiants a ete Iicencie.

The police man not arresting the students was fired

21 Unaccusative and passive verbs, that is, precisely the verbs appearing in past
participial relatives, cannot always cooccur with a present participle auxiliary in
reduced relatives for reasons which are unclear to me:
(i) ? L' etudiant etant arrete par la police est etranger.
The student being arrested by the police is a foreigner
(ii) L' etudiant ayant ete arrete par la police est etranger.
The student having been arrested by the police is a foreigner
Note also that in this respect English reduced relatives behave differently, as
shown below:
(iii) The man being hurt is a foreigner.
(iv) ? The man having read this book is my brother.

22 This suggestion may give insight into the contrast between the following
English and French paraphrases:
(i) The policeman is arresting John
(ii) • Le policier est arretant Jean
While in French the present participle morpheme is a realization of Agr. and
therefore cannot be selected by an auxiliary, in English it is associated with a
lower functional category and is compatible with the auxiliary.
Note that if the French present particiiple morphology is associated with Agr.,
there is no reason to assume that present participles involve an additional
agreement projection (AgtyP). When they contain complex verbal forms (see (53»,
it is the past participle wh ich is immediately dominated by AgrpP.

23 Thanks to Adriana Belletti for the Italian data in this section.

The test is not applicable in Hebrew, as sentences Iike (61a,c) are marginal from
the outset. As far as English is concerned, citing examples (i-iii), Burzio (1981)
reaches the opposite conclusion, namely, that reduced relatives are control
constructions just Iike infinitival clauses:
(i) The girl that John was the last one to see t

(ii) • The girl that lohn was the last one who saw t
(iii) The girl that lohn was the last one seen with t
(Burzio 1981)
However, as noted by Chomsky (1982) (who cites Adriana Belletti, personal
communication), noun phrases can be extracted more easily than PPs out of certain
islands (see also Cinque 1990). Indeed, subjacency effects do arise when a pp is
extracted (compare (iv) with (iii», and also when the present participle (v) is used:
(iv) 17 With whom was lohn the last one seen t 7
(v) 7· Who was lohn the last one discussing this matter with t 7

24 Roughly speaking, it may be observed that in some languages relatives may be

attached higher in the structure (that is, be rendered appositive relatives) only if
they are introduced by a phonetically realized element. Thus, English appositive
relatives cannot be introduced by a phonetically empty CP; semi-relatives, which
phonetically realize their 0, can function as appositive relatives (see note 7) and
reduced relatives cannot do so. I will not discuss this issue further here, because
it necessitates an extensive investigation into the nature of appositive relatives.

25 For more discussion of the relation between C and tense, see En~ (1987),
Gueron and Hoekstra (1989), Gueron (1993).

26 If D[+modl in French, English or Italian is phonetically null, one may wonder

whether it appears with modifying adjectives as weil (like its Hebrew counterpart).
I do not see what the immediate empirical consequences of this claim would be.
In this context, it is also important to recall that although the complementizer-like
o and the adjectival 0 significantly share some common nature (the feature
[+mod]), they do differ in some regards as discussed in section 4.1.3 and 4.2.
Note in addition that isolated occurrences of what can be argued to be an overt
adjectival D[+modl can be found in French:
(i) La fille la plus belle
the girl the most beautiful

27 French past participial relatives and Hebrew semi-relatives are AgrpPs; the lack
of TP is therefore straightforward. In contrast, French present participial relatives
involve an Agr.P, whose head normally selects TP as its complement (Belletti
1990, Chomsky 1991). In order to reconcile that with their non-tensed nature, two
possibilities come to mind: either an inert TP is present, whose head does not bear
tense features, or the projection is missing all together (wh ich would mean that
c1ausal reduction can skip a functional layer, contra Rizzi 1993 (see also section
4.1.1 Note that if the availability of sentential negation depends on the presence
of TP (Laka 1990 and Zanuttini 1990), the latter approach cannot be maintained,
as French present participial relatives do manifest sentential negation (section

28 For a somewbat different discussion of this issue, see Grimshaw (1991), who
proposes that verbal elements that appear as complements of 0 are unspecified for
the feature [nominal], unlike other verbal elements which are marked [-nominal].
Defining the relevant distinction using the feature [nominal] seems to me less
explanatory than defining it in terms of tense.

29 Recall that (70a) becomes grammatical if the wh-word is replaced by a null

operator (IOd). Although the occurrence of a null operator in free relatives is not
trivial, it is certainly not unique: regular free relatives containing a null operator
are well-formed in Biblical Hebrew (and also in Palestinian Arabic, according to
Shlonsky 1991). French never allows null operators in free relatives. This
difference will not be further discussed here.

30 Some configurations which iIIustrate the different distribution of variables and

lexical noun phrases are Iisted below. In Semitic a variable can occupy the doubled
position of cIitic doubIing constructions (i), while a lexical noun phrase requires
a Case marker in that position (ii) (see chapter 2 and Borer 1981, 1984). In
English, a variable can appear in the subject position of exceptional Case marking
constructions even if it is not adjacent to the verb (iii); a lexical noun phrase
cannot occupy this position (iv) (Chomsky and Lasnik 1977). And in French a
variable (v), but not a lexical noun phrase (vi), can occupy the subject position of
certain non-finite sentential complements (Kayne 1980):
(i) mij §e-'axot-oj tj 'imca kelev.
who that-sister-his adopted dog
'The one that his sister adopted a dog'
(ii) 'axot-oj ·(§el) ha-morej 'imca kelev.
sister-his (of) the-teacher adopted dog
'The teacher's sister adopted a dog'
(iii) Who j would Bill prefer very much t j to win?
(iv) ·BiIl would prefer very much John to win?
(v) L' homme OPj que je crois t j etre le plus intelligent de tous.
The man that I think to be the most intelligent 0/ all
(vi) • Je crois Jean etre le plus intelligent de tous.
I think Jean to be the most intelligent 0/ all

31 I assume Agr. of French present participles to contain covert ~-features.

Additionally, recall that a variable is not Iicensed in the subject position of the
transitive absolute construction (discussed in section 4.3.2). This is so because the
variable is not able to check its ~-features, as tbe participle bears the ~-features
of the object, with which it agrees (presumably this type of Italian participles has
the choice of bearing Agro or Agrp features; the former yields the absolute
construction and is incompatible with reduced relatives or complex tenses).
Finally, I believe that the cases mentioned in the previous note can be accounted
for rather naturally under the advanced proposal.

32 Luigi Rizzi (personal communication) suggests an alternative explanation to the

'only subject' constraint: if 0 crosslinguistically is endowed with agreement
features that render its specifier a mixed A'/A position, movement to SpecDP may
be conceived as A-movement, which can therefore take place only from the subject
position, or else Relativized Minimality would be violated, as the moved element
would skip an A-position (the subject position) (see Rizzi 1990, 1991b; see also
Taraldsen 1986, Rizzi 1990, 1991 b, Vikner 1991 for certain relative
complementizers in Norwegian, French, and Danish, respectively, which behave
on a par with 0 of participial relatives in allowing relativization only from the
subject position).



Under the approach to syntactic nominalization defended in this work,

a verb cannot incorporate into a noun following lexical insertion. N
cannot host a syntactic affixthat has a nominalizing function and
consequently subcategorizes for a VP (chapter 3). In this sense, verbs
cannot change their categorial specification. Hebrew event nominals
have been shown to be purely nominal, although at first glance they
seem to manifest some verbal properties (chapter 3).
D is typically a functional head associated with the nominal system,
but it can introduce a verbal projection if the latter is not tensed. In
other words, D can head non-tensed express ions, whether noun phrases
or non-tensed verbal projections. Participial relatives constitute an
instance of a verbal projection that is embedded under D (chapter 4).
This chapter examines another case where a verbal projection is
dominated by DP. It argues that DP-zation of a verbal projection is
what syntactic nominalization is in essence. The chapter analyzes the
syntactic behavior of Hebrew gerunds (also known as construct
infinitives).) The bulk of the chapter is devoted to the most typical
occurrence of gerunds in Hebrew, that is, gerunds that show verbal
characteristics and appear in adverbial adjuncts headed by temporal
prepositions. It is shown that this type of ge runds involves syntactic
nominalization, as it is inserted as a verb within a verbal projection that
is dominated by DP. Again, what renders this verbal projection a
legitimate complement of D is the fact that it is not tensed. An analysis
of gerunds along these lines immediately accounts for their particular
behavior and distribution.
A small sub set of verbs gives rise to gerunds that show a rather
nominal behavior (henceforth nominal gerunds). These gerunds involve
a nominal projection. I show that they are lexically analyzed as no uns
and do not involve any verbal projection in syntax, in concert with the
present approach to nominalization.
The first section of the chapter presents the properties of gerunds
appearing in adverbial adjuncts: they have verbal properties and differ
in important respects from infinitives, although the morphological form


of gerunds and infinitives is mostly identical. The second section

examines and discards sentential approaches to gerund clauses. The
third section shows that gerund clauses do not involve an NP, as
expected by the restrictive approach to syntaetie nominalization adopted
in this study. The analysis is put forward in the fourth section, where
the peeuliarities of the eonstruction are derived, as weIl as the
distinctions between infinitives and gerunds. Finally, the concluding
section suggests an analysis of nominal gerunds.


By and large Hebrew gerunds can only appear in adverbial adjuncts

introduced by prepositions expressing time relation. 2 In (la) the adjunct
is sentence initial while in (1 b) it is sentenee final:

(1) a. be-hikanes dan la-misrad, eileel ha-telefon.

in-entering Dan to+the-office, rang the-telephone
'When Dan entered the office, the telephone rang'

b. ha-dgalim hunfu, le'axar xalof ha-siyur.

the-flags were+waved, after passing the-patrol
'The flags were waved after the patrol passed'

Gerunds realize their subject either as a full noun phrase, which must
always follow the gerund (1), or as a pronominal suffix (clitic) (2):

(2) me'az hipag§-am, xayim dan ve-nili be-'o§er.

since meeting-3MsPL, live Dan and-Nili in-happiness
'Since they met, Dan and Nili live happily'

As will become clear in what folIows, gerunds show clear verbal

properties and must therefore involve a syntactically realized verbal

5.1.1. Verbal characteristics

Gerunds take accusative arguments (3a). Their accusative Case is the

ordinary accusative of transitive verbs. It does not show any of the
peculiarities exhibited by the inherent accusative of event nominals (3b)

(see discussion in chapter 3, section 3.3). Gerunds (4), unlike event

nominals (S), can take indefinite accusative arguments «4a) versus
(Sa», license accusative Case in exceptional Case marking
configurations «4b versus (Sb» and following a dative argument «4c)
versus (Sc», and realize their direct object as an accusative pronoun
«4d) versus (Sd»:

(3) a. 'im clo'ax ha-cava 'et ha-te'ala, kadru

with crossing(GERuND) the-army ACC the-canal, darkened
'When the army crossed the canal, the sky darkened'

b. clixat ha-cava 'et ha-te'ala

crossing(NouN) the-army ACC the-canal
'the army's crossing of the canal'

(4) a. bi-knot-o sefer, 'alac lib-o.

in-buying(GERuND)-3MSSG book, exulted heart-his
'When he bought a book, his heart exulted'

b. le'axar meco ha-§oftim 'et ha-ne'e§am

after jinding(GERuND) the-judges ACC the-accused
zakay, niftexu ha-dlatot.
innocent, were+opened the doors
, After the judges found the accused innocent, the doors were

c. le'axar haxzir-am le-§ula 'et ha-kesef,

after rejunding(GERuND)-3MSPL to-Shula ACC the-money,
'alac lib-am.
exulted heart-their
'After they refunded the money to Shula, their heart exulted'

d. bi-r'ot-am 'oto, yac'u ha-yeladim min

in-seeing(GERUND)-3MSPL him, went+out the-children from
'When they saw hirn, the children left the room'

(5) a. * kniyat-o sefer

buying(NouN)-his bool

b. *meci'at ha-§oftim 'et ha-ne'e§am zakay

jinding(NOuN) the-judges ACC the-accused innocent

c. * haxzarat-am le-§ula 'et ha-kesef

refund-their to-Shula ACC the-money

d. *re'iyat-am 'oto
seeing(NouN)-their him

Likewise, gerunds can take dative pronouns (6a), while event

nominals cannot do so (6b) (as already mentioned in chapter 3, note

(6) a. be-ha'anik-am 10 'et ha-pras, parac

in-awarding(GERUND)-3MSPL to+him ACC the-prize, burst
dan be-vexi.
Dan in-weeping
'When they awarded him the prize, Dan burst into tears'

b. *ha'anakat-am 10 'et ha-pras hevixa

awarding(NOUN)-their to+him ACC the-prize embarrassed

Finally, gerunds are modifiable by genuine adverbs (7a), whereas

event nominals are not (7b) (see chapter 3, section 3.2):3

(7) a. be-nos'e-nu le'at, hiclaxnu li-mco 'et ha-derex.

in-driving-1PL slowly, succeeded(wE) to-find ACC the-way
'Driving slowly, we succeeded to find the way'

b. *slilat ha-kvi§ le'at

paving the road slowly

If so, then gerund clauses contain a verbal projection, just like

participial relatives (see previous chapter). Indeed, gerunds are
completely predictable and productive verbal forms, unlike deverbal
nouns, whose productivity is restricted. By and large the morphological

form of the gerund is identical to the form of the infinitive. Neither

form is overtly specified for tense and agreement. 4 Syntactically,
however, there are important distinctions between the two, as already
discussed in Berman (1978) and Hazout (1992).

5.1.2. Gerund c1auses versus infinitivals

There are four important syntactic distinctions between gerund clauses

and infinitivals, which concern the prefix le-, their distribution, their
subject, and the availability of negation.
First, the infinitive form always appears with the marker le-, whose
vowel quality is phonologically conditioned (its unmarked value is
schwa). Gerunds never occur with this prefixal marker:

(8) a. dan bike§ la-'alot la-matos.

Dan asked to-ascend to+the-plane
'Dan asked to board the plane'

b. lifney (*la-)'alot ha-nos' im la-matos, hexe I

be/ore (to-)ascending the-passengers to+the-plane, started
la-redet gdem.
to-fall rain
'Before the passengers boarded the plane, it started raining'

The infinitive marker is homophonous with the preposition le- ('to').

Although gerunds are always introduced by prepositions, they are never
introduced by te-, as it is not apreposition that can express time
This leads os to the second difference between infinitivals and gerund
clauses, namely, their distinct distribution. As is weil known, infinitivals
can appear in various argument positions, most typically, as
complements of verbs (see (8a) above), adjectives (9a), or prepositions

(9) a. 10 keday le-'a§en gitane.

not worthwhile to-smoke Gitane
'It is not worthwhile smoking Gitane'

b. hu hevi 'et ha-sfarim kedey la-'azor

he brought ACC the-books in+order to-help
'He brought the books in order to help the children'

Gerund clauses, in contrast, constitute temporal adjuncts. They do not

appear as complements ofverbs (10a) or adjectives (lOb), nor can they
act as complements of non-temporal preposition (IOc):

(10) a. * dan bike! 'alot xayalim la-matos.

Dan asked ascending soldiers to+the-plane

b. *10 keday 'a~en dan gitane.

not worthwhile smoking Dan Gitane

c. * hu hevi 'et ha-sfarim kedey 'azor rina

he brought ACC the-books in+order help Rina

Gerunds must be introduced by temporal prepositions. Although

infinitivals may act as complements of certain prepositions (9b), they
do not appear with the same set of prepositions. While infinitivals occur
with kedey, 'al menat, bisvii (which all mean 'in order'), bli ('without'),
bimkom ('instead of'), and kmo ('like'), gerund clauses are introduced
by the following prepositions (in their temporal sense): be- ('in') and
ke- ('as') (whose vowel quality is phonologically conditioned), 'im
('with'), lifney ('before'), le'axar ('after'), 'ad ('until'), and me'az
A third distinction between infinitives and gerunds has to do with
their subjects. Infinitives cannot phonetically realize their subject.
Gerunds, in contrast, can and in fact must have an overt subject (a full
noun phrase (l1a) or a clitic (llb»:

(11) a. bi-r'ot *(dan) 'et 'im-o, 'alac lib-o.

in-seeing (Dan) ACC mother-his, exulted heart-his
'When Dan saw his mother, his heart exulted'

b. bi-r'ot·( -0) 'et 'im-o, 'alae lib-o.

in-seeing(3MSSG) ACC mother-his, exulted heart-his
'When he saw his mother, his heart exulted'

Finally, eonsider negation. While infinitives ean be freely negated

(12a), ge runds eompletely disallow negation (l2b):

(12) a. dan bike§ 'otanu 10 le-'a§en po.

Dan asked us not to-smoke here
'Dan asked us not to smoke here'

b. • 'im 10 havin-o 'et ha-maeav,

with not understanding-3MSSG ACC the-situation,

The above distinetions are summarized in the following ehart:


Infinitivals Gerund clauses

Infinitive marker obligatory impossible
Distribution argument position with temporal Ps
Overt subjeet impossible obligatory
Negation possible impossible

In seetion 5.4, onee the analysis of gerund clauses has been put
forward, these distinetions reeeive a straightforward aeeount. I now turn
to review two sentential approaehes to gerund clauses, and diseuss the
reasons for whieh I do not adopt them.


Prepositions ean take either noun phrases or sentential clauses as their

eomplements. As gerund clauses are introdueed by prepositions, apriori
they may eonstitute one of the two. Let me first eonsider the sentential

5.2.1. The P-CP analysis

One possible analysis of gerund clauses could conceive of them as non-

finite sentential structures selected by temporal prepositions. This would
immediately explain how verbal projections can be embedded under a
prepositional head (IP stands for the sentential inflectional categories):

(14) pp






However, this sentential analysis encounters serious difficulties, as

al ready observed by Hazout (1992). The difficulties stern form the
distinct syntactic behavior infinitives and gerunds show.
First, if gerunds, like infinitives, are non-finite verbal forms involving
a sentential structure, it is not clear why they cannot bear the infinitive
marker le- (see (Sb) above)?
Second, if gerund clauses are CPs, why can they not appear in
various argument positions, but must function as complements of
temporal prepositions (see (10) above)?
Third, there is no obvious reason why gerunds cannot be negated if
they entail a sentential structure (see (12b) above).
Fourth, the question arises why gerunds must phonetically realize
their subject (see (11) above). Moreover, it is not at all clear how the
subject of the gerund receives Case. Its Case is not morphologically
marked. It is clearly not accusative Case, as shown by the fact that a

definite subject is not preceded by 'et (ISa), unlike other definite

accusative arguments (see chapter 3, section 3.3), nor can it alternate
with the accusative pronoun (15b). Recall, additionally, that a transitive
gerund checks its accusative Case with its direct object; so where could
the subject have accusative Case from?

(15) a. ha-dgalim hunfu, le'axar xalof (·'et) ha-siyur.

the-flags were+waved, after passing (ACC) the-patrol
'The flags were waved, after the patrol passed'

b. *ha-dgalim hunfu, le'axar xalof 'oto.

the-flags were+waved, after passing him

The Case of the subject is not nominative either, as shown by the fact
that the subject cannot alternate with a nominative pronoun (16a),
unlike other nominative arguments in apreverbal or postverbal position
(16b-c). Indeed, given their non-finiteness, we would not expect
gerunds to be able to determine nominative Case:

(16) a. • 'im clo'ax hu 'et ha-te'ala, kadru ha-~amayim.

with crossing he ACC the-canal, darkened the-sky

b. hu calax 'et ha-te'ala.

he crossed ACC the-canal
'He crossed the canal'

c. 'etmol calax hu 'et ha-te'ala.

yesterday crossed he ACC the-canal
'Yesterday he crossed the canal'

Finally, the Case of the subject cannot be assigned by the preposition.

First, the subject must always follow the gerund and is therefore not in
a local configuration with the preposition; but Hebrew requires a local
configuration between prepositions and their assignees. Second, a
pronominal subject is attached to the gerund and not to the preposition
(see, for instance, (1Ib) above). In Hebrew, however, a pronominal
clitic is always strictly attached to its Case assigner (see chapter 2,
section 2.4.2 for some examples). If the Case assigner were the
preposition, why would the clitic be attached to the verb? In section
5.2.2 I add arguments to the same effect. The analysis proposed in
section 5.4 explains the Case assignment mechanism.

An additional problem raised by the sentential analysis, which has

not been discussed so far, has to do with object clitics. In chapter 2
(section 2.4.2), it is mentioned that verbs can realize their objects as
pronominal clitics. These clitics appear suffixed to finite (17b) and non-
finite verbs (17a). Gerunds, unlike anY other verbal form, do not permit
object clitics (17c). This is completely unexpected under any sentential

(17) a. hu roce le-hazkir-o.

he wants to-mention-him
'He wants to mention him'

b. hizkart-iv.
'I mentioned him'

c. • be-hazkir-o dan, yaca-nu min ha-xeder.

in-mentioning-him Dan, went+out-we /rom the-room

Finally, as already shown above, there is no correspondence between

the prepositions that can take gerund clauses as their complements and
those that can select non-finite CPs (see (9b) versus (10c) and the
related discussion). This would be rather unexpected if gerund c1auses,
like infinitivals, were non-finite sentential structures. Furthermore, while
most of the prepositions that introduce gerund clauses can select a finite
CP, the preposition be- ('in'), which is very common with gerunds,
never takes a CP-complement. What the initial prepositions of gerund
c1auses do have in common is their ability to select a noun phrase as
their complement (Glinert 1989). In section 5.3 lexamine and dismiss
the possibility that gerunds contain an NP in syntax. In section 5.4 I
argue that they project a DP. This explains, among other things, why
they appear only with noun phrase taking prepositions. Prior to that
however, let me examine another version of the sentential approach put
forward by Hazout (1992).

5.2.2. The CP analysis

Noting several drawbacks of the P-CP analysis, Hazout (1992)

advances a different version of the sentential approach. Hazout analyzes
the initial prepositions introducing gerund clauses as prepositional
complementizers (similar to the English prepositional complementizer

for) that have a temporal interpretation and take an IP complement,

whose specifier position is occupied by the subject. For morphological
reasons, the gerund raises to I and subsequently to C, which explains
the surface order of constituents:

(18) CP


[I+V] A
DPs I'


Under Hazout's approach, prepositional complementizers are lexically

Iisted and are different from the phonetically identical prepositions.
Thus, there is no wonder that there is no correspondence between the
initial prepositions of gerunds and those taking CP-complements,
because the former are in fact complementizers.
Moreover, Hazout notes that this also automatically explains certain
peculiarities the prepositional complementizers show in comparison with
noun phrase taking prepositions. First, the preposition be- ('in') does
not readily have a temporal interpretation when it selects a noun phrase
complement (19b); when acting as a prepositional complementizer, it
does have a temporal reading (19a). And second, the preposition
bemdex ('during') can have noun phrase complements (20a), but does
not take gerund clauses (20b):6

(19) a. be-hikare Amot ha-zoxim, yacati

in-reading(GERUND) names the-winners, went+out(I)
min ha-xeder.
/rom the-room
'When the winners' names were read, lieft the room'

b. • be-hakra'at Amot ha-zoxim yacati min

in-reading(NOUN) names the-winners went+out(I) /rom

(20) a. bemeAex ha-hafgana hunfu ha-dgalim.

during the-demonstration were+waved the-flags
'During the demonstration the flags were waved'

b. • bemeAex hikare ha-Airim, nirdamti.

during reading the-poems, Jell+asleep(I)

Hazout proposes that the subject of the gerund receives Case from
the prepositional complementizer, once gerund raising up to C has taken
place. This solves the Case problem and accounts for the distribution
of gerund clauses as temporal adjuncts: their complementizer imposes
a temporal interpretation, which hinders them from occurring in other
typical CP-positions. Further, gerunds cannot appear without the
complementizer, as their subject will not have Case, nor can they bear
the infinitive marker, because by hypothesis, it would block
incorporation of the gerund with C.
Although this analysis disposes of major problems raised by the p-
CP analysis, it still encounters important difficulties.
First, a special explanation is needed to rule out negation, which is
otherwise allowed in sentential structures.
Second, it is not at an clear why object clitics cannot be attached to
the gerund, although they can appear with other verbal forms.
More importantly, I believe, an analysis in terms of overt gerund
raising to C cannot be maintained for empirical reasons. As ilIustrated
in (21), two gerund clauses can be coordinated under a single
prepositional head. Under a generalized gerund raising analysis, there
is no adjunction site for the second gerund, as only one prepositional
complementizer is present. And for the same reason, there is no Case
available for the subject of the second gerund, as it is the complex head
created by incorporation of the gerund with the complementizer that is

responsible for the Case of the subject:

(21) a. le'axar xacot-i 'et ha-kviS va-'alot-i

after crossing(-ISG) ACC the-road and-going+up(-lsG)
la-'otobus, nirgati.
to+the-bus, calmed+down(I)
'After crossing the road and taking the bus, I calmed down'

b. 'im xalof ha-mic'ad ve-redet ha-xa§exa,

with passing the-parade and-falling the-darkness,
hitpazer ha-kabal.
dispersed the-crowd
'When the parade passed and darkness fell, the
crowd dispersed'

Mentioning examples like (22) below, Hazout concludes that

coordination of two gerund clauses under a single prepositional
complementizer is impossible, which is straight forward under his

(22) * bi-§mo'a dan 'et ha-muzika 'u-r'ot dina 'et

in-hearing Dan ACC the-music and-seeing Dina ACC
the-movie ...

Given the grammaticality of the examples in (21), the ungrammaticality

of (22) cannot reflect a standard impossibility to coordinate gerund
clauses. Rather, it ought to follow from some peculiarity of the initial
preposition in this example. One possibility that immediately comes to
mind is that it is the prefixal nature of be- ('in') which is responsible
for its distinct behavior. Indeed, the only other preposition that does not
seem to allow coordination is the other prefixal preposition, ke- ('as,):7

(23) * ka-'avor ha-mic'ad ve-redet ha-xa§exa,

as-passing the-parade and-falling the-darkness,
hitpazer ha-kabal.
dispersed the-crowd

If overt gerund raising to C cannot be generalized, then, assuming

that the subject occupies SpeclP, one can no more derive the surface
order of constituents, as schematized in (24a) (recall that the subject

always follows the gerund). Alternatively, if the subject is lower in the

structure, it is not in a local configuration with the prepositional
complementizer (24b), and the analysis encounters the same Case
problems as the P-CP analysis (see section 5.2.1 above):

(24) a. • [cp Cprep . [IP DPs Vger ... ]]

b. • [cp Cprep . [IP Vger DPs '" ]]

In conclusion, both versions of the sentential analysis raise important

problems that cast doubts on their adequacy. I therefore turn to examine
an alternative analysis arguing that gerund clauses involve a syntactic
process of nominalization.


5.3.1. Against NP

Gerund clauses display verbal properties and appear embedded under

temporal prepositions. The prepositions at stake all share the ability to
take noun phrases as their complements (Glinert 1989). Prepositions
confined to sentential complements cannot have gerund clauses as their
complements. What is the exact categorial nature of gerund clauses,
which enables them to act as complements of noun phrase taking
Under the approach to nominalization defended in this work, verbs
cannot be dominated by NPs: nouns do not subcategorize for VPs and
do not act as syntactic nominalizers. Since gerunds are inserted in
syntax as verbal forms (see section 5.1.1), we do not expect them to
project an NP. We therefore predict that gerund clauses will not show
typical nominal properties. This prediction turns out to be correct.
Gerunds do not display the most typical properties of nouns, as
already observed by Hazout (1992). First, gerunds cannot be modified
by adjectives (25a). Nouns are of course modifiable by adjectives (25b):

(25) a. • be-lext-o (ha-)mahir, ra'a dan

in-walking(GERuND)-3MSSG (the-)rapid, saw Dan

b. halixat-o ha-mehira hifti'a 'otanu.

walking(NouN)-his the-rapid surprised us
'His rapid walking surprised us'

Second, gerunds do not license sei ('of') phrases, which appear only
in NP environments, yielding either a free state (26a) or a clitic
doubling configuration (26b) (see chapter 2 for discussion). Gerunds
pennit neither a free state (26e) nor a elitic doubling eonstruction (26d):

(26) a. ha-knisa sei dan la-misrad hifti'a

the-entering(NoUN) 0/ Dan to+the-office surprised
'Dan's entering the office surprised us'

b. knisat-o sei dan la-misrad hifti'a

the-entering(NouN)-his 0/ Dan to+the-office surprised
'Dan's entering the office surprised us'

c. • be-hikanes sei dan la-misrad, eileel

in-entering(GERUND) 0/ Dan to+the-office, rang

d. • be-hikans-o sei dan la-misrad, eileel

in-entering(GERuND)-3MSSG 0/ Dan to+the-office, rang

Third, gerunds must always phonetieally realize their subjeet (27a),

unlike nouns (27b):

(27) a. * ba-'amod ba-tor, ra'a dan let

in-slanding(GERuND) in+lhe-line, saw Dan ACC

b. ha-'amida ba-tor 'iyfa toto.

the-standing(NOUN) in+the-line tired him
'Standing in line tired him'

Fourth, gerunds do not have a plural form (but some nouns eannot
pluralize either). And fifth, gerunds ean never appear with the definite
artiele (reeall that Hebrew has only adefinite artiele). Note, however,
that even if gerunds were nouns, but for some reason their Case
assignment meehanism neeessarily involved the eonstruet state and
never the Case marker sei ('of'), we would anyway not expeet them to
bear the artiele, as the head of the eonstruet state ean never realize its
artiele (see ehapter 2, seetions 2.1.1 and 2.3.3):8

(28) be-(*ha-)hikanes dan la-misrad, eileel

in-(the-)entering(GERuND) Dan to+the-office, rang
'When Dan entered the office, the telephone rang'

In addition, as is already evident, gerund elauses (29), unlike noun

phrases (30), eannot appear in regular DP positions: subjeet position,
objeet position, or as eomplements of non-temporal prepositions:

(29) a. * hikans-o la-misrad hifti'a 'oti.

entering(GERuND)-3MSSG to+the-office surprised me

b. * hi eilma 'et hikans-o

she photographed ACC entering(GERUND)-3MSSG

e. * hi him§ixa la-'avod lamrot hikans-o

she continued to-work despite entering(GERUND)-3MSSG

(30) a. knisat-o la-misrad hifti'a 'oti.

entering(NouN)-his to+the-office surprised me
'His entering the office surprised me'

b. hi cilma 'et knisat-o la-misrad.

she photographed ACC entering(NOUN)-his to+the-office
'She photographed his entering the office'

c. hi him~ixa la-'avod lamrot knisat-o

she continued to-work despite entering(NouN)-his
'She continued working despite his entering the office'

Finally, Hebrew uses a particular negation morpheme 'i- to negate

derived nouns, as shown in (31 a). Gerund clauses cannot be negated by
'i-, as illustrated by (31 b):

(31) a. 'i-havanat-o 'et ha-macav hirgiza

NEG-understanding(NOuN)-his ACC the-situation irritated
'The fact that he did not understand the situation irritated

b. • 'im 'i-havin-o 'et ha-macav,

with NEG-understanding(GERuND)-his ACC the-situation

The chart below summarizes the above distinctions:

Noun phrases Gerund clauses
Adjectives possible impossible
Sei phrases possible impossible
Overt subject possible obligatory
Article possible impossible
Distribution DP-positions with temporal Ps
Negation by 'i- possible impossible

The unavailability of adjectives and sei ('of') phrases indicates that

gerunds do not entail a structurally realized NP. Likewise, their
incompatibility with the negation prefix 'i-, which is used to negate
derived nouns, shows that gerunds are not nouns syntactically derived
from verbs. This is predicted by the- present approach to nominalization.
A detailed discussion of the additional entries in the chart is postponed
until section 5.4.
Although gerund clauses do not involve an NP, they may nonetheless
be dominated by DP, which would immediately explain why they
appear with noun phrase taking prepositions only. In the previous
chapter it is suggested that verbal projections can be headed by D if
they are not tensed. As is shown below, gerund clauses constitute
legitimate candidates, as they are not tensed.

5.3.2. Tense

As already observed by Berman (1978), gerund clauses do not have an

intemally fixed time frame. They can appear in all three tenses. Their
tense is determined by the tense value of the matrix clause. For
example, it is past in (33) and future in (34). The initial preposition
defines the exact time reference with regard to the event expressed by
the matrix clause, simultaneity in (33a) and (34a), and succession in
(33b) and (34b):

(33) a. 'im hikans-o la-misrad, kulam hitxilu

with entering-3MSSG to+the-office. everyone started
'When he entered the office, everyone started working'

b. le'axar hikans-o la-misrad, kulam hitxilu

after entering-3MSSG to+the-ofJice. everyone started
'After he entered the office, everyone started working'

(34) a. 'im hikans-o la-misrad, kulam yatxilu

with entering-3MSSG to+the-office, everyone will+start
'When he will enter the office, everyone will start working'

b. le'axar hikans-o la-misrad, kulam yatxilu

after entering-3MSSG to+the-office, everyone will+start
'After he will enter the office, everyone will start working'

Gerund clauses do not have an internally fixed tense value. This is

what they share with noun phrases and this is what renders them
legitimate complements of D. If gerund clauses are verbal projections
dominated by DP, it follows both that they have verbal properties and
that they can function as complements of DP taking prepositions.
In the subsequent section, I develop an analysis along these lines. As
will become clear in the course of this section, the distribution of
gerunds and the fact that they cannot appear with the article (chart (32»
are related to the nature of their D. The obligatoriness of an overt
subject (chart (32» follows from the fact that the configuration at stake
is that of the construct state, where the Case checking head does not
license a null element by itself.


Gerund clauses are not tensed, which suggests that they do not contain
TP. If the projection of NegP is contingent upon the presence of TP
(see Laka 1990, Zanuttini 1990), lack of the latter immediately explains
why gerund clauses do not permit sentential negation (see chart (13».
Recall that I assume that clausal reduction cannot skip an arbitrary layer
but must proceed by cutting off external layers, as hypothesized by
Rizzi (1993) (see chapter 4). I further assurne that the sentential
functional projections are hierarchically organized as follows (Belletti
1990): " ... AgrsP-(NegP)-TP ... ". It follows that gerund clauses should
contain no sentential functional category higher than TP, neither NegP
nor AgrsP. Gerunds take arguments bearing structural accusative Case,
as shown in section 5.1.1. If structural Case is always the realization of
a Spec-Agr relation, as assumed in this work, gerund clauses contain an

AgroP dominating VP.

Below lexamine the structure of gerund clauses in detail. Four major
issues deserve attention: first, the Case their subject bears (section
5.4.1), second, their internal structure (section 5.4.2), third, their limited
distribution, and fourth, the fact that they do not have a definiteness
value; the last two issues are related and discussed in section 5.4.3.

5.4.1. Construct states

In section 5.2.1, it is shown that the Case the subject of the gerund
bears is not accusative nor nominative. It is also argued that it cannot
be an oblique Case assigned by the initial preposition. The possibility
that immediately comes to mind, then, is that the subject is in a
genitival relation with the gerund. This is not really surprising, because
the configuration at stake is very reminiscent of that of the construct
state (extensively discussed in chapter 2), as abrief comparison of the
gerund clauses in (35a,c) with the noun phrases in (35b,d) already

(35) a. ha-ye~iva nin'ala, 'im hitnaged ha-~agrir

the-session was+closed, with opposing the-ambassador
'The session was closed with the ambassador's opposing
the proposal'

b. ha-ye~iva nin'ala, 'im hitnagdut ha-~agrir

the-session was+closed, with opposition the-ambassador
'The session was closed with the ambassador's opposition
to the proposal'

c. ha-ye~iva nin'ala, 'im hitnagd-o

the-session was+c1osed, with opposing-3MSSG
'The session was closed with his opposing the proposal'

d. ha-yesiva nin'ala, 'im hitnagdut-o

the-session was+closed, with opposition-his
'The session was c10sed with his opposition to the

In chapter 2, I observed that genitive Case resulting from the

construct state is available in exceptional Case marking configurations
(ECM) (see (36b) below), drawing the conclusion that it is a structural
Case that is checked within AgrgenP, unlike the inherent genitive
assigned via the Case marker sei ('of') (36a). Gerunds, just Iike nouns
in the construct state, license Case in ECM configurations, as shown in

(36) a. * ha-meci'a sei ha-ne'esam 'asem hifti'a 'otanu.

the-finding(NOUN} 0/ the-accused guilty surprised us
b. meci'at ha-ne'esam 'asem hifti'a 'otanu.
jinding(NouN} the-accused guilty surprised us

c. be-himace ha-ne'esam 'asem,

in-being+/ound(GERuND} the-accused guilty,
'When the accused was found guilty, we were surprised'

Moreover, recall that the subject of the gerund may be realized as a

suffixal clitic (35c). This clitic is identical to the nominal clitic (Rosen
1962, Berman 1978). By and large, clitics appearing on verbs and
clitics appearing on nouns are morphologically identical. However, in
the first person singular, there is a systematic difference between the
nominal clitic (37a) and the verbal clitic (37b). As shown in (37c),
when a gerund realizes its subject as a pronominal clitic, the latter is
identical to the nominal clitic (and not to the clitic appearing with
verbs), thus confirming the claim that gerunds are in a genitival relation
(in the construct state) with their subject:

(37) a. beyt-i
'my house'

b. hu raca le-vakr-eni.
he wanted to visit-me
'He wanted to visit me'

c. be-vakr-i'*-eni 'oto, ra'iti 'et bit-o.

in-visiting-JSG(=my) him, saw(I) ACC daughter-his
'When I visited hirn, I saw his daughter'

In sum, various considerations lead us to conclude that in addition to

nominal construct states, there are also construct states that comprise a
gerund and its subject.

5.4.2. Internal structure

Given the above arguments, gerund clauses must contain an Agrgen

projection, where their subject checks genitive Case, whether it is a full
noun phrase or a pronominal clitic (wh ich is argued in chapter 2 to be
a presyntactic suffix realizing the Case features of its host). It follows
that structural genitive Case can be determined not only by nominal
heads, but also by verbal heads that do not bear the morphological
features corresponding to the inflectional sentential heads T and Agrs '
More precisely, when a verb bears tense and agreement features, it must
be part of a sentential structure, which enables it to check the relevant
features. But when a verbal form bears agreement features (and no
tense features), it checks them within AgrgenP if the form is a gerund,
and within AgrpP if it is a participle (Agrp' recall, does not check Case
but only cj>-features; see chapter 4). As is already clear, both Agrl and
AgrgenP are compatible with D, as they are constituents that do not bear
tense features. 9
In brief, gerund clauses do not contain any sentential functional
material higher than AgroP' Their subject is in the construct state with
the gerund, thus checking structural genitive Case within AgrgenP. They
are not tensed phrases and therefore qualify as components of the DP
system, which immediately explains why they are introduced by DP
taking prepositions. Gerund clauses then entail the following basic

(38) pp







AgroP VP

Let us now determine the position of the different constituents within

the clause by speil-out, starting by the relative positioning of the (non-
pronominal) subjeet and the direet objeet. Reeall that in Hebrew a full
direet object surfaees in situ (Agro has weak DP-features). Trivially,
then, it follows the subjeet (39a), and its position teils us nothing about
the exact loeation of the subject. Unstressed aecusative pronouns,
however, have to raise overtly (39b). If Friedemann and Siloni (1993)
are correet in arguing that they overtly raise as high as SpecAgroP' it
follows that the subject does not overtly raise out of its base position,
as it must follow an unstressed pronominal object (the structure is
depicted only where relevant):

(39) a. bi-r'ot dan 'et 'im-o ...

in-seeing Dan ACC mother-his
'When Dan saw his mother ... '

b. bi-r'ot j [AgroP 'otaj tj [vp dan tj tj •••

in-seeing her Dan

Similarly, other unstressed pronouns (say dative pronouns) must precede

a full subject (40). According to Friedemann and Siloni (1993), they
must overtly adjoin to AgroP, which immediately explains why in
gerund clauses they, too, precede a lexical subject:

(40) 'im tet 10 'im-o let ha-matana, parac dan

with giving to+him mother-his ACC the-gift, burst Dan
'When his mother gave hirn the gift, Dan burst into tears'

As the subject does not raise overtly, the DP-features of Agrgen ought
to be weak, allowing (in fact forcing) covert checking. Note that this
way of viewing things would not be allowed under any framework
requiring Case checking earlier than LF (more specifically, the
framework developed in Chomsky 1981 and much subsequent
The positioning of unstressed pronouns also indicates that the gerund
raises at least as high as Agrgen , because it must always precede such
pronouns (as weil as the subject, of course). Ag~en then reasonably has
strong V-features. The corresponding features of D mayaiso be strong,
triggering further gerund raising. Nothing crucial hinges on this here
and no empirical evidence showing whether or not the gerund
undergoes further overt movement to D is available.
Before we conclude, a note on subject clitics is in order. Following
the discussion in chapter 2 (section 2.4), Iassume they are presyntactic
suffixes realizing the Case features of their host. By analogy with
nominal clitics, they materialize the genitive Case features of the
gerund, which are checked in Agr,en' once the gerund has raised there.
The corresponding argument position (the subject position) is occupied
by pro, which ends up in SpecAgrgenP, whose head is capable of
licensing pro: it is rich enough to do so once it has checked Case and
cj>-features with the clitic (as discussed in chapter 2, section 2.4):

(41) [pp 'im [~p [AgrgCDP proj hitnagdro j ···[vp t·I t·J
with opposing-his

In the next section, I discuss the somewhat "weak" nature of D that

heads gerund clauses. I suggest that it is the character of their D which
hinders gerund clauses form appearing in argument positions.

5.4.3. The nature 0/ D

Gerund clauses cannot occupy regular DP-positions (various argument
positions), as illustrated in (29) above and repeated here for

(42) a. • hikans-o la-misrad .hifti'a 'otL

entering-his to+the-office surprised me

b. • hi cilma 'et hikans-o la-misrad.

she photographed ACC entering-his to+the-office

c. • hi him§ixa la-'avod lamrot hikans-o la-misrad.

she continued to-wor! despite entering-his to+the-office

Under the analysis advanced here, gerund clauses are DPs. Therefore,
their inability to occupy standard argument positions is apriori
As already mentioned in the previous chapters, there are good
theoretical and empirical reasons to believe that referentiality is a
property of D. It is D that turns the expression it heads into a
referential expression, which is consequently able to occupy an
argument position (see Szabolcsi 1987, 1989, Stowell 1989, 1991,
Longobardi 1994). The described inability of gerund clauses may
therefore stern from some "weakness" of their D position.
Recall that gerunds can never be accompanied by the article (chart
(32». In this regard, they do not differ from nouns in the construct
state. A head in the construct state (that is, a head endowed with Agrgen
features) can never have its article phonetically realized. Still, nouns in
the construct state clearly have a [±definite] value, which must match
the [±definite] value of their genitive argument. In chapter 2 (section
2.3.3), it has been proposed that they realize their definiteness feature
as part of their Agrgen features. This is why they must agree in
definiteness with their genitive argument, with which they are in an
agreement relation.
Now, reasonably gerunds do not have an intrinsic definiteness value,
since as verbs they do not bear a definiteness feature (nor intrinsic <1>-

features). I suggest that in this regard their D position is inert and

therefore not referential. Having no reference, D of gerunds cannot head
an expression that can act as an argument, that is, occupy an argument
position.t 2 Of course, the question arises what nonetheless enables
gerundival DPs to appear as complements of prepositions. The fact that
these prepositions must be temporal may give us some insight into the
problem. The initial prepositions of gerund clauses denote the exact
time reference of the event expressed by the gerund; they specify it
within the time frame determined by the matrix clause (see (33)-(34)
and the related discussion). In fact, they provide the gerund (and more
precisely its D position) with reference, time reference. Acquiring
reference, D of gerunds can head a clause appearing in an argument
position; but this argument position can only be the complement
position of temporal prepositions, since temporal prepositions are
crucial for the reference of D to be determined. t3
To summarize, I have suggested that gerund clauses are DPs whose
head is not inherently referential. Hence, they must be introduced by
temporal prepositions, which provide their D position with reference.
Gerund clauses are not tensed and do not contain any sentential
functional projection higher than AgroP. Their subject checks genitive
Case in SpecAgrgenP on a par with subjects of no uns in the construct
state. The next section discusses some more differences and similarities
between gerund clauses and noun phrases.

5.4.4. More on gerund clauses and noun phrases

As already mentioned above, the prepositions that introduce gerund

clauses are all noun phrases taking prepositions, which, of course,
strengthens the analysis of gerund clauses as DPs. But if indeed these
prepositions are simply DP taking prepositions, it should be explained
why certain prepositions show some peculiarities when they appear with
gerund clauses. As observed by Hazout (1992) (and already illustrated
in (19)-(20), repeated below), the preposition be- ('in') does not readily
have a temporal interpretation when it selects a noun phrase
complement (43b); when it selects a ge rund clause, it does have a
temporal reading (43a). Additionally, the temporal preposition bemdex
('during'), which can take noun phrase complements (44a), does not
permit gerund clauses (44b):t4

(43) a. be-hikare §mot ha-zoxim, yacati

in-reading(GERuND) names the-winners, went+out(I)
min ha-xeder.
/rom the-room
'When the winners' names were read, lIeft the room'

b. * be-hakra'at §mot ha-zoxim yacati min

in-reading(NouN) names the-winners went+out(I) Jrom

(44) a. beme§ex ha-hafgana hunfu ha-dgalim.

during the-demonstration were+waved the-flags
'During the demonstration the flags were waved'

b. *beme§ex hikare ha-§irim, nirdamti.

during reading the-poems, Jell+asleep(I)

Consider first the preposition bemeSex ('during'). Although it can

take noun phrase complements (44a), it is incompatible with an event
reading. The examples in (45) are ungrammatical, because the deverbal
nouns are forced to an event interpretation by an accusative argument
(45a-b) and by an adverbial pp (45c) (see chapter 3, section 3.1.2):

(45) a. * beme§ex harisat ha-cava 'et ha-'ir

during destruction the-army ACC the-city

b. * beme§ex hakra'at-am 'et ha-§irim

during reading(NOUN)-their Ace the-poems

c. * bemdex hitkansut ha-mafginim

during gathering(NOUN) the-demonstrators
be-'itiyut hexanu krazot.
in-slowness prepared(WE) placards

The incompatibility of the preposition bemeSex ('during') with deverbal


nouns having an event structure immediately suggests an account of its

incompatibility with gerund clauses. Being verbal, gerunds always have
an event structure in the sense defined in chapter 3, and therefore
cannot cooccur with bemdex.
The preposition be- ('in') is not temporal by itself, when it selects a
noun phrase complement. It denotes a time relation only if the
associated noun designate an entity with some temporal extension.
Thus, while (46a) is grammatical, (46b) is impossible, although its
gerund counterpart is natural (46c):

(46) a. ba-boker kadru ha-samayim.

in+the-morning darkened the-sky
'In the morning the sky darkened'

b. • be-clixat ha-cava let ha-te'ala kadru

in-crossing(NouN) the-army ACC the-canal darkened

c. bi-clo'ax ha-cava let ha-te'ala, kadru

in-crossing(GERuND) the-army ACC the-canal, darkened
'When the-arrny crossed the canal, the sky darkened'

SirniIarly, the preposition ke- ('as') hardly ever has a temporal

interpretation when it selects a noun phrase complement (47a). It has
a temporal meaning when it intro duces a gerund clause, though (47b):

(47) a. • ke-yeci'at-o rnin ha-misrad hexel la-redet

as-going+out(NOUN)-his from the-office started to-fall

b. ke-cet-o min ha-misrad, hexel

as-going+out(GERUND)-his /rom the-office, started
la-redet gesem.
to-fall rain
'As he left the office, it started raining'

I would like to suggest that the two prepositions be- and ke- have a
distinet interpretation depending on whether they seleet a referential DP
as in (46a-b) and (47a), or a DP they have to render referential, that is,
a gerund clause as in (46e) and (47b). Only in the latter ease do they
express a temporal relation intrinsieally. This specifieation ean be
eneoded in the lexical information they bear.
Having diseussed these diserepancies between noun phrases and
ge rund clauses, let me add arguments to the effeet that the latter are
DPs. First, assuming that both gerund clauses and event nominals are
DPs, it is only natural that they ean be eoordinated:

(48) 'im xalof ha-siyur ve-hitkansut

with passing(GERuND) the-patrol and-gathering(NouN)
ha-mafginim, kadru ha-~amayim.
the-demonstrators, darkened the-sky
'When the patrol passed and the demonstrators gathered,
the sky darkened'

Seeond, it appears that the gerund clause ean even oeeur as the
genitive DP of eertain nominal eonstruet states. This possibility is
mainly limited to eonstruets whose head is a temporal noun 15 :

(49) hu zaxar 'et rega hikans-a la-xeder.

he remembered ACC moment entering-her to+the-room
'He remembered the moment she entered the room'

Reasonably, temporal nouns, on a par with temporal prepositions, ean

provide the D position of the gerund with referenee. Note that the
existenee of eonstruets whose genitive member is a gerund clause is
eompletely unexpeeted under any sentential analysis of gerund elauses;
ep never oeeurs as the genitive member of the eonstruet state.
In sum, like event nominals, gerund clauses are DPs and have an
event strueture. But only gerund clauses eontain a verbal projeetion and
show verbal properties. Event nominals undergo presyntaetic
nominalization and therefore do not show verbal properties. Their
formation is not really predietable, whieh is typieal of the lexieal
eomponent (for diseussion, see ehapter 3). Gerund clauses, on the
eontrary, involve syntaetie nominalization, whieh means that their
verbal form is embedded under D. Gerund clauses eonstitute legitimate
eomplements of 0 beeause they do not bear tense features. Like other
verbal forms, their formation is entirely produetive.

In the next section I show that this analysis automatically derives the
distinct syntactic behavior of infinitivals and gerund clauses (see chart
(13». As gerund clauses are DPs, while infinitivals have sentential
structure, there is no wonder that their syntactic behavior is different.

5.4.5. Gerund clauses and infinitivals

Under the DP analysis of gerund clauses, the syntactic differences

between infinitivals and gerund clauses immediately fall out.
First, infinitivals, but not gerund clauses, have a sentential structure.
If the infinitive marker le-, like its English counterpart to, is part of the
sentential inflectional morphology, we expect it to appear with
infinitives but not with gerunds (as shown in (8) above and repeated

(50) a. dan bike§ la-'alot la-matos.

Dan asked to-ascend to+the-plane
'Dan asked to board the plane'

b. lifney (*la-)'alot ha-nos'im la-matos,

be/ore (to-)ascending the-passengers to+the-plane,
hexel la-redet ge§em.
started to-fall rain
'Before the passengers boarded the plane, it started raining'

Following Stowell (1982), among others, Iassume that infinitivals

contain a tense operator (as already mentioned in chapter 4). The
marker le- is arguably the tense marker of infinitives. According to
Rizzi (1994), the tense of infinitivals is externally bound. To some
extent, their time frame hinges upon the tense of the matrix clause (in
Stowell's terms a "possible future" with respect to the matrix).
However, it clearly does not exhibit complete tense dependency on the
matrix, as do gerund clauses. In the following examples, for instance,
the tense of the infinitivals is fixed (or bound) by the temporal adverbs
'yesterday·. 'today', and 'tomorrow', (and compare (51) with (33)-(34)

(51) a. dan raca Ii-Ibo§ 'etmol xalifa.

Dan wanted to-wear yesterday suit
'Dan wanted to wear a suit yesterday'

b. dan raca li-Ibos hayom xalifa.

Dan wanted to-wear today suit
'Dan wanted to wear a suit today'

c. dan raca li-Ibos maxar xalifa.

Dan wanted to-wear tomorrow suit
'Dan wanted to wear a suit tomorrow'

Second, infinitivals and gerund clauses do not have the same

distribution, because gerund clauses must inherit their reference from
a temporal preposition (or noun); hence they are prevented from
appearing in most argument positions. Moreover, there is no wonder
that infinitivals that are introduced by prepositions do not share the
same set of initial prepositions with gerund clauses, since infinitivals
are CPs, whereas gerund clause constitute DPs that are necessarily
linked to temporal prepositions (or nouns).
Third, infinitives, unlike gerunds, cannot phonetically realize their
subject. The infinitival inflectional morphology is not rich enough to
check the Case of an overt subject. Gerunds, on the other hand, bear
Agrgen features, which allow a phonetically realized subject to check
genitive Case. Agrgen features are part of the morphological form of the
gerund. In a way, this is what renders ge runds possible complements of
D, as Agr~en enables the verbal form to check Agr-features that are not
accompamed by tense features. Agrgen by itself cannot license pro (see
chapter 2). It follows that an overt subject is obligatory simply because
Agrgen is always present (recall that Agrgen is capable of licensing a null
element, once it has checked Case and </>-features with a pronominal
clitic (see section 5.4.2».16
Finally, as infinitivals have a sentential structure, they permit
sentential negation, unlike gerund clauses (as illustrated in (12) and
repeated below):

(52) a. dan bikes 'otanu 10 le-'asen po.

Dan asked us not to-smoke here
'Dan asked us not to smoke here'

b. * 'im 10 havin-o 'et ha-macav,

with not understanding-3MSSG ACC the-situation,

In conclusion, I have proposed that gerund clauses are DPs whose

head acquires its referentiality from a temporal preposition (or noun).
Gerund clauses are non-tensed propositions and do not contain any
sentential functional projection higher than AgroP' Their subject checks
genitive Case in SpecAgrgenP, like subjects of nouns in the construct
state. The distinct behavior of gerund clauses and infinitivals
immediately follows. In the next section lexamine the behavior of a
group of gerunds that show a rather nominal behavior. I argue that these
gerunds are lexically analyzed as nouns.


As observed by Berman (1978) (see also Glinert 1989, Hazout 1990),

a small subset of unaccusative verbs gives rise to gerunds that show
some clear nominal properties, unlike the verbal gerunds discussed so
far. 17

5.5.1. Properties

First, the gerunds of this group can occupy typical argument positions:
the subject position (53a), the object position (53b), or the complement
position of non-temporal prepositions (53c):

(53) a. bo ha-xoref hifti'a 'otam.

coming the-winter surprised them
'The coming of the winter surprised them'

b. hem hisritu 'et §uv-o la-'ir.

they shot ACC returning-his to+the-town
'They shot his returning to town'

c. ha-mexkar nex§al biglal he'ader tixnun.

the-research failed because+of being+absent planning
'The research failed due to lack of planning'

Second, they can take a sei ('of') phrase, yielding a clitic doubling
configuration (recall that sei is strictly limited to nominal contexts):

(54) a. bo'-oj §el ha-xorefj hifti'a 'otam.

coming-his 0/ the-winter surprised them
'The coming of the winter surprised them'

b. hem hisritu 'et §UV-Oj §el ha-na'arj la-'ir.

they shot ACC returning-his 0/ the-guy to+the-town
'They shot the guy's returning to town'

Third, they can be negated by the nominal negation prefix 'i- (55),
just like derived nouns (31 a) and in contrast with verbal gerunds (31 b):

(55) a. 'i-bo'-oj §el ha-xorefj hifti'a 'otam.

NEG-Coming-his 0/ the-winter surprised them
'The fact that the winter did not come surprised them'

b. 'i-§uv-o hevix 'otam.

NEG-returning-his embarrassed them
'The fact that he did not return embarrassed them'

Fourth, they can be modified by adjectives:

(56)a. bo'-oj ha-mukdam §el ha-xorefj hifti'a 'otam.

coming-his the-early 0/ the-winter surprised them
'The early coming of the winter surprised them'

b. §UV-Oj ha-mafti'a §el danj hevix 'otam.

returning-his the-surprising 0/ Dan embarrassed them
'Dan's surprising returning embarrassed them'

At first glance, nominal gerunds also show some properties that are
not typical of nouns. First, nominal gerunds do not have a plural form;
but, as is weil known, some nouns do not pluralize either. Second and
more importantly, nominal gerunds (57a), unlike nouns (57b), cannot
appear in the free state (they allow a §el ('of') phrase only in a clitic
doubling construction «54»:

(57) a. • he'ader §el §lo§a marcim mi-ydivat ha-fakulta

being+absent 0/ three lecturers /rom-session the-faculty
yigrom le-dxiyat ha-ye§iva.
will+cause to-postponement the-session

b. he'adrut !el !lo!a marcim mi-ye!ivat ha-fakulta

absence 0/ three lecturers from-session the /aculty
tigrom le-dxiyat ha-ye!iva.
will+cause to-postponement the-session
'Absence of three lecturers from the faculty session will
cause postponement of the session'

Third, gerunds must overtly realize their subject (as a fuH noun
phrase (53a) or as a pronominal clitic (53b». (58a) is ungrammatical as
it contains a subjectless gerund; and compare (58a) with its grammatical
nominal counterpart (58b):

(58) a. * he'ader me-ha-tekes mutar.

being+absent from-the-ceremony allowed

b. he'adrut me-ha-tekes muteret.

absence from-the-ceremony allowed
'Absence from the ceremony is allowed'

Finally, nominal gerunds can never be accompanied by the article

(59a). However, although this is not typical of nouns (59b), it is a
characteristic of nouns whose head is in the construct state as in (59c)
(see chapter 2):

(59) a. * ha-he'ader dan hevix 'otam.

the-being+absent Dan embarrassed them

b. ha-he'adrut !el dan hevixa 'otam.

the-absence 0/ Dan embarrassed them
'Dan's absence embarrassed them'

c. * ha-he'adrut dan hevixa 'otam.

the-absence Dan embarrassed them

And indeed, just like nouns in the construct state, nominal gerunds have
a definiteness value, as shown by the behavior of their modifying
adjectives. Recall that Hebrew adjectives must agree with the head they
modify in number, gender, and definiteness. In (60a) (or (56» the
adjective must bear the article, showing that the gerund is definite. In
(60b) it cannot bear the article, showing that the gerund is indefinite:

(60) a. suv-o *(ha-)mafti'a hevix 'otam.

returning-his (the-)surprising embarrassed them
'His surprising returning em barrassed them'

b. he'ader yozma (*ha-)muxlat garam le-kiSlon

being+absent initiative (the-)complete caused to-Iai/ure
'A complete lack of initiative caused the failure of the

In brief, nominal gerunds have a definiteness value, appear in

argument positions, license seI ('of') phrases in clitic doubling
configurations, allow nominal negation, and are modifiable by
adjectives, which are all properties typical of nouns. Unlike regular
nouns, however, nominal gerunds cannot appear in the free state, must
phonetically realize their subject, and cannot realize their own article.
I now turn to explain this apparently ambiguous behavior.

5.5.2. Lexical nominalization

Nominal gerunds bear a definiteness value; hence, they are referential

and can therefore appear in argument positions. Given that and the fact
that they license seI ('of') phrases, can be modified by adjectives, and
negated by the nominal negation 'i-, they ought to contain a
syntactically present NP. According to the approach to nominalization
defended in this work, this means that nominal gerunds cannot be
inserted as verbs, because nouns cannot be derived from verbs in the
course of the syntactic derivation. In other terms, nominal gerunds are
lexically analyzed as nouns and do not involve a syntactically present
VP. If so, they should not show verbal properties.
It is impossible to prove that nominal gerunds do not license
structural accusative Case, as they are ne ver derived from transitive
verbs. However, just like event nouns (6Ib), nominal gerunds cannot be
modified by genuine adverbs (61a), which suggests that they do not
include a syntactically present verbal projection:

(61) a. * cet-o ~el dan maher min ha-xeder

going+out(GERUND)-his 0/ Dan quickly /rom the-room
hevix 'otam.
embarrassed them

b. *yeciat-o §el dan maher min ha-xe der

going+out(NOUN)-his 0/ Dan quickly /rom the-room
hevixa 'otam.
embarrassed them

In the same vein, nominal gerunds do not seem to allow (unstressed)

pronouns, which are contingent upon the presence of a verbal projection
(see section 5.1.1, examples (6a-b». (62a) contains a gerund clause
embedded under a temporal preposition; since the clause has verbal
structure, the (unstressed) dative pronoun 10 is allowed. An
homophonous gerundive form disallows 10 in (62b), as it is nominal and
therefore does not involve a verbal projection:

(62) a. 'im he'alm-am 10, parac dan be-vexi.

with disappearing-their to+him. burst Dan in-weeping
'When they disappeared, Dan burst into tears'

b. dan te'er 'et he'alm-am (*10).

Dan described ACC disappearing-their (to+him)
'Dan described their disappearing'

If indeed nominal gerunds are lexically analyzed as nouns and do not

involve a syntactically present verbal projection, the question arises why
they must reaIize their subject (S8a) and can appear neither in the free
state (S7a) nor with their own article (S9a), contrary to regular nouns.
One explanation might attempt to derive the obligatoriness of an overt
subject from the reading associated with nominal gerunds in the
following way. The subject of nominal gerunds is an internal argument
as they are always derived from unaccusative verbs. Now, if nominal
gerunds always have an event reading, it follows that their subject is
obligatory, as event nominals must realize their internat arguments (see
Grimshaw 1990 and chapter 3). But this does not explain why they can
appear neither in the free state nor with the article. Moreover, it seems
to me that at least the nominal gerund he'ader ('disappearing') can have
a result reading (it is hard to prove that this gerund has a result reading,
though, precisely because its subject is obligatory and the form cannot

pluralize, which is typical of event nominals).

The alternative solution I suggest in what follows is straightforward
and has the advantage of deriving the obligatoriness of an overt subject,
the unavailability of the free state, and the ban against the article from
one single property of the nominal gerund.
Nominal gerunds form a construct state with their subject, just Iike
verbal gerunds. When they realize their subject as a suffixal clitic (63a),
the latter is identical to the clitic that appears with nouns in the
construct state (63b) or with verbal gerunds (63c), as lalready
implicitly assumed above in example (53b) and others, where I used a
genitive pronoun to gloss the subject clitic (and see section 5.4.1):

(63) a. suv-i la-'ir hifti'a 'otam.

returning-my to+the-city surprised them
'My returning to the city surprised them'

b. beyt-i
'my house'

c. be-vakr-i 'oto, ra'iti 'et bit-o.

in-visiting-my him, saw(I) ACC daughter-his
'When I visited him, I saw his daughter'

Verbal gerunds are always in the construct state: they are obligatorily
endowed with Agr~en features, as discussed earlier. Suppose, then, that
the lexical reanalysis of certain gerunds as nouns (alongside their verbal
form) produces a set of nouns that can only be in the construct state
(that is;--endowed with Agrgen features), and do not have a free (non-
construct) morphological form. The obligatoriness of an overt subject,
the unavailability of the free state, and the ban against the article
automatically fall out. These are the basic properties of nouns in the
construct state, as extensively discussed in chapter 2 and briefly
summarized below.
First, the subject must be overt (64a), because Agr en must check its
features and does not Iicense pro (see section 5.4.5). ~econd, a nominal
gerund cannot appear in the free state, because if it realized its subject
in a seI ('of') phrase, the subject would receive Case within the seI
phrase, and the derivation would crash as Agrgen would not be able to
check its features (64b). The sole gerundive configuration that allows
the subject to surface in a seI phrase is the one in which the nominal

gerund hosts a pronominal clitic that checks Case in Agrgen' resulting

in a clitic doubling configuration (64c):

(64) a. bo • (ha-xoref)
coming (the-winter)
'the coming of the winter'

b. ·bo §el ha-xoref

coming 0/ the-winter

c. bO_IO·I §el ha-xoref,I

coming-it 0/ the-winter
'the coming of the winter'

Likewise, there is no wonder that nominal gerunds cannot bear the

article (59a), as heads of constructs can never realize their article.
Moreover, just like nouns in the construct state, nominal gerunds have
a definiteness value, as the behavior of their modifying adjectives
shows (60). This value must match the definiteness value of their
genitive subject. In (60a) the subject is pronominal and hence definite;
in (60b), in contrast, it is indefinite. Again, this is exactly what we
expect from nouns in the construct state (see chapter 2 sections 2.1.1
and 2.3.3), which corroborates the conclusion that nominal gerunds are
simply nouns that are obligatorily in the construct state.
To summarize, the gerundive form is obligatorily endowed with
Agrgen features. When it is limited to temporal contexts, its internal
structure is verbal (hence, the verbal properties), and when it can freely
occupy any DP-position, its internal structure is nominal (hence, the
nominal properties). The subject of both types of gerunds checks Case
within AgrgenP. It is obligatory because Agraen must check its features.
It is always overt because Agrgen does not license a null element. The
analysis thus captures both tne similarities and distinctions between
verbal and nominal gerunds. Under any sentential analysis of verbal
gerunds, the affinity between them and their nominal counterparts is
completely accidental, and their common properties require a different
explanation in each case.


The chapter has dealt with Hebrew gerunds. Most of the chapter has
studied gerunds that appear in adverbial adjuncts headed by temporal

prepositions. I have shown that gerunds do not involve a sentential

structure, in spite of what their morphological affinity to infinitives may
suggest; nor do they contain an NP, although the prepositions
introducing them are all DP taking prepositions. They are non-tensed
clausal structures headed by a D that is not inherently referential. Their
non-tensed nature makes them appropriate complements of D. The non-
referential character of their D restricts their distribution, as it forces
them to appear with temporal prepositions (or nouns), which provide
their D position with reference.
The last section of the chapter was devoted to nominal gerunds,
which constitute a small group of fixed elements derived from certain
unaccusative verbs. These gerunds, unlike their verbal counterparts, are
lexically analyzed as nouns, wh ich automatically derives their distinct
behavior. The symmetries the two types of gerunds display follow from
the fact that they are DPs whose head is obligatorily in the construct
state. The proposed lexical reanalysis of verbal gerunds as nominal
heads naturally captures the affinity between the two forms.
While nominal gerunds constitute an instance of lexical
nominalization, verbal gerunds instantiate syntactic nominalization,
which is, as defined throughout this work, a DP containing a non-tensed
verbal projection instead of the nominal projection it usually includes.
If this is correct, other constructions showing ambiguous categorial
nature, like Italian nominalized infinitives or English gerunds, may be
analyzed along similar lines.


1 Traditionally, the term "construct infinitive", or "infinitive construct" (Gesenius

1910) also refers to infinitives. As will become clear in section 5.1.2, infinitives
and gerunds differ in important respects.

2 Nominal gerunds, which clearly show nominal behavior, have the distribution of
noun phrases, as illustrated below (I defer discussion of nominal gerunds until
section 5.5):
(i) §uv-o hifti'a 'otL
returning-3MSSG surprised me
'His retuming surprised me'
In certain earlier stages of Hebrew, verbal gerunds had a wider distribution. In
Modem Hebrew, they appear in temporal adjuncts, as mentioned, and are typical
of written language, both in the press and in formal style.

3 Some speakers still find (7a) somewhat odd, but clearly better than (7b).

4 There are minor morphological differences between the verbal form of the
infinitive and that of the gerund. For example, stops that are fricativized in certain
environments undergo this change when they are the medial root consonant in a
gerund, but not in an infinitive (see Berman 1978, among others).

S be- ('in') is the only preposition that has some instrumental sense alongside its
temporal meaning of simultaneity.

6 I resume discussion of these two discrepancies in section 5.4.4.

7 Interestingly, prefixal prepositions do not always permit coordination of noun

phrases either:
(i) 'eßar le-hagi'a le-§am be-matos ve-sira.
possible to-arrive to-there in-plane and-boat
'It is possible to get there by plane and boat'
(ii) • ha-muzika ni§me'a ba-bayit ve-(ha-)gan.
the-music was+heard in+the-house and-(the-)garden
(iii) • hu biker be-'angliya ve-'erec ha-kode§.
he visited in-England and-Iand the-holy
I will not pursue this matter any further here. For our purposes, it suffices that
non-prefixal prepositional heads clearly permit coordination of gerund clauses.

8 This is exactly the behavior of nominal gerunds; although they are nominal, they
can never realize their artic1e, as they are always in the construct state (see section

9 Of course, languages differ on this matter. Some languages do not avail

themselves of AgrgenP, or do not have a parallel verbal form. In addition, notice
that AgrgenP, unlike Agrl, cannot appear as the complement of the auxiliary 'be'
in Hebrew complex tenses. This may be derived in at least two ways. First, if the
sentential subject checked genitive Case within AgrgenP, it would have no reason
to raise further and check the relevant features of Agra (movement being a Last
Resort operation). Second, it is plausible that AgrgenP belongs exclusively to the
DP system.

10 In chapter 2, we concluded that Agrgen appearing in noun phrases bears strong

DP-features, which trigger overt movement of the genitive DP. Given that, the
assumption that Agrgen of gerunds bears weak DP-features might seem somewhat
ad hoc. However, Agrgen of gerunds does not share with its nominal counterpart
precisely the same set of Xo_ and XP-features. First, it contains V-features rather
than N-features. Second, as will become clear below, its DP-features are different;
they do not include definiteness features, arguably the strong element.

11 In order to end up with an LF-representation in which the subject IS IR

SpecAgrgcnP and the object in SpecAgroP' we must exclude nested movements of
the subject (to SpecAgroP) and the object (to SpecAgrgenP). This can be obtained
in exactly the same way that parallel nested movements are ruled out in the
sentence (see Chomsky 1993 for discussion).

12 Italian nominalized infinitives, wh ich are infinitives embedded und er a

phonetically realized D, can appear in argument positions (thanks to Teresa Guasti
for the example):
(i) II leggere i libri fa male alla vista
the read(INF) the boolcs hurts to+the view
If in Italian definiteness is not a feature of the noun (which is checked in D) but
rather a feature of the free standing article, D can be referential even when it does
not contain a nominal head. Being referential by itself, D can head a proposition
that can freely occur as an argument. In Hebrew, in contrast, checking of
definiteness necessarily involves a nominal head.

13 The preposition can determine the time reference of D via a process of

unselective binding (in Heim's 1982 sense, and see also Pesetsky 1987). As
unselective binders may bind more than one variable, the fact that two gerund
clauses can be coordinated under one preposition is not surprising (see (21»; the
same preposition is able to bind both D positions.
Note that the fact that D cannot select tensed clauses is not incongruous with the
claim that it can acquire a temporal reference. D cannot have as complements
clauses containing tense features, but its reference can be anchored in time if it is
not otherwise specified.

14 Hazout (1992) also mentions me'az ('since') and 'ad ('until') as two additional
temporal prepositions that cannot take gerund clauses, but take noun phrases as
complements. I disagree with Hazout on this matter, as is clear from example (2)
and (i):
(i) 'ad hibaxar-o la-tafkid, 'af 'exad 10 lava§ xalifot.
until being+elected to+the-job. nobody not wore suits
'Until he was elected to the job, nobody wore suits'
(Berman 1978)

Additionally, Hazout does not mention the prefixal preposition ke- ('as'), which,
according to my judgments, is possible with gerund clauses:

(ii) ke-cet-o min ha-misrad, hexel la-redet ge§em.

as-going+out-his from the-office. started to-fall rain
'As he left the office, it started raining'
This preposition practically never has a temporal interpretation when it takes a
noun phrase complement (see the ensuing discussion).

IS Thanks to Tanya Reinhart for pointing this out to Me. Beside their occurrence
in temporal constructs, gerund clauses can also appear in constructs specifying a
non-temporal aspect of the event (i), or the event itself (ii):
(i) hu te'er 'et 'ofen xalof ha-siyur.
he described ACC manner passing the-patrol
'He described the manner in which the patrol passed'
(ii) 'ecem havi'-o 'ota le-tel-aviv sime'ax 'ota.
essence bringing-his her to-Tel-A viv made+happy her
'The very fact that he brought her to Tel-Aviv made her happy'
These constructs are Iikely to provide the gerund with reference (much Iike the
temporal construct in (49», which is needed for the gerund to appear in an
argument position.
Notice that the availability of such constructs reinforces the claim that the head
noun of the construct state is specified intrinsically with regard to definiteness
(chapter 2, section 2.3.3) rather than inherits definiteness from the genitive member
of the construct (note 21 of the same chapter), because gerunds do not have a
[±definite] value to transmit to the head of the construct.

16 Recall further that gerunds do not allow object clitics, unlike infinitives (see
section 5.2.1). One possibility that comes to mind is that these clitics can only
appear when the structure contains the higher sentential categories TP and AgrsP'
However, the fact that they are possible in semi-relatives (i), which do not contain
these sentential categories (see chapter 4), argues against this solution:
(i) yadid ha-makir-eni §anim rabot...
friend the-knowing-me years numerous
'A friend knowing me for numerous years .. .'
Rather, it seems that the unavailability of object clitics is related to the presence
of AgrgenP. In chapter 2, it is suggested that Hebrew pronominal clitics are the
materialization of the Case features of their host. Suppose a head materializing
verbal Case features (like structural accusative features) cannot be associated with
Agrgen features. The fact that gerunds cannot realize object clitics would follow.
I do not discuss this incompatibility any further here, but I think ·it offers a starting
point for an explanation. Note that if gerund clauses and infinitivals involved the
same sentential structure, the fact that gerunds disallow object clitics would be
completely unexpected.

17 Instances of nominal gerunds include: bo ('coming'), suv ('returning'), cet

('going+out'), he'ader ('being+absent'), he'a/em ('disappearing'), and a few others.

The following is a list of abbreviations used in the glosses and text:

ACC = accusative
AOR = agreement
CL = clitic
EXT = external
FM = feminine
FUT = future
OEN = genitive
INT = internal
MS = masculine
NEO = negation
NOM = nominative
0= object
OA = object agreement
P = participle
PastP = past participle
PL = plural
S = subject
SA = subject agreement
so = singular
The transcription of Hebrew reflects Israeli Hebrew (slow) speech. For
convenience, it is a compromise between phonemic and phonetic transcription. In
general, the transcriptions of other languages follow those of the cited works, with
no attempt at standardization.


Abney, Steven. 1987. The English Noun Phrase in its Sentential Aspect. Doctoral
dissertation, MIT, Cambridge, Mass.
Agmon-Fruchtman, Maya. 1982. Hayadua vehasatum. Papyrus, Tel Aviv.
Anderson, Mona. 1979. Noun Phrase Structure. Doctoral dissertation, University
of Connecticut.
Anderson, Mona. 1983-84. Prenominal Genitive NPs. The Linguistic Review 3, 1-
Baker, Mark. 1988. Incorporation: A Theory o/Grammatical Function Changing.
The University of Chicago Press, Chicago.
Bat-EI, Outi. 1986. Extraction in Modem Hebrew Morphology. Master's thesis,
UCLA, Los Angeles.
Belletti, Adriana. 1990. Generalized Verb Movement: Aspects 0/ Verb Syntax.
Rosenberg and Sellier, Torino.
Belletti, Adriana. 1993. Case Checking and Clitic Placement: Three Issues on
(ItalianlRomance) Clitics. Geneva Generative Papers 1.2, 101-118.
Belletti, Adriana and Luigi Rizzi. 1988. Psych-Verbs and Theta Theory. Natural
Language and Linguistic Theory 6, 291-352.
Belletti, Adriana and Ur Shlonsky. 1995. The Order of Verbal Complements: A
Comparative Study. Natural Language and Linguistic Theory 13, 489-526.
Ben-Shalom, Dorit and Ziva Weijler. 1991. Reflexivity in Hebrew. Ms., UCLA,
Los Angeles and Tel Aviv University.
Berman, Ruth. 1978. Modern Hebrew Structure. University Publishing Projects,
Tel Aviv.
Bernstein, Judy. 1993. Topics in the Syntax ofNominal Structure across Romance.
Doctoral dissertation, CUNY, New York.
Borer, Hagit. 1981. On the Definition of Variables. Journal 0/ Linguistic Research
1.3, 16-40.
Borer, Hagit. 1984. Parametric Syntax. Foris, Dordrecht.
Borer, Hagit. 1989. On the Morphological Parallelism between Compounds and
Constructs. In Geert Booij and Jaap van Marle, eds., Yearbook 0/
Morphology 1, Foris, Dordrecht, 45-65.
Borer, Hagit. 1995. The Ups and Downs of Hebrew Verb Movement. Natural
Language and Linguistic Theory 13, 527-606.
Borer, Hagit. In progress. Parallel Morphology. MIT Press, Cambridge, Mass.
Bottari, Piero. 1990. Livelli di rappresentazione lessicale: complementazione
nominale e complementazione frasale. Doctoral dissertation, University of
Bouchard, Denis. 1984. On the Conte nt 0/ Empty Categories. Foris, Dordrecht.
Burzio, Luigi. 1981. Intransitive Verbs and Italian Auxiliaries. Doctoral
dissertation, MIT, Cambridge, Mass.


Burzio, Luigi. 1986. Italian Syntax: A Government-Binding Approach, Reidel,

Chomsky, Noam. 1957. Syntactic Structures. Mouton, The Hague.
Chomsky, Noam. 1965. Aspects ofthe Theory ofSyntax. MIT Press, Cambridge,
Chomsky, Noam. 1970. Remarks on Nominalization. In Roderiek laeobs and Peter
Rosenbaum, eds., Readings in English Tranformational Grammar, Ginn and
Co., Waltham, Mass, 184-221.
Chomsky, Noam. 1973. Conditions on Transformations. In Stephen Anderson and
Paul Kiparsky, eds., A Festschrift for Morris Halle, Holt, Rinehart and
Winston, New York, 232-286.
Chomsky, Noam. 1975. Reflections on Language. Pantheon, New York.
Chomsky, Noam. 1976. Conditions on Rules of Grammar. Linguistic Analysis 2,
Chomsky, Noam. 1980. On Binding. Linguistic Inquiry 11, 1-46.
Chomsky, Noam. 1981. Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, Noam. 1982. Some Concepts and Consequences of the Theory of
Government and Binding. MIT Press, Cambridge, Mass.
Chomsky, Noam. 1986a. Knowledge of Language: Its Nature, Origin, and Use.
Praeger, New York.
Chomsky, Noam. 1986b. Barriers. MIT Press, Cambridge, Mass.
Chomsky, Noam. 1991. Some Notes on Eeonomy of Derivation and
Representation. In Robert Freidin, ed., Principles and Parameters in
Comparative Grammar, MIT Press, Cambridge, Mass., 417-454.
Chomsky, Noam. 1993. A Minimalist Program for Linguistie Theory. In Kenneth
Haie and Samuellay Keyser, eds., The View /rom Building 20, MIT Press,
Cambridge, Mass., 1-52.
Chomsky, Noam. 1994. Bare Phrase Strueture. MIT Occasional Papers in
Linguistics 5, MIT, Cambridge, Mass. [Publishcd 1995. In Gert Webelhut,
ed., Government and Binding Theory and the Minimalist Program,
Blaekwell, Oxford.]
Chomsky, Noam. 1995. The Minimalist Program. MIT Press, Cambridge, Mass.
Chomsky, Noam and Howard Lasnik. 1977. Filters and Contro!. Linguistic Inquiry
Chomsky, Noam and Howard Lasnik. 1993. Prineiples and Parameters Theory. In
Joachim Jacobs, Amim von Stechow, Wolfgang Sternefeld and Theo
Vennemann, eds., Syntax: An International Handbook of Contemporary
Research, Walter de Gruyter, Berlin, 506-569.
Cinque, Guglielmo. 1980. On Extraetion from NP in Italian. Journal of Italian
Linguistics 5, 47-99.
Cinque, Guglielmo. 1981. Sulla nozione di soggetto di SN in italiano. Cu/tura
Neo/atina 41.
Cinque, Guglielmo. 1988. On si Construetions and the Theory of arb. Linguistic
Inquiry 19, 521-581.
Cinque, Guglielmo. 1990. Types of A'-Dependencies. MIT Press, Cambridge, Mass.
Cinque, Guglielmo. 1992. Funetional Projections and N-movement within the DP.
aLOW News/etter 28.
Cinque, Guglielmo. 1993. On the Evidence for Partial N-movement in the
Romanee DP. Ms., University of Venice.

Cinque, Guglielmo. 1994. Posizione dei soggetto nel DP italiano. In Lorenzo

Renzi, ed., Omaggio a Florica Dimitrescu e Alexandru Niculescu, Uni press,
Clark, Robin. 1990. Structured Theta-Grids and the Syntax of Nominals. Ms.,
University of Geneva.
Cohen, Talila. 1992. A Determiner Phrase Approach to Hebrew Nominals.
Master's thesis, Bar Uan University, Ramat Gan.
Couquaux, Daniel. 1981. French Predication and Linguistic Theory. In Robert.
May and lan Koster, eds., Levels 0/ Syntactic Representation, Foris,
Dordrecht, 33-64.
Davidson, Donald. 1967. The Logical Form of Action Senten ces. In Nicolas
Rescher, ed., The Logic 0/ Decision and Action, University of Pittsburgh
Press, Pittsburgh, 81-95.
Delsing, Lars Olof. 1988. The Scandinavian Noun Phrase. Working Papers in
Scandinavian Syntax 42, 57-79.
Doron, Edit. 1983. Verbless Predicates in Hebrew. Doctoral dissertation,
University of Texas.
Doron, Edit. 1989. Derived nominals in Hebrew. To appear in lohn Payne, ed.,
The Structure 0/ Noun Phrases, Mouton, The Hague.
Dowty, David. 1979. Word Meaning and Montague Grammar. Reidel, Dordrecht.
Dowty, David. 1989. On the Semantic Content of the Notion 'Thematic Role'. In
Gennaro Chierchia, Barbara Partee and Raymond Turner, eds., Properties.
Types. and Meaning, Kluwer, Dordrecht, 69-129.
Emonds,loseph. 1978. The Verbal Complex V'-V in French. Linguistic Inquiry 9,
Emonds, loseph. 1979. Appositive Relatives Have No Properties. Linguistic
Inquiry 10, 211-243.
En~, Mürvet. 1987. Anchoring Conditions for Tense. Linguistic Inquiry 18, 633-
Epstein, Samuel. 1987. Empty Categories and their Antecedents. Doctoral
dissertation, University of Connecticut.
Everett, Daniel. 1989. Clitic Doubling, Reflexives, and Word Order Alternations
in Yagua. Language 65, 339-372.
Fassi Fehri, Abdelkader. 1989. Generalized IP Structure, Case and VS Word
Order. In Itziar Laka and Anoop Mahajan, eds., MIT Working Papers in
Linguistics 10, 75-113.
Fassi Fehri, Abdelkader. 1993. Issues in the Structure 0/ Arabic Clauses and
Words. Kluwer, Dordrecht.
Friedemann, Marc-Ariel. 1993-94. The Underlying Position of External Arguments
in French: A Study in Adult and Child Grammar. Language Acquisition 3,
Friedemann, Marc-Ariel and Tal Siloni 1993. Agrobject is not Agrparticiple' To
appear in The Linguistic Review 14.
Fukui Naoki and Margaret Speas. 1986. Specifiers and Projection. In Naoki Fukui,
Tova Rapoport and Elizabeth Sagey, eds., MIT Working Papers in
Linguistics 8, 128-172.
Gesenius, Wilhelm. 1910. Hebrew Grammar. Edited by E. Kautsch, revised by
A.E. Cowley, Clarendon Press, Oxford.

Giorgi, Alessandra and Giuseppe Longobardi. 1991. The Syntax 0/ Noun Phrases:
Configuration. Parameters. and Empty Categories. Cambridge University
Press, Cambridge.
Glinert, Lewis. 1989. The Grammar 0/ Modern Hebrew. Cambridge University
Press, Cambridge.
Grimshaw, Jane. 1990. Argument Structure. MIT Press, Cambridge, Mass.
Grimshaw, Jane. 1991. Extended Projection. Ms., Brandeis University.
Groos, Anneke and Henk van Riemsdijk. 1979. Matching Effects in Free relatives:
A Parameter of Core Grammar. In Adriana Belletti, Luciana Brandi and
Luigi Rizzi, eds., Theory 0/ Markedness in Generative Grammar, Scuola
Normale Superiore, Pisa, 171-216.
Guasti, Maria Teresa. 1992. Causative and Perception Verbs. Doctoral dissertation,
University of Geneva.
Gueron, Jacqueline. 1993. Sur la syntaxe du temps. Langue Franfaise 100, 102-
Gueron, Jacqueline and Teun Hoekstra. 1989. T-Chains and the Constituent
Structure of Auxiliaries. In Anna Cardinaletti, Guglielmo Cinque, and
Giulana Giusti, eds., Constituent Structure, Foris Dordrecht, 35-99.
Haegeman, Liliane. 1992. Generative Syntax: Theory and Description. A Case
Study /rom West Flemish. Cambridge University Press, Cambridge.
Hazout, nan. 1990. Verbal Nouns: Theta-theoretic Studies in Hebrew and Arabic.
Doctoral dissertation, University of Massachusetts, Amherst.
Hazout, nan. 1992. Tbe Verbal Gerund in Modern Hebrew. Natural Language and
Linguistic Theory 10, 523-553.
Hazout, nan. 1995. Action Nominalizations and the Lexicalist Hypothesis. Natural
Language and Linguistic Theory 13, 355-404.
Heim, ·Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases.
Doctoral dissertation, University of Massachusetts, Amherst.
Hendrick, Randall. 1990. Operator Movement within NP. In A.L. Halpern, ed.,
Proceedings 0/ WCCFL 9, 248-261.
Higginbotham, James. 1983. Logical Form, Binding, and Nominals. Linguistic
Inquiry 14, 395-420.
Higginbotham, James. 1985. On Semanties. Linguistic Inquiry 16, 547-593.
Hoekstra, Teun. 1984. Transitivity: Grammatical Relations in Government-Binding
Theory. Foris, Dordrecht.
Huang, James. 1982. Logical Relations in Chinese and the Theory of Grammar.
Doctoral dissertation, MIT, Cambridge, Mass.
Hurtado, Alfredo. 1984. On the Properties of LF. Cornell Working Papers in
Linguistics 5, 121-149.
Jackendoff, Ray. 1977. X'-Syntax: A Study 0/ Phrase Structure. MIT Press,
Cambridge, Mass.
Jaeggli, Osvaldo. 1982. Topics in Romance Syntax. Foris Dordrecht.
Jaeggli, Osvaldo. 1986. Three Issues in the Theory of Clitics: Case, Doubled NPs,
and Extraction. In Hagit Borer, ed., Syntax and Semantics 19, 15-42.
Kayne, Richard. 1975. F'rench Syntax: The Tran/ormational Cycle. MIT Press,
Cambridge, Mass.
Kayne, Richard. 1980. De certaines differences entre le fran~ais et I'anglais.
Langages 60, 47-64.
Kayne, Richard. 1984. Connectedness and Binary Branching. Foris, Dordrecht.

Kayne, Richard. 1985. L'accord du participe passe en fran~ais et en italien.

Modeles Iinguistiques VII, 73-91.
Kayne, Richard. 1989a. Facets of Romance Past Participle Agreement. In Paola
Beninca, ed., Dialect Variation and the Theory 0/ Grammar, Foris,
Kayne, Richard. 1989b. Null Subjects and Clitic Climbing. In Osvaldo Jaeggli and
Ken Safir, eds., The Null Subject Parameter, Kluwer, Dordrecht.
Kayne, Richard. 1993. Toward a Modular Theory of Auxiliary Selection. Studia
Linguistica 47, 3-31.
Kayne, Richard. 1994. The Antisymmetry 0/ Syntax. MIT Press, Cambridge, Mass.
Koopman, Hilda and Dominique Sportiche. 1982. Variables and the Bijection
Principle. The Lingustic Review 2, 139-160.
Koopman, Hilda and Dominique Sportiche. 1991. The Position of Subjects. Lingua
85, 2Il-258.
Kornfilt, Jaletin. 1984. Case Marking, Agreement, and Empty Categories in
Turkish. Doctoral dissertation, Harvard University, Cambridge, Mass.
Kuroda, Shige-Yuki. 1988. Whether We Agree Or Not: A Comparative Syntax of
Engtish and Japanese. Lingvisticae Investigationes 12, 1-47.
Lalea, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories
and Projections. Doctoral dissertation, MIT, Cambridge, Mass.
Larson, Richard. 1985. Bare NP-Adverbs. Linguistic Inquiry 16, 595-621.
Lasnik, Howard. 1988. Subjects and the 6-criterion. Natural Language and
Linguistic Theory 6, 1-17.
Lebeaux, David. 1986. The Interpretation of Derived Nominals. In Anne Farley,
Peter Farley and Karl-Erik McCullough, eds., Proceedings 0/ CLS 22, 231-
Lees, Robert. 1960. The Grammar 0/ English Nominalizations. Mouton, The
Longobardi, Giuseppe. 1994. Reference and Proper Names: A Theory of N-
Movement in Syntax and Logieal Form. Linguistic Inquiry 25, 609-665.
May, Robert. 1981. Movement and Binding. Linguistic Inquiry 12, 215-243.
McCarthy, John. 1979. Formal Problems in Semitic Phonology and Morphology.
Doctoral dissertation, MIT, Cambridge, Mass.
McCloskey, James. 1991. Clause Structure, Ellipsis and Proper Government in
Irish. Lingua 85, 259-302.
Milner, Jean-Claude. 1982. Ordre et raisons de langue. Seuil, Paris.
Mohammad, Mohammad. 1988. On the Parallelism between IP and DP. In Hagit
Borer, ed., Proceedings 0/ WCCFL 7, 241-254.
Ouhalla, Jamal. 1988. The Syntax of Head Movement: A Study of Berber. Doctoral
dissertation, University College, London.
Partee, Barbara. 1976. Montague Grammar and Transformational Grammar.
Linguistic Inquiry 6, 203-300.
Penner, Zvi and Manuela Schönenberger. 1993. The Distribution of Nominal
Agreement Features in German Dialects and the Strong DP/CP Parallelism
Hypothesis. Ms., University of Bern and University of Geneva.
Pesetsky, David. 1987. Wh-in-Situ: Movement and Unselective Binding. In Eric
Reuland and Alice ter Meulen, eds., The Representation 0/ (In)dejinitness,
MIT Press, Cambridge Mass., 98-129.

Pollock, Jean-Yves. 1989. Verb Movement, UG and the Structure of IP. Linguistic
Inquiry 20, 365-424.
Qafisheh, Hamdi. 1977. A Short Relerence Grammar 01 Gulf Arabic. The
University of Arizona Press, Tueson, Arizona.
Quicoli, Carlos. 1976. Conditions on Clitic-Movement in Portuguese. Linguistic
Analysis 2, 199-223.
Rapoport, Tova. 1987. Copular, Nominal and Small Clauses: A Study of Israeli
Hebrew. Doctoral Dissertation, MIT, Cambridge, Mass.
Rappaport, Malka and Edit Doron. 1990. Hebetim semantiyim §el §em hape'ula.
Paper presented at the Workshop on Hebrew Grammar, The Hebrew
University of Jerusalem.
Reinhart, Tanya. 1976. The Syntactic Domain of Anaphora. Doctoral dissertation,
MIT, Cambridge, Mass.
Reinhart, Tanya. 1983. Anaphora and Semantic Interpretations. The University of
Chicago Press, Chicago.
Reinhart, Tanya and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24, 657-
van Riemsdijk, Henk. 1983. Tbe Case of German Adjectives. In Frank Henny and
Barry Richards, eds., Linguistic Categories: Auxiliaries and Related Puzzles
I, Reidel, Dordrecht, 223-252.
van Riemsdijk, Henk and Edwin Williams. 1986. Introduction to the Theory 01
Grammar. Mit Press, Cambridge, Mass.
Ritter, Elizabeth. 1987. NSO Noun Phrases in Modem Hebrew. In J. McDonough
and B. Plunkett, eds., Proceedings olNELS 17, 521-537.
Ritter, Elizabeth. 1988. A Head-Movement Approach to Construct-State Noun
Phrases. Linguistics 26, 909-929.
Ritter, Elizabeth. 1991. Two Functional Categories in Noun Phrases: Evidence
from Modem Hebrew. In Susan Rothstein, ed., Syntax and Semantics 25,
Rivas, Alberto. 1977. A Theory of Clitics. Doctoral dissertation, MIT, Cambridge,
Rizzi, Luigi. 1982. Issues in Italtan Syntax. Foris, Dordrecht.
Rizzi, Luigi. 1986a. Null Objects in Italian and the Theory of Pro. Linguistic
Inquiry 17, 501-557.
Rizzi, Luigi. 1986b. On the Status of Subject Clitics in Romance. In Osvaldo
Jaeggli and Carmen Silva-Corvalan, eds., Studies in Romance Linguistics,
Foris, Dordrecht, 391-420.
Rizzi, Luigi. 1987. On the Structural Uniformity of Syntactic Categories.
Proceedings olthe Second Conference on the Basque Language, Vol.l, 89-
Rizzi, Luigi. 1990. Relativized Minimality. MIT Press, Cambridge Mass.
Rizzi, Luigi. 1991a. Residual Verb Second and the Wh-Criterion. To appear in
Adriana Belletti and Luigi Rizzi, eds., Parameters and Functional Heads:
Essays in Comparative Syntax, Oxford University Press, Oxford.
Rizzi, Luigi. 1991b. Proper Head Govemment and the Definition of A Positions.
GLOW Newsletter 26.
Rizzi, Luigi. 1993. Some Notes on Romance Cliticization. Ms., University of

Rizzi, Luigi. 1994. Some Notes on Linguistic Theory and Language Development:
The Case of Root Infinitives. Language Acquisition 3, 371-393.
Roberts, Ian. 1987. The Representation o/Implicit and Dethematized Subjects.
Foris, Dordrecht.
Roberts, Ian. 1993. Verbs and Diachronie Syntax. Kluwer, Dordrecht.
Roeper, Thomas. 1987. Implicit Arguments and the Head-Complement Relation.
Linguistie Inquiry 18, 267-310.
Rosen, Haim. 1957. Ivrit tova: iyunim betaxbir. Kiryat Sepher, lerusalem.
Rosen, Haim. 1962. A Textbook 0/ Israeli Hebrew. The University of Chicago
Press, Chicago.
Rothstein, Susan. 1983. The Syntactic Forms of Predication. Doctoral dissertation,
MIT, Cambridge, Mass.
Rothstein, Susan. 1991. Syntactic Predication: A Syntactic Primitive or a Thematic
Relation. Ms., Bar lIan University, Ramat Gan and MIT, Cambridge, Mass.
Rothstein, Susan. 1995. Small Clauses and Copular Constructions. In Anna
Cardinaletti and Maria Teresa Guasti, eds., Syntax and Semanties 28, 27-48.
Saftr, Ken. 1987. The Syntactic Projection of Lexical Thematic Structure. Natural
Language and Linguistie Theory 5, 561-601.
Salvi, Giancarlo. 1983. L'infinito con I'articolo e la struttura deI SN. Rivista di
grammatiea generativa 8, 197-225.
Shlonsky, Ur. 1987. Null and Displaced Subjects. Doctoral dissertation, MIT,
Cambridge, Mass.
Shlonsky, Ur. 1988. Govemment and Binding in Hebrew Nominals. Linguisties 26,
Shlonsky, Ur. 1990. Construct State Nominals and Universal Grammar. Ms.,
University of Haifa.
Shlonsky, Ur. 1991. Strategies of Wh-movement in Palestinian Arabic. Ms.,
University of Geneva.
Shlonsky, Ur. 1994. Semitic Clitics. Geneva Generative Papers 2.1, I-ll.
Shlonsky, Ur. to appear. Clause Structure and Word Order in Hebrew: an Essay
in Comparative Semitie Syntax. Oxford University Press, Oxford.
Shoshani, Ronit. 1980. The Object Marker in Hebrew in Intransitive Contexts.
Ms., Tel Aviv University.
Sigurösson, Halld6r. 1989. Verbal Syntax & Case in Icelandic, in a Comparative
GB Approach. Doctoral dissertation, University of Lund.
Siloni, Tal. 1990a. On the Parallelism between CP and DP: The Case of Hebrew
Sem i-Relatives. In lohn van Lit, Rene Mulder and Rint Sybesma, eds.,
Proeeedings 0/ LCJL I, 135-153.
Siloni, Tal. 1990b. Hebrew Noun Phrases: Generalized Noun Raising. To appear
in Adriana Belletti and Luigi Rizzi, Parameters and Functional Heads:
Essays in Comparative Syntax, Oxford University Press, Oxford.
Siloni, Tal. 1991a. Noun Raising and the Structure of Noun Phrases. In Jonathan
Bobaljik and Tony Bures, eds., MIT Working Papers in Linguistics 14,
Siloni, Tal. 1991b. On Variables. Ms., University of Geneva.
Siloni, Tal. 1994a. Noun Phrases and Nominalizations. Doctoral dissertation,
University of Geneva.

Siloni, Tal. 1994b. On Hebrew Event Nominals: Against a VP-analysis. In Rhonna

Buchalla and Anita Mittwoch, eds., Proceedings o/IATL 1, Akademon,
Jerusalem, 145-164.
Siloni, Tal. 1995. On Participial Relatives and Complementizer DO: A Case Study
in Hebrew and French. Natural Language and Linguistic Theory 13, 445-
Sportiche, Dominique. 1988. A Theory of Floated Quantifiers. Linguistic Inquiry
19, 425-450.
Sportiche, Dominique. 1992. Clitic Constructions. Ms., UCLA, Los Angeles.
Sproat, Richard. 1985. Welsh Syntax and VSO Structure. Natural Language and
Linguistic Theory 3, 173-216.
StowelI, Tim. 1981. Origins of Phrase Structure. Doctoral dissertation, MIT,
Cambridge, Mass.
StowelI, Tim. 1982. The Tense of Infinitives. Linguistic Inquiry 19, 561-570.
StowelI, Tim. 1989. Subjects, Specifiers and X'-Theory. In Mark Baltin and
Anthony Kroch, eds., Alternative Conceptions 0/ Phrase Structure,
University of Chicago Press, Chicago, 232-262.
StowelI, Tim. 1991. Determiners in NP and DP. In Denis Bouchard and Katherine
Leffel, eds., Views on Phrase Structure. Kluwer, Dordrecht, 37-56.
Strozer, Judith. 1976. Clitics in Spanish. Doctoral dissertation, UCLA, Los
Szabolcsi, Anna. 1983-84. The Possessor that Ran Away from Home. The
Linguistic Review 3, 89-102.
Szabolcsi, Anna. 1987. Functional Categories in the Noun Phrase. In Istvan
Kenesei, ed., Approaches to Hungarian 2, 167-189.
Szabolcsi, Anna. 1989. Noun Phrases and Clauses: Is DP Analogous to IP or CP?
To appear in John Payne, ed., The Structure 0/ Noun Phrases. Mouton, The
Szabolcsi, Anna. 1992. Subject Suppression or Lexical PRO? Tbe Case of Derived
Nominals in Hungarian. Lingua 86, 149-176.
Szabolcsi, Anna. 1994. Tbe Noun Phrase. In Ferenc Kiefer and Katalin Kiss, eds.,
Syntax and Semantics 27, 179-274.
Taraldsen, Knut Tarald. 1986. Som and the Binding Theory. In Lars Hellan and
Kirsti Koch Christensen, eds., Topics in Scandinavian Syntax, Kluwer, '
Dordrecht, 149-184.
Taraldsen, Knut Tarald. 1990. D-projections and N-projections in Norwegian. In
Joan Mascara and Marina Nespor, eds., Grammar in Progress, 419-431.
Tellier, Christine. 1991. Licensing Theory and French Parasitic Gaps. Kluwer,
Tenny, Carol 1987. Grammaticalizing Aspect and Affectedness. Doctoral
dissertation, MIT, Cambridge, Mass.
Valois, Daniel. 1991. Tbe Internal Syntax ofDP. DoctoraI dissertation, UCLA, Los
Vendler, Zeno. 1967. Linguistics in Philosophy. Cornell University Press, Ithaca.
Vergnaud, Jean-Roger. 1985. Dependances et niveau de representalton en
syntaxe. Benjamins, Amsterdam.

Vikner, Steno 1991. Relative Der and Other CO Elements in Danish. Lingua 84,
Williams, Edwin. 1975. Small Clauses in English. In lohn Kimball, ed., Syntax
and Semantics 4, 249-273.
WiIIiams, Edwin. 1981. Argument Structure and Morphology. The Linguistic
Review I, 81-114.
Williams, Edwin. 1985. PRO and the Subject of NP. Natural Language and
Linguistic Theory 3,297-315.
Zaenen, Annie, loaD Maling and Höskuldur Thräinsson. 1985. Case and
Grammatical Functions: The Icelandic Passive. Natural Language and
Linguistic Theory 3, 441-484.
Zanuttini, Raffaela. 1990. On the Relevance of Tense for Sentential Negation. To
appear in Adriana Belletti and Luigi Rizzi, Parameters and Functional
Heads: Essays in Comparative Syntax, Oxford University Press, Oxford.
Zubizarreta, Maria Luisa. 1987. Levels 0/ Representation in the Lexicon and in the
Syntax. Foris, Dordrecht.
Zucchi, Alessandro. 1989. The Language of Propositions and Events: Issues in the
Syntax and Semantics of Nominals. Doctoral dissertation, University of
Massachusetts, Amherst.

Abney, Steven 6, 9, 19, 35-36, 39, Dowty, David 3,68

63n24, 146
Agnon-Fruehtman, Maya 61n5 Emoods, Joseph 134, 144
Anderson, Mona 3, 62n16, 65, 71 En y, Mürvet 8, 156n25
Epsteio, Samuel 149
Baker, Mark 6, 11, 19, 33, 75, Everett, Daoiel 54
Bat-EI, Outi 38
Belletti, Adriana 15, 17,30,38, Fassi Fehri, Abdelkader 6, 19, 32,
64n23, 81, 83, 102n5, 106025, 61n12, 37, 41, 62n19, 62n21
115, 118, 133, 136, 143, 155n23, Friedemaoo, Mare-Ariel 15-17, 38,
156n27, 177 52,81, 105n14, 118-119, 133,
Ben-Shalom, Dorit 61 n8 136, 181-182
Berman, Ruth 22,90, 163, 176, Fukui, Naoki 6, 30
179, 190, 198n4, 199n14
Bernstein, Judy 6, 61nll, 62n13 Gesenius, Wilhelm 197n 1
Borer, Hagit 2, 4, 9, 29, 31-32, 38, Giorgi, Alessandra 7, 9, 29-30,
53, 55, 58, 61n5, 63n27, 65-66, 62016, 94, 99, 101, 102n2,
74, 79, 83, 85, 89, 91, 104n14, G1inert, Lewis 168, 172, 190
105n19, 106n20, 115, 149, 151n3, Grimshaw, Jaoe 3, 7, 9, 63n24, 65-
157n30 68, 70-71, 86,92,95-96, 101n2,
Bottari, Piero 146 102n3, 10204, 106021, 106023,
Bouehard, Denis 149 107026, 157028, 194
Burzio, Luigi 94-95, 106025, 119, Groos, Aooeke 151 n3
140, 155n23 Guasti, Maria Teresa 62n16, 199n12
Gueron, Jaequeline 156n25
Chomsky, Noam 1-3,5-6,9-15, 17,
19,32,38,41-42,58,61nl0, Haegemao, Liliane 38
62n15, 64n27, 66, 80, 83, 86, 91- Harari, Reuven 106n24
92, 140, 148-149, 156n23, Hazout, Bao 6, 9, 19, 66, 74, 77,
156n27, 157n30, 182, 199n11 89-90, 163, 166, 168-172, 184,
Cinque, Guglielmo 3, 6, 50, 61n11, 190, 199014
62n13, 63n24, 64027, 91, 156023 Heim, Ireoe 199013
Clark, Robin 10202 Hendriek, Raodall 153n9, 147
Cohen, TaIila 19 Higginbotham, James 3, 13, 67, 75
Couquaux, Daniel 64n27 Hoekstra, Teun 17, 133, 156n25
Horvath, Julia 62017
Davidson, Donald 75 Huang, James 64n27
Delsing, Lars Olof 6, 19 Hurtado, Alfredo 55
Doron, Edit 22, 60n4, 87, 112-113,
115 Jaekendoff, Ray 5, 13, 19


Jaeggli, Osvaldo 55-56 156027, 147, 158032, 177, 188

Roberts, lan 14, 106023
Kayoe, Richard 3, 14-15, 55, 64027, Roepcr, Thomas 70
94, 118, 15308, 136, 157030 Ros~o, Haim 22, 6004, 179
Koopman, Hilda 14, 30, Rothsteio, Susan 10202, 112, 128
Komfilt, Jaklio 35
Kuroda, Shige-Yuki 30 Safir, Keo 3
Salvi, Giancarlo 146
Laka,ltziar 15104, 156027, 177 Schöocobergcr, Manuela 6
Larsoo, Richard 10308 Shloosky, Ur 53, 64027, 81, 83, 99-
Lasoik, Howard 11,70, 106023, 101, 105014, 119, 124-125, 149,
149, 157030 15102, 153013, 157029
Lebeaux, David 65, 10203 Shoshaoi, Rooit 84
Lees, Robert 2, 65 Sigurössoo, Halld6r 83
Loogobardi, Giuseppe 6-7, 9, 20, Silooi, Tal 6, 15-17, 19-20,37,52,
29-30, 62016, 10202, 94, 99, 101, 62021, 64027, 81, 91, 10202,
109, 126, 183 105014, 111, 118-119, 133, 146,
149, 181-182
Maliog, loan 83 Speas, Margaret 6, 30
May, Robert 149 Sportiche, Domioique 14, 30,
McCarthy,loho 6002 64027, 119
McCloskey, James 38 Sproat, Richard 32
Miloer, Jean-Claude 3 StowelI, Tim 7-8, 20, 38, 109, 126,
Mohammad, Mohammad 6, 19 140, 145-146, 149, 183, 188
Strozer, ludith 55
Ouhalla, Jamal 6, 19 Szabolcsi, Aooa 6-8, 10, 20, 35, 39-
40, 55, 71, 91, 95-96, 10202,
Partee, Barbara 120 106021, 109, 126, 183
Peooer, Zvi 6
Pesetsky, David 199013 Taraldseo, Kout Tarald 6, 19,
Pollock, Jean-Yves 15, 39, 134, 158032
155018 Tellier, Christioe 15309
Teooy, Carol 62016
Qafisheh, Hamdi 153013 Thraiossoo, Höskuldur 83
Quicoli, Carlos 55
Valois, Daniel 6, 61011
Rapoport, Tova 81, 112, 115 Veodler, Zeoo 68
Rappaport, Malka 6004, 87 Vergoaud, lean-Roger 50
Reiohart, Tanya 28, 30, 6108, 93, Vikoer, Steo 158032
100, 200015
Reuland, Eric 6108, 93, 100 Weijler, Ziva 6108
vao Riemsdijk, Heok 64027, 83, Williams, Edwio 30, 64027, 70, 93,
15103 106023, 129
Ritter, Elizabeth 6, 19, 61011,
61012, 37-38, 82, 94 Zaeoeo, Aooie 83
Rivas, Alberto 55 Zanuttioi, Raffaela 15104, 156027,
Rizzi, Luigi I, 13-14, 30, 38, 177
63021, 56, 58, 64027, 83, 91, Zubizarreta, Maria Luisa 3
106025, 115, 126, 153011, Zucchi, Alessandro 146

Absolute construction 143, 157n31 Agent

Accusative implicit 69, 91-95, 10ln2
inherent 9, 80-89, 93-94, 96, of event nouns 4, 26-30, 50, 69,
117, 160 71, 91-95
pronouns 80-81, 83, 85, 117, see also Argument, external; 0-
161, 167, 181 roles
structural 14-17, 40-42, 80-86, Agreement
106n25, 116, 118, 119, 160, dynamic 63n21
177, 193, 200nl6 in definiteness 31,45-47, 123,
see also Case; Event nouns 127, 183
Adjacency 38-39, 51-53, 79nlO participle 15-17, 118-119, 135
Adjectives see also Case, checking; Spec-
agent oriented 69, 72 head
location of 23-25, 32-33, 37, 44- Agr-elements
45, 51-53, 77-78 versus multiple specifiers 17
modifying 5,23-25,31-33,37, Agrgon features 43-44, 46-47, 49-51,
44-45, 47, 51-53, 126-127, 183, 189, 195-196
153n12, 154n14, 156n26, 172, 'AI-yedey phrase
191-193, 196 in noun phrases 48, 50, 70-71,
referential 94-95 91, 95, 97
see also Modifier, frequency in passive 90-91, 152n5
Adjunction Anaphors
AP 32, 37, 44, 51, 53 in noun phrases 29, 34, 91-92,
mechanism 14, 32, 37-38, 153n8 99
Adjuncts in participial relatives, 119, 140
A- 101n2 Arabic
in concrete nominals 71, 96, 99 Gulf 126-128
participial 129, 131, 143-144 Palestinian 157n29
sentence initial 130, 160 Standard 8, 41, 45-47, 105n15,
sentence final 129, 160 126-128, 150
temporal 103n8, 160, 164, 170 Arbitrary interpretation 91, 93, 149
Adverbs Argument
versus adverbial PPs 9, 75-78 external 31,33-34,47-51,82-83,
agent oriented 73, 106n23 93, 10ln2 106n21, 133
as bare noun phrase 103n8 internal 31,33-34,47-51, 60n4,
VP-initial 52, 134-135 67, 82, 194; see also Event
see also Event nouns, adverbial nouns, accusative arguments
modification position 7, 126, 163, 166, 182-
Affectedness constraint 42n 16 184, 189-190, 193, 200n 15


Argument (cont.) 14, 17, 31-33, 43, 45, 47, 58,

versus semantic participant 71, 119, 133, 137, 149-150, 180,
86,92 189, 199n12; see also Case
structure 3-4, 7, 9, 65, 67-71, 75, Clausal reduction 115-116, 156n27,
96, 132; see also Event, 177
structure Clitic and cliticization
see also 6-roles; D (position) doubling 7, 9, 55-59, 100,
Article 5-6,8-10, 19,22-23,31,39, 157n30, 173, 190-191, 196
45-47 60nl, 61nl0, 102n4, and gerunds 160, 164, 168, 170,
109, 111, 116, 122-124, 174, 173, 179, 182, 190-191, 195-
183, 192-196, 199n12; see also 196, 200n 16,
Ha- and nouns 54-59, 100
Auxiliaries 16-17, 151n2, 133-136, and participles 152n6, 135-136
155n21, 155n22, 198n9 Semitic versus Romance 63n27
see also Case; Pronouns, suffix al
Beynoni 111, 151n2, 116, 118, Complex tenses 16,133, 157n31,
152n5, 152n6 198n9
Binary branching 14, 61n7 Computational system
Binary operation derivational model 12
versus singulary operation 12 levels of representation 10-12;
Binding 20, 28-30, 33, 91-93, 99, see also Logical Form;
119; see also Anaphors; Phonetic Form
Pronouns, and Principle B; Concrete no uns 22, 61n8, 33, 44,
Unselective binding 65, 78, 92, 96-101, 102n3
Biunique Relation Requirement 50, Condition on Extraction Domains
57-58, 99-101 (CED) 64n27
Burzio's generalization 94, 119 Construct state
By phrase 70, 90; see also 'AI-yedey and adjectives 24-25
phrase binding in 34
and Case 9, 37-44, 47-50, 54, 56-
Case 57,82
checking 12, 14-17,42-45,49, versus compounds 60n5
57-58, 118-119, 135, 167, 179- definiteness value of 22-23, 45-47
180, 182, 184, 189, 196 versus free state 21-26
and clitics 54-59, 167, 180, 182, and gerunds 174, 178-180, 183,
189, 196 187, 195-196
under government 14, 37-39, 41- see also Adjacency
42 Control 70, 93, 143, 149, 155n23
inherent versus structural 9, 14- Coordinate Structure Constraint 53
15,41-42,47,50,57,59,80- Coordination 170-171, 187, 199n13
82,84-85, 116-117, 160-161, Copular pronoun 112-113, 115-116,
179 152n5
null 149 Covert component: see Logical Form
and variables 148-150 Covert movement: see Movement;
see also Accusative; Adjacency; Features, weak
Biunique Relation CP
Requirement; Genitive and features of C 126-127, 147,
C-command 28-30, 62n15. 92 parallel to DP 8, 20, 121-122,
Checking and checking theory 12- 126-127

D (projection) see also Derivation, fewest steps

adjectival 127-129, 156n26 in
argumental 9, 20, 126 English 5-6, 8, 13, 32-33, 42, 45,
features of 32-33, 44, 126-128, 63n24, 70, 72, 76, 78, 84, 95,
147-148 97, 103n8, 109, 126-129,
modifying 9, 10, 126-128, 145, 155n21, 155n22, 155n23, 144,
147-148 I 56n26, 146, 157n30, 150,
and nominalization 8, 10, 159, 168, 188, 197
187, 197 'Et
and non-tensed projections x, 8, as diagnostic for definiteness 22-
10, 17, 110, 145-146, 159,176- 23,79
177, 180 in noun phrases 79, 83, 85
parallel to C 7, 8, 20, 109, 121- with passives and unaccusatives
122, 126-127 84
and reference 7, 8, 10, 20, 109, Event
183-184, 187 structure 65, 67, 186; see also
as a relative complementizer 109- Argument, structure
110, 116, 127, 145 diagnostics 3-4, 68-73
selectional properties of 144 Event nouns
its specifier as an operator accusative arguments of 9, 65,
position 121-122, 145, 147 72-74, 78-89, 93-96
and VP x, 8, 9, 110, 146, 159, adverbial modification of 9, 65-
176-177 66, 72-74, 75-78, 85-87
"weakness" of 183-184 and the lexicalist approach 7, 9,
see also Article; Ha- 66,95-96
Danish 158n32 versus result nouns 3-5, 65, 67-
Dative pronouns 104n14, 162, 182, 73, 75, 78, 85-86
194 syntactic approach to (VP-
Definiteness feature/value 22-23, 25, analysis) 4-5, 66, 74-76, 82,
31, 61nlO, 46-47, 73, 124-125, 86-89,95
183, 192-193, 196 198nl0, Exceptional Case marking (ECM)
199n12,200n15 41-42, 78-80, 83, 102n5,
Derivation 104n13, 156n30, 161, 179
crashes of 12,47,49,51, 195 External argument: see Argument
fewest steps in 13 Extended Standard Theory (EST) II
see also Economy principles Extraction: see Movement; Islands
Deverbal nouns: see Event nouns;
Result nouns Features
Disjointness 29, 92 strong 12-13, 32, 43-44, 119, 182
DP-features 14, 44, 49, 51, 119, weak 12-13, 62n18, 181-182
181-182 see also Agrgen features;
DP-zation 8, 159, 177 Definiteness feature/value; DP-
features; [±mod]; cj>-features;
Economy principles Wh-
Full Interpretation 13, Free state
Last Resort guideline 13, 61n9, and adjectives 24
45, 49, 83, 198n9 binding in 28-30
Procrastinate 13 see also Construct state; SeI
Shortest Movement condition 13 phrase; Genitive

French x, 8, 10, 13, 15-17, 40, 42, Islands: see Subjacency

45, 76, 78, 86, 95, 97-98, Italian 8, 15-16, 40, 64n27, 83, 101,
103n8, 104nl1, 109-110, 126- 126-129, 142-143, 146, 156n26,
147, 157n30, 150, 157n31, 197, 199nl2
Full Interpretation: see Economy
principles Language faculty I, 10-11
Last Resort: see Economy principles
Genitive Lexicon ix, 1-3, 5, 11-12, 66, 74-
inherent 41-42, 45, 47, SO, 57, 75, 88, 91, 95, 96, 102n2; see
59, 104n13, 179 also Nominalization, lexical;
structural 40-43, 47, SO, 54, 57, Event nouns, lexicalist
59, 179-180 approach
see also Case; Construct state; Linear order
Free state in gerund clauses 169, 171
German 83 in noun phrases 27,31, 61n12,
Gerunds 37,44
English 5, 146, 197 permutation in 81, 83, 100-101
versus infinitives 163-165, 188- universal 14
189 VSO 32
nominal 190-197 Logical Form (LF) 8,11-13,31,43,
versus noun phrases 5, 172-176, 149, 182, 199nll
184-187 Logophors 61n8, 99-100
verbal properties of 5, 160-163 Lowering 13, 28
and tense 163, 176-177
see also Clitics; Construct states; Minimalist Program 10-12, 17,33,
SeI 45
Government 14, 37-39, 42 [±mod] 126-128, 145, 147-148
Greek, Classical 126-128 Modifiers
aspectual 68, 72
Ha- frequency, 3-4, 68, 72, 76,
adjectival: see D (projection) 105n18
relative 122-125 Movement
see also Article covert 12-13
Heavy NP shift 104n 11 N-raising 6, 9, 31-34, 43-44
Hungarian 35, 39, 42-43, 54, 91, V-raising 13,32,74, 137, 155n18
95, 102n2 see also Checking and checking
theory; Subjacency
Icelandic 83 Multiple specifiers: see Agr-elements
Inalienable possession 22
Incorporation 9, 53, 74-75, 89, Negation
106n, 170 constituent 154n17, 155n20
Infinitives nominal 175-176, 191, 193
constructs 159 sentential 113, 115-116, 134-136,
nominalized 8, 146, 197, 199n 12 156n27, 163, 165, 170, 177,
and tense 8, 146, 149, 163, 188 189
see also Gerunds; Rationale Nominalization
clauses as a change in categorial status
Inherent Case: see Case 74-75, 159
Internal argument: see Argument lexical ix, 5, 187, 193

Nominalization (cont.) Prepositions

syntactic ix, x, 5, 8, 10, 159, 187 prefixal 171, 198n14
see also Event nouns temporal 159, 164, 166, 172,
Norwegian 158n32 184, 187, 189, 194
Null operator 114, 121, 153n9, 128, Principles and Parameters approach
140, 142, 144, 147 1-2
Null subject: see Agent, implicit; Process nouns: see Event nouns
PRO Procrastinate: see Economy
Operator: see Null operator Pronouns
order: see Linear order bound by a quantified noun
Overt movement: see Features, strong phrase 30, 34, 99-100
and Principle B 91-92
Participles suffixal 54, 57, 81,
active 119 see also Accusative; Clitics;
and Case capacity 17, 86, 133 Copular pronoun; Dative
passive 17, 132 pronouns
past 86, 132-135, 155n19, 137, Pro 58-59, 63n25, 182, 189, 195
139, 143 PRO 63n25, 92-93, 102n2, 140-141,
present 132-133, 135-136, 137, 143-144, 148-149; see also
13~ 156n23, 157n31 Case, null
and thematic capacity 133
see also Beynoni Quantifiers
Participial agreement: see Agreement, collective versus distributive 124
participle floating 119, 140
Participial relatives 8, 10, 109-110, tout and rien 136
137-138, 146-148, 150 155n20, see also Pronouns
155n21; see also Relatives,
reduced; Relatives, semi- Raising: see Movement
Passive Rationale c1auses 70-72, 106n23
nominal 48, 52, 83, 89-91 Reference: see D (projection)
I 05n 16, 106n20 Relative c1auses
verbal 15, 84, 89-91 appositive (versus restrictive)
see also Participles 153n7, 144
Patient 26, 28-29; see also Theme free 114, 153nl0, 154n15,
«t>-features 43,45,47,58, 61nl0, 157n28
149-150, 180, 182-183, 189 reduced 86, 109, 129-137, 139-
Phonetic Form (PF) 11-12, 32, 53, 147, 157n31
63n25, 101 regular 111-116, 120-122, 125-
Possessors 25, 35, 39, 60n4, 92, 97- 127, 129, 131, 138, 140-141,
99, 102n5, 145, 147-148
see also Construct State; Free semi- 111-122, 124-125, 127-
state; Inalienable, possession; 129, 131-132, 154n17, 135,
Semantic participant 140, 144-147
PPs see also Participial relatives
adverbial 9, 75-78, 85, 105n 18, Relativized minimality 13, 158n32
87, 185 Reordering 101
Predication 75, 102n2, 153n11 Result nouns: see Event nouns;
Concrete nouns

Selectional properties of D: see D 30,33

(projection) Theme 27, 30, 72, 84-86, 93, 95-96,
Semantic participant 3, 71, 86, 92, 103n7; see also Argument,
96 internal; Patient
SeI 6-criterion 13
in nominal gerunds 190-191, 193, 6-grid 26-27
195 6-discharge 13
in noun phrases 9, 21, 24-26, 29, 6-roles 3-4, 7-8, 10, 20, 30, 65, 67,
40-41, 45-51, 55-58, 60n4, 71, 91, 93-96 102n5, 109, 119,
61n9, 73, 77-78, 82, 97-101 152n7, 133, 135, 143, 148; see
Sm all clause 40-41, 52, 78, 83, 117 also Agent; Patient; Theme
Statives 67, 103n7 Turkish 35-36
Shortest Movement condition: see
Economy principles Unaccusatives 15, 84, 106n25, 132,
Simplicity (notion of) 10-11 155n19, 155n21, 190, 194
Singulary operation: see Binary Uniformity of Theta Assignment
operation; Movement Hypothesis (UTAH) 33
Spanish 56 Universal Grammar (UG) 1-2, 9, 11,
Spec-head (Spec-Agr) 9, 14-15, 42, 75, 89, 96, 134,
57-58, 62n20, 122-123, 128, Unselective binding 199n 13
149-150; see also Case,
checking Variables 121, 128, 140-141, 143-
Speil-out 12, 28, 32, 42-43, 119, 145, 148-150, 199n13
136, 181 Visibility Principle 148-149
Speil-out of Case features: see Case,
clitics Weak crossover: see pronouns,
Split-IP 15, 39, bound by a quantified noun
Subjacency 140-142; see also phrase
Condition on Extraction Weak features: see Features
Domains Wh-
Subject: see Agent; Null subject features 147-148
movement 63n27, 153n9
Tense phrases in relative clauses 114,
and participial relatives 112, 137- 124, 131, 144, 147-148
139, 145-146 Word order: see Linear order
and syntactic nominalization 8,
10, 159 X'-theory (x-bar) 6, 12-14 28, 61n7
see also Gerunds; Infinitives
Thematic Correspondence Hypothesis Yupic 35,43
Studies in Natural Language and Linguistic Theory

Managing Editors
Liliane Haegeman, University o/Geneva
10an Maling, Brandeis University
1ames McC1oskey, University o/California, Santa Cruz

1. L. Burzio: ltalian Syntax. A Government-binding Approach. 1986.
ISBN Hb 90-277-2014-2; Pb 90-277-2015-0
2. W.D. Davies: Choctaw Verb Agreement anti Universal Grammar. 1986.
ISBN Hb 90-277-2065-7; Pb 90-277-2142-4
3. K. E. Kiss: Corifigurationality in Hungarian. 1987.
ISBN Hb 90-277-1907-1; Pb 90-277-2456-3
4. D. Pulleyblank: Tone in Lexical Phonology. 1986.
ISBN Hb 90-277-2123-8; Pb 90-277-2124-6
5. L. Hel1an and K. K. Christensen: Topics in Scandinavian Syntax. 1986.
ISBN Hb 90-277-2166-1; Pb 90-277-2167-X
6. K. P. Mohanan: The Theory o/Lexical Phonology. 1986.
ISBN Hb 90-277-2226-9; Pb 90-277-2227-7
7. 1. L. Aissen: Tzotzil Clause Structure. 1987.
ISBN Hb 90-277-2365-6; Pb 90-277-2441-5
8. T. Ounji: Japanese Phrase Structure Grammar. A Unification-based Ap-
proach.1987. ISBN 1-55608-020-4
9. W. U. Wurzel: lnjlectional Morphology anti Naturalness. 1989
ISBN Hb 1-55608-025-5; Pb 1-55608-026-3
10. C. Neidle: The Role 0/ Case in Russian Syntax. 1988 ISBN 1-55608-042-5
11. C. Lefebvre and P. Muysken: Mixed Categories. Nominalizations in Quechua.
1988. ISBN Rb 1-55608-050-6; Pb 1-55608-051-4
12. K. Michelson: A Comparative Study 0/Lake-lroquoian Accent. 1988
ISBN 1-55608-054-9
13. K. Zagona: Verb Phrase Syntax. A Parametric Study of English and Spanish.
1988 ISBN Hb 1-55608-064-6; Pb 1-55608-065-4
14. R. Hendrick: Anaphora in Celtic anti Universal Grammar. 1988
ISBN 1-55608-066-2
15. O.1aeggli and K.1. Safir (eds.): The Null Subject Parameter. 1989
ISBN Hb 1-55608-086-7; Pb 1-55608-087-5
16. H. Lasnik: Essays on Anaphora. 1989
ISBN Hb 1-55608-090-5; Pb 1-55608-091-3
17. S. Steele: Agreement anti Anti-Agreement. A Syntax of Luisefto. 1990
ISBN 0-7923-0260-5
18. E. Pearce: Parameters in Old French Syntax. Infmitival Complements. 1990
ISBN Hb 0-7923-0432-2; Pb 0-7923-0433-0
19. Y.A. U: Order and Constituency in Mandarin Chinese. 1990
ISBN 0-7923-0500-0
Studies in Natural Language and Linguistic Theory
20. H. Lasnik: Essays on Restrictiveness and LearnabiUty. 1990
ISBN 0-7923-0628-7; Pb 0-7923-0629-5
21. M.J. Speas: Phrase Structure in Natural Langllage. 1990
ISBN 0-7923-0755-0; Pb 0-7923-0866-2
22. H. Haider and K. Netter (eds.): Representation antI Derivation in IM Theory 0/
Grammar.I991 ISBN 0-7923-1150-7
23. 1. Simpson: Warlpiri Morpho-Synta%. A Lexica1ist Approach. 1991
ISBN 0-7923-1292-9
24. C. Oeorgopoulos: Syntactie Variables. Resumptive Pronouns and A' Binding
in Palauan. 1991 ISBN 0-7923-1293-7
25. K. Leffel and D. Bouchard (eds.): Views on Phrase Structure. 1991
ISBN 0-7923-1295-3
26. C. Tellier: Ucensing Theory antI French Parasitie Gaps. 1991
ISBN 0-7923-1311-9; Pb 0-7923-1323-2
27. S.-Y. Kuroda: Japanese SyntaxandSemtmtics. CollectedPapers.l992
ISBN 0-7923-1390-9; Pb 0-7923-1391-7
28. I. Roberts: Verbs and Diachronie Syntax. A Comparative History of English
andFrench.l992 ISBN0-7923-1705-X
29. A. Fassi Fehri: Issues in lhe Structure 0/Arabie Clauses and Wortb. 1993
ISBN 0-7923-2082-4
30. M. Bittner: Case, Scope, andBintIing.I994 ISBN 0-7923-2649-0
31. H. Haider, S. Olsen and S. Vikner (eds.): Studies in Comparative Germanie
Synta%. 1995 ISBN 0-7923-3280-6
32. N. Duffield: Partieks and ProJeetions in lrish Syntax. 1995
ISBN 0-7923-3550-3; Pb 0-7923-3674-7
33. 1. Rooryck and L. Zaring (eds.): Phrase Structure and the Lexieon. 1996
ISBN 0-7923-3745-X
34. 1. Bayer: DireetionaUty and Logieal Form. On the Scope of Focusing Particles
and Wh-in-situ.I996 ISBN 0-7923-3752-2
35. R. Freidin (ed.): Current Issues in Comparative Grammar. 1996
ISBN 0-7923-3778-6; Pb 0-7923-3779-4
36. C.-TJ. Huang and Y.-H.A. Li (eds.): New HorirPns in Chinese Linguisties.
1996 ISBN 0-7923-3867-7; Pb 0-7923-3868-5
37. A. Watanabe: Case Absorption antI WH-Agreement. 1996
ISBN 0-7923-4203-8
38. H. Tbrtinsson, S.D. Epstein and S. Peter (eds.): Studies in Comparative
Germanie SyntIJX. Volume ß.I996 ISBN 0-7923-4215-1
39. CJ.W. Zwart Morphosyntax 0/ Verb Movement. A Minimalist Approach to
the Syntax of Dutch. 1997 ISBN 0-7923-4263-1; Pb 0-7923-4264-X
40. T. Siloni: Noun Phrases andNominalit.ations. The Syntax ofDPs. 1997
ISBN 0-7923-4608-4

Springer-Science+Business Media, B.V.