Universal Grammar in the Reconstruction of Ancient Languages

WDE

G
Studies in Generative Grammar 83

Editors
Henk van Riemsdijk
Jan Köster
Harry van der Hulst

Mouton de Gruyter
Berlin · New York
Universal Grammar in
the Reconstruction
of Ancient Languages

edited by
r

Katalin E. Kiss

Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter G m b H & Co. KG, Berlin.

The series Studies in Generative Grammar was formerly published by

Foris Publications Holland.

© Printed on acid-free paper which falls within the guidelines

of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data

Universal grammar in the reconstruction of ancient languages / edited

by Katalin E. Kiss.
p. cm. — (Studies in generative grammar ; 83)
Includes bibliographical references and index.
ISBN-13: 978-3-11-018550-8 (cloth)
ISBN-10: 3-11-018550-4 (cloth : alk. paper)
1. Grammar, Comparative and general. 2. Extinct languages.
I. Kiss, Katalin Ε. II. Series.
P151.U55 2005
415—dc22
2005025876

Bibliographic information published by Die Deutsche Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie;
detailed bibliographic data is available in the Internet at <http://dnb.ddb.de>.

ISBN-13: 978-3-11-018550-8
ISBN-10: 3-11-018550-4

© Copyright 2005 by Walter de Gruyter G m b H & Co. KG, D-10785 Berlin.

All rights reserved, including those of translation into foreign languages. No part of this
book may be reproduced in any form or by any means, electronic or mechanical, including
photocopy, recording, or any information storage and retrieval system, without permission
in writing from the publisher.
Cover design: Christopher Schneider, Berlin.
Printed in Germany.
Contents

Introduction 1
Katalin E. Kiss

The correlation between word order alternations,

grammatical agreement and event semantics in Older Egyptian 31
Chris H. Reintges

The nominal cleft construction in Coptic Egyptian 105

Chris H. Reintges, Aniko Liptak and Lisa Lai Shen Cheng

Genitive constructions in Coptic 137

Barbara Egedi

Left-dislocated possessors in Sumerian 161

Gäbor Zölyomi

Complex predicate structure and pluralised events in Akkadian 189

Christian Huber

VSO and left-conjunct agreement:

Biblical Hebrew vs. Modern Hebrew 239
Edit Dor on

IE *weid- as a root with dual subcategorization features 265

in the Homeric poems
Annamaria Bartolotta

The syntax of Classical Greek infinitive 295

Vassilios Spyropoulos

Latin object and subject infinitive clauses 339

Lucio Melazzo
vi Contents

Latin word order in generative perspective:

An explanatory proposal within the sentence domain 373
Chiara Polo

Some firm points on Latin word order: The left periphery 429
Giampaolo Salvi

Classical Sanskrit, "wild trees",

and the properties of free word order languages 457
Brendan Gillon and Benjamin Shaer

A particular coordination structure of Indo-European flavour 495

Emanuele Lanzetta and Lucio Melazzo

Index 519

List of contributors 525

Introduction

Katalin Ε. Kiss

1. Goals and methods

Generative linguistics aims to describe the linguistic faculty of human indivi-

duals. The generative grammar of a language models the internal grammar
of a native speaker, i.e., his/her ability to construct and understand every
possible sentence of the given language, to judge the grammaticality of
utterances, and to formulate intuitions about their structures. Generative lin-
guistic theory also formulates hypotheses about how the internal grammar
of human individuals is attained. Language faculty is seen as a genetically
given endowment of the human species. The initial state of language fac-
ulty, called Universal Grammar, determines the class of possible languages.
In the Principles and Parameters approach, Universal Grammar is a set of
universal principles and a finite array of options as to how they apply. Lan-
guage differences derive from particular choices of values of parameters.
In the course of language acquisition, values of parameters are set on the
basis of exposure to utterances of the given language.
The construction of the generative grammar of a language is a deductive
process, in the course of which hypotheses are formulated concerning the
structures of the language and the procedures of generating them, and the
predictions of the hypotheses are compared with the actual linguistic data
produced by native speakers. The non-occurrence of a predicted construc-
tion can always be an accidental gap - therefore, when testing a hypothesis,
it is crucial to check with native informants if every possible construction
derivable from, or allowed by, the given hypothesis is accepted by them,
and every construction excluded by the given hypothesis is rejected.
Starred, i.e., rejected, examples represent an indispensable element of the
deductive analysis. The deductive method, involving recurring cycles of
constructing a model, testing it by examining its positive and negative pre-
dictions, and then refining it, lead to a much higher degree of descriptive
adequacy than can be attained by traditional grammars compiled induc-
tively.
2 Katalin Ε. Kiss

Ancient languages apparently resist a generative description. If the goal of

generative language description is the modelling of an individual's internal
grammar, then it might, indeed, seem questionable if this goal can be mean-
ingfully pursued in the case of individuals deceased thousands of years ago.
It is not immediately obvious, either, that what we have to describe in the
case of an ancient language is an infinite set of sentences; in fact, we have a
finite set of written data at our disposal, which seem to be managable by an
inductive analysis. The lack of informants, i.e., the impossibility of obtain-
ing grammaticality judgments and rejected examples, also suggests that a
deductive approach would be not only unnecessary but also impossible.
The inductive analysis of closed corpora has been the methodology
adopted in traditional analyses of ancient languages. This methodology,
however, has its limitations. What it can achieve is basically the listing and
interpretation of the morphemes of a language. Thus it has been successful
in identifying grammatical categories, in compiling morphological para-
digms of verb conjugations and noun declinations, in distinguishing tenses,
moods, cases, and agreement markers. It has also been able to establish
grammatical relations, i.e., to link heads with complements, and to distin-
guish complements of different grammatical functions. However, there are
also problems which cannot be tackled by the traditional method. An obvi-
ous difficulty is the problem of gaps, possible mistakes, and idiosyncracies
in the data available. For an inductive analysis, it is often impossible to
decide whether a construction unattested is an accidental gap or something
excluded by the grammar. Similarly, it may be impossible to tell if an un-
usual piece of data is a mistake of an ancient scribe, an idiosyncracy elic-
ited by the restrictions of versification, or a rare but perfectly grammatical
construction. The greatest problem for the grammarian of a dead language
is the fact that the data - whether scarce or abundant - always underdeter-
mine grammar. This problem is particularly severe in the case of dead lan-
guages like Sumerian or Akkadian, which had to be deciphered again com-
pletely - in contrast to ancient languages which have been sustained by
cultic and scholarly traditions, or which have in some sense survived in
their descendant languages.
The difficulties of reconstructing the grammar of an ancient language
resemble the difficulties that a child experiences when reconstructing the
grammar of its mother tongue. A child acquiring its mother tongue, too, has
access only to a limited - and sometimes defective - set of positive evi-
dence (the correction of the child's mistakes by adults is by no means a
necessary element of language acquisiton). If the two processes are similar,
 Introduction 3

then the methodology adopted in the reconstruction of the grammar of an

ancient language must also be similar to that employed by a child in the
course of language acquisition. What the child does is interpret the data it
has access to on the basis of the genetically coded Universal Grammar that
it possesses. This is what the linguist setting out to reconstruct the grammar
of a dead language must do, as well; he or she must interpret the data avail-
able as indications of how the open parameters of Universal Grammar are
to be set. Naturally, a linguist is in a more difficult situation than a child
because what the linguist can work with is not Universal Grammar itself
but only a model of it whose correspondence to the implicit knowledge of a
child is an open question. Nevertheless, hypotheses concerning Universal
Grammar seem to be well-established enough to make such a project a
worthwhile experiment.
The studies of this volume aim to demonstrate that descriptive problems
which proved to be unsolvable for the traditional, inductive approach to
ancient languages can be reduced to the interaction of regular operations
and constraints of the hypothetical Universal Grammar. The proposed
analyses of ancient languages as instantiations of Universal Grammar also
bear on linguistic theory. In addition to confirming or refuting certain spe-
cific hypotheses, they provide clear empirical evidence of the perhaps most
basic tenet of generative theory, according to which Universal Grammar is
part of the genetic endowment of the human species - that is, human lan-
guages do not "develop" parallel with the development of human civiliza-
tion. The languages examined in this volume were spoken several thousand
years ago. The Egyptian and Sumerian texts to be analyzed can be as much
as 5000 years old, and the first Akkadian records are also only slightly
younger. As will become clear from the studies of this volume, the gram-
mars of these languages do not differ in any relevant respect from the
grammars of languages spoken today.

2. Predecessors

The view that ancient languages can be modelled successfully in the gen-
erative framework despite the lack of native speakers has been shared by a
number of linguists ever since the ninteen sixties. Some scholars assumed
that the lack of native speakers can be made up for by a corpus of adequate
size and variation (Ehlich 1981), while others claimed that they had devel-
oped a native-like intuition in the language studied by them (Pillinger 1980).
4 Katalin Ε. Kiss

It has been mainly Latin whose grammatical phenomena have been sub-
jected to generative analyses. The first major generative study of Latin
syntax was Robin L a k o f f s book on Latin complementation (1968), dis-
cussing various types of finite and non-finite complement clauses, and the
licensing of various moods. In the seventies and early eighties, the period
of the emergence of Government and Binding Theory, problems of Latin
infinitival phrases aroused interest, which has persisted up until today.
Pepicello (1977) analyzed accusative with infinitive as Subject-to-Object
Raising, and his views elicited alternative proposals by Bolkenstein (1979),
Pillinger (1980), and Wales (1982). Questions of accusative with infinitive
constructions, for example, the source of the accusative case, kept recurring
also in the proceedings of a series of International Colloquia on Latin Lin-
guistics, edited by Pinkster (1983), Touratier (1985), Calboli (1989), and
Herman (1994). Maraldi (1983), for example, identifyied Acl as Excep-
tional Case Marking, licensed by S' deletion; however, she also noticed
instances of Acl occurring in the context of a passive matrix verb. The
same problem was also addressed by Calboli (1983, 1989), and recently by
Cecchetto and Oniga (2002).
Areas of Latin syntax examined in the generative framework include the
case system - see Binkert (1970) and Murru (1977). Maurel (1983, 1989)
discussed problems of Latin relativization. Bertocchi and Casadio (1983),
and Bertocchi (1989) dealt with questions of binding, particularly with the
Latin se and suus, which are anaphors that can be bound not only locally
but also at a long distance. Salvi (1996) and Giusti (2001) examined the
Latin pronominal system from a historical perspective, as the predecessor
of the Romance systems of strong, weak, and clitic pronouns and the Ro-
mance article. Basic questions of Latin syntax concerning sentence struc-
ture, word order variation, its derivation, and its discourse functions were
addressed by Ostafin (1986), Salvi (1999/forthcoming), and Polo (2003).
Benucci (1996) analyzed Umber, a close relative of Latin. Recently issues
of Latin morphosyntax have gained significance, particularly in the frame-
work of Distributed Morphology - see Embick's derivation of the synthetic
and analytic forms of the perfect (2000), and Embick and Halle's analysis
of the Latin conjugation (1999). Oniga (2004) provides a generative analy-
sis of the major morphological and syntactic structures of Latin.
Studies of Ancient Greek in the generative literature had been sporadic
up until recently. The analysis of agreeing predicative adjectives in non-finite
clauses by Andrews (1971) provided important evidence for the presence of
a covert subject in infinitives, and contributed to the shaping of control and
 Introduction 5

raising theory. Lightfoot (1975) devoted a whole book to the analysis of

Classical Greek complementation and the Greek mood system. Interest in
Classical Greek has become more intensive since Modern Greek syntax has
been subjected to thorough analyses in the generative framework. The first
major result of the renewed interest in Classical Greek was Morrell (1989).
Taylor (1990, 1996) analyzed sentence structure, DP-structure, and clitic
position in Ancient Greek, whereas Taylor (1994) dealt with the change of
Ancient Greek word order from OV to VO. Philippaki-Warburton and Cat-
simali (1997), and Tantalou (2003) focus on problems of infinitival
phrases, among them the case assigned to the subject of infinitives.
The dissertation of Garret (1990) examined various syntactic problems
of Anatolian (Hittite), particularly those concerning pronominal clitics.
Kiparsky (1995) called Pänini's grammar of Sanskrit "the most complete
generative grammar of any language yet written". No wonder few Indolo-
gists felt the need of adopting the methodology and terminology of con-
temporary generative grammar. The best-known exceptions are the disser-
tations of Hale (1987) and Schäufele (1990). In addition to basic questions
of sentence structure, word order, and discourse functions, clitic placement
has been examined in a number of papers (see Hale 1990, 1991, 1993,
1996; and Schäufele 1993, 1996). The way of influence between Indo-
Iranian philology and generative theory has been bidirectional; Pänini's
grammar has also enriched generative theory - primarily through a series of
studies by Kiparsky (Kiparsky and Staal 1969; Kiparsky 1979-80, 1982,
1995).
Apart from two papers on Sumerian by Gragg, discussing the Sumerian
copula (Gragg 1968), and some basic issues of Sumerian constituent struc-
ture as well as general questions of traditional philology and transforma-
tional grammar (Gragg 1973), and apart from a book on Akkadian syntax
and morphology by Groneberg (1987), the only generative syntactic studies
of Sumerian and Akkadian are the papers of Christian Huber (1989-90;
1996; and forthcoming) and Gabor Zolyomi (1996), whose work is also
represented in the present volume.
The best-known example of the analysis of a syntactic phenomenon of
Biblical Hebrew in the generative framework is Doron (2000), a paper also
reprinted in this volume. Other major sources of the generative analysis of
Biblical Hebrew syntax are the dissertation of Vincent de Caen (1995),
examining the structural position and interpretation of the verb, and the
dissertation of Robert Holmstedt (2002), a Minimalist study of relativiza-
tion.
6 Katalin Ε. Kiss

In Egyptology, it is Chris Reintges, also an author of the present volume,

who has adopted the methodology of generative linguistics - see Reintges
(1996, 1997, 1998, 2000, 2001, 2003, etc.).

3. The problems addressed

The studies in this volume analyze syntactic and morphosyntactic phenom-

ena from various areas of grammar. Many of these phenomena represent
old problems, inexplicable idiosyncracies for the philologies of ancient
languages; others (e.g. word order variation) have not even been acknowl-
edged as questions to be accounted for in the course of grammatical de-
scription. The analyses of these problems as instantiations of options pro-
vided by Universal Grammar shed light on previously unrecognized
structures, operations, and constraints of the languages in question.
The proposed analyses also provide valuable new data and new insights
for the set of hypotheses constituting Universal Grammar. The linguistic
facts discussed, taken from Sumerian, an isolated language, Akkadian, an
East Semitic language, Biblical Hebrew, a Northwest Semitic language,
Older Egyptian and Coptic Egyptian, which represent two stages of a sepa-
rate branch of the Afro-Asiatic family, and the Indo-European Sanskrit,
Classical Greek, and Latin, most of which have been inaccessible to gen-
erative theory before, contribute to the understanding of basic questions of
syntactic theory such as the licensing of structural case, the motivation of
movement operations, the structure of coordinated phrases, the role of
event structure in syntax, etc.
The studies of the volume also set forth interesting problems for further
research, in addition to the issues addressed in detail. For example, the ex-
tended Sumerian verb projection, also harboring a great number of clitics at
various places, seems to be more complex than any other verbal projection
studied in the generative framework.
Here is a brief overview of the main empirical and theoretical problems
that the chapters of this book discuss, and the kind of analyses that they
propose.
 Introduction 7

3.1. Accusative with infinitive without ECM and without object control

In generative syntactic theory, in which nominative case is licensed - as-

signed or checked - by finite inflection, a non-finite verb can only have an
overt, case-marked subject if the subject bears accusative case licensed by a
prepositional complementizer or by the matrix verb. A condition of case-
licensing by the matrix verb is that the infinitival phrase project no CP. As
Chomsky (1981: 140) observes in a footnote, however, "the case of accusa-
tive subjects of infinitives in Greek or Latin remains more problematic".
Two papers of this volume, The syntax of Classical Greek infinitive by
Vassilios Spyropoulos, and Latin object and subject infinitive clauses by
Lucio Melazzo, are devoted to the theoretical and descriptive problems
caused by Classical Greek and Latin accusative with infinitive construc-
tions. As they demonstrate, both Classical Greek and Latin infinitives can
have an accusative subject also in constructions in which neither of the two
conditions of accusative licensing by the matrix verb is satisfied: the matrix
verb is intransitive, or the infinitive phrase is in subject or adjunct position,
and furthermore, there is evidence of the infinitival clause also projecting a
CP - even if no visible complementizer is present. The possibility of object
control, i.e., the accusative noun phrase complementing the matrix verb,
and controlling a PRO subject, can also be excluded; the accusative noun
phrase clearly occupies the subject position of the infinitival clause. In
Classical Greek, an accusative subject can even appear in the infinitival
subject position of so-called genitive with infinitive and dative with infini-
tive constructions, in addition to the genitive or dative matrix argument
normally understood to represent the subject of the infinitive. For example:

(1) deomai hymo.n [pro iastrous genesthai]

beg-I you-GEN.PL pro-ACC doctor-ACC.PL become
Ί beg you to become a doctor.'

Classical Greek being a pro-drop language, the pronominal subject of the

infinitive is phonologically empty, but its accusative case is clearly indicated
by the accusative of the predicate complement, which always agrees in case
with the subject of its clause.
For the traditional approach, the appearance of the accusative case e.g. on
the predicate complement in (1), categorized as a 'genitive with infinitive',
is inexplicable. For the generative approach, the question is what licences a
case-marked subject in the embedded CP in (1), and what is the source of
8 Katalin Ε. Kiss

its accusative case. A further theoretical problem is the apparently optional

alternation between an accusative subject and a controlled PRO in Greek
infinitival clauses. Latin infinitival clauses sometimes even display a nomi-
native subject/accusative subject alternation - as illustrated by (2a,b):

(2) a. Dicitur eo tempore matrem Pausaniae vixisse

is.said that time mother-ACC of.Pausanias to-have-lived
'Pausanias' mother(-ACC) is said to have lived in that period.'
b. quem quidem... per sequi Caesar dicitur
whom in.truth to-pursue Caesar-NOM is.said
'whom Caesar(-NOM) is said to pursue'

Spyropoulos and Melazzo seek answers to the problems raised by Ancient

Greek and Latin accusative with infinitive constructions along similar lines;
they identify a head in the complementizer domain of the infinitival clause
as a case licenser.
As Spyropoulos convincingly argues, the source of the accusative case
of the infinitival subject in ancient Greek is the infinitival complementizer,
Cfin. Greek infinitives are marked for tense, hence Cf,n, agreeing with the
infinitival inflection, also carries a [+Tense] feature. The [+Tense] feature
of Cfin can also have a parasitic [Agr] feature bundle associated with it. If
the overt DP or pro subject of the infinitive cannot establish a checking
relation either with Inflection (because it is [+Tense] but [-Agr]), or with a
matrix controller (e.g. because of an intervening CP projection), it is the
[Agr] feature bundle of Cfj„ that is at hand to licence an accusative case on it.
In Latin, the infinitive phrase is often coindexed with a 3rd person sin-
gular pronoun, which leads Melazzo to the hypothesis that the infinitival
CP (similar to an object i/zai-clause) is base-generated in the complement
position of a DP, with the pronoun (or an empty pro) occupying the speci-
fier of DP. The case assigner of the infinitival subject is the FIN head of the
CP-domain (the same head that is called Cfin by Spyropoulos). Melazzo
claims that the bundle of features substantiating FIN includes a specifica-
tion for case, which is activated when the case of the subject cannot be
licensed in the IP-layer.
Interestingly, the subject of the infinitive is in the nominative in the case
of root infinitives, or those introduced by a temporal adjunct. This is ac-
counted for by the assumption that the case-feature of FIN is [+Accusative]
only if FIN is characterized as [+complement] (representing a complement
to D); it is [+Nominative] otherwise.
 Introduction 9

Subject infinitive clauses complementing a D are also predicted to have an

accusative subject, as in (2a) above. (2b), involving a subject infinitive clause
with a nominative subject, is analyzed as a kind of subject raising. In such
structures, the specifier of the DP harboring the infinitive clause is claimed
to be generated empty, and to serve as a landing site for the subject of the
infinitive, which thereby picks up the nominative case of the matrix subject.

3.2. The structure and the motivation of possessive constructions with an

external possessor

The fact that the possessor can appear either as a determiner of the posses-
sion, or as an extracted complement, or an adjunct external to the projection
of the possession is well-known from a number of languages. Since the
case-ending or the preposition of an external possessor is often different
from that of an internal one, the recognition of an external possession con-
struction can be a very difficult descriptive problem. External possessors
also raise theoretical questions such as what triggers the external ization of a
possessor, and how an external possessor construction is derived.
Two papers of the volume are devoted to problems of the possessive con-
struction. Gabor Zolyomi's paper paper entitled Left-dislocated possessors
in Sumerian discusses two different Sumerian constructions with a topical-
ized external possessor. In one of them, called the anticipatory genitive
construction, the left-dislocated possessor bears the genitive case, and is
coindexed with a possessive pronoun internal to the projection of the pos-
session. The possessor and the possessive pronoun coindexed with it need
not be subjacent; for instance, in one of the examples discussed, the posses-
sor in the left periphery of the matrix clause is coindexed with a possessive
pronoun in a relative clause. The lack of subjacency and the resumptive
pronoun strategy are interpreted as evidence that the anticipatory genitive
construction is a base-generated construction, involving no movement.
In the other possessive construction with a left-dislocated possessor, the
so-called external possession construction, the left-dislocated possessor,
though copied by a possessive pronoun in the projection of the possession,
is not in the genitive case but bears the same syntactic case as the posses-
sion itself. This construction can be used in the case of an inalienable pos-
session. Interestingly, the verb agrees with the external possessor instead of
the head of the possessive construction.
10 Katalin Ε. Kiss

The external possession construction is often masked by various surface

phenomena. On the one hand, the identical syntactic cases of the possessor
and the possessum can be realized by different morphological cases - given
that some adverbial cases may be supplemented by the dative case on noun
phrases with a human referent. On the other hand, when the external pos-
sessor is identical with the topic of the previous sentence, it is phonologi-
cally empty, in which case it can only be reconstructed from a verbal prefix
agreeing with it. Thus, discovering the underlying syntactic structure of
sentences of this type also solves philological, interpretational problems of
Sumerian.
Zolyomi derives the properties observed of the external possession con-
struction as follows: The possessor bears the same case as the possession
because they represent the same argument of the verb, forming a kind of an
appositive construction. The external possession construction is restricted
to cases of inalienable possession because the condition of such an apposi-
tive construction is the referential non-distinctness of its members.
Barbara Egedi's paper entitled Genitive constructions in Coptic com-
pares two possessive constructions of Coptic, examining how they are
structured, and in what way their structures determine their distributions.
As Egedi demonstrates, both possessive constructions involve a 'possession,
genitive marker, possessor' string; however, in pattern A, the genitive
marker of the possessor is a genitive case-ending, while in pattern Β it is a
preposition; furthermore, in pattern A the adjacency of the possession and
the possessor is obligatory, while in pattern B, it is optional.The choice
between pattern A and pattern Β depends on the syntactic properties of the
possession. The case-marked possessor of pattern A is used if the posses-
sion has a mere definite article. If the possession is indefinite, or if it has a
demonstrative definite determiner, or is modified by an adjective, the
prepositional possessor of pattern Β must be used.
So as to derive the attested properties of pattern A, Egedi generates the
genitive-marked possessor in the specifier of the possession NP, as an al-
ternative to the indefinite article. The NP is claimed to be subsumed by a
PossP, with the Ν raised to Poss, and then to Num, and with the possessor
raised to Spec,PossP. The possessor has a weak [+def] feature, which trig-
gers no overt possessor movement to the DP domain, but ensures that the D
head be realized as a default definite determiner - thereby also excluding
the possibility of a demonstrative determiner in D. That is:
 Introduction 11

(3)

def. art. possessor] t,

possessiorij

In pattern B, on the other hand, the possessor is generated as a PP adjunct

right-adjoined to the possession noun phrase, where it can also be subjected
to extraposition. As an adjunct, it does not interact with the [+/-definitene]
feature of the possession; so it is predicted to be compatible with any type
of determiner.

3.3. Word order variation

Word order variation has been interpreted by traditional philology of Clas-

sical Greek and Latin roots as a freedom of word order, that is, a phenome-
non of language which is not regulated by rules, hence falls outside the
realm of grammar. The Päninian approach to Classical Sanskrit - though
essentially generative in its spirit - also shares this assumption of tradi-
tional philology; it does not have anything to say about the word order of
the Sanskrit sentence. Generative theory, on the other hand, has always had
doubts about the existence of genuinely free word order languages. Evi-
dence of subject-object asymmetries in English led to the hypothesis that
the subject is universally more prominent than the constituent subsuming
the verb and the object, hence it is always realized external to the V+O or
O+V complex. The apparent freedom of word order in a language is the
result of reordering rules.
Brendan Gillon and Benjamins Shaer's paper entitled Classical San-
skrit, 'wild trees', and the properties of free word order languages argues
12 Katalin Ε. Kiss

that the underlying constituent order in the Classical Sanskrit sentence is

genuinely free; at the same time, subconstituents displaced to the left and
right periphery also provide evidence of movement. That is, Gillon and
Shaer challenge both Sanskrit philologists and generative syntacticians.
They show to Sanskritologists that the generation of word order, that is, the
linearization of the verb and its complements, is a grammatical question no
matter how flexible the linearization rules be. Furthermore, Sanskrit sen-
tences with discontinuous constituents clearly display patterns of move-
ment licensed by Universal Grammar, constrained by Subjacency, to land-
ing sites on the left and right periphery of the sentence. Constituents
affected by m o v e m e n t to the peripheries appear to be associated with par-
ticular discourse functions, however, in Gillon and Shaer's view, the syn-
tactic and the information-packaging systems of grammar are autonomous,
hence it is legitimate to examine the former with no regard to the latter.
The basic new claim that Gillon and Shaer confront generative syntacti-
cians with is the assumption of a genuinely free underlying order of major
constituents in the Sanskrit sentence. Reviving and modifying the so-called
'wild tree' theory of Staal (1967), they claim that the Classical Sanskrit
clause is generated in one step, with all the major constituents merged into
an unordered flat tree. They raise the possibility that this unordered flat tree
might represent a universal underlying structure. In this view, articulating
constituents in a binary branching hierarchy is a further step licensed - but
not enforced - by Universal Grammar, which English does, and Sanskrit
does not, adopt. Structures in which the verb and all its complements are
sisters to each other have also been argued for before - see e.g. Bresnan
(1982), Hale (1983), E. Kiss (1987), etc. The Gillon-Shaer theory is more
radical than previous proposals in that it does not even assume a fixed posi-
tion for the head in the initial tree.
In current main-stream generative theory, word order variation that is
not elicited by morphological or scope requirements is seen to be motivated
by discourse considerations. The two main clues of identifying discourse
functions are prosody, which cannot be investigated in the case of dead
languages at all, and semantic interpretation, which cannot be reconstructed
with full certainty, either. Chiara Polo and Giampaolo Salvi focus on the
methodological questions of discovering discourse-motivated reordering
operations in a dead language, and also prove that surface variation in Latin
word order can be reduced to an invariant underlying order, and to topicali-
zation and focussing rules also known from other languages.
 Introduction 13

In her paper entitled Latin word order in generative perspective: An ex-

planatory proposal within the sentence domain, Chiara Polo introduces an
ingenious method of reconstructing the missing semantic and prosodic
judgments of Latin native speakers. She chooses a Latin text, Cena Trimal-
chionis by Petronius, which describes an event involving characters, situa-
tions, and happenings that are also familiar to the modern reader, and she
identifies discourse functions in its sentences on the basis of the Italian
translation. Polo's hypothesis is that the Italian translator has interpreted the
series of events described in the Latin text the same way as the ancient
author and the ancient readers did, and has formulated the corresponding
sentences of the Italian translation so as to convey the same discourse
functions, by means of the same discourse-motivated reordering rules.
As a first step of the comparison of Latin and Italian, Polo establishes
the unmarked word orders of the two languages. 70% of the 823 Latin sen-
tences examined are SOV, and 77% of the corresponding Italian sentences
are SVO. The statistical analysis is supplemented with a qualitative exami-
nation of the sentences, which shows that neutral imports are mapped into
an (S)OV structure in 90% of the cases in Latin, while they are mapped
invariably into an (S)VO structure in Italian - so these prevailing patterns
are assumed to represent the basic word orders. It is the 30% of Latin sen-
tences with a word order other than SOV which - or a subset of which - are
expected to be derived by reordering rules targeting constituents specified
as [+contrastive focus], [+emphasis], [+topic], or [+heavy]. The carriers of
these features are recognized on the basis of the context, and on the basis of
the Italian translations. Polo hypothesizes that constituents in non-basic
positions marked for these features occupy the same peripheral positions
that are licensed by Universal Grammar in present-day Italian. The reorder-
ing rules identified include right- and left-dislocation, lefthand and
righthand focalization, and heavy-NP shift, in addition to various types of
V-fronting, e.g. V-to-Foc and V-to-C movement.
Giampaolo Salvi's paper entitled 'Some firm points on Latin word order:
the left periphery' aims to identify the left-peripheral structural positions
that serve as landing sites of the movement rules described in Polo's paper.
Salvi relies on a very complex array of distributional evidence, which he
interprets and evaluates on the basis of assumptions of Universal Grammar.
Salvi demonstrates the methodology of identifying left-peripheral struc-
tural positions on material from Medieval Romance, a descendant of the
Latin language. In Medieval Romance two preverbal slots with different
properties can be distinguished: a clause-initial PI position, and an imme-
14 Katalin Ε. Kiss

diately preverbal P2 slot. A direct object in PI does, a direct object in P2

does not, trigger clitic doubling. If the preverbal constituent occupies PI,
the clitics are adjoined to the V. If, on the other hand, the preverbal con-
stituent is in P2, the clitics are adjoined to the constituent in P2. PI is typi-
cally interpreted as a topic or a frame adverbial, whereas P2 is usually a
wh-phrase or a focus, even if it can also be a topic. (Salvi raises the possi-
bility that P2 in fact corresponds to two positions, a topic slot and a focus
slot, which can be filled alternatively.)
Some facts of Latin suggest that Latin sentence structure instantiates a
version of the sentence structure that can be reconstructed for Medieval
Romance. That is, it appears that the Latin sentence contains a focus pro-
jection (FP) and a topic projection (TopP) on top of IP. The filler of
Spec,TopP/Pl can be doubled by a clause-internal pronoun, and cannot
support a clitic. The position of the clitics (the Wackernagel position) is to
be found after the constituent in Spec,FP/P2, or if no FP is projected, after
the first constituent of IP. At the same time, not all Latin sentences seem to
observe the distributional restrictions and correlations described above. For
example, the left-peripheral constituent in the following example both has a
pronominal double, like a constituent in Spec,TopP/Pl, and is immediately
followed by the pronominal double, like a constituent in Spec,FP/P2.

(4) sed urbana plebes, ea vero praeceps erat de multis causis

but urban populace it in.truth precipitate was for many.ABL reasons
'But the city populace in particular acted with desperation for many
reasons'

Salvi, attempting to assimilate the structure in (4) to the general pattern,

takes a closer look at the pronoun in it. As is well-known from the com-
parative analyses of pronominals, e.g. from Cardinaletti and Starke (1995),
pronouns occur in two or three versions in a language: they can be strong,
or weak, and some languages also have clitic pronouns. Whether or not
strong and weak pronouns differ in form, they have different distributions
and different syntactic properties. Strong pronouns function as contrastive
topics, new topics, or foci. Only anaphoric pronouns are weak. In various
Romance languages descending from Latin, strong and weak pronouns
have different forms. This difference must have had its origin in Latin. That
is, both universal considerations and historical facts suggest that Latin must
also have distinguished strong and weak pronouns, even if they appeared
identical in the written language. Thus a pronoun like ea could, in principle,
 Introduction 15

represent either a weak or a strong form. The ea in (4) must be strong, be-
cause it supports the enclitic vero.
In view of these, the structure in (4) does not contradict the hypothe-
sized sentence structure any more; the pronoun doubling the topic constitu-
ent is a strong pronoun occupying Spec,IP, hence it is not subject to the
constraints on weak pronoun placement.
The left periphery of embedded clauses is somewhat more complex than
the [ΤΟΡΡ· · · [ F P - · · [IP [VP··· structure identified in main sentences; an embed-
ded clause also contains a clause-initial relative WH-element, or a pre- or
post-topic complementizer. In order to provide place for these constituents,
Salvi supplements his left periphery-model along the lines proposed by
Rizzi (1997). In Rizzi's theory, the complementizer domain of the sentence
has three layers, to be occupied by a relative WH-element, a topic, and a
focus, respectively. The complementizer can appear in the head position of
any of these projections. (In fact, in Latin it has to be ensured that the com-
plementizer precede the filler of Spec,FP, i.e., it occupy the head of the
projection harboring the relative WH-phrase or that harboring the topic.) In
embedded clauses the complementizer also provides an additional target for
weak pronoun placement.
Lisa Lai-Shen Cheng, Aniko Liptäk, and Chris Reintges's paper entitled
The nominal cleft construction in Coptic Egyptian analyzes a focussing
device of Coptic which is also known from many present-day languages of
the world. Its discourse function being clear, the investigation aims to es-
tablish how the parameters involved in the formation of the construction are
to be set so that the restrictions attested fall out. These are the following :
(i) The cleft constituent is a DP in the left periphery of the sentence, (ii)
The backgrounded proposition is represented by a relative clause whose
relative pronominal element is coindexed with the cleft constituent. In sub-
ject relatives the relative pronoun is empty, otherwise it is a resumptive
pronoun in situ, (iii) The element linking the cleft constituent and the back-
grounded proposition is a deictic pronoun agreeing in number and gender
with the cleft constituent.
The authors claim that the Coptic nominal cleft construction is to be
derived from a small clause, with the cleft constituent functioning as its
subject, and the relative clause functioning as its predicate. This explains
why the cleft constituent cannot be anything but a DP. The relative clause
has a relative operator in situ, represented by a resumptive pronoun. The
authors adopt an analysis in which the pronominal operator undergoes
movement to Spec,CP in narrow syntax; at PF, however, the lower copy of
16 Katalin Ε. Kiss

the operator-variable chain is spelled out. Subject relatives, on the other

hand, need not move since they are local to the C position also when left in
situ. They can remain empty because the adjacent relative complementizer
in C allows them to remain empty - as is also attested in many other lan-
guages. The deictic element copying the number and gender features of the
cleft constituent is generated in the head position of the small clause (AgrP)
in which the cleft construction originates. It functions there as an agreement
clitic that overtly marks the subject-predicate relation between the clefted
NP and the relative clause.
The focus interpretation of the cleft constituent is a consequence of the
fact that the small clause constituted by the cleft constituent, the deictic
copula, and the relative clause is embedded under a focus projection. The
head of the focus phrase has an uninterpretable focus feature, which attracts
the subject of the small clause to Spec,FocP.

3.4. Morphosyntactically motivated word order variation

Two papers of the volume, The correlation between word order alterna-
tions, grammatical agreement and event semantics in Older Egyptian by
Chris H. Reintges, and VSO and Left-conjunct Agreement: Biblical Hebrew
versus Modern Hebrew by Edit Doron, deal with VSO-SVO word order
variations. Both of the studies derive the variation from a lexical idiosyn-
cracy: the lack or presence of an EPP feature on a functional category ex-
tending the verb phrase. Both of them also relate the variation to subject-
verb agreement, linking VSO to the lack of agreement, and SOV to the
presence of it. However, whereas in Reintges's theory, it is agreement that
determines SVO order (SVO being a consequence of the EPP feature of
Agreement), in Doron's approach it is the other way round; it is the SVO
order (triggered by an optional EPP feature of Tense) that elicits agreement.
Reintges argues that in Older Egyptian, the VSO-SVO word order variation
is not random but is linked to the type of the eventuality; sentences describ-
ing an event have a VSO word order, whereas those expressing a state are
SVO. This distribution of preverbal and postverbal subjects might recall a
theory elaborated by Kratzer (1995), Maleczki (1999), and others, accord-
ing to which sentences with an event variable can have a spatiotemporal
expression (whether spelled out or unarticulated) in their external argument
position, which allows the subject to remain in the VP. Sentences describ-
ing a state, which lack an event variable, on the other hand, can only have
 Introduction 17

their subject externalized. This explanation, however, would not exclude

the possibility of an SVO order in eventive sentences, which is not attested
in Older Egyptian. Furthermore, the subject in VSO sentences is shown by
Reintges to occupy Spec,IP, the position of the external argument. The
explanation to be proposed must also account for further curious correla-
tions. Namely, in stative, SVO sentences, the verb bears a personal inflec-
tion, which is underspecified in some cases, but is always spelled out. In
eventive, VSO sentences, on the other hand, the V is supplied with an
agreement element which is always specified for person, number, and gen-
der when present, but is missing if the subject is represented by a lexical
noun phrase.
In Older Egyptian, the eventive or stative nature of a sentence is deter-
mined by the choice of the verbal paradigm; it is not a consequence of the
lexical meaning of verbs. No matter if their primary meaning is eventive or
stative, verbs can appear both in the eventive and in the stative paradigm,
expressing slightly different eventualities in the two cases. Superficially, the
two paradigms display inflectional systems of a similar kind. As Reintges
shows, however, the apparent agreement markers of the eventive paradigm
are enclitic pronouns; that is why they are not spelled out in the presence of
a lexical subject. In the stative paradigm, on the other hand, the verb is truly
inflected; its agreement suffix (often underspecified) is spelled out in the
presence of either a pronominal or a lexical subject.
With these issues clarified, the scene is set for the explanation of the
VSO-SVO variation. Reintges derives both word orders from a canonical vP.
The VSO order of eventive sentences can arise either as a result of V-to-T
movement, or as a a result of (V+T) movement to a functional head in the
complementizer domain.

(5) a. [Tp V+T [vp SU [ V '...(V)...DO...]]]

b. [ f p V + T [ T P S U [r(V+T)[vP(SU)[v..V...DO...]]]]

In structure (5a), sentential adverbs, negation, and shifted objects precede

the subject; in (5b), on the other hand, they follow it. (5a) represents the
unmarked word order; (5b) arises by V movement to a functional head with
information structure (topic or focus) content.
In stative SVO sentences both the V and the subject precede TP, but
they do not reach the CP domain. In fact, no CP is projected - as is clear
from the fact that stative sentences can occur in Exceptional Case Marking
constructions. The SVO order of stative sentences is the result of V-to-Agr
18 Katalin Ε. Kiss

movement, and subject movement to Spec,AgrSP. That is, the word order
difference between eventive and stative sentences is eventually reduced to
the presence of an AgrP projection with an EPP feature in the latter.
Edit Doron demonstrates in her paper that the seemingly identical VSO
sentences of Biblical Hebrew and Modern Hebrew represent different struc-
tures; Biblical Hebrew sentences have the structure in (5a), whereas Mod-
ern Hebrew sentences display the structure in (5b). This difference is de-
rived from the assumption that in Biblical Hebrew, the Tense head does not
have an intrinsic EPP feature (although it can be supplied with an EPP fea-
ture optionally, which yields an SVO sentence). In Modern Hebrew, on the
other hand, EPP is a lexical property of Tense.
In Biblical Hebrew, the lack of the attraction of the subject to Spec,TP
also affects the Agree relation between the subject and the verb. This be-
comes evident in the case of a conjoined subject of the following type:

If Τ has an EPP feature (as is always the case in Modern Hebrew and can
also happen in Biblical Hebrew), it is the whole conjoined noun phrase (the
highlighted DP) that is subjected to movement to Spec,TP and agreement
with V+T - owing to constraints on movement. If, on the other hand, Τ
does not have an EPP feature (as is usually the case in Biblical Hebrew), Τ
can agree with the minimal D constituent closest to it which allows the
derivation to converge. That is, the VSO sentences of Biblical Hebrew dis-
play left-conjunct agreement; for example, in the equivalent of 'And lifted
David and the people with him their voice', the verb is in the singular.
In a few V S O sentences of the Bible, we attest full agreement between
the verb and a conjoined subject noun phrase following it. As Doron demon-
strates, these sentences all represent structure (5b); the verb has been raised
to a functional projection above TP, while the postverbal subject still occu-
pies Spec,TP.
 Introduction 19

3.5. The structure of coordinated phrases

Emanuele Lanzetta and Lucio Melazzo's paper entitled A particular coor-

dination structure of Indo-European flavor aims to account for the various,
often unexpected positions of the conjunction in coordinated constructions
of ancient Indo-European languages, among them Hittite, Ancient Greek,
Latin, and Vedic Sanskrit.
As is well-known, the conjunction in Latin coordinated phrases consisting
of two conjuncts is cliticized to the second conjunct (terra marique 'by.land
by.sea-and'), or to both conjuncts (remque prolemque 'wealth-ACC-and
issue-ACC-and'). However, the two conjunctions of a coordinated expres-
sion can also appear side by side, between the two coordinated nominals, as
in the following Ancient Greek example:

(7) αρό kratos te ka δτηδη

from head-GEN-and and shoulders-GEN
'from (his) head and shoulders'

Surprisingly, the conjunction may also appear cliticized after the first con-
junct - e.g. in the following Vedic Sanskrit example:

(8) devebhyas ca pitribhya

Gods-DAT-and fathers-DAT
'to the gods and the fathers'

This example seemingly contradicts not only the Latin examples cited above,
but also what we attest in present-day Indo-European languages. A con-
junction cliticized to the first conjunct is sharply ungrammatical - as illus-
trated by the English example *He left and. He didn't even say good-bye
(cf. Munn 1992).
On the basis of this seemingly contradictory array of facts, a traditional,
inductive description cannot but conclude that anything goes; the conjunc-
tion^) can stand anywhere in the coordinated phrase except in initial posi-
tion - which is not true. (For example, no 'XP and XP XP' order is attested.)
If, on the other hand, the binary branching conjunction theory following
from the principles of Universal Grammar, formulated in somewhat differ-
ent versions by Munn (1992) and Kayne (1994), is used as the framework
of the analysis, all the facts attested can be derived, and no unattested pos-
sibilities are predicted. In this framework, coordinated phrases are assigned
the following universal structure:
20 Katalin Ε. Kiss

(9) &P

& YP

If & is an enclitic, Y, the nominal constituting the second conjunct is left-

adjoined to it via head movement.
This is the underlying structure of coordinated phrases containing two
conjunctions according to Lanzetta and Melazzo - with &1P occupying the
specifier of &2P.

(10) &2P

&1 XP &2 YP

X Y

In this structure both X and Y undergo head movement to &1 and &2, re-
spectively - unless e.g. &2 is not enclitic, as happens in (7). In this frame-
work, the source of the seemingly idiosyncratic example in (8) is a version
of structure (10), in which the 2nd conjunction (&2) is phonetically null.
The benefits of deriving the coordination possibilities of ancient Indo-
European languages from the binary branching hypotheses of Universal
Grammar are mutual: not only seemingly idiosyncratic facts of a number of
languages receive a principled explanation, but also a speculative hypothe-
sis of Universal Grammar receives empirical support; ancient Indo-
European languages realize possibilities that are derivable from the Munn-
Kayne theory but have not been attested so far.

3.6. The function of the D-stem of the Semitic verb

In his paper entitled Complex predicate structure and pluralized events in

Akkadian, Christian Huber examines the seemingly contradictory functions
 Introduction 21

associated with a particular verb-stem, and manages to reduce the apparent

idiosyncracy to the interaction of argument structure, event structure, and
quantification.
The Akkadian verb has a triconsonantal root, which participates in various
fixed morphological templates called stems. One of these stems, the so-called
D-stem, seems to have different functions depending on whether the verb is
unaccusative, transitive, or unergative. In the case of unaccusative verbs,
the D-stem derives a transitive verb from the basic G-stem; thus the D-stem
of the Akkadian equivalent of ' g r o w ' means ' m a k e bigger'. In the case of
transitive and unergative verbs, on the other hand, the D-stem adds no ar-
gument; thus the D-stems of the Akkadian equivalents of the transitive
' o p e n ' or the unergative 'whisper' also mean ' o p e n ' and ' w h i s p e r ' , respec-
tively. Huber aims to clarify in what respect the D-stem of a transitive or an
unergative verb is different from the G-stem of the same verb, and whether
or not the D-stems of unaccusatives and transitives/unergatives are derived
in the same way.
As for the former question, Huber observes that the use of the D-stem is
often triggered by the presence of a plural argument or adjunct. Thus
whereas in the Akkadian equivalent of Ί opened a canal' the V occurs in
the G-stem, in the the equivalent of Ί opened canals' the D-stem is used. In
some sentences, e.g. in the eqivalent of Ί enlarged the garden of the pal-
ace'; or, The wall which R(oyal) N ( a m e ) l , RN2, RN3, RN4, and RN5 had
built...', the use of the D-stem is somehow related to the extended nature of
the process described. Assuming the theoretical framework of Pustejovsky
(1991), in which events fall into states, processes, and transitions, Huber
claims that all these sentences express a plurality of transitions. The multi-
plicity of transitions can mean a series of independent transitions, as in the
case of Ί opened canals'. In another type, the plurality of transitions is a
recursive series, with the output of one transition serving as input for the
next transition. Sentences such as Ί enlarged the garden of the palace' be-
long to this type. Finally, the plural event can consist of proto-events
(proto-transitions) which constitute a single transition in their totality,
yielding a mass-noun-like interpretation of the count noun. This type is
represented by sentences like ' T h e wall which R(oyal) N ( a m e ) l , RN2,
RN3, RN4, and R N 5 had built...'
Huber concludes that the D-stem supplies the verb phrase with a func-
tional projection, a N u m P , whose head serves to pluralize events. The reason
why the D-stem also adds an external argument to unaccusative verbs;
more precisely, why it extends the unaccusative V P into a vP projection,
22 Katalin Ε. Kiss

must be that the N u m head involved in D-stems can only merge with a vP.
Thus the interpretation of the D-stem of the Akkadian verb as a N u m P pro-
jection subsuming a vP accounts for both the seemingly idiosyncratic effect
of the D-stem on argument structure, and the sometimes delicate difference
between the interpretations of the G-stem and D-stem of one and the same
transitive or unergative verb. Huber's analyses can also be extended to
similar facts of related languages. For example, the corresponding stems of
Arabic and Hebrew verbs can have an 'intensive' interpretation. Huber
analyzes such sentences (meaning, for example, 'x bit y (fiercely or repeat-
edly)') as expressing cumulative affectedness, with either a series of transi-
tions predicated of the same entity, or the output of one transition serving
as input for the next transition.

4. New insights gained from a historical perspective

Most studies in this volume intend to provide a synchronic analysis of a

particular stage of the language examined - however, the stages they ana-
lyze typically represent an intermediate state of the given language with not
only its previous state but also its subsequent states documented. This fact
makes it possible from time to time for the reader to catch a glimpse of the
triggers and the process of language change.
For example, when Chiara Polo finds in her analysis of Latin word order
that only 9 0 % of the unmarked, neutral sentences of the Latin text exam-
ined have an SOV word order; 10% of them are SVO, it is clear in view of
the later development of Latin that the neutral SVO sentences forecast the
basic SVO word order of its daughter languages. As Polo observes, the
postverbal objects in these sentences are non-prototypical objects, ranked
high in the animacy, humanity and definiteness scale. Such objects were
presumably targets of rightward Topicalization, rightward Focalization, or
Heavy NP-shift performed on an SOV base - but the discourse motivation
for their movement was weak, therefore they could be interpreted as argu-
ments in situ when attested by a new generation of speakers acquiring their
mother tongue.
Annamaria Bartolotta's paper, entitled IE *weid- as a root with dual
subcategorization features in the Homeric poems: A Minimalist approach,
calls attention to a shift in the history of Greek, identified as a shift from
the active-stative language type to the nominative-accusative type.
 Introduction 23

The problem that Bartolotta aims to analyze is the following: even though
the Greek perfect (w)oida Ί see (with the mind's eye)' and the aorist
(w)iidon Ί saw' are verb forms derived from the same root *weid- 'to see',
associated with the same theta-grid consisting of an experiencer and a the-
me, they select different cases for their theme argument. The aorist form
takes an accusative object, whereas the perfect form occurs both with an ac-
cusative and a genitive object. The genitive appears to be the older variant; in
Iliad there are 23 occurrences of it, with only three in Odyssey. Bartolotta
wants to answer two questions: why one and the same verb had its theme
argument marked with different cases in different tenses at one stage of the
language; and why the genitive marking of the theme of the perfect verb
form eventually disappeared.
As for the first question, Bartolotta demonstrates that the Indo-European
*weid- stem underwent a meaning-split; perceptive vision, i.e., simple eye-
sight, came to be differentiated from intentional vision, i.e., internal acqui-
sition, thinking. The different tense forms were likely to elicit one or the
other of the two meanings of the verb; the [-stative] aorist form elicited the
'perceptive sight' interpretation, whereas the [+stative] perfect form evoked
the 'intentional vision' meaning.
In the early Greek represented by Iliad, the aorist (w)eidon, denoting per-
ception, always takes an accusative object because in that language variant
verbs expressing immediate contact with their object take an accusative
object. The object of the perfect (w)oida, meaning 'see with the mind', on the
other hand, is in the genitive because in that period Greek verbs denoting
mediated contact with their object take a genitive object.
The process in the course of which the genetive marking of the object of
(w)oida 'see with the mind' gradually gave way to accusative marking is
claimed by Bartolotta to be a manifestation of Greek developing from the
active-stative system of Pre-Indo-European to a nominative-accusative sys-
tem. This typology, developed in Indo-European linguistics (cf. Lehmann
1993), is recast by Bartolotta in a generative terminology. In her formula-
tion, the change from the active-stative system to the nominative-accusative
system brought about the following major changes: The semantic-lexical
relationship between the verb and its complements typical of the active-
stative type, with the N P matching the verb in its [+/-stative] feature, be-
came a structural relationship of the nominative-accusative type. Inherent
case assignment coupled by theta-role assignment gave way to structural
case assignment with no theta-role assignment involved. In the active-
stative system, the object is assigned a theta-role and case in situ; in the
24 Katalin Ε. Kiss

nominative-accusative type, on the other hand, it has to move for case. In

the active-stative type, aspect-marking plays a major role, whereas nomina-
tive-accusative languages are tense-marking languages (with the aspectual
system incorporated into the tense system). The different case-marking of
objects affected in different ways by the activity denoted by the verb is
characteristic of active-stative languages. The loss of the genitive marking
of the object of (w)oida is evidence of the loss of the theta-related case
system, and the emergence of the structural cases typical of nominative-
accusative languages.

Acknowledgments

I am grateful to my former student and present-day colleague Gabor

Zolyomi, and to my student Barbara Egedi, whose work first called my
attention to the fact that dead languages can be fruitfully studied in the gen-
erative framework. Thanks are due to all the authors of the volume - not
only for their creative contributions, but also for their participation in an
internal reviewing process. We owe thanks also to Tor Äfarli, Maya Arad,
Huba Bartos, Aniko Csirmaz, and Michael Streck for their helpful com-
ments. The editorial work was partially supported by Grant TS040705 of
OTKA, the Hungarian Scientific Research Fund.

References

Andrews, Avery D.
1971 Case agreement of predicate modifiers in Ancient Greek. Linguistic
Inquiry 2: 127-151.
Benucci, F.
1996 Studi di sintassi umbra. II verbo nelle Tavole Iguvine e nelle iscrizioni
minori. Padova: Libraria Padovana.
Bertocchi, A.
1989 The role of antecedents of Latin anaphors. In Subordination and
Other Topics in Latin, Gualtiero Calboli (ed.), 441^462. Amsterdam:
John Benjamins.
Bertocchi, A. and C. Casadio
1983 Anaphoric relations, pronouns and Latin complementation. In Latin
Linguistics and Linguistic Theory, Harm Pinkster (ed.), 27-40. Am-
sterdam: John Benjamins.
 Introduction 25

Binkert, P. J.
1970 Case and prepositional constructions in a transformational grammar
of Classical Latin. [Unpublished Ph.D. diss., University of Michigan.]
Bolkenstein, Machtelt
1979 Subject-to-object raising in Latin, Lingua 49: 15-34.
Bresnan, Joan
1982 Control and complementation. Linguistic Inquiry 13: 343-433.
Calboli, Gualtiero
1983 The development of Latin (Cases and infinitive). In Latin Linguistics
and Linguistic Theory, Harm Pinkster (ed.), 42-58. Amsterdam: John
Benjamins.
Calboli, Gualtiero (ed.)
1989 Subordination and Other Topics in Latin. Amsterdam: John Benjamins.
Cardinaletti, Anna and Michal Starke
1999 The typology of structural deficiency: a case study of three classes of
pronouns. In Clitics in the languages of Europe, Henk van Riemsdijk
(ed.), 145-233. Berlin: Mouton de Gruyter.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the analysis of Latin infinitival clauses for the
theory of case and control. Lingue e linguaggio 1: 151-189.
DeCaen, Vincent J. J.
1995 On the placement and interpretation of the verb in standard Biblical
Hebrew prose. [Unpublished Ph.D. diss., University of Toronto.]
Doron, Edit
2000 V S O and left-conjunct agreement: Biblical Hebrew vs. Modern He-
brew. In The syntax of verb-initial languages, Andrew Carnie and
Eithne Guilfoyle (eds.), 75-96. Oxford: Oxford University Press.
Eh 1 ich, Κ.
1981 Native Speaker's heritage. On philology of 'dead' languages. In A
Festschrift for Native Speaker, F. Coulmas (ed.), 153-165. The Hague:
Mouton.
Embick, David
2000 Features, syntax, and categories in the Latin perfect. Linguistic In-
quiry 31: 185-230.
Embick, David and Morris Halle
1999 The Latin conjugation. [Unpublished Ms., MIT.]
Garrett, A.
1990 The Syntax of Anatolian Pronominal Clitics. [Unpublished Ph.D.
dissertation, Harvard University.]
Giusti, Giuliana
2001 The birth of a functional category: From Latin ILLE to the Romance
article and personal pronoun. In Current studies in Italian syntax,
Essays offered to Lorenzo Renzi, Guglielmo Cinque and Giampaolo
Salvi (eds.), 157-171. Amsterdam: Elsevier.
26 Katalin Ε. Kiss

Gragg, Gene B.
1968 The syntax of the copula in Sumerian, in: J.W.M. Verhaar (ed.), The
verb 'be' and its synonyms, III (Foundations of Language Supple-
mentary Series 8). Dordrecht, 86-196.
1973 Linguistics, method, and extinct languages: The case of Sumerian.
Orientalia N.S. 42: 78-96.
Groneberg, Brigitte R. Μ.
1987 Syntax, Morphologie und Stil der jungbabylonischen hymnischen
Literatur, I—II (Freiburger altorientalische Studien 14. I—II.) Stuttgart:
Franz Steiner Verlag.
Hale, Kenneth
1983 Warlpiri and the grammar of non-configurational languages. Natural
Language and Linguistic Theory 1: 5 - 4 7 .
Hale, Mark
1987 Studies in the comparative syntax of the oldest Indo-Iranian lan-
guages. [Unpublished Ph.D. dissertation, Harvard University.]
1990 Preliminaries to the study of the relationship between syntax and
sandhi in Rigvedic Sanskrit. Münchener Studien zur Sprachwissen-
schaft 51: 77-96.
1991 Some observations on intersentential pronominalization in the language
of the Taittirlya Samhitä. In Sense and Syntax in Vedic, J. Brereton
and S. Jamison (eds.). Leiden: Brill.
1993 Tmesis and movement in Avestan. Indo-Iranian Journal 36: 29^13.
1996 Clitic placement in the language of the Rigveda. In Approaching
Second. Second Position Clitics and Related Phenomena, Aaron L.
Halpern and Arnold M. Zwicky (eds.), 165-197. Stanford: CSLI
Publications.
Herman, Jözsef (ed.)
1994 Linguistic Studies on Latin. Selected papers from the 6th Interna-
tional Colloquium on Latin Linguistics, Budapest, 23-27 March,
1991. Amsterdam: John Benjamins.
Holmstedt, Robert
2002 The relative clause in Biblical Hebrew. [Ph.D. dissertation, University
of Wisconsin at Madison. Forthcoming as The relative clause in Bib-
lical Hebrew. Linguistic Studies in Ancient West Semitic. Winona
Lake, Ind.: Eisenbrauns.]
Huber, Christian
1989/90 Ergative Sprachen in der GB-Theorie und einige Überlegungen zum
Sumerischen. Wiener Linguistische Gazette 4 5 - 4 6 : 53-78.
1996 Some notes on transitivity. Verb types, and case with pronouns in
Sumerian. Wiener Zeitschrift für die Kunde des Morgenlandes 86
(.Festschrift Hirsch)·. 177-189.
 Introduction 27

forthc. Some remarks on focus and relative clauses in Sumerian. In Work-

shop on diachronic and synchronic variations in the phonology,
morphology, and syntax of Sumerian, Papers presented at the 6th
meeting of the Sumerian Grammar Discussion Group, Oxford, 17th
and 18th September 1999, Jeremy Black and Gabor Zolyomi (eds.).
(Acta Sumerologica 22.)
Kayne, Richard S.
1994 The Antisymmetry of Syntax. Cambridge, Mass.: M I T Press.
Kiparsky, Paul
1979/80 Pänini as α Variationist. Cambridge, Mass.: M I T Press, and Poona
University Press.
1982 Some theoretical problems in Pänini's grammar. Poona, India:
Bhandarkar Oriental Research Institute.
1995 Päninian linguistics. In Concise History of the Language Sciences:
from the Sumerians to the Cognitivists, E. F. K. Koerner and R. E.
Asher (eds.), 59-65. New York: Pergamon Press.
Kiparsky, Paul and J. F. Staal
1969 Syntactic and semantic relations in Pänini. Foundations of Language
5:83-117.
Kiss, Katalin E.
1987 Configurationality in Hungarian. Dordrecht: Reidel.
Kratzer, Angelika
1995 Stage-level and individual-level predicates. In The Generic Book, G.
Carlson and F. J. Pelletier (eds.), 125-175. Chicago: The University
of Chicago Press.
Lakoff, Robin
1968 Abstract Syntax and Latin Complementation. Cambridge, Mass.:
M I T Press.
Lehmann, W. P.
1993 Theoretical Bases of Indo-European Linguistics. L o n d o n / N e w York:
Routledge.
Lightfoot, David
1975 Natural Logic and the Greek Moods. The Hague: Mouton.
Maleczki, Märta
1999 Weak subjects in fixed space. Acta Linguistica Hungarica 46: 95-118.
Maraldi, Mirka
1983 New approaches to accusative subjects: Case study vs. raising. In
Latin Linguistics and Linguistic Theory, Harm Pinkster (ed.), 177-
196. Amsterdam: John Benjamins.
Maurel, J.-P.
1983 Les relatives en Latin: 'Raising' ou 'matching'? In Latin Linguistics
and Linguistic Theory, Harm Pinkster (ed.), 167-176. Amsterdam:
John Benjamins.
28 Katalin Ε. Kiss

Maurel, J.-P.
1989 Subordination seconde du relatif en latin et theorie du 'COMP'. In
Subordination and Other Topics in Latin, Gualtiero Calboli (ed.),
181-196. Amsterdam: John Benjamins.
Morrell, K. S.
1989 Studies on the phrase structure of early Attic prose. [Unpublished
Ph.D. dissertation, Harvard University.]
Munn, Alan
1992 A null operator analysis of ATB gaps. The Linguistic Review 9: 1-26.
Murru, F.
1977 Studio di un modello semantico-generativo per l'insegnamento del
latino. II problemo dei casi. [Unpublished Ph.D. dissertation, Univer-
sity of Torino.]
Oniga, Renato
2004 Breve introduzione linguistica. Milano: Franco Angeli.
Ostafin, David Mark
1986 Studies in Latin word order: A transformational approach. [Unpub-
lished Ph.D. dissertation, University of Connecticut at Storrs.]
Pepicello, William
1977 Raising in Latin. Lingua 42: 209-218.
Philippaki-Warburton, I. - G. Catsimali
1997 Control in Ancient Greek. Greek linguistics '95. In Proceedings of
the 2nd International Conference on Greek Linguistics, Salzburg,
September 1995, G. Drachman, A. Malikouti-Drachman, J. Fykias
and C. Klidi (eds.), 577-588. Graz: W. Neugebauer Verlag.
Pillinger, Ο. Stephen
1980 The accusative and infinitive in Latin: A refractory complement
clause. Journal of Linguistics 16: 55-83.
Pinkster, Harm (ed.)
1983 Latin Linguistics and Linguistic Theory. Amsterdam: John Benjamins.
Polo, Chiara
2003 Latin, Italian and Slovene between Morphology and Syntax. [Ph.D.
dissertation, University of Padua, to be published by Unipress in
2004.]
Pustejovsky, J.
1991 The syntax of event structure. Cognition 41: 47-81.
Reintges, Chris H.
1996 Egyptian Root-and-Pattern Morphology. Proceedings of the Interna-
tional Conference on Egyptian Grammar. Crossroads 111, Yale 1994.
Lingua Aegyptia 4: 213-244.
1997 Passive voice in Older Egyptian. (HIL Dissertations 28.) The Hague:
Holland Academic Graphics.
 Introduction 29

1998 Mapping information structure to syntactic structure: One syntax for

Jn. Revue d'egyptologie 49: 195-220.
2000 The licensing of gaps and resumptive pronouns in Older Egyptian
Relatives. In Research in Afroasiatic Grammar. Papers from the
Third Conference on Afroasiatic Languages, Sophia Antipolis, 1996,
(Current Issues in Linguistic Theory 202), Jacqueline Lecarme, Jean
Lowenstamm and Ur Shlonsky (eds.), 243-262. Amsterdam: John
Benjamins.
2001 Aspects of the morphosyntax of subjects and objects in Coptic Egyp-
tian. In Linguistics in the Netherlands 2001, H. Broekhuis and T. van
der Wouden (eds.), 177-188. Amsterdam: John Benjamins.
2003 Syntactic conditions on special inflection in Coptic interrogatives. In
Research in Afroasiatic Grammar II. Papers from the Fifth Confer-
ence on Afroasiatic Languages, Paris, June 2000, Jacqueline Le-
carme (ed.), 363^408. (Current Issues in Linguistic Theory 241.)
Amsterdam: John Benjamins.
Rizzi, Luigi
1997 The fine structure of the left periphery. In Elements of Grammar,
Liliane Haegeman (ed.), 281-337. Dordrecht: Reidel.
Salvi, Giampaolo
1996 From Latin weak pronouns to Romance clitics. (Linguistica Series C
Relationes, 9.) Budapest: Research Institute for Linguistics of the
Hungarian Academy of Sciences.
forthc. La formazione della struttura di frase romanza. Ordine delle parole
e clitici dal latino alle lingue romanze antiche. [Habilitation disserta-
tion, 1999] Tübingen: Niemeyer.
Schaufele, Steven
1990 Free word-order syntax: The challenge from Vedic Sanskrit to con-
temporary syntactic theory. [Ph.D. dissertation, University of Illinois
at Urbana-Champaign.]
1993 The Vedic clause-initial string and Universal Grammar. Studies in
the Linguistic Sciences 23: 131-161.
1996 Now that we're all here, where do we sit? Phonological ordering in
the Vedic clause-initial string. In Approaching Second. Second Posi-
tion Clitics and Related Phenomena, Aaron L. Halpern and Arnold
M. Zwicky (eds.), 447-475. Stanford: CSLI Publications.
Tantalou, Niki
2003 Infinitives with overt subjects in Classical Greek. Studies in Greek
Linguistics 23: 358-365.
Taylor, Ann
1990 Clitics and configurationality in Ancient Greek. [Unpublished Ph.D.
dissertation, University of Pennsylvania.]
30 Katalin Ε. Kiss

Taylor, Ann
1994 The change from OV to VO in Ancient Greek. Language Variation
and Change 6: 1-37.
1996 A prosodic account of clitic position in Ancient Greek. In Approach-
ing Second. Second Position Clitics and Related Phenomena, Aaron
L. Halpern and Arnold M. Zwicky (eds.), 477-503. Stanford: CSLI
Publications.
Touratier, Christian (ed.)
1985 Syntaxe et latin, Actes du lime Congres International deLinguistique
Latine, Aix-en-Provence, 28-31 Mars 1983. Aix-en-Provence, Mar-
seille: Universite de Provence/J. Lafitte.
Wales, M. L.
1982 Another look at the Latin accusative and infinitive. Lingua 56: 127-
152.
Zolyomi, Gabor
1996 Genitive constructions in Sumerian. Journal of Cuneiform Studies
48: 39^15.
The correlation between word order alternations,
grammatical agreement and event semantics in
Older Egyptian

Chris H. Reintges

1. Introduction

In current research on word order typology, the systematic differences be-

tween subject and verb-initial languages are derived from a single parame-
ter of variation, viz. the active or inactive status of the 'Extended Projection
Principle' (EPP). S V O languages like English have a positive setting for
the EPP. As a result, the highest inflectional node projects an extra specifier
position into which the subject is merged. By contrast, V S O languages like
Modern Irish have a negative setting for the EPP. The highest inflectional
node does not project a specifier position for the subject, which is therefore
licensed in a lower syntactic position (see Alexiadou & Anagnostopoulou
1998 and McCloskey 2001 for representative studies).
From a comparative perspective, languages that display both V S O and
S V O patterns are of particular interest as they provide us with some insight
into the workings of the EPP and subject licensing. Older Egyptian is such
a language. 1 On the surface, it meets the syntactic profile of Greenberg's
(1966) Sixth Universal, according to which 'all languages with a dominant
V S O order have S V O as an alternative' (p. 79). Compare the verb-initial
sentence in ( l a ) with the subject-initial sentence in ( l b ) , which both contain
the epistemic verb 'to learn (about)'. 2

(1) a. The dominant V S O clausal pattern

]-'J Pjpj pn mwt-f
learn EV Pepi DEM:SM mother-3SM
η χπι Pjpj pn mwt-f ßdt sfp-t
NEG ignore E V Pepi DEM:SM mother-3SM white.crown splendid-SF
'This (King) Pepi will recognize his mother. This (King) Pepi will
not ignore his mother, the splendid white crown.'
(Pyramid Texts 910a/P)
32 Chris Η. Reintges

b. The S V O alternative
n-ntt N N pn rx(-w) rn n(j) wßi'-w (...)
since N N DEM:SM learn-3MSTAT name LINK(-SM) fowler-PM
'Since this N N (the male deceased) knows (by learning) the names
of the fowlers ( . . . ) . ' (Coffin Texts VI 2 2 o / B l B o )

The contrast between verb-initial and subject-initial word order does not
reflect discourse-configurationality, where the variable position of the sub-
ject is related to its topic or focus role. Rather, V S O and SVO structures are
associated with different aspectual viewpoints from which a given situation
is presented. The V S O clause in ( l a ) above has an event-related interpreta-
tion, describing the acquisition of some knowledge, while the SVO 'alter-
native' in ( l b ) above has a state-related interpretation, describing the pos-
session of some knowledge through learning. Apart from word order, Older
Egyptian employs two morphologically distinct finite verb conjugations,
the Eventive and the Stative, to formally distinguish event- and state-
denoting verbs that are derived from the same root. The Eventive-Stative
alternation that yields minimal pairs like j.r% 'learn about' and rx(-w)
'know (through learning)' is fully productive in various lexical classes of
transitive, unergative and unaccusative verbs.
Older Egyptian thus represents the typologically marked case of a lan-
guage where a stative-resultative verb form cannot be derivationally related
to a non-stative base form, but where the members of the opposition, stative
and eventive, are encoded by different types of inflectional paradigms
(Nedjalkov & Jaxontov 1988: 29). The aim of this study is to clarify the
complex relation between word order alternations, subject-verb agreement,
and event semantics. I will argue that the aspectual properties of verbs are
not specified at a lexical, but rather at a syntactic level. To assume an
event- or state-related interpretation, the subject and the verb must appear
in a particular hierarchical relation with one another. On the other hand, the
eventive or the stative interpretation of the main verb has a morphological
correlate in the finite verb inflection. I will show that grammatical agree-
ment in the traditional sense is only represented by the Stative verbal para-
digm, while the Eventive conjugation lacks agreement proper. In the
Stative, the presence of agreement excludes tense- and aspect morphology,
while the corresponding Eventive is compatible with the full range of Older
Egyptian tense-aspect-mood and voice marking. I will also provide a con-
figurational analysis of Eventive V S O and Stative S V O sentences. What I
want to show in particular is that V S O order does not correspond to a single
 Word order correlations in Older Egyptian 33

syntactic structure. Rather, there is syntactic variation without morphological

variation that has its roots in different subject positions and targets of verb
movement.
The organisation of the paper is as follows. Section 2 examines the se-
mantic differences between Eventive and Stative sentences. Section 3 ad-
dresses the categorial status of the concord-marking ending on Eventive
and Stative verb forms. Section 4 discusses the different routes by which
verb-initial surface order can be arrived at. The SVO alternative, by con-
trast, corresponds to a single syntactic derivation which involves verb move-
ment and subject raising to the head and the specifier position of the subject
agreement phrase. Section 5 summarises the main findings of this paper.

2. The aspectual nature of paradigmatic complexity

Finite verb inflection in Older Egyptian has a portmanteau character in that

it registers not only a particular constellation of subject and verb, but also
provides aspectual information about the type of situation that is described.
Before examining the gradual modification of lexical meaning by means of
finite verb inflection, I will first provide some background information on
situation aspect and verb classification.

2.1. Theoretical background

Situation aspect (or Aktionsart) refers to a typology of verbal predicates. A

verbal predicate may denote a more dynamic situation that involves some
kind of change or alternatively, a more static situation that requires an exter-
nal agent for change. Predicates of the former type are referred to as 'even-
tive predicates' or simply 'events' and predicates of the latter type as
'stative predicates' or 'states'. As many researchers have pointed out, the
meaning differences between event-describing and state-describing sen-
tences appear to be gradual rather than clear-cut in actual language use (see,
among various others, Comrie 1976; Mourelatos 1981; Smith 1991; Roth-
stein 2004). Nevertheless, it seems possible to distinguish between eventive
and stative predicates on the basis of their internal temporal structure. The
temporal schema for state sentences is represented in figure 1 below, which
formalises the intuition that states do not change for the time they last (e.g.
John loves Mary). They have an internal temporal structure and simply
consist of an interval in time. The initial and final endpoints (given in
34 Chris Η. Reintges

parentheses) are therefore not part of the internal temporal structure of the
state eventuality itself (Smith 1991: 32 (30); Rothstein 2004: 14-6). (I and
F represent the initial and final end points of the relevant eventuality.)

(I) (F)

Figure 1. The homogenous temporal structure of states

All non-stative situations constitute a natural class of events. While it re-

quires no special effort to remain in a state, eventive situations can only be
maintained if they are subject to 'a continuing input of energy' (Comrie
1976: 49). The occurrence of an event involves some condition when the
event begins, is terminated and replaced by another condition. The stage
property of events is particularly clear in accomplishment verbs like write
(e.g. Mary wrote the letter), which are [+dynamic], [+telic], and include in
their semantic description a culmination point. The culmination point
makes it possible to distinguish between successive stages in the develop-
ment of an accomplishment event, viz. a preparatory phase, which desig-
nates the event in progress and which reaches but does not include the cul-
mination point, and the resultant state, which designates what the outcome
of the event's termination is. The initial and final endpoints are an integral
part of the internal temporal structure of accomplishment events. See figure
2 below for further illustration (Smith 1991: 32; Kamp & Reyle 1993: 558
(5.104); Rothstein 2004: 21-2).

PREPARATORY CULMINATION RESULTANT

PHASE POINT STATE

(I) (F)

Figure 2. The tripartite structure of accomplishment events

In short, events are heterogeneous eventualities that necessarily involve

change and thus have internal temporal structure. By contrast, states have
no internal dynamism and consist of an undifferentiated period of time
without internal structure. In the rich philosophical and linguistic literature
on event ontology, more elaborate systems of verb classification have been
 Word order correlations in Older Egyptian 35

advocated, although it is generally acknowledged that the contrast between

motion and stasis, event and states, marks a core distinction in the aspectual
domain (see Bach 1986; Comrie 1976; Mourelatos 1981; Smith 1991; Kamp
& Reyle 1993; Rothstein 2004 for representative studies that go back to
Vendler's 1967 classic article). 3

2.2. The aspectual and thematic dimension of the Eventive-Stative alternation

This section examines the interaction between lexical semantics and aspectu-
al viewpoint in the derivation of verbal meaning. The following description
of the Eventive-Stative opposition is based on a simple taxonomy of verbal
classes, in which valency information is considered apart from argument
meanings. Following Grimshaw (1990) and related research, argument struc-
ture itself does not encode thematic roles like AGENT, PATIENT or THEME,
since it only represents the argument-licensing capacity of a predicate
without further specifying any semantic information about its arguments,
except for their relative prominence. The Eventive-Stative alternation in
Older Egyptian shows quite clearly that the eventuality of the verbal and
the thematic content of its arguments are not part of its lexical-semantic
frame, but are primarily determined by the morpho-syntax. The systematic
differences in meaning between Eventive-inflected and Stative-inflected
verb forms will be examined for different lexical classes of transitive-
active, unergative and unaccusative verbs.

2.2.1. Transitive-active verbs

As exemplified by the Eventive-Stative pairs in (2)-(3), transitive-active

Statives in Older Egyptian can take objects and complement clauses.
Stative formation can therefore not be equated with an argument structure
changing operation that eliminates the AGENT theta-role and creates an
unaccusative verb from a transitive base verb. 4

(2) DP OBJECTS
a. Eventive verb form
jw rx-n(-j) fik? nb ft? n(-j) gnw
AUX learn-PERF-lS EV magic every secret LINK(-SM) residence
Ί learned about every secret magic of the residence.'
(Urkunden I 143: 2)
36 Chris Η. Reintges

b. Stative verb form

jw ij-k(j) Bk? nb [CP ?χ n-f
AUX learn-1S S T at every secret be.glorious(-PTCP:SM) for-3SM
m grt-ncr ]
in necropolis
Ί know (by learning) every magic on behalf of which one becomes
glorious in the necropolis.' (Urkunden I 263: 14)

(3) CP COMPLEMENTS
a. Eventive verb form
j.mr-n(-j) [CP nd5-k jrt-k m-ΐ jr
wish-PERF-1 Sev save-2SM E v eye-2SM f r o m - a r m make(-PTCP:SM)
r-k ]
against-2SM
Ί have come to wish (that) you save your eye from the one who
acts against you.' (Ancient Hymn Da/12)
b. Stative verb form
j.mr-k(j) [cp nd'-k jrt-k m-ΐ jr
wish-lSsiAT save-2SM E V eye-2SM f r o m - a r m make(-PTCP:SM)
n-k ]
for-2SM
Ί have the wish to save your eye from the one who acts for you.'
(Coffin Texts VI 220j/L2Li)

The Eventive-Stative alternation is fully productive with verbs of creation

like ms(.i) 'to give birth' and verbs of putting into a spatial configuration
like qrs 'to bury', as seen in (4) and (5) below. According to Levin & Rap-
paport-Hovav (1995: 247-8), members of both classes qualify as accom-
plishment verbs, whose event structure comprises a process and a resultant
state component. Associating accomplishment verbs with a given aspectual
viewpoint highlights either the process or the resultative meaning of the
selected accomplishment verb. Thus, the Perfect viewpoint of the Eventive
clause in (4a) asserts the successful completion of the birth-giving event of
the deceased pharaoh, while the resultant state (the new-born state) is con-
textually implied. This contrasts with the Stative viewpoint in (4b), which
places the resultant state of some creational act into the center of attention,
without further specifying its internal development.
 Word order correlations in Older Egyptian 37

(4) VERBS OF CREATION

a. Eventive verb form
ms-n nww Mrjj-n(j)-ri Rr d'rt-f j?bt
give.birth-PERFEV ocean Meri-ni-Re on hand-3SM left
'The ocean has born (King) Meri-ni-Re on his left hand.'
(Pyramid Texts 1701a/M)
b. Stative verb form
j(w)r-kw fd(-w) ms-kw pdw
conceive-ISstat lower.sky give.birth-lSSTAT upper.sky
Ί have conceived the lower. I have born the upper sky.'
(Coffin Texts IV 51: e - f / B 3 L )

As noted by Nedjalkov & Jaxontov (1988: 24-25), transitive-active statives

may assume a possessive sense, with the subject being interpreted as the
possessor and the direct object as the possessed item. An appropriate se-
mantic paraphrase of the Stative VPs j(w)r-kw fd(-w) and ms-kw pdw in
( l i b ) above would therefore be something like Ί am the begetter of the
lower sky' and Ί am the mother of the upper sky', respectively. In a similar
vein, stativized epistemic and desiderative verbs, such as τχ 'to learn' and
mr(.i) 'to wish' convey a possessive interpretation insofar as the result of
some intellectual activity is the possession of some kind of knowledge or
the maintenance of some psychological state.
It generally appears, then, that the Eventive-Stative opposition has a
differential function, encoding alternations in the subject's relation to the
eventuality denoted by the verbal root. In transitive-active Eventive clauses,
the verbal action is construed as principally affecting the referent of a non-
subject argument (typically the direct object), while the subject is inter-
preted as the AGENT, i.e. the entity that performs some event or activity. In
the corresponding Stative, on the other hand, the subject is interpreted non-
agentively as the AFFECTEE, i.e. the entity on which the consequences of
some previous action has a positive or negative impact. To clarify this
point, consider the following Eventive-Stative pair, where the Perfect
qrs-n(-j) Ί have buried' in (5a) and the corresponding Stative qrs-k(j) Ί
had buried' in (5b) below describe what is objectively the same situation;
yet both variants differ with respect to the locus of the action's principal
effects. In (5b), the burial of the courtier is presented from the Stative point
of view, because it has a lasting effect on the first person singular subject in
the impressive rewards allotted to him by the king.
38 Chris Η. Reintges

(5) VERBS PUTTING INTO A SPATIAL CONFIGURATION

a. Eventive verb form
jw qrs-n(-j) jt(-j) pn m jz-f n(j) grt-ncr
AUX bun/-PERF-ls E v father-Is DEM:SM in tomb-3SM LINK necropolis
[n]-zp qrs-t(j) mj-t(j)-f [nb] m [mj-]t(j)-t-f
never bury-PASS2 EV like-NOM-3SM every with like-NOM-SF-3SM
Ί have buried this father of mine in his tomb of the necropolis.
Never was anyone of his rank buried with something alike.'
{Urkunden I 139: 1-2)
b. Stative verb form
qrs-k(j) ζ pn m jz-f m/?(/) N%b
bury-lSsTAT man DEM:SM in tomb-3SM north Nekheb
rd3-t(j) n(-j) ?fit sc?t 45 m t?-mß(j) fmTw (...)
give-PASS2EV to-lS field aroura 45 in Lower.Egypt Upper.Egypt
r fiz-t b?k jm
to praise-lNF servant there
Ί had this man buried in his tomb north of Nekheb (El-Kab). A field
of forty-five arouras was given to me in both Lower and Upper
Egypt (...) to reward this servant' (Urkunden I 140: 8-11)

If a transitive verb inflects in both inflectional paradigms, then the selection

of the Stative depends on whether the subject has a referent upon which the
principal effects of the action at hands devolve, while the Eventive is selec-
ted in the default case of non-affected subject-agents. In a sense, then, the
Eventive may be thought of as an 'effective' and the Stative an 'affective'
conjugation pattern. 5

2.2.2. Unergative verbs

When associated with the Eventive viewpoint, unergative verbs of bodily

expression like ßT(.i) 'to cheer' make direct reference to the physical ex-
pression of a particular emotion, as in (6a), while the concomitant mental
state is denoted by the Stative variant, as in (6b) below. Moreover, an
Eventive subject is conceptualised as a volitional AGENT that exerts control
over the bodily process that is described. Α Stative subject, on the other
hand, denotes the EXPERIENCER of the associated state of mind.
 Word order correlations in Older Egyptian 39

(6) UNERGATIVE VERBS OF BODILY EXPRESSION

a. Eventive verb form
βγ-w-f m xsf-k
cheer-PROS-3SM E V a b o u t meet(-INF)-2SM
'He will exult about meeting you.' (Pyramid Texts 656a/T)
b. Stative verb form
Nwt jM-U(j) m Xsf Nfr-k?-R Ϋ
Nut cheer-INTENS/REPET-3FSTAT about meet(-INF) Nefer-ka-Re
'(The goddess) Nüt is very exited about meeting (King) Nefer-ka-Re.'
(Pyramid Texts 1426a/N)

The exact classification of verbs like sit, stand and lie has raised some con-
troversy in the literature. Hoekstra & Mulder (1990) include such verbs of
spatial configuration in the class of (unaccusative) copular verbs, although
members of this class in Dutch appear in constructions displaying unergative
syntax. To make sense of the variable unergative-unaccusative behaviour,
Levin & Rappaport-Hovav (1995: 126-130) propose to distinguish between
three types of non-causative meanings associated with a specific spatial
configuration. The first two non-causative meanings are the 'maintain posi-
tion' and the 'assume position' sense, both of which are agentive. The third
meaning is non-agentive and describes the position of the subject with re-
spect to a particular location. In Older Egyptian verbs of spatial configura-
tion, the agentive 'maintain position' and the 'assume position' sense are
both encoded by the Eventive conjugation pattern, as shown in (7a), and the
non-agentive 'simple position' sense by the corresponding Stative, as
shown in (7b) below.

(7) VERBS OF SPATIAL CONFIGURATION

a. Eventive verb form (Agentive 'assume position' sense)
ßms Nfr-k?-RΫ jr rmn-k Fir
sitEv Nefer-ka-Re at shoulder-2SM Horus
'(King) Nefer-ka-Re will sit down besides you, Horus.'
(Pyramid Texts 2056a/N)
b. Stative verb form (Non-agentive 'simple position' sense)
j.fims-t(j) fir nst jt-k Gb m-χηί jtrt
sit-2ssTAT on throne father-2SM Geb in-front.of sanctuary
'You are seated on the throne of your father Geb in front of the
sanctuary.' (Pyramid Texts 1992b/N)
40 Chris Η. Reintges

The Eventive-Stative opposition also applies to verbs of inherently directed

motion like h?(.i) 'to descend' and j j 'to come'. These motion verbs specify
an achieved endpoint or attained location and may therefore be classified as
achievement verbs (Levin & Rappaport-Hovav 1995: 58; Talmy 1985: 72).
Their telic character is particularly clear in motion verbs like h?(.i) 'to de-
scend' that incorporate into their meaning a notion of path. A directional
phrase that indicates endpoint inherent to the verb's meaning is obligatorily
present in both the Eventive and the Stative variant.

(8) VERBS OF INHERENTLY DIRECTED MOTION [+PATH]

a. Eventive verb form

jj-n(-j) mjn m njwt(-j)
c o m e - P E R F - I S H V today from city-LS
h?-n-j m sp?t(-j)
descend-PERF-1SEV from district-Is
Ί have come today from my city, I have descended from my district'
(iUrkunden I 121: 11-12)
b. Stative verb form
jw h?-k(j) r ?bd3w ςτ Rs
AUX descend-lSsiAT to Abydos under Res
Ί descended to Abydos with Res.'
(Stele Metropolitan Museum NY no. 65.107: 4)

Verbs of inherently directed motion like j j 'to go' and jw 'to come' incor-
porate into their semantics a deictic orientation towards the speaker. The
location argument does not have to be overt, but can be semantically im-
plied, as seen in (9a-b) below.

(9) VERBS OF INHERENTLY DIRECTED MOTION [+DEITIC CENTER]

a. Eventive verb form

jj-n(-j) Trq(-w) m-<,'(-])
come-PERF-1S E V accomplish-PASSL E v through-arm-IS
Ί returned (after) (it) (the work) had been accomplished.'
{Urkunden I 220: 7)
b. Stative verb form
m-k w(j) bs-kj jj-kj
INTERJ-2SM me instal-lSsTAT come-lSSTAT
'Look, I am installed, I have arrived.'
(Pyramid Texts/Neith 831 [pi. 32])
 Word order correlations in Older Egyptian 41

Aspectual choice involves the up- or downgrading of a particular meaning

component in that class of verbs. The Perfect viewpoint asserts attainment
of a particular goal, but leaves the movement and path component of verbs
of inherently directed motion intact. This does not seem to be the case in
the corresponding Stative, where the motion and displacement sense has been
blurred. Rather, when used statively, verbs of inherently directed motion
assume a locative sense and describe the appearance of some discourse
participant on the scene related to or identified with the speaker (Levin &
Rappaport-Hovav 1995: 241-242).

2.2.3. Unaccusative verbs

When considered in isolation, the evidence presented in the previous sec-

tion might lead one to conclude that the Stative is simply the unaccusative
alternant of an unergative verb. On that view, the non-agentive reading of
Stative subjects would be readily explained. There are, however, two rea-
son for dismissing an unaccusative analysis of the Stative. Firstly, the oc-
currence of transitive-active Statives cannot be accounted for. Secondly,
bona fide unaccusative verbs fully participate in the Eventive-Stative alter-
nation, where it overtly marks the inchoative-stative contrast. As shown in
(10) and (11) below, the Eventive variant of adjectival verbs of quality,
size, and colour describes a change of state, while the corresponding Stative
indicates the target state is encoded by the Stative alternant (Levin &
Rappaport-Hovav 1995: 159-162; Doron 2003: 61-62).

(10) ADJECTIVAL VERBS

a. Eventive verb form
?X-n-f m ?χί,
be.glorious-PERF-3SM E v in horizon
d'd-n-f iv D'dwt
be.enduring-PERF-3SM EV in Djedüt
' H e has become glorious in the horizon, he has become enduring
in Djedüt (the necropolis of Heliopolis).' (Pyramid Texts 3 5 0 c / T )
42 Chris Η. Reintges

b. Stative verb form

?X-t(j) m ?χί,
be.glorious-2s S TAT in horizon
d'd-t(jj m D'dwt
be.enduring-2ssTAT in Djedüt
' Y o u are glorious in the horizon, you are enduring in Djedüt.'
(Pyramid Texts 1 2 6 1 b / N )

(11) a. Eventive verb form

wr-n-j m sf m-m wr-w
be.great-PERF-lSgv in yesterday among great.one-PM
Ί have become great among the great ones yesterday.'
(Coffin Texts II 2 6 8 e / B 9 C )
b. Stative verb form
w'l'd'-tfjJ wr-t(j) m m-k pw
be.green-2sSTAT be.great-2SsTAT in name-2SM DEM:SM
n(j) W?d3-wr
LINK(-SM) Green-Great
' Y o u are green and great in this your name Great-Green (i.e. the
Mediterranean Sea).' (Pyramid Texts 6 2 8 c / P )

The inchoative-stative alternation can also be found in various types of

copular verbs, such as verbs of appearance like χρτ 'to happen, emerge'
and verbs of existence like wnn 'to be'. Again, the Eventive form has an
inchoative sense, denoting the coming into existence of a particular entity,
while the corresponding Stative describes the existence of that entity at a
particular location. The location argument need not be overt, but may be
contextually implied (Levin & Rappaport-Hovav 1995: 1 2 0 - 1 2 1 ) .

(12) VERBS OF APPEARANCE

a. Eventive form
η m?-f Xpr-n-j m fir-f
NEG see-3SM EV exist-PERF-lS E v with face-3SM
' H e (the god N u n ) did not see with his own eyes (how) I came into
existence.' (Coffin Texts I 3 3 4 c / B 2 L )

b. Stative form
m?-n wj nw xpr-kw
see- perFev me Nun come.into.existence-lSSTAT
'(The god) N u n saw me (already) existing.'
(Coffin Texts I 3 3 4 a / B 2 L )
 Word order correlations in Older Egyptian 43

(13) VERBS OF EXISTENCE

a. Eventive form
wn Tjtj m wrwt-k
exist EV Teti in greatness-2SM
wn-n Tjtj m wrwt-k
exist-IMPERFEV Teti in greatness-2SM
'(King) Teti existed in your greatness. (King) Teti continues to
exist in your greatness.' (Pyramid Texts 719/T)
b. Stative form
wn-k(j) r-f m jw-f Ώ-f
exist-ISSTAT PCL-3SM in come-3SMEv be.great-3SMEv
Ί had, indeed, become one who gets bigger (whenever) he comes.'
(Stela British Museum 146: 4 [575])

Unlike verbs of existence and appearance, the verb corresponding to English

die does not require a location argument (either overt or covert) for its se-
mantic completeness. Older Egyptian mwt 'to die' behaves like a verb of
entity-specific change of state, with the Eventive-inchoative variant de-
scribing the process of dying and the Stative the resultant dead state.

(14) VERBS OF ENTITY-SPECIFIC CHANGE OF STATE

a. Eventive form
mr-w-f mwt-f, swt mwt-f
wish-PROS-3SM E V die-3SM E V he die-3SM E V
'If he wishes that he dies, he will die.' (Pyramid Texts 155d/W)
b. Stative form
j?m Nfib-k?-w m mtwt Hjw mwt-tj
burn EV Nekheb-kau in semen ///w-serpent die-3FSTAT
'(The god) Nekheb-kau burns the semen of the #/w-serpent dead.'
(Pyramid Texts/Neith 717 [pi. 27])

To conclude, the Eventive-Stative alternation displays a broad distribution

across various lexical classes of transitive, unergative and unaccusative
verbs. The semantic interpretation of verbal predicates is compositionally
derived from the basic lexical meaning of the root morpheme and the even-
tive or Stative viewpoint inherent to the selected inflectional paradigm. 6 The
possibility of having transitive-based statives in Older Egyptian is unex-
pected under any account that assigns to Stative-inflected verbs an unaccu-
44 Chris Η. Reintges

sative syntax. The selection of the Eventive or the Stative conjugation sets
not only a basic parameter for the eventuality described in a particular sen-
tence, but also defines the theta-role of the most prominent argument that is
realized as the surface subject.

3. Two types of agreement

The focus of this section is on the morphological and categorial properties

of person, gender and number endings on Eventive- and Stative-inflected
verb forms. In recent work on Semitic and Celtic linguistics, two different
approaches have been taken to what has been called the 'Affix Identifica-
tion Problem' (Fassi Fehri 1988). In one strand of research, concord-
marking personal inflections have been analysed in terms of grammatical
agreement (see McCloskey & Hale 1984 for Modern Irish; Stump 1984,
1989 for Breton; Borer 1981, 1995; and Shlonsky 1997 for Modern Hebrew;
and Benmamoun 2000 for Modern Standard Arabic). In another strand of
research, the concord-marking affixes on verbal stems have been identified
with incorporated subject pronouns that have been removed from their ar-
gument position and integrated into the verbal stem (see Anderson 1982,
Doron 1988 for Breton and Fassi Fehri 1988, 1993 for Modern Standard
Arabic). It is often difficult to find the relevant evidence to support either
the agreement or the pronoun incorporation analysis, since both grammati-
cal agreement and pronominal elements are characterised by the same set of
nominal functional features. Therefore, they cannot be distinguished with
respect to their syntactic category, but rather with respect to their role in
syntax (Ritter 1995). 7

3.1. The paradigm structure of the Eventive and the Stative conjugation

Older Egyptian has two exclusively suffixal conjugation patterns for finite
verb forms, which are called the Eventive and the Stative on the basis of
their primary grammatical meaning. The complete inflectional paradigms
of both verb conjugations are represented in table 1. (The triconsonantal
verb sd3m 'to hear' has been chosen to illustrate a typical paradigm).
 Word order correlations in Older Egyptian 45

Table 1. The paradigms of the Eventive and the Stative conjugation

EVENTIVE STATIVE

is 3 3
sd m(-j), sd m-j sd 3 m-k, sd 3 m-kj, sd 3 m-kw
2SM sd 3 m-k sd3m-tj, sd 3 m-c(j)
2SF sd 3 m-c, sd 3 m-cn sd3m-tj, sd 3 m-c(j)
3
3SM sd m-f sd 3 m-w, sd3m-jj,
3SF sd 3 m-s sd3m-tj, sd 3 m-c(j)
1P sd 3 m-n sd 3 m-wn, sd 3 m-wnj, sd 3 m-wjn
3
2P sd m-cn sd 3 m-tjwn, sd 3 m-tjwnj
3 PM sd 3 m-sn sd 3 m-w, sd3m-jj
3PF sd 3 m-sn sd3m-tj
2D sd 3 m-cnj, sd 3 m-cny sd 3 m-tjwnj
3
3 DM sd m-sny sd 3 m-wjj, sd 3 m-wj
3DF sd 3 m-sny sd3m-tjj, sd 3 m-tj

DPmasc. sd3m-jj, sd 3 m-w,

D P SUBJECT sd3m DP
DPf e m sd'm-tj, sd 3 m-c(j)

As we can see from the left-hand column of table 1, Eventive-inflected

verbs may appear in two distinct forms: a synthetic form containing a con-
cord-marking personal affix and a 'bare' analytic form. The selection of
either form is dependent on the (pro)nominal status of subject: synthetic
forms can only appear in the absence of a subject DP. By contrast, no such
alternations apply to the Stative conjugation in the right-hand column of
table 1, where the same form occurs with nominal or pronominal subjects.
On the other hand, there is an exponent of every person, gender and number
combination in the Eventive paradigm, while two or more paradigmatic
cells share one exponent in the corresponding Stative paradigm. For in-
stance, the third person masculine inflection -w and its variant -jj are not
specified for grammatical number, while the second person dual and plural
marker -tjwn(.y) is not differentiated with respect to grammatical gender.
The inflectional ending -tj is even more radically underspecified, realising
second person singular as well as third person singular and plural feminine.
On the face of it, the paradigmatic split of the verbal inflectional system
looks like another instance of the familiar asymmetry between 'rich' and
'poor' agreement as found in Modern Standard Arabic (see, among various
others, Mohammad 1990; Fassi Fehri 1993; Ouhalla 1994; Aoun, Benmamoun
46 Chris Η. Reintges

& Sportiche 1994; and Benmamoun 2000). I will argue, however, that we
are dealing with an agreement asymmetry of a rather different kind. The
personal inflections on Eventive verb forms do not instantiate subject-verb
agreement at all, but represent enclitic subject pronouns that correspond to
an argument position. By contrast, grammatical agreement proper is only
instantiated by the featurally less coherent Stative paradigm.

3.2. An agreement analysis of the Stative paradigm

This section argues against the commonly held view in Egyptological lin-
guistics that Stative inflections represent incorporated subject pronouns (see
Allen 1984: 6 §11 and 384 §564 and Schenkel 1997: 199 for representative
views). The first argument concerns the role of locality in distinguishing
between pronoun incorporation and grammatical agreement. The second
argument relates to the limited distribution of pro-drop.

3.2.1. Locality

Older Egyptian has 'Exceptional Case Marking' ( E C M ) constructions with

finite complement clauses. The E C M complement is compatible with both
Eventive and Stative-inflected verb forms. T w o examples of Stative E C M
constructions are shown in ( 1 5 a - b ) below. That the embedded subject is,
indeed, exceptionally accusative-case marked by the matrix V P is disclosed
when the lexical D P is replaced by the corresponding pronoun, since the
direct object pronoun siv ' h i m ' rather than the corresponding nominative
subject pronoun - / ' h e ' is selected.

(15) ECM CONSTRUCTIONS WITH EMBEDDED STATIVES

a. E C M subject = D P
gm-n(-j) [agssp fiq? F?m fm(-w) r-f r t? Cmfi ]
find-PERF-lS E v ruler Yam leave-3MSTAT PCL-3SM to land Libyan
Ί found (that) the ruler of Yam had departed to the land of the
Libyan.' {Urkunden I 125: 15-16)

b. E C M subject = direct object pronoun

gm-n-f [ A G S S P SW d-jj fir gs-f m Gßst ]
find- PERF-3SM EV him p l a c e - 3 M s T A T on side-3SM in Gazelle-ville
' H e (Geb) found him (Osiris) placed on his side in Gazelle-ville.'
(Pyramid Texts 1033b/P)
 Word order correlations in Older Egyptian 47

In view of the fact that the ECM complement does not exceed the domain
of the subject agreement phrase AGRSP, it is hard to see how the co-
occurrence of ECM subjects and Stative verb forms could be reconciled
with the traditional pronoun incorporation analysis of Stative inflections,
given that there would be two subjects for which only a single theta-role
and structural case would be available. If we were to assume that these
personal inflections are incorporated pronouns, we would expect them to
conform to the general principles of the Binding Theory of Chomsky
(1981). In other words, they should be free in their governing category. In
the above examples, the purported subject pronoun is bound within its gov-
erning category (i.e. the subject agreement phrase AGRSP) by either a refer-
ring expression or another pronoun. The grammaticality of (15a-b) above
vis-0-vis the Binding Theory is expected under an agreement analysis of the
Stative paradigm. On this analysis, Stative inflections redundantly express
the person and gender features of the preverbal subject, but do not occupy a
separate structural position (Bresnan & Mchombo 1987: 752ff). As pointed
out by the reviewer, a clitic-doubling account might save the traditional
pronoun incorporation analysis. Such a clitic-doubling analysis would,
however, be at odds with the limited distribution of null subjects, as we will
see next.

3.2.2. Partial pro-drop

Although locality considerations argue against a pronoun incorporation

analysis of the Older Egyptian Stative, the intuition behind this analysis
seems to be correct, namely that Stative endings have anaphoric properties.
Compare example (16a) below, in which the first person singular Statives
pr-kj Ί have come forth' and wSb-kj Ί am pure' appear without a subject
pronoun with example (16b) below, in which the third person masculine
Stative jj(-w) 'has arrived' is construed with the third person singular mas-
culine pronoun -f 'he' to the right of the complementizer writ 'that'.

(16) ANAPHORIC AGREEMENT

a. First person singular Statives
pr-kj r-j wib-kj
come.forth-lSsTAT PCL-lS be.pure-lS S TAT
Ί have come forth pure.' (Coffin Texts VI 136o/M22C)
48 Chris Η. Reintges

b. Third person masculine Statives

d'd-cn χι- R? [CP wnt-f jj(-w) m ncr ]
say-3PEV to Re C0MP-3SM come-3M S TAT as god
'You should announce to (the sun-god) Re that he (the deceased
king) has arrived as a god.' (Pyramid Texts 1862a/N)

In the longstanding research tradition on the null subject parameter, the

licensing of covert subject pronouns (small pro's) has been related to the
amount of featural information encoded by verbal inflection. That is,
grammatical agreement has to be specified beyond a certain threshold to
recover the referential content of the null subject (see Borer 1981, 1986;
Rizzi 1982, 1986; Huang 1984; Jaeggli & Safir 1989; and much subsequent
research). The availability of pro-drop in the various paradigmatic cells of
the Stative conjugation is exemplified in (17) below.

(1 7) THE PRO-DROP PARADIGM OF THE STATIVE

IS a. pro jj-k(j) m fitp r fmΫ

come-lSsTAT in peace to Upper Egypt
Ί have arrived in Upper Egypt in peace.'
(stela Leiden V88: 10 [pi. 10])
b. m-k w(j) jj-k(j)
INTERJ-2SM me come-lS S TAT
'Look, I have arrived.' (Coffin Texts V 7 8 a / T l C )
2S c. pro m-tj χίί RT pr-f m j?bt
stand-2SsTAT before Re come-3SM EV from east
' Y o u are standing in front of Re (when) he comes from the East.'
(Pyramid Texts 7 4 3 b / T )
d. cwt fir-t(j) r-f
you:SM stand.up-2sSTAT be.far away-2sSTAT from-3SM
' Y o u are standing far away from him.' (Pyramid Texts 2 5 1 c / W )

3SM e. m sw j-jj
INTERJ him come-3M S T AT
'Look, he (the deceased king) has arrived.'
(Pyramid Texts 1495a/P)
3SF f. wnt-s sr-tj n-k r-s
COMP-3SF foretell-3F S T AT to-2SM about-3SF
'(To inform you) that she (the deceased female) has foretold about
herself to you.' (Coffin Texts I 140g/B3Bo)
 Word order correlations in Older Egyptian 49

IP g. pro s-d5f?-wn
CAUS-provide-1 PSTAT

' W e are provided (with food).' (Bissing, Gem-ni-kai I, pi. 18, 2 nd reg.)

h. jw-n s?-wn
AUX-lPEv satisfy-1 P S T A T
' W e are satisfied.' (stele Mehu-akhti (Reisner G 2375),
MIO 1 (1953), p. 328 B: 2, fig. 1)
2P/D i. pro ßr-tjwn r b?-j pn
be.far.away-2D/P S TAT from soul-lS DEM:SM
'You should be far away from this soul of mine.'
(Coffin Texts VI 7 6 c / B 3 L )
j. m-χί cny gm-tjwny
COMP you:D find-2P/D S T AT
'after the two of you (i.e. the two chopped-off hands of the god
Horus) have been found.' (Coffin Texts II 350a/B4L b )
3PM k. sk sn Γχ-jj s(j) η χτη-sn s(j)
COMP they understand-3MsTAT it NEG ignore-3P EV it
'While they (the demons) know it (the eye of Horus), they are not
unaware of it.' (Coffin Texts VII 111 j / SQ1OC)

The above data illustrate the limited distribution of pro-drop in the Stative
paradigm. Null subjects are licensed in the context of first and second person
singular and plural reference, where the inflectional endings are unambigu-
ously specified for [person] and [number], but excluded from third person
contexts where agreement lacks an explicit morphological representation of
the features [number] and, in the feminine, [person]. Furthermore, note that
the availability of pro-drop serves as a means for disambiguating the refer-
ence of the polysemous ending -tj, informally stated in (18).

(18) CONSTRUAL RULE FOR THE POLYSEMOUS STATIVE ENDING - t j

The Stative inflection -tj has second person singular reference in null
subject clauses.

A similar paradigmatic split between first and second person on the one
hand and third person on the other has been observed for verbal tenses in
Modern Hebrew (Borer 1986; Ritter 1995; Shlonsky 1997). If the inflec-
tional endings on Stative verb forms represented incorporated pronouns, it
would be a complete mystery why clitic-doubling is obligatory in third
50 Chris Η. Reintges

person contexts, while it is available in first and second person contexts

only when the subject pronoun is contrastively focused. Moreover, the pic-
ture is complicated by the fact that third person singular masculine Statives
can occur without an overt third person pronoun in impersonal sentences,
such as (19a-b) below. In line with Rizzi (1986), Borer (1986), Shlonsky
(1990), and Franks (1990), I assume that the preverbal subject position of
impersonal Stative sentences is occupied by a phonologically null expletive
pronoun (indicated as proexp\).

(19) IMPERSONAL STATIVE SENTENCES

a. jw proexpi d5w-w
AUX be.bad-3MsTAT
'(It) has turned out bad.'
(Lepsius, Denkmäler II [pi. 63, 4th register])
b. jw proexpι sfp(-w) d3d NN pn
AUX be.bright-3MSTAT speak EV NN DEM:SM
jw proexp/ knß(-w) d3d NN pn
AUX be.dark-3MSTAT speak EV NN DEM:SM
'(It) dawns (when) this NN (the deceased male) speaks, (it) gets
dark (when) this NN speaks.' (Coffin Texts IV 29e/Sq6C)

According to Chomsky (1996: 342, 364), expletive subject pronouns in the

specifier position of AGRSP have only the categorial feature [D], Due to the
lack of referential content, the null expletive in the preverbal subject posi-
tion of impersonal Statives need not be identified by 'rich' agreement in-
flection, yet triggers the spell-out of third person masculine agreement by
default. Since it has an effect on the PF output, the pleonastic null pronoun
should be included in the numeration (Chomsky 1996: 294).

3.2.3. The complementary distribution between aspect and agreement

morphology

The Stative represents a morphologically restricted verbal inflectional para-

digm not only in the sense of the limited distribution of pro-drop, but also
because the presence of grammatical agreement blocks the selection of in-
dependent tense and aspect morphology. In Older Egyptian, the perfective-
imperfective opposition is lexicalized by means of different stem classes
(Reintges 1996). While there is no specialized morphology to mark perfec-
 Word order correlations in Older Egyptian 51

tive or neutral aspect, imperfective aspect is marked by reduplicating the

final root consonant, e.g. fiz 'to praise' (perfective, neutral) vs. fiz-z (imper-
fective) 'continue to praise, praise a lot'. 8
As already noted by Edel (1959), imperfective verb stems do not occur
in the Stative. This distributional gap cannot be related to the reduplicative
expression type of imperfective aspect, since Statives can freely be formed
from intensive-repetitive stems that express verbal plurality, as shown in
(20a). Moreover, Stative VPs can be modified by high degree adverbs like
T? wrt 'very much', as in (20b), and iterative temporal adverbs like rf nb
'every day', as in (20c) below. Both adverbial phrases express quantifi-
cation over states (Parsons 1990: 199-200). 9

(20) QUANTIFIED STATIVES

a. With intensive-repetitive stems
tm fm(-w) j.s-xd-xd(-w)
NEG:AUX walk(-GER:SM) CAUS-turn.down-INTENS/REPET-3M S T AT
jn b? m grt-ncr
FM soul in necropolis
'Not to walk turned upside down in the necropolis by a soul.'
(Coffin Texts I l a / T I L )
b. With high degree adverbs
jn-k(j) jm r i?f wrt
b r i n g - I S s t a t there PREP great much
Ί imported therefrom (the Wadi Hudi) very much.'
(Wadi Hudi no. 14: 10)
c. With iterative adverbs
psd3-kj m ΗΫ ή' nb
shine-lSsTAT as Re day every
Ί shine as (the sun-god) Re every day.' (Coffin Texts II 149a/G2T)

Reintges (1997: 98-100) proposes to derive the complementary distribution

between grammatical agreement and independent tense and aspect mor-
phology from semantic restrictions on Stative formation. In asserting that a
given eventuality has not reached its natural endpoint, both the imperfective
and the stative viewpoint convey an unbound or cumulative reading, yet
differ systematically from each other in that imperfectives, but not statives,
imply a certain degree of dynamism and change over time. As a result, the
combination of imperfective morphology with Stative agreement inflection
would yield an aspectually uninterpretable form. By the same token, the
52 Chris Η. Reintges

principle of full interpretation would rule out a combination of the Perfect

and the Stative, since the former has a bound event reading, asserting that
some eventuality has reached a closure, whereas the latter lacks a clearly de-
fined endpoint as part of its internal temporal structure (see above, section
2.1). To recapitulate, the unavailability of imperfective and Perfect/Anterior
aspect with statives is due to a general restriction on states, which lack an
internal temporal structure that could be referred to by means of verbal
aspect. This clearly shows that the inflectional endings on Stative verb
forms instantiate a portmanteau morpheme in which aspect and agreement
are fused.

3.3. A pronominal argument analysis of the Eventive paradigm

This section presents a pronominal argument analysis of the concord-

marking personal inflections on Eventive verb forms. On this account, the
Eventive paradigm does not manifest subject-verb agreement proper, but is
rather an instance of pronominal enclisis. The pronominal argument analy-
sis offers not only a straightforward account for the 'Complementarity
Principle', i.e. the incompatibility of synthetic verb forms with lexical DPs,
but can accommodate the apparently idiosyncratic distribution of inflected
and uninfected forms in a range of impersonal sentences, co-ordinate struc-
tures and relative clause constructions without stipulating construction-
specific agreement rules.

3.3.1. The Complementarity Principle

Eventive verb forms come in two varieties: synthetic forms with concord-
marking person inflections and analytic forms without such inflections. The
distribution of both forms is strictly regulated: finite clauses with postver-
bal DP subjects require the analytic form, as (21a) below illustrates. The
corresponding synthetic form is selected in the context of pronominal sub-
jects, as (21b) exemplifies. Although analytic forms lack discrete agree-
ment markings, they can still be inflected for tense and aspect, witness the
presence of the Perfect marker n-.
 Word order correlations in Older Egyptian 53

(21) THE DP-AFFIX COMPLEMENTARITY

a. Analytic form (lexical DP)
jj-n m/Ϋ pn m fitp
come-PERF E v army DEM:SM in peace
'This army has returned in peace.' ( U r k u n d e n I 103: 7)
b. Synthetic form (pronominal reference)
b?-n -f t? ßr-y-w fi'w
destroy-PERF-3SM EV land upon-NOM-PM sand
'It (the army) has destroyed the land of the sand-dwellers (i.e. the
Bedouins).' ( U r k u n d e n I 103: 8)

As we can see from table 2, the Complementarity Principle also applies to

non-verbal categories, such as nominal constructs and basic prepositions.

Table 2. The cross-categorial applicability of the Complementarity Principle

ANALYTIC FORM SYNTHETIC FORM

EVENTIVE VERB FORMS sd3m DP DP hears sd'tn-f he hears

BASIC PREPOSITIONS η DP for DP n-f for him

NOMINAL POSSESSION pr DP the house (of) DP pr-f his house

Similar distributional patterns have been observed in the Celtic V S O lan-

guages (Welsh, Breton, Modern Irish), where the person inflections on
verbs, nouns and prepositions have been identified with morphologically
'rich' agreement. To account for the Complementarity Principle, a special
inflectional rule is devised, according to which the governee of the agree-
ment-marking element has to be drawn from the set of empty categories
(see Stump 1984, 1989 for Breton; but see Doron 1988 for an opposite
view). 1 0 According to Baker & Hale (1990), these distributional facts can
readily be explained by the categorial status of concord-marking personal
inflections: as pronominal arguments, such inflections compete with full
DPs for the same structural slots in the syntactic representation: the selec-
tion of one category will therefore automatically exclude the selection of
the other.
54 Chris Η. Reintges

3.3.2. Impersonal sentences

The Complementarity Principle does not cover the entire spectrum of

agreement p h e n o m e n a in Eventive V S O clauses. A family of constructions
that falls outside of the scope of this descriptive generalization comprises
impersonal sentences of the following kind:

(22) IMPERSONAL SENTENCES

a. Impersonal unaccusatives with dative-beneficiaries

nfr-n proSXJ)ι η Wnjs ΗηΫ k?-f
be.good-PERFEv for Unas with soul-3SM
'(It) goes well (lit. it has become good) with (King) Unas and his
soul.' (Pyramid Texts 3 3 8 a / W )

b. Impersonal passives with dative-beneficiaries

ßw-t(j) proexpl n-k jr-f m jfst
beat-PASS2 EV for-2SM PCL-3SM with what
' H o w could (grain) be threshed for you?' (Coffin Texts III 1 3 7 j / S l C )

In impersonal unaccusatives and impersonal passives with dative-bene-

ficiaries, the finite verb assumes an analytic form, even though there is no
DP subject. There is reason to assume that the postverbal subject position is
occupied by a null expletive pronoun (proexp\). Older Egyptian has a modal
particle jr-f (short form r-f) 'indeed, really, actually', which is homo-
phonous with the directional preposition jr-f 'towards him, as for him'.
Akin to Hebrew reflexive datives, jr-f highlights the autonomy of the event,
i.e. its occurrence without the intervention of an outside agent (Berman
1981; Borer and Grodzinsky 1986), but it is also associated with corrobora-
tive focus, stressing the truth-value of a given proposition. What interests
us here is the syntactic behaviour of this particle, whose co-pronoun repro-
duces the person, number and gender specification of the subject constitu-
ent. In this respect, it behaves like a reflexive clitic. When the subject is a
null expletive pronoun, the reflexive clitic assumes third person singular
masculine value jr-f by default, as in the impersonal passive construction in
(22b) above. This clearly shows that the null expletive pronoun of imper-
sonal clauses is syntactically active: it can act as a controller of reflexive
clitics or trigger third person masculine inflection on impersonally used
Statives. The reason why null expletives are left unidentified by verbal
inflection in impersonal Eventive clauses is simply that there is no subject-
verb agreement to begin with.
 Word order correlations in Older Egyptian 55

3.3.3. Subject ellipsis

In Older Egyptian, there is a kind of ellipsis process that allows the subject
DP of asyndetically coordinated clauses to be deleted under identity with
the subject of the preceding clause (McCloskey 1991). The main point to be
observed here is that the Eventive verb appears in its analytic form, i.e.
without a personal affix.

(23) SUBJECT ELLIPSIS

s-Ϋηχ-η cw ßr m m-k pw n(j)

CAUS-live-PERF E V you:SM Horus in name-2sm D E M : S M L I N K ( - S M )
?nd3tj rcf-n n-k ßr jrt-f rwd3-t
Anejti give- P E R F e v to-2 S M Horus eye-3 S M s t r o n g - P T C P : S F
d-n n-k s(j) jm-jm-k
place-PERF E V tO-2SM it.SF be.strong-INTENS/REPET-2SM EV
'(The god) Horus has nourished you in this your name of (the god)
Anejtj. Horus has given to you his strong eye. <He> placed it to you
that you become very strong.' (Pyramid Texts 614: a - c / T )

All that needs to be said under the pronoun incorporation analysis is that
the subject position of the third conjunct contains a non-pronominal empty
category or gap (indicated as ' '), which is co-indexed with the preceding
DP subject (Huang 1984). Under the competing agreement analysis, one
has to stipulate that there is no agreement with vv/7-subject traces in co-
ordinate structures (see Borsley & Stephens 1989 for an analysis along
these lines for Breton subject ellipsis). On this view, one would expect
analytic forms to be consistently used in subject extraction contexts. This
prediction is not borne out by the empirical evidence, as the distribution of
analytic forms in relative clauses shows.

3.2.4. Subject relatives

In Older Egyptian finite relative clauses are introduced by a lexical com-

plementizer that agrees in number and gender with the relative head. As
shown by the contrast between (24a) and (24b) below, relative extraction of
the subject DP leaves a gap in the embedded subject position, while a re-
sumptive subject pronoun is introduced in the Eventive counterpart. On an
agreement analysis, the synthetic verb form Γχ-sn 'they know' would identify
56 Chris Η. Reintges

the referential content of a w/z-subject trace. This raises the question why the
subject gap of co-ordinate structures is never governed by an agreeing verb
form. Under the pronoun incorporation analysis, the personal affix -sn of the
embedded Eventive verb form would be a resumptive pronoun.

(24) FINITE SUBJECT RELATIVES

a. Subject gaps in relativised Stative clauses

χΐΏ-Wi [Cp jwt-w, _ j r/(-w), sw? ßr-s ]
demolisher-PM C O M P : N E G R E L - P M know-3M S T AT pass.by(-LNF) at-3SF
'the demolishers who do not know (how) to pass it (the first gate)'
(Coffin Texts VII 4 3 6 M 3 7 a / B 9 C )
b. Resumptive subject pronoun in relativised Eventive clauses
χτη-wι [cp jwt-w, r/sn, sw? ßr-s ]
demolisher-PM COMP:NRGR[.,-PM know-3PEV pass.by(-INF) at-3SF
'the demolishers who do not understand (how) to pass it'
(Coffin Texts VII 4 3 6 M 3 7 a / B 3 C )

Reintges (2000) offers a locality explanation for the distribution of gaps and
resumptive pronouns. Since the embedded subject position is adjacent to
the relative complementizer, the wh-trace is properly head-governed by an
agreeing C O M P . That locality is, indeed, the relevant factor is evident in
Eventive subject relatives, where the initial verb intervenes between C° and
the subject position, thereby blocking government from C°. To rescue the
derivation, the gap has to be replaced by a resumptive pronominal. As we
can see from the contrast between (25a-b) and (25c-d) below, gaps and re-
sumptive pronouns may vary in direct object relatives.

(25) FINITE DIRECT OBJECT RELATIVES

a. Direct object gaps in relativised Stative clauses

rmc-wΙ [CP nt(j) Mrjj-nj-rt r/-j _, ]
man-PM C O M P R E L ( - S M ) Meri-ni-Re know-3M ST AT
'the people whom (King) Meri-ni-Re knows'
(Pyramid Texts 1223b/M)
b. Direct object gaps in relativised Eventive clauses
ncn pw [ C P nt(j) η iy-k ]
god DEM: SM COMPREL(-SM) NEG know-2SMEV
'this god whom you do not know' (Coffin Texts V 11 l d / T I C )
 Word order correlations in Older Egyptian 57

c. Resumptive direct object pronouns in relativised Stative clauses

ncn pw [Cp nt(j)-k i%-t(j) .s'w, ]
god DEM:SM C0MPREL(-SM)-2SM know-2SSTAT him
'this god w h o m you k n o w ' (Coffin Texts V 111 d / M 2 C )

d. Resumptive direct object pronouns in relativised Eventive clauses

jr-w-cnj [Cpjwt(j) IJ st, xpr-w jm-y-cn ]
contour-PM-2p COMP:NEG REL (-SM) k n o w E V t h e m being-PM in-NOM-2P
'your contours that the beings inside you do not k n o w '
(Coffin Texts IV 61: b - c / L 2 L i )

Relative extraction from the derived subject position of a passive relative

clause patterns alike with relative extraction from the direct object position
of the corresponding transitive active relative in that either a gap or a re-
sumptive pronoun may appear at the extraction site. This distributional
pattern can be directed related to the fact that the derived subject of a pas-
sive relative clause is an underlying object and hence both gaps and re-
sumptive pronouns are licensed. The possibility of having either gaps or
resumptive pronouns in the derived subject position of passive relative
clauses is illustrated in (26a-b). Both examples are free relative clauses
with a covert relative head (indicated as [Dp 0 ] ) , whose gender and number
features are morphologically marked on the agreeing relative complemen-
tizer. (t marks the 'trace' left behind by the object-to-subject movement in
the course of passivisation).

(26) a. sm-n(-j) [[ DP 0 ] , [ CP ntyx η sm-t(j)-fx i,]]

succour-perf-1 s EV COMPREL(-SM) NEG succour-PASS2-3 SMEV
Ί succoured who had not been succoured.' (stela Leiden V4: 8 - 9 )

b. jw grt jr-n(-j) [[DP 0], [nt-t, η jr-t(j) t,

AUX PCL do-PF-lSnv C O M P ^ - S F N E G do-PASS2 EV
jn k(j)-w rmc-w [CP j-w gr fi?t-j ]]]
FM other-PM man-PM come-PTCP:PM u n d e r front-Is
' B u t I achieved what had not been achieved by any other man who
had come before m e . ' (Hatnub Inscription 22: 17-18)

It is hard to see how the availability of analytic and synthetic verb forms in
classical w/z-constructions can be reconciled with an agreement analysis of
the personal inflections on Eventive verb forms. On an agreement account,
one would have to stipulate a variety of construction-specific agreement
58 Chris Η. Reintges

rules to accommodate the absence of agreement inflection in the context of

w/i-subject traces. N o such problem arises under the competing pronominal
argument analysis where the co-occurrence of analytic and synthetic forms
can be captured in a principled way in terms of licensing conditions on gaps
and resumptive pronouns.

3.3.5. COMP-cliticisation

A final piece of evidence for the pronoun incorporation analysis comes

from examples like ( 2 7 a - b ) in which the pronominal suffix appears inde-
pendent of the finite verb to the right of a lexical complementizer. As noted
by Shlonsky (1997: 185-187), such complementizer cliticisation processes
are not available in languages like Berber, where grammatical agreement is
not a clitic.

(27) COMP-CLITICISATION
a. Finite subordinate clauses
r rd3-t ij-t(j) [Cp wnt-k h?-t(j% m Γηρ
t o give-INF understand-PASS2 COMP-2SM descend-2s S T AT in p e a c e
m Y?m ]
from Y a m
'to let it be known that you descended in peace from (the land) Y a m '
(Urkunden I 128: 7 - 8 )
b. Finite relative clauses
ncn pw [Cp nt(j)-k rx-t(j) siVj ]
god DEM:SM COMP REL (-SM)-2SM know-2sSTAT him
'this god whom you know'
(Coffin Texts V 11 l d / M 2 C ) (cf. ex. (25c))

Despite the surface similarities with complementizer agreement in various

types of subordinate clauses in West-Flemish and Southern Dutch dialects,
we seem to be dealing with a different type of phenomenon. If the subordi-
nating complementizer were, indeed, endowed with grammatical agree-
ment, there would be an unexpected mismatch between the alleged Even-
tive inflection on clause-initial complementizer and the Stative inflection
on the embedded finite verb (see Carstens 2003 and van Craenenbroeck &
van Koppen 2002 for relevant discussion). If, on the other hand, the second
person singular masculine suffix -k ' y o u ' is an enclitic subject pronoun
 Word order correlations in Older Egyptian 59

plain and simple, its appearance in the preverbal subject position of an em-
bedded Stative clause can be accounted for in a natural and unconstrained
way. W e return to the syntax of pronominal enclisis in section 4.2.2.3.

4. Deriving the VSO-SVO word order contrast

This section addresses the question how the verb-initial and subject-initial
word orders of Eventive and Stative clauses are syntactically derived from
the interaction of verb raising and subject positioning. In Reintges' (1997:
2 8 7 - 2 9 8 ) analysis, the finite verb in Stative SVO clauses raises to the highest
inflectional position (but no further). By contrast, the finite verb in Eventive
V S O clauses always raises out of the inflectional domain, targeting the com-
plementizer position or a lower left-peripheral functional head. A crucial as-
sumption of this approach was that Eventive and Stative subjects appear in
the same structural slot, to wit, the Specifier position of the IP. The analysis
just outlined gives a correct description of the word order and agreement
facts of Stative S V O order, but does not present a complete picture of the
flexible syntax of Eventive V S O clauses, in particular, the availability of
more than one hierarchical position for the postverbal subject.

4.1. Theoretical background

This section outlines the main assumptions on VP-internal structure and

verb-initial order that underlie the following configurational analysis of the
Older Egyptian V S O - S V O word order contrast.

4.1.1. Assumptions about VP-internal structure

My point of departure is the view that V S O and S V O orders are both de-
rived f r o m the same V P shell, a plausible assumption given the broad lexi-
cal distribution of Eventive and Stative verb forms. The shared VP-internal
structure encodes valency and argument hierarchy information, but is void
of semantic information about event types and argument meaning. This
must be so, since eventuality and theta roles are not part of the lexical-
semantic template of the root itself, but are specified at the next deriva-
tional level or phase, in C h o m s k y ' s (2000, 2001) system. 1 1 I adopt the main
60 Chris Η. Reintges

assumptions about VP shells found in important work by Larson (1988),

Hale & Keyser (1993, 2001), Kratzer (1994), Chomsky (1996), Marantz
(1997), and Doron (2003). In particular, I assume that the main arguments
of the verbs are licensed by distinct verbal heads. The external argument
that surfaces as the clausal subject is introduced by a light verb ν that takes
VP complement. The internal argument that surfaces as the direct object in
transitive-active clauses is located in the specifier position of the VP com-
plement. Double object constructions involve yet another verbal layer.
Consider the Eventive-Stative pair in (28a-b) with the vP-internal structure
shown in (29).

(28) TRANSITIVE-ACTIVE CLAUSES

a. Eventive VSO clauses
jw rx-n(-j) fik? nb ft? n(-j) gnw
AUX learn-PERF-lS E v magic every secret LINK(-SM) residence
Ί learned about every secret magic of the residence.'
0Urkunden I 143: 2)
b. Stative SVO clauses
jw τχ-kQ) iik? nb [CP?Z n-f
AUX l e a r n - l S s T A T every secret be.glorious(-PTCP:SM) for-3SM
m grt-ncr ]
in necropolis
Ί know (by learning) every magic on behalf of which one be-
comes glorious in the necropolis.' (Urkunden I 263: 14)
(cf. exx. (2a-b))

The category-neutral root appears as a complement of the lowest verbal

head. It takes on its specific verbal meaning in the context of the first verbal
head that merges with the ROOTP (see Marantz 1997, 2000).

(29)

Spec
DO
fik? nb V ROOTP
V/r-x/
 Word order correlations in Older Egyptian 61

The skeletal structure of VPs shown in (29) is incompatible with the claims
of the so-called Split-VP Hypothesis, according to which one or more func-
tional functions, for instance, the Object Agreement Phrase AGROP, may
intervene between the subject and direct object licensing verbal heads ν and
V, respectively (see Harley 1995 and the references cited therein). The
departure of relatively well-established principles of clausal organisation
(in particular, the bifurcation between a lexical and functional layer) is,
however, contradicted by the empirical evidence from Older Egyptian ob-
ject shift, where the DP subject occupies a position below the target posi-
tion of the shifted object pronoun (see McCloskey 1997: 221-225, 2001:
179-189 for counterevidence from Irish object shift and dative subjects).

4.1.2. Previous analyses of verb-initial order

In generative analyses that derive VSO order via verb fronting, it is gener-
ally agreed upon that the subject occupies the specifier position below the
functional head in which the finite verb appears. 12 In a left-headed lan-
guage, this will always mean that the subject will appear to the right of the
finite verb. Within this broad consensus, the exact position of the subject
and the scope of verb movement are topics of controversy. One school of
thought represented by Emonds (1985), Sproat (1985), Stowell (1989) and
Aoun, Benmamoun & Sportiche (1994) holds that V S O order is derived
from SVO order through verb movement and adjunction to COMP in a
manner reminiscent of den Besten's (1983) influential analysis of the Verb
Second phenomenon in West-Germanic. Under the V-to-C analysis, the
subject occupies the inflectional subject position [Spec, IP], See the dia-
gram in (30) below for further illustration.

(30) VSO ORDER AS THE RESULT OF V°-TO-C° MOVEMENT

I I I
[cpC° [IP S U [R 1° [ V P . . . v ° ... ]]]]

Another school of thought represented by Chung & McCloskey (1987),

Koopman & Sportiche (1991), Mohammad (1990), McCloskey (1991,
1996, 2001), Ouhalla (1994), and Benmamoun (2000) argues that the finite
verb is not in COMP at all, but rather appears in the highest functional head
of the inflectional domain. 1 3 Under this approach, too, V S O is derived from
62 Chris Η. Reintges

an underlying SVO order, but the subject and the direct object are both
case-licensed in their base-position in a configuration of government by the
finite verb, as schematically represented in diagram (31) below.

( 3 1) VSO ORDER AS THE RESULT OF V°-TO-I° MOVEMENT

I I
[r 1° [VP su [v v° DO ]]]

Both approaches assume, explicitly or implicitly, that there is only a single

route for deriving verb-initial order. I will argue, however, that Older Egyp-
tian V S O clauses show a considerable degree of syntactic variation without
morphological variation. The different routes for deriving verb-initial sur-
face order stem from the complex interaction between multiple subject
positions and different targets of verb fronting processes. Akin to Biblical
Hebrew (Doron 2000, this volume) and Old Irish (Carnie, Harley & Pyatt
2000), Older Egyptian provides another example of an ancient language in
which short verb movement (V —> T) and Verb Second (V —» C) exist side
by side.

4.2. The diversity of Eventive VSO clauses

This section argues that Eventive VSO order can be arrived at by different
routes. There are two parameters of variation, viz. (i) the availability for
more than one clausal position for the licensing of the subject and (ii) dif-
ferent targets for verb movement.

4.2.1. Evidence for verb raising

The finite verb in Eventive VSO clauses displays the characteristic behaviour
of verb raising. First, the verb may appear to the left of sentential adverbs and
negation, suggesting that it has moved out of the vP shell. As a result of verb
movement, adverbial particles like (j)r-f 'indeed' in (32a) and js 'really' in
(32b) below occur in clause-second position, from where they precede the
subject DP. Evidently, there is no strict adjacency requirement in Older
Egyptian that holds between the finite verb and the nominal subject.
 Word order correlations in Older Egyptian 63

(32) THE POSTVERBAL POSITION OF ADVERBIAL PARTICLES

a. j j r-f Wnjs pn ?χ j.xm(-w) sk
come EV PCL-3SM Unas DEM:SM spirit not.know(-PTCP:SM) destruction
'This (King) Unas arrives, indeed, (as) an indestructible spirit (lit.
which does not know destruction).' (Pyramid Texts 1 5 3 b / W )

b. wn-n js Pjpj [p]n m-^b-sn ncr-w jm(-y)-w pt

be-iMPERFgy E M P H Pepi d e m : s m among-3p god-PM in-NOM-PM heaven
'This (King) Pepi is among them, (namely) the celestial gods.'
(Pyramid Texts 1490a/P)

Since Pollock (1989, 1997) and Cinque's (1999) seminal work on adverbial
distribution and clausal architecture, the placement of negative elements
has been used as a diagnostic tool for verb movement in various languages.
Older Egyptian has two negative markers, the clause-initial negative parti-
cle η (var. njj) and the clause-internal negation w 'not' (see Loprieno 1995;
Kammerzell 1993 for further discussion). As illustrated in (33) below, the
verb raises out of the vP to the left of the negative adverbial w.14

(33) THE POSTVERBAL POSITION OF NEGATION

fzp w fimn zfct-f hrw n(j) fim nb(-j)
accept E V NOT Hemen offering-3SM day LINK(-SM) majesty lord-Is
'(The god) Hemen will not accept his offering (meat) on the day of
the Majesty of My Lord.' (Mo°alla Inscription nr. 8, III. 5)

Second, in Older Egyptian VP coordination, 'Across the B o a r d ' (ATB) ex-

traction is permissible when the coordinate VPs are headed by identical verbs
(see McCloskey 1991: 2 6 6 - 2 7 2 for comparable phenomena in Modern Irish).
As shown by the contrast between (34a) and (34b) below, A T B extraction
of the finite verb strands DP subjects, but does not strand subject clitics.

(34) ACROSS THE BOARD EXTRACTION

a. sd3r-n k?-w-s ijs(-w)
spend.night-PERF EV bull-PM-3SF slaughter-3M STA T
smn-s npd(-w)
fowl-3SF cut.off-3MSTAT
'Her (the ferry boat) bulls spent the night slaughtered and her fowl
cut up.' (Coffin Texts V 9 7 g / T l B e )
64 Chris Η. Reintges

b. jw rd3-n(-j) t n- fiqr fibs n- fi?jj

AUX give-PERF-lS EV bread to hungry.one clothes to naked.one
Ί have given bread to hungry and clothes to the naked one.'
(Urkunden I 122: 6 - 7 )

Third, there is in Older Egyptian an ellipsis process that targets only the
subject, but not the complement and the finite verb of asyndetically coordi-
nated clauses. The verb bears tense and aspect inflection, indicating that the
subject is elided after the verb has moved to Tense (McCloskey 2001: 161-3).

(35) S U B J E C T ELLIPSIS

s-ίηχ-η cw fir m m-k pw n(j) S'nd^tj

CAUS-live-PERF E V y o u : S M Horus in name-2sm DEM:SM LINK(-SM) Anejti
rd'-n n-k fir jrt-f rwd'-t
give- P E R F e v to-2SM Horus eye-3SM s t r o n g - P T C P : S F
d-n n-k s(j) jm-jm-k
place-PERFEV to-2SM it be.strong-INTENS/REPET-2SMEV
'(The god) Horus has nourished you in your name Anejti. Horus has
given to you his strong eye. <He> placed it to you so that you become
very strong.' (Pyramid Texts 614: a-c/T) (cf. ex. (23))

The syntactic facts considered so far support a verb (X°) raising account of
verb-first sentences. It remains to be seen whether an alternative analysis in
terms of remnant verb phrase movement (with previous extraction of the
direct and indirect object) would account equally well for the complexities
of Older Egyptian VSO order. I leave this issue for future research.

4.2.2. Floating subjects and the in/active status of the EPP

Two interrelated questions arise from the previous discussion on verb raising:

(i) How far does the verb raise in Eventive VSO clauses?
(ii) Which positions are available for the postverbal subject?

The focus of this section is on the syntactic distribution of Eventive subjects,

while the next section focuses on the different landing sites of verb move-
ment. Subject constituents are licensed in at least three distinct hierarchical
positions, to wit the highest specifier positions of the vP, the specifier position
of the Tense Phrase (TP), and the cliticisation site of pronominal subjects.
 Word order correlations in Older Egyptian 65

4.2.2.1. vP-internal DP subjects

In the aftermath of Pollock's (1989) SPLIT-IP Hypothesis, it has become

increasingly difficult to find clear cases where the subject of a VSO clausal
pattern can be shown to remain within the vP. A decompositional view of
the former IP makes available a number of potential licensing positions for
the subject, with transitive-expletive constructions being a particular well-
studied case (e.g. Bobalijk & Jonas 1996). The subject of verb-initial
clauses may therefore be quite deeply embedded in the inflectional domain
without actually occupying the vP-internal subject position (McCloskey
1997: 216-7). Although the issue remains murky, standard tests for subject
positioning divulge the existence of at least one type of VSO order in Older
Egyptian, which is derived from verb fronting alone. In this minimal deri-
vation, the EPP is not active, meaning that the TP does not project a speci-
fier (see Doron 2000, this volume; and McCloskey 2001 for the absence of
EPP effects in Modern Hebrew and Irish). The structure for VSO clauses
with unraised subjects is shown in (36) below.

(36) V S O C L A U S E S W I T H N P - I N T E R N A L S U B J E C T DPS

I I I 0
I
[τ Τ [vp D P S U [v. ν [Vp D P d o [ν' V [ROOTP ROOT ]]]]]]

As the locus of tense and event structure, the T°-node qualifies as a sub-
stantive rather than a functional category, falling together with V and Ν
(Chomsky 2001: 9). With respect to its featural content, one may plausibly
assume that its temporal, aspectual, and modal specifications instantiate
interpretable features, which remain accessible throughout the derivation.
Since there is no morphological correlate of subject-verb agreement on
Eventive-inflected verb forms, the stipulation of uninterpretable phi-features
on Tense is empirically unsubstantiated. The trigger for V—>T raising is
therefore not feature co-valuation, but rather the need for providing Tense
with an uncontroversial lexical root at some point in the derivation (Nash &
Rouveret 2003).
In Older Egyptian, Tense is morphologically expressed by means of
suffixes, which need to be attached to a verbal host. Verb raising is a pre-
condition for combining the functional features of Tense with the lexical-
semantic features of the verbal root. What evidence is there for the subject-
in-situ derivation in (36) above? The first argument for the location of DP
66 Chris Η. Reintges

subjects in the specifier position of vP concerns the distribution of adver-

bial elements. A s illustrated in (37) below, the subject appears following
the clause-internal negation w and the emphatic particle js, indicating that it
occupies a lower position in the syntactic structure.

(37) a. V E R B > N E G W > DP S U > DP D O

fzp w Rmn jft-f nb

accept E V NOT Hemen thing-3SM every
'(The god) Hemen will not accept any of his property.'
(Mo°alla Inscription nr. 8, III.6)
b. V E R B > E M P H js > DP S U > DP D 0

η d3d-n js N N pn r? [c?w]-w [p]w

N E G say- PERFEV E M P H N N D E M : S M spell w i n d - P M D E M : S M
[CP Γχ-w rmc-w m-sn ]
k n o w ( - P T C P ) - P M people n a m e - 3 P
'This N N (name of the deceased) did not say the spell of the winds
whose name people know.' (Coffin Texts VI 2 5 2 d / S q 4 C )

This low subject position cannot be identified with a functional projection

on top of the vP. The reason for this is that such a functional projection
serves as a target for pronominal object shift. Older Egyptian has a produc-
tive rule of object shifting that forces direct object pronouns to move out of
the vP to the left of the DP subject (see Collins & Thräinsson 1996 and much
related research on Icelandic object shift). The sentence pair in ( 3 8 a - b )
below exemplifies the contrast between canonical V S O and the derived
V O S order of shifted object pronouns.

(38) OBJECT SHIFT WITH PRONOMINAL DIRECT OBJECTS

a. V E R B > DP S U > DP D 0 > DP I 0

wd'-n Nwn Tjtj n- Jtm

order-PERFnv N u n Teti to Atum
'(The god) N u n has commended (King) Teti to (the god) Atum.'
(Pyramid Texts 6 0 4 a / T )
b. V E R B > C L D O > DP S U > DP I 0

wd3-n sw Pg? n- fw
order- PERFev him Open.one to- Shü
' T h e Open One has commended him (the king) to (the god) Shü.'
(Pyramid Texts 6 0 4 b / N )
 Word order correlations in Older Egyptian 67

Object shift also affects the placement of dative pronoun, as shown in (39a)
below. When both objects are pronominal arguments, indirect object pre-
cedes the direct object in linear order, as seen in (39b) below.

(39) OBJECT SHIFT WITH PRONOMINAL INDIRECT OBJECTS

a. VERB > CLI0 > DPSU > DPD0

fiz w(j) Ν χ η , ίηχ ηί-jj fimn
praise EV me Nekhen liveEV for-lS Hemen
'(The god) Nekhen praises me, (The god) Hemen lives for me.'
(Mo c alla Inscription 12, V ß l )
b. VERB > CL I 0 > C L d o > DP SU
jr-n n-f sw Ri ?χ-ί jm-f
do- perFEV for-3SM it Re be.glorious-3SM E V in-3SM
'Re has done it for him (the deceased) that he may become glorious
there.' (Coffin Texts VI 315j/BlBo)

There is good reason to believe that Older Egyptian object shift is a syntac-
tically rather than prosodically driven reordering process. To begin with, it
is possible for the negation adverb w and the emphatic particle js to inter-
vene between the verb and the shifted object pronoun, as exemplified by
(40a-b) below. It is clear, then, that shifted object pronouns do not form a
prosodic constituent with the main verb.

(40) INTERVENING ADVERBIAL PARTICLES

a. VERB > NEG w > CL D 0 > DP SU

jwi w sw jwi-f
suceedEV NOT him heir
'His heir shall not succeed him.' (Mo°alla Inscription nr. 8, III.7)
b. VERB > EMPH js > CL D 0 > DP SU
η jr-r js sw N N pn sbj(-w) jm-f
NEG do-IMPFEV EMPH it.SM NN DEM:SM rebel-PASSl EV in-3SM
'This NN (name of the deceased) will not do it when one rebels
against him.' (Coffin Texts VI 3 1 5 k / B l B o )

When associated with contrastive focus or emphasis, pronominal objects

exhibit some of the distributional freedom of full DPs. The marked word
order patterns of such emphatically used pronouns are shown in (41a-c)
below. In (41a), the pronominal object form kw 'you (man)' replaces the
68 Chris Η. Reintges

null subject of a Stative clause, appearing in clause-initial position. In

(41b), on the other hand, the pronominal clitic wj 'me' is base-generated in
the left-peripheral topic position of an embedded Clitic Left-Dislocation
construction. Dative pronouns, too, may assume a contrastive reading.
When this happens, they are exempt from object shift and appear to the
right of the direct object DP in clause-final position, as (41c) exemplifies.

(41) THE EMPHATIC USE OF PRONONOMINAL OBJECT FORMS

a. CL D O > V E R B > C L I 0 > D P D 0
kw j.d3d-t(j) n-f m-f pw n(j) s/Ϋ
you:SM say-2sSTAT to-3SM name-3SM DEM:SM LINK(-SM) hare
' Y o u should tell him this name of his "hare"!'
(Pyramid Texts 1269b/P) 15
b. C O M P > C L D 0 > V E R B > CL S U > C L D 0
sk ? wj ms-n-cn wj
COMP PCL me give.birth-PERF-2P EV me
'While (concerning) me, indeed, you have given birth to me.'
(Coffin Texts III 156c/B2Bo a )
c. V E R B > CL S U > D P D 0 > C L I 0
jm wn(-w)-k ?-w(j)-k n-f
NEG:IMP open-PROSP-2SM EV arm-DM-2SM for-3SM
'Do not open your arms for him!' (Pyramid Texts 1269b/P)

The distribution of pronominal clitics is determined by operations of the

syntax rather than by prosodic phrasing. 16 One of these operations is pro-
nominal object shift, which forces direct and indirect object pronouns to
move out of the vP domain. The shifted pronominal object occurs to the
right of negation or focus particles, and as the subject DP must occur to the
right of the raised direct object pronoun, it must be lower in the tree than
either of these two. If we assume that the shifted object pronoun targets a
functional projection above the vP and below the NEGP as well as the
EMPHP, we are left with the vP-internal subject position as the one clausal
position in which the subject DP can be licensed. 17
On this analysis, Older Egyptian might provide a potential counter-
example to the Subject-in-Situ Generalization, according to which no more
than one DP argument can stay in the vP position in which they are merged
externally (Alexiadou & Anagnostopoulou 2001). In canonical VSO senten-
ces of the kind in (38a) above, there is no evidence for phrasal pied-piping
of the subject and direct object argument. The possibility of argument
 Word order correlations in Older Egyptian 69

crowding within the vP cannot be related to rich agreement inflection, how-

ever, since there is no agreement morphology to begin with.

4.2.2.2. Externalisation of the DP subject to Spec, TP

Leftward raising of the verb past the vP-internal subject represents one
route to the verb-initial surface order in Older Egyptian, but not the only
such route. A more complex derivational route by which VSO order can be
arrived at involves subject raising to the specifier of TP and verb movement
to a functional head F° at the left edge of the inflectional domain. It looks
as if the F°-node has information structure content, being involved in the
licensing of left-peripheral topic and focus phrases (see Aoun, Benmamoun
& Sportiche 1994: 204, fn.8; and Uriagereka 1995: 9 5 - 9 7 for related ideas).
Reintges (1997: 127ff) furthermore suggests that auxiliary verbs are merged
into this position, since such verbs play a crucial role in the temporal orga-
nisation of narrative discourse, introducing foregrounded chains of events.
The optional activation of the EPP feature on Tense gives rise to a VSO
variant with raised subjects.

(42) VSO CLAUSES WITH RAISED SUBJECT DPS

[ F F7AUX [TP D P s u [ r T° [vP ? su [v. Ü [ VP D P d o [ v ' V°[ROOTPROOT]]]]]]

ΐ I
The syntactic distribution of negation and information-structuring particles
has been used as a diagnostic tool for both verb movement and subject po-
sitioning. Granted that adverbial elements have a fixed position in the syn-
tactic structure, the surfacing of both the finite verb and the subject DP to
the left of the negation w and the emphatic particle js in examples like (43)
below clearly shows that both constituents have vacated the vP.

(43) V E R B > D P S U > N E G w > E M P H js > CLI0 > DPdo

wnm NN w js fis-w
eat EV N N NOT EMPH faeces-PM
swr NN w js n-sn wzft
drink EV N N NOT EMPH for-3P urine
' N N (name of the deceased) will surely not eat faeces, N N will
surely not drink urine for them.' (Coffin Texts VII 115: i-j/B4Bo)
70 Chris Η. Reintges

Further support for the vP-external location of the subject DP comes from
the stranding of postnominal adjectives. In Older Egyptian, universally
quantified subjects must undergo Quantifier Raising in the narrow syntax.
This is illustrated in (44) below, in which the subject χί nb 'everything'
precedes the object shifted dative pronoun r-f ' f r o m him'. The original
position of the raised quantifier is marked by the stranded adjective d3wt
'evil' (Borer 1995: 535). Further note that universal quantifiers are inter-
preted as free choice items in negative contexts.

(44) STRANDING OF POSTNOMINAL ADJECTIVES

Xw-k [CP χρτ χί nb r-f [vP /Su dhv-t n- cPtcf't \

prevent-2SM EV happen E v thing every to-3SM evil-SF for eternity
' M a y you prevent that anything evil ever happens to him (the deceased
king).' (Pyramid Texts 1654c/N)

E C M constructions with Eventive complement clauses display the hallmarks

of clitic left-dislocation. As we can see from the contrast between ( 4 5 a - b )
and ( 4 6 a - b ) below, the left-dislocated subject appears in a preverbal posi-
tion and is connected to the following clause via a co-referential subject
pronoun (indicated as CL S u)· Under Rizzi's (1997) left-periphery analysis,
the C L L D e d subject D P in ( 4 5 a - b ) below would be located in the specifier
position of a lower Topic Phrase whose head is occupied by the fronted verb.

(45) SUBJECT CLITIC LEFT-DISLOCATION

a. Main clauses
DP S U > V E R B > CL S U > PP
wj? Mi s-qd-d-fi m-gnw jgpt
barque Re CAUS-row- IMPERF-3smEv in-inside cloudy.sky
' T h e barque of (the sun-god) Re, it sails inside the cloudy sky.'
(Coffin Texts IV 1 2 5 c / S l P )
b. Subordinate clauses (reported speech)
C O M P > DPsu > V E R B > CL S U
dM-n-k n- R? [ CP ntt Nt, jw-S; ]
say-PERF-2SM E V to R e COMP N e i t h come(-PROS)-3SF E V
' Y o u told (the sun-god) Re that (Queen) Neith would come.'
(Pyramid Texts/Neith 4 0 - 4 1 [pi. 7])
 Word order correlations in Older Egyptian 71

(46) EC Μ CONSTRUCTIONS WITH EMBEDDED EVENTIVE VPs

a. D P S U > V E R B > C L S U > PP
njj gm-n(-j) [ΐΡ rmc-w nb, fm-sn, m w?-wt n- d3i ]
NEG find-PERF-lSEV man-PM every walk-3PEV on road-PF for storm
Ί did not find any people walking on the roads because of the
storm.' (Urkunden I 182: 15-16)

b. C L D O > V E R B > CLSU > PP

m?-?-sn [ TP w(j) { iq(-j), m rwt ]
see-IMPERF-3PEV me proceed-lS E v from door-gate
'(Whenever) they (the people) saw me coming out of the door-
gate (...)' (Inscription of Bia, 1. 7 [Edel ZÄS 83 (1958): 8])

In ECM constructions with Eventive clause complements, the CLLDed

subject DP subject is accusative-case marked by the matrix verb, witness
the presence of the direct object pronoun w(j) 'me' in (46b) above. Unlike
embedded CLLD of the kind in (45b) above, subordinating 'THAT'-
complementizers like ntt appear to be systematically absent. I interpret this
to mean that ECM constructions never exceed the domain of a Tense
Phrase (TP). The simplest assumption to account for the CLLD properties
of the ECM subject is to say that the EPP feature on the Tense head may be
activated, projecting a specifier for the raised subject.
Further evidence for the optional activation of the EPP feature on Tense
in Eventive VSO clauses comes from auxiliary verb constructions. The
contrast between vP-internal and vP-external subject DPs is illustrated in
(47a-c) below. In (47a), the subject DP Rm(-j) 'My Majesty' is preceded by
both the main verb m?-n 'has seen' and the auxiliary verb jw. By contrast,
the subject DPs rwd3 'the strong one' and jnb-w-s 'its wall' in (47b) and
(47c) appear in a higher subject position to the right of the auxiliary jw and
to the left of the main verb.

(47) LOW AND HIGH SUBJECTS IN AUXILIARY VERB CONSTRUCTIONS

a. AUX > VERB > DP SU
jw m?-n fim(-j) md3?t-k tn (...)
AUX see-PERF E v m a j e s t y - I S letter-2SM DEM:SF
'My Majesty has seen this letter of yours (...)'
{Urkunden I 60: 16)
72 Chris Η. Reintges

b. A U X > DPsui > V E R B > CL S U i

jw rwd'' j d-n-fl r t?
AUX strong.one place-PERF-3SM EV to earth
' T h e strong one has placed (himself) to earth.'
(Lepsius, Denkmäler II 104b)
c. A U X > DP S u j > V E R B > VBL j
jw jnb-w-sj dm-n j pt
AUX wall-PM-3SF scratch-PERF E V sky
'Its (the temple's) walls scratched the sky.' (Stela Louvre C3: 6)

The raising of the subject displays some of the properties of A'-movement:

the subject D P in Spec, TP binds either a resumptive pronoun, as in (47b),
or a gap, as in (47c) above. The optional activation of the EPP feature on
Tense falls into place if one assumes with Chomsky (2001: 3 3 - 3 5 ) that the
EPP position is assigned a special interpretation. In the examples consid-
ered so far, raised external DP arguments are non-canonical subjects: they
are either universally quantified DPs or sentence foci. It is therefore natural
to relate the active status of the EPP to the information structure of the
clause. In line with Zubizaretta (1998: 99-11 8), I assume that Older Egyp-
tian Tense may be a syncretic category, encoding not only temporal deixis
and mood, but also focus and emphasis (whence the A'-properties of its
specifier).

4.2.2.3. The cliticisation site of subject pronouns

In this section, I briefly discuss the syntax of pronominal enclisis. Although

I do not deny the importance of morpho-phonological constraints on the
distribution of enclitic subject pronouns, I will take a primarily syntactic
view on cliticisation, relating the different positions of pronominal clitics to
movement operations that apply before spell-out. From this perspective,
clauses containing clitics have essentially the same syntax as clauses with
moved D P arguments (see Kayne 1989, Uriagereka 1995 and, more re-
cently, Nash & Rouveret 2003).
As already pointed out in section 3.2.1, the Complementarity Principle
can be directly related to the argument status of subject and object pronouns:
as pronominal arguments, they compete with full DPs for the same struc-
tural slots in the syntactic representation. Nominal and pronominal argu-
ments differ, however, systematically from one another with respect to their
 Word order correlations in Older Egyptian 73

licensing requirements. While full DP subjects and objects can be licensed

in-situ in vP-internal argument positions, their pronominal counterparts must
always vacate the vP. Thus, weak direct and indirect object pronouns must
undergo object shift to left of the vP, appearing in front of the vP-internal
subject DP. In a similar vein, subject clitics raise out of the vP to the higher
subject position Spec, TP while the verb moves higher up to a functional
position F°. Doron (1996, 2000) proposes to analyse pronominal clitics as
anaphors bound by discourse operators. Such operators may be left unpro-
nounced, when the identity of the antecedent is recoverable from the pre-
ceding discourse. The derivation for V S O clauses with pronominal clitics
will therefore be entirely parallel to the one with raised D P subjects.

(48) VSO CLAUSES WITH SUBJECT CLITICS

I I0
I I I
[fp O p [ F F° [ T P CL S U [ T ' T [ v p ? c l [ v v [Vp D P d o [v V° [ROOTP ROOT ]]]]]]

ΐ I
The raising analysis of enclitic subject pronouns is empirically well moti-
vated. As shown by the contrast between the a- and the b-examples of ( 4 9 ) -
(50) below, subject DPs and clitics appear on opposite sites of the clause-
internal negation w and the emphatic particle js. This mirror image pattern
follows from their respective vP-internal and vP-external location.

(49) a. V E R B > N E G w > D P S U > DPdo

fzp w Rmn jft-f nb
acceptnv NOT Hemen thing-3SM every
'(The god) Hemen will not accept any of his property.'
(Mo°alla Inscription nr. 8, III.6) (cf. ex. (37a))
b. V E R B > CLsu > N E G w > D P d 0 > PP
fi?m-cn w N N pn m j?dt-cn tw
catch-2P EV NOT N N DEM:SM with net-2P DEM:SF
'You will not catch this N N (name of the deceased) with this your
net.' (Coffin Texts VI 231/B1 Bo)
74 Chris Η. Reintges

(50) a. N E G > V E R B > E M P H js > DP S U

η jw-n js Wnjs d3d-f
NEG come-PERF EV EMPH Unas self-3SM
'(King) Unas has not come by himself.' (Pyramid Texts 333b/W)

b. N E G > V E R B > CL S U > E M P H j s

η jw-n-f js d?d-f
NEG come-PERF-3SM EV EMPH self-3SM
' H e has not come by himself.' (Pyramid Texts 333b/T)

That subject clitics occupy a high position in the syntactic tree is also evi-
dent from their appearance in front of shifted pronominal objects, as shown
in (51).

(51) V E R B > CL S U > C L I 0 > C L D 0 > PCL jr-f > f f / / - A D V E R B

s-rs-j n-k sw jr-f jr- j f s t
CAUS-wake-1 SEV for-2SM him PCL-3SM for what?
'For what reason should I wake him up for y o u ? '
(Pyramid Texts/Jbj 5 9 4 - 5 9 5 [pi. 11])

In auxiliary verb constructions, focused subject DPs raise to Spec, TP for

feature-checking purposes and appear - as a consequence - to the right of
the auxiliary jw and to the left of the main verb. Subject clitics are licensed
in exactly the same syntactic positions, as shown in (52) below. It is clear,
then, that enclitic subject pronouns have the syntactic profile of pronominal
arguments.

(52) A U X > DPsu/CLsu > V E R B > CL S U

jw-j sc-j jw b?-ji sc-f,
AUX-1S beget-1S EV AUX soul-lS beget-3SM EV
Ί (usually) beget (and) my soul usually begets.'
(Coffin Texts I 3 6 4 b / S l C )

In auxiliary verb constructions with an EPP position, the moved subject is

cross-referenced by a clitic double on the main verb. If this clitic double
were left behind in the vP-internal subject position, it should be lower in the
syntactic structure than shifted object pronouns, but it is not. As we can see
from (53) below, clitic doubles always appear preceding pronominal ob-
jects, indicating that they no longer reside in the vP domain.
 Word order correlations in Older Egyptian 75

(53) A U X > D P S U j > V E R B > C L S U i > CL D O > A D V E R B

wn ßm-fi fiz-f, w(j) wr r χί nb
AUX majesty-3SM praise-3SM EV me great than thing every
'His Majesty used to praise me to the highest degree (lit. more than
everything).' {Urkunden I 194: 10)

To answer the question where the clitic double has been raised to, a com-
parison between Stative and Eventive complement clauses is instructive.
The embedded subject position of Eventive and Stative clauses can be oc-
cupied by an enclitic pronoun, which leans on the preceding lexical com-
plementizer. But while there are two instances of the subject clitic in the
embedded Eventive clauses in (54a) below, the presence of independent
agreement morphology on the finite verb in the Stative counterpart in (54b)
automatically excludes such clitic doubling.

(54) EMBEDDED EVENTIVE AND STATIVE CLAUSES

a. C O M P > CL S U > E V E N T I V E V E R B > CL S U

dd'-sn yj ßm-f [CP wnt-i\ db?-f, ]
say-3P EV under majesty-3SM COMP-3SM faint-3SM EV
'They reported to His Majesty that he (the physician) had fainted.'
(•Urkunden I 42: 11)
b. C O M P > CL S U > STATIVE V E R B
r nP-t rx-t(j) [CP wnt-k, h?-t(j)t m ßtp
to give-INF understand-PASS2 COMP-2SM descend-2s ST AT in peace
m Y?m ]
from Y a m
'to let it be known that you descended in peace from (the land)
Yam' ( U r k u n d e n 1 1 2 8 : 7 - 8 ) (cf. ex. (27a))

This asymmetry is expected when Stative inflections and clitic doubles

compete for the same position. At the present state of research, it is not
entirely clear whether the clitic moves to Tense independently of the verb,
or whether enclisis takes place when the verb resides in a lower functional
head and the VERB-CLITIC complex moves to Tense as a unit (see Adger
1997 for an analysis along these lines).
76 Chris Η. Reintges

4.2.3. The Verb Second Condition

The previous discussion on floating subjects has identified three different

routes by which by which verb-initial surface order can be arrived at. The
availability of the inflectional subject position Spec, TP could be shown to
derive from an optional activation the EPP feature on Tense. When this
happens, the finite verb raises out of the TP to a higher functional head F,
which relates to the information structure of the clause. Next to V —> Τ and
V —> F movement, Eventive-inflected verb forms may also undergo V —> C
fronting of the kind attested in productive Verb Second languages. In Older
Egyptian, however, the Verb Second Condition is parameterised in the sense
that not all finite verb patterns have to cover the entire path leading Τ to C.
Verb Second (henceforth V2), as it is known from West Germanic lan-
guages like German, concerns the different position of the finite verb in main
and subordinate clauses. In tensed clauses without an overt complemen-
tizer, the verb must occupy a position immediately following the sentence-
initial constituent, whatever the status of that constituent is. In (55a), for
instance, the auxiliary verb hat 'has' appears in the second position of the
clause, while the clause-initial position is occupied by the object DP diesen
Film 'this movie'. The subject Katja and the remaining sentence constituents
follow the finite verb. In embedded clauses with an overt complementizer,
by contrast, there is no V2 ordering and the finite verb appears at the end of
the clause, as in (55b) below. The combination of V2 ordering with the com-
plementizer daß 'that' yields ungrammatical structures like (55c) below.

(55) VERB SECOND IN STANDARD GERMAN

a. Diesen Film hat Katja schon gesehen.
this(-ACC) movie(-ACC) has Katja(-NOM) already seen
' K a t j a has already seen this movie.'
b. Chris glaubt, [cp daß Katja diesen
Chris(-NOM) believes that Katja(-NOM) this(-ACC)
Film schon gesehen habe],
movie(-ACC) already seen has:SUBJ
'Chris believes that Katja has already seen this movie.'
c. * Chris glaubt, [cp daß Katja habe diesen
Chris(-NOM) believes that Katja(-NOM) has:SUBJ this(-ACC)
Film schon gesehen ].
movie(-ACC) already seen
'Chris believes that Katja has already seen this movie.'
 Word order correlations in Older Egyptian 77

In most generative analyses going back to den Besten's (1983) seminal

paper, V2 is derived from verb raising to Comp and an additional require-
ment that the specifier of the matrix clause be filled by some XP constituent.
This analysis accommodates rather naturally the observed complementary
distribution of lexical complementizers and fronted finite verbs in embedded
contexts (see Vikner 1995 for detailed discussion on the V2 phenomenon).
An obvious extension of this approach is to posit a set of VSO languages
where the clause-initial position of the finite verb derives from V —> C raising,
but where the requirement on filling the specifier of CP is not imposed (see
Carnie, Harley & Pyatt 2000: 41). On the face of it, Older Egyptian does
not seem to qualify as such a 'weak' V2 language. In embedded clauses
without an overt complementizer, the Eventive verb raises to the vacant C°-
node, as in (56a) below. However, note that embedded VSO clauses are
equally well formed, when a subordinating complementizer like wilt 'that'
is present, as seen in (56b) below. Recall that in German V2 does not raise
when the C°-node is occupied by a complementizer.

(56) EMBEDDED VSO CLAUSES

a. With vacant C°
sd3m-n-sn [CP d-d Nfr-k?-M m?Tt <m st jsfO ]
hear-PERF-3PEV place-iMPERFEV Nefer-ka-Re justice in place injustice
'They heard (that) (King) Nefer-ka-Re places justice <instead of
injustice^' (Pyramid Texts 1775b/N)
b. With lexical C°
d'd-n-k r md3?t-k tn [CP wnt jn-n-k
say-PERF-2SM EV at letter-2SM DEM:SF COMP bring-PERF-2SM EV
dncj jb?-w ncr m t? Ϋχ-tj-w
d w a r f dance-PM god f r o m land horizon-NOM-PM
'You said in this your letter that you brought a dwarf for the
dances of the god from the land of the Horizon-inhabitants.'
(Urkunden I 128: 14-16)

Closer inspection reveals, however, that certain verb-initial clausal patterns

are subject to the Verb Second Condition. To illustrate this point, the con-
trastive behaviour of eventive passives has been chosen for sample treatment.
The most productive passive patterns are the apophonic passive 1, which is
formed with a vocalic passive morpheme /u/ (rendered as - w in hiero-
glyphic writing), and the affixal passive 2, where the passive morpheme
-t(j) (var. -tw) follows the verbal stem and its tense/aspect formatives. As
78 Chris Η. Reintges

we can see from the following minimal sentence pair, passive Is and pas-
sive 2s appear as functional equivalents of one another in main clauses.

(57) THE PASSIVE 1/PASSIVE2 ALTERNATION IN ROOT CLAUSES

a. η ms-jj-j js ms-jj-t
NEG bear-PASSl-lS E V EMPH bear-GER-SF
Ί w a s really not born in a birthing.' (Coffin Texts I 3 4 4 c / G l T ) 1 8

b. η ms-n-t(j)-j js ms-jj-t
NEG bear-PERF-PASS2-lS EV EMPH bear-GER-SF
Ί was really not born in a birthing.' (Coffin Texts I 3 4 4 c / S l C )

Both passive patterns occur in free variation in embedded clauses without

an overt complementizer.

(58) THE PASSIVE 1/PASSIVE2 ALTERNATION IN EMBEDDED CONTEXTS

a. jw wd3-n n-j RY [Cp rd^(-w) n-k tp-k ]
AUX order-PERF EV t o - l S Re give-PASSl EV to-2SM head-2SM
'(The sun god) Re has given an order to me (that) you should be
given your head.' (Coffin Texts I 1 9 9 f / B 1 2 C )

b. jw wcf'-n n-j R? [CP rd'-t(j) n-k tp-k ]

AUX order-PERF EV to-lS Re give-PASS2EV to-2SM head-2SM
' ( T h e sun god) Re has given an order to me (that) you should be
given your head.' (Coffin Texts I 1 9 9 f / B 1 0 C b )

The symmetry breaks down in embedded clauses with a filled complemen-

tizer position. In such complement clauses, passive 2s seem to be system-
atically absent and the corresponding passive Is are selected in their place.
This selectional restriction can be accounted for in a principled way, if the
Verb Second Condition is operative on the syntactic derivation of passive 2
clauses. As a result, the passive verb must always move all the way up to
C°, whence the complementary distribution with lexical complementizers.
The Verb Second Condition does not apply in clauses with passive 1 s that
may but need not undergo V —> C fronting.
 Word order correlations in Older Egyptian 79

(59) PASSIVE 1 SELECTION IN EMBEDDED CONTEXTS WITH FILLED COMP

a. jr nf cl'd-w χτ fim(-j)
as.for DEM:SM say-PASSl-(PTCP:SM) under m a j e s t y - l S
[cpwrit xtm(-w) wcPf-w) n(j)-w njswt r fmTw
COMP seal-PASSl ev order-PM LINK-PM king about Upper.Egypt
r jr-t h? n(j) k?t n- njswt (...)]
to do-INF corvee LINK-SM work for king
' A s for what has been said on the part of My Majesty that royal
orders have been sealed concerning Upper Egypt to do the corvee-
service for the king (...)·' ( U r k u n d e n I 282: 15-16)

b. jsc wb?(-w) n-j dw?t

COMP open-PASSl E V for-lS netherworld
'while the Netherworld is opened for m e '
(Coffin Texts VI 9 1 q / M 2 2 C )

Superficially similar distributional patterns obtain in auxiliary verb con-

structions. Edel (1955: 451, §891) was the first to observe that either an
aspectual auxiliary or a passive 2 marked lexical verb appears in the front
position of the clause. Thus, compare (60a) with (60b) below.

(60) PASSIVE 1 SELECTION IN AUXILIARY VERB CONSTRUCTIONS

a. jw rd3(-w) [cp jw-t smr wft(j) ßmj Jnj-jt-f r-s ]
AUX give- PASS 1EV come-SUBJEV friend sole HemiAntef for-3SF
'(It) has been arranged (that) the sole friend (honorary title) Antef,
(son of) Hemi, comes for it.' ( U r k u n d e n I 296: 16)

b. rd'-t(j) [cp jwt <fi?tj-?> smr wWj)

g i v e - PASS2 E V come-SUBJ E V prince friend sole
jmy-r pr-wj qbfixwnj m %sf m iiw'i-w (...) ]
superintendent house-DM coolKhuni in proceed(-lNF) in ship-PM
'(It) is arranged (that) the prince, sole friend, superintendent of the
two cool-houses, Khuni advances with ships ( . . . ) . '
{Urkunden I 127: 13-15)

At this juncture, one might suspect that the auxiliary and the passive 2 verb
compete for the same syntactic position, viz. the C°-node. This cannot be
80 Chris Η. Reintges

the whole story, however. As we can see in (61) below, an auxiliary verb
like wn 'be' may appear in a conditional adjunct clause formed with the pre-
positional complementizer jr 'as for, i f . I take this to mean that auxiliary
verbs are not subject to the Verb Second Condition.

(61) A U X I L I A R Y V E R B S IN C O N D I T I O N A L A D J U N C T CLAUSES

jr wn mr-jj <n->cn [Cpfiz cn njswt (...)]

a s . f o r AUX l o v e - P A S S l E V t o - 2 p praiseEV you:P king
Yq-cn w jr jz pn zbt-tjn
enter-2pEV N O T to tomb D E M : S M be.unclean-2p STAT
'If it pleases you (that) the King praises you (...), you should not
enter this tomb in an unclean state.' (Urkunden 1218: 8-10)

What blocks passive 2 selection in auxiliary verb constructions is probably

not a filled complementizer position. Rather, V —> C fronting of the passive
is blocked by the Head Movement Constraint, given that an intermediate
landing site on the path of verb movement is occupied by the auxiliary
verb. No such minimality violation occurs when the auxiliary verb itself
undergoes passive 2 formation, as seen in (62) below. Again, the main verb
occurs in the passive 1.

(62) PASSIVISED AUXILIARY VERB CONSTRUCTIONS

γβγ-η-tw ίϊΫ-w jm wr r χί nb
AUX-PERF- rejoice- P A S S 1EV there great than thing every
PASS2 e v
'Then one rejoiced (there was rejoicing) very much (lit. more than
anything).' (Stele Louvre C 11: 17)

The occurrence of examples like (62) above is in line with our expectations,
since the auxiliary verb and the passive 2 formative -tw are both located in
their designated landing site, which is the C°-node. While Verb Second
applies in full force in passive 2s, passive Is may, but need not, move all
the way up to C° and may eventually be positioned in a lower functional
head. The parameterisation of the V2 feature correctly predicts that pas-
sives 1 and passives 2 occur in free variation in main and embedded clauses
with a vacant C°-node. It is worth pointing out that weak V2 effects are by
no means a hallmark of the passive paradigm, but can also be observed in
other verbal-inflectional patterns, such as subjunctive or imperfective verb
forms (see Reintges 1997: 79ff). Future research has to clarify the morpho-
syntactic factors that underlie the parameterisation of the Verb Second
Condition.
 Word order correlations in Older Egyptian 81

4.3. The agreement configuration of Stative SVO clauses

The focus of this section is on the subject-initial syntax of Stative clauses.

The SVO surface order of Stative constructions corresponds to a single
structure, which derives from two independent though related movement
operations: both the verb and its subject move out of the vP into the inflec-
tional domain, targeting the head and the specifier position of the subject
agreement phrase AGRSP. As we have seen in section 3.2, Stative inflec-
tions are portmanteau morphemes that express stative semantics and sub-
ject-verb agreement at the same time.The conspiracy of V—> AgrS 0 move-
ment and the merging of the subject in the Spec of AGRSP give rise to a
canonical agreement configuration of the following kind:

(63) STATIVE SVO CLAUSES

UgrSP DPsu [AgrS' A G R S [ v P ? s u [ν· V [Vp DPd0 [v' V 0

[ ro u r Ρ ROOT ]]]]]]

ΐ I
Despite their derivational connection, the merging of the subject into the
Specifier of AGRSP and the adjunction of the verb into the AgrS°-head
satisfy different syntactic constraints. The interpretable features on the
AGRS°-node are [+stative], [+finite], while the φ-features and the EPP fea-
ture represent uninterpretable features. In Chomsky's (2000, 2001) system,
the EPP, like other selectional features, seeks an XP to merge with the
category it heads. The uninterpretable φ-features act as a probe that seeks a
goal with the corresponding interpretable features to establish agreement.
The AGREE relation removes the uninterpretable features from the narrow
syntax, making it possible for the narrow syntactic derivation to converge at
LF. As we will see next, verb movement is triggered by the need to combine
the interpretable semantic features on the AGRS°-node with a verbal root.

4.3.1. Independent V-to-Agrl·f movement of Stative-inflected verbs

Given the portmanteau character of Stative inflections as fused aspect and

agreement markers, one might plausibly assume that these markers (or the
feature matrixes that they represent) are merged directly into the designated
inflectional head. These features are evidently interpretable, encoding tense,
finiteness and stative semantics. Stative AGRS0 is therefore a substantive
82 Chris Η. Reintges

category just like Eventive Tense, which must be combined with a verbal
root at some point in the derivation. I assume modulo Baker's (1988)
'Mirror Principle' that verb movement and adjunction to AGRS0 have a mor-
phological correlate in the templatic structure (VERB + SUFFIX) of Stative-
inflected verb forms.
If the clause-second position of Stative verb forms can be identified
with the AGRS°-node, the preverbal subject is in all likelihood located in
the specifier position of that node. This account predicts - c o r r e c t l y - that
the subject DP appears in front of adverbial particles and shifted pronomi-
nal objects. In (64) below, both the ECM subject fiq? Y?m 'the rule of
Yam' and the embedded Stative fm(-w) 'had departed' precede the subject-
oriented particle (j)r-f, which is left-adjoined to the vP.

(64) DP SU > STATIVE > PCL r-f > PP

gm-n(-j) L G S S P ßq? Y?m fm(-w) r-f r t? Cmfi ]
flnd-PERF-1 SEV ruler Yam leave-3MSTAT PCL-3SM to land Libyan
Ί found (that) the ruler of Yam had departed to the land of the Libyan.'
(Urkunden I 125: 15-16) (cf. ex. (15a))

In examples like (65a) below, the postverbal position of direct object pro-
noun cw 'you (man)' is by itself not indicative of object shift, given the
SVO surface order, but would still indicate that both the subject and the
verb have raised out of the vP domain. The applicability of pronominal
object shift to a Stative clause is, however, evident from the CL 1 0 > DP DO
order of indirect object pronouns in examples like (65b) below. Recall that
the canonical order of nominal object is DP DO > DPI 0 .

(65) OBJECT SHIFT

a. DP SU > STATIVE > CL D 0
sc Pjpj pn rx(-w) cw
COMP P e p i DEM:SM learn-3M ST AT you:SM
'while this (King) Pepi knows you' (Pyramid Texts 329b/P)
b. S T A T I V E > C L I 0 > DPD0
s-h?-k(j) n-f fitp pn n- hrw 17
CAUS-descend-lSsTAT for-3SM altar DEM:SM in d a y 17
wß?(-w) m ßt-nbw
quarry-3M S TAT in Hatnub
Ί sent him this altar in 17 days after it had been quarried in Hatnub.'
0Urkunden I 108: 1-2)
 Word order correlations in Older Egyptian 83

It is clear then that the preverbal subject and the finite verb of Stative
clauses have both raised out of the vP. That Stative subjects are, indeed,
located in the Specifier position of AGRSP and not higher up in the syntac-
tic tree is evident from their occurrence as ECM subjects as well as their
intermediate position between the auxiliary jw and the main verb, since
auxiliary verbs appear in a preclausal functional position F°.

(66) A U X > D P S U > S T A T I V E > D P D O

jw NN tn r%-c(j) fdw jpw r<n-w-cn>
AUX N N DEM:SF learn-3f S T A T four DEM:P name-PM-2P
n-w jrt Rr
LINK-PM eye Horus
'This N N (the female deceased) knows these four names of you be-
longing to the eye of (the god) Horus.' (Coffin Texts III 339f/Sq3C)

To conclude, considerations of affix order as well as syntactic placement

facts provide empirical support for the synchronisation of independent V —>
AGRS0 movement and subject raising to Spec, AGRSP in the derivation of
Stative clauses. There is some cross-linguistic evidence for the correlation
between the occurrence of finite verb movement to the highest inflectional
node and the explicitness of verbal paradigms. It generally seems to be the
case that if a language has more verbal inflection, it is more likely to have
independent V-to-I movement (Vikner 1995: 163).
In Older Egyptian, this correlation is quite straightforward, since only
Stative verb forms are inflected for agreement. It may very well be the case
that richness of agreement is based on the existence of distinct person in-
flections in both singular and plural, as suggested by Roberts (1993: 2 6 6 -
267). The Older Egyptian Stative data seem to be less compatible with
Rohrbacher's (1994: 109, 119) view that independent V-to-I movement is
contingent on the presence of a complete or full paradigm, meaning that
agreement of the first and second person differ from each other as well as
from the forms of the third person. This is evidently not the case in the
Stative paradigm, where the syncretic ending -t(j) is used for the second
person singular as well as the third person feminine singular and plural,
with the pro-drop serving as a disambiguating device. I leave a more pre-
cise formulation of the relation between inflectional morphology and verb
positions for future research.
84 Chris Η. Reintges

4.3.2. EP Ρ-induced subject merge

According to Chomsky (2000: 102), EPP features are uninterpretable (non-

semantic), although the configuration they establish has effects for interpre-
tation. Particularly well-studied cases are the additional specifier positions
on top of C and little v. In this section, I will show that the merge of the
subject into the specifier position of AgrS 0 may display similar surface
semantic effects. To be more precise, AgrS 0 is a syncretic category, which
is endowed with features of the peripheral system (topic, focus). As a re-
sult, the EPP-induced preverbal subject position has mixed A/A'-properties
(Zubizaretta 1998: 9 9 - 1 0 8 , Chomsky 2000: 108). From the perspective of
information structure, Stative sentences are for the most part backgrounded
propositions, in which the subject constitutes the sentence topic about
which the Stative V P makes a comment. As sentence topics, Stative sub-
jects correspond to information that is given or available from the preced-
ing discourse and are therefore part of the presupposed portion of the
clause. The familiarity condition on Stative subjects is particularly clear in
the case of subject pronouns like -f ' h e ' in (67a) below, but sense-of-
identity anaphors like k j j 'other' in (67b) are equally presuppositional in
that the identity of their referent can be inferred from the preceding clause.

(67) THE FAMILIARITY CONDITION

a. d3d-cn χΓ R? [Cp wnt-f jj(-w) m tier]
say-3P EV to Re COMP-3SM come-3MSTAT as god
' Y o u should announce to (the sun-god) Re that he (the deceased
king) has arrived as a god.' (Pyramid Texts 1 8 6 2 a / N )

b. jr-n jt Nfr-k?-R'i jb-f

make-PERF EV father Nefer-ka-Re heart-3SM
kjj fd-jj n-f g?k-f
other remove-3MSTAT for-2SM be.hostile-3SM EV
' T h e father of (King) Nefer-ka-Re (re-)made his heart. The other
one is removed from him (when) it became hostile.'
(Pyramid Texts 1 1 6 2 a / N )

The familiarity condition does not exclude Stative subjects from being as-
sociated with different types of semantic focus (E. Kiss 1998). In (68) be-
low, the interjection m- ' l o o k ' signals new information focus on the adja-
cent subject. Definite expressions (proper names, personal pronouns) are
admissible as presentational foci, since they correspond to discourse-new
 Word order correlations in Older Egyptian 85

information, meaning that the relation its referent enters into is novel with
respect to the preceding stretch of discourse.

(68) PRESENTATIONAL FOCUS

m-k Tjtj pr(-w)
INTERJ-2SM Teti come.forth-3MSTAT
m-k Tjtj jw-f
INTERJ-2SM T e t i come-3SMEV
'Look (King) Teti has come forth, look, Teti comes (here).'
(Pyramid Texts 333a/T)

In (69) below, the sentence negation η takes wide scope over the indefinite
N P z? (J)tm 'a son of (Atum)', which therefore receives the semantic inter-
pretation of a downward entailing quantifier (NOT > A SON = NO SON). As
non-referential expressions, such quantifiers are excluded as topics, although
they do function as subjects (Alexiadou & Anagnostopoulou 1998: 5 0 4 -
511; Alexopoulou, Doron & Heycock 2003: 9-12).

(69) DOWNWARD ENTAILING QP

cz-jj Ϋβ-w zm?-jj mgn-wt η z? (J)tm
tie-IMP:P rope-PM make.ready-IMP:P ferryboat-PF for son Atum
η z? (J)tm jw-jj
NEG s o n A t u m be.boatless-3M S TAT
'Tie the ropes and make ready the ferryboats for the son of (the god)
Atum. No son of Atum is left without a boat.'
(Pyramid Texts 1742: b - c / M )

The preverbal subject position hosts not only presentational foci and raised
quantifiers, but is equally well suited as a landing-site for w/?-phrases, as
seen in (70) below.

(70) WH-IN-SITU SUBJECT QUESTIONS

m twt(-w) n-f
who resemble-3MsTAT to-3SM
' W h o resembles him?' (Pyramid Texts 1689b/M)

The dual nature of the preverbal subject position as an A/A'-position has

been attested for such diverse languages as Modern Spanish (Zagona 2002),
Yiddish (Diesing 1990), and Modern Hebrew (Borer 1995). Older Egyptian
86 Chris Η. Reintges

differs from these languages in that the preverbal subject position is re-
stricted to nominative arguments. It looks as if the EPP position of Stative
clauses is targeted as a landing site by various types of operators, indicating
that the AgrS 0 conveys some left-peripheral functional features (Topic,
Contrast, Focus). This provides an explanation for the absence of topicali-
sation and clitic-lefit-dislocation structures in Stative clauses.

4.3.3. A note on the Broad Subject construction

In this section, I will show that there are Stative clauses that look superfi-
cially like CLLDs, but actually instantiate a sentence pattern with two
nominative subjects. In the so-called Broad Subject construction, the A-
and A'-properties of EPP-induced subject merge are distributed over two
specifier positions, as exemplified in (71a-b) below.

(71) THE BROAD SUBJECT CONSTRUCTION

a. psd't jb-sn nd3m-w fir χην
ennead heart-3P be.sweet-3MsTAT about voice
qr-r n(j) nf?-jj
stormy-INTENS/REPET(-PTCP:SM) LINK(-SM) e x h a l e - P T C P : S M
' T h e ennead (i.e. a constellation of nine gods), their heart is de-
lighted about the stormy voice of the exhaler.'
(Coffin Texts IV 9 3 j / B 5 C )
b. jw b?kt tn n(j)-t Mrrj jb-s ϊηχ('\ν)
AUX m a i d . s e r v a n t DEM:SF LINK-SF M e r e r i h e a r t - 3 S F live-3m S TAT
r-f m??-s wpwty n(j) nb-s
PCL-3SM see-3sf E V messenger LINK-SF lord-3SF
' N o w , this maid servant of Mereri, her heart is alive (whenever)
she sees a messenger of her lord.' (Pap. Cairo CG 58043: 8 - 9 )

The Broad Subject construction has been extensively studied by Doron

(1988), Doron & Heycock (1999), and Alexopoulou, Doron & Heycock
(2003). Their research shows that for all intents and purposes the two non-
adverbial D P s behave like nominative subjects. In examples ( 7 1 a - b ) above,
the lower narrow subject and the Stative V P form a semantic predicate for
the clause-initial broad subject, with the open position provided by the re-
sumptive pronominal on the narrow subject. The principal difference be-
tween broad and narrow subjects is that the former are merged directly into
 Word order correlations in Older Egyptian 87

the higher specifier, while the latter are derived subjects that originate
within the vP. Since the broad subject may be either topic or focus, it seems
natural to classify the higher specifier as an A'-position and the lower speci-
fier as an Α-position. K u r o d a (1988) draws a connection between the mul-
tiple filling of the specifier of IP in Japanese and the impoverishment of
agreement. On this account, one would expect, - contrary to the facts - that
the multiple subject construction would be another syntactic variant of an
Eventive V S O clause that lacks subject-verb agreement altogether. It rather
looks that the possibility of filling and licensing multiple subjects depends
on the active status of the EPP feature on the relevant inflectional node
rather than the richness of subject agreement itself.

5. S u m m a r y and conclusions

This study discussed some core phenomena related to the verbal-

inflectional system of Older Egyptian. In this language, the event- or state-
related interpretation of the verbal predicate is therefore not lexical infor-
mation preserved in the syntax, but emerges from operations of the syntax:
the merger of the verbal complex with inflectional morphology that takes
place in the cyclic domain on top of the vP shell. The semantic regularities
between Eventive and Stative alternants of a given root in various lexical
classes of transitive, unergative and unaccusative verbs follow from the
uniformity with which the syntax combines verbal stems with inflectional
morphemes, leaving little or no space for lexical idiosyncrasy.
Comparative syntactic research on verb movement has disclosed an
impressive amount of evidence for the crucial role of inflectional morphol-
ogy as a trigger for the fronting of the finite verb to a particular syntactic
position. The Older Egyptian Eventive conjugation distinguishes uniquely
between person, gender and number features, while several paradigmatic
cells of the corresponding Stative conjugation share the same morphologi-
cal exponent. Yet, it could be shown that subject-verb agreement proper is
only represented by the Stative paradigm, while the concord-marking per-
sonal suffixes on Eventive verb forms are actually enclitic pronouns that
correspond to argument positions. In the Stative paradigm, the presence of
morphologically discrete agreement marking excludes independent tense
and aspect morphology. This contrasts with the Eventive paradigm, where
absence of agreement inflection correlates with the combinability with the
whole range of tense-aspect-mood markers. Given that the personal inflec-
88 Chris Η. Reintges

tions on Eventive verb forms instantiate fully fledged pronominal argu-

ments, Older Egyptian provides evidence against Alexiadou & Anag-
nostopoulou's (1998: 518) claim that ' V S O languages are also pro-drop
languages, because only pro-drop languages have the option to check the
EPP feature through the verbal agreement morpheme'. Table 3 summarises
the clustering of morpho-syntactic and semantic properties associated with
the Stative and the Eventive paradigms.

Table 3. Formal parameters of the Eventive-Stative contrast in Older Egyptian

TYPE OF pro- ASPECT WORD SUBJECT VERB

AGREEMENT DROP ORDER POSITIONS RAISING

EVENTIVE Pronominal No Events VSO Spec, vP V^T

PARADIGM enclisis Spec, TP V —> F
V ^ C

STATIVE Subject-verb Yes States svo Spec, V — AGRS

PARADIGM agreement AgrsP

The V S O clausal pattern exhibits syntactic variation without morphological

variation. The availability of more than one clausal position for the licens-
ing of the subject can be related to the optional activation of the EPP.
Moreover, Eventive-inflected verbs raise and adjoin to different functional
heads in the inflectional and the complementizer domain, including T, F,
and C. The absence of syntactic variation in Stative clauses is due to the
active status of the EPP, forcing the merging of the subject into the speci-
fier of AGRSP and the absence of attracting functional heads in the left
periphery that trigger Verb Second fronting processes. Older Egyptian
therefore provides another example of the complexities of word order phe-
nomena in a predominantly verb-initial language.
 Word order correlations in Older Egyptian 89

Notes

1. Older Egyptian represents the earliest stage of the Ancient Egyptian-Coptic

language, which is the oldest written language of the Afro-Asiatic phylum; it
covers a period of more than six hundred years (ca. 2650-1990 BCE). Despite
its internal differentiation into Old and Early Middle Egyptian, there is good
reason to believe that Older Egyptian constitutes a linguistic continuum, with
only minor structural changes in the course of generation-to-generation trans-
mission of the language. Loprieno (1995) provides an up-to-date overview of
Ancient Egyptian language history.
2. The following abbreviations are used in the glosses. (Glosses are given in paren-
theses for morphemes that have no surface-segmental shape): 1 'first person';
2 'second person'; 3 'third person'; AUX 'auxiliary verb'; COMP 'subordinating
complementizer'; COMPRA;1, 'relative complementizer'; D 'dual'; DEM 'demon-
strative article'; EMPH 'emphatic particle'; EV 'Eventive inflection' (in sub-
script), F 'feminine gender', FM 'focus marker'; GER 'gerund'; LINK 'linkage
marker'; IMP 'imperative'; IMPERF 'imperfective aspect'; INF 'infinitive';
INTENS/REPET 'intensive/repetitive stem pattern', Μ 'masculine gender'; NEG
'negative scope marker'; NOM 'nominalizing suffix'; Ρ 'plural'; PASSL 'passive
Γ, PASS2 'passive 2', PCL 'particle'; PERF 'Perfect'; PROS 'prospective
tense/aspect'; PTCP 'participle'; S 'singular', STAT 'Stative inflection' (in sub-
script); SUBJ 'subjunctive'; TERM 'Terminative'. Epigraphic conventions: [...]
passage is heavily damaged and cannot be reconstructed, ( . . . ) left out for the
sake of brevity, < . . . > one or more characters have been erroneously omitted
by the ancient scribe. NN stands for the proper name of the owner of a Middle
Kingdom coffin.
3. Some of the semantic oppositions advanced in more elaborate taxonomical
systems are grammaticalised in the so-called Benoni and the progressive con-
struction. The Benoni construction has an equative format consisting of a
nominal or pronominal subject and an active or passive participle functioning
as a nominal predicate. In this construction, predicate fronting is obligatory,
the result being predicate-initial order. Semantically, the Benoni construction
describes individual-level predication, which attributes permanent properties to
individuals (Kratzer 1989).
(i) THE BENONI CONSTRUCTION
rx-w(j) cw tr cf'd mr-r(-w)-t
learn(-PTCP:SM)-PCL you:SM PCL say(-lNF) love-lMPERF-PASSl-PTCP:SF
Izzj r χι nb
Izezi than thing every
'How much you understand to say what is beloved by (King) Izezi more
than everything!' {Urkunden 11 63: 4)
90 Chris Η. Reintges

The progressive is a locative construction formed with an infinitival verb in a

small clause configuration and describes dynamic situations that have not
reached their culmination at the time of the utterance,
(ii) THE PROGRESSIVE CONSTRUCTION
jw t? x?st fir mwt m-ΐ fiqr
AUX DEM:SF f o r e i g n . c o u n t r y at die(-iNF) f r o m - a r m hunger
'This foreign country is dying from starvation.'
(Semnah Dispatches 4 x+10)
The progressive and the Benoni construction fall outside the scope of the pre-
sent study, since they are derived from locative and nominal predication, re-
spectively (see Reintges 1997: 96ff for a more detailed discussion).
4. See Mchombo (1983) and Dubinsky & Simango (1996) for an unaccusative
analysis of Chichewa (Bantu) statives.
5. See Klaiman (1991: 69-82) for comparable phenomena in Tamil, which are
analysed in terms of active-middle voice alternations.
6. A different stance is taken by Loprieno (1995: 76), who argues that the verbal
root itself specifies 'a further temporal dimension, called Aktionsart'. The in-
ternal semantic structure of the lexeme is said to 'bear heavily on the spectrum
of semantically acceptable combinations for each root, restricting the number
of choices by the speaker'. Since there is no clear-cut evidence for lexical re-
strictions on either the Stative or the Eventive conjugation, we may plausibly
assume that the underlying verbal root is unspecified for the event-state di-
chotomy. This raises an important question as to what extent Vendler's (1967)
classical typology is applicable to the Older Egyptian verbal lexicon.
7. The minimal differences between pronouns and grammatical agreement mark-
ers may have a diachronic source insofar as agreement systems evolve histori-
cally from the incorporation of pronouns into verbal stems (see Givon 1976;
Bresnan & Mchombo 1987; Lehmann 1988; and much related research).
8. By 'neutral aspect' I mean the absence of a morphologically marked aspectual
category on the verb. In this case, the aspectual interpretation of the verb is
determined contextually by means of adverbial modification. See Reintges
(1997: 104-109) for some discussion on the ternary distinction between imper-
fective, perfective, and neutral aspect in Older Egyptian.
9. Since imperfective aspect is expressed by means of reduplicative copying, a
number of idiosyncratic factors come into play. Firstly, the verbal lexicon
comprises a considerable number of roots that simply lack a reduplicated alter-
nant. Secondly, those roots that undergo partial or full morpheme reduplication
appear in three derivational classes, viz. (i) the imperfective (e.g. mr-r 'con-
tinue to love, love a lot' < mr(.i) 'to love'), (ii) the intensive-repetitive (e.g.
wn-wn 'to hasten to and fro' < wn 'to hurry'), and (iii) the future passive (e.g.
jwT-r 'will be conceived' < jwr 'to conceive'). The class membership of a
given reduplicant is also not predictable from the morphological properties or
 Word order correlations in Older Egyptian 91

the lexical-semantic template of the underlying root. Thirdly, a given root can
appear in only one reduplicative pattern. Reintges (2003) conjectures that the
non-cyclicity of reduplicative copying has its basis in some economy principle
preventing the generation of strings of two adjacent copies of the root mor-
pheme, which would leave room for too many parsing options.
10. There are some remarkable structural differences in the synthetic paradigms of
Older Egyptian and Modern Celtic languages. First, in contradistinction to
Modern Irish, separate affixes do exist for every person, gender and number
distinction in Older Egyptian verbal paradigms. Second, inflected prepositions
in Irish, unlike their Older Egyptian counterparts, exhibit a high degree of
morpho-phonological irregularity (McCloskey & Hale 1984: 506). Third, in
Breton, it is possible to construe a subject clitic with an analytic verb form, as
in (ia), or to attach a subject clitic to a co-referential synthetic verb form, as in
(ib). In line with Borsely & Stephens (1989), one might analyse (ia) as an in-
stance of a topicalisation construction in which the analytic form agrees with
the trace left behind by subject extraction. In (ib), we seem to be dealing with a
special type of clitic-doubling, where the second (clitic) pronoun is added to
the synthetic verb form for contrastive purposes. Contrary to what is said in
Stump (1984: 333-7), the absence of resumptive pronominalisation in the con-
text of direct object fronting does not provide compelling evidence against the
pronoun incorporation analysis of synthetic verbal paradigms, but rather shows
that Breton displays the standard subject-object asymmetries.

(i) BRETON SUBJECT-VERB AGREEMENT

a. Analytic form + subject clitic
(subject topicalisation and agreement with a gap)
int a lenn (*lennont) le vrioit
they PCL read (read-3p) books
'They read books.' (Stump 1984:291 (2f))
b. Synthetic form + subject clitic (clitic doubling)
levrioü a lennan-me (*lenn-me)
books PCL read-ls-clitic.lS (*read-clitic.ls)
Ί read books.' (Stump 1984: 302 (27))
See sections 2.2.3 and 2.2.4 of the main text for the selection of analytic forms
in the context of w/z-subject extraction. The clitic doubling option (ib) is, how-
ever, systematically absent in our documentation of Older Egyptian.
11. The reviewer suggests to account for the difference between Eventive- and
Stative-inflected verb forms by associating the Eventive and Stative conjuga-
tion pattern with distinct ν heads. Under an analysis along these lines, an
Eventive ν head would assign an agent theta role, while a Stative ν head could
not do so. This analysis would, however, not capture the agreement asymme-
tries between Eventive and Stative clauses: Stative SVO clauses instantiate a
specifier-head configuration with agreement morphology being morphologically
92 Chris Η. Reintges

realised on the finite verb, whereas Eventive VSO clauses have no agreement
at all, but distinct tense and aspect projections. In short, the difference between
Eventive and Stative clauses is due to the functional superstructure above the
v P domain.
12. For Chamorro (Mariana Island, Western Austronesian), Chung (1998, chapter
4) presents detailed arguments to show that the pragmatically unmarked VSO
order cannot be derived from SVO order, but rather derives from VOS order
via lowering of the subject and adjunction to the right of the verbal predicate.
This generally shows that verb-initial languages are typologically as diverse as
subject-initial ones.
13. To account for the agreement mismatches between VSO and SVO clauses in
Modern Standard Arabic, Mohammad (1990) and Ouhalla (1994) propose that
the postverbal subject of VSO sentences is located in the VP-internal subject
position, while the structural subject position [Spec, IP] is occupied by a pho-
nologically null expletive pronoun that induces impoverished agreement in-
flection. At LF the null expletive is replaced by its associate, the subject DP in
the thematic subject position [Spec,VP], As pointed out by Alexiadou &
Anagnostopoulou (1998) and Doron (2000, this vol.), the /7ro exp |-VSO analysis
is intrinsically problematic, since VSO clauses in Modern Standard Arabic and
Modern Hebrew do not show the indefiniteness restriction. This criticism is
not entirely justified, however, as defmiteness effects are systematically absent
in Arabic VSO structures, even when the expletive pronoun is lexicalised. The
spell-out of the expletive pronoun hu 'it' is obligatory in the context of the
subordinating finite complementizer ?anna 'that'.
(i) OVERT EXPLETIVE CONSTRUCTIONS WITH DEFINITE DP SUBJECTS
a. öann-a [CP ?anna-hu wasal-a ?at-tulaabu ]
think-PERF.3SM that-ζ'ί arrive-PERF.3SM the-male.student.PM.NOM
b. Sann-a [c? ?anna-hu wasal-at ?at-taalib-aat
think-PERF.3SM that-it arrive(-PERF)-3sf the-female.students-PF.NOM
'He thought that the male/female students had arrived.'
In ?anna 'that' complement clauses, the finite verb agrees in person and gender
with the postverbal DP subject rather than the expletive pronoun singular mascu-
line hu 'it'. This is particularly clear in the context of feminine DPs, as in (ib)
above. According to Ouhalla (1994: 70, fn.2), the agreement of the embedded
verb with the postverbal subject is only apparent. His main argument is that the
third person singular masculine form of the finite verb can also appear before
feminine subjects. This agreement pattern is instantiated in Classical Arabic
(Fischer 1972: 165, §356), but judgements about its grammaticality status in
Modern Standard Arabic vary. Moreover, if gender agreement with the post-
verbal subject were, indeed, only apparent, it remains to be explained why the
selection of the third person singular feminine form of the verb in the context
of masculine subjects results in ungrammaticality. See Aoun, Benmamoun &
 Word order correlations in Older Egyptian 93

Sportiche (1994) for additional arguments against the Spec-Head Agreement

with expletive analysis.
14. It looks as if the negative adverb w was originally part of a syntactically dis-
continuous negation η ... w, reminiscent to the ne ... pas negation in Standard
French.
(i) DISCONTINUOUS NEGATION η ... W
<
η Xtm-k w i?-w(j) pt
NEG c l o s e - 2 S M E V NOT gate-DM h e a v e n
η xsf-k w xsf-w(j)-s
NEG l 0 c k - 2 S M E V NOT bolt-DM-3SF
d3r pr-t k? Pjpj Nfr-k?-RΫ jr pt
until come-TERMEV soul Pepi Nefer-ka-Re to heaven
'You should not close the two gates of heaven, you should not lock its two
bolts until the soul of (King) Pepi Nefer-ka-Re comes to heaven.'
(Pyramid Texts/Pepi II 1055+44)
Notice, however, that (i) is the only example attested for the discontinuous
negation η ... w. In all the other examples, the negative adverb w does not op-
erate under the scope of the clause-initial negation marker η and had become
the sole carrier of negative force.
15. Edel (1955/64: 431, §858a) analyses the kwas a particle or interjection, which
would, however, be attested in the context of this spell only. Since the adhorta-
tively used Stative j.dJd-t(j) 'you should say' is specified for the addressee-
oriented second person singular, it is tempting to regard kw as the original,
non-palatalized form of the co-referential clitic pronoun cw 'you (man)'.
Edel's argument against a pronominal analysis of the form kw rests entirely on
the assumption that clitics either precede or follow their phonological hosts.
Examples like (41a) of the main texts are indicative for the more flexible syn-
tax of weak pronouns.
17. Cf. Chung (2003), where the same tests support a purely prosodic analysis of
the weak pronoun placement in Chamorro.
18. See Reintges (1997: 262-4) for some discussion of the -wl-jj alternation in
apophonic passive Is in the context of the first person singular pronoun -j.

References

Andersen, S. R.
1982 Where's morphology? Linguistic Inquiry 13: 571-612.
Adger, D.
1997 VSO and Weak Pronouns in Celtic. Canadian Journal of Linguistics
42: 9-31.
94 Chris Η. Reintges

Alexiadou, A. and Ε. Anagnostopoulou

1998 Parametrizing AGR: Word Order, V-movement and Epp-checking.
Natural Language and Linguistic Theory 16: 491-539.
2001 The Subject-in-Situ Generalization and the Role of Case in Driving
Computations. Linguistic Inquiry 32: 193-231.
Alexopoulou, Th., E. Doron and C. Hey cock
2003 Broad Subjects and Clitic Left Dislocation. In Peripheries, D. Adger,
C. de Cat and G. Tsoulas (eds.). Dordrecht: Kluwer.
Allen, J. P.
1984 The Inflection of the Verb in the Pyramid Texts. Bibliotheca Aegyptia
2. Malibu: Undena Publications.
Aoun, J., E. Benmamoun and D. Sportiche
1994 Agreement, Word Order, and Conjunction in Some Varieties of Arabic.
Linguistic Inquiry 25: 195-220.
Bach, E.
1986 The Algebra of Events. Linguistics and Philosophy 9: 5-16.
Baker, M. C.
1985 The Mirror Principle and Morphosyntactic Explanation. Linguistic
Inquiry 16: 373-416.
1988 Incorporation: A Theory of Grammatical Function Changing. Chicago:
The University of Chicago Press.
Baker, M. C. and K. Hale
1990 Relativized Minimality and Pronoun Incorporation. Linguistic Inquiry
21: 289-297.
Baker, M. C., K. Johnson and I. G. Roberts
1989 Passive Arguments Raised. Linguistic Inquiry 20: 219-251.
Benmamoun, E.
2000 The Feature Structure of Functional Categories: A Comparative Study
of Arabic Dialects. Oxford Studies in Comparative Syntax. Oxford/
New York: Oxford University Press.
Berman, R. A.
1981 Dative Marking of the Affected in Modern Hebrew. In Theoretical
Issues in the Grammar of Semitic Languages, H. Borer and J. Aoun
(eds.). MIT Working Papers in Linguistics 3: 150-179. Cambridge,
Mass.
Besten den, H.
1983 On the Interaction of Root Transformations and Lexical Deletive
Rules. On the Formal Syntax of the Westgermania, W. Abraham (ed.),
47-137. Amsterdam/ Philadephia: John Benjamins.
Bobalijk, J. P. and D. Jonas
1996 Subject Positions and the Roles of TP. Linguistic Inquiry 27: 195-236.
 Word order correlations in Older Egyptian 95

Borer, H.
1981 Comments on Pro-Drop Phenomena. MIT Working Papers in Lin-
guistics 3: 127-149. Cambridge, Mass.
1986 I-Subjects. Linguistic Inquiry 17: 375-416.
1995 The Ups and Downs of Hebrew Verb Movement. Natural Language
and Linguistic Theory 13: 527-606.
Borer, H. and Y. Grodzinsky
1986 Syntactic Cliticization and Lexical Cliticization: The Case of Hebrew
Dative Clitics. In The Syntax of Pronominal Clitics, H. Borer (ed.),
175-217. Syntax and Semantics 19. New York: Academic Press.
Borsley, R. D. and J. Stephens
1989 Agreement and the Positions of Subjects in Breton. Natural Language
and Linguistic Theory 7: 407^127.
Bowers, J.
2002 Transitivity. Linguistic Inquiry 33: 183-224.
Bresnan, J. and S. A. Mchombo
1987 Topic, Pronoun, and Agreement in Chichewa. Language 63: 741-782.
Carnie, Α., Η. Harley and E. Pyatt
2000 VSO Order as Raising Out of IP? Some Evidence from Old Irish. In
The Syntax of Verb-initial languages, A. Carnie and E. Guilfoyle
(eds.), 39-59. Oxford Studies in Compartive Syntax. Oxford Univer-
sity Press.
Carstens, V.
2003 Rethinking Complementizer Agreement: Agree with a Case-Checked
Goal. Linguistic Inquiry 34: 393-412.
Chomsky, N.
1981 Lectures on Government and Binding. Dordrecht: Foris Publicatons.
1996 The Minimalist Program. Current Studies in Linguistics 28. Cam-
bridge, Mass.: MIT Press.
2000. Minimalist Inquiries: The Framework. In Step by Step: Essays
on Minimalist Syntax in Honour of Howard Lasnik, R. Martin, D.
Michaels and J. Uriagereka (eds.), 89-155. Cambridge, Mass.: MIT
Press.
2001 Derivation by Phase. In Ken Hale: A Life in Language, M. Kenstowicz
(ed.), 1-52. Current Studies in Linguistics 36. Cambridge: MIT Press.
Chung, S.
2003 The Syntax and Prosody of Weak Pronouns in Chamorro. Linguistic
Inquiry 34: 547-599.
Chung, S. and J. McCloskey
1987 Government, Barriers, and Small Clauses in Modern Irish. Linguistic
Inquiry 18: 173-237.
96 Chris Η. Reintges

Cinque, G.
1990 Types of A'-Dependencies. Cambridge, Mass.: MIT Press.
1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective.
Oxford Studies in Comparative Syntax. Oxford/New York: Oxford
University Press.
Collins, C.
2002 Multiple Verb Movement in =j= Hoan. Linguistic Inquiry 33: 1-29.
Collins, C. and H. Thräinsson
1996 VP-Internal Structure and Object Shift in Icelandic. Linguistic In-
quiry 27: 3 9 1 ^ 4 4 .
Comrie, B.
1976 Aspect - An Introduction to the Study of Verbal Aspect and Related
Problems. Cambridge: Cambridge University Press.
Craenenbroeck, J. van and M. van Koppen
2002 The Locality of Agreement and the CP-Domain. Paper Presented at
GLOW 2002 Conference in Amsterdam, The Netherlands.
Diesing, M.
1990 Verb Movement and the Subject Position in Yiddish. Natural Lan-
guage and Linguistic Theory 8: 41-79.
Doron, E.
1988 On the Complementarity of Subject and Subject-Verb Agreement. In
Agreement in Natural Language: Approaches, Theories, Descriptions,
M. Barlow and C. A. Ferguson (eds.), 201-218. Papers Presented at
a Conference Held at Stanford University in October 1984. Stanford:
CSLI.
1996 The Predicate in Arabic. In Studies in Afroasiatic Grammar, Papers
from the Second Conference on Afroasiatic Languages Sophia Anti-
polis 1994, J. Lecarme, J. Lowenstamm and U. Shlonsky (eds.), 7 7 -
104. The Hague: Holland Academic Graphics.
2000 VSO and Left-conjunct Agreement: Biblical Hebrew vs. Modern
Hebrew. In The Syntax of Verb-initial languages, A. Carnie and E.
Guilfoyle (eds.), 75-96. Oxford Studies in Compartive Syntax. Ox-
ford University Press. [Reprinted in this volume.]
2003 Agency and Voice: The Semantics of the Semitic Template. Natural
Language Semantics 11: 1-67.
Doron, E. and C. Hey cock
1999 Filling and Licensing Multiple Specifiers. In Specifiers - Minimal-
ist Approaches, D. Adger, S. Pintzuk, B. Plunkett, and G. Tsoulas
(eds.), 69-89. Oxford: Oxford University Press.
Dubinsky, S. and S. R. Simango
1996 Passive and Stative in Chichewa: Evidence for Modular Distinctions
in Grammar. Language 72: 749-781.
 Word order correlations in Older Egyptian 97

Edel, Ε.
1955/64 Altägyptische Grammatik. Analecta Orientalia 34/39, Rome: Ponti-
ficial Biblical Institute.
1959 Beiträge zur ägyptischen Grammatik 1. Zum angeblich geminierenden
Pseudopartizip der Verben III. infirmae. Zeitschrift für Ägyptische
Sprache und Altertumskunde 84: 105-108.
Emonds, J. E.
1985 A Unified Theory of Syntactic Categories. Studies in Generative
Grammar 19. Dordrecht: Foris Publications.
Fassi Fehri, A.
1988 Agreement in Arabic, Binding and Coherence. In Agreement in Natural
Language: Approaches, Theories, Descriptions, M. Barlow and C. A.
Ferguson (eds.), 107-158. Papers Presented at a Conference held at
Stanford University in October 1984, Stanford: CSLI.
1993 Issues in the Structure of Arabic Clauses and Words. Dordrecht:
Kluwer.
Fischer, W.
1972 Grammatik des Klassischen Arabisch. Porta Linguarum Orientalium
NS 11. Wiesbaden: Harrasowitz.
Franks, S.
1990 The Status o f N u l l Expletives. Lingua 81: 1-24.
Fukui, N. and M. Speas
1986 Specifiers and Projection. In MIT Working Papers in Linguistics 8,
Ν. Fukui, Τ. R. Rapoport and E. Sagey (eds.), 128-172.
Givon, T.
1976 Topic, Pronoun, and Grammatical Agreement. In Subject and Topic,
C.N. Li (ed.), 149-188. New York/London: Academic Press.
Greenberg, J. H.
1966 Universals of Language. Cambridge, Mass.: MIT Press.
Grimshaw, J.
1990 Argument Structure. Linguistic Inquiry Monograph 18. Cambridge,
Massachusetts: MIT Press.
Guilfoyle, Ε., H. Hung and L. Travis
1992 Spec of IP and Spec of VP: Two Subjects in Austronesian Lan-
guages. Natural Language and Linguistic Theory 10: 375-414.
Hale, K. and S. J. Keyser
1993 On Argument Structure and the Lexical Expression of Syntactic
Relations. In The View from Building 20: Essays in Linguistics in
Honour of Sylvain Bromberger, K. Hale and S. J. Keyser (eds.), 5 3 -
109. Current Studies in Linguistics 24. Cambridge, Mass.: MIT Press.
1997 On the Complex Nature of Simle Predicators. In Complex Predicates,
A. Alsina, Joan Bresnan, and Peter Sells (eds.), 29-65. CSLI Lecture
Notes 64. CSLI Publications: Stanford.
98 Chris Η. Reintges

Hale, Κ. and S. J. Keyser

2001 Prolegomenon to a Theory of Argument Structure. Linguistic Inquiry
Monographs 39, Cambridge, Mass.: MIT Press.
Hoekstra, T. and R. Mulder
1990 Unergatives As Copular Verbs: Locational and Existential Predication.
The Linguistic Review 7: 1-79.
Holes, C.
1995 Modem Arabic. Structures, Functions and Varieties. London/New
York: Longman.
Huang, J. C.-T.
1984 On the Distribution and Reference of Empty Pronouns. Linguistic
Inquiry 15: 531-574.
Jaeggli, O. A. and K. J. Safir
1989 The Null Subject Parameter and Parametric Theory. In The Null Sub-
ject Parameter, O. A. Jaeggli and K. J. Safir (eds.), 1-44. Dordrecht:
Kluwer.
Kammerzell, F.
1993 Die altägyptische Negation w: Versuch einer Rekonstruction. Lingua
Aegyptia 3: 151-175.
Kamp, Η. and U. Reyle
1993 From Discourse to Logic - Introduction to Modeltheoretic Semantics
of Natural Language, Formal Logic and Discourse Representation
Theory. Studies in Linguistics and Philosophy 42, Dordrecht: Klu-
wer.
Kayne, R.
1989 Null Subjects and Clitic Climbing. In The Null Subject Parameter,
O. A. Jaeggli and K. J. Safir (eds.), 239-261. Dordrecht: Kluwer.
Kiss, Κ. E.
1998 Identificational focus versus information focus. Language 74: 245-273.
Klaiman, Μ. H.
1991 Grammatical Voice. Cambridge Studies in Linguistics 59. Cambridge:
Cambridge University Press.
Koopman, H. and D. Sportiche
1991 The Position of Subjects. Lingua 85: 211-258.
Kratzer, A.
1994 The Event Argument and the Semantics of Voice. Ms., University of
Massachusetts at Amherst.
2000 Building Statives. In Proceedings of the 26th Annual Meeting of the
Berkeley Linguistic Society, L. J. Conathan et al. (eds.). Berkeley:
Berkeley Linguistic Society.
Kuroda, S.-Y.
1988 Whether we agree or not: A Comparative Syntax of English and
Japanese. Linguisticae Investigation's 12: 1-47.
 Word order correlations in Older Egyptian 99

Larson, R.
1988 On the Double Object Construction. Linguistic Inquiry 19: 335-391.
Lehmann, C.
1988 On the Function of Agreement. In Agreement in Natural Language:
Approaches, Theories, Descriptions, M. Barlow and C. A. Ferguson
(eds.), 55-65. Papers Presented at a Conference held at Stanford
University in October 1984, Stanford: CSLI.
Levin, B. and M. Rappaport-Hovav
1986 The Formation of Adjectival Passives. Linguistic Inquiry 17, 623-661.
1995 Unaccusativity At the Syntax-Lexical Semantics Interface. Linguistic
Inquiry Monographs 26. Cambridge, Mass.: MIT Press.
Loprieno, A.
1995 Ancient Egyptian - A Linguistic Introduction. Cambridge: Cambridge
University Press.
Marantz, A.
1997 No Escape From Syntax: Don't Try Morphological Analysis in the
Privacy of Your Own Lexicon. In Proceedings of the 2Γ' Annual
Penn Linguistics Colloquium, University of Pennsylvania Working
Papers in Linguistics 4.2, A. Dimitriadis, L. Siegel, C. Surek-Clark
and A. Williams (eds.), 201-225. University of Pennsylvania, Phila-
delphia.
2000 The Universality of Root and Pattern Morphology. Paper Presented
at the 5th Conference on Afroasiatic Linguistics (June 2000), Paris.
McCloskey, J.
1991 Clause Structure, Ellipsis and Proper Government in Irish. Lingua
85: 259-302.
1996 On the Scope of Verb Movement in Irish. Natural Language and
Linguistic Theory 14: 47-104.
1997 Subjecthood and Subject Positions. In Elements of Grammar,
L. Haegeman (ed.), 197-235. Dordrecht: Kluwer.
2001 The Distribution of Subject Properties in Irish. In Objects and Others
Subjects: Grammatical Functions, Functional Categories, and Confi-
gurationality, W. D. Stanley and St. Dubinsky (eds.), 157-192. Studies
in Natural Language and Linguistic Theory 52. Dordrecht: Kluwer.
McCloskey, J. and K. Hale
1984 On the Syntax of Person-Number Inflection in Modern Irish. Natural
Language and Linguistic Theory 1: 487-533.
Mchombo, S. S.
1993 A Formal Analysis of the Stative Construction in Bantu. Journal of
African Languages and Linguistics 14: 5-28.
100 Chris Η. Reintges

Mohammad, Μ. Α.
1990 The Problem of Subject-Verb Agreement in Arabic: Towards a Solu-
tion. In Perspectives on Arabic Linguistics I, Papers from the First
Annual Symposium on Arabic Linguistics, M. Eid. Current (ed.),
95-125. Issues in Linguistic Theory 63. Amsterdam/Philadelphia:
Benjamins.
Mourelatos, A. P. D.
1981 Events, Processes and States. In Tense and Aspect, P. Tedechi and A.
Zaenen, 191-212. Syntax and Semantics 14. New York: Academic
Press.
Nash, L. and A. Rouveret
2003 Cliticization as Unselective Attract. Ms., Universite Paris-7.
Nedjalkov, V. P. and S. J. Jaxontov
1988 The Typology of Resultative Constructions. In Typology of Resulta-
tive Constructions, V. P. Nedjalkov (ed.), 3-62. Typological Studies
in Language 12, 1988. Amsterdam/ Philadelphia: Benjamins.
Ouhalla, J.
1994 Verb Movement and Word Order in Arabic. In Verb Movement,
D. Lightfoot and N. Hornstein (eds.), 41-72. Cambridge: Cambridge
University Press.
Parsons, T.
1990 Events in the Semantics of English: A Study in Subatomic Semantics.
Current Studies in Linguistics 19. Cambridge, Mass.: MIT Press.
Perlmutter, D. M.
1978 Impersonal Passives and the Unaccusative Hypothesis. In Proceed-
ings of the Fourth Annual Meeting of the Berkeley Linguistic Society,
J. Jaeger (ed.). Berkeley Linguistic Society: University of California,
Berkeley.
1983 Personal and Impersonal Constructions. Natural Language and Lin-
guistic Theory 1: 141-200.
Perlmutter, D. M. and P. M. Postal
1983 Toward a Universal Characterisation of Passivisation. In Studies in
Relational Grammar 1, D. M. Perlmutter (ed.), 3-29. Chicago: The
University of Chicago Press.
Pollock, J.-Y.
1989 Verb Movement, Universal Grammar, and the Structure of IP. Lin-
guistic Inquiry 20: 365-424.
Reintges, C. H.
1996 Egyptian Root-and-Pattern Morphology. Lingua Aegyptia 4. Proceed-
ings of the International Conference on Egyptian Grammar. Cross-
roads III (Yale 1994), 213-244.
1997 Passive Voice in Older Egyptian: A Morpho-Syntactic Study. HIL
Dissertations 28. The Hague: Holland Academic Graphics.
 Word order correlations in Older Egyptian 101

1998 Mapping Information Structure to Syntactic Structure: One Syntax

for Jn. Revue d'Egyptologie 49: 195-220.
2000 The Licensing of Gaps and Resumptive Pronouns in Older Egyptian
Relatives. In Research in Afroasiatic Grammar, Papers from the
Third Conference on Afroasiatic Languages, Sophia Antipolis (1996),
J. Lecarme, J. Lowenstamm and U. Shlonsky (eds.), 243-262.
Current Issues in Linguistic Theory 202. Amsterdam/Philadelphia:
John Benjamins.
2003 The Reduplicative Passive in Older Egyptian: Description and Typo-
logy. In Afrasian: Selected Comparative-Historical Studies in Memory
of Igor M. Diakonoff M. L. Bender, D. Appelyard, M. Lamberti and
G. Takäcs (eds.), 175-186. LINCOM Studies in Afro-Asiatic Lin-
guistics 14. Lingcom: Munich.
Ritter, Ε.
1995 On the Syntactic Category of Pronouns and Agreement. Natural
Language and Linguistic Theory 13: 405-443.
Rizzi, L.
1982 Issues in Italian Syntax. Dordrecht: Foris Publications.
1986 Null Objects in Italian and the Theory of Pro. Linguistic Inquiry 17:
501-557.
1990 Relativized Minimality. Linguistic Inquiry Monograph 16. Cam-
bridge, Mass.: MIT Press.
1997 The Fine Structure of the Left Periphery. In Elements of Grammar,
L. Haegeman (ed.), 281-337. Dordrecht: Kluwer.
Roberts, I. G.
1993 Verbs and Diachronic Syntax: A Comparative History of English
and French. Dordrecht: Kluwer.
Rohrbacher, Β. W.
1994 The Germanic VO Languages and the Full Paradigm: A Theory of V to
I Raising. PhD dissertation. University of Massachusetts at Amherst.
Rothstein, S.
1983 The Syntactic Form of Predication. PhD dissertation, Massachusetts
Institute of Technology, distributed 1985. Indiana University Lin-
guistic Club: Bloomington (Indiana).
2004 Structuring Events: A Study in the Semantics of Lexical Aspect.
Explorations in Semantics 2. Blackwell: Oxford.
Schenkel, W.
1997 Tübinger Einführung in die klassisch-ägyptische Sprache und Schrift.
Tübingen: pagina GmbH.
Shisha-Halevy, A.
1986 (j)rf in the Coffin Texts: A Functional Tableau. Journal of the Ameri-
can Oriental Society 106: 641-658.
102 Chris Η. Reintges

Shlonsky, U.
1990 Pro in Hebrew Subject Inversion. Linguistic Inquiry 21: 263-275.
1997 Clause Structure and Word Order in Hebrew and Arabic: An Essay
in Comparative Semitic Syntax. Oxford Studies in Comparative Syn-
tax 11. Oxford/New York: Oxford University Press.
Smith, C. S.
1991 The Parameter of Aspect. Studies in Linguistics and Philosophy 43.
Dordrecht: Kluwer.
Sproat, R.
19B5 Welsh Syntax and VSO Structure. Natural Language and Linguistic
Theory 3: 173-216.
Stowell, T.
1989 Raising in Irish and the Projection Principle. Natural Language and
Linguistic Theory 7: 317-359.
Stump, G. T.
1984 Agreement vs. Incorporation in Breton. Natural Language and Lin-
guistic Theory!·. 289-348.
1989 Further Remarks on Breton Agreement. Natural Language and Lin-
guistic Theory 7: 429-471.
2001 Inflectional Morphology: A Theory of Paradigm Structure. Cam-
bridge Studies in Linguistics 93: 57-149.
Talmy, L.
1985 Lexicalization Patterns: Semantic Structure in Lexical Forms. In
Language Typology and Syntactic Description, vol. III. Grammatical
Categories and the Lexicon, T. Shopen (ed.), 57-149, Cambridge:
Cambridge University Press.
Tenny, C. L.
1994 Aspectual Roles and the Syntax-Semantics Interface. Studies in Lin-
guistic and Philosophy 52. Dordrecht.
Uriagereka, J.
1995 Aspects of the Syntax of Clitic Placement in Western Romance.
Linguistic Inquiry 26: 79-123.
Vendler, Z.
1967 Linguistics in Philosophy. Ithaca (New York): Cornell University
Press.
Vikner, S.
1994 Finite Verb Movement in Scandinavian Embedded Clauses. In Verb
Movement, D. Lightfoot and N. Hornstein (eds.), 117-147. Cam-
bridge: Cambridge University Press.
1995 Verb Movement and Expletive Subjects in the Germanic Languages.
Oxford Studies in Comparative Syntax. Oxford: Oxford University
Press.
 Word order correlations in Older Egyptian 103

Wasow, Th.
1977 Transformation and the Lexicon. In Formal Syntax, P. Culicover, Th.
Wasow and J. Bresnan (eds.), 327-360. New York: Academic Press.
Zagona, K.
2002 The Syntax of Spanish. Cambridge Syntax Guides. Cambridge: Cam-
bridge University Press.
Zubizaretta, M. L.
1998 Prosody, Focus, and Word Order. Linguistic Inquiry Monographs
33. Cambridge, Massachusetts: MIT Press.
The nominal cleft construction in Coptic Egyptian

Chris H. Reintges, Aniko Liptdk and

Lisa Lai Shen Cheng

1. Introduction

Cleft sentences belong to a larger family of focusing constructions. In a

language like English, clefts are formed by dividing a simple clause into
two parts, namely an initial focus (the cleft constituent) and a back-
grounded proposition, which is expressed by a relative clause (see Huddle-
ston & Pullum 2002: 1414-20; Lambrecht 2001).

(1) English it-clefts

a. It was a red wool sweater that I bought.
b. It was the wording of the question that confused me.

This paper examines the corresponding construction in Coptic Egyptian,

which represents the latest developmental stage of the Ancient Egyptian
language. The modern term Coptic is derived from Middle Arabic qubtj,
itself a corruption of the Greek word (ai)gypt(ios) 'Egyptian'. At the roots
of the literary production lies the translation of the Greek Bible into Coptic
Egyptian. The earliest Biblical manuscripts date from about 350 CE, but go
back to the third century. Coptic Egyptian continued to be in use well be-
yond its extinction as a spoken language. 1
Coptic is actually a dialect cluster, consisting of at least six regional
varieties, two of which gained supra-regional importance: Sahidic, the lan-
guage of the whole Nile valley above the Delta, and Bohairic, the language
of the Nile Delta. Prior to the Arabic conquest in 641 CE, Sahidic was the
predominant literary dialect of Coptic. Its supremacy became challenged by
Bohairic Coptic from the 9 th century onwards. By the 11th century, Bohairic
had replaced Sahidic as the official church language and become the sole
representative of Coptic Egyptian, which survived as the liturgical language
of the present-day Coptic Orthodox Church. The language material of this
paper is exclusively drawn from Sahidic Coptic, the main reference dialect. 2
106 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

Turning to the topic of our paper, cleft sentences are very common in the
syntactic patterns of Sahidic Coptic to indicate the focal status of a nominal
argument. As with English clefts, Coptic cleft constructions have a bi-
clausal syntax, yet express a single proposition. The focused DP p-tfoejs
'the Lord' in (2a), the ννΛ-subject nim 'who' in (2b), as well as the contras-
tively stressed pronoun ntof 'HE' in (2c) are all associated with an out-of-
focus relative clause that is introduced by the relative complementiser et
'that'. (Relative gaps are indicated as ' ').3

(2) a. DP-clefts
p-tfoejs gar p(e) [Cp et ο m-metre
DEF:SM-lord PCL PRON:SM CREL(PRES) be as-witness
n-ta-senedis ].
to-DEF:SM: 1 S-conscience
'For (it is) the Lord who is witness to my conscience.'
(V. Pach. 89: 1-2)
b. Wh-clefts
nim tenu p(e) [CP et sorm m-p-meefe ]?
who A D V P R O N : S M CREL ( P R E S ) misguide DO-DEF:SM-crowd
'Who (is it) now that is misleading the crowd?'
(Acts of Andrew and Paul 212: 231)
c. Pronoun clefts
ntof gar pe [cp et shai gn-fatfe nim
HE PCL PRON:SM CREL (PRES)write DO-word every
[e-nere Jesus tfo ommo-u ]]
REL-PRET Jesus say D O - 3 P
'Since (it is) him who wrote down every word that Jesus had said.'
(Pistis Sophia 71: 5-6)

Different from English //-clefts, there is no copular verb in the Coptic coun-
terpart that connects the cleft constituent to the subordinate relative clause.
Instead, a deictic pronoun (glossed as P R O N ) is employed to perform this
linkage function. The deictic pronoun recurs in predicational and identifica-
tional nominal sentences, as shown by the contrast between (3 a) and (3 b)
below.4
 The nominal cleft construction in Coptic Egyptian 107

(3) a. Nominal clefts [DP PRON CP]

an ok gar p(ej [et η uhorn am mo-ton eßol
I PCL PRON:SM CREL (PRES) save DO-2P PCL
hon thlipsis nim ]
from trial every
'For (it is) me who saves you from every trial.'
(Eudoxia 50: 22-23)
b. Identificational nominal sentences (equatives) [DP PRON DP]
anok pe p-fere m-p-nute
I PRON:SM DEF:SM-son of-DEF:SM-god
Ί (am) the son of God.' (Abbatön 240: 13)

The nominal cleft construction has received a considerable amount of atten-

tion in Coptic language studies. As already established in Polotsky (1962),
nominal clefts are not pseudo-clefts in which a free (headless) relative
clause functions as the primary predicate of the entire construction. As of yet,
the structural analogy between cleft and nominal predicational sentences as
well as the different semantic types of focus that are expressed through
clefting have not received a principled explanation. By making use of the
analytical tools made available by generative syntax, we can make the rela-
tionship between the constituents of nominal clefts, their structural position,
and their meaning and function more precise and draw parallels with other
constructions, in particular, nominal predicational sentences and relative
constructions.
The aim of the present study is therefore two-fold: to clarify the descrip-
tive facts and to develop a configurational analysis of Coptic cleft sentences
that captures the relationship between sentence form and information struc-
ture in a straightforward way. The key idea is that clefts have an underlying
small clause configuration headed by the deictic pronoun. As the subject of
the small clause, the cleft constituent underlies strict categorial restrictions:
it must be a nominal category (either a DP or an NP). As a focus phrase, it
cannot remain in-situ within the small clause, but must move to a desig-
nated focus projection in the left periphery of the clause.
The analysis of Coptic nominal clefts pursued in this paper contributes to
the ongoing research on the mapping between discourse properties and left-
peripheral functional superstructure. The Coptic facts provide evidence for
the existence of a type of clefting that involves Α-bar movement. They also
show that focus fronting of the cleft constituent correlates with contrastive
focus interpretation, but may also express new information focus.
108 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

The organization of this paper is as follows. Section 2 is concerned with the

range of focus interpretations that the cleft constituent may assume. Section
3 provides an overview of the main syntactic properties of nominal cleft
sentences, with particular attention for their biclausal structure and agree-
ment facts. Section 4 presents a configurational analysis of Coptic clefts. We
will argue that nominal clefts have an underlying small clause structure,
upon which an articulated left periphery is erected. Cleft formation involves
focus fronting of the clefted DP into the specifier position of a designated
focus phrase. Section 5 summarizes the main results of the paper.

2. Types of focus in nominal clefts

Coptic cleft sentences are typically used for contrastive emphasis with
various degrees of strength. Following Rooth (1992) and subsequent re-
search, we assume that contrastive focus specifies a subset of the given set
of discourse entities for which the backgrounded proposition holds true, but
at the same time implies a set of alternatives for which that proposition
does not hold true. The set of alternatives brought into play by contrastive
focus may be overtly given, as in (4). The most natural interpretation of this
example is one in which the contrast set on which focus operates consists of
the clefted DP A ugustos Kostantinos and the DP complement Dioklctianos
of the complex preposition e-p-ma 'instead' (lit. in the place of).

(4) Contrasted discourse referents

A UQUStQS Kostantinos p-rro n-dikajos pe
Augustus Constantine DEF:SM-king of-righteous PRON:SM
[Cp nt-a p-tfoejs pe.n-nute tunos-f na-n n-rro
REL-PERF DEF:SM-lord DEF:SM. 1 P-god raise-3SM for-lP as-king
e-p-ma m-p-anomos n-at-fipe Dioklctianos
at-DEF:SM-place of-DEF:SM-criminal of-shameless Diocletian
p-rro m-p-tfi-n-kyons ]
DEF:SM-king of-DEF:SM-take-DO-injustice
'(It is) Augustus Constantine, the righteous king, who the Lord, Our
God, has raised as a king for us instead of the shameless criminal
Diocletian, the king of injustice.' (Eudoxia 38: 1-3)

The strongest form of contrast is corrective or counter-assertive focus, used

in contexts where the speaker contradicts a previous utterance. Counter-
 The nominal cleft construction in Coptic Egyptian 109

assertive focus is exemplified by pairs of negative and affirmative clefts, in

which the contrasted foci specify different values for the same proposition,
as shown in (5).

(5) Juxtaposition of negative and affirmative cleft

m- p-rome an p(e) [cp et t f i kßa
NEG DEF:SM-man NOT PRON:SM CREL(PRES) take vengeance
nhet-s hn u-kyont ]
by-3SF in INDEF: S M - w r a t h

η tos ntof p(e) [Cp et kto m-p-rome

SHE ADV PRON:SM CREL(PRES) turn DO-DEF:SM-man
e-p-ma [Cp eta-s-wof e-kons mmau]]
at-DEF:SM-place REL-(PRES)-3SF-want to-wound there
'(It is) not the man who takes vengeance by it (the sword) wrathfully.
Rather (it is) it (the sword) that turns the man to the place where it
wants to wound.' (Shenoute IV 12: 10-12)

In the following set of data, the contrastive focus reading is evident from
the use of dedicated focus particles. The emphatic reflexives mawaa-n
'ourselves' in (6a) and ho 'myself in (6b) below impose an exclusive inter-
pretation on the cleft constituent. It is asserted that none of the alternatives
could provide a value for the open predication contained in the restrictive
relative clause.

(6) Modification of the defied DP by emphatic reflexives

a. anon de mawaa-n p(e) [Cp et loop m-pei-ma ]
WE PCL OWN-LP P R O N : S M CREL ( P R E S ) reside i n - D E M : S M - p l a c e
'(It is) only us who reside here.' (Budge, Martyrd. 221: 2-3)
b. anok ho pe [Cp nt-a-i-r noße ero-k
I SELF-IS PRON:SM REL-PERF-1 S-do sin against-2SM
n-u-sop n-wot ]
in-INDEF:S-time of-single
'(It is) me, however, who sinned against you just once!'
(Acts of Andrew and Paul 200: 87)

Consider next the cleft construction in (7), in which the long discourse
topic DP p-nute nt-a-f-fope mn 'God who has been with (DP)' functions as
the antecedent of the clefted pronoun ntof' he'. The interpretation of this
110 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

example is clearly not contrastive, as can be seen from the presence of the
additive focus particle on 'also, too'.

(7) Non-contrastive clefted pronoun

p-nute [cp nt-a-f-fope mn n-genea [cp nt-a-u-fope
DEF:SM-god REL-PERF-3SM-be with DEF:P-generation REL-PERF-3P-be
e-a-u-wejne etße te.u-mnt-akaireos
REL-PERF-3P-pass.by because.of DEF:SF:3P-NOM-unharmed
mn te.u-pistis e-p-nute ]]
with DEF:SF-faith towards-DEF:SM-god
ntof on p(e) [CP EF na-Jope mn n-genea
HE PCL PRON:SM CREL FUT-be with DEF:P-generation
[CP et neu ]]
CRELCPRES)come
'God who has been with the generations that existed and passed by
because of their unharmed state and their faith in God, (it is) him, too,
who will be with the generations to come.'
(Testament of Isaac 228: 13-16)

As we can see from (8) below, no definiteness restriction is operative on

the cleft constituent in Coptic Egyptian. 5 The cleft sentence in (8a) occurs
in the opening sequence of a new narrative unit. We may therefore plausibly
assume that the indefinite DP u-hoß nte p-nute 'a divine matter' corresponds
to presentational focus, which introduces a new topic into the discourse. In
(8b-c), on the other hand, the indefinite NPs hen-kuwi 'a few (things)' and
u-fere 'a son' are modified by the scalar focus particle emate 'only' and
the degree adjective wot 'single', respectively, and are both presentational
and contrastive foci.

(8) Clefted indefinite DPs

a. u-hoß nte p-nute p(e) [Cp e-ti-fatfe nmme-tn
INDEF-.S of DEF:SM-god PRON:SM REL-(PRES)-lS-speak with-2P
etßect-f ]
about-3SM
'(It is) a divine matter that I am speaking to you about.'
(Eudoxia 60: 24-25)
 The nominal cleft construction in Coptic Egyptian 111

b. pi en e f t f e hen-kuwi emate n(e) [Cp eta-n-na-tfoo-u

ADV if INDEF:P-small PCL PRON:P REL-lP-FUT-say-3P
hn ne.f-katorthoma ]
from DEF:P.3SM-achievement
'Yet even though (it is) only a few (things) that we are going to say
about his achievements.' (Zenobius 201: 13)
c. u-fcre n-wot p(e) [Cp et foop na-i ]
INDEF:S-child of-single P R O N : S M C R E L (PRES)be for-lS
'(It is) a single son that I have.' (Acts of Andrew and Paul 194: 22)

In summary, Coptic nominal clefts are not a purely contrastive focus device,
but allow for a much broader range of semantic focus interpretations (cf.
Doetjes, Rebuschi & Rialand 2004 for similar observation on the French
c 'estXP que/qui 'it/that is XP who' construction).6

3. Syntactic properties of Coptic clefts

This section reviews the main structural properties of Coptic nominal clefts.
We will show that clefts minimally involve two clausal projections: the
matrix clause is a nominal predicational sentence into which a relative clause
is embedded: [CPi DP PRON [CP2 RELATIVE CLAUSE]]. In this structure, the
relative clause lacks nominal properties, i.e. it is not a free relative clause.
We will also look at the categorial restriction on the cleft constituent and
the agreement behaviour of the deictic pronoun.

3.1. The nature of the relative clause

In the literature on nominal cleft constructions, the categorial status of the

relative clause has been an issue of major importance. For Coptic nominal
clefts, it has been argued in the descriptive literature that the relative clause
is not a free relative with nominal properties (e.g. Polotsky 1962; Layton
2000). An analysis along these lines has, however, been argued for by
Byrne (1990) and Ouhalla (1999) for the structurally similar nominal clefts
of Saramaccan and Arabic. Saramaccan has a focus structure (typified by
sentence (9a)), which shares important properties with Coptic nominal
clefts: the initial focus phrase precedes an open sentence (where the open
position or gap is indicated as e). Moreover, a pronominal element may be
112 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

inserted after the focus phrase, which copies its person and number feature.
Byrne (1990) provides a single-clause analysis for that construction, in
which the focused constituent (NP) is base-generated in an adjoined posi-
tion to the CP. According to this analysis, the optionally present pronoun
represents a pronominal operator that has moved from the gap position, as
schematically represented in (9b) (labels are from Byrne (1990). 7

(9) Saramaccan focus structures with optionally present pronouns

a. DEE FOUj (DEj) di womi sei ex a di wojowojo
the (pi) bird THEY the man sell LOC the market
' T h e man sold the birds at the market.' (Byrne 1990: 67 (16a))
b. [ s . . . N P j [ C pSPEC (Υ;) Comp ( X ° ) I s . . . c , ...111

At first glance, a mono-clausal analysis of Coptic clefts seems to be sup-

ported, albeit indirectly, by the broad distribution of relativizing markers.
As we can see from (lOa-d), the occurrence of relative markers like ant-
does not indicate relative embedding per se, since such markers appear not
only in DP-internal relative clauses and clefts, but also in main clause wh-
in-situ questions and declarative focus sentences.

(10) a. Relative clauses

e-po-ma [CP ant-a-k-k*nta-f mheta-f ]
to-DEF:SM-place REL-PERF-2SM-find-3SM inside-3SM
'to the place where you have found it (the boat)'
(Acts of Andrew and Paul 204: 145-146)
b. W h - c l e f t s
nim n-rome eneh pe [CP nt-a-f-ei e-rat-f
w h o of-man ever PRON:SM REL-PERF-3SM-come to-foot-3SM
m-pe.n-jot e-f-mokh n-het
as-DEF:SM.IP-father REL(-PRES)-3SM-be.sad of-heart
e-me-f-ßok e-f-rafe ] ?
REL-NEG.HAB-3SM-go REL(-PRES)-3SM-rejoice
'Which man ever (is there) that had gone to our father (Matthew)
disheartened and did not go away rejoicing?' ( K H M L II 11: 8 - 9 )

c. Wh-in-situ questions
ant-a u f°Pe mmo-k pe.n-tfoejs pa-no ?
REL-PERF what happen to-2SM DEF:SM.lP-lord DEF:SM-king
' W h a t happened to you, our lord and king?' (Eudoxia 36: 24)
 The nominal cleft construction in Coptic Egyptian 113

d. Focus-in-situ declaratives
ont-a-f-fiße gar an nkyi p-nute han ne.f-hap
REL-PERF-3SM-change PCL NOT FM DEF:SM-god in DEF:P.3SM-law
'God has not changed in his laws. (Shenoute De Iudicio, 31: 9 - 1 1 )

In Reintges (2004a: ch. 4) and Reintges, LeSourd & Chung (2005), the
relative marking of the tense-aspect word is analysed as w/z-agreement
morphology, which flags classical w/z-constructions (relative clauses, wh-
questions, and focusing constructions) and sets them apart from pragmati-
cally neutral declaratives. Notice, however, that operator-variable construc-
tions are not automatically flagged by special inflectional morphology: the
relevant operator must also be in the appropriate configuration. As shown in
(1 la), relative marking is obligatory when the w/z-phrase appears in-situ in a
clause-internal argument or adverb position. By contrast, no such special in-
flection is resorted to when the w/z-phrase appears ex-situ in a left-peripheral
focus projection to the left of the perfect marker a-, as shown in (1 lb) below.

(11) a. Wh-in-situ question with relative marking

awo nt-a-u-ei eßol ton ?
and REL-PERF-3 P - c o m e PCL where
'From where did they come?' (Apocalypse 7: 13, ed. Budge)
b. Wh -fronting question without relative marking
eßol ton a-tetsn-ej e-pej-ma ?
PCL where PERF-2P-come to-DEM:SM-place
'From where did you come here?' (Budge, Martyrd. 220: 8)

Given the complementary distribution of w/z-fronting and w/z-agreement

marking, the co-occurrence of the w/z-phrase in Spec-FocusP and relative
markers in w/z-clefts like (10b) above would be somewhat mysterious under
a mono-clausal analysis, but receives a straightforward explanation if wh-
clefts involve minimally two clauses.
Further evidence for the biclausal structure of Coptic clefts comes from
the limited distribution of the relative complementiser el·. This comple-
mentiser is licensed in subject relatives with present or future time reference,
as (12a-b) below illustrates. In such relative constructions, the relative
complementiser must be adjacent to a gap (as opposed to a resumptive pro-
noun). In view of the fact that it is excluded from arguably mono-clausal
w/z/focus-in-situ constructions, we can safely assume that the cleft predicate
instantiates a relative clause proper.
114 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

(12) a. Subject relative with et-complementiser

anok pe p-aggelos [CP et diakonej
I PRON:SM DEF:SM-angel CREL (PRES) serve
e-pe.k-jot Aßraham ]
to-DEF:SM.2SM-father Abraham
Ί (am) the angel who serves your father A b r a h a m '
(Testament of Issac 229: 19-20)
b. DP-clefts
p-tfoeis gar pie) [Cp et ο m-metre
DEF:SM-lord PCL PRON:SM C Y P R E S ) be as-witness
n-ta-senedis ].
to-DEF:SM: 1 S-conscience
'For (it is) the Lord who is witness to my conscience.'
(V. Pach. 89: 1 - 2 )

If the nominal cleft sentences with ef-marked CP-predicates would be mono-

clausal structures, there would be nothing in the sentence that the gap would
correspond to.
Ouhalla (1999) argues that the nominal cleft sentences in Modern Standard
Arabic are simple equative copular sentence with the free relative function-
i n g as t h e D P p r e d i c a t e : [F(OCUS)-XP PRON [ D P RELATIVE CLAUSE]]. C o m -
pare (13a) with (13b), where the underlined string has the referential inter-
pretation of a definite noun phrase (Ouhalla 1999: 343ff.).

(13) a. Nominal clefts in Modern Standard Arabic

ZAYNAB-u hiyya llatii ?allaf-at l-riwaayat-a
Zaynab-NOM PRON.she REL wrote-she the-novel-ACC
'It was ZAYNAB who wrote the novel.'
(Ouhalla 1999: 341 (5a))
b. Free/headless relative clauses
wasal-a lladhi haddath-ta-nni 'an-hu
arrived-he REL.the-he talked-you-to-me about-him
' T h e one you talked to me about has arrived.'
(Ouhalla 1999: 344 (6a))

In the Coptic counterpart, the predicate has no such DP layer, which would
give it the syntactic distribution and referential properties of free relative
clauses. In the examples considered so far, we have treated the deictic pro-
noun and the relative clause as independent sentence constituents. The very
 The nominal cleft construction in Coptic Egyptian 115

same surface string can also be found in free relative clauses, however.
Compare pronominal pe in the cleft construction in (14a) with the formally
identical definite article pe- 'the' that introduces the free relative clause
pe-nt-a-i-meewe 'what I had thought' in (14b) below.

(14) a. Pronoun cleft

anok ho pe [cp nt-a-i-r noße ero-k
I SELF-Is PRON:SM REL-PERF-1 S-do sin against-2SM
n-u-sop n-wot ]
in-INDEF:S-time o f - s i n g l e
'(It is) me, however, who sinned against you just once!'
(Acts of Andrew and Paul 200: 87)
b. Free relative clauses
[DP pe-nt-a-i-meewe de ero-f ] a-i-aa-f
DEF:SM-REL-PERF-1 S-think PCL about-3SM PERF-lS-do-3SM
Ί did what I had thought about.' (Budge, Martyrd. 207: 10)

If the nominal cleft construction consisted only of a DP and a free relative

clause, we could not explain the following facts. First, in free relatives the
prenominal definite article may vary with a freestanding demonstrative
pronoun, as seen in (15a-b). No such variation is attested for nominal cleft
sentences. ([NP0] represents the empty nominal head of the free relative).

(15) a. Free relative clauses with definite article p(e)-

Judas p-apostolos [ DP p- [NP0] [CP EF moofe
J u d a s DEF:SM-apostle DEF:SM C REL (PRES) w a l k
mn pe.n-tfoejs ]]
with DEF:SM.lP-lord
'the apostle Judas, who accompanied our Lord'
(Acts of Andrew and Paul 200: 74)
b. Free relative clauses with demonstrative pronoun pai
pe.n-jot Abraham [ Dp pai [CP nt-a-f-mu ]]
DEF:SM.lP-father A b r a h a m DEM:SM REL-PERF-3:SM-die
'Our father Abraham, who had died' (John 8: 53)

Second, when free relative clauses occur in nominal predicational structures,

the element pe shows up twice, namely as a determiner on the free relative
clause and as a deictic pronoun linking the subject to the nominal predicate,
see (16a-b) below. No such repetition of pe is found in nominal cleft sen-
tences, however.
116 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

(16) Equative copular sentences with free relatives

a. u-athet de pe | DP ρ- [ N P 0] [Cp et moste
INDEF:S-stupid PCL PRON:SM DEF:SM CREL (PRES) hate
n-ne-tfpio ]]
DO-DEF:P-criticism
'The one who hates criticism (is) stupid.' (Proverbs 12: 1)
b. LDPZ?«?- [NP0] I CP nt-a-u-diakonei de kahs)] ne
DEF:P-REL-PERF-3P-serve PCL ADV PRON:P
[ DP ne- [ N P 0] [CP nt-a-u-ah(e) e-rat-u
DEF :P-REL-PERF-3 P-stand on-foot-3 Ρ
hm p-fi n-ne-graphe ]]
in DEF:SM-height of-DEF:P-scripture
'The ones who served well (are) the ones who stand within the
confines of the Scriptures.' (Praec. et institut. Pach. 33: 30-31)
c. ntok pe [ Dp p- [NP0] [CP et neu ]]
you:SM PRON:SM DEF:SM CREL (PRES) come
'Are you the one who comes?' (Luke 7: 20)

Third, as we can see from (17a-b) below, the deictic pronoun may be
dropped when the clefted constituent is an independent pronoun. By con-
trast, it is never possible for free relative clauses to omit the definite article
or the demonstrative pronoun.

(17) Pronoun clefts with omission of the pronominal copula

a. anok [cP et na-kategori ammo-tgn anahrom p-jot ]
I CREL FUT-accuse DO-2P before DEF:SM-father
'(It is) I who will accuse you to the Father.' (John 5: 45)
b. ntok [cp et na-ti logos m-p-nute ha ta-psykhe ]
you:SM CREL FUT-give account to-DEF:SM-god for DEF:SF.lS-soul
'(It is) you who will account to God for my soul.' (Hilaria 5: 28)

A final piece of evidence against a DP [Dp FREE RELATIVE] analysis

comes from the possibility of parenthetical expressions like vocative phrases
to disrupt the syntactic continuity between the pronominal copula and the
relative clause predicate. It is, however, never possible for vocatives to
appear between the D°-head and the embedded relative clause.
 The nominal cleft construction in Coptic Egyptian 117

( 1 8 ) COPULA > VOCATIVE > RELATIVE CLAUSE

pe.k-fatfe pe pa-tfoejs [Cp et tutfo
DEF:SM.2SM-word PRON:SM DEF:SM(-1 S)-lord CREL (PRES) save
n-won nim ]
DO-one every
'(It is) your word, oh Lord, that saves everyone.' (Sapientia 10: 12)

In this section, we discussed and refuted a free relative clause analysis of

the cleft predicate. The main argument concerns the systematic structural
differences between nominal clefts and equative copular constructions.

3.2. The categorial restriction on the clefted constituent

Clefts in Coptic are more restricted than the English ones in that the clefted
constituent cannot be of any other category than a DP. This categorial re-
striction is illustrated in the data in (19). Notice that Coptic has both argu-
ment (subject, direct, indirect and prepositional object) and adjunct clefts,
but due to the categorial restriction it is never possible to cleft the entire
prepositional or adverbial phrase, only the DP part of it:

(19) ARGUMENT AND ADJUNCT CLEFTS

a. Subject
ta-feere tu-pistis te
DEF:SF(-1 S)-daughter DEF:SF(-2SF)-faith PRON:SF
[cp 9nt-a-s-nahme ]
REL-PERF-3 SF-save(-2SF)
'My daughter, (it is) your faith that has saved you.' (Luke 8: 48)
b. Direct Object
u-hoß gar [CP e-nanu-f ] pe
INDEF:S-thing PCL REL(-PRES)-be.fair-3sm PRON:SM
[cp nt-a-f-aa-f ]
REL-PERF-3 SM-make-3SM
'(It is) a beautiful thing that he did.' (AP Chaine no. 17,3: 26)
118 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

c. Prepositional object
u-mnt-at-noße tie) [Cp et-u-fine nso-s
INDEF:S-NOM-sinless PRON:SF C REL -PRES-3P-seek for-3SF
n-toot-n ]
from-hand-lP
'(It is) a sinless (life) which is requested (lit. which they request)
from us.' (Testament of Isaac 233: 21)

d. Locative adverb
p-kah on pe [CP nt-a-u-kto-u ero-f ]
DEF:SM-earth PCL PRON:SM REL-PERF-3P-turn-3P to-3SM
'It (is) again the earth that they (i.e. the birds and the fish) turn
themselves to.' (Zenobius 202: 15-16)

e. Cause/reason adverb
awo nai n-tei-he mn [DP ne- [CP et eine
and DEM:P of-DEM:SM-kind with DEF:P-C REL (PRES) resemble
mmo-u ]] n(e) [cp etere p-Jatfe
DO-3P PRON:P REL(-PRES) DEF:SM-word
[Cp et seh ] tfo mmo-s etßeet-u (...)]
CREL (PRES) be.written say DO-3SF because.of-3P
'(It is) such kind of people and those who resemble them that the
Scripture word (lit. the word that is written down) says about them
(...)' (Shenoute III 151: 2 6 - 2 7 )

In these examples the clefted constituent is always a DP, and it is linked

either to a subject gap or a resumptive object pronoun inside the relative
clause. W e will relate this categorial restriction to the configuration we
assign to nominal cleft sentences in section 4.1.

3.3. The agreement behaviour of the deictic copula

The deictic pronoun agrees with the clefted DP in number and gender, but
not in person. Consider the data in (20) below, all of which involve clefted
DPs.
 The nominal cleft construction in Coptic Egyptian 119

(20) NP clefts with gender and number agreement of the pronominal copula
a. SINGULAR MASCULINE pe
arcu p-haqios A pa Mena pe
perhaps DEF:SM-holy Apa Mena PRON:SM
[Cp nt-a-f-kyont ero-j etbe p-eret
REL-PERF-3SM-be.wrathful about-lS for DEF:SM-vow
Lcp nt-a-i-eret mmo-f na-f ]]
REL-PERF-1S-VOW DO-3SM for-3SM
'Perhaps (it is) the holy Apa Mena who has become wrathful about
me because of the vow that I made to him.'
(Mena, Mir. l i b : 25-29)
b. SINGULAR FEMININE te
te.k-hmhal Eudoxia t[e)_ [Cp et wof e-ei
DEF:SF.2SM-servant Eudoxia PRON:SF CREL (PRES) want to-go
ehun e-pe.k-aspasmos ]
PCL to-DEF:SM-2SM-greeting
'(It is) your maid servant Eudoxia who wishes to enter to greet you.'
(Eudoxia 56: 1-2)
c. PLURAL ne
ftow n-oeik n(e) [Cp et tef na-n mmeene]
four of-bread P R O N : P C R E L ( P R E S ) be.portioned for-LP daily

'(It is) four (loaves) of bread that are portioned to us daily.'

(Budge, Martyrd. 218: 25)

Clefted third person pronouns show the same agreement behavior as clefted
DPs, as seen in (21):

(21) Third person pronouns clefts with gender and number agreement of the
pronominal copula
a. SINGULAR MASCULINE pe
ne ntof mawaa-f pe [cp e-f-anakhorej
PRET HE alone-3SM PRON:SM REL(-PRES)-3SM-retreat
hm p-ma etmmau ]
in DEF:SM-place that
'(It) was him alone who was living as a hermit in that place.'
(AP Chaine no. 181, 43: 21 - 2 2 )
120 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

b. PLURAL ne
[DP ne- FCp nt-a-u-mu mn pe-Khristos ]]
DEF:P REL-PERF-3P-die with DEF:SM-Christ
ntou n(e) [cp et na-onh mn pe-Khristos ]
THEY PRON:P CREU FUT-live with DEF:SM-Christ
' ( A s for) those who died with Christ, (it is) them w h o will also
live with Christ.' (Shenoute IV 4: 1 8 - 1 9 )

The systematic absence of person agreement is evident in the context of

clefted first and second person pronouns, where the deictic pronoun only
reproduces the gender and number specification. This is why the third per-
son pronouns p(e) ' h e ' and n(e) 'they' are selected in (22a,b) rather than
the clitic counterparts of the clefted pronominal. The unattested examples
with full agreement between the clefted pronoun and the deictic copula are
given in (22a',b'). W e take the non-occurrence of such examples to stem
from a grammaticality restriction.

(22) First and second person pronoun clefts with gender and number, but
not person agreement of the pronominal copula

a. SINGULAR MASCULINE pe
ontok pfe) [cpef neu ]
YOU(-SM) PRON:SM CREL (PRES) come
'(Is it) you w h o will c o m e ? ' (Luke 7: 20)
a.' *antok antak [Cp et neu ]
YOU(-SM) PRON:2SM C r e l (PRES) come

b. PLURAL n e
onto ton nie) [Cp et tmayo ammo-tan ]
YOU(-P) PRON:P CREL (PRES) justify DO-2P
'(It is) you that justify themselves.' (Luke 16: 15)
b.' * an to ton on to ton [Cp et tmayo ammo-ton ]
YOU(-P) PRON:2P C r e l (PRES) justify DO-2P

Furthermore, clefted personal pronouns may be construed with an invariant

third person singular masculine pronominal pe, which gives rise to agree-
ment mismatches of the following kind.
 The nominal cleft construction in Coptic Egyptian 121

(23) Pronoun clefts with invariant pe

a. ntos 'SHE' (Mismatch in gender)
η tos gar p(e) [CP e-ne-s-moone n-n-esou
SHE PCL PRON:SM REL-PRET-3SF-pasture DO-DEF:P-sheep
m-pe.s-jot ]
of-DEF:SM.3SF-father
'Since (it) was her who pastured the sheep of her father'
(Genesis 29: 6)
b. anon 'WE' (Mismatch in number)
anon de mawaa-n p(e) | C p et foop m-pei-ma ]
WE P C L alone-LP PRON:SM CREL ( P R E S ) be in-DEM:SM-place
'(It is) us alone who reside here.' (Budge, Martyrd. 221: 2-3)

Finally, pronoun clefts allow for the deletion of the pronominal copula
when the clefted pronoun is the antecedent of a subject relative clause. 8

(24) Pronoun clefts with copula deletion

a. anok [cp et na-kategori ammo-tan anahram p-jot ]
I CREL FUT-accuse DO-2P before DEF:SM-father
'(It is) I who will accuse you to the father.' (John 5: 45)
b. ntok [Cp et rofe ]
YOU:SM CREL(PRES)be.responsible
'(It is) you who is responsible.' (Matthew 27: 4)
c. er-wof e-feene pu-tfaj nsa u
R E L - 2 S M - w a n t t o - g i v e . a w a y DEF:SM:2SF-boat f o r what
ntof [cp et saanf mmo ]
HE CREL ( P R E S ) keep.alive DO-2SF
'For what (purpose) do you (woman) wish to give away your boat?
(It is) it (the boat) that keeps you alive!' (KHML II 17: 1-3)

d. me an to tan an [Cp et tso ammo-s (...)]

Q YOU:Ρ NOT CREU ( P R E S ) say DO 3 S F
'(It is) not you that say (...)' (John 4: 35)

The main agreement patterns discussed so far are summarized in table 1. As

evidenced by clefted first and second person pronouns, the deictic pronoun
only express number and gender, but not person agreement with the clefted
constituent. In other words, the absence of a morphological representation
of grammatical person translates into a default third person form.
122 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

Table 1. Subject agreement patterns in Sahidic nominal clefts

CLEFT NUMBER A N D GENDER DEFAULT COPULA

CONSTITUENT AGREEMENT AGREEMENT DELETION

DP yes no no

3 r d PRONOUNS yes yes yes

JST/2ND P R O N O U N S
yes yes yes

The impoverishment of agreement can even be taken further as to assume

an invariant third person singular masculine value, which leaves number
and gender features unspecified. The deletion of the deictic pronoun repre-
sents the most radical case of featural underspecification.

4. The configurational analysis of Coptic nominal clefts

In this section, we present a configurational analysis that captures the inter-

pretational and syntactic properties of nominal clefts in a straightforward
way. We will argue that cleft formation in Coptic involves a two-step deri-
vation, consisting of the merger of a small clause structure, and the subse-
quent w/z-/focus-movement of its subject into a designated focus projection.
We will also provide an explanation for the distribution of the gaps and
resumptive pronouns in the relative clause predicate and its interaction with
the alternations in the form of the relative complementiser.

4.1. The small clause kernel

As was shown in the previous section, Coptic nominal clefts instantiate a

subject-predicate relationship, which is mediated by the deictic copula that
links the clefted constituent and the relative clause. In structural terms, this
means that the clefted element serves as subject to predication by the deic-
tic pronoun + relative clause complex. Neither this complex, nor the rela-
tive clause itself is a free relative with a DP-layer. Given that the relative
clause never occurs at the sentential level in other contexts beside clefting,
we deduce that it by itself cannot function as a predicate. For the relative
clause to assume such a predicative function, the deictic copula has to be
introduced into the structure. Its role is that of spelling out the head of a
 The nominal cleft construction in Coptic Egyptian 123

small clause (SC) structure (see Stowell 1981; Moro 1997 on small clause
structures in general, and Rothstein 1995; Doron 1986 on the role of pro-
nominal copulas therein).

(25) SC

{pe, te, n e } [CP CREL/REL...]

In Coptic, small clauses only license DP subjects, as the following exam-

ples of locative and nominal predicates illustrate.

(26) a. Locative predication

ti-hem-pa-jot
lS-in-DEF:SM-l S-father
Ί am in my father.' (John 14: 11)

b. Identificational nominal sentences (equatives)

anok pe p-fere m-p-nute
I PRON:SM DEF:SM-son of-DEF:SM-god
Ί (am) the son of G o d . ' (Abbatön 240: 13)

The categorial restriction on cleftability in Coptic to DPs can be directly

related to the small clause structure in (25). In restricting cleft constituents
to DPs, Coptic nominal clefts differ from their counterparts in more famil-
iar languages like English and French, which allow for a much broader
class of cleftable elements, including adverbial phrases and clausal comple-
ments (see, a m o n g various other, Huddleston & Pullum 2002: 1417-1419
and Doetjes, Rebuschi & Rialand 2004, but cf. Heggie 1993 for thematic
and discourse restrictions on English clefts). W e hypothesize that this typo-
logical variation reflects the fact that Coptic and English clefts are derived
from different underlying structures, although they share the basic ingredi-
ents of this construction, viz. an initial focus constituent and a relative
clause that contains the presupposition against which focusing takes place.
124 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

4.2. The left-peripheral position of the clefted DP

Following E.Kiss's (1998) influential analysis, we present several arguments,

based on the order of topics, complementisers and question particles, to
show that the cleft constituent does not stay in the subject position of the
small clause, but rather moves to the specifier position of a designated fo-
cus projection in the left periphery.
Consider first the data in (27a-b) below, which show that the clefted DP
is not always the first element of the clause, but may be preceded by the
finite 'THAT'-complementiser tfe and dedicated question particles like eye.
We assume without further discussion that such clause-typing elements are
merged into the exposition. This indicates that the cleft constituent is not
located in the operator position Spec, CP, but rather in the specifier of a
lower functional projection.

(27) COMP » CLEFTED D P

a. Finite subordinating complementiser tje
eßol tfe p-rro pe [cpnt-a-f-tamio-f na-s ].
PCL C DEF:SM-king PRON:SM REL-PERF-3SM-furnish-3SM for-3SF
'because (it was) the King who had furnished it (the chamber) for
her.' (Eudoxia 50: 8 - 9 )

b. Q-PCL » CLEFTED W H
eye u p(e) [Cp et na-fope ham p-et-fufu ]?
Q what? PRON:SM CREL FUT-happen to DEF:SM-CREU-dry
'What (is) it that will happen to the dried out one (the tree)?'
(Luke 23: 31)

In (28a-b), we encounter topicalized constituents that precede the clefted

DP. Their discourse status is indicated by the topicalizing particle de. This
shows that the clefted DP is below higher topics. The clefted DP must
therefore be located lower in the left periphery than discourse topics.

(28) DISCOURSE TOPIC » CLEFTED D P

a. etfn nai de u p(e) [cp ete-f-na-aa-f
besides DEM:Ρ PCL what PRON:SM REL-3SM-FUT-do-3SM
nkyi p-diaßolos ]?
FM DEF:SM-devil
'Besides these (things), what (is it) that the devil will (manage) to
do?' (Zenobius 203: 27-28)
 The nominal cleft construction in Coptic Egyptian 125

b. mmnsa p-hap de t-fote m-p-nun

after DEF:SM-law PCL DEF:SF-pit of-DEF:SM-Abyss
t(e) [Cp et na-fope na-k m-ma-m-moone ]
PRON:SF CREL FUT-become for-2SM a s - d w e l l i n g . p l a c e
'But after the (day of) judgement, (it is) the pit of the Abyss that
will become your dwelling place.' (Eudoxia 38: 24-25)

However, as we can see in (29) below, topic phrases can also occur to the
right of the clefited constituent. This lower topic position hosts various types
of elements: left-dislocated personal pronouns (29a), emphatic reflexives
(29b), topicalised time adverbials (29c).

(29) CLEFTED W H » LOWER TOPIC

a. u ntjtn pic) [CP ete-tn-wef tre-n-aa-f
w h a t YOU:P PRON:SM REL(-PRES)-2P-want CAUS:INF-lP-do-3SM
nnahm naj ]?
about DEM:P
'(As for) you, what (is it) that you want us to do about these
(things)?' (Cambyses 11:9-10)
b. u hoo-n on p(e) [cp eta-n-na-aa-f ]?
w h a t self-1Ρ PCL PRON:SM REL-lP-FUT-do-3SM
'What (is it) that we, on our part, shall do?' (Luke 3:14)
c. nim tenu p(e) [ C p et sorm m-p-meefe ]?
who A D V C O P : S M C R E L ( P R E S ) misguide DO-DEF:SM-crowd
'Who (is it) now that is misleading the crowd?'
(Acts of Andrew and Paul 212: 231)

We thus propose that the left peripheral position of the cleft constituent is the
result of w/z/focus-movement in the overt syntax. See diagram (30) for fur-
ther illustration.

(30) [CP[τορΡ [FOCP DPi [FOC· [τορΡ [sc ti [sc· {pe, te, n e } [ R C ... ]]]]]]]]

ΐ I
The proposed focus fronting process is cross-linguistically well attested as
a general strategy in languages like Hungarian (E.Kiss 1987), Basque
(Ortiz de Urbina 1989), Greek (Tsimpli 1995), and Hausa (Green & Jaggar
2003). Focus fronting is also available as a marked alternative to in-situ
focus in Coptic Egyptian (Reintges 2003, 2004a). 9
126 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

4.3. The representation of the clefted constituent in the relative clause

predicate

In this final section, we spell out how the cleft constituent is referentially
linked to a gap or a resumptive pronoun in the associated relative clause.
The broad syntactic distribution of resumptive pronouns in relative clauses
is illustrated in (19b-e) above. Coptic resumptive pronouns behave differ-
ently from the ones found in English type languages in that they are not
restricted to the context of island violations (see, among various others,
Demirdache 1991; Tellier 1991; Shlonsky 1992). 10
Following Demirdache (1991, 1997) we analyze resumptive pronominals
as in-situ relative operators on a par with w/z-in-situ phrases. The relative
operator undergoes w/z-movement to the specifier of CP thereby creating
the relevant operator-variable dependency. On this view the resumptive
pronoun represents the spell-out of the lower copy of the operator. The
movement configuration of resumptive pronominalization is represented in
diagram (31). See Browning (1997) for an early account of relative opera-
tors as null pronominals (OPPR0N).

(31) C P (RELATIVE CLAUSE)

. . . OPpRoN · · ·

There is only one position where a gap appears instead of a resumptive

pronoun, namely the highest subject position of the embedded clause (see
Reintges 1998). The obligatory presence of a gap in this position has been
attested in a variety of languages (see Borer 1984; McCloskey 1990;
Demirdache 1997 for representative views on the "Highest Subject Restric-
tion"). The presence of the gap in the subject position has a morphological
correlate in the complementizer allomorph et-. In this case, we contend that
the relative operator stays in-situ in the subject position. This raises the
question about how the operator-variable dependency is created. We resort
to a locality explanation along the lines of Agbayani (2000), who argues
that the w/z-subject of English questions is licensed in the Spec,TP position
because it is locally adjacent to C°, the clause-typing head just as local as
the Spec-CP position is. On this account, the operator binds a variable only
after LF-raising. The configuration for Coptic subject relatives with an in-
situ relative operator would look like (32) below.
 The nominal cleft construction in Coptic Egyptian 127

(32) C P (RELATIVE CLAUSE)

That locality is indeed the relevant licensing condition for the in-situ
placement of the relative operator is evident from subject relative clauses in
which intervening material disrupts the adjacency between the relative
complementizer and the subject position. Consider the stacked relative
clause in (33), where the possessive auxiliary verb wanta ' H A V E ' appears
sandwiched between the complementizer allomorph ete- and the embedded
subject position. Notice that in this context a resumptive pronoun appears
in the embedded subject position.

(33) Stacked relative clause with both subject gap and resumptive pronoun
pei-rome, [CP1 et , waaß [CP2 ete-wsnta-f, mmau
DEM:SM-man CREL (PRES) be.holy REL-HAVE-3SM there
rn-pei-tncefe n-arete j]
DO-DEM:SM-multitude of-virtue
'this holy man who possessed such a multitude of virtues'
(Budge, Horn. 2: 1-2)

A full analysis of the distribution of gaps and resumptive pronouns in Coptic

relative constructions, which clearly depends on one's theory of locality,
falls outside the scope of the present paper. We hope to tackle this issue in
future research.

5. Concluding remarks

This paper dealt with the syntactic and semantic aspects of the nominal
cleft construction in Coptic Egyptian. Nominal clefts in this language were
shown to belong to a larger family of focusing constructions, where the
fronted cleft constituent corresponds to contrastive (identificational) focus
in the majority of cases, although non-contrastive, presentational focus
readings are also available. As we have shown, Coptic clefts partake in the
grammar of copular constructions. The deictic copula is merged into the
128 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

head position of a small clause. It links the small clause subject, the clefted
NP, to its predicate, the associated relative clause. The clefted DP cannot
remain in-situ in the subject position of the small clause, but has to undergo
focus fronting to the specifier position of a designated focus phrase in the
left periphery. The underlying small clause configurationality of Coptic
nominal clefts provides a principled explanation for the categorial restric-
tion on electable elements, which can only be DPs.

Acknowledgements

This paper has benefited from detailed comments by Marcel den Dikken,
Barbara Egedi and an anonymous reviewer. The work of the first two authors
was supported by the Netherlands Organisation for Scientific Research
(NWO). All remaining errors are ours.

Notes

1. Since the Arab conquest of Egypt in 641 CE, Coptic was gradually replaced by
Arabic for most practical purposes and reduced to a mainly ecclesiastical lan-
guage, cultivated only by educated members of the Christian minority. The
appearance of Coptic grammars, vocabularies, and textual editions written in
Arabic in the 13th century CE signal a revived interest in Coptic philology by
Egyptian Christian scholars, but also the disappearance of Coptic as a spoken
language.
2. Coptic Egyptian is the linguistic outcome of widespread bilingualism within a
speech community, with Greek as the politically and culturally predominant
language. Greek was not only the language of the literate elite, but also the
language of the Holy Scriptures and the new religion and therefore a language
of great cultural importance. Although no clear statistics are available at pre-
sent, it is estimated that approximately forty percent of the Coptic vocabulary
consists of Greek loan words. The transfer of Greek lexical material was not
confined to lexical items, but also involved a considerable amount of function
words, such as sentence conjunctions, discourse markers, manner and time ad-
verbiale, and even some prepositions. Language contact phenomena at all
grammatical levels (lexicon, syntax, discourse structure) show that Coptic
should be classified as a bilingual language variety with two parent languages,
Egyptian and Greek (see Reintges 2004b for a more detailed discussion).
 The nominal cleft construction in Coptic Egyptian 129

3. The following abbreviations are used in the glosses. (Glosses are given in paren-
theses for morphemes that have no surface-segmental shape): 1 'first person';
2 'second person'; 3 'third person'; ADV 'adverb', C 'subordinating com-
plementiser'; CRQ 'relative c o m p l e m e n t e r ' ; CAUS.INF 'causative infinitive';
CONJ 'conjunctive'; DEF 'definite article', DEM 'demonstrative article'; DO
'direct object marker'; F 'feminine'; FM 'focus marker'; FUT 'future'; HAB
'habitual aspect'; INDEF 'indefinite article'; INF 'infinitive'; Μ 'masculine';
NEG 'negative scope marker'; NEG.PERF 'negative perfect'; NOM 'nominalising
prefix'; PCL 'particle'; PERF 'perfect'; Ρ 'plural'; PRES 'present tense'; PRET
'preterit'; PRON 'deictic pronoun/pronominal copula'; Q 'question particle';
REL 'relative marker'; s 'singular'; TEMP 'temporal conjugation'. We distin-
guish relative markers from relative complementizers, since the former but not
the latter can also appear in main clauses. See Reintges (2004a) for the text
editions used in this article.
4. Note that the pronominal copula in nominal clefts is phonologically reduced
(as indicated by the parentheses), but not in nominal sentences. This is because
the deictic copula of clefts is a proclitic element that is attached phonologically
to the adjacent relative clause, while its counterpart in equatives is a clause-
second enclitic. The phonological reduction of the deictic copula stems from
an optional process of vowel elison (cf. (2a,c)): pe, te, ne —> p, t, η / CR|:1
et, ete, e, ere (cf. Polotsky 1962: 414 and Layton 2000: 371, §464).
5. As we can see from the grammaticality contrast between the (a) and the (b)
examples of (i) and (ii), there is a definiteness restriction in the corresponding
cleft constructions in Morrocan Arabic and Modern Hebrew.
(i) Definite restriction in Moroccan Arabic
a. L-WLAD huma lli sarrd-at (-hum) Nadia.
the-children PRON.they RM sent-she (-them) Nadia
' It was the CHILDREN that Nadia sent.' (Ouhalla 1999: 341 (5b))
b. *WLAD huma lli sarrd-at (-hum) Nadia.
children PRON.they RM sent-she (-them) Nadia
*'It was CHILDREN that Nadia sent.'
(ii) Definite restriction in Modern Hebrew
a. Dani hu Se 'azar le Dina
Dani he that helped to Dina
' It is Dani who helped Dina.' (Doron & Heycock 1999: 77 (21))
b. ??veled hu Se 'azar le Dina
boy he that helped to Dina
'It is a boy/one boy who helped Dina.' (Edit Doron, p.c.)
Ouhalla (1999: 341) hypothesizes that "the restrictive scope of the focus position
in Arabic clefts may well have to do with the involvement of the pronominal
copula", a plausible assumption given the mismatch in definiteness that would
130 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

arise when an indefinite DP is clefted. Notice, however, that Coptic nominal

clefts do not display any defmiteness restrictions, yet employ the same type of
pronominal copula to link the focus item to the cleft predicate.
6. In Coptic new information focus can be in-situ as shown by the question-answer
pair in (i).
(i) Q: e-tetn-tfi m-pej-rome e-ton ?
REL (-PRES)-2p-bring DO-DEM:SM-man to-where
A: e-n-tfi mmo-f e-p-topos n-Apa Mena
REL (-PRES)-LP-bring DO-3SM to-DEF:SM-shrine of-Apa Mena
'Where do you bring this (sick) man to? - We bring him to THE SHRINE
OF APA MENA' (Mena, Mire., 24b: 1 - 6 )
Focus-in-situ constructions may also express explicit contrast, for instance, in
the "NOT X BUT Y" construction, see (ii).
(ii) NEG: ompor pa-fere amp-u-tof-ak gar
no DEF:SM:ls-son NEG.PERF-3P-destine-2SM PCL
e-ti-oikonomia
for-DEM:SF-career
AFF: alia ant-a pa-1foe is tof-ak e-u-solsal
but REL-PERF DEF:SM-lord destine-2SM for-lNDEF:s-comfort
on-ne-sneu [CP et waaß [Cp et foop ham
for-DEF:P-brothers CR,L(PRES) be.holy CREI (PRES) live in
pa-tfaye ]]
DEF:SM-desert
'No, my son! For you have not been destined (lit. they have not destined
y o u ) FOR THIS CAREER (as a hermit), but the Lord has appointed y o u AS A
COMFORT for the holy brothers who live in the desert'
(Budge, Martyrd 216: 33-217: 1)
The availability of identificational and new information focus fronting in in-
situ constructions has also been observed for Hausa (cf. Green & Reintges
2004 for a comparative analysis).
7. In an earlier study, Koopman (1982) proposed a similar topicalisation analysis
for Haitian cleft sentences, which was later refuted by Lumsden (1990) on the
basis of the distribution of tense and negation. Naturally, a monoclausal analy-
sis that involves adjunction of a focus constituent to CP would no longer be
feasible in contemporary theory.
8. Coptic dialects display a considerable amount of variation with respect to
copula deletion. The language variety represented by Papyrus Bodmer VI
permits copula deletion not only in the context of clefted pronouns, but also in
the contexts of clefted indefinites DPs, but only if the cleftee is not the subject
of the embedded relative clause, as exemplified in (i) (see Polotsky 1962:425,
fn. 1 for some discussion).
 The nominal cleft construction in Coptic Egyptian 131

(i) u-dororii gar [Cp e-nanu-f

iNDEFiS-gifit PCL REL(-PRES)-be.beautiftil-3SM
[cp ete-ej-ti mmo-fj ne-tn ]]
REL(-PRES)- 1 s - g i v e D 0 - 3 S M
to-2P
'(It is) a beautiful gift that I give to you.'
(Proverbs 4: 2 [Papyrus Bodmer VI])
Clearly, future research needs to clarify the parametric differences of Coptic
language varieties with respect to copular agreement in nominal clefts.
9. In line with recent developments (Chomsky 2001), one could interpret this
movement as being motivated by an unvalued operator feature on the attract-
ing Foc°-head. When the cleftee is moved to the Spec, FocP position, the un-
valued operator feature of Foc° is valued, i.e. it is interpreted as [+identifi-
cational] or [+presentational] focus (see above, section 2 for the different
semantic types of focus in Coptic clefts).
10. Naturally, Coptic resumptive pronouns will also occur as last resort devices in
islands. This is illustrated by the following example, in which an entire cleft
sentence is embedded inside a relative clause. The resumptive pronoun is
found in the locative PP mmo-s 'in it' within the relative clause predicate,
which is an island.
(i) Resumptive pronominalisation in Islands
u-poleiSj (...) [Cp e-hen-soßt fem nie)
lNDEF:s-city REL(-PRES)-lNDEF:P-wall small COP:P
[CP et _ mmo-s, ]]
CK,.:1. (PRES) in-3SF
'a city (about which holds that) (it is) small walls that (there are) in it'
(Shenoute IV 24: 27)

References

Adger, David, Susan Pintzuk, Bernadette Plunkett and George Tsoulas (eds.)
1999 Specifiers: Minimalist approaches. New York: Oxford University
Press.
Agbayani, Brian
2000 if/z-subjects in English and the vacuous movement hypothesis. Lin-
guistic Inquiry 31: 703-713.
Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds.)
1997 Materials on left-dislocation. Linguistik Aktuell / Linguistics Today
14. Amsterdam/Philadelphia: Benjamins.
Barbiers, Sjef, Johan Rooryck and Jeroen van de Weijer (eds.)
1998 Small words in the big picture: Squibs for Hans Bennis. HIL Occa-
sional Papers 2, Leiden: Holland Institute of Generative Linguistics.
132 Chris Η. Reintges, Anikö Liptäk and Lisa Lai Shen Cheng

Borer, Hagit
1984 Restrictive relatives in Modern Hebrew. Natural Language and
Linguistic Theory 2: 219-260.
1986 The syntax of pronominal clitics. Syntax and Semantics 19: 313-332.
Orlando: Academic Press.
Browning, Maggie
1987 Null operator constructions. Ph.D. dissertation, Massachusetts Insti-
tute of Technology.
Byrne, Francis
1990 Toward an account of preclausal focus in some Creole languages.
Linguistics 28: 661-688.
Cheng, Lisa Lai-Shen and Norbert Corver (eds.)
2005 Wh-Movement moving on. Cambridge, Mass.: MIT Press.
Chomsky, Noam
2001 Derivation by phase. In Ken Hale: A life in language, M. Kenstowicz
(ed.), 1-52. Cambridge, Mass.: MIT Press.
Demirdache, Hamida
1991 Resumptive chains in restrictive relatives, appositives and dislocation
structures. Ph.D. dissertation, Massachusetts Institute of Technology.
1997 Dislocation, resumption and weakest crossover. In Materials on left-
dislocation,, Elena Anagnostopoulou, Henk van Riemsdijk and Frans
Zwarts (eds.), 193-231. Linguistik Aktuell / Linguistics Today 14.
Amsterdam/Philadelphia: Benjamins.
Doetjes, Jenny, Georges Rebuschi and Annie Rialand
2004 Cleft sentences. In Handbook of French semantics, Henriette de
Swart and Francis Corblin (eds.). Stanford: CSLI.
Doron, Edit
1986 The pronominal copula as agreement clitic. In The syntax of pro-
nominal clitics, Hagit Borer (ed.), 313-332. Syntax and Semantics
19. Orlando: Academic Press.
Doron, Edit and Caroline Heycock
1999 Filling and licensing multiple specifiers. In Specifiers'. Minimalist
approaches, David Adger, Susan Pintzuk, Bernadette Plunkett and
George Tsoulas (eds.), 69-89. New York: Oxford University Press.
Green, Melanie and Philip Jaggar
2003 Ex-situ and in-situ focus in Hausa. In Research in Afroasiatic grammar
II, Papers from the fifth conference on Afroasiatic languages (Paris,
June 2000), Jacqueline Lecarme (ed.), 189-214. Current Issues in
Linguistic Theory 241. Amsterdam/Philadelphia: John Benjamins.
Green, Melanie and Chris H. Reintges
2004 Syntactic anchoring in Hausa and Coptic w/?-constructions. In Papers
of the Berkeley Linguistic Society, Special Session on Afro-Asiatic Lan-
guages, Andrew Simpson (ed.), 61-72. Berkeley: Linguistic Society.
 The nominal cleft construction in Coptic Egyptian 133

Heggie, Lorie
1993 The range of null operators. Natural Language and Linguistic Theory
11: 45-84.
Huddleston, Rodney and Geoffrey K. Pullum
2002 The Cambridge grammar of the English language. Cambridge:
Cambridge University Press.
Kenstowicz, Michael (ed.)
2001 Ken Hale: A life in language. Current Studies in Linguistics 36.
Cambridge, Mass.: The MIT Press.
Kiss, Katalin E.
1987 Configurationality in Hungarian. Budapest: Akademiai Kiado.
1998 Identificational focus versus information focus. Language 74: 2 4 5 -
273.
Kiss, Katalin E. (ed.)
1995 Discourse configurational languages. New York: Oxford University
Press.
Koopman, Hilda
1982 Les questions. In Syntaxe de l'Haitien, Claire Lefebvre, Helene
Magloire-Holly and Nannie Piou (eds.). Ann Arbor, Michigan:
Karoma.
Lambrecht, Knud
2001 A framework for the analysis of cleft constructions. Linguistics 39:
463-516.
Layton, Bentley
2000 A Coptic grammar with chrestomathy and glossary. Porta Lingua-
rum Orientalium N.S. 20. Wiesbaden: Harrassowitz.
Lecarme, Jacqueline (ed.)
2003 Research in Afroasiatic grammar II. Papers from the fifth confer-
ence on Afroasiatic languages (Paris, June 2000). Current Issues in
Linguistic Theory 241. Amsterdam/Philadelphia: John Benjamins.
Lefebvre, Claire, Helene Magloire-Holly and Nannie Piou (eds.)
1982 Syntaxe de l'Haitien. Ann Arbor, Michigan: Karoma.
Lumsden, John S.
1990 The biclausal structure of Haitian clefts. Linguistics 28: 741-759.
McCloskey, Jim
1990 Resumptive pronouns, A'-Binding, and levels of representation in
Irish. Syntax and Semantics 23.
Moro, Andrea
1997 The raising of predicates: Predicative noun phrases and the theory of
clause structure. Cambridge: Cambridge University Press.
Ortiz de Urbina, Jon
1989 Parameters in the grammar of Basque. Dordrecht: Foris.
134 Chris Η. Reintges, Aniko Liptäk and Lisa Lai Shen Cheng

Ouhalla, Jamal
1999 Focus and Arabic clefts. In The grammar of focus, Georges Rebuschi
and Lauri Tuller (eds.), 335-359. Amsterdam/Philadelphia: Benjamins.
Polotsky, Hans Jacob
1962 Nominalsatz und Cleft Sentence im Koptischen. Orientalia 3 1 : 4 1 3 -
430.
Rebuschi, Georges and Laurie Tuller (eds.),
1999 The grammar of focus. Amsterdam/Philadelphia: Benjamins.
Reintges, Chris H.
1998 The 'Highest Subject Restriction' in Coptic relative clauses. In Small
words in the big picture: Squibs for Hans Bennis, Sjef Barbiers,
Johan Rooryck, and Jeroen van de Weijer (eds), 83-89. HIL Occa-
sional Papers 2, Leiden: Holland Institute of Generative Linguistics.
2003 Syntactic conditions on special inflection in Coptic interrogatives. In
Research in Afroasiatic grammar II, Papers from the fifth conference
on Afroasiatic languages (Paris, June 2000), Jacqueline Lecarme (ed.),
363^108. Amsterdam/Philadelphia: John Benjamins.
2004a Coptic Egyptian (Sahidic dialect): A learner's grammar. Afrikawis-
senschaftliche Lehrbücher 15. Cologne: Rüdiger Köppe Verlag.
2004b Coptic Egyptian as a bilingual language variety. In Lenguas en con-
tacto de la Antigüedad a la Edad Media [Languages in contact in
Antiquity and the Middle Ages], S. Torallas Tovar (ed.). Madrid:
Consejo Superior de Investigaciones Cientificas.
Reintges, Chris H., Philip LeSourd and Sandra Chung
2005 Movement, Wh-Agreement, and apparent Wh-m-s\t\i. In Wh-Move-
ment moving on, Lisa Lai Shen Cheng and Norbert Corver (eds.).
Cambridge, Mass.: MIT Press.
Rooth, Mats
1992 A theory of focus interpretation. Natural Language Semantics 1: 7 5 -
116.
Rothstein, Susan
1995 Small clauses and copular constructions. Syntax and semantics 28:
Small clauses, 27-48. New York: Academic Press.
Shlonsky, Ur
1992 Resumptive pronouns as a last resort. Linguistic Inquiry 23: 443-468.
Simpson, Andrew (ed.)
2004 Proceedings of the twenty-seventh annual meeting of the Berkeley
Linguistics Society, March 22-25, 2001, Special Session on Afroasi-
atic Languages. Berkeley: Linguistic Society.
Stowell, Tim
1981 Origins of phrase structure. Ph.D. dissertation. Massachusetts Insti-
tute of Technology.
 The nominal cleft construction in Coptic Egyptian 135

de Swart, Henriette and Francis Corblin (eds.)

forthc. Handbook of French semantics. Stanford: CSLI.
Tellier, Christine
1991 Licensing theory and French parasitic gaps. Dordrecht: Kluwer Aca-
demic Publishers.
Torallas Tovar, S. (ed.)
2004 Lenguas en contacto de la Antigüedad a la Edad Media (Languages
in contact in Antiquity and the Middle Ages). Madrid: Consejo Su-
perior de Investigaciones Cientificas.
Tsimpli, I. Μ
1995 Focusing in Modern Greek. In Discourse configurational languages,
Katalin E. Kiss (ed.), 176-206. New York: Oxford University Press.
Genitive constructions in Coptic

Barbara Egedi

1. Introduction

1.1. Definition of 'Coptic'

Coptic is the language of Christian Egypt (4 th to 14th century) written in a

specific version of the Greek alphabet. It was gradually superseded by Ara-
bic from the ninth century onward, but it survived to the present time as the
liturgical language of the Christian church of Egypt. In this paper I examine
only one of its main dialects, the Sahidic Coptic and I use a transcription
which simply reflects the Coptic letters irrespectively of phonological de-
tails. 1

1.2. UG in the reconstruction of dead languages

Natural languages are claimed to have universal properties or principles

which constitute what is referred to as Universal Grammar. Accepting cer-
tain universal principles and observing the corresponding parameters in
Coptic, we can also analyse a language without living native speakers, and
explain its structural relations with the help of coherent models.
For example, it is considered a universal principle that the projections of
lexical heads are extended by one or more functional projections. If we
assume that it can be demonstrated in many languages, why could not we
suppose the same in the case of Coptic? Indeed, as it will be shown in chap-
ter 4, there are at least two functional projections above the Coptic lexical
noun phrase as well.
The aim of this paper is to provide an adequate account of the basic
structure of the Coptic N P within the theoretical framework of the Mini-
malist Program (a short summary of which will be found in the following
section); at the same time, I intend to find the answer to unsolved questions
related to genitive constructions. In fact, in many cases even the appropriate
questions have not been posed yet. Coptic has two different genitive con-
138 Barbara Egedi

structions: one with simple definite possessions and one for indefinite, modi-
fied or deictically marked head nouns. After establishing the distribution and
the use of the two genitive constructions, I will point out the differences
between their internal structures.
The paper is organised as follows: section 2 expounds the theoretical
assumptions regarding the noun phrase research and the minimalist pro-
gram; section 3 examines the Coptic data, the distributional conditions of
noun phrases and the internal structure of the two genitive constructions
with particular attention to the nature of linking elements between the pos-
session and the possessor. In the last section, I analyse the derivation of
simple NPs and that of genitive constructions in the framework of a modi-
fied checking theory.

2. Theoretical background

2.1. Layered Noun Phrase

The research of the noun phrase does not have as old a tradition in generative
theories as the research of the verbal constructions and the sentence as a
whole. (On the history of this research see Giorgi-Longobardi 1991.) How-
ever, since the end of the eighties several suggestions have appeared for an
analysis of noun phrases similar to that of sentences, supposing functional
categories (whose role is essentially grammatical) above lexical nouns and
their phrases. In other words, the noun phrase is actually a DP projected
from a Det(erminer), whose complement is an NP. The DP hypothesis tra-
ditionally has been ascribed to Abney (1987) but the comparison of NPs
with VPs can be traced back to Chomsky's earlier works. 2
Once this analysis was widely accepted, functional heads appeared in a
greater number in the noun phrase corresponding to the multiplied number
of functional projections in the clause. Bernstein (1991), Ritter (1991) and
others suggested that this functional FP between the lexical NP and the
functional DP should be a NumP, a projection responsible for number
specifications, and this view is going to be adopted in this paper as well.
The main motivation for supposing an intermediate FP was the require-
ment to explain the word order variations within the noun phrase across
languages and to provide a landing site for N-movement. As the approach
presented here intends to be a minimalist one, the notion of movement needs
a more accurately formulated motivation 3 in terms of checking theory. In
 Genitive constructions in Coptic 139

the next section I give a short summary of the principles and methodology
of the minimalist program.

2.2. The minimalist theory

Chomsky's minimalist theory claims that the syntax (the computational

system of Human Language) is a minimally complex system, which maps a
numeration of elements from the lexicon to a pair of interface representa-
tions at PF and LF, respectively. Syntactic derivation consists of two opera-
tions: Merge and Move. Through Merge, uninterpretable features like case
and agreement can also get into the syntax. Unlike interpretable elements,
uninterpretable features must be erased by LF to satisfy Full Interpretation.
In order to be erased, these features must be checked, that is, paired with a
corresponding feature in an appropriate structural relation (specifier-head
relation, adjoined heads).
The other operation is Move, which applies exactly in order to create this
local checking relation: a certain functional category with uninterpretable
features attracts the features of a "lower" category. Movement can be visible,
in that case it has to take place before Spell-Out in overt syntax, or it can
happen covertly, not affecting phonological representation (Chomsky 1993,
1995; Lasnik 1999).
Minimalist feature checking was developed first of all to account for
sentence-level agreement, but it can be extended to the N P as well, since
concord inside a noun phrase is actually a type of agreement among the co-
features (number, gender, case) of the noun and the determiners and adjec-
tives modifying it. Carstens (2000) provides a detailed account of minimal-
ist analysis to concord - slightly modifying the original theory of move-
ment - , and I rely on her study in many respects. 4

3. The Coptic NP

3.1. Noun phrases

Traditional grammars (e. g., Till 1986: §§44-122; Lambdin 1986: passim·,
Vergote 1950) confine themselves to list exhaustively the possible orders of
constituents, the types of articles and other determiners, without trying to
explain the relationship between certain phenomena or establishing the
140 Barbara Egedi

basic phrase structure of the NP. Two exceptions to this generalisation are
Shisha-Halevy (1986, §5.1.1. and passim) - however, his view is entirely
different from my own - and Layton (2000), who is concerned with article
phrases and specifier phrases. 5
According to the DP hypothesis, argument noun phrases are DPs,
maximal projections of the functional category D (determiner). What is more,
the complement of the D is another functional category, an intermediate
projection, often associated with the number specification of the lexical NP.
However, only necessary functional projections must be built (Grimshaw
1991).
This supposition of different NP types (NPs, NumPs, DPs) is not an
arbitrary classification, it is also motivated by the syntactic distribution of
noun phrases: the various noun phrase projections distribute differently
across sentence positions (E. Kiss 2000: 124) as the evidence in 3.2.2 will
show it.

3.2. The Coptic data

3.2.1. Gender and number

Nouns have masculine or feminine grammatical gender, but in lack of any

modifying element, this feature is usually not expressed morphologically.
The formal marking of gender is limited to a class of exceptional nouns
which have two related forms corresponding to male and female biological
sex (son/söne = brother/sister). In most cases it is the article or some other
determiner that shows the gender and also the number specification:

(1) röme söse

man field
(2) p-röme t-söse "n-röme/söse
def:sg.m.-man def:sg.f.-field def:pl.-man/field
'the m a n ' 'the field' 'the men/fields'
(3) pei-rdme tei-söse, nei-röme/söse
dem:sg.m.-man denr.sg.f.-field dem:pl.-man/field
'this m a n ' 'this field' 'these men/fields'
(4) pef-röme tef-söse nef-röme/söse
poss:sg.m.-man poss:sg.f.-field poss:pl.-man/field
'his m a n ' 'his field' 'his men/fields'

(5) ou-•rome/sose hen-röme/söse

indef:sg.-man/field indef:pl.-man/field
'a man' 'men/fields'
 Genitive constructions in Coptic 141

The definite article and the demonstrative have three distinct forms: in the
singular there is a masculine as well as a feminine form, while in the plural
no such morphological distinction can be found. The indefinite article has a
singular and a plural variant. The possessive article does not only mark the
number and gender of the possessed noun like the article but also the per-
son, the number and - in 2. and 3. sg. - the gender of the possessor. 6

3.2.2. Distribution

The distribution of definite and indefinite noun phrases is clearly different

in three (morpho)-syntactic cases:

a.) in the so-called 'first present' type sentences, the subject cannot be in-
definite, more precisely, an indefinite subject must be preceded by an
existential particle:

(6) p-röme rime

def:sg.m.-man cry
'the man is crying'

(7a) *ou-röme rime (7b) oun ou-röme rime

indef:sg.-man cry ex.prt. indef:sg.-man cry
'a man is crying' 'a man is crying/there is a man crying'

b.) indefinite noun phrases cannot be modified by the relative converters 7

et-/ent-, but by the circumstantial converters only. In other words, if the
antecedent of a relative clause is indefinite, a circumstantial verbal form
follows:

(8) p-röme et-rime

def:sg.m.-man conv:rel.ps-0-cry
'the man who is crying'

(9) ou-röme e-f-rime

indef :sg.-man con ν: circ.-proN :3 sg. m. -cry
'a man who is crying'

c.) Coptic has two types of genitive constructions, in one of which definite-
ness is obligatory. In fact, the situation is more complicated, since in
this pattern the possessed noun must have a definite article but cannot
have any other determiner or modifier.
142 Barbara Egedi

3.2.3. Genitive constructions

Let us call the two genitive constructions pattern A and pattern B, exempli-
fied in ( 1 0 M 1 3 ) :

Pattern A
e
(10) p-sere m-p-röme
def:sg.m.-son o/-def:sg.m.-man
'the man's son'

Pattern Β
e
(11) ou-sere nte-p-röme
indef:sg.-son o/-def:sg.m.-man
'a son of the man'
e
(12) p-sere "n-sabe nte-p-röme
def:sg.m.-son adj.prt.-clever o/-def:sg.m.-man
'the clever son of the man'

(13) pei-sere "nte-p-röme

dem:sg.m.-son o/-def:sg.m.-man
'this son of the man'

In both patterns the order of the constituents is: Possessed N P + morpheme

expressing genitive relationship, genitive marker + Possessor NP.
The difference between the two constructions is that in pattern A the pos-
sessed N P is introduced by a definite article, and the possessor expression is
marked by a genitive marker "n- Cm-)? while in pattern B, the possessed NP
may be indefinite, preceded by a demonstrative, 9 or modified by an adjective,
and the possessor expression is introduced by a preposition ente~, enta- 10
Before turning to structural investigation, we have to say a few words
about the status and nature of these morphemes marking the possessive
expression, and about other controversial aspects present also in traditional
Coptic grammars.
It is clear that contrary to certain opinions (Lambdin 1986: 6), the en-
element in pattern A is not a preposition. It has no prepronominal form (as
it is expected from a standard preposition). In fact, pronominal possessors
are expressed by what we call the possessive article:

(14a) pef-sere (14b) nef-sere (14c) tek-maay

poss.art.-son poss.art.-son poss.art.-mother
'his son' 'his sons' 'your mother'
 Genitive constructions in Coptic 143

It is somewhat embarrassing that apparently numerous function words have

the same form. For example an e n - introduces the direct object 11 as well,
and there is also a partitive preposition en~ illustrated in (15), which is a
real preposition with two forms. One has to distinguish it from the genitive
marker in (16), functioning more like a case marker morpheme:

c e
(15a) oua n-nen-eiote (15b)owa mmo-ou
one of(part)-our-fathers one of(part)-them
' o n e of our fathers' 'one of them'

(16a) p-nobe "n-nen-eiote (16b) *p-nobe "mmo-ou (16c) peu-nobe

def:sg.m.-sin of(poss)-our fathers def:sg.m.-sin of-them poss.art.-sin
'our f a t h e r ' s sin' *the sin of them 'their sin'

An en- morpheme introduces adjectives, too, as it can be seen in (12): p-sere

e
n-sabe 'the clever son'. This "n- obviously lacks a prepronominal form,
too. Observe the following phrases, 12 the first of which is a genitive, while
the second and the third are attributive constructions:

e C
(17a) p-höb n-nek-cig (17b) hen-höb n-cig
the-work of-your(poss.art)-hand indef:pl.-work adj.prt.-hand
'the work of your hands' 'handiworks'
e
(17c) pa-höb n-cig
poss.art.-work adj.prt.-hand
' m y handiwork'

What is it that shows us that the in (17b) is not a genitive marker? First,
bare nouns have a very restricted use in Coptic, some kind of an article
usually appears with them. Adjectives, on the other hand, do not need any
article or determiner, apart from this linking element. 13 Loprieno considers
it a determinative pronoun, comparing it with that in Hebrew attributive
constructions (Loprieno 1995: 56), but let me call the reader's attention to
the fact that while after a definite noun there appears another definite article
before the adjective in Hebrew (hä- 'is hag-gädöl the-man the-great i.e. 'the
great man'), after an indefinite noun the article (or determinative pronoun)
is absent before the adjective, too ('is gädöl man great i.e. 'a great man'). In
my opinion this phenomenon can be explained as an agreement in definite-
ness, and it is far from being similar to the Coptic adjective particle, whose
appearance is indifferent to the definiteness of its noun, as can be seen in
(17b). (17a) would not be grammatical with an indefinite article. (17c) pro-
vides a clearer case where the presence of the possessive article evidently
excludes the genitive interpretation of the adjective linking particle.
144 Barbara Egedi

Perhaps this ambiguous status is responsible for the fact that some Coptic
grammars refer to both en- morphemes with the same rather neutral names
as mark of relationship (Layton 2000: §203) or nota relationis (Shisha-
Halevy 1986: 20). I prefer calling it a genitive particle (cf. Till 1986: §113:
"Genitivpartikel") or rather marker, thus indicating that it functions as a
case-marker or much alike.
It is worth mentioning that the notion of "case" in Coptic does not go
without saying. The only obvious structural case is the nominative (for its
position see Reintges 2001: 100). The object of the verb can be expressed
either by a prepositional phrase or by forming a construct state with the
verb. The latter is a procedure, which is not possible in every conjugation
type. Apart from these cases all of the relations within the sentence are
expressed by prepositions. That is why the case-marker status of the e n -
morpheme is not self-evident for me.
The "nte-, enta- of pattern Β is a better candidate for being a preposi-
tion: it has two allomorphs, the first of which was exemplified in (11)-(13).
The second form serves for introducing pronominal possessors.

e
(18) ou-sere nta-f
indef:sg.-son o/proNiB.sg.m.
'a son of his'

This view is shared by many authors (Till 1986: §113; Lambdin 1983).
Nevertheless, Layton (2000: §204) finds it questionable, arguing that it
never modifies a preceding verb or verbal clause. On the one hand there is
no doubt about its prepositional origin (Loprieno 1995: 71), on the other the
above criterion does not seem to be relevant in defining the preposition as a
category.
In the Coptic data, there is a well-observable complementary distribu-
tion between the two patterns. Although earlier grammars pointed out the
syntactic conditions of this distribution, Layton explains it on semantic
grounds. He speaks about a restrictive expansion, a construction of two
entity terms such that one restricts the meaning of the other by limiting the
number of referents to which it applies (Layton 2000: §146). In his termi-
nology, restrictive expansion has two types: the general (possessive) rela-
tionship and the appurtenance construction, but they do not correspond
exactly to our pattern A and pattern Β division. Shisha-Halevy argues that
the original opposition was essential possession vs. incidental possession or
appurtenance but it is usually neutralised and maintained in isolated cases
only (Shisha-Halevy 1986: 21). I maintain that the factors determining the
 Genitive constructions in Coptic 145

choice between them are purely syntactic: the choice depends on what kind
of other N P modifiers are present. In 3.2.3., it has been shown that pattern
A is used with simple definite possessions only, while pattern Β appears
practically in every other case.
As a matter of fact, the distribution seems unusual at first sight, but we
find a similar phenomenon in the English preposed/postposed genitive alter-
nation. The preposed possessive construction {John's book, a teacher's work)
is always understood as definite. Lyons explains it by the fact that the geni-
tive phrase is in Det position, and the filling of this position forces a definite
interpretation (Lyons 1986: 138-140). In the case of an indefinite head noun,
another construction has to be used (a friend of mine), typically a PP com-
plement. This PP construction occurs when the head Ν is modified by a
demonstrative (this friend of mine). As a general rule, it is available to
make possible anything other than a simple definite possessive, which is all
the preposed construction can express. 14
The resemblance (of the English constructions) to the distribution of
Coptic genitive constructions is striking. Though Coptic has no preposed
constructions (except with a pronominal possessor), pattern A implies the
same simple definite interpretation. While in English the preposed genitive
and the definite article cannot co-occur (*the John's book) - presumably
because they occupy the same structural position, 15 the Coptic genitive
phrase in pattern A is in a postposed position, and the noun phrase has an
overt definite article.
I claim that there is a correlation between the English and Coptic data,
and in the next section I intend to provide an adequate account of this fact
and determine the exact structure of Coptic possessive patterns.

4. The structure

4.1. A modified checking theory

In 2.2. we have asserted that for being checked, formal features must be in
some local structural relation and it is this requirement that motivates
movement. First, Chomsky (1993) formulated the principle of Greed: Move
a applies to an element a only if morphological properties of a itself are
not otherwise satisfied. In other words, movement is triggered by the mor-
phological requirements of the moved element. Lasnik (1999) suggests a
revision of Greed in terms of Enlightened Self Interest (ESI): The morpho-
146 Barbara Egedi

logical requirement can be either one of the moving element (as with Greed)
or one of the position it is moving to.
At the same time Chomsky (1995) modifies his principle apparently to
the contrary, his "Suicidal Greed" insists that only the target is relevant:
Movement happens only when a category with an uninterpretable feature
attracts some feature into its checking domain, (cf. The principle Attract:
Chomsky 1995: 297)
Carstens, giving a version of Lasnik's ESI, replaces Chomsky's Attract
with the principle Move (Carstens 2000: 324), which essentially suggests
the same: the features of the target have no special status in the motivation
of movement; if a category has uninterpretable features, it may either raise
itself or "attract" raising for checking purposes. The features may be
"strong", requiring immediate checking via overt movement, or "weak", in
which case checking takes place only in covert syntax - obeying the econ-
omy principle of Procrastinate.
This extended version of the checking theory will fit our analysis of
genitive constructions, but first of all I intend to show the base structure of
the simple noun phrase in Coptic.

4.2. N P without possession

The bare noun [ NP [ N röme]] without any determiner is underspecified for

number and in cross-reference it can be either plural or singular (Layton
2000: §48). Meanwhile gender must be an intrinsic feature of nouns, testi-
fied by adjectives agreeing with it. Adopting the idea that there is a func-
tional projection NumP above the lexical NP, I consider the Num head to
be responsible for number specification.
As the noun precedes the genitive expression and the adjectives' 6 in the
word order, we can assume that Num° has a strong category feature for N°,
hence Ν overtly raises and adjoins to it.

(19) (20)

NumP Num°

Num NP N° Num°
[β number] [a gender] [β number]

Ν
[a gender]
 Genitive constructions in Coptic 147

Overt movement is supported by 'irregular' morphological plural forms, 17

where formal features are visible on nouns (21), hence the morphological
requirement of movement is obvious. It is interesting to observe that adjec-
tives, which - in my opinion - normally do not leave the NP projection,
may have exceptional plural forms, but these forms occur mainly in nomi-
nalized usage (22).

(21) en-sney e
n-saben (22) c
n-sabeeye]9
def:pl.-brother:pl. wise def:pl.-wise.pl.
'the wise brothers' 'the wise ( m e n ) '

The indefinite articles (see the forms in (5)) are merged probably in Spec,NP,
but they have to check their uninterpretable number features, hence they
raise to the specifier of NumP, where they are in appropriate spec-head
checking relation with the N+Num head.
The definite articles are heads of the DP determiner phrase or more pre-
cisely - adopting Lyons's theory (1999) - of the definiteness phrase. As we
could see in (2), the Coptic definite article shows agreement in gender and
number features with the noun. As a matter of fact, agreement (i.e. concord)
takes place between D°, whose uninterpretable gender and number features
motivate raising to check them, and the features of N+Num head. The raising
of features takes place in covert syntax as no overt movement can be ob-
served. 20

(23) DP
148 Barbara Egedi

Demonstratives (see (3)) seem to have a [definite] feature. Identifiability is

somehow part of their semantic content, so demonstratives always imply
definiteness (Lyons 1999: 18-21). For this reason, they appear in D° like the
definite article, and this is why they - demonstratives and definite articles -
can never co-occur in the same phrase.
Demonstratives can check their uninterpretable gender and number fea-
tures in the same manner as the articles: the corresponding features of the
N+Num head are attracted to check them.

4.3. Genitive constructions: pattern A

The pattern A exemplified in (10) provides a real challenge for our investi-
gation. We have to face several questions: what happens during the deriva-
tion ensuring that the possessed noun always be definite and at the same
time excluding any other determiner from the phrase? Where is the genitive
phrase base-generated?
Derived (deverbal) nouns (especially those expressing a process or an
event) inherit the theta-grid of the verb. The arguments of the noun should
be generated in the corresponding external/internal argument positions
(Ritter 1991; Siloni 1996). For example, the agent argument is assigned
nominative case in verbal constructions and genitive case in nominal con-
structions. The theme argument is assigned accusative case in both con-
structions. This is the case in Hebrew (26) and in Coptic, as well: 21

e
(24) p-ei ebol m-p-israel hen-keme22
the-come:inf. forth of-the-Israel from-Egypt
'the coming forth of Israel from Egypt'

(25a) p-koos-t23 (25 b) *pa-köös

the-bury:inf.-l.sg. poss.art.-bury:inf.
' m y burial (the burying-me)' ' m y burial'

(26) axilat dan et ha-tapuax (Hebrew)24

eating Dan acc.the-apple
' D a n ' s eating of the apple'

If we suppose that the base position of the phrase expressing the possessor
is in the SpecNP, the combination of a genitive construction with an adjec-
tival modifier (possibly adjoined to NP 2 5 ) would yield the Possession +
Adj(s) + Possessor order:
 Genitive constructions in Coptic 149

e e
(27a) *p-sere n-cabe m-p-röme
def:sg.m.-son clever o/-def:sg.m.-man
'the man's clever son'

(27b) *DP

D NumP
P-
Num NP
sere.
AdjP NP
L
n-cabe
Spec N'
e
m-p-röme
Ν
ti

However, in Coptic the possession and the possessor require direct adja-
cency in pattern A, therefore adjectives must follow both of them (which
otherwise results in structural ambiguity, since the adjective may be the
modifier of either of the nouns):

e e
(28) p-sere m-p-r5me n-cabe
def:sg.m.-son o^defisg.m.-man clever
a) 'the m a n ' s clever son'
b) 'the clever m a n ' s son'

To ensure the direct adjacency and the licensing of genitive case we have to
assume a PossP projection with an abstract possessive morpheme 26 in its
head, and with the possessor phrase in its specifier. It would explain the ex-
ceptional behaviour of the genitive marker en-, em-, which can be regarded
as a case marker on the possessor.
But we have not given account of the fact that the head noun in pattern
A cannot be indefinite or modified. It seems reasonable to assume that the
possessor phrase works as a determinant, it has a [definite] feature. There is
a similar assumption for Hungarian caseless possessor phrases, which are
claimed to have a [determininer] feature and move to the DP domain to
have it checked. 27
In Coptic, the possessor itself does not raise. Its [def] feature is weak, and
there is no overt movement - as is attested in Hungarian or English {a boy's
book). Nevertheless, the [definite] feature of the possessor has to be raised
in order to be checked by the D head. The definite article - as a default
150 Barbara Egedi

determiner - seems to be the phonological realization of the [def] feature in

the DP, otherwise lacking lexical content. As a consequence, the presence
of the definite article excludes the demonstrative and the possessive article
in the same structure.

(29) DP

D NumP
def.art.
Num PossP

— Spec Poss'
possessor
[def] Poss ΝΡ
' 1
T ^
Ν
possession

The uninterpretable gender and number features of the definite article will
be checked as illustrated in (23).
Returning to the original assumption that the possessor phrase originates
in Spec,NP we could also exclude the appearance of an indefinite possession
in pattern A. I supposed earlier that Spec,NP is the starting-point also for
the indefinite articles, so the indefinite noun and the possessor of type A
mutually exclude each other. However, the possessor merged in Spec,NP
has a further [gen] case feature (beside his [def] feature), which can be
checked in Spec,PossP. The possessor phrase is raised to the PossP domain
overtly, the proof of which is given by the possession-possessor-modifier(s)
word order discussed above (see (27) and (28)).
 Genitive constructions in Coptic 151

(30) DP

D NumP
def.art.
Num PossP

— Spec Poss'
possessor
[defj Poss NP

AdjP NP
modifier
Spec Ν'
possessor
[gen],[def] Ν
- possession

The possessive article which encodes the pronominal possessor has an in-
terpretable [definite] feature (see (4) and (14)) and its derivation is parallel
in manner with that of demonstratives. They are in the D-head and combine
with the N P by merge.
However, the definite feature assignment to the possessor is not so self-
evident if we take into consideration languages classified as adjectival-
genitive by Lyons (1999). From (31) it can be seen, that in Italian the pro-
nominal possessor can co-occur with a definite/indefinite article.

(31 a) il mio amico (31 b) un mio amico

the my friend a my friend

Two conclusions can be drawn from (31): the pronominal possessor does
not occupy the D° position since it seems to be reserved for the definite
article il, moreover, the possessiveness itself does not involve defmiteness
as a rule. However, this divergence can be due to parametric variation among
languages.
152 Barbara Egedi

4.4. Genitive constructions: pattern Β

As it was stated earlier, pattern Β constructions appear with indefinites,

demonstratives, and other modifiers in the noun phrase, as was exemplified
in (11)—(13). The possession-possessor sequence can be interrupted by
certain elements in the phrase, but their relative order is retained.
An exception to this constraint could be the predication of possession, a
special grammaticalised structure in Coptic. Existence or non-existence are
predicated by oun "there is" and rnn "there is not", respectively, followed
by the subject, indefinite 28 as can be expected. Their combinations with the
preposition "nte-, enta- yields the combinations ounte-, ounta-; rnnte-,
rnnta- expressing possession and its negation respectively (Till 1986:
§§289-294; Lambdin 1983: 90-91; Layton 2000: §§383-390). In these
constructions the ente- phrase seems to move out from its postposed posi-
tion, but the occasional appearance of the object-marker on the possession
(32) suggests that the ounte- combination has been reinterpreted as a verb,
and the original subject has become the object of the compound verb denot-
ing "have" (Polotsky 1987: §40). So this pattern is no more suitable for a
target of our investigation.

(32) ounte-p-röme (n-)ou-shime

have-def:m.sg.-man (obj.)-indef:sg.-wife
'the man has a w i f e '

Returning to genitive constructions, we can state that the cnte- phrase follows
the possession, although it can be separated from it even by a short embedded
relative clause (33).

e
(33) p-moou et-onh nte-p-ouoin29
def:sg.m.-water rel.conv.-0-alive of- def:sg.m.-light
'the living water of the light'

Moreover, it can leave the noun phrase domain:

(34) ere hen-enka calöou ero-f e

nte-ou-römei0
foe. indef:pl.-thing entrust:stative to-3.sg.m. of-indef:sg.-man
' s o m e o n e ' s things were entrusted to him'

This mobile character of the possessive in pattern Β and the prepositional

nature of W e - discussed in 3.2.3. indicate that this genitive cannot be a
structural case. Anyway, it does not have any kind of [defj feature, so it
 Genitive constructions in Coptic 153

cannot be generated in the same position as pattern A. The genitive case

seems to be an inherent one, licensed by the preposition "nte-.
For a similar phenomenon, namely the Hebrew postnominal seZ-phrase
(an alternative genitive construction beside the so-called Construct State),
Ritter suggested that it should be a base generated adjunct attached to the
matrix DP. 3 1 Providing a similar derivation for Coptic, the phrase could ad-
join either to N u m P or to D P - or in exceptional cases to CP, as well (34).

(35a) NumP (35b) DP

NumP PP DP PP
e e
nte-p-röme nte-p-röme
ou-sere pei-sere/
p-sere e n-sabe

An alternative analysis could be that the ente- phrase is directly adjoined to

N P , similarly to the adjective-adjunction. However, as we have seen, adjec-
tives modifying the same head noun precede the "nte- possessor in linear
order (see (12)). What kind of constraint should be introduced into our syn-
tax to ensure the proper order? 3 2

(36) NumP

Num NP
sere
AdjP NP
e
n-sabe
PP NP
nte-p-röme

For lack of such a constraint the former adjunction-hypothesis seems more

plausible.
The last question that remained is that if the cnte- phrase can be adjoined
to a D P as well, what prevents it from appearing with the simple definite
possession. Provided that the above derivations are right, the preference for
pattern A seems to be due to economy conditions only.
154 Barbara Egedi

5. Conclusion

In this paper I intended to provide an account of the basic nominal construc-

tions in Coptic with special attention to the two different genitive structures,
which I called pattern A and pattern B. The former had restrictions in re-
spect of the definiteness of the head noun, and I suggested an analysis ac-
cording to which the definite feature of the possessor checked in the covert
syntax assures these restrictions to be satisfied. The latter construction is a
looser relation between the possession and the possessor, and several facts
indicate that basically it has a different structure, probably an adjunction on
distinct levels of the noun phrase.
Many questions remained open regarding the Coptic noun phrase struc-
ture which I have not even attempted to answer for the moment. The sys-
tematic investigation of the adjectives and their relative order within the
noun phrase may reveal new results not only for Coptic studies but for Uni-
versal Grammar, as well. It may also modify the issue of the present paper,
and I hope to continue an instructive inquiry - for the linguistics of dead
languages, too.

Acknowledgements

For helpful comments and suggestions on an earlier version of this paper I

owe thanks to Huba Bartos, Katalin E. Kiss and Chris H. Reintges. I am
also grateful to Zsuzsa Kertesz for correcting my English. All errors are, of
course, my own.

Notes

1. e.g. rules like the pronunciation of the ou sequence that has two values: a con-
sonantal [v] and a vocal [u] depending on its position within a syllable. The
transcription is based on Lambdin (1983: χ.) I initiated only one more diacritic
sign: e , which corresponds to the Coptic supralinear stroke. It is used to indi-
cate either a syllabic consonant or a short schwa before the letter in question.
The abbreviations used for grammatical expressions: acc. = accusative; circ. =
circumstantial; conv. = converter; def.= definite; dem. = demonstrative article;
foe. = focus-marker; gen. = genitive; indef. = indefinite; poss. = possessive
article; proN = pronoun; prt. = particle; ps. = present; rel. = relative.
 Genitive constructions in Coptic 15 5

2. Szabolcsi argues for an INFL node in Hungarian noun phrases as early as in

1981, claiming that NPs have a sentence-like structure with agreement features.
3. "if there is a leading technical idea in Minimalism, it is that movement is a last
resort, taking place only when triggered by a driving force" (Lasnik 1999: 2.)
4. As Huba Bartos pointed out to me (pers. comm.) concord is an "internal" cor-
relation between potentially more than two items - sharing in a feature that is a
property of the whole category containing them, unlike in the case of sentence-
level agreements, where we always find "external" correlation between two
items. So the comparison of concord and agreement raises some theoretical
problems.
5. The nucleus of the article phrase is the article itself, expressing grammatical
and referential information, the expansion element expresses lexical content.
(Layton 2000: §43) He treats indefinite and definite determiners in the same
way, without making the obvious distinction, otherwise relevant in syntax, be-
tween them. In an earlier study he speaks about "determinator syntagm"
(Layton 1990).
6. In the examples below the -f element i n p e f - is a 3. sg. m. marker.
7. Converter is the traditional name for a group of Coptic complementisers. The
two converters in (8) and (9) show some syntactic difference: after the circum-
stantial converter a resumptive pronoun always has to be present in subject po-
sition, while in the case of the relative clause - in well definable conditions -
a null-form can be attested.
8. The variant "m- is the result of a simple phonological rule, namely, that before
bilabial stops (p, m) η becomes m, and on certain morpheme boundaries this
assimilation was reflected in the spelling.
9. Shisha-Halevy (1986) seems sceptical whether demonstrative-modified nouns
fall under this case. Opinions vary in this respect.
10. Prepositions in Coptic have two distinct forms depending on the category
(noun vs. clitic-like personal pronoun) of the following complement. This fact
will be relevant when arguing that the n- introducing the possessor phrase is
not a preposition.
11. But in that case it behaves like a preposition having two forms, one for
prenominal and one for prepronominal use.
12. All of them from the same text (The Life of Apa Onnophrios, Till, 1986: 282)
13. Attributive constructions show certain variation, the discussion of which does
not concern us here. See Layton (2000, §§96-103)
14. Lyons (1986, 140) introduces a new distinction: languages of the above type
are determiner-genitive (DG) languages, while languages like Italian, where
genitives can co-occur with the definite and also with the indefinite article (il
mio libro, un mio libro) are adjectival-genitive (AG) languages. For further
details: Lyons (1986 and 1999, Ch. 8.).
15. More precisely, the genitive is in SpecDP and the article in D°, but only one of
the two positions can be filled within the phrase.
156 Barbara Egedi

16. Adjectives can precede their nouns but its conditions and problems do not con-
cern us here. I suppose for now that the Ν Adj sequence is the unmarked one.
17. About one hundred nouns have a distinct plural form as well (according to
Layton to express the category of individual concrete plurality, Layton 2000:
§ 108b).
18. From son 'brother' in singular
19. From sabe 'wise' in singular (Lambdin 1983: 58)
20. This analysis follows that of the Italian determiner le in Carstens (2000: 329)
21. In reality, Coptic uses the infinitive form of the verb in both constructions. An
alternative analysis for DPs supposes an nP "shell" above the core NP - fol-
lowing the VP-shell hypothesis - and its specifier provides the position where
the possessor or the agent of derived nouns can be generated (Carstens 2000;
Radford 2000). This analysis seems to be redundant for Coptic because of the
above-mentioned infinitival character of the nominal constructions. One hardly
finds any deverbal derived noun with its full argument structure.
22. Ps 113(114): 1. (Layton 2000: §105)
23. Matt 26:12. (Layton 2000: §105) The personal pronoun is in canonical object
position. Its realization is clearly different from that of pronominal possessors
of simple nouns lacking arguments. (25b) would be grammatical only with this
latter type.
24. Ritter 1991, 39 (2b)
25. There is a general uncertainty concerning the generation of adjectives: are they
adjunctions or generated in SpecXP? For the moment I accept the common as-
sumption that they are adjoined to a maximal projection. About this problem
and a different point of view: Cinque (1995)
26. Radford (2000); this view is supported by the fact that in Hungarian (and in
Turkish) there is a visible agreement inflection on the possessed noun. Cf. E.
Kiss (2000)
27. E. Kiss (2000: 134). The Hungarian noun phrase has a relatively complex struc-
ture with real agreeing properties in genitive constructions. Its demonstration
would need a more detailed analysis than seems relevant to our understanding
of Coptic NPs. Beyond a desultory comparison, there seem to be more differ-
ences than similarities between the constructions of the two languages. (For
references, see: Szabolcsi 1994; E. Kiss 2000, 2002: Ch. 7. among others.)
28. Polotsky (1960: §33) exhaustively enumerates the possible subject types, but
these cases are practically all indefinite.
29. From p.Beroliensis 8502, 26, 20. after Till 1986, § 113
30. After Reintges (2002, 350. (9))
31. And it freely receives a default possessor interpretation (Ritter 1988: 921). In
Ritter (1991) she modifies her theory, and supposes two sources of sei: either a
realization of case assigned by Ν to an argument inside NP or the head of a DP
adjunct. (Ritter 1991: 48) The Coptic "nte-phrase does not seem to appear for
expressing an argument of the infinitival nominal constructions.
 Genitive constructions in Coptic 1 57

32. In the theory of Cinque (1995) every adjective and attributive modifier in the
phrase has its own projection with the adjectives in the specifier position and
these projections can be ordered hierarchically. If there were also a so-called
c
«/e-projection, it could be positioned directly over the NP.

References

Abney, Steven
1987 The English noun phrase and its sentential aspect. Ph. D. diss., Cam-
bridge, Mass.: MIT.
Bernstein, Judy
1991 DPs in French and Walloon: evidence for parametric variation in
nominal head movement. Probus 3(2): 101-126.
Carstens, Vicki
2000 Concord in Minimalist Theory. Linguistic Inquiry 31:319-355
Chomsky, Noam
1993 A minimalist program for linguistic theory. In The Minimalist Program,
1995, Noam Chomsky, 167-217. Cambridge, MA: MIT Press.
1995 Categories and transformations. In The Minimalist Program, 1995,
Noam Chomsky, 219-394. Cambridge, MA: MIT Press.
Cinque, Guglielmo
1995 On the evidence for partial N-movement in the Romance DP. In Ital-
ian Syntax and Universal Grammar, Guglielmo Cinque (ed.), 2 8 7 -
309. Cambridge: Cambridge University Press.
E. Kiss, Katalin
2000 The Hungarian noun phrase is like the English noun phrase. In Papers
from the Pics Conference, Approaches to Hungarian 7, Alberti Gabor
and Kenesei Istvän (eds.), 121-149. Szeged: JATE Press.
2002 Hungarian Syntax. Cambridge: Cambridge University Press.
Giorgi, Alessandra and Giuseppe Longobardi
1991 The Syntax of Noun Phrases: Configuration, Prameters, and Empty
Categories. Cambridge: Cambridge University Press.
Grimshaw, J.
1991 Extended projection. Ms., Brandeis University.
Lambdin, Thomas O.
1983 Introduction to Sahidic Coptic. Macon: Mercer University Press.
Lasnik, Howard
1999 Minimalist Analysis. Oxford: Blackwell.
Layton, Bentley
1990 The Coptic Determinator Syntagm and its Constituents. Journal of
Coptic Studies 1: 79-97.
158 Barbara Egedi

Layton, Bentley
2000 A Coptic Grammar with Chrestomathy and Glossary. Sahidic Dialect.
Wiesbaden: Harrassowitz Verlag.
Loprieno, Antonio
1995 Ancient Egyptian. A Linguistic Introduction. Cambridge: Cambridge
University Press.
Lyons, Christopher
1986 The syntax of English genitive constructions. Journal of Linguistics
22: 123-143.
1999 Definiteness. Cambridge: Cambridge University Press.
Polotsky, Hans Jacob
1960 The Coptic conjugation system. Orientalia 29: 392-422.
1987 Grundlagen des koptischen Satzbaus. Erste Hälfte. American Studies
in Papyrology 28. Decatur, GA: Scholars Press.
Radford, Andrew
2000 NP shells. Essex Research Reports in Linguistics 33: 2-20.
Reintges, Chris H.
2001 Agreement marking, case assignment and the composition of the Coptic
clause. Göttinger Miszellen 180: 97-102.
2002 A configurational approach to Coptic second tenses. Lingua Aegyptia
10: 343-388.
Ritter, Elisabeth
1988 A head-movement approach to construct-state noun phrases. Linguistics
26: 909-929.
1991 Two functional categories in noun phrases: Evidence from modern
Hebrew. In Perspectives on phrase structure: heads and licensing,
Susan Rothstein (ed.), 37-62. (Syntax and semantics 25.) San Diego:
Academic Press.
Shisha-Halevy, Ariel
1986 Coptic Grammatical Categories. Structural Studies in the Syntax of
Shenoutean Sahidic. (Analecta Orientalia 53.) Roma: Pontificium In-
stitutum Biblicum.
Siloni, Tal
1991 Noun raising and the structure of NPs. MIT Working Papers in Lin-
guistics 14: 255-270.
1996 Hebrew noun phrases. In Parameters and Functional Heads: Essays in
Comparative Syntax, Belletti, Adriana and Luigi Rizzi (eds.), 239-267.
Oxford University Press.
 Genitive constructions in Coptic 159

Szabolcsi, Anna
1981 The possessive construction in Hungarian: a configurational category
in a non-configurational language. Acta Linguistica Academiae Scien-
tiarum Hungaricae 31: 261-289.
1994 The noun phrase. In The Syntactic Structure of Hungarian. Kiefer,
Ferenc and E. Kiss Katalin (eds.), 179-274. (Syntax and Semantics
27.) San Diego: Academic Press.
Till, Walter C.
1986 Reprint. Koptische Grammatik (Saidischer Dialekt). 2. Aufl. 1961.
Leipzig: VEB Enzyklopädie Verlag.
Vergote, J.
1950 La phrase nominale en copte. In Coptic Studies in Honor of Walter
Ewing Crum, 229-242. (The Bulletin of the Byzantine Institute 2.).
Boston: The Byzantine Institute.
Left-dislocated possessors in Sumerian

Gabor Zolyomi

1. Introduction

Sumerian was spoken in the southern part of ancient Mesopotamia, an area

which roughly corresponds to today's Iraq. Sumerian is not related geneti-
cally to any known language. It is a mainly agglutinative language, charac-
terized by ergativity with a split according to the semantics of the NP and
to the tense and modality of the finite verb. 1 It has a system of grammatical
gender based on a distinction between human and non-human referents. It
is a verb final language, the word order before the verb is determined by the
information structure of the sentence. It is an extinct language, which can
be studied today solely from written sources which were recorded using a
cuneiform writing which was a mixed logographic-phonographic writing
system. 2 The first Sumerian texts which use enough phonographic signs to
facilitate linguistic analysis date from around the middle of the 3rd millen-
nium BC. Sumerian was replaced by Semitic Akkadian as a vernacular
during the first part of the second millennium BC. After around 1600 BC
Sumerian remained to be taught and learnt only for the purposes of the
cultic, literary and scholarly tradition. Sumerian texts were continued to be
written until the 1st c. AD. 3
The present paper aims to discuss two types of genitive constructions of
Sumerian: the anticipatory genitive and the lexical external possession con-
struction. In both constructions the possessor occupies a position before the
head of the NP, which is a marked position of the possessor in Sumerian.
They differ, however, in the case-marking of the possessor. In the anticipa-
tory genitive construction (henceforth, AGC), the possessor is case-marked
with the genitive, while in the lexical external possession construction
(henceforth, EPC), the possessor is case-marked with a case governed by
the predicate. It will be argued that in both of them the possessor is left-
dislocated to topicalize a cognitively accessible but inactive participant.
The difference in their case-marker will be connected with a difference in
the marking of internal and external possessors in Sumerian.
162 Gabor Zolyomi

Section 2 gives a very short overview of the parts of Sumerian grammar

which play a role in the constructions described here. Section 3 describes
the structure and the function of the anticipatory genitive construction.
Section 4 describes the external possession constructions of Sumerian. The
last section analyzes the differences and similarities between AGCs and
lexical EPCs.
In addition to some structural and functional similarities, the two con-
structions show some similarities also in the way they are treated in the
standard grammatical descriptions of Sumerian. They can be described and
explained only with reference to linguistic concepts which are missing
from the inventory of traditional descriptions. Consequently they are either
misinterpreted or overlooked.
The anticipatory genitive is a structurally very visible phenomenon. Its
existence has been therefore noted from the very beginning. Its function is,
however, either not discussed, 4 or explained as some sort of emphasis. A
typical statement is: "we deal here ... with an emphatic way of expression"
(Poebel 1923: 60 [§161]). 5 There also exists an explanation which considers
the anticipatory genitive as an archaic construction reflecting an earlier
word-order of the language. 6 Attinger (1993: 228, 259) was the first con-
necting the function of the anticipatory genitive with the information struc-
ture of the sentence.
As regards the Sumerian external possession construction, its most strik-
ing property is that a pronominal affix of the predicate agrees in person,
gender, and number with the possessor instead of the possessum. Its recog-
nition therefore depends on certain expectations about the structure of the
finite verb. The finite verbs are the most controversial part of Sumerian
grammar. Many of its traditional descriptions simply lack the analytic tools
to recognize the mismatch between the possessum and the pronominal affix
of the predicate. 7
EPCs of Type C (see Section 4.3. below) are typically translated with
constructions like the following: "As regards your middle (i.e., inside), you
are a 'Giant', on your outside, you carry fearsome radiance (Ludwig 1990:
101, 8 a translation of our [21]). Translations like this do justice to the sense
of the constructions but the construction itself is left unexplained and has
never been connected with the other types of EPCs in Sumerian.
 Left-dislocated possessors in Sumerian 163

2. Sumerian in a nutshell

2.1. The Sumerian noun phrase

The Sumerian noun phrase (henceforth, N P ) consists of five structural posi-

tions:

Table 1. The Sumerian nominal template

1. 2. 3. 4. 5.
HEAD MODIFIER POSSESSOR PLURAL-MARKER CASE-MARKER

This apparently simple structure may produce very complicated construc-

tions, primarily due to the range of structural units that may occur in posi-
tion 3 (henceforth, P3). This position, the position of the POSSESSOR, may
be occupied by two kinds of elements: by a possessive pronominal enclitic,
or by a N P in the GENITIVE case. The N P occupying P3 can have elements
in all of its five positions, and then there may be five structural units be-
tween the HEAD and the CASE-MARKER of the main NP:

(1) piama p 3 [pidumu p2zid=p 3 ani=p4ene=p5ak]=p 5 ra

mother son true=POSS.3SG=PL=GEN=DAT.H
'for the mother of his true sons'

The embedded noun in turn may also contain a N P in P3:

(2) P 1 ama P 3 [ P idumu P2 zid p3 [pilugal=p5ak]=p4ene=p5ak]=p 5 ra

mother son true king=GEN=PL=GEN=DAT.H
'for the mother of the k i n g ' s true sons'

As the elements in P4, P5, and the possessive pronominal in P3 are encli-
tics being attached to the final word-level constituent of the NP, all these
elements cumulate at the right end of the phrase in double and triple geni-
tive constructions like (1) and (2).
The Sumerian noun phrase projects the same functional projections in
the same order as has been attested in various present-day languages, e.g. in
Hungarian (cf. Bartos 2000), in such a way that complements preceede
their head, while specifiers follow it. The N P in (1) can be represented by
the tree in Figure l : 9
164 Gabor Zolyomi

K(ase)P

Κ
PossP
fDATl

Poss' Spec

NP Poss

am a zid =ani =ene =ak =ra

mother true =3SG.POSS =PL =GEN =DAT.H

Figure 1. The structure of (1)

The N P in (2) can be represented by the tree in Figure 2:

 Left-dislocated possessors in Sumerian 165

K(ase)P

PossP Κ
IDATI

ama dumu =ak =ene =ak =ra

mother son =GEN =FL -GEN =DAT.H

Figure 2. The structure of (2)

The last surface position of the Sumerian noun phrase accommodates the
case-markers. Ten case-markers can be distinguished in Sumerian: -/o/, -/e/,
-/'a/, -/ra/, -/ta/, -/da/, -/se/, -/ak/, and -/gin/. The case-markers are enclitics
that function to distinguish cases. In Sumerian cases are distinguished,
however, not solely by nominal case-markers, the verbal affixes also play
an essential role in the identification of cases.10 Three of the nominal case-
markers (-/ra/, -/'a/, and -Id), and one of the verbal affixes (/i/ in S9) are
166 Gabor Zölyomi

used as markers of more than one case. On the basis of correspondences

between nominal case-markers and verbal affixes 11 cases can be distin-
guished in Sumerian:

Table 2. Sumerian cases

CASE NOMINAL CASE-MARKER VERBAL AFFIX

human non-human
ERGATIVE -Id -Id final pronominal prefix
(S10)11 and pronominal
suffix (SI3)
ABSOLUTIVE -lei -lei final pronominal prefix (S10)
and pronominal suffix (SI3)
ADVERBIAL CASES

DATIVE -/ra/ -Id /a/ (S6)

OBLIQUE-DIRECTIVE -Ira/ -Id l\l (S9)
OBLIQUE-LOCATIVE -/ra/ -/'a/ l\l (S9)
LOCATIVE — -/'a/ /ni/ (S9)
TERMINATIVE -Is e/ -Isd /si/ (S8)
ABLATIVE — -/ta1 /ta/ (S8)
COMITATIVE -/da/ -/da/ /da/ (S7)
ADNOMINAL CASES

GENITIVE -/ak/ -/ak/ —

EQUATIVE -/gin/ -/gin/ —

The eleven cases can be classified into three groups: i) ergative and absolu-
tive, encoders of the subject and the object, the primary syntactic functions;
ii) adverbial cases; iii) adnominal cases which relate to no corresponding
verbal affixes.

2.2. Sumerian finite verb

Finite verbal forms in Sumerian are distinguished by the large number of

affixes which can be attached to a verbal stem. These affixes and the verbal
stem can be arranged into fourteen structural positions or slots (see Table 3
below). There are ten slots before the stem, which is in slot 11 (henceforth,
 Left-dislocated possessors in Sumerian 167

S i l ) ; and there are three slots after the stem. Unlike in German or English
where among the participants of a verb only the subject is cross-referenced
with an affix on the verbal form, Sumerian verbal forms may cross-
reference up to four participants of the verb.

Table 3. Sumerian verbal template

Slot 1 MODAL prefixes

Slot 2 COORDINATOR prefix
Slot 3 CISLOCATIVE prefix
Slot 4 MIDDLE prefix or 3NH PRONOMINAL prefix ( s p e c i f y i n g the person, gender
and number of the first in the sequence of adverbial prefixes)
Slot 5 INITIAL PRONOMINAL prefix {specifying the person, gender and number
of the first in the sequence of adverbial prefixes)
Slot 6 Adverbial I: DATIVE prefix
Slot 7 Adverbial II: COMITATIVE prefix
Slot 8 A d v e r b i a l III: ABLATIVE or TERMINATIVE prefix
Slot 9 Adverbial IV: LOCATIVE, OBLIQUE-LOCATIVE, or OBLIQUE-DIRECTIVE
prefix
Slot 10 FINAL PRONOMINAL prefix (referring to A or P, depending on the tense)
Slot 11 STEM

Slot 12 PRESENT-FUTURE MARKER (in intransitive verbs)

Slot 13 PRONOMINAL suffix (referring A, S, or Ρ depending on the tense)
Slot 14 SUBORDINATOR

The Agent, the Subject and the Patient of a verb are cross-referenced with
pronominal affixes in S10 and SI 3. The syntactic function of their referents
is indicated mainly by their position, so, e.g., in the present-future conjuga-
tion the pronominal suffix in S13 cross-references A and S, while the final
pronominal prefix in S10 cross-references P. In the case of the other par-
ticipants, the syntactic function is indicated with one of the prefixes in the
ADVERBIAL slots (S6-S9), while the person, number, and gender of the par-
ticipant may be marked with an initial pronominal prefix in S5. The set of
pronominal prefixes in S5 contains no morpheme cross-referencing the 3rd
ps. sg. non-human participant. The pronominal prefix referring to 3rd ps.
sg. non-human participants developed from the middle-marker in S4. The
cislocative prefix in S3 functions as the 1 st ps. sg. pronoun before a dative
(see, e.g. [18] below), an oblique-locative, or an oblique-directive-prefix.
16 8 Gabor Zölyom i

If there is more than one adverbial prefix in a verbal form, then the pro-
nominal prefix in S5 or S4 specifies the person, number, and gender of the
one which is the nearest to it. The other adverbial prefixes refer to a
3rd.ps.sg. non-human participant by default. The verbal affixes cross-
referencing verbal participants function similarly to the pronouns of other
languages, so verbal participants introduced earlier into the discourse may
be encoded only with the verbal affixes in Sumerian.
The string of 14 slots identified by Sumerian philology, represented in
Table 3, suggests a left-branching, head-final sentence structure involving a
CP, an AgrSp, a TenseP, and a VP projection (and probably many more). 12
The subordination suffix in S14 functions as the head of the CP (see below
(8) for an example with this morpheme). The verbal complex is preceded
by focus and topic positions, presumably left-adjoined to CP.

CP

AgrSP C

TenseP AgrS

Figure 3. The Sumerian sentence structure

3. The anticipatory genitive

The possessor (henceforth PR) normally occupies P3 of the NP. Under

certain conditions, however, the PR may have a position before the head of
the NP, presumably before the projection of the possessed Ν P. Sumerian
has two type of genitive constructions with the PR left to the head: (i) an-
ticipatory genitive construction; (ii) external possession construction. This
section discusses the structure and the function of AGC, while the EPC will
be the subject of the next section.
 Left-dislocated possessors in Sumerian 169

In an AGC the PR positioned before the head of the NP is in the genitive

case, while P3 is occupied by a pronominal enclitic which is co-indexed
and therefore agrees in person, gender, and number with the PR. A simple
example is (3) and (4) (the anticipated PR and the co-referent resumptive
pronominal enclitics are underlined in the glosses):

(3) Ent. 28 5:12-13 1 3

nam-nurt-da-ki-gar-ra, ur2-bi na4-a
pi namnundakigara=p s ak p,ur=p3bi=p50 na='a
GN=GEN base=POSS.3NH=ABS stone=LOC
mu-na-ni-du3
S3 mu- S5 nn- S6 a- S9 ni- S10 n- S , idu- s 130
CISL-3SG-DAT-LOC-A.3SG-build-P.3NH
'He built him the base of the Namnunda-kigara from stone'

In (3) the possessive pronominal enlitic -/bi/ agrees in person, gender, and
number with the left dislocated non-human PR "Namnunda-kigara", which
is marked with the genitive case-marker -/ak/. In (4) the possessive pro-
nominal enlitic -/ani/ agrees in person, gender, and number with the left
dislocated human PR "Ur-tukula", which is marked with the genitive case-
marker -/ak/.

(4) TCS 1 36:

Ur-glstukul-ka, gu4-a-ni ga-na-ab-zig3
pi urtukulak=p 5 ak P1gu=p3ani=P50 siga-s 5 nn- s6 a- si() b- sn zig
PN=GEN OX=POSS.3SG=ABS MOD-3SG-DAT-P.3NH-issue
Ί want to issue Ur-tukula's ox for him.'

The PR and the possessum (henceforth, PM) may be separated by another

argument of the verb like in (5) and (6) below. In (7) the PM functions as
the possessor of another NP. In (8) the PM is the participant of a relative
clause whose head functions as the object of another main clause. The left
dislocated PR "Gudea, the ruler of Lagas" precedes the head noun of the
relative clause and stands at the beginning of the whole main clause.
170 Gäbor Zölyomi

(5) GudeaCyl. A 17:11 (2.1.7)

e2-a ^en-ki-ke4 gis-hur-be2
p,e=p5ak enkik=e P1gishur=P3bi=p5e
temple=GEN DN=ERG plan=3NH.POSS=OD.NH
si mu-na-sa2
si=0 s3 mu- s5 nn- sft a- s9 -j- s 1()n-s,, sa- s , 3 0
horn=ABS CISL-3SG-DAT-OD-A.3SG-straighten-P.3NH
'The god Enki put right the design of the temple for him.'

(6) Bar-Am 366/p 1-3 1 4

Ur-^ul-pa-e3-ka, ensi2-ke4, gu4 apin l-am3
pi ursulpaek=p 5 ak ensik=e gu apin=ak l=am-0
PN=GEN ruler=ERG ox plough=GEN > su-ne 2 ba-an-sum2
pisu=p3ani=p5e S4 b- S6 a- Sio n- S11 sum- S13 0
hand=3SG.POSS=DAT.NH 3NH-DAT-A.3SG-give-P.3NH
'The ruler entrusted (lit.gave to his hand) Ur-Sulpae with one
plough-ox'

(7) GudeaCyl. A 17:11 (2.1.7)

d
en-lil2-la2 lu2 sa3-ga-na-kam
Pl enlil- P5 ak P1lu
p3[P1sag-p3ani-p5ak]-P50-am-0
DN-GEN man heart-3SG.POSS-GEN-ABS-COP- S.3SG
'He is a man of Enlil's heart = he is a favourite of Enlil.'

(8) Gudes Statue Β 8:39-42

gu3-de2-a, ensi2, lagask'-ka,
P l gudea P1 ensi [n[pilagas=p^ak]=p^ak
PN Ruler GN=GEN=GEN
lu2 inim-ni ib-kur2-a
P1 lu p2[piinim=p3ani=p50 sinib-snkur-^ae-s^'a]^^
man word=3SG.POSS=ABS P.3NH-change-A.3SG-SUB=ABS
'the man who changes the orders of Gudea, the ruler of Lagas

The Sumerian AGC, the left-dislocation of the PR, is a pragmatically moti-

vated structural device whose primary function is to announce a new topic or
to mark a shift from one topic to another by promoting a participant from a
cognitively accessible but inactive state to an active state in the discourse.
 Left-dislocated possessors in Sumerian 171

Having been promoted, the topical participant will be expressed by a pro-

nominal which is the most preferred topic expression for an active topical
participant (see Lambrecht 1994: 172-184). Consider the following exam-
ple of an A G C :

(9) G u d e a C y l A 29: 1 4 - 1 7 ( 2 . 1 . 7 )
e2-a ni2 gal-bi,
P1 e= P5 ak P1 ni P2 gal= P3 bi=p 5 0
house=GEN fear great=3NH.POSS=ABS
kalam-ma mu-ri,
kalam=' a S3 mu- S9 n- S11 ri- S13 0

land=LOC CISL-LOC-settle-S. 3NH

ka-tar-ra-bi, kur-re ba-ti
pi katara= p3 bi=p50 kur=e S4 b- S6 a- S11 ti- S13 0
praise=3NH.POSS=ABS highlands=DAT.NH 3NH-DAT-reach-S. 3NH
' T h e h o u s e ' s great awesomeness settles upon the Land, its praise
reaches to the highlands.'

In (9) the first clause contains an A G C introducing the " h o u s e " as the topic.
In the following clause the same participant remains the topic, but being
cognitively active it is expressed only with a pronominal enclitic on the PM.
The announcement of a new topic as the function of the A G C is espe-
cially clear in (4) and (6). These examples come f r o m an administrative
letter and a legal document respectively, which represent a simpler dis-
course situation than the royal inscriptions and the literary texts. (4) is the
first sentence in the message part of a letter, preceded only by the address
formula. The participant functioning as the PR, "Ur-tukula", must be there-
fore in a cognitively inactive state. The sender of the letter, however, must
have assumed that the receiver can identify it, so it should be cognitively
accessible. The same applies to (6), the very first sentence of a legal docu-
ment about Ur-Sulpae.
Syntactically, the left-dislocated PR is in one of the topic positions situ-
ated at the left periphery of the sentence. 1 5 The fact that the left-dislocated
PR is associated with a resumptive pronoun in the projection of the posses-
sor suggests that it is generated in its adjoined position, involving no
movement. This assumption is corroborated by (8) above, where the re-
sumptive pronoun is inside a relative clause showing that the topicalized
P R does not have to observe subjacency in AGCs.
172 Gabor Zolyomi

4. External possession constructions

External possession is the grammatical phenomenon that in a semantic

possessor-possessum relationship the PR is expressed externally to the
constituent that contains the PM, and the PR occurs as a separate clause-
level constituent that is not required by the verb's argument structure. A
familiar manifestation of external possession is the following German ex-
ample:

(10) Die Mutter wusch dem Kind die Haare.

'The mother washed the child's hair'

In (10) the PR ("das Kind") is expressed by an N P external to the PM ("die

Haare") and is case-marked with the dative case. 16 The PR is human and
the PM is inalienable, both of which are kinds of nouns cross-linguistically
most accessible to external possession constructions (henceforth, EPC) (see
Payne and Barshi (eds.) 1999 (a): 14).
Grammatical research of the last two decades has shown "that there is
no geographical area of the world where th[is] phenomenon does not occur,
and it is hardly exotic. Its sheer ubiquity suggests it must be a linguistically
natural phenomenon, serving some central human communicative need —
as equally central as that served by active-passive-antipassive choices,
causativization, and a host of other well-documented alternations." (Payne
and Barshi (eds.) 1999 (a): 6) The phenomenon shows a large range of
typological variation in its actual manifestation across the languages of the
world. Languages may differ in the way the external possessor is coded, in
the type of accessible possessa in terms of their grammatical function
and/or semantics, in the range of predicate types allowing external posses-
sion, and in the kind of accessible possessors (see Payne and Barshi (eds.)
1999 (a): 6-14).

4.1. External possession constructions of type A

Zolyomi (1999: 231-237) showed that EPCs are also present in Sumerian.
Consider the following example (the external PR, the co-indexed resump-
tive pronominal enclitic, and the cross-referencing verbal pronominal and
adverbial prefix are underlined in the glosses):
 Left-dislocated possessors in Sumerian 173

(11) Ean. 6 4: 13-15

d a
nin-gir2-su-ra, -~sa&*asag ki ag2-ni,
P1 ningirsuk= P 5ra asag kiag=ani=0
DN=OL.H field beloved=POSS.3SG=ABS
su-na mu-ni-gi4
n
P1 su= p3 ani=p5' a S3 mu- S5 nTi- S9 i- s ιo "s11 gi"s130
hand=POSS.3SG=OL.NH CISL-3SG-OL-A.3SG-return-P.3NH
" H e ( = E a n n a t u m ) has returned (lit. 'returned to his h a n d ' )
his beloved field to Ningirsu"

In (11) the P R of the word su " h a n d " is the god Ningirsu. The P R is, how-
ever, not in P3 of the N P whose head is the word "hand", but in a sentence
initial position. P3 is filled with a pronominal enclitic co-indexed and
therefore agreeing in gender, person, and number with the PR. The PR is
cross-referenced with the verbal prefix l\l of S9 glossed as oblique-locative
here. The pronominal prefix preceding the oblique-locative prefix in S5 is
co-indexed with the PR, but not with the PM. The PR and the PM are case-
marked with the case-markers -/ra/ and -/'a/ respectively which correspond
to the human and non-human markers of the oblique-locative case.
The same verbal expression is used in (12), but here the word " h a n d "
has no possessor. Accordingly the pronominal prefix in S4 agrees in gen-
der, person, and number with the word "hand".

(12) A W L 81 5: 1 - 4
en-ig-gal, nu-banda3, e2 zag iri-ka-ka,
eniggal nubanda=e e zag iri=ak=ak= > a
PN overseer=ERG house border city=GEN=GEN=LOC
su-a bi2-gi4
su= 5 a s4b-s9i-sion-sngi-si30
hand=OL.NH 3NH-OL-A.3SG-return-P.3NH
'Eniggal, the overseer has delivered (lit. 'returned to h a n d ' ) them
(= various sorts of wood) in the house at the border of the city.'

Examples (13)—(17) show further examples of EPCs. In all these examples

the left-dislocated PR and P M are case-marked with the same case. This
can be taken for granted even if the PR is not present overtly (as in [ 1 4 ] -
[17]) because the verbal prefix cross-referencing the PR confirms its case.
Also, in all these examples the verbal pronominal prefix is co-indexed with
174 Gäbor Zolyomi

the PR but not with the PM as is usual. This type of EPC will be referred to
as Type A EPC.

(13) Ean. 1 4 : 2 4 - 2 6
Min-hur-sag-ra, dug3 zid-da-na,
ninhursag=ra dug zid-a=ani= 'a
DN=OL.H knee right-PT=POSS.3SG=OL.NH
mu-ni-tus
s3mu-s5nn-s9i-sion-siitus-si30

CISL-3SG-OL-A3.SG-sit-P.3SG
'She (= Inana) has made him (= Eanatum) sit on the right knee of the
goddess Ninhursag.'

(14) AWL 183 3: 1 - 4

subur, nu-banda3, gu2-ne-ne-a,
subur nubanda=e gu=anene='a
PN overseer=ERG neck=POSS.3PL=OL.NH
e-ne-gar
S5 enne- S9 j,- Sio n- S11 gar- S13 0
3PL-OL-A.3SG-put-P.3NH
'Subur, the overseer, has entered it (= a given amount of silver) as
their (= 3 fishermen) debt (literally, "put them on their neck").'

(15) En. 129 10: 1 - 2

e2-sag4 ni2-ga2-se3
e.sag ni=gu=ak=se
house.heart REFL. 1SG-GEN-TERM
mu-se3-gen-na-am6
S3 mu- S5 j- S8 si-Si igen- sl3 0-am-0
CISL-1 SG-TERM-gO-S. 3SG-C0P-S .3SG
'It was the case that he (= the Ummaite) even entered (lit. come to)
the innermost part of my (= Ningirsu's) own temple.'
 Left-dislocated possessors in Sumerian 175

( 1 6 ) En. 1 33 2: 1 3 - 3 : 6

kur-kur su-ni-se3,
kur~kur=0 su=ani=se
country~PL=ABS hand-POSS.3SG-TERM
mu-se3-gar-ra-a
S3 mu- S 5n- S8 si- S9 n- S 1() gar- s 13 0- s 14 ' a-' a
CISL-3SG-TERM-A.3SG-put-P.3NH-SUB-LOC
' W h e n h e ( = L u g a l - U R U x K Ä R ) has given all the
foreign lands into his ( = Enannatum's) hands.'

(17) Luzag. 1 1: 4 4 - 4 5

kur-kur giri3-na,
kur~kur=0 giri=ani='a
country~PL=ABS foot-POSS.3SG-OL.H
e-ni-se3-ga-a
S5 enn- S9 i- S1() n- S11 seg- S13 0- S i4 a- a
3SG-OL-A.3SG-throw-P.3SG -SUB-LOC
' ( w h e n ) h e ( = Enlil) made all the countries serve him
( = L u g a l z a g e s i ) (lit. threw all the countries to his f e e t ) . '

4.2. External possession constructions o f type Β

Examples ( 1 8 ) and ( 1 9 ) b e l o w represent a slightly different type o f EPC. In

these e x a m p l e s the PR and the PM are case-marked with different cases:
the PR is case-marked with the dative, while the PM is case-marked with
the locative. The different cases o f the PR and PM are to be explained with
the semantics o f the locative in Sumerian: this case can only be used with
non-human nouns, the locative verbal prefix always stands without a pro-
nominal prefix in the verbal prefix-chain. Consequently, the human exter-
nal possessor cannot be cross-referenced with the locative, it can be marked
only with another case. 1 7 External possessors case-marked with the dative
are cross-linguistically a c o m m o n type, especially in Europe. 1 8 Haspelmath
(1999: 125) pointed out that "the external possessor is semantically related
to other typical dative uses, such as the marking o f a beneficially affected
participant (benefactive) ...".
Only one e x a m p l e with an overt PR is known from this type, (20); and
except for (20), all attested PRs are in the 1st or 2nd ps. The German trans-
176 Gäbor Zolyomi

lations are provided with (18)-(19) to demonstrate how easily these exam-
ples translate with German EPCs. This type of EPC will be referred to as
TypeBEPC.19

(18) Isme-Dagan A (2.5.4.01) 90

c
iitu nig2-si-sa2 inim gen6-na
utu=e nigsisa inim gen- 1 a=0
DN=ERG justice word firm-PT=ABS
ka-ga2 ha-ma-ni-in-gar
ka=gu= 1 a s iha- s3 m- s6 a- s9 ni- s 10 n- s , ,gar- s , 3 0
mouth=POSS. 1 SG=LOC MOD-CISL-DAT-LOC-A.3SG-put-P.3NH
'The god Utu put justice and reliable words in my mouth.'
'Utu legte mir Gerechtigkeit und zuverlässige Worte in den Mund'.

(19) Isme-Dagan Q (2.5.4.17) 5

gidru kug su-za
gidru kug=0 su=zu= 'a
sceptre holy=ABS hand=POSS.2SG=LOC
ma-ra-ni-in-ge-en
S3 ma- S5 r- S6 a- S9 ni- S 1{)n-s n g e n - s l 3 0
CISL-2SG-DAT-LOC-A.3SG-firmly.put-P.3NH
'He has firmly put the holy sceptre in your hand.'
'Er legte dir das heiligen Zepter fest in die Hand.'

(20) Enmerkar and the lord of Aratta (1.8.2.3) 433-434

en-e kig2~gi4-a kur-se3 du-ur2
en=e kiggia kur=se du=ra
lord=ERG messenger mountain=TERM go.PF-DAT.H
gidru su-na
gidru=0 su=ani= 1 a
sceptre=ABS hand=POSS.3SG=LOC
mu-un-na-ga2-ga2
S3mu-S5nn-S6a-S9n-S1,ga~ga-S13e
CISL-3SG-DAT-LOC-put~PF-A. 3SG
'The lord placed the sceptre in the hands of the messenger going
to the mountains.'
 Left-dislocated possessors in Sumerian 177

Type A and Β EPCs are similar in that their predicate is a verb denoting a
motion event with a spatial endpoint. In their literal meaning these verbs
imply a physical contact. The PM of these EPCs is marked with one of the
adverbial cases whose meaning implies a motion towards an entity.

4.3. External possession constructions of type C

Examples (21)—(23) represent yet another type of EPC in Sumerian. They

differ from the previous examples in having a non-verbal predicate. The
predicate of these examples consists of the copula and a NP or a non-finite
verbal form. The subject of these copular clauses is possessed by a 2nd ps.
sg. participant expressed by a pronominal enclitic. 20 The copula agrees in
gender, person, and number with the PR referred to by the pronominal en-
clitic and not with 3rd ps. sg. subject as is usual. This type of EPC will be
referred to as Type C EPC.

(21) Isme-Dagan W (2.5.4.23) A38-40

sag4-zu u!8-ru-me-en, bar-zu,
sag=zu=0 uru=me-en bar=zu=0
heart=POSS.2SG=ABS huge=COP=S.2SG outside=POSS.2SG=ABS
ni2 gur3-ru-me-en
ni gur-ed=me-en
fear carry-PF=COP-S.2SG
'Your (= Nibru) interior is enormous, your exterior is imbued with
fearsomeness'

(22) A 7479 iv:3'^t' 2 1

ur2-zu he2-gal2-la he2-me-en,
ur=zu=0 hegal=am=0 he-me=en
root-POSS.2SG-ABS abundance-COP-S.3SG
pa-zu girizal x (KAxNI)-am 3 he2-me-en
pa=zu=0 girizal=am=0 he-me=en
branch-POSS.2SG-ABS luxuriance-COP-S.3SG MOD-COP-S.2SG
'It is abundance that your (= a temple) roots indeed are, it is luxuriance
that your branches indeed are.' 2 2
178 Gäbor Zolyomi

(23) Iddin-Dagan D (2.5.3.4) 30

igi-zu hus-me-en
igi=zu=0 hus=me-en
face=POSS.2SG=ABS awesome=S.2SG
za-pa-rag2-zul rmah?-me-enl
zapag=zu=0 mah=me-en
cry=POSS.2SG=ABS majestic=COP-S.2SG
'Your face is awesome, your cry is majestic'.

4.4. External possession constructions of type D

A fourth type of EPCs involves verbs of perception or experience. This

type of EPC will be referred to as Type D EPC. The predicate in (24) is a
verb of perception. In (24), the pronominal prefix in S5 before the termina-
tive prefix /si/ of S9 agrees in person and number with the 2nd ps. sg. PR
("your") and not with the 3rd ps. sg. PM ("word"). 23

(24) Isme-Dagan Q (2.5.4.17) 15

inim kug an-gin7 su
inim kug an=gin su=0
word holy sky=EQU hand=ABS
nu-te-ge76-zu-se3
nu-teg-ed=zu=se
NEG-approach-PF=POSS.2SG=TERM
igi-bi su-mu-e-si-gal2
igi=bi=0 S1 su- s3 mu- s5 e- s8 si- si0 b- si , gal- s , 3 0
eye=POSS.3PL=ABS MOD-CISL-2SG-TERM-A.3N-be.at-P.3SG
'They always pay attention to your holy words, which, like the
heaven, can never be understood fully.'

Type C and D EPCs differ from the Type A and Β EPCs in that their predi-
cate does not imply any tangible affectedness. All occurrences of these two
type come from literary texts or incantations in which a deity, a city, a tem-
ple, or a deified object is praised and is addressed directly in 2nd ps.
 Left-dislocated possessors in Sumerian 179

4.5. The shared features of external possession constructions in Sumerian

All four types of EPC are characterized by inalienable possession, which

comprises "either inextricable, essential or unchangeable relations between
'possessor' and 'possessed' - that is relations over which possessors exer-
cise little choice or control." (Chappel and McGregor 1996: 4). In the great
majority of the attested EPCs the PM is a body part or part of inanimate
wholes, which are cross-linguistically the most accessible type of possessa
in EPCs.
Haspelmath (1999: 119) classifies the EPCs in non-European languages
into three broad categories in terms of the grammatical coding of the PM in
an EPC: 24 (i) EPCs with PM demotion; (ii) EPCs with PM incorporation;
(iii) EPCs with applicative marking on the verb. Type A, C, and D EPCs of
Sumerian belong to category (i). In the Sumerian constructions the PM
retains its case, but it is not cross-referenced by a bound pronominal on the
verbal form or the copula. It is therefore demoted in the sense that it will no
longer belong the core participants which should be cross-referenced in the
verbal prefix-chain or on the copula. What is common in all types of EPCs,
including Type B, is, however, the promotion of the PR: it will be the PR
but not PM which is cross-referenced with a pronominal on the predicate.
To sum it up, the most important licensing condition of EPCs in Sume-
rian appears to be the semantics of the PM: external possession is possible
only in the case of inalienable possession. The PRs are as a rule human,
often in the 1 st or 2nd ps. The case of the PM and the type of the predicate
do not seem to play a role. No PM functioning as the agent, the object, or
indirect object is, however, attested with external PRs.
Without informants or relevant contrasting examples it is difficult to see
what exactly influenced the choice between an EPC and an internal posses-
sive construction. Examples like (6) above which involve a body part and
yet do not use an EPC may suggest that inalienable possession was only a
necessary but not a sufficient condition. It seems likely that the affected-
ness of the possessor might have also played a role, i.e., the possessor had
to be construed as the primarily affected and not his or its part.

5. Lexical external possession constructions and anticipatory genitives

A further distinction can be made among the EPCs in terms of the presence
or the absence of a left-dislocated lexical PR. EPCs where the external PR
is overtly present as a lexical N P will be referred to as lexical EPCs, while
180 Gäbor Zölyomi

EPCs where the PR is present only as a bound pronominal on the PM and

on the verbal form or the predicate will be referred to as pronominal EPCs.
Lexical EPCs and AGCs show a number of structural similarities. In
both constructions the PR is left-dislocated, and a pronominal enclitic is
attached to the PM co-indexed with the PR. Both constructions present a
structural configuration in which the PR is no longer behind the PM, but is
external to the projection of the possessor occupying a topic position of the
sentence. There exist, however, a number of differences. The PR of an
AGC is marked with the genitive, while the external PR bears the case of
the PM or the dative and is cross-referenced by a pronominal and an adver-
bial prefix on the verb. Consider Table 4 which summarizes the formal
features of the normal and anticipatory genitive construction, and the lexi-
cal EPC.

Table 4. Types of genitive constructions in terms of structure

P R ' S POSITION P R ' S CASE VERBAL CROSS-

REFERENCING

NORMAL behind the PM genitive no

AGC in topic position genitive no
LEXICAL EPC in topic position that of PM, or dative yes
(in type B)

What Table 4 shows is that in comparison to a normal genitive construc-

tion, the AGC and the lexical EPC differ in the number of the distinguish-
ing formal features. The AGC differs from the normal genitive construction
only in the position of the PR, whereas the overt EPC differs in the position
and the case of the PR.
The left-dislocation of the PR in the AGC has been identified after
Lambrecht (1994) as a pragmatically motivated structural device to topical-
ize an accessible but inactive participant. It seems reasonable to assume
that left-dislocation has the same function in the lexical EPC. At the same
time lexical EPCs are a sub-type of the external possession constructions
which share with the pronominal EPCs the promotion of the PR, and the
inalienability of the possession. The formal features of Table 4 may there-
fore be "translated " into the notional features of Table 5 below:
 Left-dislocated possessors in Sumerian 181

Table 5. Types of genitive constructions in terms of notional features

TOPICALITY INTERNAL EXTERNAL

POSSESSION POSSESSION

NORMAL - + -

AGC + + -

LEXICAL EPC + - +

Table 5 suggests that the A G C is a topicalized internal PR. Sumerian thus

appears to distinguish formally between topicalized internal possessors and
lexical external possessors: the former retains its genitive case-marker, while
the latter will be case-marked with a case governed by the predicate. One way
of explaining that in lexical EPCs of Type A, C, and D the left-dislocated
P R bears the same case as the P M is to assume that they represent the same
argument of the verb, forming a kind of an appositive construction. 2 5 The
members of this construction are not, as expected, juxtaposed, because the
topical P R cannot but be located in the sentence initial topic position. The
construction is restricted to cases of inalienable possession because the con-
dition of such an appositive construction is the referential non-distinctness
of its members. The case-marking of EPCs of Type Β follows from the
semantic incompatibility of the locative with human nouns, and from the
cross-linguistically, and also in Sumerian attested use of dative to mark a
beneficially affected participant.
The question of the topicality of the P R in pronominal EPCs still needs
to be addressed. As EPCs seem to signal that instead of his or its part, the PR
is affected by the predicate, i.e. it signals a shift in the aboutness relation
between a participant and the predicate, it is plausible to assume that EPCs in
Sumerian always involve the topicality of the PR. A pronominal E P C thus
appears to differ from a lexical EPC in the cognitive accessibility of the ref-
erent of the external PR. In a lexical EPC the referent of the PR is accessible
but inactive, while in a pronominal EPC it is accessible and active. Their
difference is therefore similar to the difference between e 2 - a n i 2 g a l - b i
"the house, its great awesomeness" and k a - t a r - r a - b i "its praise" in (9).

7. Conclusion

The present paper attempted to show that Sumerian has two different con-
structions with a left-dislocated possessor. In both of them the possessor is
positioned before the head of the possessum in one of the topic positions of
182 Gäbor Zolyomi

the sentence, while in a normal genitive construction the possessor occu-

pies P3 in the N P , a position behind the head noun. They, h o w e v e r , differ
in the case-marking of the left-dislocated possessor. In the anticipatory
genitive construction the possessor is in the genitive case, while in the other
construction its case is governed by the predicate. It w a s shown that their
case-marking is different because the former is a preposed internal posses-
sor, while the latter is an external possessor which is case-marked in Sume-
rian either with the case of the possessum (Type A, C, D) or with the dative
(Type B).

Acknowlegements

The present paper w a s written while I was the holder of a H u m b o l d t Re-

search F e l l o w s h i p between M a y 2003 and March 2 0 0 4 at the Institute of
Assyriologie and Hethitologie in M ü n c h e n . I gratefully a c k n o w l e d g e the
support of the A l e x a n d e r von H u m b o l d t Stiftung, and thank the Institute for
the excellent research conditions.

Notes

1. See Michalowski 1980.

2. For the writing system see Krebernik and Nissen 1994, Wilcke 1994, and
Coulmas 2002. The writing system used for recording Sumerian does not re-
late to the spoken language in the same way as the modern alphabetical scripts.
For a grammatical description it is necessary to keep separate the language and
the medium through which we have access to it. Accordingly, in the examples
quoted in this chapter, the first line represents the utterance in standard gra-
phemic transliteration; the second, a segmentation into morphemes which in-
terprets the first line; the third, a morpheme by morpheme glossing; and the
fourth, a translation. In the graphemic transliteration subscript numerals distin-
guish homophonic graphemes; graphemes that constitute a word are linked by
hyphens, as are enclitics to their host; superscript graphemes are semantic clas-
sifiers. The morphemic glossing follows the conventions of "The Leipzig
Glossing Rules" (http://www.eva.mpg.de/lingua/files/morpheme.html). The
sign "=" links enclitics to their hosts, the sign indicates reduplication.
Three special characters are used in transliterating Sumerian: g (pronounced as
in sing), h (as in loch), s (as in ship).
 Left-dislocated possessors in Sumerian 183

Abbreviations used in the morphemic glosses: A = Agent, the subject of a transi-

tive verb (in the glosses represents its marker); ABS = absolutive case-marker;
CISL = cislocative prefix; DAT = dative case-marker or prefix; DN = divine name;
EQU = equative case-marker; ERG = ergative case-marker; GEN = genitive case-
marker; GN = geographical name; Η = human; LOC = locative case-marker or
prefix; MIDDLE = middle prefix; MOD = modal prefix; Ρ = patient (in the
glosses represents its marker); NH = non-human; OD = oblique-directive case-
marker or prefix; OL = oblique-locative case-marker or prefix: PF = present-
future tense (in the glosses represents its marker); PL = plural marker; PN =
personal name; PR = pronominal enclitic; REFL = reflexive pronoun; s = subject
of an intransitive verb (in the glosses represents its marker); SUB = subordinator
suffix; TN = temple name; LSG = first person singular; 3NH = third person non-
human; 3SG = third person singular human; 3PL = third person plural human.
In the morphemic glosses the initial pronominal prefix is specified according to
person and gender (3SG, 3NH etc.) and the final pronominal prefix according to
person, gender and syntactical role as agent, subject, or object (A.3SG, 0.3N,
etc).
3. See Geller 1997.
4. See Thomsen 1984: 91, Edzard 2003: 39.
5 "Es handelt sich hierbei ... um eine emphatische Ausdrucksweise".
6. See Hayes 1991. There is no evidence supporting Hayes' argument that the
anticipatory genitive was "in the process of becoming limited to certain stock
expressions" (Hayes 2000: 133).
7. See, however, Krecher 1985: 14832 who in connection with our (13) and (14)
and Attinger 1993: 198 (ex. 62) who in connection with our (13) take notice of
the phenomenon which is analyzed here as external possession.
8. "Was dein Mitte betrifft (d.h. innen), so bist du ein 'Riese', an deiner Außen-
seite trägst du furchterregenden Glanz".
9. For the sake of simplicity the structure of noun phrases will continue to be ana-
lyzed with reference to the positions of Table 1, and not to the tree diagrams.
10. In the standard grammars of Sumerian (e.g., Thomsen 1984) cases are identi-
fied according to their nominal case-markers. The system used in this paper
identifies cases on the basis both of their nominal and verbal marking which
results in a considerably different system. For a description of the Sumerian
cases in these terms see Zolyomi 1999: 224-230 (using slightly different la-
bels) and Zölyomi 2005.
11. Verbal structural positions will be referred to as "slots" (= S) throughout this
paper to distinguish them form the structural positions of the noun phrase, re-
ferred to as "positions" (= P).
12. The analysis of the preverbal prefix-chain is controversial and will not be
discussed here as it bears no relevance to the subject of the present paper.
13. The inscriptions of Ean. (= Eanatum), En. I (= Enanatum), Ent. (= Entemena),
and Luzag. (= Lugalzagesi) are numbered here according to Steible 1982. The
184 Gäbor Zölyomi

administrative texts quoted as A W L + no. are n u m b e r e d according to Bauer

1972. T h e letter-order T C S 1 36 is f r o m Sollberger 1966. Literary texts are
quoted after the edition of the Electronic Text Corpus of Sumerian Literature
(Black et al. 1 9 9 8 - ) ; the number in square brackets is the c o m p o s i t i o n ' s cata-
logue n u m b e r in the corpus.
14. The text w a s published in Malul 1989: 146-147.
15. See Zolyomi 1996: 4 5 (exx. 6 2 - 6 3 ) for probable attestations of multiple topics.
16. See H a s p e l m a t h (1999) w h o sees the dative-marked external possessor as a
characteristic feature of a European linguistic area.
17. T h e dative functions as a kind of "substitute" case also in another construction
of Sumerian involving the locative. The adverbial prefixes cross-referencing
the participant in the oblique-locative, the oblique-directive, and the locative
occupy the same slot, S9, in the verbal prefix-chain. W h e n a transitive verbal
form, which otherwise case-marks one of its participants (= X) with either the
oblique-locative or the oblique-directive, also has a participant in the locative,
then X will be case-marked with the dative. The participant in the locative thus
appears to take precedence over that in the oblique-locative or the oblique-
directive to be cross-referenced in S9, and the "ousted" participant will be case-
m a r k e d with the dative similarly to the PR of type Β EPC. See Attinger 1993:
199 (§129c), 2 3 3 - 2 3 4 (§148), 2 8 1 - 2 8 2 (§182b) and Zolyomi 1999: 2 3 8 - 2 4 2 for
these constructions, labelled as "four participant verbal f o r m s " in Zolyomi 1999.
18. See H a p e l m a t h 1999: 125-131.
19. Further examples in literary texts are: "Enki and the world order" (1.1.3), 3 9 1 -
392, " I s m e - D a g a n A " [2.5.4.01] 7 6 - 7 7 , "Ur-Ninurta C " (2.5.6.3) 3 1 - 3 2 , " E n -
lil-bani A " (2.5.8.1) 1 4 3 - 1 4 6 and 163-167, " D u m u z i d - I n a n a C " (4.08.26)
S e g m e n t B13, " N e r g a l C " (4.15.3) 4, "Ninazu A " (4.17.1) 23, " T h e temple
h y m n s " (4.80.1) 1 2 - 1 3 .
20. T h e subject is in the absolutive case which is normal with the copula. T h e only
e x a m p l e s w h e r e the case of the P M in a Type C E P C seems to be different
f r o m the absolutive c o m e f r o m the literary composition Sulgi R (2.4.2.18),
which contains a series of Type C EPCs in 11. 1 0 - 3 9 . Here the possessa are all
written as N O U N - z u - u 3 , which is normally a writing for the m o r p h e m e se-
quence N O U N - z u - e = NOUN-POSS.2SG-ERG or NOUN-POSS.2SG-OD (see
Klein 1990: 97 1 7 7 ), where the case-marker contracts with the /u/ of the pro-
n o m i n a l enclitic resulting in a long /u:/. Neither of these cases are expected
here, so the lenghtening of the vowel remains to be a problem.
21. T h e Old Babylonian tablet containing incantations is edited in Färber and
Färber 2003. The editors read the third sign in 1. 4 ' as s u 6 ( K A x S A ) , but their
translation "an deinen Zweigen bist du ein Barf" (italics are the a u t h o r s ' ) is
considered odd by the authors themselves. The n u m e r o u s passages in which
the w o r d g i r i 17 - ζ a 1 is used in pair with words m e a n i n g " a b u n d a n c e " (e.g.,
h e 2 - g a l 2 , n a m - h e 2 ) (cf. Sjöberg 1963: 1 - 4 ) m a k e the reading g i r i -
z a l x ( K A x N I ) almost certain.
 Left-dislocated possessors in Sumerian 185

22. The enclitic copula after the words h e 2 - g a l 2 and g i r i z a l x are considered
to function here as focus marker. The cleft structure of the English translation
attempts to render this.
23. Another possible example of this type is Gudea Cyl. A (2.1.7) 8: 2 3 - 9 : 4,
which uses the verb ζ u "to know", and the external possessor is marked with
the comitative. See Zolyomi 1999 (a): 183 for the interpretation of this passage.
24. He uses in fact the term 'possessor raising'.
25. Constructions similar to Sumerian lexical EPCs are described by Heine (1997:
158-161) as the result of possessor specification involving the Topic Schema.
Heine states that these "constructions are occasionally described as instances of
possessor-possessee apposition. ... in languages having a system of overt case
marking, such constructions are likely to be characterized by case agreement in
that the possessee (= the specifier) receives the same case marking as the posses-
sor (= the specified). Such a situation appears to obtain, for example in many
Australian languages (cf. Dixon 1980: 293), even if possessor specification of
this type tends to be confined to 'inalienable' possession." (Heine 1997: 159)

References

Attinger, Pascal
1993 Elements de linguistique sumerienne. La construction de dui/e/di
'dire'. (Orbis Biblicus et Orientalis. Sonderband.) Fribourg, Suisse/
Göttingen: Editions Universitaires/Vandenhoeck & Ruprecht.
Bauer, Josef
1972 Altsumerische Wirtschaftstexte aus Lagasch. (Studia Pohl, 9.) Rome:
Biblical Institute Press.
Bartos, Huba
2000 Az inflexiös jelensegek szintaktikai härtere. In Strukturälis magyar
nyelvtan 3: Morfologia [Structural Hungarian Grammar, v. 3:
Morphology], Ferenc Kiefer (ed.), 653-762. Budapest: Akademiai
Kiado.
Black, Jeremy, Graham Cunningham, Jarle Ebeling, Esther Fltickiger-Hawker,
Eleanor Robson, Jon Taylor and Gabor Zolyomi
1998- The Electronic Text Corpus of Sumerian Literature. Oxford,
(www-etcsl.orient.ox.ac.uk)
Chappell, Hilary and William McGregor
1996 Prolegomena to a Theory of Inalienability. In The Grammar of In-
alienability. A Typological Perspective on Body Part Terms and the
Part-Whole Relation, Hilary Chappel and William McGregor (eds.),
3 - 3 0 . (Empirical Approaches to Language Typology, 14.) Berlin/
New York: Mouton de Gruyter.
186 Gäbor Zölyomi

Coulmas, Florian
2002 Writing Systems. An Introduction to Their Lingusitic Analysis.
(Cambridge Textbooks in Linguistics.) Cambridge, etc.: Cambridge
University Press.
Dixon, R.M.W.
1980 The Languages of Australia. Cambridge: Cambridge University Press.
Edzard, Dietz Otto
2003 Sumerian Grammar. (Handbook of Oriental Studies, Sect. I, 71.)
Leiden/Boston: Brill.
Farber, Gertrud and Walter Färber
2003 Vor einem, der auszog, ein g u d u 4 zu werden. In Sallaberger, Volk,
and Zgoll (eds.)2003: 99-114.
Geller, Mark
1997 The Last Wedge. Zeitschrift für Assyriologie 87: 43-95.
Haspelmath, Martin
1999 External Possession in a European Areal Perspective. In Payne and
Barshi (eds.) 1999: 109-135.
Hayes, John
1991 Some Thoughts on the Sumerian Genitive. Acta Sumerologica 13:
185-194.
2000 A Manual of Sumerian Grammar and Texts. (ARTANES 5.) 2nd
revised and expanded ed. Malibu: Undena Publications.
Heine, Bernd
1997 Possession. Cognitive Sources, Forces, and Grammaticalization.
(Cambridge Studies in Linguistics, 83.) Cambridge, etc.: Cambridge
University Press.
Klein, Jacob
1990 Sulgi and Ismedagan: Originality and Dependence in Sumerian
Royal Hymnology. In Bar-Ilan Studies in Assyriology Dedicated to
Pinhas Artzi, Jacob Klein und Aaaron Skaist (eds.), 65-136. (Bar-
Ilan Studies in Near Eastern Languages and Culture.) Ramat Gan:
Bar-Ilan University Press.
Krebernik, Manfred and Hans J. Nissen
1994 Die sumerisch-akkadische Keilschrift. In Schrift und Schriftlichkeit /
Writing and Its Use, 1 (1), H. Günther and O. Ludwig, (eds.), 2 7 4 -
288. Berlin/New York: Walter de Gruyter.
Krecher, Joachim
1985 Die /m/-Präfixe des sumerischen Verbums. Orientalia NS 54: 133-181.
Lambrecht, Knud
1994 Information Structure and Sentence Form. Topic, Focus, and the
Mental Representations of Discourse Referents. (Cambridge Studies
in Linguistics, 71.) Cambridge etc.: Cambridge University Press.
 Left-dislocated possessors in Sumerian 187

Ludwig, Marie-Christine
1990 Untersuchungen zu den Hymnen des läme-Dagan von Isin. (SANTAG,
2) Wiesbaden: O. Harrassowitz.
Malul, Meir
1989 An Ur III Legal Document in the Possession of the Museum of the
Kibbutz of Bar-Am, Israel. Acta Sumerologica 11: 145-154.
McGregor, William
1999 External Possession Constructions in Nyulnyulan Languages. In
External Possession, Doris L. Payne and Immanuel Barshi (eds.),
4 2 9 ^ 4 8 . (Typological Studies in Language, 39.) Amsterdam/Phila-
delphia: John Benjamins.
Michalowski, Piotr
1980 Sumerian as an Ergative Language. Journal of Cuneiform Studies
32: 86-103.
Payne, Doris L. and Immanuel Barshi (eds.)
1999 External Possession. (Typological Studies in Language, 39.) Amster-
dam/Philadelphia: John Benjamins.
1999 (a) External Possession. What, Where, How, and Why. In Payne and
Barshi (eds.) 1999: 3-29.
Poebel, Arno
1923 Grundzüge der sumerischen Grammatik. (Rostocker orientalistische
Studien, 1.) Rostock: Selbstverlag des Verfassers.
Sallaberger, Walther, Konrad Volk and Anette Zgoll (eds.)
2003 Literatur, Politik und Recht in Mesopotamien. Festschrift für Claus
Wilcke. (Orientalia Biblica et Christiana, 14) Wiesbaden: Harrassowitz
Verlag.
Sjöberg, Äke
1963) giri x (= KA)-zal. Zeitschrift für Assyriologie 55: 1-10.
Sollberger, Edmond
1966 The Business and Administrative Correpondence under the Kings of
Ur. (Texts from Cuneiform Source, 1.) Locust Valley, NY: J. J.
Augustin Publisher.
Steible, Horst
1982 Die altsumerischen Bau- und Weihinschriften. 2 vols. (Freiburger
altorientalische Studien, 5.) Wiesbaden: Franz Steiner Verlag.
Thomsen, Marie-Louise
1984 The Sumerian Language. An Introduction to Its History and Gram-
matical Structure. (Mesopotamia, 10.) 3rd ed. Copenhagen: Akade-
misk Forlag.
Wilcke, Claus
1994 Die Keilschriftkulturen des Vorderen Orients. In Schrift und Schrift-
lichkeit /Writing and Its Use, 1, 1, H. Günther and O. Ludwig, (eds.),
491-503. Berlin/New York: Walter de Gruyter.
188 Gabor Zölyomi

Zolyomi, Gabor
1996 Genitive Constructions in Sumerian. Journal of Cuneiform Studies
48: 39^15.
1999 Directive Infix and Oblique Object in Sumerian: An Account of the
History of their Relationship. Orientalia N S 68: 215-253.
1999 (a) Review of D. O. Edzard, Gudea and His Dynasty. Orientalische
Literaturzeitung 94: 178-189.
2005 Sumerisch. In Schriften und Sprachen des Alten Orients, Michael
Streck (ed.), 11-43. Darmstadt: Wissenschaftliche Buchgesellschaft.
Complex predicate structure and pluralised events
in Akkadian
Christian Huber

Introduction

Semitic languages are known to employ fixed morphological templates for

derivation from abstract, in the standard case triconsonantal, roots. Word
forms are built by means of vowel patterns, affixes and special CV-
patterns. For verbs, the resulting forms are usually referred to as stems. In
this paper I will be concerned with some stems in Akkadian and the syntac-
tic structure underlying them.
The Akkadian language with its main dialects Assyrian and Babylonian
is a Semitic language of ancient Mesopotamia. It is preserved in a vast
number of written documents from the mid-third millenium BC until the
first century AD. It is believed, however, that the language ceased to be
spoken around the mid-first millenium BC. Akkadian texts are written in
cuneiform script, which is a mixed syllabic-logographic writing system.
The main writing material was clay, the cuneiform signs were impressed
with a stylus. The texts thus recorded on clay tablets encompass genres of
all kinds (letters, administrative, legal and business documents, literary
compositions, juridical and mathematical works, etc). 1
Contrary to Arabic and Hebrew, the two Semitic languages studied most
in the generative literature, Akkadian is an SOV language and has three
prefix conjugations (traditionally termed present, preterite, perfect) and one
suffix conjugation (traditionally termed stative, a predicative construction
that can also be used with nouns). There are four main stems, named after
their morphological characteristics, viz. the basic G-stem (cf. German
Grundstamm) and the derived D-stem (doubling stem with geminated mid-
dle root consonant), S-stem (ί-prefix), and N-stem («-prefix). The G-stem
corresponds to Arabic form I and Hebrew Qal, the D-stem corresponds to
Arabic form II and Hebrew Pi'el, the S-stem corresponds to Arabic form
IV and Hebrew Hif ( il, the N-stem corresponds to Arabic form VII and
Hebrew N i f ' a l . There is no "internal" passive. From these four main stems
190 Christian Huber

further byforms could be derived, among them iterative stems which are not
found in Arabic, Hebrew, or Ethiopian Semitic. The present paper concen-
trates on the G-stem, D-stem and S-stem, whose morphological properties
are briefly summarised in Table 1, the radicals (i.e. root consonants) are
schematically represented by CiC 2 C 3 (final -u in the infinitive forms is the
nominative case marker).

Table 1.

G-stem D-stem S-stem

basic form gemination of middle radical prefix s

vowel pattern partly de- specific uniform vowel pattern irrespective of (thematic/
pendent on (thematic/lexical) lexical) 'root vowel' (same in D- and S-stem; also same
'root v o w e l ' CV-template) 2
3.m.sg. pret.: iC 1 C 2 a/i/uC 3 3.m.sg. pret.: uC|aC 2 C 2 iC3 3.m.sg. pret.: usaCiC 2 iC 3
(infinitive: C|aC 2 aC 3 -u) (infinitive: C|UC 2 C 2 uC 3 -u) (infinitive: suC,C 2 uC 3 -u)

The D-stem displays an erratic behaviour with different verb types which had
always made it a troublesome task to assign it a uniform function. For ex-
ample, it brings about a transitivity or valency asymmetry, illustrated in (1).

(1) G-stem D-stem S-stem

a. Vrbj rabit rubbü surbü

'to g r o w ' (intrans.), 'to make big(ger)' 'make/let/cause (to)
'to become big(ger)' be(come) big(ger)'

b. V p t ^ petu puttü suptü

'to o p e n ' (trans.) 'to open' (trans.) 'make/let/causey
(to) open jc'

In ( l a ) the D-stem derives a transitive verb from an intransitive verb. In

( l b ) , however, the valency of the derived verb remains unchanged. Appli-
cation of the S-stem, in contrast, is valency-increasing in both ( l a ) and
( l b ) . The asymmetry in (1) raises the question whether the D-stem in ( l a )
is derived in the same way as the D-stem in ( l b ) or whether there are dif-
ferent mechanisms at work in ( l a ) and ( l b ) , for some reason or other still
producing morphologically identical surface forms.
 Complex predicate structure andpluralised events in Akkadian 191

The fact that from a great number of G-stem verbs (or their roots) a D-stem
can be derived systematically according to ( l a ) or ( l b ) (see below footnote
4) suggests that the D-stem does not come about by mere coincidence in
these cases but that it is an important derivational tool of the language.
Synchronically, thus, it would not seem unreasonable to assume that there
are not two distinct morphosyntactic processes at work in ( l a ) and ( l b )
which only happen to share the same morphological guise but that it is the
same process that applies in ( l a ) and (lb).
In the present paper I will mainly be concerned with the asymmetry in
(1) and the properties of predicates that participate in it either in the way of
( l a ) or as in (lb). As I hope to show, general assumptions in generative
grammar about the syntactic structure underlying different verb types to-
gether with certain assumptions about event structure allow to formulate
the properties of the D-stem in a way that accounts for both ( l a ) and ( l b )
while still deriving both ( l a ) and ( l b ) from the same set of properties.
In section 1 I present some data that exemplify the basic syntactic be-
haviour of unaccusative, transitive and unergative verbs in the G-stem, D-
stem and S-stem. On the basis of the observations made in section 1 I make
a proposal in section 2 as to the structural configurations underlying the D-
stem and S-stem and how they differ from each other and from the G-stem.
I suggest, in particular, that the D-stem involves a v-VP configuration (Hale
& Keyser 1993; Chomsky 1995) together with an additional property,
which later will turn out to be verbal plural. Section 3 deals with a number of
verbs or verb classes that undergo a G-stem/D-stem transitivity alternation,
with particular attention paid to the behaviour of so-called psychological
predicates. In section 4 I investigate the properties of D-stem verbs derived
from basic transitives that do not undergo a transitivity alternation and il-
lustrate the phenomenon of verbal plural as recognised e.g. by Poebel
(1939), Greenberg (1991), and Kouwenberg (1997). Section 5 is concerned
with verbal plural and its properties with different types of predicates. I
attempt to make the notion of verbal plural more precise and suggest that
verbal plural should be analysed as eventual plural. I propose an approach
to pluralisation of events inspired by Pustejovsky (1991) that recognises
different kinds of event pluralisation and pluralised events and thereby allows
us to account for the discussed instances of verbal plural in Akkadian as well
as to articulate a difference between verbal plural in the Akkadian D-stem
and in the Arabic and Hebrew intensive stems. The D-stem phenomena
discussed, among them the transitivity asymmetry sketched in (1), are ar-
gued to reduce to the type of event involved, the way of pluralising it and
192 Christian Huber

the fact that the D-stem involves a v-VP configuation. In section 6, finally,
I briefly treat some questions pertaining to the morphological make-up of
the D- and S-stems.
A few remarks need to be added here. The present article is mainly con-
cerned with the G- and D-stem and, to a lesser degree, the S-stem. I will not
deal with other stems. Likewise, I will not deal with the corresponding
stems in other Semitic languages apart from occasional remarks. It must
also be noted that in the stages of the language that are accessible to us
through the texts, the D-stem is not a fully productive grammatical cate-
gory. D-stem forms are sometimes lexicalised to a varying degree and have
acquired specialised meanings. 3 From certain verbs apparently no D-stem
could be derived at all, whereas a few other verbs seem to be attested only
in the D-stem but not in the G-stem. Lexicalised or otherwise exceptional
D-stem instances I largely ignore. 4 Nevertheless I can discuss only a lim-
ited choice of the relevant data. Due to space limitations data will some-
times be presented only schematically. The present article is thus by no
means an exhaustive account of all phenomena relevant in connection with
D-stem and S-stem.
Another point that should be kept in mind is that working with data from
a dead language is always problematic. As native speakers are no longer
available to provide judgements or comments and all evidence comes from
the surviving (and, discovered, evaluated, and accessible) written texts, a
good deal of the finer (or even coarser) grained nuances will almost neces-
sarily escape us and many a question regarding interpretation, productivity,
grammaticality or the like will remain open to speculation.
Finally, I will not speculate on the historical development of the D- and
S-stems.

1. Syntactic effects of D-stem and S-stem with different verb classes

In this section I will briefly illustrate some basic syntactic properties of the
D- and S-stems when applied to intransitive and transitive G-stem verbs.
With intransitives I will make the familiar distinction between unaccusa-
tives and unergatives. Unaccusatives are intransitive verbs whose surface
subjects are underlyingly objects. Unergatives are intransitive verbs whose
surface subjects are also underlyingly subjects. The D-stem is sensitive to
this distinction in that it brings about a transitivity alternation with unaccu-
satives that is not met with transitives and unergatives.
 Complex predicate structure andpluralised events in Akkadian 193

The adjectival roots in (2) surface in the G-stem prefix conjugations as in-
transitive predicates denoting the process that leads to, or increases the de-
gree of, the state described by the respective adjective. Both the D-stem and
the S-stem render 2-participant forms. The D-stem gives rise to transitive
predicates. The S-stem, employing the causative morpheme s, yields causa-
tive predicates.

(2) G-stem D-stem S-stem

V'bl 'abälu 'be(come) dry' 'ubbulu 'dry χ sübulu ' l e t / c a u s e χ (to)

be(come) dry'

V>rk 'aräku 'be(come)/ 'urruku ' l e n g t h e n x' süruku Ί./c. χ (to)

last l o n g ' be(come) long'

Vlbr labäru 'be(come) old' lubburu 'make χ old' sulburu Ί . / c . χ (to)

be(come) old'
5
VmM mädu 'be(come) sum^udu Ί . / c . χ (to)
numerous, much' be(come)
numerous, much'

Vmrs maräsu ' b e ( c o m e ) ill' murrusu ' ( m a k e χ ill=) sumrusu Ί . / c . χ (to)

o f f e n d x' b e ( c o m e ) ill'

Vrps rapäsu 'be(come) wide' ruppusu 'widen x' surpusu Ί . / c . JC to b e ( c o m e )

wide'

Vslm salämu 'be(come) dark' sullumu 'make χ dark' suslumu Ί . / c . χ to b e ( c o m e )

dark'

Vshn sahänu 'become warm' suhhunu ' w a r m , heat x ' —

Vslm salämu 'be(come) sound, sullumu 'make χ sound,

( c o m e t o ) b e in a make χ (come
good condition' to) be/stay in a
good condition'

Vspl sapälu 'be(come) deep, suppulu 'deepen, l o w e r * ' suspulu Ί./c. χ (to) be(come)
low' deep, low'

Vsrh (sarähu) (Stative o n l y : surruhu ' m a k e χ glorious, susruhu Ί./c. Λ: (to) be(come)
be glorious, magnificent' glorious, magni-
magnificent) ficent'

There is no "ergative" alternation in the basic stem. In order to derive a

transitive/causative version from the G-stem unaccusatives in (2), the D-
stem or the S-stem has to be applied.

(3) rapäsu * 'widen x/cause χ to be(come) wide' etc.

salämu * 'make χ dark/cause χ to be(come) dark' etc.
etc.
194 Christian Huber

From the roots in (4) both the G-stem and the D-stem yield transitive predi-
cates featuring a subject and a direct object, whereas the S-stem yields dou-
ble-transitive, causative predicates with the causer surfacing as the subject
in nominative case and the causee and the object surfacing with accusative
case.

(4) G-stem D-stem S-stem

•/btq batäqu 'cut (off) *' buttuqu 'cut(off)*'
y }
VM1 edelu 'lock *' uddulu 'lock*'
3
V'ps 'epesu 'do, make x' uppusu 'do, make*' süpusu 'let/make/cause y
(to) do, make *'
Vrks rakäsu 'bind*' rukkusu 'bind*' surkusu 'let/make/causey
(to) bind *'
Vshp sahäpu 'overwhelm *, suhhupu 'overwhelm*, sushupu 'let/make/causey
lay χ flat' lay * flat' (to) overwhelm *,
lay * flat'
Vsbt sabätu 'seize*' subbutu 'seize*' susbutu 'let/make/cause y
(to) seize*'
Vsql saqälu 'weigh, pay x' suqqulu 'weigh, pay *' susqulu 'let/m./c. y (to)
weigh, pay *'
Vsbr seberu 'break*' subburu 'break*'

With these verbs there is no increase of valency in the D-stem.

(5) rukkusu * 'let/make/cause y (to) bind x'

}
uppusu * 'let/make/cause y (to) make x'
etc.

However, valency is always increased in the S-stem.

(6) surkusu * 'bind*'

susbutu * 'seize x'
etc.

The roots in (7) give rise to verbs denoting body functions or utterances of
sound, thus providing good candidates for unergatives. These verbs pattern
with transitives in that there is no increase of valency in the D-stem but
only in the S-stem. 6
 Complex predicate structure andpluralised events in Akkadian 195

(7) G-stem D-stem S-stem

Vdmm damamu 'wail' _ sudmumu 'let/make/cause^

(to) wail'
vV§ — gu'^usu 'puke' —

Vghb — guhhubu 'cough' —

/gnh ganähu 'cough' gunnuhu 'cough' —

/hbb habäbu 'whisper, hubbubu 'whisper, suhbubu Ί./m./causey (to)

hiss, ...' hiss, ...' whisper, hiss, ...'
\Aibh nabähu 'bark' nubbuhu 'bark' —

Vngg nagägu "cry' —

VsM sa 'älu "cough' su ' 'ulu 'cough' (St: 'be caused to cough')7
Vsrt sarätu 'fart" surrutu 'fart' —

Among the unergatives listed in (7) some appear in both the G-stem and the
D-stem, whereas others are attested only in one or the other. In those cases
where the verb can appear in both stems, however, the D-stem never gives
rise to causativised versions of the G-stem, which can only be derived by
the S-stem.
The comparison of (2) and (4)/(7) shows that the D-stem turns the G-
stem unaccusatives into transitives, while it does not extend the valency of
unergative verbs or verbs which are already transitive in the G-stem. This
sets it apart from the S-stem, which derives causative forms from G-stem
unaccusatives as well as G-stem transitives and unergatives. The resulting
surface patterns are summarised in Table 2.

Table 2.

G-stem D-stem S-stem

unaccusatives (cf. (1)) intransitive => transitive causative ('bivalent')
unergatives (cf. (7)) intransitive => intransitive causative ('bivalent')
transitives (cf. (4)) transitive => transitive causative ('trivalent')
196 Christian Huber

2. Configurations of D-stem and S-stem

Considering the data in section 1 it can be observed that the D-stem is pre-
sent only with an external argument. Since unaccusatives become transitive
in the D-stem whereas the valency of transitives and unergatives is not ex-
tended it can be concluded that the D-stem adds something to the structure
of unaccusatives which allows the introduction of an external argument and
which is already present with basic transitives and unergatives. However,
the fact that the D-stem can also be applied to verbs that already have an
external argument in their basic form suggests that it provides not only
some element that allows for an external argument but that the D-stem has
some additional property.
In sum this indicates that the D-stem is a complex phenomenon that
involves some element which is only present with an external argument but
also involves yet another element or feature. I will consider both issues in
turn. In the remainder of this section I will discuss the configurational as-
pects of the D-stem and the S-stem. What sets apart G-stem transitives
from their D-stem counterparts will be discussed in section 4.

2.1. Configuration of the D-stem

An element that is present only with an external argument is proposed in

Chomsky's (1995) version of Hale & Keyser's (1993) approach to Theta
theory, namely little ν in a v-VP configuration. Chomsky (1995: 315f) states
that "The internal arguments occupy the positions of specifier and comple-
ment of V. Accordingly, the external argument cannot be lower than [Spec,
v]. If it is [Spec, v], as I will assume, then the v-VP configuration can be
taken to express the causative or agentive role of the external argument. It
would be natural to extend the same reasoning to transitive verb construc-
tions in general, assigning them a double-VP structure ... If intransitive
(unergative) verbs are hidden transitives, as Hale and Keyser suggest, then
only unaccusatives lacking agents would be simple VP structures."
According to Chomsky there are two basic configurations for verbs,
depicted in (8).
 Complex predicate structure andpluralised events in Akkadian 197

(8) a. unaccusatives b. transitives, unergatives

VP vP

EXT

VP
(internal structure of VP omitted)

I will therefore propose that in Akkadian, the D-stem always involves a v-VP
configuration. It transfers G-stem unaccusatives into a v-VP configuration,
thereby making them into transitives with an external argument. G-stem
transitives and unergatives, in contrast, already involve a v-VP configura-
tion and an external argument in their basic form. Therefore their configu-
ration remains unchanged in the D-stem and valency is not increased.

(9) G-stem D-stem

unaccusative => transitive
transitive => transitive

The fact that unaccusatives are turned into transitives in the D-stem but the
valency of transitives remains unchanged suggests that in the latter case some
additional property of the D-stem (to be dealt with in section 4) visibly takes
effect, hinting at its complex nature. Configurationally, this complex nature
of the D-stem can be captured in two ways: (i) as involving another func-
tional head in addition to little ν that is responsible for some additional
property of the D-stem, or (ii) as a little ν with some special feature.
There are two options for a v-VP configuration in combination with
another head, depending on whether all non-theta-related functional projec-
tions are required to be above vP (Chomsky) or whether a non-theta-related
functional projection is allowed to intervene between the two portions of a
double VP (e.g. Aspect in Baker 1996, among others).
In (10a) a functional head F selects a vP complement, whose head ν in
turn takes VP as a complement. In (10b) ν selects FP and F selects VP so
that F is sandwiched between the upper and the lower portion of the v-VP
configuration. However, the approach (10c) would also allow for F to take
a VP complement without having been selected by v, resulting in an unac-
cusative VP that is complement of F. This raises the questions of what the
appearance and properties of such a construction would be. The issue will
be returned to in section 4.
198 Christian Huber

Alternatively the properties of ν and F could be collapsed in a single head.

If ν is allowed to bear semantic content beyond what is needed for its role
in Theta theory, the D-stem might involve a little ν that differs from the
little ν of the G-stem (to the general effect that there could be different
types of little v). Resulting structures are those in (11), with unergatives
subsumed under transitives.

(11) a. unaccusative b. transitive c. transitive

with unmarked ν with marked ν (= v D )

vP vDP

EXT EXT vD

VP VP vD VP

In both approaches, crucially, the D-stem involves a v-VP configuration.

2.2. D-stem vs. S-stem

Taking the D-stem to involve a v-VP configuration also accounts for the
difference between the D- and S-stems with respect to their causativising
effect. Little ν is a theta-relevant head that allows the addition of an exter-
nal argument. Causation via ν is relevant only in that part of the structure
that affects the thematic properties of a predicate. In contrast, causation via
the S-stem need not interact with the thematic properties of a predicate. It
may thus embed predicates that appear thematically "closed" in the G-stem
under a causative event and introduce a causer argument, irrespective of
whether or not the embedded predicate contains an external argument. With
 Complex predicate structure andpluralised events in Akkadian 199

transitive and unergative verbs, therefore, the S-stem simply functions as a

causative of the respective G-stem verb, whereas with unaccusatives it may
also replace the D-stem to render a causativised form. Conversely, however,
the D-stem cannot replace the S-stem to render a causativised version of a
G-stem transitive or unergative verb. In this respect the S-stem is somewhat
reminiscent of causative verbs like English let that take a thematically
"closed" predicate as a complement which may be transitive or intransitive
(minimally, vP or VP), i.e. have or not have an external argument. 8
Consider (12) and (13): (12) provides examples of a transitive verb
(saqälum 'pay, weigh o u t ' ) in the G-stem (12a), D-stem (12b) and S-stem
(12c); in (13) examples of an unaccusative verb {rabü 'be(come) big(ger)')
in the G-stem (13a), D-stem (13b) and S-stem (13c) are given.

(12) a. (,..)10siqil kaspam (...) isaqqal

10 shekel.(ACC).abs silver.ACC pay(G).3sgm.pres
' ( . . . ) h e will pay 10 shekel of silver.' ( K H §209)

b. sa saqqulim usaqqil-ma
R E L pay(D).inf.GEN pay(D).pret.3sgm-MA
' H e paid what (there was) to pay.' ( O I P 27: 57 2 5 )

c. (...JawTlam kaspam lusasqil

man.ACC silver.ACC p a y ( S ) . p r e c . l s g c
'(I will t a k e the tablet and) make/have/let the m a n pay the silver!' ( C C T 2 22: 4 6 )

(13) a. summa tulJmum eli miniäti-su irabbi

if spleen.NOM on ( n o r m a l ) size. be-big(G).pres.3sgm
plf.GEN.cstr-possCL3sgm
' I f t h e spleen is larger than usual, ( . . . ) ' ( R A 6 44: 4 3 )

b. me-sunu plqäte urabbi-ma

spring.pi. ACC.cstr-possCL3plm narrow.plf.ACC m a k e - b i g ( = D ) . p r e t . 1 sgc-MA
Ί enlarged their n a r r o w springs.' ( O I P 2 144 viii 35)

c. ekallu ma-hir-ti9 magal usarbi

palace. ACC earlier.sgf. greatly make-big(=S).pret.lsgc
' T h e earlier p a l a c e I greatly enlarged.' ( O I P II 133, 85)

(12a/b) and (13a/b) illustrate once again the G-stem/D-stem valency asym-
metry, (12c) and (13c) show that the predicate embedded under the S-stem
causative may be transitive (in (12c)) as well as unaccusative (in (13c)).
Unergatives, for which I omit examples here (see (7) above), pattern with
the transitives in (12). Since the predicate embedded under the S-stem
200 Christian Huber

causative may be transitive, unergative or unaccusative, i.e. contain an ex-

ternal argument or not, the structure of the S-stem can be approximated as
in (14). 10

(14) sP

In general, the difference in interpretation between D-stem causatives and

S-stem causatives might thus have resembled the semantic difference be-
tween internal and external causation as in English intransitive vs. transitive
sink as opposed to read vs. let read.

(15) a. The boat sank.

b. John sank the boat.

(16) a. Mary read the book.

b. John let/had/made Mary read the book.

In (15b), John is interpreted as the agent or causer involved in the sinking

event. In (16b), John does not directly participate in the reading event but is
involved only in the causing event (i.e. in bringing about the reading event).
Interpreting John as directly taking part in the reading event is blocked by
the intervening agent Mary.
From (2) above it can be observed that unaccusatives may have either
only a D- or a S-stem, or both. Kouwenberg (1997) argues that in the latter
case, unless an instance of dialectal variation, one of the two often is
merely a literary, stylistic variant (with deadjectivals, mostly the S-stem),
since such variation takes place in similar contexts without evident interpre-
tative differences. For example, S-stem forms of rabü 'be(come) big(ger)' as
in (13c) do not seem to occur outside literary compositions. At present it
cannot be said what led the scribes to use also the S-stem with verbs for
which outside literary texts only the D-stem is attested (or vice versa"), and
what factors may have conditioned the choice of forms.' 2 At any rate, it shows
that these scribes could (at least to some extent) make productive use of the
underlying morphosyntactic means. Crucially, however, even in literary
 Complex predicate structure andpluralised events in Akkadian 201

contexts the D-stem could not be used as a matrix causative for verbs that
are transitive in the G-stem, which again indicates that the D-stem is re-
stricted to a thematic configuration with only a V P complement, whereas
the S-stem is not restricted to a thematic configuration in that it may also
take complements larger than VP, such as vP.
The fact that the D- and S-stems seemingly could be interchanged with a
number of G-stem unaccusatives (beside the fact that some take only one or
the other) also indicates that with unaccusatives, the S-stem apparently did
not bar an agentive interpretation of the causer argument and that in such
cases a s - V P configuration could be interpreted thematically equivalent to a
v-VP configuration, both introducing an agent/causer argument where there
is none in the G-stem.
With verbs that appear to be ambiguous between a non-agentive and an
agentive interpretation in the G-stem such as elü 'rise, go up, be(come)
high(er)' in (17), causativisation seems to differentiate between the readings
in so far as the agentive version, which arguably contains an external argu-
ment in Spec,vP, is targetted by the S-stem and not by the D-stem. To convey
causative meanings of the verb in (17a) like ' m a k e someone go up (to a
higher location)', ' m a k e a person come forth (as a witness in a law suit)',
etc., apparently only the S-stem could be used. D-stem forms as in (17b)
rather seem to pattern with D-stems of unaccusative deadjectivals.

(17) a. inanna 4 ammätim Habur Ilcim-ma (...)

now 4 cubit.plf.ACC ( n a m e ) rise(=G).pret.3sgm.vent-MA
' n o w the (river) H a b u r has risen 4 cubits ( . . . ) . ' (Syria 19, 123)

b. resT-su lü ulli
head.ACC.cstr-possCL3sgm indeed raise(=D).pret.lsg
Ί raised its (=wall) s u m m i t . ' ( L I H 57 i 17)

c. lü zikara lü sinnista ana üri tuselli-ma (...)

either man.ACC or woman.ACC to roof.GEN let-go-up(=S).
pres.2sgm-MA
' Y o u m a k e a m a n or a w o m a n g o u p on the roof ( . . . ) . ' ( Z A 3 2 , 172: 15)

Interestingly, there seem to be no Akkadian equivalents of English sen-

tences such as (18).

(18) a. John let/had/made Mary sink the boat.

b. John let/made/had Mary enlarge the springs.
202 Christian Huber

That is, the D-stem version rubbu 'enlarge x' of the G-stem rabü 'be(come)
big(ger)', for example, or the D-stem ullü 'raise x' of the G-stem elü 'rise,
go up' apparently could not be morphologically further causativised,
making it impossible in Akkadian to express something like John
had/let/made Mary enlarge the springs or John let/made/had Mary raise
the wall's summit by means of verbal morphology. 13 The issue will be re-
turned to in section 6.14

3. G-stem/D-stem transitivity alternations

In section 1 it was shown that the D-stem transitivises unaccusative de-

adjectival verbs. However, G-stem/D-stem transitivity alternations are
found also with other verbs. The present section will briefly illustrate such
alternations in some other verb classes.
As the data in (19) show, G-stem unaccusatives other than apparent
deadjectivals become transitives in the D-stem as well:

(19) G: belü ' g o out (fire, light)' D: bullü 'extinguish x'

G: narätu 'shake (intrans.), D: nurrutu 'shake x'
undergo shaking'
G: pahäru 'gather, D: puhhuru 'gather,
come together' bring together x'
G: tebü 'sink (intrans.)' D: tubbü 'sink (trans.)'

Only few motion predicates appear to have a D-stem. If so, they are turned
into transitives with an agent argument. One example is given in ( 2 0 ) . 1 5

(20) G: akäsu 'go' D: ukkusu 'drive away λ ' ( S : 'cause x to g o ' )

An interesting case is posed by psychological predicates (c. Belletti & Rizzi

1988) of the fear/frighten type since the Akkadian equivalents of fear and
frighten (and comparable pairs) both are derived from the same root. While
subject experiencer verbs such as fear surface in the G-stem, the D-stem
yields object experiencer verbs such as frighten from the same root. It thus
appears as if the D-stem adds, or allows to realise, an agent/causer argu-
ment that is not present in the G-stem, to the effect that the experiencer
surfaces as the verb's object.
 Complex predicate structure andpluralised events in Akkadian 203

(21) G : galätu 'tremble, be nervous, D : gullutu 'frighten, scares'

be afraid o f y (=ACC)'
G : hadü 'be joyful, be glad D : huddü 'gladden, delight*'
(abouty(=PP, DAT))'
G : palähu 'be afraid o f > (=ACC)' D: pulluhu 'frighten .r
G : sahatu 'be afraid ( o f y (=ACC))' D : suhhutu 'frighten, intimidate x'

However, as in English and other languages, also in Akkadian a causer

argument and the target/subject matter of emotion (cf. Pesetsky 1995)
seemingly cannot co-occur (see (22)). This shows that the D-stem does not
simply just add a (VP-external) agent/causer argument to some possibly
complex (VP-internal) structure, suggesting that the analysis of pairs such
as those in (21) should not differ substantially from languages like English.
As the issue of psych-verbs is still much debated (for some additional dis-
cussion see e.g. Pesetsky 1995 and Baker (1997), I will not pursue it further.

( 2 2 ) pulluhu * 'frighten χ of_y' (hypothetical meaning: 'make χ fear/be afraid o f y ' )

huddü * 'gladden χ about y' (hypothetical meaning: 'make χ glad about y')
(etc.)

Verbs of knowledge also exhibit a G-stem/D-stem alternation where an

agent or causer argument is added in the D-stem. With the dynamic predi-
cate lamädu 'learn' the D-stem adds an agent argument. In contrast to the
psychological predicates in (21), however, the original two arguments of
the verb may be retained, yielding an agentive predicate with a complex
internal domain.

(23) G : lamädu 'learn, come to know, experience χ (=ACC)'

D : lummudu 'inform^ (=ACC) about.ν (=ACC)\ "teach y (=ACC) * (=ACC)'

The stative predicate edü/wadü 'know' does not allow for a double object
construction in the D-stem in most dialects. Again the D-stem adds an
agentive argument but the resulting verb's internal domain may contain
only one of the two arguments found in the G-stem (predominantly the
theme).

(24) G: edü, wadü 'know*'

D : (w)uddü 'informy' (only Middle- and Neo-Assyrian),
'identify/mark/assign x\ 'explain x'
204 Christian Huber

The verb hasäsu 'remember' allows for two different D-stem constructions.
One appears to be an extension of the G-stem version, to which an agent/
causer argument is added (see (25 i)). The D-stem verb here takes as its object
the entity whose memory is being affected (and occasionally the target of
remembering as a second object). The other D-stem construction, (25 ii), still
involves an agentive subject but rather denotes a conscious mental activity
directed at some target.

(25) G : hasäsu 'remember* (=ACC)'

D: hussusu (i) 'remind y ( = A C C ) ( o f * (=ACC))'
(ii) ' c o n s i d e r * (=ACC)'

In all three cases, lamädu 'learn', edu 'know' and hasäsu 'remember', the
G-stem versions are non-agentive verbs whose valency seems increased in
the D-stem by addition of an agent/causer argument with the option of re-
taining the original arguments of the G-stem verbs in at least two cases.
This might suggest that in the G-stem versions none of the arguments oc-
cupies a VP-external position, be it a bare VP or a v-VP structure lacking
an argument in the specifier of vP. The case of edü ' k n o w ' is rather remi-
niscent of the psych-verbs in (21) in that in the general case only one of the
two arguments of the G-stem version may be retained in the D-stem. Ques-
tions arise also with respect to the second D-stem construction of hasäsu
'remember' in (25ii). All mentioned cases need closer investigation. Both
the psych-verbs in the D-stems and the D-stem versions of the verbs of
knowledge, however, clearly come in transitive structures with an agent or
causer argument as the subject that is not present in the G-stem.
The question of the actual G-stem structure(s) also arises with verbs
such as those in (26) that seem to oscillate between a reflexive agentive and
a non-agentive or stative reading. In (26) the D-stem targets only the
stative/non-agentive reading of the corresponding G-stem verb, to which it
adds an agent or causer argument, allowing D-stem forms with a complex
internal domain. 1 6

(26) G: sebiι D: subbü

(i) 'be(come) satiated (with y (=ACC))' 'satiate * (with y (=ACC))'
(ii) 'satiate oneself ( w i t h y (=ACC))' (* ' m a k e * satiate x-self (with y)')

G: labäsu D: lubbusu
(i) ' w e a r y (=ACC)' 'dress * (=ACC) (with y (=ACC))'
(ii) 'put o n y ( = A C C ) \ (* ' m a k e * put o n / )
'dress oneself ( w i t h y (=ACC))' (* ' m a k e * dressx-self ( w i t h y ) ' )
 Complex predicate structure andpluralised events in Akkadian 205

4. G-stem transitives vs. D-stem transitives: verbal plural

Let us now return to the question of what sets D-stem transitives like those
in (4), partly repeated in (27), apart from their G-stem counterparts. Recall
that G-stem transitives such as those in (4)/(27) are not causativised in the
D-stem but that their valency remains unchanged.

(27) G: batäqu 'cut (off) Λ·' D: buttuqu 'cut (off) x'

G: 'edelu 'lock x' D: Juddulu 'lock x'
G: }epesu 'do, make x' D: }uppusu 'do, make x

One label frequently found in treatments of Semitic languages to characterise

the function of the corresponding stem (e.g. Arabic stem II, Hebrew Pi'el)
is "intensive". Already Poebel (1939: 5 n . l , 65-68) argued that this charac-
terisation be erroneous and that the respective stem express not the idea of
intensity but that of plurality. Also Greenberg (1991) argues that the D-
stem and its congeners in other Semitic languages have, in addition to other
functions, characteristics of verbal plurality similar to those found in other
languages that exhibit verbal plural phenomena. 17 Kouwenberg (1997) points
out that in Akkadian in cases like (4)/(27) the D-stem never yields an "in-
tensive" reading of the corresponding G-stem but that in these cases the D-
stem is predominantly employed when the sentence contains a plural element
or when the action itself is pluralic in some sense.

4.1. Verbal plural with the D-stem of transitive verbs

Kouwenberg observes that the D-stem is frequently used with plural (or
mass-noun) direct objects, but that it occurs also with plural indirect objects
or adjuncts, occasionally also with plural subjects or in the case of durative
or repeated action. Yet, the use of the D-stem is not obligatory in these
cases. The G-stem is free to occur with a plural subject, object or adjunct.
However, the D-stem is always used in case of a contrast of singular vs.
plural. Kouwenberg gives examples like those in (28) to (34).
In (28) the contrast of the use of the G-stem with a singular direct object
(28a) and the use of the D-stem with a plural direct object (28b) is illustrated.
206 Christian Huber

(28) a. näram epte-sum-ma

river.sg.ACC open(G).pret.lsg-CL3sgm.DAT-MA
Ί opened a canal for it (= a city) (and called i t . . . (name)).' (RIME 4, 603: 26f.)

b. närätim upetti
river.plf.ACC open(D).pret. lsg
Ί opened canals.' (RIME 4, 603: 47f.; Kouwenberg p. 120)

The same contrast is also found under passivisation, where after NP-
movement an underlying object NP surfaces as the subject in nominative
case. In (29a) the root Vbtq 'cut (off/through)' occurs in the N-stem, which
serves as a passive of the G-stem, whereas in (29b) it occurs in the Dt-stem,
which serves as a passive of the D-stem.

(29) a. butuqtum ibbattaq (bb < nb)

sluice channel(?).sg.NOM cut(N).pres.3sgc
'the sluice channel(?) will be cut through' (YOS 10, 16: 5; Kouwenberg p. 123)

b. butuqätu ubtattaqä
sluice channel(?).plf.NOM cut(Dt).pres.3plf
'the sluice channels(?) will be cut t h r o u g h "
(YOS 10, 26 iii 29; Kouwenberg p. 123)

Also note that the D-stem is not barred by countability.

(30) adi sibi-su u 'allat

until 7 th .cstr.GEN-CL3sgm swallow(D).pres.3sgm
' H e swallows 7 times.' (BKMB 30: 43, cf. Kouwenberg p. 127)' 8

(31) a. üma 1 atäna(EME5) ätamar ina qät assuräye

today 1 she-ass.ACC see(G).perf.lsg in hand.cstr(GEN) Assyrian.GEN
assabat
seize(G).perf.lsg
'Today I saw one she-ass and caught (G) (her) in the hand of an Assyrian.'
(ZA 73, 78: 18, cf. Kouwenberg p. 136)
b. 1 urita(EME5) sa imeri(ANSE) adi 3 imeri(ANSE.MES) (...)
1 mare.sg.ACC REL donkey.GEN besides 3 donkey.pl.GEN (...)
ussabbit ilte-su
seize(D).perf. lsg with-CL3sgm
'One mare of a donkey together with three donkeys (...) I caught (D) in his posses-
sion (lit. 'with h i m ' ) . ' (ZA 73, 77: 6f., cf. Kouwenberg p. 136)
 Complex predicate structure andpluralised events in Akkadian 207

However, the D-stem is not restricted to co-occurence with a plural direct

object. Occasionally the use of the D-stem seems to be connected with a
plural beneficiary.

(32) (...) 2 PI.TA.AM3 ana märe Ninua pilku

(...) at 2 PI each for son.pl.cstr.GEN Nineveh.(GEN) partition.sg.ACC
upallik
partition(D).pret. 1 sg
' ( . . . ) at 2 PI each I m a r k e d out a plot of land for the citizens of N i n i v e h . ' (i.e. plots of
land at a m e a s u r e of 2 PI each w e r e marked out for the citizens)
(Sn 97: 88, cf. K o u w e n b e r g p. 150)

There are only few attestations of the D-stem with a plural subject and a
singular direct object, as in (33) (see also (34) below).

(33) ina dabäbim umahhisü-su-ma

in law suit.GEN beat(D).pret.3plm-CL3sgm-MA
' i n / d u r i n g the trial they beat (pret.) h i m ' ( A b B 1 2 , 65:29; cf. K o u w e n b e r g p. 165)

In (34) the D-stem is employed to refer to the wall-building activities by

various former kings. When referring to the wall's rebuilding by the present
king or by a future king, in contrast, the G-stem is used.

(34) durum sa RN, RN2 RN3 RNj RN5 mär RN6

wall.sg.NOM REL R N , R N 2 R N 3 R N 4 RN5 son.cstr RN6
abbü-ya uppisüni enah-ma
father.plurxstr-possCLlsg make(D).pret.3plm.sub tire(G).pret.3sgm-MA
(...) epus (...) rubä^u urkiu enüma
(...) make(G).pret.lsg (...) prince.sg.NOM later.NOM when
durum set enuhu-ma eppusu
wall.sg.NOM this.f.NOM tire(G).pret.3sgm.sub-MA m a k e ( G ) . pres. 3 s g m . sub
DNj u DN2 ikribi-su isamme 'ü-su
D N ] and D N 2 prayer.cstr-possCL3sgm hear(G).pres.3plm-CL3sgm
' t h e wall w h i c h R ( o y a l ) N ( a m e ) , R N 2 R N 3 R N 4 (and) R N 5 , the son of R N 6 , m y fore-
fathers, had built (D), dilapidated (lit.: w a s b e c o m i n g tired) ( . . . ) and I (re)built (G)
(it); (as f o r ) a f u t u r e prince, w h e n that wall has b e c o m e dilapidated and he rebuilds
( G ) (it), D ( i v i n e ) N ( a m e ) i and D N 2 will listen to his prayer(s).'
( R I M A 1, l O l f : 5ff, cf. K o u w e n b e r g p. 148)

Another case of "plural action" is perhaps found in the D-stem of verbs

such as natälu 'look at', if e.g. nuttulu 'inspect' can be taken to derive from
something like 'look at χ many times'. Apart from that, the D-stem seems
to indicate a conscious or wilful action here.
208 Christian Huber

(35) G \ natalu "look a t * ( = A C C ) ' D: nuttulu 'inspect*'

No clear contrasts, however, are detectable with unergatives in those cases

where a verb is attested in both the G- and the D-stem. On the whole, rele-
vant unergative data are rather scarce.

(36) a. summa kalbu (...) inabbuh

if dog.sg.NOM ( . . . ) bark(G).pres.3sg
' i f a dog (...) barks' ( C T 39, 2: 98)

b. kalbu sa pahhäri (...) unambah (< unabbah)

dog.sg.NOM REL potter.sg.GEN ( . . . ) bark(D).pres.3sg
'The potter's dog (...) barks.' ( A B L 4 0 3 : 7)

4.2. Case alternation in a stative construction - a contrast

The assumption that the D-stem involves a v-VP configuration featuring an

external argument may also shed some light on a change in construction
encountered with certain stative constructions in the G- and D-stems.
The so-called stative conjugation is a predicative construction that can
be applied to nouns and verbs. It is formed, roughly, by adding a conjuga-
tional pronominal suffix to an adjective, verbal adjective, or noun. Statives
like (37a, b) are thus reminiscent of adjectival passives. With transitive
verbs, the stative conjugation predicates of its subject a state resulting from
the event denoted by the verb. The subject N P of a dyadic predicate in the
stative conjugation corresponds thus to the verb's internal argument but
surfaces with nominative case like the subject in a passive construction (see
(29) above).

(37) a. damäqu G: 'become good': NP-NOM damiq

' N P is g o o d ' (c. damqu 'good')
b. satäru G: 'write':
NP-NOM satir
' N P is w r i t t e n ' (c. satru 'written')

c. sarräku
king.statlsg
Ί am k i n g ' (c. sarru ' k i n g ' , anäku Τ)

A number of verbs exhibit two stative constructions. In addition to the

regular stative construction (illustrated in (38a)) they also allow for what
 Complex predicate structure andpluralised events in Akkadian 209

Assyriologists usually call an "active" stative construction. 19 With so-called

"active" statives, illustrated in (38b), it is the argument that corresponds to
the verb's surface subject of which a state resulting from the event denoted
by the verb is predicated. This yields a stative clause containing both argu-
ments of the verb, in (38b) the holder as well as the holdee.20

(38) sabätu G 'hold, seize':

a. "regular" stative: NP-NOM sabit 'NP is held/seized'
b. " a c t i v e " stative: ΝΡΓΝΟΜ NP2-ACC sabit 'NP, holds/is holding NP 2 '

On the surface, at least, this suggests that in contrast to (38a) the "holder"
argument in (38b) is no external argument but originates internally or per-
haps expresses a different theta role. Note that this "active" use of the stative
conjugation is not possible with a verb like write.

(39) satäru G 'write':

a. HP-Norn satir 'NP is written'
b . *NPRNOM NP2-ACC satir 'NP, is writing/has written NP 2 '

Under the assumption that the D-stem creates basically agentive, transitive
forms, no "active" D-stem statives are expected to occur, D-stem statives
are expected to appear in the "regular" stative construction (then resembling
passives in that the patient/theme cannot remain in the object position).
That is, if (38b) were put in the D-stem one would expect a case alternation
to the effect that the holdee appears as the surface subject in nominative case
and the holder would be allowed to surface only as a όν-phrase-type adjunct.

(40) D: subbutu 'hold, seize': a. "regular" stative:

NP2-NOM NPI-OBL subbut
'NP 2 is held by N P , '

rather than b. "active" stative:

(*) NPJ-NOM NP2-ACC subbut
'NP, holds/is holding NP 2 '

This state of affairs is indeed found. The G-stem example (41a) is an in-
stance of the "active" stative construction presented schematically in (38b).
In (41a) the "holder", filament (singular), appears in the nominative and the
"holdee", epigastrium, in the accusative case. In (41b) however, which
differs from (41a) only in that the "holder" is plural and accordingly the
210 Christian Huber

verb is in the D-stem, we find the reverse situation. N o w the "holdee", epi-
gastrium, surfaces in the nominative case whereas the "holder", filaments
(plural), appears as a fry-phrase type adjunct in the instrumental accusative,
as schematically illustrated in (40a). 2 '

(41) a. summa res libbi qüm sabit

if head.constr(ACC) inside.GEN filament.sg.NOM seize(G).stat.3sgm
' i f a f i l a m e n t h o l d s t h e epigastrium (lit. ' h e a d of the i n s i d e ' ) ' ( Y O S 10, 4 2 II 35)

b. summa res libbi qe subbut

if head.constr(NOM) inside.GEN filament.pl.ACC.iNSTR seize(D).stat.3sgm
' i f t h e e p i g a s t r i u m (lit. ' h e a d of the inside') is held b y filaments'
( Y O S 10, 4 2 II 33)

The data in (41) merit some comments. In (41a) the stative verb agrees with
qü ' f i l a m e n t ' . In (41b) the stative verb agrees with res libbi 'epigastrium',
which follows from the fact that it is inflected for third person masculine
singular and not plural, as would be required for agreement with a plural
'filaments'. There is no morphological marker of nominative or accusative
visible on res libbi 'epigastrium (lit. 'head of the inside')' because the noun
on which it would appear (resu ' h e a d ' ) is in the construct state. Instru-
mental accusative is morphologically identical to the 'regular' accusative as
the object case, which gives the construction a strange flavour at first sight,
together with the fact that word order remains unaffected. The case (and
number) alternation, however, is visible on qu. 'filament'. Word order re-
mains unaffected by the case alternation because res libbi 'epigastrium' is
topicalised in both (41a) and (41b), preceding the focus portion of the
clause (minimally, qu/qe 'filament(s)').
The change of construction in (41) seems to be due to the fact that in the
D-stem version of the verb the "holder" argument must originally occupy
the Spec,vP position (whereas the subject-NP of "active" statives perhaps
represents an argument that cannot occupy Spec,vP). Due to the passive-
like properties of the stative construction the "holder" can appear only as an
adjunct whereas the object, the "holdee", as the internal argument of the
corresponding verb becomes the surface subject.
 Complex predicate structure andpluralised events in Akkadian 211

4.3. Verbal plural and unaccusatives

Crucially, the D-stem never derives unaccusative f o r m s of G-stem unaccu-

satives, as predicted under the assumption that the D-stem always yields a
transitive (v-VP) configuration. Thus, the D-stem of aräku "become long"
does not yield an unaccusative verb with verbal plural properties assuming
a meaning such as, say, "become very long" or "becoming long of many x"
but becomes transitive "lengthen". Likewise, the D-stem of kasädu "arrive,
reach" surfaces as an agentive form with a lexicalised meaning "chase
(away)" but not as "arrive many times" etc.

(42) a. G: rabü D: rubbü

' b e c o m e large/great; grow" 'make χ large/great, raise'
(not: 'growing of many x, grow often/at many
occasions, grow extremely large', etc.)

G: aräku D: urruku
' b e c o m e long' 'lengthen x'
(not: ' b e c o m i n g long of many
'become extremely long', etc.)

c. G: kasädu D: kussudu
'arrive, reach y ( = A C C ) ' 'chase (away) JC' (lexicalised) 2 2
(not: 'arriving of many x\ 'arrive many
times/at many places', etc.)

The fact that there is no principled reason apparent why verbal plural
should be excluded with unaccusatives like the verbs in (42) (say, with a
plural surface subject) supports the assumption that the verbal plural prop-
erty of the D-stem is linked to a v-VP structure. 23
At this point we are able to address a question raised in section 2. In
section 2 it was argued that the D-stem is a complex phenomenon that in-
volves a v-VP configuration and yet another property, which in the present
section was identified as verbal plural. Two ways were sketched regarding
how the complex nature of the D-stem could be captured configurationally:
(i) as involving another (non-theta-relevant) functional head F in addition
to little v, or (ii) as a little ν with some special feature, indicated as v D , F or
VD now assumed to be responsible for "verbal plural". While v D is perfectly
compatible with the claim that a v-VP configuration is an integral part of
the D - s t e m ' s outfit there remains the question of the structural position of a
potential head F. Two options were outlined, repeated here for conven-
ience: merger of F above vP (10a), or a head F sandwiched between the
212 Christian Huber

two portions of the ν-VP configuration (10b). It was noted that option (10b)
would also allow for F to take a V P complement without having been se-
lected by v, resulting in an unaccusative V P that is complement of F, rais-
ing the question what the properties of such a construction would be.

Identifying F as the head that encodes verbal plural, a hypothetical structure

(10c) would surface as an unaccusative D-stem verb displaying verbal plural
properties. That is, one would expect to find unaccusative D-stem versions of
unaccusative G-stem verbs precisely like the starred hypothetical renderings
in (42) above that require a plural (or mass noun) surface subject (e.g. "be-
coming big(ger) of many x"), or denote a larger-than-usual degree (e.g.
"grow extraordinary large") or otherwise multiple taking place of the re-
spective event (e.g. "grow large many times") or the like. However, as
pointed out above, D-stem versions of G-stem unaccusatives do not have
these properties but surface as transitive verbs, ruling out the structure
(10c). If (10c) is ruled out, also option (10b) must be discarded. 2 4
This still leaves us with the question of the nature of F or v D , respec-
tively, and which a m o n g the two should be preferred. Assuming, as I will
propose in section 5, that the D-stem pluralises events (or sub-events) it
seems apt to suggest that F or vD are part of the N u m b e r system. F might
thus be a N u m b e r head that expresses verbal (or eventual) number (perhaps
an aspectual category or at least expected to interact with the aspectual
system in some way). Considering approaches that propose that noun
phrases and clauses have analogous structures and noun phrases contain a
N u m b e r projection (e.g. Szabolcsi 1994), some such projection would be
expected also with verb phrases or clauses. Under the assumption that
N u m b e r is an interpretable feature on noun phrases (Chomsky 1995) rather
than a separate functional projection, verbal plural as a number-related
category might be an interpretable Number feature on verb phrases (in the
present case, vP) as well, constituting vD.
 Complex predicate structure andpluralised events in Akkadian 213

Another issue that becomes relevant at this point is the fact that beside the
D-stem Akkadian has another set of candidates for some verbal plural-like
property, namely the so-called iterative stems that can be derived from any
of the four main stems G, D, S and N. The properties of iterative stems
differ somewhat from those of the D-stem. For example, in contrast to D-
stem versions of G-stem unaccusatives, iterative stem versions of G-stem
unaccusatives remain unaccusative, see e.g. the iterative versions of abälu
'become dry' and maräsu 'become ill' in (43), predicating of some entity
repeated becoming dry or becoming ill, respectively.

(43) a. summa liq pi-su Ttanabbal

if ....cstr.NOM mouth.cstr.GEN-possCL3sgm become-dry(Gtn).pres.3sgc
'if his palate b e c o m e s dry again and a g a i n ' ( A M T 76, 1, 4)

b. ana minim libba-ki imtanarras

to what.ACC heartxstr.NOM-possCL2sgf become-ill(Gtn).pres.3sgc
' W h y d o e s y o u r heart b e c o m e ill again and a g a i n ? ' ( T L B 4, 17, 3)

However, the relation between the D-stem and the iterative stems as well as
many issues regarding their properties and lexical distribution still remain
to be investigated in more detail. I will therefore not further deal with issues
such as more precise positions of functional projections, potential inter-
actions, or feature checking or agreement mechanisms.
Before proceeding another fact needs to be mentioned: while in the gen-
eral case the appearance of the D-stem of transitive verbs is visibly connected
to the presence of some pluralic element, say, a plural object, there is no
such requirement for the D-stem versions of unaccusatives, as illustrated in
(44), featuring rubbü 'make big, raise (a child)', which is the D-stem of
rabü 'become big', and urruku 'lengthen', which is the D-stem of aräku
'become long'. In both cases the verb appears with a singular object (and
subject).

(44) a. kisalla-sa magal urabbi-ma

courtyard.(ACC).cstr.-possCL3sgf greatly m a k e - b i g ( = D ) . p r e t . 1 sgc-MA
Ί greatly e x p a n d e d its (= palace) c o u r t y a r d . ' ( B o r g e r Esarh. p. 26 §27 vi 32)

b. Samas hatta-su lirrik

Samas.(NOM) sceptre.(ACC).cstr-possCL3sgm lengthen(=D).prec.3sgm
' M a y (the g o d ) S a m a s m a k e his sceptre (i.e. rule) l o n g ! ' ( K H 26r 15)

The question why there is no plural requirement for the object of a D-stem-
transitivised former G-stem unaccusative will be addressed in section 5.
214 Christian Huber

5. Verbal plural as eventual plural

The D-stem effects demonstrated in sections 3 and 4 can be summarised as

follows:

(45) D-stem effects (i) with G-stem unaccusatives: transitive/causative

(ii) with G-stem transitives: verbal plural

The crucial question now is: how can (45i) and (45ii) be related? The answer
must have to do with the mechanisms behind "verbal plural" and its effects
with different types of predicates.
Following Greenberg 1991 and Kouwenberg 1997 I took the D-stem to
be associated with "verbal plural". In section 2 it was proposed that the D-
stem reflects a functional head that co-occurs with an external argument,
either some functional head F that is associated with ν in a v-VP configura-
tion or a special type of v, indicated as vD. Under this assumption "verbal
plural" as met in the Akkadian D-stem must therefore be encoded in vD or
F, which in turn must associated with the Number system, perhaps also
associated or interacting with the aspectual system.
In the present section I will try to make the notion of verbal plural more
precise and propose an approach that allows us to account for the properties
of D-stems of transitive and unaccusative verbs as well as for some differ-
ences in the properties of the Akkadian D-stem and its counterparts in Arabic
and Hebrew. For the moment I will largely limit my attention to transitive
verbs as discussed in section 4 and unaccusatives as presented in section 1
(e.g. grow, lengthen, etc.), admitting that widening the scope of investiga-
tion will certainly refine the picture. I will also point out some differences
between Akkadian on the one hand and Arabic and Hebrew on the other
hand.

5.1. Pluralising events

First, let us address the question of what exactly is pluralised in "verbal

plural". Take "plural" to express a number >1 of (instances of) entities be-
longing to the same category. If these entities are, say, objects (as expressed
as noun phrases), nominal plural will be interpreted as "many objects of the
same type". Let us assume now that such entities may not only be of nomi-
nal category, e.g. objects of the same type, but that they may also be events.
 Complex predicate structure andpluralised events in Akkadian 215

Verbal plural will then be interpreted as "many events of the same type". I
will therefore propose that the D-stem pluralises (sub-)events. Morphologi-
cally the pluralic value of the D-stem is iconically encoded via gemination
of C 2 . Gemination may be regarded as an instance of reduplication, i.e., in
the simplest case, repetition.
Recall from section 4 that the basic G-stem may occur freely with a plural
subject, object or adjunct. That is, it is unspecified with respect to "verbal
number". The difference between the G-stem and the D-stem can thus be
characterised in the following way:

G-stem vs. D-stem:

for some (sub-)event(-type) e, G-stem: unspecified w.r.t. number of e

D-stem: plural of e (many e)

D-stem phenomena are now expected to be dependent on the semantic type

of NP (count noun, mass noun, ...), the type of event (or resulting state),
and the manner of pluralising the relevant (sub-)event.
Pustejovsky (1991) proposes a system of eventual decomposition that
recognises three basic event types: S(tate), P(rocess), T(ransition). In what
follows I will make use of his state and transition but I will leave aside
compositional questions. Pustejovsky defines a state (S) as a single event
which is evaluated relative to no other event (46a), a transition (T) is an
event identifying a semantic expression which is evaluated relative to its
opposition (46b).

(46) a. S b. Τ
I A
e -'e e

For present purposes let me propose that the D-stem pluralises transitions, i.e.
it creates a series of transitions of the same type. In addition, I will assume
that transitions may differ with respect to the (type of) state in which they
result so that there may be several types of simple transitions. For exposi-
tion, let us consider three types of predicates.
First, consider predicates such as (i) build, break(-in-two) or (ii) open,
close. These predicates express a binary opposition. The transition brings
about a terminal state predicated of some entity that is evaluated relative to
its opposition and cannot be further subdivided. The resulting state has to be
reversed before it can be brought about anew with respect to the same entity.
216 Christian Huber

(47) (i) build, break(-in-two), (ii) open, close:

Τ

-s

With predicates like build or break(-in-two) the transition results in a ter-

minal state (existence or non-existence) of the entity of which the resulting
state is predicated. With predicates like open or close the transition results
in a terminal state with respect to some entity, but leaves the entity intact
(that is, the change of state expressed by the transition is not from existence
to non-existence or vice versa). Informally put, once some object is built
(and thereby created) or broken (and thereby destroyed), the same object
cannot be built or broken anew (unless un-built or un-broken before).
Likewise, once a door is opened, it has to be closed again before it can be
opened anew. And if John gives a book to Mary, Mary has to return the
book to John before he can give it to her one more time.
Next, consider predicates such as hit, beat, bite, affect. Again they ex-
press a binary opposition, the transition brings about a state with respect to
some entity that is evaluated relative to its opposition and cannot be further
subdivided. However, the resulting state does not have to be reversed be-
fore it can be brought about anew with respect to the same entity.

(48) hit, beat, bite, affect:

Τ

ι affected affected

That is, some entity may be affected by, say, hitting, and its having been
affected by hitting will remain part of its further history, but nevertheless
that entity can be affected by hitting again, which will again remain a part
of that entity's further history, and so on. Thus, it is possible to say that
John beat the dog three times whereas it is not possible to say that *John
killed the dog three times, assuming the dog stands for the same individual
in all three killing instances.
Finally, consider predicates such as grow and lengthen. These predicates
do not express a binary opposition (in absolute terms). The transition does
not result in a state that can be evaluated relative to its opposition, but
rather results in a (change of) degree which is evaluated relative to another
degree on a scale, predicating of an entity a (change of) degree (i.e. progress
 Complex predicate structure andpluralised events in Akkadian 217

on a scale), whereby the interval between the degrees can be infinitely sub-
divided.

(49) grow, lengthen:

degree a degree b > a

Informally put, again, growing of, or lengthening, some object does not
necessarily imply that the object has been short before, it may still have
been longer than any other object in a given comparison set. However, its
degree of longitudinal extension increased (or keeps increasing), resulting
in an interval, which, if mapped onto a scale at some time of inspection,
will yield a distance between two points respresenting the degrees of longi-
tudinal extension at the beginning and at the end of the inspection interval
such that degree a < degree b whereby the interval between the two degrees
can be infinitely subdivided. Thus, no absolute state (like e.g. existence) is
evaluated relative to a corresponding non-state (like e.g. non-existence) but
rather one degree of longitudinal extension relative to another degree of
longitudinal extension.
In Akkadian, verbs of types (47) and (48) are generally realised as tran-
sitives in the basic stem, verbs of type (49) are always realised as unaccusa-
tives in the basic stem. 25

Table 3.

Akkadian: verb type G-stem (basic form)

build/open etc.-type verbs (47): transitive
hit, bite etc.-type verbs (48): transitive
grow, lengthen etc.-type verbs (49): unaccusative

Let us now turn to the question how transitions may be pluralised. In the
simplest case, pluralising transitions has the effect that the number of rele-
vant transitions be not restricted to one. We may thus think of a series of
transitions of the same type. 26 However, there may be series with different
properties. Consider the following three types of transition series. Type I,
illustrated in (50), is a series of independent transitions, the output of one
transition does not serve as input for the next transition.
218 Christian Huber

Type II, illustrated in (51), is a recursive series, the output of one transition
serves as input for the next transition.

Type III, illustrated in (52), is a series of proto-events (proto-transitions)

which in their totality constitute a transition, rendering a mass-noun-like
interpretation with a count noun. 27

(52) type III

("• e ) (e) (-e) (e) (-e) (e) -e e

Verbs may of course be assigned to more than one event type. For example,
a verb like punch may be treated according to (48) predicating affectedness
by punching or as "punch-do, give a punch" etc, then corresponding to (47).
With (overt or covert) three-place predicates, thus, a number of options
come into play. Through interaction with the particular way of pluralising
the relevant (sub-)event various interpretative possibilities arise.
Instances of verbal plural as illustrated in section 4 result from which
type of transition is pluralised by means of which type of series. However,
not all possible combinations occur in Akkadian. Semitic languages appar-
ently differ as to what combinations they allow. As I will argue later, this
accounts for the fact that "intensive" D is largely absent in Akkadian.
 Complex predicate structure andpluralised events in Akkadian 219

5.2. Brief review of some instances of eventual plurality as met in Akkadian

Let us n o w briefly review some instances of verbal - or eventual - plural as

met in the Akkadian D-stem. With build/open!break-type verbs eventual
plural in Akkadian brings about a type I-series of independent, binary, ter-
minal transitions. The state resulting from such a transition can be predi-
cated of an entity only once (otherwise it had to be reversed before being
brought about anew) so that such a series is interpretable only with plural
count nouns or mass nouns: each transition (or the state resulting from it,
respectively) must be predicated of a different entity (in the case of change
of location/possession verbal plural of this type may also yield a distribu-
tive interpretation). Below I repeat relevant examples from section 4 (trans-
lation only).

(28) b. I opened canals.

(29) b. The sluice channels(?) will be cut through.

(30) He swallows 7 times.

(31) b. One mare of a donkey together with three donkeys ( . . . ) I caught

in his possession.

(32) ( . . . ) at (a measure of) 2 PI each I marked out a plot of land for the
citizens of Nineveh.

Occasionally eventual plural may also bring about a type-III-series of inde-

pendent, binary, terminal transitions, then interpreted as a series of activi-
ties which in their totality constitute a transition. The resulting state is
predicated of an entity that receives mass-noun-like interpretation, e.g. a
series of building-events that result in a wall, as in (34). 28

(34) the wall which R(oyal) N(ame), RN 2 (...) (and) RN 5 (...) had built (...)

With grow-type verbs eventual plural in Akkadian brings about a type-II-

series of degree transitions such that the output of one transition serves as
input for the next transition, the last degree of each inspection interval
evaluated relative to the first. However, since what is predicated of an en-
tity is still merely a (progress in) degree and all intervals can be infinitely
subdivided, the plural event need not yield a different interpretation (that is,
220 Christian Huber

the resulting 'degree-state(s)' can still be predicated of a singular entity

since no binary, terminal state results). 29 Examples are (42a, b) and the
transitivised unaccusatives in (2).

(42) a. enlarge χ
b. lengthen χ
etc.

Consider now eventual plural with affect-type verbs. The state resulting
from an affect-type predicate does not have to be reversed before it can be
brought about anew with respect to some entity. In Akkadian eventual plural
mostly brings about a type-I-series here that is interpretable only with plural
count nouns (or mass nouns), to the effect that the affected state is predi-
cated of a multitude of entities as the affected state resulting from each
transition must be predicated of a different entity. The option of multiply
predicating an affected state of the same entity seems to be realised only
marginally in Akkadian. One possible instance is (33), which is a case of
the D-stem with a plural subject and a singular object. Another potential
instance is (35), if to be understood as something like "look at χ many
times".

(33) they beat (pret.) him

(35) inspect (D, vs. G: 'look at')

For unergatives several analyses are available. In Hale and Keyser's (1993)
analysis of unergatives as concealed transitives involving an incorporated
bare N, unergative verbs such as work or dance are underlyingly composed
of something like 'work-do' or 'dance-dorespectively. Thus, a terminal
change of state may be predicated of the underlying nominal component of
an unergative, which however has mass noun-like properties in that it re-
sists counting. This might explain why discernible differences between the
G- and D-stem of unergatives are largely absent, and may also be the rea-
son why many unergatives only occur either in the G-stem or the D-stem.
Alternatively, an unergative verb may be treated as involving a terminal
change-of-state transition predicated of a single entity such as 'act of swal-
lowing', 'act of barking', etc., whose coming about is denoted by the verb.
In this case pluralising the event will yield a repetitive interpretation ('swal-
low, bark, etc. several times'). Potential instances are (36b) (bark) and (30)
(swallow).
 Complex predicate structure andpluralised events in Akkadian 221

5.3. Eventual plural as "Intensive": Arabic and Hebrew vs. Akkadian

Above it was said that in Akkadian, the D-stem generally does not yield an
"intensive" interpretation as is often found with the corresponding stems in
Arabic and Hebrew. The approach just outlined allows us to make the
somewhat intuitive notion "intensive" more precise and at the same time
determine in which respects Arabic and Hebrew differ from Akkadian. I
will consider two cases. First consider the Arabic data in (53), where x
stands for a singular count noun object.

(53) Classical Arabic: 3 0 a. stem I: (

adda 'bite Λ'
qarasa 'tweak x'
(
b. Stem II: addada 'bite χ fiercely or frequently'
qarrasa 'tweak χ sharply'

In (53b) the "intensive" interpretation may come about in two ways: (i) by
predicating a series of independent "affected" states ("affected by biting/
tweaking") of a singular count noun entity, which yields an interpretation of
repeated affectedness by biting or tweaking as opposed to a single affected-
ness by biting or tweaking, or (ii) the output of one transition serves as
input for the next transition so that the resulting states (or rather, degrees)
add up, yielding an interpretation of increased affectedness by biting or
tweaking (ie. an increased degree of affectedness).
Next, consider the Arabic and Hebrew data in (54).

(54) Classical Arabic Hebrew

a. I: kasara Qal: savar 'break*(intwo)'
b. II: kassara P i ' e l : sibber 'break χ (into many pieces)', 'shatter x'

Under the approach taken here, the "intensive" interpretation in (54b) re-
sults from a series of binary terminal state transitions predicated of a singu-
lar count noun entity, whereby (i) each resulting state is predicated of the
same singular count noun entity, or (ii) the output of one transition serves
as input for the next transition (i.e. in both cases the singular count noun
entity is interpreted in a mass noun fashion). 3 1
Akkadian seems to lack options (53bii) and (54b) completely and to
allow for option (53bi) only marginally. The difference between Akkadian
and Arabic/Hebrew can therefore roughly be stated as follows. Akkadian
does not allow the predication to a singular count noun entity of the state(s)
222 Christian Huber

resulting from a type II series of terminal or affected state transitions (i.e.

Akkad. D = * break χ into many pieces', *beat χ hard etc.). Furthermore,
Akkadian only marginally allows the predication to a singular count noun
entity of the state(s) resulting from a type I series of terminal or affected
state transitions {do χ many times, beat χ many times, etc., otherwise ex-
pressed via iterative stems).

5.4. The Puzzle, revisited:

Let us now return to the question how the D-stem effects, for convenience
repeated in (55), can be related.

(55) D-stem effects (i) with G-stem unaccusatives: transitive/causative

(ii) with G-stem transitives: verbal plural

In particular, the following questions need to be addressed:

(56) (i) Why do transitivised unaccusatives surface in the D-stem?

(ii) Why is there no plurality requirement for the object of a D-stem-
transitivised (former) unaccusative?

Question (56i) seems to be connected to question (57):

(57) Why do Akkadian G-stem transitive verbs have no unmarked unaccu-

sative G-stem form?

As illustrated in Table 4, there are no transitive versions of unaccusatives

such as grow or lengthen in the G-stem. Transitive versions of such verbs
occur in the D-stem only. Conversely, verbs of the build/open type have no
intransitive G-stem alternant. To render intransitive versions of verbs of
this type intransitivising morphology has to be applied. 32

Table 4.
intransitive transitive
grow-type verbs G-stem D-stem no transitive graw-type
verbs in the G-stem

build!open-type verbs (mid./pass.) G-stem no intransitive/unaccusative

build/open-type verbs in the
G-stem
 Complex predicate structure andpluralised events in Akkadian 223

This indicates that there is a distributional restriction, due to the derivational

properties of the stems: Akkadian allows no - or tries to avoid an - (un-
marked) "ergative" alternation in a verb class within the same stem. Valency-
changing operations must be reflected in morphology, thus necessitating a
different stem. Therefore - answering questions (56i) and (57) - , the transi-
tive alternant of an unaccusative must appear in a different stem. However,
this stem must meet two requirements: it must (i) allow for transitives, and
(ii) be compatible with the semantics of the predicate.
Under the present analyses the D-stem meets both requirements with
respect to G-stem unaccusatives. First, the D-stem is transitive in that it
always comes in a v-VP configuration. Secondly, with grow-type verbs, the
D-stem brings about a series of degree-transitions but what is predicated of
an entity is still merely a (progress in) degree so that the series of events
can still be predicated of a single entity and no plural object is forced. This
answers the question (56ii) why there is no plural requirement for the ob-
ject of a D-stem verb. In contrast, with verbs like build, break or open,
which are always transitive in their basic form and involve a binary termi-
nal transition, eventual plural creates a multitude of building, breaking or
opening-events resulting in a multitude of respective terminal states that
cannot be predicated of a semantically singular entity. Accordingly, they
must be predicated of mass nouns or plural count nouns.

6. Some morphological issues

In section 2 it was observed that the - transitive - D-stem versions of G-

stem unaccusatives, e.g. D-stem rubbü 'enlarge, make big(ger)' vs. G-stem
rabü 'be(come) big(ger)', apparently could not be further causativised in
order to derive a form with a hypothetical meaning such as "cause y to en-
large x". A connected question is whether a verbal plural form of a basi-
cally transitive verb, e.g. D-stem puttü 'open many x' vs. G-stem petü
'open x\ could be embedded under a matrix causative to derive a form with
a hypothetical meaning such as "cause y to open many x". In both cases it
would be necessary to derive a causativised version of a D-stem form. On
the face of it one might thus expect that a combination of D-stem and S-
stem would produce forms with the desired properties.
Interestingly, Akkadian has a stem that seems to combine the morpho-
logical features of D-stem and S-stem: the so-called SD-stem. The SD-stem
takes a s-prefix and exhibits gemination of the middle radical as well as the
224 Christian Huber

uniform vowel pattern. SD-stem attestations mostly come from literary

contexts. They are quite rare and are found only with a handful of verbs and
possibly do not have the full paradigm of forms that exist in the regular
stems. In (58) the forms of preterite and present are given schematically,
the three root consonants for convenience represented in capitals by the
root Vptr 'loosen', which is also met in the SD-stem.

(58) SD-stem: usPaTTiR (preterite, 3sgm)

usPaTTaR (present, 3 sgm)

Despite the fact that the SD-stem combines the morphological characteristics
of both the D- and S-stems, however, the respective forms never yield an
interpretation of a pluralic transitive verb embedded under a matrix causa-
tive, nor do they yield causativised versions of transitivised unaccusatives.
Rather, the attested instances merely correspond to the D- or S-stems of the
respective verbs (see Kouwenberg 1997: 336ff for a detailed discussion).
To the extent that the issue is real, I will submit tentatively that the reasons
for this might have to do with morphology and the interpretability of mor-
phological encoding. Let us therefore take a brief look at the morphological
composition of the D- and S-stems.
The fact that the D-stem combines two seemingly unrelated properties -
transitivity and eventual plural - strongly suggests that it results from the
conflation of two originally distinct categories. This view receives support
from the D-stem's morphological make-up, although I will not speculate
here about what historical development(s) may have led to this situation.
Let us consider the morphological characteristics of the D-stem: (i) it
displays a specific "^-coloured" vowel pattern, and (ii) it displays gemina-
tion of the middle radical, which occurs in all tenses. However, neither of
these characteristics is restricted to the D-stem. As illustrated in Table 5,
gemination of the middle radical is found not only in the D-stem but also in
"present tense" of G- and N-stems. 33 The specific "w-coloured" vowel pat-
tern is common to both D- and S-stems.

Table 5.
gemination of C2
(G present) iPaRRa/i/uS
u-coloured (D, S present) uPaRRaS usaPRaS
vowel pattern (D, S preterite) uPaRRiS usaPRiS
s prefix
 Complex predicate structure andpluralised events in Akkadian 225

From this it follows that (i) gemination of Cj does not automatically trigger
the special «-coloured vowel pattern, and (ii) the special «-coloured vowel
pattern does not automatically trigger gemination of C 2 . That is, «-vocali-
sation and gemination of C2 do not depend on each other.
Also, each seems to represent a different property. The fact that gemina-
tion of C 2 is availabe for present tense also of unaccusatives and other types
of non-agentive and stative predicates suggests that gemination is not a
morphological indicator of a transitive or causative relation or the presence
of an external argument or some other causer, despite being compatible with
its presence. Similarly, the fact that the «-vocalisation pattern is also a mor-
phological feature of the S-stem, which serves as a causative of transitives,
unergatives and unaccusatives and does not display verbal plural properties
as those witnessed in section 4, suggests that the «-vocalisation is not a
morphological indicator of verbal plural, despite being compatible with it.
However, it can be seen that the «-vocalisation occurs when some notion of
causativity is involved, as it is found in both the D- and S-stems.
In sum all this indicates that the two morphological characteristics of the
D-stem do not form a unit in only together expressing a single, unique
property but rather hint at the D-stem being a conflation of two originally
distinct categories. It may therefore be assumed that the two effects of the
D-stem - transitive/causative, eventual plural - correspond to the two mor-
phological characteristics under discussion, namely gemination of the middle
radical and the specific vowel pattern, as indicated in (59).

(59) eventual plural = morphologically gemination of C 2

encoded by:

transitive/causative = morphologically «-coloured vowel pattern

encoded by: (apophony)

In a non-linear model of morphology that recognises distinct morphological

tiers (e.g. McCarthy 1981; Guerssel & Lowenstamm 1993) the verb forms
uparris (D-stem preterite) and usapris (S-stem preterite) from a root Vprs (in
the G-stem: paräsu 'separate') could - somewhat simplified - be analysed
as in (60) and (61), respectively (irrelevant details omitted). (60) and (61)
make also use of Guerssel & L o w e n s t a m m ' s (1993, 1996) apophonic path
0=>i=>a=>u=>u. The valency-sensitive apophonic vowel [u] in the prefix
is derived from the prefix-vowel [a] of the G-stem and indicates an increase
of valency (i.e. causative), 34 the apophonic vowel [i] is "tense"-sensitive and
changes to [a] when (morphologically) deriving the present forms uparras
226 Christian Huber

and usapras from the preterite forms uparris and usapris, respectively. In
(60) eventual plural is morphologically encoded by gemination of the middle
radical, the transitive/causative value is encoded by the vowel pattern, more
precisely the prefix vowel [u]. In (61), where no gemination is present, 35
the causative value is doubly encoded: once by the vowel pattern (again,
more precisely the prefix vowel [u]) and once by the causative prefix s.

(60) D-stem (irrelevant details omitted)

plural of (sub-)event
iconically encoded by: gemination of C2

vprs
••' valency-sensitive \
apophonic vowel: tense"-sensitive
a=>u: increase of apoph. vowel: )
valency i=>a: preterite => present ./'
(causative) ...·•''

transitive/causative encoded in: vowel pattern (more precisely, [u])

(61) S-stem (irrelevant details omitted)

Vprs ρ r s
/ valency-sensitive \ | | | ..·•• ....
/ apophonic vowel: \ CV-CVCCVC / "tense"-sensitive
!
1 a =fu: increase of ].... I l l I > apoph. vowel: )
\ valency / u ''; s a 'ι \ . i=>a: preterite => present ../
(causative) ...·••'' '•

causative encoded in: s + vowel pattern (more precisely, [u])

With regard to the morphological marking of the causative properties of the

D- and S-stems we observe the situation sketched in (62):

(62) D-stem: vP-internal causative morphologically simple indication by

(transitive): w-vocalisation (apophony)

S-stem: vP-external causative morphologically complex indication by

(matrix causative): 5-prefix + «-vocalisation (apophony)

I would like to propose tentatively that the apparent impossibility of inter-

preting SD-stem forms as causativised D-stem forms (i.e. transitives derived
 Complex predicate structure andpluralised events in Akkadian 227

from G-stem unaccusatives on the one hand and verbal plurals derived from
G-stem transitives on the other hand) might have to do with the fact that the
causative value of the S-stem is morphologically doubly encoded. Suppose
the D-stem requires u of the vowel pattern to be interpreted as internal
causative (i.e. transitive), whereas the S-stem requires u of the vowel pat-
tern to be interpreted as one component of the external (i.e. matrix) causa-
tive, the other component being the prefix s. Interpreting a SD form as a
causativised D-stem form would then require to interpret u at the same time
as indicating internal causative (i.e. transitive) as well as external (i.e. ma-
trix) causative. Assuming that the morphological marker u (or the operation
of apophony behind it) could not receive such simultaneous double inter-
pretation (i.e. be employed to signal two different derivational operations)
would explain for the apparent absence of SD forms being interpreted as a
causativised version of the corresponding D-stem form. In addition, it fol-
lows that in Akkadian any morphologically indicated causativising opera-
tion could be applied only once.

7. Some concluding remarks

In this paper I have proposed an analysis of the Akkadian D-stem that allows
to account for its effects when applied both to unaccusative and transitive
verbs. I suggested that the D-stem represent a transitive (v-VP) configuration
with some additional head or feature that pluralises events, other effects fol-
lowing f r o m the type of event involved and how it is pluralised. Mapping
transitives onto that configuration therefore does not increase their valency
but endows them with the verbal, or rather eventual, plural property as il-
lustrated in section 4. Mapping unaccusatives onto that configuration makes
them transitive but the verbal plural property is concealed by the fact that
pluralising the event does not result in a multitude of states here that need
to be predicated of different entities. As only a limited set of data and rele-
vant questions could be treated here, a wider scope of investigation will
certainly yield a richer picture. The issue of verbal plural of the type en-
countered in Akkadian does not seem to have received much attention in
the generative literature and clearly merits further research.
In the majority of cases, there obtains a predictable and systematic relation-
ship between a v e r b ' s G- and D-stem. As is the case with the corresponding
stems in other Semitic languages, however, also the Akkadian D-stem is not
a fully productive category but may be subject to lexicalisation in varying
228 Christian Huber

degrees, beside some dialectal variation. Not all Akkadian verbs may occur
in the D-stem, the D-stem may also be restricted to special meanings or
usages. One does thus not expect a fully coherent picture. To provide just
one example, consider (63), which is a case of a verb that has no G-stem
but occurs in the D-stem only and allows for a reading that is neither causa-
tive nor displays any verbal plural properties (see (i)). In addition, it allows
for a valency alternation without a change of stem.

(63) zummü (i) 'lack χ (=ACC), be deprived of χ (=ACC)'

(ii) 'deprive y (=ACC) of χ (=ACC), make y lack x'

Deviations from the regular pattern are also found in the corresponding stems
of other Semitic languages, e.g. in Hebrew, where some instances of unac-
cusative verbs can be found in the corresponding stem (data from Joosten
1998: 223):

(64) Biblical Hebrew:

Qal (=G-stem) Pi'el (=D-stem)

kahah 'grow dim, grow faint' kihhah 'grow faint'

samah 'grow, sprout' simmeh 'grow abundantly'
qahah 'become blunt' qehhah 'become blunt'

In modern Arabic dialects some grammaticalisation processes can be ob-

served. Greenberg (1991: 580) mentions that in peninsular Arabian dialects
form II, which corresponds to the Akkadian D-stem, replaced form IV,
which corresponds to the Akkadian S-stem. Kaye and Rosenhouse (1997:
296) report that modern Arabic dialects often replace stem IX, which builds
unaccusative colour predicates, by stem II, which otherwise yields transitive
predicates.

(65) Yemenite Arabic (Kaye and Rosenhouse 1997:296):

II: hammar 'turn red' (< IX ('i)hmcirra)

Whether there is some development in the D-stem discernible in the history

of the Akkadian language remains yet to be investigated.
 Complex predicate structure andpluralised events in Akkadian 229

Acknowledgements

I would like to thank the people w h o over the years have discussed with me
issues related to this paper, notably Henry Davis, Jean L o w e n s t a m m , David
Pesetsky, D o m i n i q u e Sportiche, Tim Stowell, and especially Friedrich
Neubarth. Parts or versions of this paper w e r e presented at the 35th
C o l l o q u i u m of Linguistics (2000, University of Innsbruck); at Ö L T 2000
(University of Graz); at the 2001 Conference on the Syntax and Semantics
of Semitic L a n g u a g e s at the University of Southern California, Los
Angeles; and at Ö L T 2 0 0 2 (University of Innsbruck). Let m e again express
my thanks to the organisers of C S S S L for reimbursing m y travel expenses
and providing m e a c c o m m o d a t i o n in Los Angeles. I am also grateful to the
two reviewers for their c o m m e n t s and to Gillian R a m c h a n d for additional
comments.

Notes

1. For details on the language see the grammars of Soden (1995), Buccellati (1996),
and Huehnergard (1997). (Note however that Buccellati and Huehnergard do
not always indicate attestations for the data they present.) I give data in bound
transcription, using standard semitistic transliteration symbols. By convention,
s = [J], h = [x], 5 = [?]; s and t are emphatics; ä, e, Τ, 0 represent long vowels;
ä, e, i, ü represent long vowels resulting from contraction. For some sugges-
tions as to the actual realisation of the corresponding Akkadian graphs see e.g.
Diakonoff 1992 and Soden 1995. Further symbols are: d (emphatic), 1 = [Ϊ], h
= [h]. As a citation form I use the infinitive, which also was the citation form
in ancient Mesopotamian lexical and grammatical works. Abbreviations used
in glosses and tables are as follows. V = root; C, V = consonant, vowel; EXT =
external argument; DN = devine name, RN = royal name; cases: NOM = nomi-
natve, GEN = genitive, ACC = accusative, ACC.INST = instrumental accusative;
DAT = dative, OBL = oblique; PP = preposition phrase; abs = absolute state,
cstr = construct state; verb stems: G = G-stem, D = D-stem, S = S-stem, Ν =
N-stem; trans. = transitive, intrans. = intransitive; REFL = reflexive; inf = in-
finitive, sub = subjunctive, prec = precative; gender: f = feminine, m = mascu-
line, c = communis; number: sg = singular, pi = plural (with nouns only plural
will be indicated in glosses); CL = clitic(ised pronoun), possCL = possessive
pronominal clitic ('your', 'his', etc. case of clitics will be indicated explicitly
only with dative clitics); persons: 1, 2, 3 = 1st, 2nd, 3rd person; (thus,
'possCL3sgm' = 'possessive pronominal clitic 3rd person singular masculine');
vent = ventive; tenses: pres = present, pret = preterite, perf = perfect, stat =
Stative; NEG = n e g a t i o n , REL = relative particle; MA: t h e enclitic particle -ma,
230 Christian Huber

which I will not discuss, is glossed as MA. Although Akkadian data are given
in bound trascription, a clitic will be separated from its host by a hyphen in or-
der to make glosses more transparent. In glosses to verb forms, the basic stem
meaning is followed by the indications of the stem, tense, and per-
son/number/gender, separated by a hyphen from cliticised pronouns (i.e. pro-
nominal affixes) or other clitics. The form umahhisüsuma 'they beat him' will
thus be spelled out as umahhisü-su-ma and glossed as beat(D).pret.3plm-
CL3sgm-MA, indicating that the verb is a D-stem form in the preterite, 3rd per-
son plural masculine, followed by the 3rd person singular masculine pronomi-
nal affix (as its cliticised object), which in turn is followed by the enclitic par-
ticle -ma. Example sentences are accompanied by indications of the textual
source, given in standard Assyriological abbreviations. For a key to these ab-
breviations as well as for general quotations of Akkadian vocabulary items
outside a context (mainly verbs quoted merely as infinitives) see the dictionar-
ies AHw and CAD.
2. In contrast to Arabic and Hebrew the D-stem template cannot host quadriliteral
(or reduplicated biliteral) roots in Akkadian. Quadriliteral roots appear with a
prefix η or s instead, e.g. Vblkt N: nabalkutu 'to cross, pass over', S: subalkutu
'to let/make cross, pass over' (*bulkutu).
3. That verbs may acquire special meanings is not uncommon among languages.
In some instances relevant here it is also doubtful whether the G-stem and the
D-stem belong to the same root. However, the possibility of a D-stem does not
depend on the existence of a corresponding G-stem verb, as can be seen from
the fact that the D-stem may also be employed in denominal verbs, and that a
number of verbs are not found in the G-stem but only in the D-stem (see foot-
note 4). This might indicate that the D-stem (and perhaps also other stems)
could be derived directly from the root.
4. See Kouwenberg (1997) for an excellent survey of actually attested D-stem
forms and their distribution among the dialects/language stages and text gen-
res, also in relation to other verb stems.
Of course, no exact numbers can be given of attested D-stem forms and their
distribution. For mere orientation purposes, the following - approximate -
figures may be gleaned from Kouwenberg's survey (note, however, that
Kouwenberg himself hesitates to give any figures): ca. 165 G-stem verbs (ex-
cluding unergatives) do not undergo a valency alternation in the D-stem but
display verbal plurality instead. Among these are ca. 9 verbs whose D-stems
can be used not only to underline plurality but, in another use, may also ex-
press a specialised or somewhat different meaning (in some cases indicating a
higher degree of agentivity). In ca. 13 further verbs that do not undergo a va-
lency alternation, the semantic relationship between the G- and D-stems is gen-
erally not, or not fully, predictable. The lexicalised forms are usually restricted
to a specific dialect. With a handful of G-stem verbs, the D-stem primarily
 Complex predicate structure andpluralised events in Akkadian 231

seems to express a conscious action (perhaps in combination with plural action).

For ca. 109 further verbs, the situation is not clear, mainly because they are
only poorly attested (in ca. 13 of these, there is no observable difference in the
use of the G-stem vs. the D-stem). In ca. 216 verbs (deadjectival, change of
state, motion, and psychological predicates) the D-stem brings about a transitive/
causative alternation (in ca. 20 of them, the D-stem is associated with a lexical-
ised meaning). Finally, ca. 64 verbs occur in the D-stem but apparently have
no corresponding G-stem. The D-stem is also used in a number of denominal
verbs. Some D-stems appear to occur only in hendiadys or serial verb con-
structions. It must be stressed again, however, that the above figures are given
merely for a coarse orientation and certainly do not represent the actual situa-
tion that prevailed in any of the dialects at the times when the language was
still spoken.
5. Blank entries in tables indicate that a verb is not attested in the respective stem.
6. As expected, the verbs in (7) differ from the transitives in (4) in that no adjec-
tive can be derived from them. That is, while for a verb such as rakäsu 'bind'
there is an adjective raksu "bound" denoting the state of some entity (i.e. the
verb's object) resulting from the event expressed by the verb, no adjective such
as the hypothetical *nabhu 'barked' can be derived from nabähu 'bark'. The un-
accusative verbs in (2) naturally come with an adjective, e.g. rapäsu 'be(come)
wide(r)' and rapsu 'wide(r)'.
7. Although no S-stem is attested of this verb there is however one attestation of
the St-stem (sutas^ulu), i.e. a passivised version of the S-stem meaning 'be
caused to cough'.
8. The causative element s could be etymologically related to the possessive verb
isu 'have' (lassu 'have-not', Assyrian 'there is/are no', < lä (=Neg.)+m/; cf.
also Arabic laysa 'is not', Hebrew yes 'there is', etc.). Historically, thus, the
Akkadian S-stem and its cognates in Semitic (and Afroasiatic) may perhaps
have originated from a causative /wve-construction like e.g. English John had
the students write three papers.
9. I take the spelling ma-hir-ti to be connected to mahritu 'earlier, former (sgf)',
as ekallu mahritu 'earlier palace' is also attested elsewhere (see e.g. CAD M,1
p. 110).
10. I leave open the question whether there may be some (more) internal structure
also to sP. Also note that the S-stem of transitive verbs often appears without
an embedded subject, raising the question of restructuring. Demotion of the
embedded subject gives such a construction a passive flavour. The issue cannot
be pursued here.
(i) summa assat awilim (...) mussa (<mut-sa) usdXk (...)
if wife.cstr.NOM man.GEN (...) husband.cstr.ACC-possCL3sgf kill(S).pret.3sgc (...)
'If a m a n ' s wife had her husband killed (because of another man, ...)' (KH §153)
232 Christian Huber

11. The reverse case is found e.g. in maräsu 'be(come) ill, distressed', where
causativisation appears to be via the S-stem in the regular case and only excep-
tionally via the D-stem.
12. It is also unclear why for example the transitivised/causativised version of
mädu 'be(come) (more) numerous, much/more' appears in the S-stem only and
no D-stem seems to have been possible.
13. Note that a construction such as John had/let/made Mary enlarge the springs
could not be yielded by applying the S-stem form of rabü because the S-stem
simply causativises the respective G-stem verb and unaccusatives such as rabü
'be(come) big(ger)' cannot undergo an "ergative" alternation, i.e. do not allow
for a transitive alternant in the G-stem which the S-stem could causativise.
Akkadian seemingly had no general periphrastic causative construction com-
parable to e.g. the English ma£e-causative.
Also note, in this connection, the proposal in Streck (1994: 177f.), where it
is suggested that the St2-stem of the verb kämt 'be(come) firm' correspond to a
causative of that verb's Dt-stem ('be made firm'=) 'be checked', yielding the
meaning 'have χ checked', 'let jc be checked' ("veranlassen, daß [χ] überprüft
wird"). AHw renders the meaning of the form in question as "überprüfen" ("to
check"). If Streck's assumption is on the right track, the St 2 -stem of the verb
känu constitutes an exceptional case in Akkadian. Perhaps restructuring is an
issue here as well.
14. An anonymous reviewer points out to me Doron's (2003) work on verbal tem-
plates in (Modern) Hebrew, of which I was not aware. Doron's analysis, which
cannot be discussed here in any greater detail, assumes that two additional
kinds of functional heads beside little ν are involved in the syntactic construc-
tion of verbs: agency heads and voice heads. In that approach, the Hebrew in-
tensive template (i.e. the Hebrew counterpart of the Akkadian D-stem) in-
volves an agency head that is a modifier of the root. It classifies the event as
action, licenses ν and determines that the role of the external argument be Ac-
tor (c. also Kouwenberg's (1997) "high transitivity" approach). The characteri-
sation of the external argument as Actor (in the weak concept of action of
Doron 2003: 19) seems by and large compatible also with the Akkadian D-
stem evidence. The Hebrew causative template (i.e. the Hebrew counterpart of
the Akkadian S-stem) involves an agency head that merges with a fully con-
structed verb and introduces its own argument.
However, Doron's approach cannot be straightforwardly applied to Ak-
kadian. The properties of the Akkadian template system are somewhat differ-
ent from the Hebrew system of templates (in the sense of what is traditionally
referred to as stems). For example, iterative templates such as the Akkadian it-
erative stems are altogether absent from Hebrew. Also the verbal plural prop-
erty of the Akkadian D-stem does not seem to have a counterpart in the Modern
Hebrew intensive template and therefore plays no role in Doron's approach.
 Complex predicate structure andpluralised events in Akkadian 233

As for voice morphology, in contrast to Hebrew and Arabic, Akkadian has no

"internal passive" (i.e. passive morphologically signalled only by a different
vowel pattern) but resorts solely to templates involving the elements n- or -t-
(any discussion of voice-related phenomena in Akkadian is far beyond the
scope of the present article).
Nevertheless Doron discusses many contrasts with respect to which it would
be interesting to know what the situation in Akkadian would be. Yet they often
involve data of a kind that are hard (if not impossible) to retrieve in a dead lan-
guage, where native speakers are not available and all data have to be culled
from texts.
15. Most motion predicates are causativised by the S-stem, which would be com-
patible with an unergative analysis (i.e. involving an external argument in
Spec,vP). Cases like ukkusu 'drive away' in (20) are somewhat reminiscent of
e.g. the use of march in march the soldiers round the block.
16. Note that the G-stem object (i.e. garment, etc.) and correspondigly the second
object in the D-stem could be analysed both as object accusative or instrumen-
tal accusative. Also, verbs like labäsu often occur in the so-called "active"
stative construction (see below sub-section 4.2).
(i) subätam lä labsaku
garment, ACC NEG wear(G).stat.lsg
Ί am not dressed with a garment/wearing a garment' (TCL 18, 84: 7)
17. That Semitic "intensive" stems are at times used with some plural element was
already recognised by Brockelmann (1908:508) and others. For hints at a ver-
bal plural use of the Akkadian D-stem see also Ungnad (1949: 56 §35a) and
Soden (1995: 116 §88f).
18. Also see Soden, AHw pp. 1033. Note, however, that the interpretation of sibisu
as a variant of sebisu '7x' is not beyond doubt. Another option is 'until his be-
ing sated' (cf. sebu 'be(come) sated', see CAD S p. 204). Yet, this would not
affect the argument that (30) is an instance of "plural action".
19. Mostly these seem to be verbs of carrying, holding, wearing, accepting, receiv-
ing, or verbs of knowledge. That is, verbs that also allow for (or are restricted
to) a non-agentive or stative reading. See Rowton (1962) for discussion and a
collection of attestations.
20. Also note that (38b) seemingly cannot be derived from the semantics of the
verbal adjective.
(i) sabtu "seized", *"holding/seizing"
21. Except from one single Dt-stem instance there seem to be no attestations of a
passive D-stem form of sabätu "seize, hold" outside the stative conjugation.
22. The verb kasädu exists also with the fully agentive meaning "conquer". The D-
stem does not access this meaning, however.
23. The fact that the verbal plural property is retained under passivisation (see e.g.
(29)) may suggest that the relevant configuration is present also in passives
234 Christian Huber

and other constructions with passive-like properties that display verbal plural.
This is compatible with Chomsky's (1999: 6f.) view that also passives involve
a v-VP configuration, however with ν being phi-incomplete (as opposed to phi-
complete ν in the transitive/active case). See also Chomsky 1999: 14 and fn. 35
and 36 on participles.
24. For some - exceptional - pluralic unaccusative D-stem cases in Biblical Hebrew
see below footnote 31 and ex. (64).
25. Exceptions to this rule are rare and often occur only in technical language (for
example, an intransitive version of petü 'open' may denote the becoming dis-
tant of stars). I will ignore them here.
26. Since the D-stem is not barred by countability (see e.g. (31) above in section 4)
it must be assumed that such a number or series of transition may be explicitly
limited by a numeral.
27. Taking place of proto-events/proto-transitions without constituting a transition
will yield an atelic reading (e.g. conative {try, ...), etc.); possibly effected by
rendering the series incomplete). Note that (52) could also be seen as a transi-
tion series applying to proto-entities which in their totality constitute an entity
(e.g. an object).
28. Another relevant case here would be a series of (distributive) benefactive events
involving a mass noun(-like) direct object.
29. Another scenario might yield a change of degree with respect to progress. In a
type I-series progress in degree would be predicated of different entities, an
option not found in Akkadian.
30. Arabic form I/Hebrew Qal = G-stem; Ar. form II/Hebr. Pi'el = D-stem; cited
forms: perf 3sg masc.
31. Consider also the possibly exceptional case in (ib) from Hebrew, involving a
mass noun. If Joosten's (1998) rendering of (ib) is correct, this could be a case
of a type II series of grow-type events whereby multiple embedding is inter-
preted as a larger-than-usual degree.
(i) Biblical Hebrew: a. Qal: samah 'grow, sprout'
b. Pi c el: simmeh 'grow abundantly'(?) (said of hair)
(cf. Joosten 1998: 223)
32. See also above footnote 25.
33. Gemination of the middle radical might be seen as akin to or an instance of
reduplication. For the development of reduplicated roots from a 'repetition'
meaning towards 'imperfective/present' see e.g. Bybee, Perkins and Pagliuca
1994 ch 5.1 will not discuss here the morphology of iterative stems.
34. The prefixes in the G-stem are a-, ta-, i- {< yi- < ya-) and ni- (< na-). By apo-
phonic vowel change they become u-, tu-, u-, and nu-. In the case of i- and ni-
I take the (at least historically) underlying forms ya- and na- to serve as input
for the apophonic step a=>u.
 Complex predicate structure andpluralised events in Akkadian 235

35. Note, however, that there still seems to be an underlying geminate in the pre-
sent tense form of the S-stem which cannot surface with sound verbs only for
templatic reasons. Evidence for this comes from verbs with a so-called "weak"
first radical. Consider e.g. the S-stem forms usähiz (preterite) and usahhaz
(present) from the root V 'hz 'take' with the weak first radical ' (= [?]). The
weak first radical is deleted, allowing in the preterite form the vowel [a] to
spread to the empty position, resulting in a long vowel [aa] (= ä) (usa > hiz =>
usaahiz). In the present form, however, deletion of the weak first radical al-
lows to restore the underlying geminate. Here, [h] is attracted to the empty
slot, resulting in the geminate [hh], and [a] remains short (usa'haz =>
usahhaz). I assume this hidden geminate to be due to a morphological site in
the template of Akkadian verbs originally identified by Guerssel & Lowen-
stamm (1993) for Arabic which they call "Derivational Syllable". However, is-
sues of Akkadian templatic morphology cannot be discussed here any further.

References

Baker, Mark
1996 The Polysynthesis Parameter. New York/Oxford: Oxford University
Press.
1997 Thematic roles and syntactic structure. In Elements of Grammar,
L. Haegeman (ed.), 73-137. Dordrecht: Kluwer.
Belletti, Adriana, and Luigi Rizzi
1988 Psych-verbs and θ-theory. Natural Language and Linguistic Theory
6: 291-352.
Brockelmann, Carl
1908 Grundriß der vergleichenden Grammatik der semitischen Sprachen.
1. Band: Lautlehre und Formenlehre. Berlin.
Buccellati, Giorgio
1996 A Structural Grammar of Babylonian. Wiesbaden: Harrassowitz.
Bybee, Joan, Revere Perkins and William Pagliuca
1994 The Evolution of Grammar. University of Chicago Press.
Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
1999 Derivation by Phase. MIT Occasional Papers in Linguistics 18.
Diakonoff, Igor
1992 Proto-Afrasian and Old Akkadian. A study in historical phonetics.
Journal of Afroasiatic Languages 4: 1-133.
Doron, Edit
2003 Agency and Voice: The Semantics of the Semitic Templates. Natural
Language Semantics 11: 1-67.
236 Christian Huber

Greenberg, Joseph
1991 The Semitic "Intensive" as Verbal Plurality. In Semitic Studies in
honor of Wolf Leslau, Alan S. Kaye (ed.), 576-587. Wiesbaden: Har-
rassowitz.
Guerssel, Mohand and Jean Lowenstamm
1993 Apophony in Classical Arabic. Unpublished manuscript.
1996 Ablaut in Classical Arabic Measure I Active Verbal Forms. In Stud-
ies in Afroasiatic Grammar, Jacqueline Lecarme, Jean Lowenstamm
and Ur Shlonsky (eds.), 123-134. The Hague: Holland Academic
Graphics.
Hale, Kenneth L. and Samuel J. Keyser
1993 On Argument Structure and the Lexical Expression of Syntactic
Relations. In The View From Building 20; Essays in Linguistics in
Honor of Sylvain Bromberger, K. Hale and S. J. Keyser (eds.), 5 3 -
109. Cambridge, Mass.: MIT Press.
Huehnergard, John
1997 A Grammar of Akkadian. Atlanta, Georgia: Scolars Press.
Joosten, Jan
1998 The Functions of the Semitic D-stem: Biblical Hebrew Materials for
a Comparative-Historical Approach. Orientalia N.S. 67: 202-230.
Kaye, Alan and Judith Rosenhouse
1997 Arabic Dialects and Maltese. In The Semitic Languages, Robert
Hetzron (ed.), 263-311. London/New York: Routledge.
Kouwenberg, N. J. C.
1997 Gemination in the Akkadian Verb. Assen, The Netherlands: Van
Gorcum.
McCarthy, John
1981 A prosodic theory of nonconcatenative morphology. Linguistic In-
quiry 12: 373-418.
Pesetsky, David
1994 Zero Syntax. Experiencers and Cascades. Cambridge, Mass.: MIT
Press.
Poebel, Arno
1939 Studies in Akkadian Grammar. [Assyriological Studies 9] Chicago,
111.: The University of Chicago Press.
Rowton, M.B.
1962 The Use of the Permansive in Classical Babylonian. Journal of Near
Eastern Studies 21: 233-303.
Pustejovsky, James
1991 The syntax of event structure. Cognition 41: 47-81.
Soden, Wolfram v.
1995 Grundriß der akkadischen Grammatik, 3. Auflage. Wiesbaden: Har-
rassowitz.
 Complex predicate structure andpluralised events in Akkadian 237

Streck, Michael
1994 Funktionsanalyse des akkadischen St2-Stamms. Zeitschrift für Assyrio-
logie 84: 161-197.
Szabolcsi, Anna
1994 The Noun Phrase. In The Syntactic Structure of Hungarian, Syntax
and Semantics 27, Katalin E. Kiss and Ferenc Kiefer (eds.), 179-
274. San Diego and London: Academic Press.
Ungnad, Arthur
1949 Grammatik des Akkadischen. 3. Auflage. München: Biederstein
Verlag.

Bibliographical abbreviations

AHw = Wolfram ν. Soden, Akkadisches Handwörterbuch. 1965-1981. Wies-

baden: Harrassowitz
CAD = A. Leo Oppenheim, Erica Reiner et al., The Assyrian Dictionary of
the Oriental Institute of the University of Chicago. 1956-.
VSO and left-conjunct agreement:
Biblical Hebrew vs. Modern Hebrew

Edit Doron

VSO word order may on principle be the reflection of very different clause
structures. This chapter argues that for many languages where VSO order is
attested (Semitic, Celtic, Romance, Greek), it reflects a particular clause
structure, which I will call a "VSO clause," where the subject does not raise
out of the c-command domain of the tense head of the clause:

τ A
...s ...o...
A VSO clause is derived only if Τ has the following property:

(i) Τ does not have the EPP feature.

V S O clauses exhibit the pattern of "left-conjunct agreement" observed in

Irish by McCloskey (1986), and in Arabic by Aoun, Benmamoun, and
Sportiche (1994), where the verb agrees with the leftmost conjunct of a
postverbal conjoined subject, rather than with the full conjoined subject.
Example (1) shows left-conjunct agreement in Biblical Hebrew 1 :

(1) way-yiqqafi sem wa:-yepet ?et-hassimla:

and-tookJMS Shem and-Japheth ACC-the.garment
'And Shem and Japheth took a garment.' (Genesis 9: 2 3 f

This chapter shows that left-conjunct agreement follows from (i) above,
and more precisely from the weaker (ii):

(ii) Τ does not attract DP.

240 Edit Doron

I assume (following Borer 1986 and Chomsky 1998) that what triggers the
attraction of DP to Τ is the Agree relation between Τ and DP. Therefore, if
Τ attracts DP, then they must be related by Agree. But if Τ has property (ii) -
i.e., it does not attract DP - then Τ need not be related by Agree to DP.
Rather, I propose that the Agree relation be defined to hold between Τ and
the minimal D constituent closest to Τ which allows the derivation to con-
verge. In a derivation which requires D to move, the minimal D that does
not lead to a violation of the constraints on movement (such as the Coordi-
nate Structure Condition) is the full subject DP. But in a derivation where
D is not required to move, the minimal D constituent is a D head. In section
3, I show that in clauses with a conjoined subject, it is the head of the left
conjunct that is the D head closest to T.
VSO clauses share property (ii) with another very different type of
clause, those with an expletive subject. In this type of clause, Τ has the EPP
feature - i.e., it does not have property (i). The EPP feature of Τ is satisfied
by merging an expletive rather than by attracting the subject; that is, clauses
with expletive subjects have property (ii). Yet both VSO clauses (which
also have property (i)) and clauses with expletive subjects (which do not
have property (i)) exhibit left-conjunct agreement, which shows that this
kind of agreement follows from a property weaker than (i). It will be shown
in section 3 that left conjunct agreement follows from property (ii) - i.e.,
from the lack of DP-raising to Τ - either because Τ does not have the EPP
feature (property (i)), or because the EPP feature is satisfied by merging an
expletive.
Left-conjunct agreement is illustrated below for English, in a clause
with an expletive subject, and for Modern Hebrew, in a clause with a null
expletive:

(2) a. In the school there was a library and a terminal room.

b. ??In the school there were a library and a terminal room.

(3) Modern Hebrew

a. hayta li sipriya ve-fiadar mafisebim
w a s J F S DAT.me library.F and-room.M (of)computers
Ί had a library and a terminal room.'
b. ??hayu li sipriya ve-fiadar mafisebim
w e r e J p DAT.me library.F and-room.M (of)computers
Ί had a library and a terminal room.'
 VSO and Left-Conjunct Agreement 241

In general, VSO clauses are not expletive constructions, since they do not
show the defmiteness effect found in (2) and (3). (1) above, for example, is
a VSO clause which is clearly not an expletive construction, since its sub-
ject is definite. In addition, not every VSO sequence is a VSO clause. VSO
word order is found in Modern Hebrew, but not VSO clauses. First, VSO
word order occurs in Modern Hebrew in a null expletive structure such as
(3) above. In addition, Modern Hebrew has VSO sequences following any
preverbal constituent. 3 But there is no VSO sequence in Modern Hebrew
which by itself constitutes a clause. Example (4) below shows that clause
initial VSO is ungrammatical in non-expletive constructions, irrespective of
the agreement features of the verb. (5) shows that VSO is possible only
when some other constituent precedes the verb. The contrast between (5a)
and (5b) indicates that left-conjunct agreement is disallowed. These con-
structions are therefore different in some crucial way from the expletive
construction in (3), which allows left-conjunct agreement:

(4) Modern Hebrew

a. *yiqafi sem ve-yepet ?et-hasimla
will-take.iMS Shem and-Japheth ACC-the.garment
'Shem and Japheth will take the garment.'
b. *yiqfiu sem ve-yepet ?et-hasimla
will-takeJ/» Shem and-Japheth ACC-the.garment
'Shem and Japheth will take the garment.'

(5) a. *mafiar yiqafi sem ve-yepet ?et-hasimla

tomorrow will-take J M S Shem and-Japheth ACC-the.garment
'Shem and Japheth will take the garment tomorrow.'
b. mafiar yiqfiu sem ve-yepet ?et-hasimla
tomorrow will-takeJp Shem and-Japheth ACC-the.garment
'Shem and Japheth will take the garment tomorrow.'

Left-conjunct agreement in non-expletive constructions seems to be attested

in languages which, like Biblical Hebrew, have VSO clauses:

(6) a. Standard Arabic

laiibat maryam wa- zayd fi 1-bayt
played.JFS Mariam.F and-Zayd.M in the.house
'Mariam and Zayd played in the house.' (Rana Fahoum, p.c.)
242 Edit Doron

b. Modern Irish
da mbeinn -se agus tusa ann
if be.COND.Is EMPH and you there
'if you and I were there' (McCloskey and Hale 1984: 31a)
c. Spanish
Estaba abierta la tienda y el mercado
was,3S open.FS the shop.F and the market.Μ
'The shop and the market were open.' (Rodrigo Gutierrez, p.c.)

d. Greek
Irthe ο Pavlos kai ο Giannis sto parti
c a m e J S the Paul and the John to-the party
'Paul and John came to the party.' (Anastasia Giannakidou, p.c.)

1. Word Order in Biblical Hebrew and Modern Hebrew

The prevalent word order in Biblical Hebrew is verb initial, as shown again
in (7) below. In Modern Hebrew, on the other hand, the prevalent word
order is SVO, as shown in (8a), whereas verb initial sentences are in gen-
eral ungrammatical, as shown again in (8b):

(7) Biblical Hebrew

V S O
hir?a-ni: yhwh ?o:tka: melek ?al ? a ra:m
showed-me God ACC.you king over Syria
'The Lord hath showed me that thou shalt be king over Syria.'
(2 Kings 8: 13)
(8) Modern Hebrew
S V O
a. haseret her?a-li ?et-dani menaceafi ba-tafiarut
the.movie showed-me ACC-Dani winning in-the.race
'The movie showed me Dani winning the race.'
V S Ο
b. *her?a-li haseret ?et-dani menaceafi ba-tafiarut
showed-me the.movie ACC-Dani winning in-the.race
'The movie showed me Dani winning the race.'
 VSO and Left-Conjunct Agreement 243

SVO is also often attested in Biblical Hebrew. Example (9), exactly as is, is
a perfect grammatical sentence of both Modern and Biblical Hebrew:

(9) Biblical Hebrew

S V Ο
u-mo:se: ha:ya: ro?s: ?et-so:n yitro: fiotno:
and-Moses was.iMS keeping.MS' ACC-sheep (of)Jethro father-in-law.his
' N o w Moses kept the flock of Jethro his father-in-law.' (Exodus 3 : 1 )

In Biblical Hebrew, the only sentences not introduced by an overt comple-

mentizer are direct quotations, such as (7) above. Other main clauses are
always introduced by the complementizer w- 'and', also realized phonologi-
cally as u- , as in (9) above, or as way- in (10): 4

(10) V S O
way-yiqqafi mo:se ?et-matte ha:? E lo:hi:m b°- ya:do:
and-tookJA/S Moses ACC-rod (of)the.God in-hand.his
'And Moses took the rod of God in his hand.' (Exodus 4: 20)

In verb initial sentences such as (10), the complementizer ' a n d ' cliticizes to
the verb, yet it does not follow that the verb raises to C in Biblical Hebrew.
Rather, the complementizer in both (9) and (10) lowers to cliticize to the
left edge of the clause, similar to what is argued by Shlonsky (1988) for the
Modern Hebrew complementizer se 'that' and by McCloskey (1996b) for
Irish. Indeed, when sentences like (9) and (10) are preceded by adverbial
clauses (themselves introduced by complementizers), then the main-clause
complementizer follows the adverbial clause, as shown in ( l l a - b ) . Notice
that it should not be inferred from the syntax of the corresponding King
James translations that the clause following the adverbial clause is an em-
bedded clause. In Biblical Hebrew, unlike English, a clause with an overt
complementizer is possible as a main clause. In fact, the complementizer
'and' never introduces an embedded clause (cf. footnote 4):

(11) a. wa-yhi: 1- sibTat hayya:mi:m u- me hammabu:l

and-was to-seven the.days and-waters (of)the.flood
ha:yu: i a l ha:?a:rss
were upon the.earth
'And it came to pass after seven days, that the waters of the flood
were upon the earth.' (Genesis 7: 10)
244 Edit Doron

b. wa-yhi: miq-qes ya:mi:m way-ya:be qayin mip-pri:

and-was in-end (of-)days and-brought Cain from-fruit
ha:? a da:ma:
(ο f)the. ground
'And in process of time it came to pass, that Cain brought of the
fruit of the ground.' (Genesis 4: 3)

2. Conjoined Subjects in Biblical Hebrew

As noted in the standard grammars of Biblical Hebrew (e.g., Gesenius

1910, Joüon 1923), the verb agrees fully with a conjoined subject in SV
clauses, as in (12), but it agrees with the left conjunct in VS clauses, e.g.,
(13). I list here more examples of left-conjunct agreement simply because it
is more exotic:

(12) u-mo:se: ?ah a ro:n w-fiu:r ?a:lu: ro:s h a g g M a :

and-Moses Aaron and-Hur climbedJA/P head (of)the.hill
'And Moses, Aaron and Hur went up to the top of the hill.'
(Exodus 1 7 : 1 0 )

(13) a. wat-tasar d 3 bo:ra u:-ba:ra:q ben ? a bi:no:?am

and-sang.3FS Deborah and-Barak son (of)Abinoam
'Then sang Deborah and Barak the son of Abinoam.' (Judges 5 : 1 )

b. way-ya:mot na:da:b wa-? a bi:hu: lißne ? a bi:hem

and-died.3MS Nadab and-Abihu before father.their
'But N a d a b and Abihu died before their father.'
(1 Chronicles 24: 2)

c. u-ba:ta: ?atta: w-ziqne yisra:?el ?el mslsk

and-will.come.2MS' you and-elders (of)Israel to king
misrayim wa-? a mart8m ?ela:w
(of)Egypt and-will.say.2iW to.him
'And thou shall come, thou and the elders of Israel, unto the king
of Egypt, and ye shall say unto him . . . ' (Exodus 3 : 1 8 )

d. way-yiqafi ?abra:m w 3 -na:fio:r la:hem na:si:m

a n d - t o o k J M S Abram and-Nahor DAT.themselves wives
'And Abram and Nahor took them wives.' (Genesis 11: 29)
 VSO and Left-Conjunct Agreement 245

e. wat-ta:qa:m ribqa: w"-naS a ro:teyha wat-tirkabna: Tal

a n d - r o s e J f S Rebecca and-maids.her a n d - r o d e . i F P on
hagg 3 malli:m
the. came Is
' A n d Rebeka arose, and her damsels, and they rode upon the
camels.' (Genesis 24: 61)

f. wat-taYan ra:fiel w n -le?a: wat-to:marna: lo:

a n d - a n s w e r e d J F S Rachel and-Leah and-said.JFP to.him
ha-9o:d la:nu: fieleq w°-nafi a la: b-bet ?a:bi:nu:
Q-yet to.us portion and-inheritance in-house (of)father.our
' A n d Rachel and Leah answered and said unto him, Is there yet
any portion or inheritance for us in our father's house?'
(Genesis 31: 14)

g. way-yissa: da:wid w a -ha:?a:m ? a ssr ?itto:

a n d - l i f t e d J M 5 David and-the.people that with.him
?st-qo:la:m way-yibku:
ACC-voice.their and-wept.3MP
' T h e n David and the people that were with him lifted up their
voice and wept.' (1 Samuel 30: 4)

In the examples in (13), the form of the verb is singular, yet the subject is
clearly plural. It is implausible to analyze these examples as containing a
singular subject combined with a comitative phrase, though this is the in-
terpretation often offered by traditional interpreters of the Bible, such as
Rashi (Rabbi Shlomo Yitzhaki, 1040-1105), and traditional translations
such as the King James Bible (see in particular the translation of 13c and
13e). 5 (13d), for example, contains a plural reflexive dative bound by the
subject. (13g) contains the idiom ' X lifted up X ' s voice' which is obligato-
rily reflexive: There is no lifting up anybody's voice but o n e ' s own. But
this entails, since X is marked as plural in the second part of the idiom, that
the subject is plural as well. The examples in (13) are, therefore, examples
with plural subjects.
The contrast between full agreement of the preverbal subject and partial
agreement of the postverbal subject is also found in the other languages
vith left-conjunct agreement. Postverbal left-conjunct agreement, as in the
(a) sentences below, alternates with full preverbal agreement, in the (b)
sentences:
246 Edit Doron

(14) Standard Arabic

a. laYibat maryam wa-zayd fi 1-bayt
played J F S M a r i a m . F and-Zayd.Af in the.house
'Mariam and Zayd played in the house.'

b. maryam wa-zayd laTiba:/* la?iba/*la?ibat fi 1-bayt

Mariam and Zayd played.3MDUAL/*.3MS /*.3FS in the.house
'Mariam and Zayd played in the house.'

(15) Spanish
a. Estaba abierta la tienda y el mercado
w a s . i S open.FS the shop.F and the market.M
'The shop and the market were open.'

b. La tienda y el mercado estaban abiertos/*estaba

the shop and the market were.3Ρ open.A//V*wasJs
abierto/abierta
open*FM/*FS
'The shop and the market were open.'

(16) Greek
a. Irthe ο Pavlos kai ο Giannis sto parti
c a m e J s the Paul and the John to-the party
'Paul and John came to the party.'

b. Ο Pavlos kai ο Giannis irthan/*irthe sto parti

the Paul and the John c a m e J M P / * c a m e J 5 to-the party
'Paul and John came to the party.'

3. V-Raising and Left-Conjunct Agreement in VSO Clauses

First, we must establish that V S O word order in Hebrew is indeed a case of

V-raising, similar to Irish (Chung and McCloskey 1987; McCloskey 1991;
1996a, 1996b; Koopman and Sportiche 1991; Duffield 1991, 1995;
Guilfoyle 1993; Bobaljik and Carnie 1996) and Arabic (Mohammed 1990;
Benmamoun 1992; Fassi Fehri 1993), and unlike other types of V S O lan-
guages such as Chamorro (Chung 1990), where the subject lowers to VP
rather than V raising.
 VSO and Left-Conjunct Agreement 247

It has already been argued extensively that the verb in Modern Hebrew
raises out of the VP. 6 First, there is evidence (Shlonsky 1987) that the verb
can precede sentential adverbs, like bevaday 'certainly':

(17) Modern Hebrew

a. hamore bevaday yasbir ?et-hasi?ur
the.teacher certainly will.explain ACC-the.lesson
' T h e teacher will certainly explain the lesson.'

b. hamore yasbir bevaday ?et-hasiiur

the.teacher will.explain certainly ACC-the.lesson
' T h e teacher will certainly explain the lesson.'

Second, there is evidence (Shlonsky 1991) that quantifiers are floated from
a postverbal position:

(18) hayeladim hebinu kulam ?et-hasi?ur

the.children understood &W.3MP ACC-the.lesson
'The children have all understood the lesson.'

Third, VP-ellipsis strands the verb (Doron 1990, 1999), which shows that
the verb has raised out of the VP. The second reading available for (19)
shows that VP-ellipsis has applied, stranding the verb in T:

(19) ?im misehu yedaber ?al abodato, gam dani yedaber

if someone will.speak about work.his, also Dani will.speak
a. 'If someone will speak about his work, Dani will speak too.'
b. 'If someone will speak about his work, Dani will too.'

In Biblical Hebrew, it is possible to show that the verb may be found to the
left of sentential adverbs, which indicates that it raises out of VP:

(20) Biblical Hebrew

we-lo: yiqqa:re: *io:d ?6t-simka: ?abra:m
and-NEG will.be.called any more ACC-name.yours Abram
'Neither shall thy name any more be called Abram.' (Genesis 17: 5)
248 Edit Doron

In addition, at least for the absolute form (ABS), the verb can be shown to
raise out of VP, since ABS precedes not only Τ (i.e., the tensed form of the
verb), but negation (NEG) as well, as mentioned by Levin (1971) 7 :

(21) a. ra:?o: ra:?i:ti ? s t - ? ° n i : Tammi: ?a§8r b 3 -misra:yim

see.ABS saw .Is ACC-affliction (of)people.my that in-Egypt
Ί have surely seen the affliction of my people which are in Egypt.'
(Exodus 3: 7)
b. ba:ko: lo tibke:
cry.ABS NEG will.cry.2M5
' T h o u shalt weep no more.' (Isaiah 30: 19)

The original position of ABS is within VP, as can be seen in untensed clauses.
The order ABS -Τ observed in (21) is not found in imperative clauses, which
are untensed. Rather, the order found in imperative clauses is T- ABS, as in
(22):

(22) simiu: samo:?a w-?al ta:bi:nu:

hear.lMP.MP hear.ABS and-NEG will.understand.2Λ//>
u-r?u: ra:?o: w-?al teda:?u:
and see.IMP MP see.ABS and-NEG will.know.2MP
' H e a r ye indeed, but understand not; and see ye indeed, but perceive
not.' (Isaiah 6: 9)

I will therefore assume that V raises in tensed clauses in Biblical Hebrew,

and I will attempt to answer the question posed as a consequence: How far
does V raise in a V S O clause? The answer that Aoun, B e n m a m o u n and
Sportiche (1994) have given for Arabic is that V raises to functional head F
beyond T. F is lower than C, since V S O order is possible in embedded
clauses introduced by an overt C. The motivation for Aoun, Benmamoun,
and Sportiche's answer is theory internal: Subject-verb agreement, accord-
ing to them, is a relation which holds between T, the head of the clause, and
its specifier, as shown in (23) and, exemplified in (24). Therefore, the sub-
ject must be in the specifier of TP even when the verb precedes it. Accord-
ingly, the verb must have raised to a functional head F higher than Τ in a
V S O clause:
 VSO and Left-Conjunct Agreement 249

(23) FP

V+T+F TP

SUBJ TP

agreement
tv+r VP

VP

tv OBJ

(24) FP

V+T+F TP
la'iibat
playedJre S U B J TP
Mariam

tv+T VP

tsuBJ VP

tv PP

fi 1-bayt
in the-house

Yet if agreement is a relation which invariably holds between Τ and its

specifier, then there is no structural difference between examples with left-
conjunct agreement, such as (14a), and examples with full agreement, such
as (14b). Indeed, Aoun, Benmamoun, and Sportiche deny the existence of
left-conjunct agreement, and argue that such examples involve sentence
conjunction where ellipsis of the verb and Right Node Raising have also
taken place:
250 Edit Dor on

(25) FP

FP PP

FP FP fi 1-bayt
in the-house
V+T+F TP V+T+F TP
e
laiibat / \
playedJra SUBJ T P SUBJ χρ
Mariam Zayd

tv+T VP tv+T VP

tsUBJ V P t S U B J V P

tv PP tv PP
I I
tpp tpp

In (25), the second occurrence of the verb 'play' is ellided, whereas the PP
'in the house' is Right Node Raised (RNR). Yet this account is problematic
even for Arabic. It predicts that the RNR constituent should show singular
agreement, since this constituent supposedly originates from two singular
clauses. This prediction cannot be tested with an RNR constituent such as
'in the house' in (25), which does not exhibit agreement. Yet in examples
where the RNR constituent is a predicate with overt agreement, this predic-
tion is systematically falsified. In (26), for example, the RNR constituent
yallabaini fi l-bayt 'play.3MD in the house' is obligatorily marked as dual
(D), but if it were raised from two singular clauses, it should be marked as
singular:

(26) Standard Arabic

ka:nat maryam wa-zayd yal?aba:ni fi 1-bayt
was.3FS Mariam and-Zayd play.3MD in the.house
'Mariam and Zayd used to play in the house.' (Rana Fahoum, p.c.)

Aoun and Benmamoun (1999) deny the existence of this problem by show-
ing that it is not attested in either Lebanese or Moroccan Arabic. Yet this
problem arises for Standard Arabic, as (26) shows, and moreover, it is also
found in Irish and Biblical Hebrew. The relevant example from Irish is
 VSO and Left-Conjunct Agreement 251

shown in (27), where the constituent which is Right Node Raised from two
singular clauses (according to Aoun, Benmamoun, and Sportiche), 'ndr sui,
is nevertheless plural:

(27) Irish
Bhinn fein agus an seanduine 'när sui.
was IS EMPH and the old-fellow IP sitVN
'The old fellow and I used to be sitting.' (McCloskey 1986: ex. 37)

In Biblical Hebrew as well, the putatively RNR constituents contain plural

anaphors - e.g., la:hem 'for themselves' in (13d), and qo:la:m 'their voices'
in (13g), which is an anaphoric part of the expression 'raise their voices'.
The plurality of these anaphors would be completely unexpected if they
originated from singular clauses.
Since the only argument in Aoun, Benmamoun, and Sportiche for rais-
ing V beyond Τ is to afford a unified account of agreement, and since this
attempt is not successful, we are left with no reason to assume that V raises
beyond Τ in VSO structures. Indeed, in McCloskey (1986, 1996a, 1996b),
it is argued with respect to Irish that the verb does not raise beyond T.
Rather, the verb is in T, while the subject is in a specifier of a head lower
than T. For simplicity's sake, I assume that this lower head is V, though it
is argued in Bobaljik and Carnie (1996) and McCloskey (1996b) that this
head is actually a functional head below Τ and above temporal adverbs
which follow the verb. Agreement in Irish holds between Τ and the lower
subject, as shown in (28):

(28) TP

V+T VP
agreement
SUBJ VP

tv OBJ

Something needs to be said about how the structure in (28) satisfies the
EPP. 8 According to Chomsky (1995), the EPP is satisfied by covert raising
of the features of the subject to T. Chomsky (1998) argues against covert
feature movement, and moreover proposes to view the EPP as a feature not
just of Τ but of functional heads in general. This feature is not necessarily a
252 Edit Dor on

lexical property of functional heads, but can be added independently into the
derivation. I propose that in a strictly VSO language such as Irish, Τ is not
compatible with the EPP feature, neither as part of its lexical specification
nor as an addition by the derivation. In Biblical Hebrew, Arabic, Romance,
and Greek, on the other hand, an EPP feature may be added to Τ in some
derivations, though it is not part of the lexical specification of T. Accord-
ingly, in these languages, a V S O structure such as (28) is derived with a
numeration which does not include the EPP feature. An SVO structure such
as (29) is the outcome of a different derivation, one which includes an EPP
feature but no expletive:

(29) TP

SUBJ TP

V+T VP
agreement
VP

tv OBJ

The two derivations in (28) and (29) have different agreement patterns,
based on the same operation Agree:

The operation AGREE (adapted from Chomsky 1998) 9

(a) The relation AGREE holds between the φ-features of Τ and the φ-
features of D which is closest to Τ (in terms of c-command) in T ' s
domain (all the nodes dominated by its sister).
(b) The values of φ-features are copied to Τ from the D related to it by
AGREE.
(c) If Τ has an EPP feature, D is raised to T.

In (30), if Τ has an EPP feature not satisfied by the pure merging of an ex-
pletive, it enters into the AGREE relation with the highlighted DP, since this
is the DP closest to T, and since this is the minimal constituent within the
closest DP that may move without violating the constraints on movement:
 VSO and Left-Conjunct Agreement 253

(30)

D NP CONJ DP

Yet, if Τ in (30) does not have an EPP feature (or if the EPP feature can be
satisfied by a pure merging of an expletive), no movement is forced, and
the AGREE relation holds with the closest D head, which is the boldfaced D
in (30). Crucially, I assume the asymmetric structure of coordinate structures
argued for by Larson (1990). where conjunctions head coordinate struc-
tures. I also assume that conjunctions lack any formal features, from which
it follows that the category they project is that of the conjuncts themselves.
Moreover, the number specification of a conjoined DP is not a morphosyn-
tactic feature of the conjunction head, as argued hy Farkas and Zee (1995).

4. Subject-Verb Agreement in OVS Clauses

In section 3, I showed that left-conjunct agreement to postverbal subjects

is a motivation for assuming VSO clause structure - i.e., a structure where
the subject does not raise beyond T. The question now is whether such a
sructure is also motivated in clauses where V is not clause initial, e.g., OVS
order. At first sight, it seems reasonable to continue assuming that in OVS
clauses as well, the subject does not raise beyond T. OVS clauses are sim-
ply VSO clauses with subsequent fronting of the object. By this reasoning,
the structure of (31a) should be (31b), which is derived from a structure
like (28) by raising OBJ to T:

(31) Biblical Hebrew

a. Ο V S
a
we-?et-? ma:sa: sa: m ?absa:lomtafiat yo?a:b
and-ACC-Amasa appointed Absalom instead (of)Joab
?al hassa:ba:
on the.army
'And Absalom made Amasa captain of the host instead of Joab.'
(2 Samuel 17:25)
254 Edit Doron

b. TP

tv toBJ

OVS sentences like (31a) are also attested in Modern Hebrew. As men-
tioned in the introduction, Modern Hebrew allows postverbal subjects on
condition that some other constituent - e.g., the object - precedes the verb.
Indeed, (31b) is proposed by Borer (1995) as the structure of Modern
Hebrew OVS sentences. Yet if this is the right structure for OVS sentences
in Modern Hebrew, and if left-conjunct agreement is accounted for struc-
turally, then the prediction is that (31 b) should exhibit left-conjunct agree-
ment in Modern Hebrew as well. But this is not the case, as already men-
tioned. In Modern Hebrew, only full agreement is attested:

(32) Modern Hebrew

a. *?et-haseper sama dvora ve baraq al hamadap
ACC-the.book p u t J F S Deborah and Barak on the.shelf
'Deborah and Barak put the book on the shelf.'
b. ?et-haseper samu d°vora ve baraq al hamadap
ACC-the.book p u t J P Deborah and Barak on the.shelf
'Deborah and Barak put the book on the shelf.'

The lack of left-conjunct agreement in Modern Hebrew is puzzling in view

of the fact that it exists in Biblical Hebrew. This leads us to suspect that the
structure in (31b) cannot be the right structure for OVS sentences in Modern
Hebrew. I claim that neither is it the right structure for Biblical Hebrew.
Despite the grammar-book generalization concerning Biblical Hebrew
left conjunct agreement, there are conjoined postverbal subjects in Biblical
Hebrew where full agreement is attested:

(33) Biblical Hebrew

w 9 -hanno:tErst mimmenna: yo:klu: ?ah a ro:n u-ba:na:w
and-the.remainder from.it will.eatJMP Aaron and- sons.his
'And the remainder thereof shall Aaron and his sons eat.'
(Leviticus 6: 9/16)
 VSO and Left-Conjunct Agreement 255

(34) way-yhi: ?afiar hadd°bari:m ha:?ell8: fia:t?u: masqe mslsk

and-was after things these o f f e n d e d J M P butler (of)king
misra:yim w 3 -ha:?o:pe: la-? a do:nehem l-mslek misra:yim
(of)Egypt and-the.baker to-lord.their to-king (of)Egypt
'And it came to pass after these things, that the butler of the king of
Egypt and his baker had offended their lord the king of Egypt.'
(Genesis 40: 1)

(35) w 3 -ta:psu: bo: ?a:bi:w w a -?immo:

and-will.hold.JjW at.him father.his and-mother.his
'Then shall his father and his mother lay hold on him.'
(Deuteronomy 21: 19)

Full agreement is specially puzzling in examples such as (36a), since in the

same chapter, three verses earlier, the same postverbal subject does trigger
left-conjunct agreement, as shown in (36b):

(36) a. way-yerdu: ?ela:w melek yisra:?el w-i:ho:sa:pa:t

and-descendedJA/P to.him king (of)Israel and-Jehoshaphat
u- melek ? E do:m
and-king (of)Edom
' S o the king of Israel and Jehoshaphat and the king of Edom went
down to him.' (2 Kings 3 : 1 2 )

b. way-yelek mslsk yisra:?el u-mslsk y 3 hu:da: u-mslsk

and-went.3A/5 king (of)Israel and-king (of)Judah and-king
? E do:m way-ya:so:bbu: dsrsk s M a t ya:mi:m
(of)Edom and-circledJA// 1 road seven days
' S o the king of Israel went, and the king of Judah, and the king of
Edom; and they fetched a compass of seven days' journey.'
(2 Kings 3: 9)

We must ask whether there is a structural difference between sentences

with full agreement and sentences with left-conjunct agreement. The answer
is provided by Moreshet (1967). In a comprehensive study of the complete
prose of the Bible (excluding poetry), Moreshet found 235 sentences with a
conjoined postverbal subject. In 210 of these sentences, agreement is with
the left conjunct, whereas in 25 examples the verb fully agrees with the
conjoined postverbal subject. Moreshet was able to discover a descriptive
generalization which captures the distribution of full versus left-conjunct
256 Edit Dor on

agreement. His generalization constitutes a necessary condition for full

agreement:

THE MORESHET GENERALIZATION (adapted from Moreshet 1967)

The verb in Biblical Hebrew agrees with the leftmost conjunct of a post-
verbal conjoined subject, unless either (I) or (II) hold:
(I) The verb is preceded in the clause by some constituent.
(II) A clitic is attached to the verb.

Examples (33) and (34) above fall under clause (I) of this generalization,
whereas (35) and (36a) fall under clause (II). 10 The problem is that the two
clauses of the Moreshet Generalizaton do not seem to constitute a natural
class of syntactic environments.
Fortunately, the analysis proposed for clitics in Semitic by Doron (1996)
and Doron and Hey cock (1999) makes it possible to subsume condition (II)
of the Moreshet Generalization under condition (I). According to this
analysis, clitics may be viewed as anaphors bound by preverbal constitu-
ents, as in (37):

(37) kol habben hayyilo:d hay?o:r-a: tasli:ku:-/iH.·

every son born the.river-ALLAT you.will.throw-/w/n
'Every son that is born ye shall cast into the river.' (Exodus 1: 22)

(37) is not an example of left-dislocation, as the constituent binding the

clitic is the quantifier kol habben hayyilo.d 'every son that is born'. Quanti-
fiers with 'every' do not undergo left-dislocation.
Crucially, the constituent binding the clitic is possibly empty, if previ-
ously mentioned in the discourse. This is clearly the case in (36a), for ex-
ample, where the object clitic refers to the prophet Elisha, mentioned earlier
in the text. The same is true for (35), where there is previous mention of a
rebellious son, to which the clitic is anaphoric. These clitics may therefore
be analyzed as bound by a preverbal constituent, just as in (37). The only
difference is that the preverbal constituent is overt in (37), but it is null in
(35) and (36a). Condition (II) therefore does not characterize any examples
that do not already fall under (I). The two necessary conditions can be col-
lapsed to a single one, which coincides with condition (I):
 VSO and Left-Conjunct Agreement 257

THE MORESHET GENERALIZATION ( r e v i s e d )

The verb in Biblical Hebrew agrees with the leftmost conjunct of a
postverbal conjoined subject, unless the verb is preceded in the clause
by some constituent.

The reformulation of the Moreshet Generalization is an improvement over

the original formulation for yet another reason. It now accounts as well for
cases where the anaphoric element which licenses full agreement is not a
clitic attached to the verb, but some other anaphor. An example is shown
in (38), where the anaphor is part of the conjoined subject itself (additional
examples are Exodus 29: 15, Leviticus 8: 19,22, Numbers 20: 10):

(38) way-yo:klu: way-yistu: hu: w-ha:? a na: si:m ? a ser ?immo:

and-ateJAfP and-drank JA/P he and-the.men that with.him
'And they did eat and drink, he and the men that were with him.'
(Genesis 24: 54)

The Moreshet Generalization in conjunction with our previous discussion

provides a necessary condition for the raising of the verb beyond T. Ac-
cording to the conclusions of section 3, full agreement is the result of the
subject raising to specifier of TP. Accordingly, if in a full-agreement struc-
ture the subject is found following the verb, then it must be that the verb
has raised beyond T. The Moreshet Generalization states a condition on
such a movement very similar to the condition on V-raising beyond Τ
known from the discussion of Germanic languages (den Besten 1983, Plat-
zak 1986a, and many others). This condition can be formulated as a V2
condition on V-raising beyond T:

THE V 2 CONDITION
The verb in Biblical Hebrew does not raise beyond Τ unless preceded
by some constituent.

In other words, the raising of V to a functional projection F beyond Τ is

possible only if some constituent - e.g., the object - occupies the specifier
position of FP: 11
258 Edit Dor on

(39) FP

tv toBJ

Descriptively, as we have seen, Modern Hebrew differs from Biblical He-

brew in two relevant respects:

(i) The verb is always preceded by some constituent, not necessarily the
subject.
(ii) There is no left-conjunct agreement when the subject follows the verb,
other than in expletive constructions.

The difference may be reduced to a single factor if we simply assume that

in Modern Hebrew, Τ has the EPP feature as a lexical property. This is why
in Modern Hebrew, unlike Biblical Hebrew, there is always some constitu-
ent preceding the verb, and there is no left-conjunct agreement. Even where
the object precedes the verb, the subject is in the specifier of TP and there-
fore triggers full agreement. Indeed, (39) is the structure proposed by
Shlonsky and Doron (1992) for Modern Hebrew OVS sentences. In that
chapter, independent evidence was presented in favor of (39) over (31b) as
the structure for OVS sentences in Modern Hebrew. The distribution of
left-conjunct agreement in Biblical Hebrew coupled with the lack of left-
conjunct agreement in Modern Hebrew, is an additional argument to the
same effect. The structure of Modern Hebrew SVO sentences, on the other
hand, is argued by Shlonsky and Doron (1992) to be as in (29), the struc-
ture proposed here for Biblical Hebrew SVO sentences as well. A similar
asymmetry between the position of a preverbal subject and a preverbal
object is argued for in Germanic by Zwart (1993a).
 VSO and Left-Conjunct Agreement 259

A d o p t i n g t h e f r a m e w o r k o f C h o m s k y ( 1 9 9 8 ) has m a d e it possible to ac-

count f o r t h e d i f f e r e n t distribution o f the E P P feature in the d i f f e r e n t lan-
guages. Τ in M o d e r n H e b r e w is assigned the E P P feature lexically (as in
English). In Biblical H e b r e w , it is not, but Τ m a y b e e n r i c h e d with an E P P
feature as part o f s o m e derivations but not o f others (the s a m e is true of
Arabic, G r e e k , and R o m a n c e ) . In Irish, in contrast, Τ is i n c o m p a t i b l e with
the E P P feature, w h i c h cannot be added to it either lexically or by the deri-
vation.

Notes

1. All the Biblical Hebrew translations are from the King James Bible (1611),
which is generally more literal than the other translations.
2. I am very grateful to Shraga Assif for the phonetic transcription of the Biblical
Hebrew data.
3. The preverbal constituent may be null even when it is not expletive, as is
generally assumed for "narrative inversion" and for "all-focus sentences," e.g.:
(i) Modem Hebrew
hitqaser ?aba selka
called father yours
'Your father called.'
4. The main clause complementizer 'and' has an important role for text cohesion
(for recent discussion see de Caen 1995 and Hatav 1997). It is in complementary
disiribution with other complementizers - e.g., ki: 'for/that', pen 'lest', h a ' Q '
(a yes-no interrogative complementizer) - which only introduce embedded
clauses:
(i) ?al ti:r?i: ki: sa:ma? ? e lo:hi:m ?el qo:l hannaTar
1
n e g fear,2FS for heard.iMS God to voice (of)the.lad
'fear not; for God hath heard the voice of the lad.' (Genesis 21: 17)
(ii) ki: ?a:mru glisti:m ρεη yaVasu: ha:?ibri:m
for said.iMP Philistines lest will.make.3MP the.Hebrews
fisrsb ?o: fiani:t
sword or spear
'For the Philistines said, lest the Hebrews make them swords or spears.'
(1 Samuel 13: 19)
(iii) way-yo:msr h a tafiat ? e lohi:m ?a:no:ki:
and-said.iMS β. instead God I
'and he said, Am I in God's stead.' (Genesis 30: 2)
260 Edit Doron

5. Genuine comitative phrases in Biblical Hebrew seem to be small clause ad-

juncts of the form [ S c and DP with him], e.g.:
(i) way-ya:bo: no:afi [ S c u-ba:na:w w-?isto: u-nse
and-cameJAiS" Noah and-sons.his and-wife.his and-wives
ba:na:w ?itto:] ?ε1 hatteba: mippne me hammabbu:l
(of)sons.his with.him to the.ark because waters (of)the.flood
'And Noah went in, and his sons and his wife and his sons' wives with
him into the ark, because of the waters of the flood.' (Genesis 7: 7)
6. Borer (1995) presents arguments that V does not always raise in Modern He-
brew, but the validity of these arguments is disputed in Doron (2000).
7. ABS is the absolute (i.e. non-construct) form of the verbal gerund used to redu-
plicate the verb for the purpose of strengthening the affirmative force of the ut-
terance. It is usually translated as 'surely' or 'indeed'.
8. One approach to the question of how the structure in (28) satisfies the EPP is
that of Alexiadou and Anagnostopoulou (1998). According to them, in pro-drop
languages, V-raising to Τ itself satisfies the EPP, since V in these languages
carries a clitic which has the feature [+D], I will not discuss this proposal in
the text, since it makes several wrong predictions. First, it predicts that every
pro-drop language has VSO clauses, a prediction clearly falsified by Modern
Hebrew. Second, by this approach, the subject cannot raise to the specifier of
TP in pro-drop languages. Rather, any preverbal subject is purely merged to
the specifier of TP, an A' position, and binds the thematic subject pro situated
in the specifier of VP (or some other projection lower than T). Yet it can be
shown that the subject does raise to the specifier of Τ in pro-drop languages.
Raised subjects differ in many of their syntactic and semantie properties from
purely merged constituents, as amply shown for Standard Arabic, Modern He-
brew and Japanese by Doron and Heycock (1999). For a very different view of
the EPP in VSO languages, see Massam (2000).
9. Chomsky's definition also includes an additional clause regarding the erasure
of the non-interpretable φ-features of an agreeing T.
10. There exist a couple of apparent counterexamples to this generalization, yet it
seems that even those can be explained away:
(i) wa-ykah a nu: 'MYa:za:r w°-?i:ta:ma:r
and-served.iA// 5 Eleazar and-Ithamar
'Eleazar and Ithamar executed the priest's office.' (1 Chronicles 24: 2)
The verb in (i) shows full agreement to the postverbal subject, yet neither does
it contain an object clitic, nor is it preceded by any constituent. But notice that
(i) appears in the context of a detailed list of all the temple officials in King
David's administration, at the point where the divisions of high priests, the de-
scendents of Aaron, are listed:
 VSO and Left-Conjunct Agreement 261

(ii) wMi-bne ?ah a ro:n mafilqo:ta:m bne ?ah a ro:n na:da:b

and-to-sons (of)Aaron divisions.their sons (of)Aaron Nadab
wa-? a bi:hu: ?Elia:za:r w s -?i:ta:ma:r
and-Abihu Eleazar and-Ithamar
way-ya:mot na:da:b wa-? a bi:hu: li^ne ? a bi:hsm u-ba:ni:m
and-died.iMS Nadab and-Abihu before father.their and-sons
lo: ha:yu: la:hsm wa-ykah a nu: 'islTaizair w n -?i:ta:ma:r
NEG were to.them and-servedJa/p Eleazar and-Ithamar
'Now these are the divisions of the sons of Aaron. The sons of Aaron;
Nadab and Abihu, Eleazar and Ithamar. But Nadab and Abihu died before
their father, and had no children: therefore Eleazar and Ithamar executed
the priest's office.' (1 Chronicles 24: 1-2)
The passage in (ii) is an explanation for why there are only two divisions of
high priests listed, not four, in spite of the fact that Aaron, the forefather of all
high priests, had had four sons (about two centuries prior to David's time). The
clause in (i), which is the last clause in (ii), is therefore probably not conjoined
to the preceding clause, but rather contains it as an adverbial clause: 'Because
Nadab and Abihu had died leaving no children while Aaron was still in office,
only Eleazar and Ithamar succeeded him as high priests.' If this is so, then (i)
does after all fall under the Moreshet Generalization.
11. In Arabic, Spanish, and Greek, left-conjunct agreement is optional, which may
indicate that in those languages, T-raising to F is independent of the fronting
of a constituent to the specifier of FP.

References

Abraham, Werner (ed.)

1983 On the Formal Syntax of Wesgermania.
Adger, David, S. Pintzuk, B. Plunkett and G. Tsoulas (eds.)
1999 Specifiers: Minimalist Approaches. Oxford: Oxford University Press.
Alexiadou, Artemis and Elena Anagnostopoulou
1998 Parametrizing AGR: word order, V-movement and EPP-checking.
Natural Language and Linguistic Theory 16.3: 491-539.
Aoun, Joshep and Elabbas Benmamoun
1999 Gapping, PF merger and patterns of partial agreement. In Fragments:
Studies in Ellipsis and Gapping, Elabbas Benmamoun and Shalom
Lappin (eds.), 175-192. Oxford: Oxford University Press
Aoun, Joshep, Elabbas Benmamoun and Dominique Sportiche
1994 Agreement, word order and conjunction in some varieties of Arabic.
Linguistic Inquiry 25: 195-220.
262 Edit Doron

Benmamoun, Elabbas
1992 Structural conditions on agreement. Proceedings ofNELS 22: 17-32.
Benmamoun, Elabbas and Shalom Lappin (eds.)
1999 Fragments: Studies in Ellipsis and Gapping. Oxford: Oxford Univer-
sity Press.
den Besten, Hans
1983 On the interaction of root transformations and lexical deletive rules.
In On the Formal Syntax of Wesgermania, Werner Abraham (ed.),
47-131. Amsterdam: John Benjamins.
Bobaljik, Jonathan D. and Andrew Carnie
1996 A minimalist approach to some problems in Irish word order. In The
Syntax of the Celtic Languages. Robert Borsley and Ian Roberts (eds.).
Cambridge: Cambridge University Press.
Borer, Hagit
1986 I-subjects. Linguistic Inquiry 17: 375-416.
1995 The ups and downs of Hebrew verb movement. Natural Language
and Linguistic Theory 13.3: 527-606.
Borsley, Robert and Ian Roberts (eds.)
1996 The Syntax of the Celtic Languages. Cambridge: Cambridge Univer-
sity Press.
Carnie, Andrew and Eithne Guiloyle (eds.)
2000 The Syntax of Verb Initial Languages. Oxford: Oxford University
Press.
Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
1998 Minimalist inquiries: the framework. MIT Occasional Papers in
Linguistics 15.
Chung, Sandra
1990 VPs and verb movement in Chamorro. Natural Language and Lin-
guistic Theory 8: 559-620.
Chung, Sandra and James McCloskey
1987 Government, barriers and small-clauses in Modern Irish. Linguistic
Inquiry 18: 173-237.
Cinque, Guglielmo and G. Giusti (eds.)
1995 Advances in Roumanian Linguistics. Amsterdam: John Benjamins.
DeCaen, Vincent
1995 On the Placement and Interpretation of the Verb in Standard Biblical
Hebrew Prose. Ph.D. dissertation, University of Toronto.
Doron, Edit
1990 VP-ellipsis in Hebrew. MS., The Hebrew University of Jerusalem.
1996 The Predicate in Arabic. In Studies in Afroasiatic Grammar,
Jacqueline Lecarme, Jean Lowenstamm and Ur Shlonsky (eds.), 7 7 -
87. The Hague: Holland Academic Graphics.
 VSO and Left-Conjunct Agreement 263

1999 V-movement and VP-ellipsis. In Fragments: Studies in Ellipsis and

Gapping, Elabbas Benmamoun and Shalom Lappin (eds.), 124-140.
Oxford: Oxford University Press.
2000 Word order in Hebrew. In Studies in Afroasiatic Grammar, Jacque-
line Lecarme, Jean Lowenstamm and Ur Shlonsky (eds.), 41-56. The
Hague: Holland Academic Graphics.
Doron, Edit and Caroline Heycock
1999 Filling and licensing multiple specifiers. In Specifiers: Minimalist
Approaches, David Adger, S. Pintzuk, B. Plunkett and G. Tsoulas
(eds.), 69-89. Oxford: Oxford University Press.
Duffield, Nigel
1991 Particles and Projections. Ph.D. Dissertation, USC.
1995 Particles and Projections. Dordrecht: Kluwer.
Eid, Mushira (ed.)
1990 Perspectives on Arabic Linguistics I. Amsterdam: John Benjamins.
Fassi Fehri, Abdelkader
1993 Issues in the Structure of Arabic Clauses and Words. Dordrecht:
Kluwer.
Farka§, Donca and Draga Zee
1995 Agreement and pronominal reference. In Advances in Roumanian
Linguistics, Guglielmo Cinque and G. Giusti (eds.). Amsterdam:
John Benjamins.
Gesenius, Friedrich Η. W.
1910 Hebrew Grammar. Edited and enlarged by E. Kautzsch, second
English edition revised by A.E. Cowley. Oxford: Clarendon Press.
Guilfoyle, Eithne
1993 Nonfinite clauses in Modern Irish and Old English. Proceedings of
the 29th Regional Meeting of the Chicago Linguistic Society, 199-214.
Hatav, Galia
1997 The Semantics of Aspect and Modality; Evidence from English and
Biblical Hebrew. Amsterdam: John Benjamins.
Joüon, Paul
1923 Grammaire de l'Hebreu Biblique. Rome: Institut Biblique Pontifical.
Koopman, Hilda and Dominique Sportiche
1991 On the position of subjects. Lingua 85: 211-258.
Larson, Richard
1990 Double objects revisited: reply to Jackendoff. Linguistic Inquiry 21:
589-593.
Lecarme, Jacqueline, Jean Lowenstamm and Ur Shlonsky (eds.)
1996 Studies in Afroasiatic Grammar. The Hague: Holland Academic
Graphics.
Lecarme, Jacqueline, Jean Lowenstamm and Ur Shlonsky (eds.)
2000 Research in Afroasiatic Grammar. Amsterdam: John Benjamins.
264 Edit Doron

Levin, Saul
1971 The Indo-European Languages. Albany, NY: SUNY University Press.
Massam, Diane
2000 VSO and VOS: Aspects of Niuean word order. In The Syntax of Verb
Initial Languages, Andrew Carnie and Eithne Guilfoyle (eds.), 97-
116. Oxford: Oxford University Press.
McCloskey, James
1986 Inflection and conjuction in Modern Irish. Natural Language and
Linguistic Theory 4: 245-281.
1991 Clause structure, ellipsis and proper government in Irish. Lingua 85:
259-302.
1996 a Subjects and subject positions. In The Syntax of the Celtic Languages,
Robert Borsley and Ian Roberts (eds.). Cambridge: Cambridge Uni-
versity Press.
1996b On the scope of verb movement in Irish. Natural Language and
Linguistic Theory 14: 47-104.
McCloskey, James and Kenneth Hale
1984 On the syntax of person-number inflection in Modern Irish. Natural
Language and Linguistic Theory 4: 487-533.
Mohammad, Mohammad
1990 The problem of subject-verb agreement in Arabic: towards a solution.
In Perspectives on Arabic Linguistics /., Mushira Eid (ed.). Amster-
dam: John Benjamins.
Moreshet, Menachem
1967 Hanasu hakodem lishney nos'im bilshon hamikra. Leshonenu 31: 1-10.
Platzak, Christer
1983 Germanic word order and the Comp/Infl parameter. Working Papers
in Scandinavian Syntax 2. Trondheim.
Shlonsky, Ur
1987 Null and Displaced Subjects. Ph.D. Dissertation, MIT.
1988 Complementizer cliticization in Hebrew and the Empty Category
Principle. Natural Language and Linguistic Theory 6: 191-205.
1991 Quantifiers as functional heads: a study of quantifier float in Hebrew.
Lingua 84: 159-180.
Shlonsky, Ur and Edit Doron
1992 Verb-second in Hebrew. Proceedings ofWCCFL 10: 431-446.
Zwart, Jan Wouter
1993 Dutch Syntax: A Minimalist Approach. Groningen Dissertations in
Linguistics 10.
IE *weid- as a root with dual subcategorization
features in the Homeric poems

Annamaria Bartolotta

This paper is organized as follows: the first section sketches the theoretical
background involved in the case study of Old Greek eidon/oida. As is well
known, the aorist eidon takes only an accusative DP-object, while the perfect
oida can take either a genitive or an accusative DP-object. Sections 2-5 aim
to prove that the diachronic development of the root *weid- in early Greek
must be take into consideration to explain the synchronic phenomenon of
dual subcategorization features. This root proves indeed to be polysemous
and is split into two different meanings which are lexicalised by means of
different bridging contexts and different morphological developments. In
section 6 the peculiar evolution of oida from a genitive/accusative to an ac-
cusative syntax is considered from a typological point of view. The genitive/
accusative syntax is held to be a residual trace of a previous active-stative
system which had characterised the Proto-Indo-European before it adopted
a nominative-accusative system. Section 7 examines the available evidence
about the inherent status of genitive and accusative in Homeric Greek. In
section 8 the modular conception of grammar is discussed by considering
the role played by the morpho-semantic properties of the root when dealing
with a THETA-related case system. Sections 9 - 1 0 are concerned with a
minimalist representation of the syntactic structure of the early Greek verb.
Against the LATE INSERTION principle, a model endowed with three func-
tional heads (TP, AspP, vP) is put forward which requires the root features
to be visible in syntactic derivation before Spell-out. Thus, the syntactic
object case feature at that stage of the language is taken to have been inter-
pretable in LF (Object-in-situ hypothesis). Section 11 examines the Greek
sentence structure after the nominative-accusative type had been stabilized:
once case had turned to a purely formal feature, it became uninterpretable;
so, the semantic properties of the root were not able to determine the inher-
ent case assignment any longer, and tense tended to incorporate all tempo-
ral distinctions related to the verb (including aspect).
266 Annamaria Bartolotta

1. Greek *weidon / wöida: a case study

In the Homeric poems, the Greek perfect (w)oida Ί see with the m i n d ' s
eye, I k n o w , ' shows a peculiar behaviour as regards its morphosyntactic
structure. Although the aorist (w)eidon Ί saw', derived from the same root
*weid- 'to see', is characterized by an identical lexical entry, we can observe
a different case mapping between the two verb forms when they are bivalent
and select their internal argument (i.e. a nominal direct object). Regardless
of the same categorization features [ ] v , the same sub-categorization
features [+ N P ] Vp, the same selective restrictions as for the subject
N P [+an] , and the same theta-grid structure <Experiencer, Theme>,
the aorist is seen to regularly assign the accusative case, while the syntax of
the perfect alternates between an assignment of the accusative and genitive
cases in an apparently unpredictable way. 1 Compare, for example, the fol-
lowing passages quoted f r o m the Iliad, where oida sometimes assigns the
genitive case, as in Iliad X V , 4 1 1 - 1 2 , 2

(1) äs rhä te pases eu

who-NOM indeed-PRT+CONJ all-GEN.SING well-ADV
eide sophies hypothemosynesin Athenes
knows-PERF.SUBJ skill-GEN promptings-DAT Athene-GEN
' w h o (viz. a cunning workman) indeed well knows (sees in his mind)
the whole skill by the promptings of Athene'

and sometimes the accusative case, as in Iliad VII, 237:

(2) autär egon eu oida mtikhas t'

nay-ADV I-NOM well-ADV know-PERF battles-ACC and-CONJ
androktasias te
manslayings-ACC and-CONJ
'nay, I know (see in my mind) well battles and slayings of m e n . ' 3

At first glance, this outline may seem very unusual if we refer to the tradi-
tional generative thematic theory, according to which properties of verb
selection and valence are invariable for all the tenses, moods and persons of
the inflected verb. Furthermore, it should be noted that inherent case assign-
ment occurs simultaneously with internal thematic role assignment, which
is specified at the lexical entry of the verb itself (Haegeman 1996: 149).
 IE *weid- as a root with DS features in the Homeric poems 267

A complicating factor in this study is the apparent unpredictability of this

alternation in linking the accusative or the genitive case to the nominal ar-
gument selected by the perfect. There is an accusative in most of the occur-
rences with oida, while the genitive is less and less present in the passage
from the Iliad (23 χ) to the Odyssey (3 χ), until it disappears definitively in
the Greek Classical age where the finite forms of oida are shown to assign
only the accusative. This seems to occur without any apparent reason due to
contextual differentiations. Furthermore, the same dual subcategorization is
noticeable for many other Greek verbs as well (from akouo Ί h e a r ' to ρίπο
Ί drink', from mäiomai Ί seek after' to pynthänomai Ί learn', etc.), but it
does not seem possible to ascribe to oida the same account that has been
traditionally given for those verbs. Many scholars have indeed described
the genitive case selected by verbs of perception as a PARTITIVE GENITIVE
which alternates with the accusative. According to this hypothesis, the action
expressed by a verb of perception like oida could involve the direct object
totally (accusative) or only partially (partitive genitive). 4 If this explanation
could probably fit some other verbs of perception, the same cannot be said
for oida. In fact, as the analysis of the Homeric verses indicates, the alter-
nation between accusative and genitive does not seem to be connected with
the partitive meaning that is usually ascribed to the genitive, since such a
partitive meaning is never noticeable in the perfect oida occurrences. 5
I will argue that this phenomenon has to be accounted for by reapprais-
ing, on the one hand, the semantic values involved by the original Indo-
European root, and, on the other, the role of diachronic change also in a
synchronic framework.

2. The meaning-split of the IE root *weid-

The first step of the differentiation between *weidon and *woida has to be
found in the meaning split that the original IE root *weid-, from which both
verbs derive, undergoes. The split is motivated by the linguistic sign's
POLYSEMOUS nature: originally, the root *weid- 'to see' w a s a lexeme with
more than one related sense, denoting a kind of vision that could be both
perceptive and intentional. 6 As in most Indo-European languages, also the
Greek-derived stem *wid- / *weid- / *woid- bears more than one meaning,
with a semantic extension from 'see' to ' k n o w ' . 7 In the Homeric text, the
zero-grade stem of the aorist *wid- 'see' meant not only a kind of perception
which consisted of simple eyesight, but it also denoted the beginning of an
268 Annamaria Bartolotta

internal acquisition process after working-out data that had been seen, to-
wards a brief form of thought. This is what the analysis of eidon occurrences
in the Homeric poems tells us clearly (see below § 4). Soon, probably al-
ready at a Proto-Indo-European stage, as the comparison with the Vedic
developments of the same root lets us reconstruct (Bartolotta 2002a: 17-20),
this semantic complexity found a lexical counterpart in a new stem-formation
which entailed a vowel change (ablaut). Therefore, it is possible to distin-
guish in Greek the results of such semantic splitting:

(i) the PERCEPTIVE side of sight may be attributed to the zero-grade stem
*wid-\ the aorist eidon denotes a punctual action characterized by an
Aspect feature [-Stative]. This verb assigns the accusative case to its
internal argument, a case which semantically reflected an IMMEDIATE
contact with the object of sight;
(ii) the INTENTIONAL side of perception may be instead attributed to the
o-grade stem *woid-: the perfect oida denotes a process characterized
by an Aspect feature [+stative] (see below § 4). This verb originally
assigns the genitive case to its internal argument, a case which se-
mantically reflected a MEDIATED contact with the object of sight. 8

In the following sections I shall try to examine the way this semantic split
will condition the morphosyntactic structure of the verb.

3. Distributional contexts

In order to understand the gradual steps of the semantic change which is

hidden behind the morphosyntactic structure synchronically sketched, let us
first examine the role of distributional contexts in which the two verb forms
occur in the Homeric poems. It appears that most of the aorist and perfect
forms follow a regular - if not straight formulaic - contextual distribution
including a PP with a substantive that refers to the physical organ involved
during the action/state described by the verb. The recurring syntactic links
are as follows (table l ) : 9
 IE *weid- as a root with DS features in the Homeric poems 269

Table 1. The physical organ involved in the action/state described by the aorist
eidon and the perfect oida.

aorist eidon Iliad Odyssey

+ ophthalmoisin eyes-DAT.PL 'in (with) the eyes' 21 χ 17 χ

+ en ophthalmoisin in-PREP eyes-DAT.PL 'in the eyes' 3 χ 1 χ
+ opsei sight-DAT.SG 'with the sight' 1 χ -

perfect öida Iliad Odyssey

+ thymö mind-DAT.SG 'in (with) the mind' 2 χ 3 χ

+ eni thymö in-PREP mind-DAT.SG 'in the mind' - 1 χ
+ katä thymön in-PREP mind-ACC.SG 'in the mind' 3 χ 1 x
+ thymos mind-NOM.SG 'the mind' I χ -

+ phresin mind-DAT.PL 'in (with) the mind' 3 χ 4 χ

+ eniphresi(n) in-PREP mind-DAT.PL 'in the mind' 3 x 4 χ
+ katäphrena in-PREP mind-ACC.SG 'in the mind' 3 χ 1 χ
+ peri phresin in-PREP mind-DAT.PL 'in the mind' - 1 χ
+ prapidessi mind-DAT.PL 'in the mind' 4 χ 1 χ

This distribution confirms the semantic hypothesis outlined above: on the

one hand, the aorist denotes a kind of perception that merely consists of an
immediate eyesight (dative ophthalmoisin 'eyes' - or equivalent locative
periphrastic expressions); on the other, the perfect denotes a perception that
goes beyond the physical sighting, since it being mediated by a process of
elaboration which takes place in the mind (dative thymö ' m i n d ' - or equi-
valent locative periphrastic expressions).

4. Gradualness of linguistic change: the role of polysemy

Now, focusing on the context allows us to distinguish two crucial steps of the
semantic change, (1) the so-called STAGE OF OVERLAP and (2) the STAGE
OF LEXICALIZATION. According to Evans and Wilkins (2000: 549), in the
first stage, the semantic change from a meaning A to a meaning Β involves a
transitional stage of POLYSEMY where only one form takes both the meanings
(stage of overlap). 1 0 At this stage, the new meaning is not yet lexicalised.
On the basis of these claims, we would have the following evolution path:
270 Annamaria Bartolotta

1. Stage of overlap
(a) initially, one polysemous form has in nuce both the overlapping
meanings A and B: the aorist eidon denotes the two senses inherited
from the IE root *weid-, i.e. 'to see with the eyes' (MEANING A) and
the extended form 'to see with the mind' (MEANING B);
(b) the primitive meaning A is enriched by a regular contextual support
(bridging context) which enables the speakers to distinguish clearly a
new meaning B: some Homeric passages testify the presence of such
BRIDGING CONTEXTS which support the semantic extension of eidon
(see below).

2. Stage of lexicalization
The contextual sense (MEANING B) becomes then lexicalised after
creating a different morphological status of the root. This new status
is here created by deriving the Aktionsart category of Perfect through
the change of the root's vowel (*woid-). *woida is in fact character-
ized by a strong paradigmatic autonomy and an early morphological
lexicalization. Evidence for this early lexicalization comes from the
absence of reduplication (which normally marks the perfect), the
presence of proper personal endings, the presence of the vowel
change (apophony) and the presence of a new categorization form
(Aktionsart), which denotes the so-called 'resulting state', i.e. a state
resulting from a process. 11

Evidence for the first stage comes from Homeric passages that seem to be
an exception to the regular distribution discussed in §3 (table 1). As stated
above, the dative ophthalmoisi 'eyes' occurs in fact with the aoristic forms
since they properly refer to the original sense of 'seeing' related to eye-
sight; whereas datives such as thymö, phresin, etc. 'mind', are linked to the
perfect, since this latter relates to a cognitive process which takes place
only in the mind. Now, the BRIDGING CONTEXT is instead represented by
those passages in which the aorist forms occur with the syntactical expres-
sions usually proper to the perfect. Let us analyse the text. In Iliad XXI, 61,
as shown in (3),
 IE *weid- as a root with DS features in the Homeric poems 271

(3) ophra idömai eni phresin

(in order) that-CONJ see-AOR.SUBJ.lSG in-PREP mind-DAT
ede daeiö
and-CONJ know-AOR.SUBJ.lSG
'that I see in my mind and k n o w '

the locative expression eni phresin 'in the mind', which occurs with the
aorist subjunctive idömai Ί see', represents a bridge between the first
meaning 'to see with the eyes', normally conveyed by eidon from *wid-,
and the extended sense 'to see with the mind', that involves a cognitive
inferential process triggered by ocular sight. A confirmation to what has
been claimed comes from the presence, in the context, of the subjunctive
daeiö (derived from the root Vda 'to know') which is closely related to the
subjunctive idömai Ί see': in this way a form of knowledge which com-
pletes the idea of a whole cognitive process is expressed. The other exam-
ple is in Odyssey VIII, 450, as shown in (4),

(4) hod' [...] ide thy mo

he-NOM+PRT [...] saw-AOR.IND. 1SG mind-DAT
'he (viz. Odysseus) saw in his mind.'

where the irregular coexistence of the dative thymö 'mind' with the aorist
ide 'he saw' is again the residual trace of a semantic change in progress.
Further supporting evidence for bridging contexts which function as
catalysing support for linguistic change is to be found in Homeric passages
where the aorist eidon is immediately followed by cognitive verbs such as
noeö Ί perceive', gignoskö Ί come to know', phräzomai Ί think, I per-
ceive', hormainö Ί revolve in the mind', etc. Here it is again possible to
distinguish two steps of the same process which begins with (i) the visual
stimulus and ends with (2) an internalization of knowledge (see Bartolotta
2002b: 25).

5. The splitting traces

To summarize, the Homeric poems testify to a gradual splitting of the root

*weid-, both semantically and morphosyntactically:

- different distributional contexts that indicate the different semantic

value of eidon and oida (see table 1);
272 Annamaria Bartolotta

- morphosyntactic and aspectual differentiation, evidenced by alternating

case assignment: aorist [-stative] + accusative vs. perfect [+stative] +
genitive;
- presence of bridging contexts (aorist eidon + distributional context proper
to the perfect) which support the extended sense of the root (*wid-) (ex-
amples (3) and (4)).

6. The morphologic-syntactic-semantic interface:

a typological perspective

After delineating this situation, let us return to the peculiar morphosyntactic

behaviour of the perfect oida. The variation from a genitive to an accusative
syntax - which characterizes many other knowledge verbs of the Homeric
poems, like pynthdnomai Ί learn', didäskö Ί teach', 4da Ί learn, I k n o w ' ,
etc. as well as oida - is noticeable in the transitional period between the
Iliad (genitive 23 x) and the Odyssey (genitive 3 χ ) and, finally, to the Greek
classic age, where the accusative case definitively prevails over the genitive
case. Further investigations allow us to hypothesize that this morphosyntac-
tic variation is part of a broader diachronic change that involves the whole
linguistic system, a typological change of which this variation is an impor-
tant piece of evidence. As is well-known, many Indo-European scholars
pointed out that it is possible to reconstruct, for the late stage of Indo-
European, a change from an active-stative to a nominative-accusative system
(see, a m o n g others, Lehmann 1999: 318). According to this assumption
based on comparative studies, the Indo-European language undergoes a
complex systemic change involving morphological, syntactical, semantic
and lexical structures. To quote just the more relevant here, it is interesting
to note that, after this change, the relationship between verb and noun (or
between noun and noun) turns out to be based on a structural (i.e. a mere
inflectional) regimen rather than on one of semantic concord, with the fol-
lowing results: 1 2

(i) the semantic-lexical relationship between verb and noun (proper to an

active-stative type), according to which the noun meaning was aimed
to complete the meaning of the verb depending on its [± stative] fea-
ture, becomes just a verb/noun structural relationship in the nomina-
tive-accusative type;
 IE *weid- as a root with DS features in the Homeric poems 273

(ii) the semantic value of the so-called INHERENT case, which is assigned
by a lexical head V to the N P it governs, in conjunction with θ-role
assignment, now makes way for a new purely syntactic evaluation of
the STRUCTURAL case, which is not assigned in conjunction with Θ-
role assignment by a lexical head. 13

Furthermore, as is well known in Indo-European studies, this change means

that the verbal system has been shifted from an aspectual model, which is
characteristically proper to an active-stative type, to a temporal model,
which is a subsequent acquisition of the nominative-accusative type, where
Tense distinctions begin to play an increasingly important role in respect of
Aspect distinctions. On the basis of these assumptions, I will argue that the
oida genitive-marked construction is a residual trace of an early active-
stative stage of the IE language - a stage which early Greek inherited -
since the genitive was the inherent case which best adapted itself semanti-
cally in order to complete the verb aspectual meaning. 14 In other words, a
genitive of INFERENCE would semantically reflect the mediated relationship
which links the stative verb oida to its object, since at the early stage there
still was no trace of transitivity in the PIE verbal system. 15 The oida accusa-
tive-marked construction, which originally had its own semantic value ex-
pressing an immediate relationship verb/noun, would have instead gradually
reflected a transitive structural relationship between the verb and its object,
having just an abstract syntactic-position value (i.e. the accusative case marks
the noun which follows the verb immediately). In this way, the accusative-
marked construction, which now falls within a nominative-accusative sys-
tem, makes regular the verb-noun relationship in a temporal model where
the aspectual differences, which were once essential in the option for a se-
mantically related case, began to be less and less important. Such a devel-
opment would explain why an accusative case had to replace the genitive
case in the syntax of oida during the shifting to the Greek classical age. To
recap the most important features involved in the linguistic change just
sketched, here is a summary table (table 2):
274 Annamaria Bartolotta

Table 2. Indo-European typological variation: some important morphological-syn-

tactical-semantic features involved in oida linguistic change.

active-stative system nominative-accusative system

semantic-lexical relationship verb/noun —» syntactic-structural relationship
verb/noun
semantic value of the inherent case —> syntactic value of the structural case
aspectual system —» temporal system
oida + genitive (θ-related case) —> oida + accusative (no θ-related case)

7. The syntax of oida between genitive and accusative

From a synchronic point of view, one of the most striking pieces of evi-
dence of the semantic value related to case-assignment comes from those
passages where oida assigns different cases practically to the same DP-
object. This case assignment is not random, for, as I shall argue, it is
closely connected to the semantic relationship between verb and noun.
Compare, for example, Iliad XII, 100 (= V, 11; V, 549)

(5) Arkhelokhos t' Akamas te, mdkhes

Archelochus-NOM and-CONJ Acamas-NOM and-CONJ battle-GEN.SG
eu eidote pdses
well-ADV know-PERF.PART.DL whole-GEN
'both Archelocus and Acamas, who well know (see in their mind) the
whole battle'.

to Iliad VII, 237, here repeated as (6):

(6) autär egon eu oida miikhas t'

nay-ADV I-NOM well-ADV know-PERF battles-ACC.PL and-CONJ
androktasias te
slayings.of.men-ACC.PL and-CONJ
'nay, I know (see in my mind) well battles and slayings of men'.

In (5) the genitive singular adjective pdses, which refers to mdkhes, means
'all' taken as a whole, and not 'each' as some classical translations give. 16
It refers to a knowledge as a result of a whole complex of military opera-
 IE *weid- as a root with DS features in the Homeric poems 275

tions of which a battle consists. The same explanation is indeed valid for
the genitive pases sophies in (1), where pases refers to a kind of knowledge
which is the result of a complex of techniques taken as a whole, the fruit of
an internalization and re-elaboration of data from which a sort of science or
art is inferred (mediated relationship between verb and its DP-object). More-
over, it should be noted that, morphologically speaking, the genitive is not
due to the fact that oida is inflected as a nominal form, i.e. the participle
eidote in (5), because the same construction is to be found in (1), where eide
is a subjunctive.
In (6), instead, the plural accusative mäkhas refers to a series of military
operations considered not as a whole, but as single concrete events, directly
experienced without referring to a general abstract form of knowledge
(non-mediated relationship between verb and its DP-object). The plural
forms are often preferred in similar constructions, as we can see in Iliad XI,
719, where Nestor is too young and has not learnt yet the art of war deriv-
ing from the internalization of the concrete experience of a series of mili-
tary actions:

(7) ou gär ρό ti m' iphe

not-NEG in fact-CONJ yet-ADV I-PRON.ACC said-AOR.3SG
idmen polemeia erga
to know-PERF.INF warlike-ACC.PL deeds-ACC.PL
'he (Neleus) said in fact that I (Nestor) did not know (see in my mind)
as yet deeds of w a r ' .

These considerations should also explain every alternating construction of

oida, which assigns either genitive or accusative. Thus, on the one hand, we
have the genitive case in expressions such as the following (8):

(8) a. eidote theres

knowing-PERF.PART.DL hunt-GEN.SG
' k n o w i n g (seeing in the mind) the art of hunting' (II. X, 360)

b. eidos pugmakhies
knowing-PERF.PART.DL boxing-GEN.SG
' k n o w i n g (seeing in the mind) the art of boxing' (II. XXIII, 665)

c. aikhmes eu eidos
spear-GEN.SG well-ADV knowing-PERF.PART.SG
'knowing (seeing in the mind) well the art of the spear' (II. XV, 525)
276 Annamaria Bartolotta

d. eidote khärmes
knowing-PERF.PART.DL fight-GEN.SG
'knowing (seeing in the mind) the art of the fight' (II. V, 608)

e. oiönön sdpha eidos

birds-GEN.PL clearly-ADV knowing-PERF.PART.SG
'knowing (seeing in the mind) clearly the (divinatory) art of the birds'
(Od. I, 202)
On the other hand, the same inflected forms of oida assign the accusative
case, as in (9):

(9) a. apatilia eidos

deceiving-ADJ.NEU.ACC.PL knowing-PERF.PART.SG
'knowing (seeing in the mind) deceiving things' (Od. XIV, 288)

b. eidos pantoious te
knowing-PERF.PART.SG each.kind.of-ADJ.ACC.PL and-CONJ
dölous käi medea pyktui
crafts-ACC.PL and-CONJ devices-ACC.PL cunning-ACC
'(viz. Odysseus) knowing (seeing in the mind) each kind of crafts
and cunning devices' (II. Ill, 202)

c. thymos eni stethessi epia

mind-NOM in-PREP breast-DAT kindly-ADJ.ACC
denea oide
thoughts-ACC.NEU.PL knows-PERF.IND
'(your) mind in the breast knows (sees) kindly thoughts' 1 7
(II. IV, 360)
d. olophoia eidos
pernicious-ADJ.NEU.ACC.PL knowing-PERF.PART.SG
'knowing (seeing in the mind) pernicious things' (Od. XVII, 248)

As is definitively shown by all the occurrences of oida in Homeric poems, 18

the inherent status of genitive can be namely defined as GENITIVE OF
INFERENCE. Therefore, both Jannaris' (1897) hypothesis on the Greek per-
ception verbs (genitive = direct perception vs. accusative = indirect percep-
tion), and the partitive genitive hypothesis (see § 1) are to be discarded for
the morphosyntax of oida}9 Furthermore, since there is no semantic neu-
tralization of the two cases at the synchronic level, the idea that the genitive
case is an alternative spelling-out of the abstract objective case feature
 IE *weid- as a root with DS features in the Homeric poems 277

should be rejected. To recapitulate, in the early stage of Greek, case assign-

ment was a matter of inherent meaning based upon the semantic relation-
ship between the verb and its DP-object.

8. The modularity of the grammar

After this analysis of the status of the features involved in the oida perfect,
it is possible to clarify a synchronic account of the early Greek verb struc-
ture. The presupposition behind this is that the peculiar situation of eidon
and oida, which are tenses belonging to the same verb paradigm, is ex-
plainable only by considering the idiosyncratic lexical and morphological
properties of the root-stem. These properties play a role in syntax since
only they can account for certain future developments responsible for any
morphosyntactic structure of the verb. The morphological and lexical fea-
tures of the root have to be visible in the syntactic derivation before and not
after Spell-Out: only in this way will it be possible to give a predictive syn-
chronic explanation of all the linguistic phenomena related to the verb. That
is why in this treatment I follow Embick (2000: 187-188) when he argues
against the Distributed Morphology principles such as those of the LATE
INSERTION a n d t h e FEATURE DISJOINTNESS e x p r e s s e d in M a r a n t z ( 1 9 9 4 ,
1995). According to these principles, which defend a modular concept of
grammar, the morphological features and the arbitrary properties of vo-
cabulary items are irrelevant for syntax. They must not be present in the
syntactic derivation, with the result that the lexical Roots and the functional
morphemes have to be inserted after Spell-Out. The *wiidon / *woida case
study seems instead to highlight the role played in structuring syntactic
configurations by the idiosyncratic semantic properties of the verbal Root
with its morphological features, such as Aspect and Aktionsart (carried by
apophony). 20 This claim is valid for the inherent case-selection and, as we
shall see in the next section, for the related feature-checking process.21
Moreover, it is worth noting that dead languages such as Sanskrit, for
example, allow us to reconstruct a Proto-Indo-European language whose
lexicon was essentially based on a root-items system.

9. Syntactic structure of the Early Greek verb

In order to describe the structure of the Homeric Greek syntax, I shall assume
a clause structure endowed with three functional heads, Tense, Aspect and
278 Annamaria Bartolotta

Light verb, within the framework of the Agr-less checking theory of the
Minimalist model. Following some suggestions coming from Embick's
analysis of Latin verb, the notation V in (10) is for the Root, «that is, the
member of the open-class vocabulary appearing in this position» (Embick
2000: 195). Moreover, according to Stroik (2001: 368), it should be noted
that 'the light verb v, which bears the [Vform] feature, is a functional cate-
gory that has both morphophonetic and semantic justification'. The [Vform]
contains all the morphological features of the root, such as Aktionsart, etc.,
which are responsible for the case selection of the verb. 22 These morpho-
logical features, that have to be visible in the syntax by virtue of the EARLY
INSERTION of the root (i.e. from the outset of the derivation), will be
checked in the head v. The structure will be then represented as follows:

(10) TP

Τ AspP

V DP

The TP bears the temporal feature, but it has been hypothesized that an
Agr(eement) node can be added in the morphological component to form a
T-Agr position (see Chomsky 1995). Here it is assumed that the verb under-
goes the checking of φ-features such as person, number, gender, including
thematicity feature [±th] (Embick 2000: 197), i.e. a morphological feature
which reveals the presence of a theme vowel as in weid-o-n. However, it
should be noticed that tense inflection and agreement inflection seem to be
two quite different systems, as pointed out, to quote just a few, by Friedmann
and Grodzinsky (1997, 2000) in their neurophysiological analysis, or by
Guasti and Rizzi (2000: 1). It could be therefore assumed that agreement is
checked in a mechanism different from that for tense. So, since agreement
may not have a node of itself, it may check in one of the other checking
points below Τ (Friedmann and Grodzinsky 2000: 99). The AspP projects
the perfective or imperfective aspectual feature, which is related to the verb
and then to v. It deals with a basic aspectual opposition, which plays a fun-
damental role in the syntax of early Greek, as already pointed out. Then,
the vP head contains all the morpholexical properties ([Vform]) which the
 IE *weid- as a root with DS features in the Homeric poems 279

verb is provided with. Among them, the most important is the feature related
to the Aktionsart category, including notions such as stativity, causativity,
agentivity, punctuality, and so on. As for eidon and oida, the [Vform] fea-
tures relate respectively to a punctual action (aorist) and a resulting state
(perfect). 23 As regards the Λ/Ρ, it should be noted once again how important
it is that the verbal Root is present in the syntax from the outset of the deri-
vation. In fact, the case difference between accusative or genitive overtly
manifested by the morphology depends on the lexical values of the Root: 24
both cases had originally a semantic (inherent) value which was available
in order to make the verb semantically complete. Moreover, the presence of
the Root playing a role in Homeric Greek also finds a diachronic justifica-
tion, as results from the well known fact that in ancient languages such as
Vedic and even Proto-Indo-European (from which Homeric Greek directly
derived) the lexical items were memorized as roots. Finally, as for the verb's
arguments, I follow Lopez (2001: 714) in distinguishing the different proper-
ties of subject and object feature checking. In other words, while for the ex-
ternal argument, i.e. the subject, it has been hypothesized that, after origi-
nating from the VP shell (the Specv position), it checks its formal features
with T, a different treatment seems to characterize object feature checking.
This is what I shall try to argue in the next section.

10. The Object-in-situ hypothesis: Case assignment in an active-stative

language

On the basis of the analysis of many modern Indo-European languages,

such as English, which does not maintain a morphological case system, the
Minimalist Case theory (Chomsky 1995) is mainly concerned with abstract
case. Thence, when it does not involve a covert operation, case assignment
is considered as a pure syntactic specifier-head relationship between a noun
phrase and a functional head (see, among others, Giorgi and Pianesi 1997:
10). However, this assumption cannot be valid for languages whose typology
is active-stative or in a transitional stage from an active-stative to a nomina-
tive-accusative type, as is seen in Homeric Greek. A Spec-Head relation-
ship could be more successfully maintained for nominative, which is the
case related to the structural position of the external argument subject, but
the same cannot be valid for an inherent semantic case related to an internal
argument whose principal aim was to complete the meaning of the verb. In
other words, it could be suggested that in a general theory the inherent-case
280 Annamaria Bartolotta

assignment should be more emphasized. The oida case study has shown
that alternating accusative or genitive is not a mere structural fact, but a
matter of morphological-syntactic-semantic interface. That is to say, a verb
with dual subcategorization features, as is seen to occur with oida in the
early stage of Greek, points out that case assignment of internal arguments
depends on the morphosyntactic properties of the lexical verb. In structural
terms, object can also check its features in situ, against the same verbal head
that assigns it a θ-role (Lopez 2001: 699) without moving anywhere. 2 5 This
is indeed the result of the fact that accusative case is here a feature which is
interpretable at the semantic component LF. Only this hypothesis, which is
consistent with some recent suggestions about the role of semantically un-
interpretable morphological features in dislocation property (Chomsky
2000: 26), allows us to highlight the typological difference between a
nominative-accusative type, where the accusative case has only a structural
configurational value (non θ-related), and an active-stative type, where the
case still involves a semantic value which plays an important role in identi-
fying the verb's meaning. 2 6 This solution also maintains strict locality of Θ-
marking: an Object-in-situ hypothesis points out indeed the close semantic
relationship between the verb and its internal argument through maintaining
the highest level of adjacency between the two constituents. Crucially, de-
spite some widely accepted assumptions of Chomsky's theory (see Adger
et al. 1999: 5), in certain languages - for example Homeric Greek - case
becomes an interpretable feature which does not need movement for check-
ing since it survives in the mapping to the semantic component and is not ' a
formal feature that plays a role only in the syntactic derivation itself (Hale
1998: l l ) . 2 7 Under these assumptions, it could therefore be hypothesized
that a language which is typologically classified as active-stative is a [+0PC]
language. According to this THETA POSITION CHECKING Parameter, elements
can undergo feature checking at their θ-position (Ura 2000: 38), that is the
same position where they receive a θ-role. It could be hypothesized that
object enters into a checking relation with ν without moving anywhere, that
is, following Ura (2000: 220), the morphosyntactically tight relation be-
tween a verb and its object is connected to an obligatory checking relation
between them. Accordingly, it has been proposed that 'the [±0PC]-ness
lies in v's lexical properties' and that 'v's property concerning formal fea-
tures checking depends upon the aspectual-head that selects ν as its com-
plement' (Ura 2000: 219). 28 As a consequence, in the proposal made here,
the role of such a Parameter turns out to be fundamental in the analysis of
Homeric Greek, which still shows traces of an active-stative type. Hence, I
 IE *weid- as a root with DS features in the Homeric poems 281

argue against M a r a n t z ' s proposal that features such as accusative do not

belong to the computational system, which sees only the formal feature
[+Case], so that spelling out this case feature as accusative is just a matter
of PF (Alexiadou and Anagnostopoulou 2001: 21 1).29 If this statement could
in fact refer to an abstract case accusative, the same is not valid for an in-
herent case accusative. In other words, such a generalization does not take
minimally into account the inherent-case system, which, on the contrary,
should be more emphasized in a M P general theory. Indeed, this explana-
tion does not fit the analysis of a language like early Greek and other an-
cient dead languages w h o s e verbal system shows that it maintains traces of
an ancient active-stative type. In structural terms, this m e a n s that the Greek
sentence structure, during a gradual transitional stage w h e r e its typology
w a s not completely nominative-accusative but it was still maintaining some
active-stative type peculiarities, probably had the following characteristics:

(i) the verbal system turns out to be provided with both AspP and TP
since the Indo-European active-stative type, which is directly inher-
ited in Greek, shows that the aspectual distinctions play a role more
important than the tense distinctions;
(ii) the vP contains all the morphosemantic features, such as [±stative],
proper to the verbal Root (i.e. Aktionsart), which are f u n d a m e n t a l for
case assignment (the option for genitive or accusative); this is the rea-
son w h y we have to hypothesize an Early Insertion of the Root, which
is responsible for the lexical entry of the verb and its dual subcatego-
rization features;
(iii) the verb has to be represented as a Root, since are the semantic prop-
erties of this latter which are relevant in assigning the inherent case:
w h e n the Root is *wid- [-stative] the aorist eidon assigns the accusa-
tive; w h e n the Root is *woid- [+stative] the perfect oida is able to as-
sign also the genitive;
(iv) the internal argument (object) checks its formal feature of inherent
case (genitive or accusative) in situ (without involving a movement),
i.e. where it is base-generated and receives its θ-role f r o m the related
lexical head ( * w i d - / *woid-). It is in fact semantically interpreted at
the LF. This solution fits the hypothesis that the N o u n and the Root
are listed together in a derived lexical entry;
(v) the external argument (subject) checks regularly its formal feature
(nominative case) against the functional head T.
282 Annamaria Bartolotta

On the basis of these considerations, the Homeric Greek verb structure re-
lated to eidon and oida results as follows: 3 0

(11) TP

DP, Τ

q-JJ
AspP
[aor] / [perf]
[φ-features] Asp 1
(V+v)i
Asp 0 vP
[±perfective]
t (V + v)j t. v

,0
VP
[Aktionsart]
t(^+v)i t (V Root), DP

According to T a n a k a ' s hypothesis (1999), it could be also proposed that

AspP is inside the lexical domain of vP, sandwiched between vP and VP.
As the anonymous reviewer has pointed out, this hypothesis would suit
with the fact that semantic case is related with the aspectual and Aktionsart
properties of the lexical entry and respects adjacency. Furthermore, if ν is
the locus of Aktionsart, then incorporation of Asp to ν can derive the rela-
tion between Aktionsart and aspect in the language. 3 1

11. Greek sentence structure after change

After Greek gradually developed into a nominative-accusative language,

Verbs and N o u n s of its Lexicon need not be listed together in a derived
lexical entry. In accordance with the nominative-accusative type, there is in
fact a mere structural relation of transitivity between a verb and its internal
argument, without involving any semantic concord. Thus, being an abstract
case which is not associated inherently with the verbal Root, the accusative
is not semantically interpretable at LF. Consequently, the feature checking
requires the intermediacy of a Specifier-Head relation with a functional
category. That is why the inherent genitive case with oida is increasingly
less c o m m o n in the Greek classical age, until it disappears and is totally
replaced by the accusative, the structural position case par excellence.
 IE *weid- as a root with DS features in the Homeric poems 283

Therefore, once the nominative-accusative system has stabilized, the sen-

tence of a structure having a form of the paradigm of oida as a predicate
shows that:

(i) the internal argument (object) has to move in order to check its [-inter-
pretable] feature of accusative structural case, since it does not receive
a case in situ from the lexical head V (oida + accusative in Classical
Greek);
(ii) it becomes less and less important that the semantic properties of the
Root have to be visible in a pre-syntactic level, since they do not deter-
mine an inherent case assignment (oida does not show case alternation
between accusative and genitive in Classical Greek);
(iii) both AspP and vP, which bear the morphosyntactic characteristics of
the verb, play now a marginal role in respect of TP, since Tense tends
to incorporate every temporal distinctions related to the verb.

12. Conclusion

The implications of this analysis seem to be straightforward. Through an

examination of the *weidon / *woida case in the Homeric poems, it can be
seen that the lexical properties of the Root in a presyntactic lexicon deter-
mine the syntactic behaviour. Some recent results in the field of the brain
sciences are unexpectedly found to support in the view we have recon-
structed. As pointed out by Penke (2003: 50), problems with inflectional
morphology have consequences for syntactic representations, that is, the
morphosyntactic features of the selected lexical items determine the build-
ing up of syntactic representations (Penke 2003: 51). The background as-
sumption here is that in languages with an active-stative typology, case
plays a role in syntax. In fact, having an inherent semantic meaning that
indicates the close relationship between a verb and its internal argument,
case can be a feature which is interpretable at the LF. In other words, the
direct θ-role of the verb is associated in the lexicon with a direct argument
that receives a precise lexical case.32 This close semantic relationship be-
tween verb and noun finds a syntactic counterpart in the fact that the direct
internal argument NP checks its object-case feature within the lexical verb,
without involving a movement. According to Pesetsky (1996: 223), 'the
availability and syntactic positioning of arguments is not a matter of
chance, but arises from laws governing the structure of lexical entries, from
284 Annamaria Bartolotta

laws regulating the projection of these entries onto syntactic structure'.

Moreover, as pointed out again in Pesetsky (1996: 291):

(12) The meaning of an item determines most (if not all) of its syntactic
properties.

This perspective attributes a fundamental role to the semantic/syntactic inter-

face, to the extent that the centrality of the morphosemantic properties of the
Root-item turns out to be relevant in the syntactic derivation. This conclusion
is in line with recent research on the strategic role of the interaction be-
tween subcategorization information and semantics of the verb in order to
know how to use a verb in deriving a syntactic structure (see Hare, McRae
and Elman 2003: 281). 33 Furthermore, since changing from semantic to
syntactic values of the case drives movement operations in order to check
the uninterpretable formal features, it could be said that 'changing mor-
phology has certainly consequences for syntax' (Lightfoot 2002: 18).
It should also be noticed that the [±0PC] Parameter and the role of case
are both due to the typological structure of a language. Unlike an active-
stative language, where the [+0PC] Parameter is activated, in a nominative-
accusative language the licensing of the object-case checking feature (when
not inherent) normally requires the intermediacy of a Specifier-Head relation
with a functional category, since the case generally has a mere abstract con-
figurational positioning value.
The *weidon / *woida case study here outlined shows some characteris-
tics of the syntactic behaviour proper to Early Greek, but it is noteworthy
that the same phenomenon can be easily found in Vedic as well (Bartolotta
2002a: 114), which shows an identical morphosyntactic development, both
synchronically and diachronically, in the dual subcategorization features of
the root vid- (Speijer 1886 [1988]: 90). This could lead us to hypothesize
that IE languages underwent an important typological change, whose residual
trace is represented by such inherited phenomena as the dual subcategoriza-
tion features of the verb oida in Homeric poems. Old Greek at its earliest
stage maintains traces of a gradual typological change involved by Indo-
European languages. The synchronic discrepancies can be accounted for by
diachrony: the dual subcategorization features which characterized part of
the morphosyntactic properties of the early Greek verbal system would
testify for a change from a previous active-stative type to a more recent
nominative-accusative type. After an initial overlapping period of the two
types, the latter would have been definitively stabilized during the classical
 IE *weid- as a root with DS features in the Homeric poems 285

age. Hence, unlike theories of linguistic change which suggest that mor-
phosyntactic change is abrupt and total, because it is the result of the reset-
ting of certain parameters (Timberlake 1977, Lightfoot 1991, 1998, Harris
and Campbell 1995), the line of research proposed here is more consistent
with some recent approaches, according to which parameter settings do not
change abruptly. On the contrary, change occurs as a result of competition
between alternative parameter settings during periods of syntactic variation
(Pintzuk 2003). Typology and semantics in a diachronic perspective have
had to be taken into account to put forward an explanation of peculiar case
assignments by different forms of one verbal paradigm in a synchronic
state of language. 34 My working hypothesis has involved diachrony in the
interpretation of this phenomenon. As for the implications and the motiva-
tions of the linguistic change that I have supposed, further investigation is
required.

Notes

1. From the analysis of all the occurrences of oida in the Homeric poems, the use
of genitive or accusative case due to the definite or indefinite nature of the ob-
ject NP involved does not seem likely (cf. Belletti 1988: 31). For the analysis of
the oida occurrences and such related matters, see Bartolotta (2002a: 57-70).
2. The following abbreviations are used in the paper: AN = animate; AOR = aorist;
PERJF = perfect; IND = indicative; SUBJ = subjunctive; PART = participle; INF =

infinitive; CONJ = conjunction; GEN = genitive; ACC = accusative; NOM =

nominative; DAT = dative; SG = singular; PL = plural; DL = dual; NEU = neuter;
A D V = adverb; PREP = preposition; PRON = pronoun; ADJ = adjective; NEG =
negation; PRT = particle.
3. It is worth noting that in the Homeric poems the perfect oida occurs very often
with adverbs such as eu 'well', like in these first examples (with the same role
of sapha 'clearly' in example 8e). The presence of such adverbs means that the
vision expressed by the lexical root is a particular one, i.e. it deals with a vision
deeper than a mere perception (see § 2 in the text). Indeed, such adverbs never
appear in the poems with the aorist eidon.
4. See, to quote just a few, Speijer (1886 [1988]: 90), Brugmann-Delbriick (1893:
308), Schwyzer (1953: 105), Chantraine (1963: 55), Kurilowicz (1964: 186),
Lasso de La Vega (1968: 418-426), Luraghi (1996: 53), etc.
5. Further analysis of the oida occurrences in Homeric poems is discussed in
Bartolotta 2002a: chapter 8.
286 Annamaria Bartolotta

6. For more discussion about this definition initially introduced by Vendryes, see
De Boel (1987: 20).
7. Compare Old Indie (yeda), Avestan (vaedä), Gothic (wait), Old Irish (ro.fetar),
Armenian (gitem), Old Slavic (vede), etc., which all reveal a deep connection
between the meaning of 'knowing' and the meaning of 'seeing', from which
the former originated. For a detailed description of the semantic extension
among different domains see Sweetser (1998).
8. In a previous work (Bartolotta 2002a: 116 ff.) I have hypothesized the existence
of a GENITIVE OF INFERENCE. See also below, note n. 15.
9. It can be easily assumed that the three nouns thymos, phren and prapides refer
to the human mind, whose seat is respectively placed in the heart, in the dia-
phragm and in the midriff.
10. It has been recently pointed out that 'the great majority of cases where different
meanings of a verb are associated with different subcategorization frames in-
volve polysemy. That is, these verbs exhibit highly related meanings, often with
a more concrete physical sense and extensions to more abstract and metaphorical
uses' (Hare, McRae and Elman 2003: 283).
11. See Di Giovine (1996, 2: 127-131) for a detailed discussion on each piece of
evidence regarding this early morphological lexicalization of the perfect oida.
12. In an active-stative type, as has been supposed for the Proto-Indo-European
language, to assign an argument depends on the semantic properties of the
lexical head, that is, the choice of a Noun and its related case strictly depends
on the [± stative] feature of its head.
13. The inherent (theta-related)-structural opposition of generative grammar corre-
sponds to the concrete-abstract opposition of historical linguistics (see Haudry
1977).
14. The genitive case was traditionally defined in historical linguistics as the case
of 'general determination' (Adrados 1992: 152), since it is shown to hold a
wide range of different meanings in respect of the other cases. For the original
functions of the genitive case, see Morani (1992: 221).
15. This terminology is taken from a previous work (Bartolotta 2002a: 113). The
INFERENCE refers to the peculiar cognitive process involved with oida where
knowledge is not immediate but inferential, that is, the result of a complex ac-
tivity of the mind including more than one step, from the visual perception to
the internal acquisition (memorization) and finally to the elaboration of the
data. The option for the genitive case would reflect the fact that, as underlined
also in Belletti (1988: 3), 'the most typical instance of an inherent Case is one
where a particular Case is associated with a particular θ-role'.
16. See, among others, Murray (1924): 'well skilled in all manner of fighting',
Mazon (1957-61): 'experts ä tous les combats', Calzecchi Onesti (1950): 'es-
perti d'ogni battaglia'. Note also that the adjective pases seems to be dislo-
cated on the right peryphery in order to emphasize by means of topicalization
the comprehensive meaning of 'all' as a whole.
 IE *weid- as a root with DS features in the Homeric poems 287

17. It has to be noticed that in the Homeric poems, even when the accusative refers
to an abstract object, e.g. 'thoughts' in (9c), every mental experience is experi-
enced in a very concrete way, as results from a wide literature on this topic.
See Bartolotta (2002a: 123) and references cited there.
18. Further details and occurrences in Bartolotta (2002a: 116-124).
19. Both hypotheses are to be discarded also for other verbs of perception like
mnäomai (see Bartolotta 2003: 54).
20. On the role of Aktionsart in the verbal system see, among others, Van Valin and
La Polla (1999: 92): '[...] verbs have a basic Aktionsart type, which is how they
are represented in the lexicon', and Gennari (2003: 35), who underlines that
we need not only the tense morphemes, but also the Aktionsart morphemes in
order to determine the exact temporal interpretation of a sentence.
21. Although the dissociation between the lexicon and the computational system
has been pointed out in order to corroborate the thesis of the modularity of the
Grammar (Anderson and Lightfoot 2000: 20), the close interaction between the
lexicon and morphological features on the one hand and the computational op-
erations on the other should not be underestimated, since the latter are often
triggered by the former during the syntactic derivation (Lorenzo and Longa
2003: 651). Also White (2003: 3) points out that 'properties of items that enter
ino a computation may vary in feature composition and feature strength, with
associated syntactic consequences'.
22. With regard to the distribution of more than one morphological feature related
to one Head 'little v', it should be said that this is a first attempt to include
those features in the syntactic configuration. Subsequent thorough examinations
of such relationships should lead us to a more homogeneous representation,
which perhaps will include a one-to-one relationship between a Head and its
related features.
23. It has been shown that 'little v' morphemes with semantic contents as agentive,
Stative, inchoative, causative, etc., which can form different types of verbs from
the same root, are responsible for many syntactic effects (see Arad 2002).
24. Maling (2001: 458) shows that 'although there is no neat one-to-one correlation
between m-case and thematic role, case does serve to identify verbal arguments
on a verb-by-verb basis'.
25. The theory which distinguishes subject and object Feature checking is in line
with Alexiadou and Anagnostoupolou's generalization (2001) according to
which 'by Spell-Out VP can contain no more than one argument with an un-
checked Case feature'. Furthermore, some current issues offer new inputs
against movement for feature checking. Among others, Picallo (2002: 118-
119) assumes 'following Chomsky 1999, that Case-φ checking (i.e. the opera-
tion Agree) is a syntactic process that applies independently of movement';
Moro (2004: 5) points out that 'movement of noun phrases in languages with
fully overt Case morphology (such as Latin) is hard to explain' if maintaining
that movement is triggered by uninterpretable features.
288 Annamaria Bartolotta

26. Evidence for the relevance of the object noun phrase to the aspectual interpre-
tation of the verb with which it is closely integrated and which denotes a com-
plex activity is to be found also in non-Indo-European languages such Hungarian
(see Kiefer 1994: 185-205).
27. Ormazabal (2000: 236) has demonstrated that in constructions using certain
verbs the object movement is triggered by an animacy feature; differently it
stays in situ.
28. This parameter has been introduced by Ura in order to characterize the ergative
languages compared to 'languages like English or other European languages:
[...] whereas Checking may coincide with θ-assignment in some languages, it
may not in others. [...] Put differently, [±0PC] is the key to the distinction be-
tween nominative-accusative languages and ergative-absolutive languages, a big
topic to which numerous studies have been devoted'. Given this background,
referring to the comments of the anonymous reviewer, the [±©PC] Parameter
could be considered as a macroparameter (in the sense of Baker 1996), to the
extent that its implications have a fundamental impact on the typology of lan-
guages such as Old Greek or Vedic at earlier stages.
29. A confirmation to what has been here hypothesized comes also from the studies
of non-Indo-European languages such as Finnish, where 'aspect is not defined
in terms of a simple, one-to-one correspondence between form and meaning
but rather as a result of the contribution of one or more of the following factors:
the lexical semantics of the verb, derivational morphology, case variation of
the object [...]' (Bache, Basboll and Lindberg 1994: 11).
30. It should not be disregarded that, as some recent hypotheses might suggest (see
Lorenzo and Longa 2003: 648), not only the case but also morphological fea-
tures such as Tense, Aspect and Aktionsart could be considered [+Interpretable],
and then they would not need to be moved in order to be checked.
31. Notice that, differently from here, Tanaka does not take into account any vP,
but he simply splits the VP in two shells, inside which an AspP is assumed to
be placed.
32. With regard to this topic, see Maling (2001: 436^137), who has pointed out the
role of the differences between a direct θ-role which is associated with a direct
argument that receives structural case and one which receives lexical case in
German.
33. Following Pinker, Hare, McRae and Elman (2003: 283) point out that 'subtle
semantic distinctions between otherwise similar verbs often determine the sorts
of syntactic structure in which the verbs may appear'.
34. As stated in Hale (1998: 1), who follows a Minimalist perspective, 'the study
of change may provide valuable insights into the proper characterization of
certain syntactic phenomena within current syntactic theory'.
 IE *weid- as a root with DS features in the Homeric poems 289

References

Adger, David, Susan Pintzuk, Bernadette Plunkett and George Tsoulas

1999 Specifiers. Minimalist approaches. Oxford: Oxford University Press.
Adrados, Francisco R.
1992 Nueva Sintaxis del Griego antiguo. Madrid: Gredos.
Alexiadou, Artemis and Elena Anagnostopoulou
2001 The Subject-in-Situ Generalization and the Role of Case in Driving
Computations. Linguistic Inquiry 32/2: 193-231.
Anderson, Stephen R. and David W. Lightfoot
2000 The Human Language Faculty as an Organ. Annual Review of Physi-
ology 62: 697-722.
Arad, Maya
2002 Universal features and language-particular morphemes. In Theoretical
Approaches to Universals, Artemis Alexiadou (ed.), 15-39. Amster-
dam/Philadelphia: John Benjamins.
Bache, Carl, Hans Basbell and Carl-Eric Lindberg
1994 Tense, Aspect and Action. Empirical and Theoretical Contributions
to Language Typology. Berlin/New York: Mouton de Gruyter.
Baker, Mark
1996 The Polysynthesis Parameter. Oxford: Oxford University Press.
Bartolotta, Annamaria
2002 a L'occhio della mente. Un'ereditä indoeuropea nei poemi omerici.
Palermo: Circolo Glottologico Palermitano.
2002b Generativismo e cognitivismo: teorie grammaticali a confronto. In
Idee e parole. Universi concettuali e metalinguistici, Vincenzo Orioles
(ed.), 9-42. Roma: II Calamo.
2003 Towards a Reconstruction of Indo-European Culture: Semantic
Functions of IE *men-. In Proceedings of the Fourteenth Annual
UCLA Indo-European Conference, Karlene Jones-Bley, Martin E.
Huld, Angela Delia Volpe and Miriam Robbins Dexter (eds.), 37-62
(Journal of Indo-European Monograph Series 47). Washington, DC:
Institute for the Study of Man.
Belletti, Adriana
1988 The Case of Unaccusatives. Linguistic Inquiry 19/1: 1-34.
Brugmann, Karl and Berthold Delbrück
1893 Grundriss der vergleichenden Grammatik der Indogermanischen
Sprachen. III (Erster Teil). [Outline of the Comparative Grammar of
the Indo-Germanic Languages III, Part 1] Strassburg: Trübner.
Chantraine, Paul
1963 Grammaire Homerique. Paris: Klincksieck.
290 Annamaria Bartolotta

Chomsky, Noam
1995 The Minimalist Program. Cambridge, Mass.: MIT Press.
2000 Linguistics and Brain Science. In Image, Language, Brain, Alec
Marantz, Yasushi Myashita and Wayne O'Neil (eds.), 13-28. Cam-
bridge, Mass./London, England: MIT Press.
De Boel, Gent
1987 La syntaxe des verbes 'voir' chez Homere. Glotta 65/1-2: 19-32.
Di Giovine, Paolo
1996 Studio sulperfetto indoeuropeo. II. Roma: II Calamo.
Embick, David
2000 Features, Syntax, and Categories in the Latin Perfect. Linguistic
Inquiry 31/2: 185-230.
Evans, Nicholas and David Wilkins
2000 In the mind's ear: the semantic extensions of perception verbs in
Australian languages. Language 76/3: 546-592.
Friedmann, Na'ama and Yosef Grodzinsky
1997 Tense and Agreement in Agrammatic Production: Pruning the Syn-
tactic Tree. Brain and Language 56: 397-425.
2000 Split inflection in neurolinguistics. In The Acquisition of Syntax,
Marc-Ariel Friedemann and Luigi Rizzi (eds.), 84-104. Harlow,
England: Longman.
Gennari, Silvia P.
2003 Tense Meanings and Temporal Interpretation. Journal of Semantics
20/1: 35-71.
Giorgi, Alessandra and Fabio Pianesi
1997 Tense and Aspect. From Semantics to Morphosyntax. New York/
Oxford: Oxford University Press.
Guasti, Maria Teresa and Luigi Rizzi
2000 Agreement and Tense as distinct syntactic positions: evidence from
acquisition, in: http://www.ciscl.unisi.it/persone/rizzi.htm.
Haegeman, Liliane
1996 Manuale di Grammatica Generativa. Milano: Hoepli. (Introduction
to Government and Binding Theory. 2nd Edition. UK: Blackwell
Publishers).
Hale, Mark
1998 Diachronic Syntax. Syntax 1/1: 1-18.
Hare, Mary, Ken McRae and Jeffrey Elman
2003 Sense and structure: Meaning as a determinant of verb subcategoriza-
tion preferences. Journal of Memory and Language 48 (2): 281-303.
Harris, C. Alice and Lyle Campbell
1995 Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cam-
bridge University Press.
 IE *weid- as a root with DS features in the Homeric poems 291

Haudry, Jean
1977 L 'emploi des cas en vedique. Introduction a l'itude des cas en indo-
europeen. Lyon: L'Hermes.
Jannaris, Antonius Ν.
1897 An Historical Greek Grammar Chiefly of the Attic Dialect. Hildes-
heim: Georg Holms.
Kiefer, Ferenc
1994 Some peculiarities of the aspectual system in Hungarian. In Tense,
Aspect and Action. Empirical and Theoretical Contributions to Lan-
guage Typology, Carl Bache, Hans Basb0ll and Karl-Eric Lindberg
(eds.), 185-205. Berlin/New York: Mouton de Gruyter.
Kurylowicz, Jerzy
1964 The inflectional categories of Indo-European. Heidelberg: Carl Winter
Universitätsverlag.
Lasso de La Vega, Jose S.
1968 Sintaxis Griega I. Madrid: Patronato Menendez y Pelayo.
Lehmann, Winfred P.
1999 La linguistica indoeuropea. Bologna: II Mulino (Theoretical Bases
of Indo-European Linguistics. 1993: London-New York: Routledge).
Lightfoot, David W.
1991 How to set Parameters: Arguments from Language Change. Cam-
bridge, Mass: MIT Press.
1998 The Development of Language: Acquisition, Change and Evolution.
Oxford: Blackwell Publishing.
2002 Syntactic Effects of Morphological Change. Oxford: Oxford Univer-
sity Press.
Lopez, Luis
2001 On the (Non)complementarity of Θ-Theory and Checking Theory.
Linguistic Inquiry 32/4: 694-716.
Lorenzo, Guillermo and Victor M. Longa
2003 Minimizing the genes for grammar. The minimalist program as a bio-
logical framework for the study of language. Lingua 113: 643-657.
Luraghi, Silvia
1996 Studi su casi e preposizioni nel greco antico. Milan: Franco Angeli.
Maling, Joan
2001 Dative: The heterogeneity of the mapping among morphological case,
grammatical functions, and thematic roles. Lingua 111: 419^164.
Marantz, Alec
1994 A late note on Late Insertion. Cambridge, Mass: MIT Press.
1995 'Cat' as a phrasal idiom: Consequences of Late Insertion in Distrib-
uted Morphology. Cambridge, Mass.: MIT Press.
292 Annamaria Bartolotta

Morani, Moreno
1992 Linee di storia della flessione nominale indoeuropea. Alessandria:
Edizioni dell'Orso.
Moro, Andrea
2004 Linear Compression as a Trigger for Movement. In Triggers, Henk
van Riemsdijk and Anne Breitbarth (eds.), 387-429. Berlin: Mouton
de Gruyter.
Ormazabal, Javier
2000 A Conspiracy Theory of Case and Agreement. In Roger Martin,
David Michaels and Juan Uriagereka (eds.), Step by Step. Essays on
Minimalist Syntax in Honor of Howard Lasnik, 235-260. Cam-
bridge, Mass./London: MIT Press.
Penke, Martina
2003 On the morphological basis of syntactic deficits. Brain and Language
87 (1): 50-51.
Pesetsky, David
1996 Zero Syntax. Experiencers and Cascades. Cambridge, Mass./London,
England: MIT Press.
Picallo, M. Carme
2002 Abstract agreement and clausal arguments. Syntax 5 (2): 116-147.
Pintzuk, Susan
2003 Variationist Approaches to Syntactic Change. In Joseph D. Brian and
Richard D. Janda (eds.), The Handbook of Historical Linguistics,
509-528. Oxford: Blackwell Publishing.
Schwyzer, Edward
1953 Griechische Grammatik I-II. München: C. Η. Beck.
Speijer, Jacob S.
1886 Sanskrit Syntax. Leiden. [Reprint 1988.] Delhi: Motilal Banarsidass.
Stroik, Thomas
2001 On the Light Verb Hypothesis. Linguistic Inquiry 32 (2): 362-369.
Sweetser, Eve
1998 From etymology to pragmatics. Metaphorical and cultural aspects of
semantic structure. 6th Edition. Cambridge: Cambridge University
Press.
Tanaka, Hidezaku
1999 Raised Objects and superiority. Linguistic Inquiry 30: 317-325.
Timberlake, Alan
1977 Reanalysis and Actualization in Syntactic Change. In Mechanisms of
syntactic change, Charles N. Li (ed.), 141-177. Austin: University of
Texas Press.
Ura, Hiroyuki
2000 Checking Theory and Grammatical Functions in Universal Gram-
mar. New York, Oxford: Oxford University Press.
 IE *weid- as a root with DS features in the Homeric poems 293

Van Valin Robert D. and Randy J. La Polla

1999 Syntax. Structure, meaning and function. 2nd Edition. Cambridge:
Cambridge University Press.
White, Lydia
2003 Second Language Acquisition and Universal Grammar. Cambridge:
Cambridge University Press.

Texts

Monro, David B. and Allen W. Thomas

1902 Homeri Opera. [Reprint 1978] Oxford: Clarendon Press
Murray, Albert T.
1924 Homer, The Iliad. Vol. I—II. Loeb Classical Library. [Reprint 1978]
Cambridge: Harvard University Press
Mazon, Paul
1957-61 Homere. Iliade. Tome I : chants I—VI; tome II: chants VII—XII; tome
III: chants XIII-XVIII; tome IV: chants XIX-XXIV. Paris: Les
Belles Lettres.
Calzecchi Onesti, Rosa
1950 Omero. Iliade. Torino: Einaudi.
The syntax of Classical Greek infinitive

Vassilios Spyropoulos

1. Introduction

Generative approaches to language assume the existence of a Universal

Grammar (UG) as a set of general principles that are common to all lan-
guages and define the Computational System of the Language (Chomsky
1981, 1995). These principles interact with a number of parameters, which
express language specific properties, so as to produce linguistic structures.
The principles of UG, therefore, offer a powerful tool for not only describ-
ing, but also explaining linguistic structure. In addition, they can also pro-
vide with a method for improving our knowledge about the structure of
dead languages. In many cases information about a dead language is so
sporadic that linguists rely on the comparative method1 in order to recon-
struct its missing structures. UG may also help linguists reconstruct such
languages: based on what they know about a language and on the general
principles of the UG, linguists can formulate hypotheses about what may
constitute a possible structure in that language. On the other hand, there are
dead languages for which there is sufficient information, so that our knowl-
edge about their grammar is pretty much complete. In such cases, UG can
help linguists to account for certain idiosyncrasies of the language under
examination and, ultimately, to explain why certain structures and phenom-
ena exist in this language.
In this paper we examine infinitival clauses in Classical Greek (CIGr).2
The infinitive is a common verbal category in the languages of the world,
and its syntax has been studied a lot. As a result, it is well established that
infinitival syntax is governed by a number of universal principles (common
to all clausal structures) regarding the following modules:

(1) a. The structure of INFL- and C-layers

b. Case-assignment
c. The licensing of null-elements
d. Theory of movement
e. Control module
296 Vassilios Spyropoulos

There are a number of parameters associated with these principles, which de-
termine the internal structure of an infinitival clause and the constructions
in which this infinitival clause can participate. A major parameter is con-
cerned with the issue of finiteness, i.e. the ability of the functional category
INFL to assign case to the subject of the clause. The issue of finiteness is
closely associated with the feature specification of INFL, i.e. the existence
of tense and agreement features, the parameters of which are argued to be
morphosyntactically determined (Chomsky 1993, 1995). Another parameter
is concerned with the categorial status of the infinitival clauses, i.e. whether
they are INFLPs or CPs. The interaction of these parameters determines the
conditions under which an infinitival clause can have an overt DP-subject
or participate in control and raising structures.
We argue that CIGr. infinitival clauses are not finite, in the sense that
their INFL does not assign case to their subjects and, consequently, CIGr.
exhibits raising and control phenomena. We also claim that CIGr. infinitival
clauses have an articulated C-domain, the lower functional projection of
which has the option of assigning accusative case to the DP-subject of the
infinitival clause, when it is properly activated. This analysis of CIGr. in-
finitives provides an explanation for a number of idiosyncratic properties of
infinitival syntax in this language, namely the occurrence of an overt DP-
subject in exactly the same environments where raising and control phe-
nomena are also found. In addition, it offers a classification of infinitival
constructions based on the properties of the infinitival clause itself and not
on the characteristics of their surface structure, as descriptive analyses of
traditional grammars do.
The paper is organized as follows: In section 2 we present some of the
main morphological and syntactic properties of CIGr infinitives. In section
3 we examine the so-called Accusativus cum Infinitivo construction and we
argue against an analysis that suggests that CIGr. infinitival INFL, being
specified for [T], is able to assign accusative case to the DP-subject of the
infinitival clause (Philippaki-Warburton and Catsimali 1989, 1991, Sevdali
2003). Instead, in section 4, it is argued that this DP-subject is assigned
accusative case by a relevant head inside the C-domain. Section 5 outlines
how the interaction of UG principles with the language specific parameters
can account for the whole range of CIGr. infinitival clauses. Section 6 con-
cludes the paper.
 The syntax of Classical Greek infinitive 297

2. Some facts about CIGr infinitives

2.1. Morphosyntax

CIGr. infinitive inflects for four of the traditional tenses (present, aorist
(past), future, (present) perfect) of the CIGr. verbal system and for three
voices (active, passive and medio-passive). The relevant forms are shown
in the following table:

(2) Infinitive forms of the verb lyo: 'set free'

Voice
Tense
active passive medio-passive
present lyein lyesthai lyesthai
future lysein lythe:sthai lysesthai
aorist lysai lythe:nai lysasthai
(present) perfect lelykenai lelysthai lelysthai

It has been noticed that tense distinctions in CIGr. infinitive mask an aspect
distinction (Jannaris 1897; Schwyzer 1950; Binnick 1991): the present in-
finitive denotes durative aspect, the aorist infinitive punctual aspect and the
perfect infinitives perfect aspect.3 The future infinitive is partly out of this
system4 for the following reason: It is considered to be a novelty in the
CIGr. verbal system that resulted from the grammaticalization of direct
speech (Schwyzer 1950). As a consequence, it has a very limited distribu-
tion and can be mainly used after the so-called verba dicendi et sentiendi,5
where it substitutes for the future indicative of the direct speech:6

(3) tous nomizontas ikanous esesthai ta symferonta

the think-PART.ACC.PL7 able-ACC.PL be-INF.FUT the useful-ACC.PL
didaskein tous politas
teach-lNF.PRS the citizen-ACC.PL
'those w h o think that they will be able to teach the citizens their own
interests' (X. Mem. 1.2, 10)

Interestingly, in the structures consisting of a verbum dicendi/sentiendi and

an infinitive, the infinitive may vary for all possible tenses (present, past,
future, perfect), with each form referring to a different temporal point in
relation to the tense of the main clause:
298 Vassilios Spyropoulos

a. elegon ton kyron poiein touto

said-3PL the Cyrus-ACC.SG do-INF.PRS this-ACC
b. elegon ton kyron poie.sein touto
said-3PL the Cyrus-ACC.SG do-INF.FUT this-ACC
c. elegon ton kyron poie:sai touto
said-3PL the Cyrus-ACC.SG do-INF.AOR this-ACC
d. elegon ton kyron pepoie:kenai touto
said-3PL the Cyrus-ACC.SG do-INF.PRF this-ACC
'They were saying that Cyrus is doing/will do/did/have done this'

On the basis of this variability, it has been suggested that CIGr. infinitive
carries morphological specification for tense, and thus its INFL is specified
as [+T] (Philippaki-Warburton and Catsimali 1989, 1991; Tantalou 2003;
Sevdali 2003). However, this variability is not always possible. As it was
mentioned above, the future infinitive has a very limited distribution and it
is mainly found after the verba dicendi/sentiendi, where a full sequence of
tenses is established. However, in other embedded clauses the future infini-
tive is not standardly used and the different infinitival forms denote only
aspect: present - durative, aorist - punctual, perfect - perfect:

(5) a. tois allots past pare:ggellen eksoplizesthai

the other-DAT.PL all-DAT.PL was.ordering give.arms-INF.MED.PRS
' H e was ordering all the others to take the arms' (X. An. 1.8, 3)

b. l*eksoplisesthai /eksoplisasthai
give.arms-INF.MED.FUT /give.arms-INF.MED.AOR
/eksoplisthai
/give.arms-INF.MED.PRF

In addition, there are also embedded infinitival clauses where the infinitive
cannot vary for tense, and only one infinitive form (usually the present in-
finitive) is possible. In these cases, the infinitive and the matrix verb denote
the same event:

(6) arkhomai epainein /*epaine:sein /*epaine:sai

start-1SG praise-INF.PRS / praise-INF.FUT / praise-INF.AOR
Ί start praising'

We will return to this issue later in section 3.2.

 The syntax of Classical Greek infinitive 299

2.2. Syntactic distribution

The syntactic distribution of the infinitive in CIGr is the following:

- Infinitive in subject clauses:

(7) tis alke: (estin) ton thanonta epiktanein?
what help-NOM (is) the dead-ACC kill.again-INF.AOR
'What would be the benefit of killing again a dead man?' lit.
'What kind of help is to kill again a dead man?'

- Infinitive in complement clauses:

(8) a. meno.n ebouleto ploutein
Menon-ΝΟΜ wanted-3SG get.rich-INF.PRS
'Menon wanted to get rich'
b. omologo: meidian ... lamprotaton gegene.sthai
admit-lSG Meidias-ACC glorious-ACC become-INF.PRF
Ί admit that Meidias has become the most glorious...' (Z>. 21, 153)
c. sou ... deomai akolouthein
you-GEN beg-lSG follow-INF.PRS
Ί beg you to follow...' {PI. Prt. 336a)

- Infinitive in temporal and result adjunct clauses introduced with the

complementizers prin 'before' and o:ste 'so that' respectively:
(9) a. ou proteron epausato makhomenos ..., prin
not before stopped-3SG fight-PRT.MS.NOM.SG before
helein to basileion
conquer-INF.AOR the kingdom-ACC.SG
'He didn't stop fighting until he conquered the kingdom'
(Isoc. 9,32)
b. te:i orge:i houto: khalepe:i ekhre.to es pantas,
the anger-DAT so tough-DAT used-3SG to all-ACC.PL
o.ste me.dena dynasthai prosienai
so.that nobody-ACC.SG be.able-INF.AOR get.close-INF.PRS
'He was behaving so badly to everybody, so that nobody could stay
with him' (Th. 1.130, 2)
300 Vassilios Spyropoulos

- Infinitive in main clauses instead of the imperative:

(10) sy de moi ta khre.mata deksai...

you-NOM PRT I-DAT the money-ACC.PL accept-INF.AOR
'You, on the other hand, accept the money from m e . . . ' (Hdt. VI.86, a)

- Infinitive as a noun, introduced with the definite article and functioning

as a proper D P (the articular infinitive)·.

(11) a. to so.fronein timo.sa

the-ACC be.wise-INF.PRS value-PART.FEM.NOM.SG
tou biou pleon
the-GEN life-GEN more
'...(she) valuing wisdom more than life' (A. Supp. 1013)

b. hina apisto.si to:i eme tetime:sthai

COMP doubt-SBJ.3PL the-DAT 1-ACC honour-PASS.INF.PRF
hypo daimono.n
by gods-GEN
'in order to doubt that I have been honoured by the gods'
(X. Mem. 14)
c. arksantes tou diabainein
begin-PART.AOR.MSC.NOM.PL the-GEN cross-INF.PRS
' . . . ( t h e y ) starting to c r o s s . . . ' (X. An. 1.4, 15)

- Infinitive in fixed expressions (absolute infinitive)·.

(12) ale:thes ge, ho:s epos eipein, ouden

true-NTR.ACC.SG PRT so word-ACC.SG say-INF.AOR nothing
eire.kasin
have.said-3PL
' S o to speak, not a single word of what they have said is true'

Leaving aside the so called 'absolute infinitive' and the infinitive in main
clauses, we move on to examine the infinitival syntax in CIGr. by focusing
on the case properties of the subject of the infinitival clause.

2.3. Traditional grammar description of infinitival syntax

Traditional grammars describe infinitival syntax by focusing on the mor-

phological case of the DP that is interpreted as the subject of the infinitive.
 The syntax of Classical Greek infinitive 301

They refer to four possible 'syntaxes': nominativus cum infinitivo, accusa-

tivus cum infintivo, genitivus cum infinitivo and dativus cum infinitivo.

(13) nominativus cum infinitivo (NomI)

a. (houtos) ouketi edynato bioteuein

(he-NOM) no more was.able-3SG live-INF.PRS
' H e was not able to live any more'
a', he/proi edynato [PROj bioteuein]

b. meno.n ebouleto ploutein

Menon-ΝΟΜ wanted-3SG get.rich-INF.PRS
' M e n o n wanted to get rich'
b'. meno:nj ebouleto [PRO] ploutein]

c. ο assyrios eis te:n kho:ran embalein

the Assyrian-NOM in the country-ACC invade-INF.AOR
aggelletai
announce-PASS.3SG
'It is announced that the Assyrian has invaded the country'
c'. [o assyrios]j aggelletai [tj embalein eis te:n kho:ran]

(14) accusativus cum infinitivo (AccI)

a. es touton ton kho.ron legetai
in this the place-ACC say-PASS.3SG
aphikesthai ton straton
arrive-INF.AOR the army-ACC
'It is said that the army has arrived in this place' (Hdt. 111.26)
a'. protxpi legetai [[ton straton] aphikesthai...]

b. omologo: meidian ... lamprotaton gegene.sthai

admit-lSG Meidias-ACC most.glorious-ACC become-INF.PRF
Ί admit that Meidias has become most glorious' (D. 21, 153)
b'. proi omologo: [[Meidian] gegenesthai lamprotaton]

c. dedidakhen ... home ros

has.taught Homer-NOM
kai tous allous pseude: legein
and the others-ACC lies-ACC say-INF.PRS
' H o m e r has taught the others also to tell lies'
c'. home:roSj dedidakhen [tous allous]j [PROj legein pseude:...]
(Arist. Po. 1460 a 1 8 - 1 9 )
302 Vassilios Spyropoulos

(15) genitivus cum infinitivo (GenI)

a. sou ... deomai akolouthein
you-GEN b e g - l S G follow-INF.PRS
Ί beg you to follow...'
a', proj deomai souj [ P R O j akolouthein] (PL Prt. 336a)

(16) dativus cum infinitivo (DatI)

a. tois allois pa:si pare:ggellen eksoplizesthai
the other-DAT.PL all-DAT.PL was.ordering give.arms-lNF.MED.PRS
'he w a s ordering all the others to take the arms'
a', proj pare:ggellen [tois allois pa:si]j [PROj eksoplizesthai]

b. edoksen autois ... eksoplisamenois

thought-3SG they-DAT give.arms-PRT.MED.DAT.MSC.PL
proienai
walk.forward-INF.PRS
'They decided to take the arms and march forward' lit. 'It was de-
cided by them, after they take the arms, to march forward'
b'. proexpi edoksen autoisj [PRO, proienai] (X. An. II. 1, 2)

However, none of these 'syntaxes' describes a homogeneous set of struc-

tures. Thus, it is obvious that the nominativus cum infinitivo syntax is a
cover term for obligatory subject control (13a), optional subject control
(13b) and raising-to-subject structures (13c). Accusativus cum infinitivo
syntax describes both object control structures (14c) and structures with an
infinitival clause containing a DP-subject marked with accusative case
(14a,b). Genitivus cum infinitivo and dativus cum infinitivo syntaxes de-
scribe control structures, with the DP in genitive/dative being an argument
of the matrix verb.
In what follows we examine the properties of CIGr. infinitival syntax, in
order to discover the principles that underlie this divergent set of structures.
We will be mainly concerned with a construction that is considered to be an
idiosyncrasy of CIGr. infinitival syntax, namely the AccI syntax.

3. Accusativus cum infinitivo syntax

Accusativus cum infinitivo (AccI) syntax describes a syntagm that consists

of an infinitive and a DP in accusative which is understood as the subject of
the infinitive. However, the syntactic status of this DP varies, as it can be
 The syntax of Classical Greek infinitive 303

either the subject of the embedded infinitival clause or the DP-object of the
matrix predicate. Recall examples (14b,c) repeated as (17a,b):

(14) a. omologo: meidian ... lamprotaton gegene:sthai

admit-lSG Meidias-ACC most.glorious-ACC become-INF.PRF
Ί admit that Meidias has become most glorious' (D. 21, 153)
b. dedidakhen... home.ros kai tous allous pseude: legein
has.taught Homer-NOM and the others-ACC lies-ACC say-FNF.PRS
'Homer has taught the others also to tell lies'
CArist. Po. 1460 a l 8-19)

In (17a) the DP in the accusative is not theta-marked by the matrix verb

and, therefore, it is the structural subject of the embedded infinitival clause
(18a). On the other hand, the DP in the accusative in (17b) realizes an argu-
ment of the matrix verb, from which it receives structural accusative case.
The subject of the embedded infinitival clause is a controlled empty cate-
gory, namely a PRO (18b).

(18) a. pro\ omologo: [[meidian] gegenesthai lamprotaton]

b. home:rosj dedidakhen [tous allous]j [PRO; legein pseude:...]

The AccI sequences of the second type constitute object control constructions
and should not be confused with those of the first type. From now on, we will
use the term AccI syntax for referring to the constructions of the first type
only, i.e. for the structures in which the DP in the accusative is the structural
subject of the embedded infinitival clause.
Given that the DP in the accusative case in AccI syntax is the structural
subject of the infinitival clause, the source of its accusative case calls for an
explanation. The principles of case assignment present us with two possi-
bilities: (a) AccI syntax involves Exceptional Case Marking (ECM), i.e. the
DP-subject receives its accusative case from the matrix predicate; and (b)
the accusative case is assigned by a head inside the infinitival clause.

3.1. The ECM analysis

According to an ECM analysis of the AccI syntax, the DP-subject of the

embedded infinitive receives its accusative case by participating in a syn-
tactic relation with the matrix predicate. There are different proposals about
304 Vassilios Spyropoulos

the exact syntactic mechanism that assigns this accusative case, 8 but there
is a consensus about the conditions such a mechanism should obey: (a) the
matrix predicate must be able to assign structural case; and (b) the embedded
infinitival clause should project only to the level of INFLP, i.e. there should
be no CP intervening between the matrix predicate and the embedded DP-
subject.
However, AccI syntax appears in constructions that do not obey these
conditions: First, the infinitival clause in AccI constructions is a CP, as co-
ordination with an o/7'-clause shows (Sevdali 2003; Tantalou 2003):

(19) a. hymas pantas eidenai e:goumai oti ego: men

you-ACC.PL all-ACC.PL know-INF.PRS think-1SG that I-NOM PRT
ortho:s lego: touton de skaion einai
correctly say-lSG he-ACC PRT unkind be-INF.PRS
Ί think that you all know that I, on the one hand, speak correctly,
while he, on the other hand, is unkind'

b. e:goumai [hymas pantas eidenai [cp oti ego: men lego: ortho:s]
[infCL touton de einai skaion]]

In (19) the infinitival clause (InfCL) is co-ordinated with a clause intro-

duced by the complementizer oti 'that'. The latter clause is therefore a CP
and since the infinitival clause can be co-ordinated with such a clause, it
must also be a CP.
The CP status of infinitival clauses with AccI syntax is also evident by the
fact that such clauses occur in environments where only a CP is allowed.
Thus, an infinitival clause with AccI syntax can appear as clausal associate
to an zY-expletive:

(20) a. khre:n kai tous allouspotamous... homoio:s paskhein

was.necessary and the other rivers-ACC similarly suffer-INF.PRS
'The other rivers should also suffer in the same way' (Hdt. 11.20)
b. proexpi khre:n [kai tous allous potamous homoio:s paskhein]

Clausal associates of /f-expletives are considered to be CPs (Chomsky 1981,

2001). CIGr. is a null-subject language, thus /i-expletives have the syntactic
status of a null-pronoun proexp\ specified as 3SG. Its clausal associate, there-
fore, should have the status of a CP, which means that the infinitival clause
in (20b) is a CP. In addition, infinitival clauses with AccI syntax can appear
as subjects:
 The syntax of Classical Greek infinitive 305

(21) a. tro.as algea paskhein ou nemesis (estin)

Trojans-ACC sufferings-ACC suffer-lNF.PRS NEG just-NOM (is)
'It's not right for the Trojans to suffer in this way'
b. [[tro:as paskhein algea] ou estin nemesis]

Infinitival clauses in subject position have been considered to be CPs

(Chomsky 1981, Manzini 1983, Haegeman 1994). Thus, we conclude that
the infinitival clause in (21) is a CP.
Second, AccI syntax occurs with infinitival clauses that are complements
of verbs with no case assigning properties. More specifically, consider ex-
ample (22) where the infinitival clause is a complement of an 'impersonal'
verb:

(22) synebe: ... gelo:na nika:n amilkan

happened-3SG Gelon-ACC defeat-INF.PRS Amilcas-ACC
'It happened that Gelon defeated Arnikas' (Hdt. VII. 166)

Impersonal verbs do not assign an external theta role and take a clausal
complement that realizes a 'theme' theta-role. Thus, they are not able to
assign accusative case by Burzio's Generalization (Burzio 1986). Conse-
quently, the DP-subject of the infinitival clause in these structures cannot
receive its accusative case from the matrix verb. Similarly, the AccI syntax
in an infinitival clause introduced by a verbum dicendi is maintained even
when passivization has turned the matrix verb into an 'impersonal' verb.
Consider example (14a) repeated here as (23a) and its active equivalent
(23b):

(23) a. es touton ton kho.ron legetai

in this the place-ACC say-PASS.3SG
aphikesthai ton straton
arrive-lNF.AOR t h e army-ACC
'It is said that the army has arrived in this place'
b. es touton ton kho. ron legousin aphikesthai ton straton
in this the place-ACC say-3PL arrive-lNF.AOR the army-ACC
'They say that the army has arrived in this place'

Example (23a) is the passive equivalent of (23b). The relevant structures

are shown in (24). Notice that in an ECM construction, passivization of the
matrix verb makes ECM unavailable (25):
306 Vassilios Spyropoulos

(24) a. proi legousin [[ton straton] aphikesthai]

b. proexpi legetai [[ton straton] aphikesthai]

(25) a. They consider [Zidanne to be the best football player]

b. *It is considered [Zidanne to be the best football player]

Since the external argument of a passive verb is not syntactically realized,

such verbs are not able to assign accusative case. Thus, ECM is banned in
this environment. Examples (24) and (25) constitute a minimal pair, which
shows that in CIGr. the AccI syntax does not involve ECM. In addition,
infinitival clauses with AccI syntax can appear as complements of nouns:

(26) he: aggelia tous helle:nas ta hopla

the announcement-NOM the Greeks-ACC the weapons-ACC
tithes thai...
put-INF.MED.PRS
'the announcement that the Greeks hand over the weapons...'

Nouns are assumed to assign inherent case (Chomsky 1981, 1986), and
thus they cannot assign case to the DP-subject of an embedded infinitival
clause, because they do not theta-mark it. This is the reason why ECM is
banned in infinitival clauses that are complements of nouns as the un-
grammaticality of (27) shows:

(27) *the belief of Zidanne to be the best football player

Thus, the grammaticality of example (26) indicates that AccI syntax does
not involve ECM.
Let us summarize so far: We presented evidence that AccI syntax can
appear in environments which do not respect the conditions of ECM. Thus,
it was shown that infinitival clauses with AccI syntax (a) are CPs and (b)
appear as complements to matrix predicates that are not able to assign
structural accusative case. All these facts indicate that AccI syntax does not
involve ECM and the source of the accusative case of the DP-subject of the
infinitival clause is clause internal. This conclusion is reinforced by the
following examples, where no case assignment relation can be established
between the matrix predicate and the DP-subject of the infinitival clause
because of the syntactic status of the infinitival clause:
 The syntax of Classical Greek infinitive 307

(28) dareios, prin men aikhmalo. tous genesthai

Darius-NOM before PRT captives-ACC become-INF.AOR
tous eretrieas, eneikhe sphi deinon kholon
the Eretrians-ACC had-3SG them-DAT terrible-ACC anger-ACC
'Darius was galled with Eretrians, before they were taken captives'

(29) a. hina apisto.si to:i eme tetime.sthai

COMP doubt-SBJ.3PL the-DAT I-ACC honour-PASS.INF.PRF
hypo daimono:n
by gods-GEN
'in order to doubt that I have been honoured by the gods'
b. apisto:si [ DP to:i [Cp eme tetime:sthai ]]

In example (28) the infinitival clause is a temporal clause introduced with

prin 'before'. Such clauses are adjuncts and therefore the matrix verb has
no access to their internal structure (Condition on Extraction Domains',
Huang 1982). Yet, AccI syntax is possible as (28) shows. Similarly, AccI
syntax occurs in nominalized infinitival clauses. Such infinitival clauses are
introduced with the definite article to 'the' in the appropriate case and have
the status of a DP. Case assignment from the matrix predicate is blocked by
the intervention of the DP node (a Relativized Minimality or Minimal Link
Condition effect; Chomsky 1986, 1995; Rizzi 1991).
Another argument against the ECM analysis of AccI syntax comes from
a very interesting idiosyncrasy of the GenI and DatI constructions. In sec-
tion 2.3, we mentioned that these constructions typically involve a control
structure, with the DP in the genitive/dative case realizing an argument of
the matrix predicate and controlling an empty subject in the infinitival
clause (30b)-(31b):

(30) a. sou ... deomai akolouthein

you-GEN beg-lSG follow-INF.PRS
Ί beg you to follow...'
b. pros deomai souj [PROj akolouthein]

(31) a. tois allots pa:si pare.ggellen eksoplizesthai

the other-DAT.PL all-DAT.PL was.ordering give.arms-INF.MED.PRS
'he was ordering all the others to take the arms'
b. proj pare:ggellen [allois pa:si]j [PROj eksoplizesthai]
308 Vassilios Spyropoulos

The controlled empty subject of the embedded infinitival clause is a PRO

that carries no case. This analysis is supported by the following facts:

(32) a. kyrou edeonto o:s prothimotatou pros ton polemon

Cyrus-GEN begged-3SG so most.willing-GEN to the war-ACC
genesthai
become-lNF.AOR (X. HG. 1.5, 2)
'They were begging Cyrus to become most willing for war'
b. proj edeonto [kyrou-GEN]j [PROj genesthai o:s prothimotatou-GEN]

(33) a. ο de:mos synekho:re:sen auto.i

the public-NOM permitted-3SG he-DAT
pro.to.i graphe.nai
first-DAT write-INF.PASS.AOR
'the public permitted him to register first' (Aeschin. III. 186)
b. ο de:mos synekho:re:sen [auto:i-DAT]j [PROj graphe:nai proto:i-DAT]

In (32)-(33) the infinitival clauses contain nominal predicate modifiers that

modify their empty subjects. In CIGr., predicate modifiers, as all nominal
modifiers, obligatorily agree in case with the elements they are predicated
of or modify (Andrews 1971; Morel 1989).9 Thus, the case of such modifi-
ers can be used as a diagnostic for the case of the element they modify,
when this element is phonologically empty. The predicate modifiers in
(32)-(33) modify the empty controlled subject of the embedded infinitival
clause and agree in case with the DP that is coindexed with this empty sub-
ject. This means that the empty subject of the embedded infinitival clause
carries no case (i.e. it is a PRO), so that case transmission from the matrix
controller is not blocked. But, one can easily find examples such as the
following, in which the predicate modifier appears in the accusative case
instead of the expected genitive/dative:

(34) a. deomai hymo:n ... iatrous genesthai

beg-lSG you-GEN doctor-ACC become-INF.AOR
Ί beg you to remedy...'
b. deomai [hymo:n-GEN]j [ec{ genesthai iatrous-ACC]
 The syntax of Classical Greek infinitive 309

(35) a. oude autois the.baiois epetrepete

not these-DAT Thebans-DAT were.permitting-2PL
autonomous einai
independent-ACC be-INF.PRS {X. HG. VI.3, 9)
'Not even Thebans were you permitting to be independent'
b. epetrepete [the:baiois-DAT]i [ec, einai autonomous-ACC]

That the predicate modifier is marked with accusative case indicates that
the empty subject of the embedded infinitival clause, which the predicate
modifier modifies, is also marked with accusative case, so that it blocks
case transmission from the controller in the matrix clause. Thus, examples
(34) and (35) involve a hidden AccI construction, in which the infinitival
clause contains a controlled null-subject pro in the accusative case:

(36) a. deomai [hymo:n-GEN]j [[/?ro-ACC]i genesthai iatrus-ACC]

b. epetrepete [the:baiois-DAT]j [ [ p r o - A C C ] j einai autonomous-ACC]

Since the matrix predicate has already assigned genitive/dative case to its
DP-object, ECM cannot be the explanation for the accusative case of the in-
finitival empty subject.
To sum up, the structures examined above clearly show that AccI syntax
appears in environments where ECM is not possible. All these facts indi-
cate that there is a head inside the infinitival clause that is able to assign
accusative case to its subject. In what follows we will present and assess
two hypotheses: (i) the Tense-hypothesis, according to which the source of
this accusative case is the [+T] feature of infinitival INFL, (ii) the C-
hypothesis, which claims that the accusative case is assigned by a null
prepositional complementizer.

3.2. The role of Tense and the Tense-Hypothesis

CIGr. infinitives inflect for tense and in this respect they differ from the
infinitives of languages such as English, Italian etc. Although not all the
infinitival clauses exhibit the same temporal properties, as we will see be-
low, CIGr. infinitival INFL may be considered to be specified as [+T], as a
reflex of this morphological marking. From this point of view, the [T] fea-
ture of the INFL is assumed to be morphologically defined. Philippaki-
Warburton & Catsimali (1989, 1991) and Sevdali (2003) capitalize on this
310 Vassilios Spyropoulos

tense property of the infinitival INFL and suggest that the infinitival INFL
in CIGr has case assigning properties, because it is specified as [+T]. 10 Let
us call this approach as the Tense-Hypothesis. Significantly, this account
associates the case assignment ability of INFL with the existence of a mor-
phologically defined [T] feature." Thus, an INFL head may have the fol-
lowing case assigning properties according to its specification:

(37) a. [+T, +AGR] nominative case (finite clauses)

b. [-T, -AGR] no case (non-finite clauses)
c. [+T, - A G R ] accusative case (CIGr. infinitives)
d. [-T, +AGR] nominative case (Portuguese infinitives' 2 )

Strictly speaking, such an analysis makes some very clear predictions:

CIGr. infinitives are [+T], because of their paradigmatic alternation for
morphological tense. If case assignment properties are associated with
[+T], then all infinitival clauses in CIGr. are expected to be finite and have
either an overt DP-subject or a null pro subject marked with accusative.
Consequently, CIGr. infinitival clauses are expected not to allow for raising
and syntactic control in terms of a PRO. The data presented so far show
that this prediction is falsified. We will return to the discussion of these
data in the next section.
There is a worrying aspect in this analysis: It is mainly based on the
AccI constructions with verba dicendi et sentiendi, where the infinitive can
vary for all tenses and a full sequence of tenses is established:

(38) a. elegon ton kyron poiein touto

said-3PL the Cyrus-ACC.SG do-INF.PRS this-ACC
b. elegon ton kyron poie.sein touto
said-3PL the Cyrus-ACC.SG do-INF.FUT this-ACC
c. elegon ton kyron poie.sai touto
said-3PL the Cyrus-ACC.SG do-INF.AOR this-ACC
d. elegon ton kyron pepoie.kenai touto
said-3PL the Cyrus-ACC.SG do-INF.PRF this-ACC
'They were saying that Cyrus is doing/will do/did/have done this'

In such examples the infinitival clause has both morphologically and inter-
pretatively full temporal properties. However, not all infinitival clauses
share the same temporal properties. In examples such as the ones in (39),
 The syntax of Classical Greek infinitive 311

the infinitival clause has a fixed temporal interpretation (the event time of
the infinitival clause is future oriented with respect to that of the matrix
clause) and the variation among the infinitive forms simply denotes aspec-
tual variation. Crucially, the future infinitive is not normally found in such
examples: 1 3

(39) a. tois allots pasi pare.ggellen eksoplizesthai

the other-DAT.PL all-DAT.PL was.ordering give.arms-INF.MED.PRS
' H e was ordering all the others to take the arms'

b. /*eksoplisesthai /eksoplisasthai
give.arms-INF.MED.FUT/give.arms-INF.MED. AOR
leksoplisthai
give.arms-lNF.MED.PRF

There are also examples in which the infinitival clause can be argued to
have no temporal properties at all:

(40) arkhomai epainein l*epaine:sein l*epaine:sai

start-lSG praise-INF.PRS/ praise-INF.FUT/ praise-INF.AOR
Ί start praising'

In such examples, the event time of the infinitive is identical or simultaneous

with that of the matrix predicate. As a result no variation of infinitive forms
can be found, and the infinitive can appear only in one form, mainly that of
the present infinitive.
Putting these facts together, we can conclude that there is a three level
pattern as far as the temporal properties of the infinitival clauses are con-
cerned: (a) The infinitival clause has full temporal properties, which are
determined by the tense morphology of the infinitive (free infinitives (FI)),
(b) The infinitival clause has a fixed tense value, which is imposed by the
semantics of the matrix predicate. The tense morphology of the infinitive is
in this case interpretatively transparent (dependent infinitives (DI)). (c) The
infinitival clause has no tense value at all, or it is anaphoric to that of the
matrix predicate (anaphoric infinitives (AI)).
These facts show that there is an asymmetry between the paradigmatic
value of the [T] feature and its interpretative value. An analysis that associ-
ates case assignment with the temporal properties of the INFL should de-
cide and clearly explain whether case assignment properties are associated
with the paradigmatic or the interpretative value of this [T] feature. In what
312 Vassilios Spyropoulos

follows, we will show that we need not worry about that, since there is
enough evidence that shows that the assignment of accusative case to the
DP-subject of the infinitive is not determined by the [+T] specification of
the infinitival INFL.

3.3. Evidence against the Tense-hypothesis

Let us examine the following examples:

(41) a. oude phthora outo.s anthro:po:n oudamou

no destruction-NOM so people-GEN.PL no.place
emne:moneueto genesthai
mention-PASS.PST.3 SG become-INF.AOR
'Never was such a slaughter mentioned anywhere to have taken
place' (Th. 11.51)
b. ephe: ... aksein lakedaimonioius zo:ntas
said-3SG bring-INF.FUT Lakedaimonians-ACC live-PRT.PRS.ACC.PL
' H e said that he would bring back the Lakedaimonians alive'
(Th. IV.28, 4)

Examples (41a) and (41b) involve subject-to-subject raising and subject

control respectively:

(42) a. [DP oude phthora outo:s anthro:po:n]j emne:moneueto [IP tj gen-

esthai] 1 4
b. proj ephe: [cp PROj aksein lakedaimonious]

Both examples (41a) and (41b) involve an infinitival clause as a comple-

ment to a verbum dicendi. The infinitival clause is able to establish a full
sequence of tenses as the examples in (43) show:

(43) a. oude phthora emne:moneueto

no destruction-NOM mention-PASS.PST.3SG
genesthai /gignesthai lgene:sesthai
become-INF. AOR/become-lNF.PRS/become-lNF.FUT
b. ephe: agein /aksein lagagein te.ksein
said-3 SG bring-INF.PRS/bring-INF.FUT/bring-INF.AOR/bring-INF.PRF
 The syntax of Classical Greek infinitive 313

According to the Tense-hypothesis, this tense variation indicates that the

infinitival I N F L is specified as [+T] and as a consequence it has the ability
of assigning accusative case. Thus, neither can subject-to-subject raising
take place nor can a P R O be licensed as the subject of the infinitival clause.
The examples in (41) would then be incorrectly predicted to be ungram-
matical. In fact, in CIGr., Accl syntax in infinitival clauses after a verbum
dicendi/ sentiendi is in free variation with a N o m I syntax, which involves
either subject-to-subject raising or subject-control:

(44) a. N o m I syntax involving subject-to-subject raising

ο assyrios eis te:n kho.ran embalein
the Assyrian-NOM in the country invade-INF.AOR
aggelletai
anounce-PASS.3SG
' T h e Assyrian is announced to have invaded the country'
(X. Cyr. V.3, 30)
b. N o m I syntax involving subject-control
ο assyrios eis te:n kho. ran embalein aggellei
the Assyrian-NOM in the country invade-INF.AOR anounce-3SG
' T h e Assyrian announces that he invaded the country'

c. Accl syntax
ton assyrion eis te:n kho. ran embalein aggellousi
the Assyrian-ACC in the country invade-INF.AOR anounce-3SG
'They announce that the Assyrian invaded the country'

d. Accl syntax after a passive verbum dicendi

ton assyrion eis te:n kho:ran embalein
the Assyrian-ACC in the country invade-INF.AOR
aggelletai
anounce-PASS.3SG
'It is announced that the Assyrian invaded the country'

It is obvious that the Tense-hypothesis has no account for the examples with
N o m I syntax, because in these examples it is evident that the infinitival
INFL, despite its [+T] specification, assigns no case. This conclusion gains
further support f r o m the following examples:
314 Vassilios Spyropoulos

(45) a. e.ggeltai ... he: ... makhe: pany iskhyra

announce-PASS.PRF.3SG the battle-NOM very strong-NOM
gegonenai
become-INF.PRF (PL Chrm. 153b)
'The battle has been announced to have become very fierce'
b. [Dp e: makhe:-NOM]j e:ggeltai [ [ec]j gegonenai iskhyra-NOM]

(46) a. ksynebe: san de kai byzantioi ... ype.kooi einai

agreed-3PL PRT and Byzantines-NOM vassals-NOM be-INF.PRS
'The Byzantines also agreed to be vassals'
b. [DP byzantioi-NOM]j ksynebe:san [ [ec]j einai ype:kooi-NOM]

In examples (45) and (46) the predicate modifier, which modifies the empty
subject of the embedded infinitival, is marked with nominative case. Given
that predicate modifiers always agree in case with the element they modify,
the nominative case of the predicate modifiers in (45) and (46) indicates
that no case assignment has taken place in the embedded infinitival clause
and that the ec-subject is an NP-trace and a PRO respectively.
The facts presented above show that AccI syntax is in free variation
with control and raising phenomena in FIs. Interestingly, AccI syntax is in
free variation with a control structure in DIs and after a big variety of
predicates as well, which indicates that this variation is a generalized phe-
nomenon of CIGr. complementation structures:

(47) a. AccI syntax

ton men synebe: apothanein
the-ACC.SG PRT happened-3SG die-INF.AOR
'It happened that this one died...'
a'. proexΡ, synebe: [ C P fop-subject ton men] apothanein]
(Arist. Po. 1452 b 28)
b. DatI syntax involving control
ksynebe: hymi:n peisthe.nai akeraiois
happened-3SG you-DAT.PL persuade-PASS.INF.AOR intact-DAT.PL
'It happened to you to be persuaded when you were safe...'
b'. proexpi ksynebe: hymi:nj [Cp PRO; peisthe:nai akeraois]
(:Th. 11.61,2)
 The syntax of Classical Greek infinitive 315

(48) a. N o m I involving control

eime: sy boule:i apokrinasthai
if NEG you-NOM want-3SG reply-INF.AOR
'if you don't want to reply...'
a', ei syj me: boule:i [cp PRO] apokrinasthai ]

b. AccI syntax
kindyneuein ouk eboulonto .... autous
risk-INF.PRS NEG wanted-3PL these-ACC
'They didn't want for them to r i s k . . . '
b'. pro\ ouk eboulonto [Cp autouSj kindyneuein] (Th. IV. 15)

(49) a. DatI syntax involving control

to is allois pa.si pare.ggellen eksoplizesthai
the other-DAT.PL all-DAT.PL was.ordering give.arms-INF.MED.PRS
'he was ordering all the others to take the arms'
a', proj pare:ggellen [tois allois pa:si]j [Cp PROj eksoplizesthai]
(X. An. 1.8, 3)
b. AccI syntax
pare.ggeile ta hopla tithesthai tous helle. nas
ordered-3SG the arms-ACC put-INF.MED.PRS the Greeks-ACC
' H e gave the order that the Greeks should hand over the arms'
b'. proi pare:ggeile [ C p bp-subject tous helle:nas], tithesthai ta hopla]
(X.An. 11.2,21)

The constructions examined above show that both a DP-subject in accusa-

tive case and a caseless null-subject (PRO or NP-trace) may appear in the
same environment. Given that the properties of the infinitival I N F L remain
stable - i t is specified as [ + T ] - we conclude that INFL is not responsible for
this case variation. More specifically, it is evident that infinitival INFL
does not assign accusative case, otherwise this variation w o u l d n ' t have
existed. Thus, we conclude that infinitival INFL in CIGr. is non-finite and
has no case assigning properties. This hypothesis accounts for the raising
and control p h e n o m e n a in a straightforward way. Thus, AccI syntax is sug-
gested to be an optional Last Resort operation for the licensing of an overt
DP-subject in the infinitival clause, when required.
316 Vassilios Spyropoulos

4. AccI syntax and the C-Hypothesis

In the previous section, we concluded that AccI syntax is mainly used as a

Last Resort operation that licenses an overt DP-subject in the infinitival
clause. A similar mechanism is used in English infinitival constructions,
when suspension of control is required:

(50) a. Bill, wants [PROj to go shopping]

b. Bill wants [for Mary to go shopping]

(51) a. John, tried hard [PROj to get this job]

b. John tried hard [for Mary to get the job]

The verbs want and try are control verbs and take as a complement an in-
finitival clause with a P R O subject (50a) and (51a). The control pattern is
suspended in (50b) and (51b) by the insertion of the preposition for as C.
The preposition assigns accusative case to the subject of the infinitival
clause, allowing for it to be overtly realized, and thus control is suspended.
The similarity with the corresponding constructions from CIGr. seems not
to be accidental:

(52) a. ei me: sy boule:i apokrinasthai

if NEG you-NOM want-3SG reply-INF.AOR
'if you don't want to reply...'
b. kindyneuein ouk eboulonto ... autous
risk-INF.PRS NEG wanted-3PL these-ACC
'They didn't want for them to r i s k . . . '

CIGr. uses AccI syntax in the same way as English uses ybr-infinitivals.
This similarity indicates that the accusative case in AccI syntax is deter-
mined by the C-layer. W e will call this analysis the C-Hypothesis (Tantalou
2003).
AccI syntax is also found in Latin (Bolkestein 1976; Pillinger 1980;
Cecchetto and Oniga 2002):

(53) laudavisse hasce ait architectonem

praise-INF.PAST these-ACC.PL say-3SG architect-ACC.SG
' H e says that an architect has praised these'
 The syntax of Classical Greek infinitive 317

Cecchetto and Oniga (2002) examine the properties of AccI syntax in Latin
and conclude that it does not involve ECM. They propose that the DP-
subject of the embedded clause is assigned accusative case by an empty
prepositional complementizer, something like a null for. Tantalou (2003)
claims that CIGr. AccI syntax should be analyzed in the same way and she
argues that this analysis is supported by the CP-status of the infinitival
clause in an AccI construction. We believe that this analysis is on the right
track, mainly because it can account for the optional assignment of the ac-
cusative case. If this null complementizer is an optional element, as for is in
English, control and raising are also possible with the same infinitival
clause, when this null-C is absent. But, such an analysis requires a number
of issues to be addressed: (i) the licensing of this element and (ii) its exact
position and status.

4.1. AccI and the licensing of the null element

Let us assume that the DP-subject in an infinitival clause with AccI syntax
is assigned accusative case by a null prepositional complementizer. We will
call this element 0for. 0Jor as a null element requires licensing. Empty com-
plementizers have been assumed to have an affix status and, therefore, they
are licensed by getting incorporated into the matrix verb (Stowell 1981;
Kayne 1984; Pesetsky 1991; Ormazabal 1995). This kind of incorporation
is considered to be a strictly local operation, so that it is banned when adja-
cency is not respected (Boskovic 1997; Martin 1996). This analysis ex-
plains why ίΛαί-deletion in English is blocked, when (i) the embedded
clause is in subject position (57b), (ii) an adverbial element intervenes be-
tween the main predicate and the embedded clause (57c), or (ii) the embed-
ded clause has been extraposed (57d):

(54) a. John believes that/0,Mary will take the job

b. That/*0,ha, Mary took the job made John happy
c. John believes without a doubt that!* 0,ha, Mary will take the job
d. It didn 7 surprise John that/*0thai Mary took the job

In CIGr., infinitival clauses with AccI syntax can be found in all the envi-
ronments mentioned above:
318 Vassilios Spyropoulos

(55) a. Infinitival clause with AccI in subject position

tro.as algea paskhein ou nemesis (estin)
Trojans-ACC sufferings-ACC suffer-INF.PRS NEG just-NOM (is)
'It's not right for the Trojans to suffer in this way'

b. Extraposition
kindynos oun pollous apollysthai
danger-NOM then many-ACC loose-PASS.lNF.PRS
'Then, there is a danger for many people to get killed'
(X. An. V . l , 6)
c. Intervention of adverbials
synebe: te:s aute.s he:mera:s en te:i sikelia. i
happened-3SG the-GEN this-GEN day-GEN in the-DAT Sicily-DAT
gelo:na nika.n amilkan
Gelon-ACC defeat-INF.PRS Amilcas-ACC
'It happened that on the same day in Sicily Gelon defeated Amilcas'
cHdt. VII. 166)

The data in (55) show that the licensing of CIGr. & f o r is different from that
of English 0,/, a( , in that no adjacency with the matrix verb is required. This
means that the relevant mechanism for the licensing of 0/„,. is not incorpo-
ration/affixation into the matrix verb. Rather, CIGr. 0 / o r is licensed inside
its infinitival clause by a clause internal mechanism.
The answer to the licensing issue lies behind the distribution of AccI
syntax. The data presented in sections 3.2 and 3.3 shows that AccI syntax is
possible with FIs and DIs but not with Als. Als occur after the so-called
'verbs of obligatory coreference' (translation of the Greek term). These are
mainly obligatory subject control verbs such as the following:

(56) a. ouketi edynato bioteuein

no more was.able-3SG live-INF.PRS
'He was not able to live any more'
b. houto: kalein eio.thasin
so call-INF.PRS are.used.to
'They are used to call (it) with this name' {Arist. Po. 1447 b 17)

The main characteristic of these verbs is that they take an infinitival com-
plement that does not have temporal properties, since its event point is the
same or simultaneous with that of the matrix predicate. This is because the
matrix and the embedded predicate in fact express the same event. Based
 The syntax of Classical Greek infinitive 319

on this distributional evidence, we propose that 0f O r is only licensed in en-

vironments where tense is licensed. Thus, 0/„ r is only licensed in FIs and
DIs, because in these infinitival clauses either a full sequence of tenses is
established (FIs: 57) or there is a fixed temporal reference (DIs: 58). Cru-
cially, the facts presented in section 3.3 show that AccI syntax is always
possible in these environments:

(57) omologo: meidian ... lamprotaton gegene:sthai

admit-lSG Meidias-ACC most.glorious-ACC become-INF.PRF
Ί admit that Meidias has become the most glorious'

(58) a. pare.ggeile ta hopla tithesthai tous helle.nas

ordered-3SG the arms-ACC put-MED.INF.PRS the Greeks-ACC
'He gave the order that the Greeks should hand over the arms'
b. kindyneuein ouk eboulonto ... autous
risk-INF.PRS NEG wanted-3PL these-ACC
'They didn't want for them to risk...'

Theoretically speaking, such an analysis is closely associated with the pro-

posals that associate the temporal properties of a clause with the C-domain
(Stowell 1982; Boskovic 1997; Martin 2001; Chomsky 2001; Pesetsky and
Torrego 2001). According to these analyses, the temporal properties of
INFL are selected by a tense feature [T] of C and are licensed by means of
a covert syntactic operation (covert movement - feature checking) that
guarantees their matching. Longobardi (1996) proposes that such a covert
operation licenses the accusative case of the DP-subject in English gerunds:

(59) a. Him sleeping in the bedroom, Mary...

b. While PRO/*him sleeping in the bedroom, Mary...

He claims that English gerunds have temporal properties and proposes that
(covert) V-to-C raising licenses these temporal properties. This (covert)
movement also licenses the accusative case of the gerundival DP-subject.
The existence of an overt complementizer blocks the (covert) movement
and consequently, the licensing of the accusative case. Thus, in example
(59b), a PRO is licensed as the subject of the infinitival clause. Similarly,
Cecchetto and Oniga (2002) associate covert V-to-C movement for tense
checking with AccI syntax in Latin infinitival clauses. They propose that the
infinitive covertly moves to C in order to license its temporal properties,
320 Vassilios Spyropoulos

and by doing so it also licenses a null preposition 0for in C, which in turn

assigns accusative case to the embedded DP-subject of the infinitival
clause.
Returning to CIGr. infinitives, recall that AccI syntax is possible only
when the embedded infinitival exhibits temporal properties and is banned
from Als, which carry no temporal properties. Thus, we propose that FIs and
DIs have a [T] feature in their C-domain, which expresses their temporal
properties and that 0fOr is parasitic to it. Thus, it is licensed only when this
[T] feature is also licensed. However, it should be clarified that we do not
propose that, whenever [T] is licensed in the C-domain, accusative case is
assigned to the subject of the infinitival clause. This would be a notational
variant of the Tense-Hypothesis and it would leave unaccounted for the
possibility of raising and control in the environments where AccI syntax
occurs, contrary to the facts. What we argue for is that 0for can optionally
appear only when a [T] feature is licensed in the C-domain, but not neces-
sarily vice-versa.

4.2. The position of 0for in CIGr.

Let us return to the English examples in (59) repeated here as (60):

(60) a. Him sleeping in the bedroom, Mary...

b. While PROf*him sleeping in the bedroom, Mary...

Longobardi (1996) claims that the ungrammaticality of (60b) indicates that

the existence of a complementizer blocks the V-to-C movement for tense
checking and, consequently, the licensing of accusative case. Given this
observation, 0for and, consequently, AccI syntax will be in complementary
distribution with an overt complementizer. Nevertheless, in CIGr., AccI
syntax is possible with infinitival clauses introduced with overt comple-
mentizers:

(61) a. te:iorge:i houto: khalepe.i ekhre:to es pantas,

the anger-DAT so tough-DAT used-3SG to all-ACC.PL
o.ste me.dena dynasthai prosienai
so.that nobody-ACC.SG be.able-lNF.AOR get.close-lNF.PRS
'He was behaving so badly to everybody, so that nobody could
stay with him'
 The syntax of Classical Greek infinitive 321

b. dareios, prin aikhmalo.tous genesthai

Darius-NOM before captives-ACC become-INF.AOR
tous eretrieas, eneikhe sphi deinon kholon
the Eretrians-ACC had-3SG them-DAT terrible-ACC anger-ACC
'Darius was galled with Eretrians, before they were taken captives'

Such examples are unexpected, if 0{nr merges with the C position. These
examples involve an infinitival clause with AccI syntax, which is intro-
duced by an overt complementizer. The availability of the AccI syntax in
such an environment indicates that 0for occupies another position than the
normal C position. The pattern is explained by assuming the Split-C-
Hypothesis proposed by Rizzi (1997). Rizzi proposes that the functional
category C is decomposed to two basic C heads: The first carries informa-
tion about the clause-typing properties of the clause and is called Cforce.
Proper complementizers are assumed to merge with this position. The sec-
ond carries information about the finiteness of the clause and is called Cfin.
The feature content of this head is therefore relevant to the grammatical
properties of INFL (mood, aspect, tense, agreement). Topic and Focus pro-
jections are also assumed to occur between these two heads:

(62) [cPl Cf orce [TopP/FocP Top/FoC [cP2 Cfi„ [iNFLP INFL ... ]]]]

We propose that 0for merges with Cfin and, thus, it is compatible with other
complementizers that merge with the Cforce position. Significantly, the posi-
tion of 0for and its licensing are now straightforwardly related. We argued
in the previous section that 0for is licensed when a [T] feature is licensed in
the C-domain. Given that the locus of this feature is the Cfin head, it follows
that this head is also the host of the 0f„r\

(63) [CPI Cforce •••[CP2 Cfm[tense] [INFLP ]]]

prin/oste 0for [DP-subject]-ACC

Thus, the existence of a [T] feature, which expresses the temporal properties
of the infinitival clause, activates the C^,, head in the C-domain. Being acti-
vated, Cfln can host 0/ o r , which is responsible for the AccI syntax. In other
words, the relation between [T] and the licensing of 0for is indirect: [T]
licenses the position that 0for occupies and not the element itself.
322 Vassilios Spyropoulos

4.3. 0/or as [Agr]

The discussion of the properties of AccI syntax has led to an analysis that
incorporates the following assumptions:

- AccI syntax in CIGr is optional;

- It involves a syntactic operation that licenses an overt DP-subject in an
infinitival clause by assigning accusative case to it;
- The element that assigns this accusative case is a 0/„ r ;
- This 0f n r merges with the lower Cfin head, when the latter is activated/
licensed by the existence of a [T] feature, which expresses temporal
properties;
- Cfin head is a bundle of features relevant to the inflectional properties of
the clause.

In the Minimalist Program the existence of null elements is questioned on

the grounds of the principles of Inclusiveness and Full-Interpretation
(Chonsky 1993, 1995). Only bundle of features are supposed to exist as
null-elements in syntax. Thus, the status of 0for as an empty preposition
that assigns accusative case is questionable. Based on the theoretical assump-
tions that case is assigned by agreement checking (Chomsky 2000, 2001),
we will propose that this 0for is in fact a bundle of [Agr] nominal features,
which constitutes a part of the feature specification of Cfin.
Let us discuss the properties of the Cfin head. Cfm is assumed to host the
finiteness properties of the clause and to select the appropriate INFL head.
Thus, a Cfin that selects for a finite INFL supposedly carries [T] and [Agr]
features; these elements are proposed to be licensed by means of an operation
Agree with INFL and, by this means, to acquire morphological manifesta-
tion. The assumption that a finite Cf,n carries [Agr] features is supported by
the complementizer-subject agreement data found in languages such as
French and Flemish (Rizzi 1991, Haegeman 1992). In a non-finite clause,
such as the English infinitive, the [T] and [Agr] features of Cfin cannot be
licensed by establishing a relation with INFL, since the latter is negatively
specified for [T] and [AGR], Nevertheless, we saw that in CIGr. infinitival
clauses a [T] feature can be licensed in C^,, when (i) a full sequence of
tenses is established after verba dicendi (FIs), and (ii) a fixed tense value is
selected by the matrix predicate (DIs). Given that [Agr] features can be pa-
rasitic to [T] features (Landau 2004), whenever [T] is licensed in Cfin, [Agr]
may also occur. These [Agr] features cannot be licensed by establishing a
checking relation with INFL, because the latter is negatively specified for
 The syntax of Classical Greek infinitive 323

[Agr], since CIGr. infinitives do not inflect for subject agreement. However,
they can establish a checking relation with the DP-subject of the embedded
clause. As a result of this checking relation, [Agr] is licensed and the DP-
subject of the infinitival clause acquires accusative case. CIGr. infinitival
clauses with AccI syntax are therefore proposed to have the following
structure:

(64) [CP, C force ... [CP2 Cfin[T, Agr] [INFLP DP-subject INFL[+T] ...]]]

We therefore conclude that the infinitival DP-subject in AccI syntax ac-

quires its accusative case by establishing a checking relation with an [Agr]
element in the Cfin head, which is a part of its feature constitution. This pro-
posal has the following theoretical advantages: (i) it subsumes AccI syntax
to the unified agreement system of case assignment, (ii) it does not face any
problems with the Inclusiveness and Full Interpretation conditions, since
[Agr] is a bundle of features which is subject to the principles and opera-
tions that license the constitution of the Cfin and the licensing of its features.

5. Principles and Parameters of the CIGr infinitival syntax

On the basis of (i) the complex data reviewed in the previous sections and
(ii) the analysis of AccI syntax outlined above, we can now formulate a full
account of CIGr. infinitival syntax, which is based on the general principles
that define the feature constitution of the functional heads and their licensing.
More specifically, we argue that CIGr. infinitival clauses have the universal
clause structure, which consists of the following functional heads:

(65) [cPl Cforce ··• [cP2 Cfln [iNFLP INFL ...]]]

The feature constitution of each functional head is subject to language spe-

cific and construction specific parameters. Thus, INFL is suggested to be
specified as [+T, -AGR], because CIGr infinitives inflect for tense but not
for subject-agreement. Such a specification implies that INFL carries a [T]
feature that is interpretable because of its morphological strength, and no
[Agr] features. Because of the lack of [Agr] features, infinitival INFL can-
not establish a checking relation with the infinitival subject, and thus no case
assignment that involves this INFL and its subject takes place. The feature
constitution of INFL is therefore determined by the language specific mor-
324 Vassilios Spyropoulos

phological properties of CIGr. infinitives. The feature constitution of C fin on

the other hand is mainly determined by the specific properties of the con-
structions it participates in. Cfi„ is the locus of finiteness, and therefore carries
grammatical information concerning mood, tense, agreement, etc. We claim
that Cf in can be either positively or negatively specified for tense, depending
on the selectional requirements of the matrix predicate. When Cf in is posi-
tively specified for tense, it carries a [T] feature, which can be either inter-
pretable or not. When non-interpretable, the [T] feature of C fin establishes a
checking relation with INFL and gets its value from the [T] feature of the
INFL. This checking relation results in establishing a full-sequence of
tenses, since CIGr infinitival INFL can be morphologically specified for all
tenses. This is the structure underlying FIs (67d). On the other hand, Cf in
has an interpretable [T] feature, when its temporal properties are fixed be-
cause of the semantic and selectional properties o f the matrix predicate.
This feature need not enter into a relation with the [T] feature of INFL in
order to acquire content; the interpretable value of the [T] feature of INFL
is, therefore, neutralized. This is the structure of the DIs (67b). A C fin that is
negatively specified for tense can be suggested to carry no [T], resulting in
the structure of Als (67a). Positive specification for tense is a prerequisite
for the activation of the C fin head. When properly activated, Cfi„ can also be
specified for agreement features and carry an [Agr] bundle of features.
[Agr] features are considered to be universally non-interpretable (Chomsky
1995, 2000), so they establish a checking relation with a nominal element
in their domain, in order to acquire value. As a result, the nominal element
acquires case, more specifically accusative case. This is the structure of
AccI syntax (67c, e). The general CIGr. infinitival structure is presented in
(66) and (67) summarizes the resulting structures from the combination of
the feature specifications of the Cfi„ and INFL heads (plus and minus marks
represent interpretability and non-interpretability respectively):

(66) cfin ([ [±T], ([-Agr])]) INFL[+T]

(67) a. [CPI C f o r c e ... [CP2 C f i n [,NFLP I N F L [ + T ] ...]]] (Als)

b. [CPI C f o r c e ... [cp2 Cfi n [+T] [INFLP I N F L [ + T ] ...]]] (DIs)
c. [CPI C f o r c e ... [CP2 C f i n [[+T], [-Agr]] [INFLP I N F L [ + T ] ...]]]
(DIs with AccI)
D. [CPI C f o r c e ... [CP2 Cfi n [-T] [INFLP I N F L [ + T ] ...]]] (FIs)
e. [CP, C f o r c e ... [CP2 C f i n [[-T], [-Agr]] [INFLP I N F L [ + T ] ...]]]
(FIs with AccI)
 The syntax of Classical Greek infinitive 325

To be more concrete, the structure of an AI excludes the possibility of hav-

ing an [Agr] feature in C fin , because Cf,„ carries no [T], Given that infiniti-
val INFL cannot assign case to its subject, the latter must either be a PRO
or establish a checking relation with a case assigning head in the matrix
clause, resulting in an overt/covert raising structure. Both control and rais-
ing structures surface as NomI constructions.

(68) a. houto: kalein eio:thasin

so call-INF.PRS are.used.to
'They are used to call (it) with this name'
b. proi eio:thasin [ C P [INFLP PROj INFL[+T] kalein houto:]

DI structure allows for two options: When C fin is not specified for [Agr], a
control structure is established and the subject of the infinitival clause is a
caseless PRO. Depending on the case of the controller, this structure gets
the surface form of NomI, GenI, DatI and AccI constructions (69a,b,c, and
d respectively):

(69) a. ei me: sy boule.i apokrinasthai

if NEG you-NOM want-3SG reply-INF.AOR
'if you don't want to reply...'
a', ei sy; me: boule:i [ CP C fm [+T] [ INFLP PROj INFL[+T]
apokrinasthai]]
b. sou ... deomai akolouthein
you-GEN beg-lSG follow-INF.PRS
Ί beg you to follow...'
b'. proj deomai souj [Cp Cfj„[+T] [iNFLP PROj INFL[+T] akolouthein]]
c. to is allois pa:si pare:ggellen eksoplizesthai
the other-DAT.PL all-DAT.PL was.ordering give.arms-INF.MED.PRS
'he was ordering all the others to take the arms'
c'. proj pare:ggellen [tois allois pa:si]j [ c p Cf,„[+T] [ I N F L P PRO]
INFL[+T] eksoplizesthai]]
d. dedidakhen ... home.ros kai tous allous
has.taught Homer-NOM and the others-ACC
pseude: legein...
lies-ACC say-INF.PRS
'Homer has taught also the others to tell lies'
d'. home:roSj dedidakhen [tous allous]j [CP Cf ln [+T]
[INFLP PROj INFL[+T] legein pseude: ... ]]
326 Vassilios Spyropoulos

When Cfi„ is specified for [Agr], then an overt DP-subject in the accusative
case is licensed as a result of the checking relation established between the
[Agr] and the DP-subject. As a consequence, control is suspended and ob-
viation is established. The structure surfaces as an AccI construction:

(70) a. kindyneuein ouk eboulonto .... autous

risk-INF.PRS NEG wanted-3PL these-ACC
'They didn't want for them to risk...'
a', proj ouk eboulonto [Cp Cf in [[+T], [-Agr]]
[inflp autousj INFL[+T] kindyneuein]]
b. pare.ggeile ta hopla tithesthai tous helle:nas
ordered-3SG the weapons-ACC put-MED.INF.PRS the Greeks-ACC
'He gave the order that the Greeks should hand over the weapons'
b'. proj pare:ggeile [ CP C fin [[+T], [-Agr]] [[nflp [tous helle:nas]j
INFL[+T] tithesthai ta hopla]]

However, control is not always suspended in the presence of an [Agr] fea-

ture in Cfin. In other words, [Agr] feature can license an overt pronominal
which is controlled by a matrix argument:

(71) a. bouloime.n an eme tykhein

want-OPT. 1SG PRT 1-ACC.SG come.across-INF.PRS
ho:n boulomai
which-GEN.PL want-1 SG
Ί would like to get whatever I want' (D. 24, 8)
b. pro, bouloime:n an [Cp C fin [[+T], [-Agr]]
[inflp ernej INFL[+T] tykhein]]

(72) a. bouloime.n an tykhein ho:n boulomai

want-OPT. 1 SG PRT come.across-INF.PRS which-GEN.PL want-lSG
b. proj bouloime:n an [Cp C fin [+T] [iNFlp PRO; INFL[+T] tykhein]]

Example (71) differs minimally from example (72), in that the controlled in-
finitival subject surfaces as an emphatic pronominal in the accusative case,
which is controlled by the matrix subject. Example (71) shows that CIGr re-
sorts to AccI syntax whenever emphasis demands that the infinitival subject
be overt. Theoretically speaking, by being assigned accusative case the em-
bedded subject escapes the case-filter (Vergnaud 1980; Chomsky 1981) and
is licensed so as to surface as an emphatic pronominal. Consider now the
examples (73) and (74), which involve GenI and DatI syntax respectively:
 The syntax of Classical Greek infinitive 327

(73) a. kyrou edeonto o:s prothimotatou pros ton polemon

Cyrus-GEN begged-3SG so most.willing-GEN to the war-ACC
genesthai
become-INF.AOR
'They were begging Cyrus to become most willing for war'
a', proj edeonto [kyrou-GEN]j [PRO, genesthai
o:s prothimotatou-GEN] (X. HG. 1.5, 2)
b. deomai hymo.n ... iatrous genesthai
beg-lSG you-GEN.PL doctor-ACC.PL become-INF.AOR
Ί beg you to remedy...'
b'. deomai [hymo:n-GEN]j [ec, genesthai iatrous-ACC]

(74) a. ο de .mos synekho:re:sen auto.i pro.to.i

the public-NOM permitted-3SG he-DAT first-DAT
graphe:nai
write-INF. PA SS. A OR
'the public permitted him to register first'
a', ο de:mos synekho:re:sen [auto:i-DAT]j [PRO] graphe:nai
proto:i-DAT] (Aeschin. III. 186)
b. oude autois the:baiois epetrepete
not these-DAT Thebans-DAT were.permitting-2PL
autonomous einai (X. HG. VI.3,9)
independent-ACC.PL be-INF.PRS
'Not even Thebans were you permitting to be independent'
b'. epetrepete [thebaiois-DAT]f [ec, einai autonomous-ACC]

These examples were discussed in section 3.1 and it was concluded that the
(b) examples involve a hidden AccI syntax with a pro-ACC subject in the
infinitival clause, as evident from the accusative case of the predicate modi-
fier. It is therefore suggested that [Agr] in C fin can also license a null-
subject pro-ACC in the infinitival clause, which is controlled by a matrix
argument. Noticeably, morphological manifestation of [Agr] and pro is
achieved by means of the agreement relation established between pro and
the predicate modifier. As a consequence this structure is detectable only in
the presence of a nominal modifier that modifies the empty subject of the
infinitival clause, and the structure surfaces as an AccI construction. The
relevant structures are presented in (75):
328 Vassilios Spyropoulos

(75) a. deomai [hymo:n-GEN]i [ CP C fm [[+T], [-Agr]]

[INFLP P R O - A C Q INFL[+T] genesthai iatrous-ACC]]

b. epetrepete [thebaiois-DAT]i [Cp C fin [[+T], [-Agr]]

[INFLP pro-ACQ INFL[+T] einai autonomous-ACC]]

Finally, FIs also allow for two options, depending on the presence vs. ab-
sence of [Agr] in Cfj„. In the absence of [Agr], the subject of the infinitival
clause is assigned no case, allowing for control and raising. 15 These struc-
tures surface normally as NomI constructions (76a-b) and occasionally as
DatI constructions (76c):

(76) a. edoksen auto.i ... ske:ptos pesein

thought-3SG he-DAT thunder-NOM fall-INF.AOR
eis te:n patro.ian oikian
in the family house
'He assumed that a thunder hit his family house' (X. An. III. 1, 11)

a', skeptosj edoksen auto:i [INFLP t, INFL[+T] pesein]

b. edokoun aeton ferein
were.thinking-3PL eagle-ACC bring-lNF.AOR
'They were thinking of bringing an eagle' {Ar. V.15)
b'. pro, edokoun [cp Cf in [-T] [INFLP PROj INFL[+T] ferein aeton]]

c. edokei to:i nikia:i ... ikano:s paraskeuasthai

was.thinking-3SG the Nikias-DAT sufficiently prepare-MED.lNF.PRF
'Nikias was believing that he has been adequately prepared'
c'. proeKp, edokei [to:i nikia:i], [ CP C fin [-T] [ INFLP PRO; INFL[+T]
paraskeuasthai]] (Th. VI.75)

In the presence of [Agr] an agreement relation is established with the infini-

tival subject and the latter is assigned accusative case. The structure surfaces
as an AccI construction:

(77) a. es touton ton kho.ron legetai aphikesthai ton straton

in this the place-ACC say-PASS.3SG arrive-INF.AOR the army-ACC
'It is said that the army has arrived in this place' (Hdt. 111.26)
a'. proexPI legetai [ CP C fln [[-T], [-Agr]] [INFLP [ton straton]
INFL[+T] aphikesthai]]
 The syntax of Classical Greek infinitive 329

b. omologo: meidian ... lamprotaton gegene:sthai

admit-lSG Meidias-ACC glorious-ACC become-INF.PRF
Ί admit that Meidias has become most glorious' (D. 21, 153)
b'. proi omologo: [ C P C f i n [[-T], [-Agr]] [ 1NFLP [meidian] INFL[+T]
gegenesthai lamprotaton]]

The following diagram summarizes the relation between the possible infini-
tival structures with the feature specification of the relevant functional
categories and their surface manifestation in terms of the traditional gram-
mar descriptions:

(78) a. [CP Cfin [INFLP INFL[+T] ...]] -> control NomI

b. [CP Cfi„[+T] [INFLP INFL[+T] ...]]

control -> Noml/Genl/Datl/AccI

C. [cp Cfi„[[+T], [-Agr]] [INFLP INFL[+T] ...]]

obviation -> AccI
control with overt emphatic pronominal/pro Datl/GenI

d. [cpCfi„[-T] [INFLP INFL[+T] ...]]

control NomI
raising NomI

e. [CP2 C f i n [[-T], [-Agr]] [,NFLP INFL[+T] ...]] obviation AccI

6. Conclusions
In this paper, w e investigated the properties of infinitival clauses in CIGr. It
was shown that infinitival syntax in CIGr. is subject to the general principles
of U G that define the clause structure, the licensing of empty elements, and
the licensing of the features of the functional and lexical categories. On the
other hand, certain parameters determine the feature content of the relevant
functional categories of the C- and INFL-domains. More specifically, it
was shown that CIGr. infinitival INFL is specified as [+T, - A G R ] , because
infinitives carry tense but no agreement inflection. As a consequence, CIGr
infinitival INFL does not establish a checking relation with its subject,
which is therefore allowed to establish a checking relation with another
element, or to be a PRO. When not a PRO, the infinitival subject can estab-
lish a checking relation with either an element in the matrix clause (overt/
330 Vassilios Spyropoulos

covert raising) or an element in the C-domain of the infinitival clause. In

the latter option, the subject acquires accusative case and is therefore li-
censed as an overt DP-subject or a pro. We argued that the element, with
which the infinitival subject in accusative establishes a checking relation, is
an [Agr] bundle of features situated in a separate Cfi„ head in the C-domain.
This head hosts the grammatical properties of the clause and is therefore
argued to be activated/licensed by the existence of a [T] feature, which ex-
presses the temporal properties of the clause. Thus, it is concluded that the
interaction of the parameters that define the language specific morphosyn-
tactic structure of CIGr. interact with the general principles of UG that de-
termine the licensing of categories and features, so as to allow for CIGr.
infinitival clauses to have the option of having both null and overt subjects
and to freely exhibit coreference and obviation.

Appendix

Key to the abbreviations of Classical Greek texts

Α., Supp. Aeschylus Tragicus, Supplices
Aeschin. Aeschines Orator
Ar., V. Aristophanes Comicus, Vespae
Arist. Po. Aristoteles Philosophus, Poetica
D. Demosthenes Orator
Hdt. Herodotus Historicus
Isoc. Isocrates Orator
Pi., O. Pindarus Luricus, Olympian Odes
PL, Chrm. Plato Philosophus, Charmides
PL, Prt. Plato Philosophus, Protagoras
S., Ant. Sophocles Tragicus, Antigone
Th. Thucydides Historicus
X., An. Xenophon Historicus, Anabasis
X., Cyr. Xenophon Historicus, Institutio Cyri (Cyropaedia)
X., HG Xenophon Historicus, Historia Graeca (Hellenica)
X., Mem. Xenophon Historicus, Memorabilia
 The syntax of Classical Greek infinitive 331

Notes

1. See Anttila (1989: Part III), Campbell (1998: Chapter 5), Szemerenyi (1999)
and Fox (1995) for overviews of the principles of the comparative method.
2. The term Classical Greek refers to a specific period of the Ancient Greek lan-
guage that spans between the 6 th and the 3 rd centuries BC. It describes the variety
used in the literature production of this period, which is in most cases written
in the Attic and Ionic dialects. Some representative grammatical descriptions of
the syntax of CIGr. can be found in Jannaris (1897), Smyth (1918), Schwyzer
(1950). Our data are drawn from the texts of this period. However, for the ease
of exposition, in some cases we also use elementary textbook examples, either
taken from existing grammar textbooks, or invented by ourselves. In the latter
case, it has been verified that the constructions represented in these examples
occur in the texts of the period. CIGr. examples are transliterated in a broad way,
which relies on the main phonological values represented by the graphemes
and abstracts away from phonological and phonetic details, such as length and
tone, when they are not represented by separate graphemes and are not gram-
matically significant.
3. Despite its significance in determining the morphosemantics of the CIGr. verbal
system, aspect seems to have no effect on the fmiteness of a clause and subse-
quently on the case marking of the subject of the infinitival clause. We will
therefore ignore the role of aspect in the CIGr. verbal system, because it is not
related with the issues of finiteness examined in this paper.
4. Future infinitive is suggested to denote punctual aspect (Schwyzer 1950;
Binnick 1991).
5. Literally, the Latin term is translated as 'verbs of saying and feeling'. In fact,
this category of verbs includes report-verbs and judge-verbs, such as lego: 'to
say', aggello: 'to bring the news', nomizo: 'to decide, to judge', dokeo: 'to be-
lieve, to decide', etc.
6. Sometimes, a future infinitive may be found after verbs that express promise,
will, hope, desire or expectation, as a result of denoting an inherently future
event:
(i) epaggellometha Ariaio.i... eis ton thronon
promise-1 PL Ariaeus-DAT in the throne-ACC
ton basileion kathiein {X. An. 11.1,4)
the royal-ACC sit-iNF.FUT
' W e promise to Ariaeus that he will ascend to the royal throne'
7. The following abbreviations are used in the paper. A O R : aorist, COMP: comple-
mentizer, FEM: feminine, IMP: imperative, INF: infinitive, MED: mediopassive,
M S C : masculine, N E G : negation, NTR: neuter, OPT: optative, PART: participle,
P A S S : passive, PL: plural, PRF: perfect, PRS: present, PRT: particle, PST: past,
SBJ: subjunctive, SG: singular.
332 Vassilios Spyropoulos

8. In the Government & Binding framework government from the matrix predicate
is proposed to assign the accusative case to the DP-subject of the infinitival
clause (Chomsky 1981; Kayne 1984). In the Minimalist Program government
has been dispensed with as a syntactic notion and case assignment has been
subsumed under the more general checking theory (Chomsky 1993). Two dif-
ferent analyses have been proposed in this framework: (i) ECM involves covert
case assignment either by means of covert feature movement to the vP projec-
tion of the matrix verb (Chomsky 1995), or by means of operation Agree es-
tablished for agreement checking between the matrix ν and the embedded DP-
subject (Chomsky 2000). (ii) ECM involves overt movement of the DP-subject
of the infinitival clause to a specifier position in the vP-domain of the matrix
predicate (Lasnik 2001).
9. This agreement pattern is retained to Modern Greek as well (see Spyropoulos
1998, 1999, 2005)
10. To be more specific, Philippaki-Warburton and Catsimali (1991) attribute the
case assigning properties of INFL to the positive specification of a finiteness
feature [fin]. They claim that [fin] in CIGr. infinitives can be positively speci-
fied, and thus assign accusative case, because they inflect for tense. This
analysis can be considered as the INFL variant of the analysis we will propose
in sections 4 and 5.
11. Philippaki-Warburton and Catsimali (1989, 1991) and Sevdali (2003) also pro-
pose that the infinitival INFL cannot assign nominative case because it is speci-
fied as [-AGR]. This constitutes an argument for the hypothesis that nomina-
tive case is, in fact, the result of subject-agreement checking (Chomsky 2000,
2001). Evidence for the agreement-nominative association also comes from
Portuguese inflected infinitives and Modern Greek subjunctives: (a) Portuguese
infinitives may inflect for subject-agreement; when such an infinitive is used
the subject of the infinitival clause appears in nominative case (Raposo 1987).
(b) Modern Greek lacks the category of infinitive and uses the subjunctive in-
stead in complementation. MG subjunctives inflect for subject agreement and
as a result nominative case is always assigned to its subject (Philippaki-
Warburton and Catsimali 1999; Philippaki-Warburton and Spyropoulos 2002,
in prep.).
12. See Raposo (1987).
13. But, see note 6.
14. Here, we follow the widely accepted assumption that the infinitival clause in
raising constructions is deprived of its CP status and it projects only to the
INFLP level (Chomsky 1981, 2001).
15. In raising structures the infinitival clause is deprived of its CP-status (see also
note 14). That CIGr. infinitival clauses may not sometimes project to the level
of CP is also evident from examples such as the following, which involve
ECM:
 The syntax of Classical Greek infinitive 333

(i) kroisos enomizen heo.uton einai

Kroisos was.thinking-3SG himself-ACC be-lNF.PRS
anthro:po:n olbio.taton (Hdt. 1.34)
people-GEN most.blissfull-ACC
'Kroisos believed himself to be the most blissful of all men'
Given that in CIGr clauses anaphors cannot appear in subject position, that the
anaphor heo. uton 'himself appears as the subject of the infinitival clause indi-
cates that the infinitival clause does not constitute the governing category (or
the complete functional complex) for the anaphor licensing. This is indicative
of a construction involving ECM (see also the English translation).

References

Alexiadou, Artemis, Geoffrey Horrocks and Melita Stavrou (eds.)

1999 Studies in Greek syntax. Dordrecht: Kluwer.
Andrews, A. D.
1971 Case agreement of predicate modifiers in Ancient Greek. Linguistic
Inquiry2: 127-151.
Anttila, Raimo
1989 Historical and Comparative Linguistics. Amsterdam: John Benjamins.
Binnick, Robert I.
1991 Time and the Verb: A Guide to Tense and Aspect. Oxford: Oxford
University Press.
Bolkenstein, A.M.
1976 A.c.i. and ut-clauses with verba dicendi in Latin. Glotta 54: 263-291.
Boskovic, Zeljko
1997 The Syntax of Non-finite Complementation. Cambridge, Mass.: The
MIT Press.
Burzio, Luigi
1986 Italian syntax. Dordrecht: Kluwer.
Campbell, Lyle
1998 Historical Linguistics. Edinburgh: Edinburgh University Press.
Chomsky, Noam
1981 Lectures on Government and Binding. Dordrecht: Foris.
1986 Barriers. Cambridge, Mass.: The MIT Press.
1993 A Minimalist Program for Linguistic Theory. In The view from Build-
ing 20, Ken Hale and Samuel J. Kayser (eds.), 1-52. Cambridge,
Mass.: The MIT Press.
1995 The minimalist program. Cambridge Mass.: The MIT Press.
334 Vassilios Spyropoulos

Chomsky, Noam
2000 Minimalist inquiries: The framework. In Step by step: Essays on
minimalist syntax in honor of Howard Lasnik, Roger Martin, David
Michaels and Juan Uriagereka (eds.), 89-115. Cambridge Mass.: The
MIT Press.
2001 Derivation by phase. In Michael Kenstowicz (ed.), 1-52.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the analysis of Latin infinitival clauses for the the-
ory of Case and control. Lingue e Linguaggio 1: 151-189.
Davies, William D. and Stanley Dubinsky (eds.)
2001 Objects and other Subjects: Grammatical Functions, Functional
Categories and Configurationality. Dordrecht: Kluwer.
Drachman, Gaberel, Angeliki Malikouti-Drachman, John Fykias and C. Klidi (eds.)
1997 Greek Linguistics '95: Proceedings of the 2nd International Confer-
ence on Greek Linguistics, Salzburg, September 1995\ Vols. I—II.
Graz: W. Neugebauer Verlag GmbH.
Fox, Anthony
1995 Linguistic Reconstruction: An Introduction to Theory and Method.
Oxford: Oxford University Press.
Haegeman, Liliane
1992 Generative Syntax: Theory and Description. A case study of West
Flemish. Cambridge: Cambridge University Press.
1994 Introduction to Government and Binding Theory. (2nd edition.) Ox-
ford: Blackwell.
Haegeman, Liliane (ed.)
1997 Elements of Grammar: A Handbook in Generative Syntax. Dordrecht:
Kluwer.
Hale, Ken and Samuel J. Keyser (eds.)
1993 The view from Building 20. Cambridge, Mass.: The MIT Press.
Huang, Cheng-Teh James
1982 Logical relations in Chinese and the theory of Grammar. [Unpub-
lished Ph.D. dissertation, MIT.]
Jannaris, Antonius Ν.
1897 An Historical Greek Grammar, chiefly of the Attic dialect. London:
McMillan.
Joseph, Brian D., Geoffrey Horrocks and Irene Philippaki-Warburton (eds.)
1998 Themes in Greek Linguistics II. Amsterdam: John Benjamins.
Kayne, Richard
1984 Connectedness and Binary Branching. Dordrecht: Foris.
Kenstowicz, Michael (ed.)
2001 Ken Hale: A life in language. Cambridge Mass.: The MIT Press.
Landau, Idan
2004 The scale of finiteness and the calculus of control. Natural Language
and Linguistic Theory 22: 811-877.
 The syntax of Classical Greek infinitive 335

Lasnik, Howard
2001 Subjects, objects and the EPP. In Objects and other Subjects: Gram-
matical Functions, Functional Categories and Configurationality,
William D. Davies and Stanley Dubinsky (eds.), 103-121. Dordrecht:
Kluwer.
Longobardi, Giuseppe
1996 The Syntax of N-Raising: A Minimalist Theory. (OTS Working Papers
in Linguistics 96-005.)
Manzini, Rita
1983 On control and control theory. Linguistic Inquiry 14: 421-^146.
Martin, Roger
1996 A Minimalist Theory of PRO and Control. Unpublished Ph.D. dis-
sertation, University of Connecticut.
2001 Null case and the distribution of PRO. Linguistic Inquiry 32: 141-166.
Martin, Roger, David Michaels and Juan Uriagereka (eds.)
2000 Step by step: Essays on minimalist syntax in honor of Howard Lasnik.
Cambridge Mass.: The MIT Press.
Morrell, K. S.
1989 Studies on the Phrase Structure of Early Attic Prose. Unpublished
Ph.D. dissertation, Harvard University.
Ormazabal, Javier
1995 The Syntax of Complementation: On the Connection between Syn-
tactic Structure and Selection. Unpublished Ph.D. dissertation, Uni-
versity of Connecticut.
Pesetsky, David
1991 Zero Syntax, vol. 2: Infinitives. Unpublished MS.
Pesetsky, David and Esther Torrego
2001 T-to-C Movement: Causes and consequences. In Ken Hale: A life in
language, Michael Kenstowicz (ed.), 335-426. Cambridge Mass.:
The MIT Press.
Pillinger, Ο. S.
1980 The accusative and infinitive in Latin: A Refractory Complement
Clause. Journal of Linguistics 16: 55-83.
Philippaki-Warburton, Irene and Georgia Catsimali
1989 'Accusativus-cum-infinitivo' sta arxea elinika (Accusativus-cum-in-
finitivo in Ancient Greek). Studies in Greek Linguistics 10: 89-108.
1991 The theory of control and raising in Ancient Greek and the status of
PRO and NP-trace: Implications for the definition of subject. Unpub-
lished MS.
1997 Control in Ancient Greek. In Greek Linguistics '95: Proceedings of
the 2nd International Conference on Greek Linguistics, Salzburg,
September 1995: Vols. I—II, Gaberel Drachman, Angeliki Malikouti-
Drachman, John Fykias and C. Klidi (eds.), Vol. II, 577-588. Graz:
W. Neugebauer Verlag GmbH.
336 Vassilios Spyropoulos

Philippaki-Warburton, Irene and Georgia Catsimali

1999 On Control in Greek. In Studies in Greek syntax, Artemis Alexiadou,
Geoffrey Horrocks and Melita Stavrou (eds.), 153-168. Dordrecht:
Kluwer.
Philippaki-Warburton, Irene and Vassilios Spyropoulos
2002 Infinitely intriguing: The question of finiteness of Greek subjunc-
tives. Paper presented at the Workshop on Greek Syntax: The Mini-
malist Seduction, University of Reading.
in prep. That's how you get it under control: Greek subjunctives and con-
trolled pro. Unpublished MS.
Raposo, Eduardo
1987 Case theory and Infl-to-Com: The inflected infinitive in European
Portuguese. Linguistic Inquiry 18: 85-109.
Rizzi, Luigi
1990 Relativized Minimality. Cambridge, Mass.: The MIT Press.
1997 The fine structure of the left periphery. In Elements of Grammar: A
Handbook in Generative Syntax, Liliane Haegeman (ed.), 281-337.
Dordrecht: Kluwer.
Schwyzer, Eduard
1950 Griechische Grammatik. Zweiter Band: Syntax und syntaktische
Stilistik. München: C.H. Beck'sche Verlagsbuchhandlung.
Sevdali, Christina
2003 On the properties of Ancient Greek infinitive. Paper presented at the
6'h International Conference on Greek Linguistics, University of
Crete, Rhethymnon.
Smyth, Herbert Weir
1918 Greek Grammar. Harvard University Press.
Spyropoulos, Vassilios
1998 The structure of small clauses in Modern Greek. In Themes in Greek
Linguistics II, Brian D. Joseph, Geoffrey Horrocks and Irene Philip-
paki-Warburton (eds.), 169-196. Amsterdam: John Benjamins.
1999 Agreement relations in Greek. Unpublished Ph.D. dissertation, Uni-
versity of Reading.]
2005 Agreement and multiple case licensing in Greek. In Advances in
Greek Generative Grammar, Melita Stavrou and Arhonto Terzi
(eds.), 15-39. Amsterdam: John Benjamins.
Stavrou, Melita and Arhonto Terzi (eds.)
2005 Advances in Greek Generative Grammar. Amsterdam: John Benja-
mins.
Stowell, Tim
1981 Origins of Phrase Structure. Unpublished Ph.D. dissertation, MIT.
1982 The tense of infinitives. Linguistic Inquiry 13: 561-570.
 The syntax of Classical Greek infinitive 337

Szemerenyi, Oswald J. L.
1999 Introduction to Indo-European Linguistics. Oxford: Oxford Univer-
sity Press.
Tantalou, Niki
2003 Infinitives with overt subjects in Classical Greek. Studies in Greek
Linguistics 23: 358-365.
Latin object and subject infinitive clauses

Lucio Melazzo
The main Harvard Professor Joshua Whatmough, a
rather pompous character, got up, interrupted after
10 minutes or so: "How would you handle..." and
then he mentioned some obscure fact in Latin. I
said I didn't know and tried to go on, but we got
diverted and that's what we talked about for the
rest of the time...
(N. Chomsky, On Nature and Language)

The paper is organized as follows. In section 1 some data are presented

which form the basis of my argument. Section 2 discusses the hypotheses
that some scholars have made about the 'origin' of the accusative-and-
infinitive construction. Section 3 examines the reasons which have been
given for the accusative instead of the nominative as the case of the subject
in infinitive clauses. Section 4 is devoted to the claim that object infinitive
clauses are CPs. In section 5 a new structure for object infinitive clauses is
proposed. Section 6 contains a first attempt to explain the case used for the
subject in object infinitive clauses. Section 7 analyses subject infinitive
clauses, which are held to have the same structure as object infinitive
clauses. Section 8 makes some conclusive remarks and describes the new
directions of research that the author proposes to follow.

1. The data

Should any demonstration be needed, sentences (1) - (3) show that infini-
tive clauses in Latin exhibit a subject in the accusative.1

(1) a. Nam constitui cum quodam hospite

forCONJ agreePF IND ι SG with a-certainABL f o r e i g n e r ^
me esse ilium conventuram
I A C C be PRES .INF he ACC meet FUT PPLE

'For I agreed with a certain foreigner that I would meet him.'

(TER. Hec. 195-196)
340 Lucio Melazzo

b. Teque cognomen non solum Athenis deportasse sed

you A C C -and c o g n o m e n A C C not only Athens A B L bring PF INF but
humanitatem et prudentiam intellego
culture A C C and w i s d o m A C C be-aware P R E S . I N D .i. S G
'and I am aware that y o u brought h o m e from Athens not only a
c o g n o m e n but culture and (practical) w i s d o m too.' (CIC. Sen. 1.1)

(2) a. Romam erat nuntiatum [...] fugisse Antonium

R o m e A C C be [M p F I N D 3 SG announce P F PPLE N N 0 M f!ee P P INF Antony A C C
'It had been announced that Antony had fled to R o m e . '
(CIC. Fam. 11.12.1)

b. Difficile est amicitiam manere,

hard ADJ .N.NOM be PRES .IND.3.SG friendship A C C remain P R E S INF

si a virtute defeceris
if from virtue^ departPF.SUBJ.2.sG
'It is hard for friendship to remain if you have forsaken virtue.'
(CIC. Amic. 11.37)

(3) a. Sed ego id respondeo [...] te dolorem ferre

but InOM rtpRON.N.ACC replypREs .IND.l.SG y ° u A c c grief A C C ucdl
PRES. INF
moderate, nec potuisse non
in-a-dignified-manner A D V , not-and be-able P F [NF not
commoveri, nec fuisse id humanitatis tuae
move
PREs.iNF.PAss not-and be P F J N F it ACC humanity G E N your G E N
'But I reply that you bear grief in a dignified manner, you could
not have been unmoved by it nor was this (part) o f your humanity.'
(CIC. Amic. 2.8)
b. Illud quidem nec faciendum est
thatpRQM.N.NOM indeed not-and to-be-done G D V E N N0M be PRES [ND.3.SG

nec fieri potest, me diutius [...] in

not-and happen P R E S 1 N F be-able P R E S . [ N D . 3 . S G I ACC longer A D V in PREP
tam misera tamque turpi vita
s o A D V wretchedABL.sa s o ^ y - a n d ignominiousABL.so life A B L
commorari
linger PRES .INF
'Indeed this neither must nor can be done: that I (should) linger
(any) longer in a life so wretched and ignominious.'
(CIC. Q. f r . 1.3.6)
 Latin object and subject infinitive clauses 341

The subject is always in the accusative, regardless of the type of the infini-
tive clauses. In conformity with traditional classification these are divided
into object clauses, e.g. those in (1); subject clauses, e.g. those in (2); and
epexegetical clauses, e.g. those in (3).
Besides infinitive clauses, there was another type of clause which may
be held equivalent to the finite ίΛαί-clause of English, as exemplified in (4).

(4) a. Prae.tereo quod [...] earn sibi domum

pass-overpRESjND.i.sc that C0 Nj that ADJ F.ACC herself DAT home ACC
sedemque delegit
residence A C C -and choose PF-IND .3 SG
Ί pass over (the fact) that (she) chose for herself that residence
and home.' (CIC. C/w.188)
b. Accidit perincommode, quod eum nusquam
happen PF jND 3 SG unfortunately ADV that C0NJ he ACC nowhere
vidisti
See
PF.IND.2.SG
'It happened most unfortunately that you did not see him any-
where.' (CIC. Att. 1.17.2)
c. Hoc reprehendis, quod solere
thispRQN.N.ACc blamePRES IND 2.SG thatC0NJ be-in-the-habitPRES 1NF
me dicas de me ipso
Sa
IaCC yPRES.SUBJ.2.SG a b o u t p R E p 'ABL ( m y ) s e l f A B L
gloriosius praedicare
boastfullyadv.compar praisePRESINF
'You blame (me for) this: that you say I am (i.e. that I am, as you
say,) in the habit of speaking about myself too boastfully.'
(CIC. Dom. 93.1-2)
d. Quod omnes Siculi[...J Romae sunt,
that C0NJ alljNDEp M N 0 M PL Sicilian M N 0 M PL Rome GEN be PRES ^ 3 PL
nihilne id valebitl
nothing 1NTERRPTCI it N0M be-worth F U T J N D 3 S G
'That all Sicilians are in Rome, will this be worth nothing?'
(CIC. Ver. 1.1.20)

In this type the clause introduced by quod had its verb in the indicative as
in (4a), (4b) and (4d), or in the subjunctive as in (4c) and its subject, when
phonetically realized, in the nominative. It is precisely the continuation of
342 Lucio Melazzo

this type with the so called declarative quod that is found in the Romance
languages. 2 In these languages, on the other hand, there exist no infinitive
clauses corresponding precisely to the Latin infinitive clauses of the type
exemplified in ( l ) - ( 3 ) . This situation is related to a diachronic change
which affected this area of Latin syntax: quod-clauses of the kind presented
in (4) had begun to supplant infinitive clauses in Latin since the 2 nd century
AD. 3 Bearing in mind this circumstance and considering that both infinitive
clauses and quod-clauses in Latin could occur in the same contexts intro-
duced by the same verbs, school grammarians speak of object quod-c\msQS,
subject quod-c\auses, epexegetical object quod-clauses, and epexegetical
subject quod-clauses. These types are exemplified in 4(a-d). They also de-
fine clauses like those in (4) introduced by quod as explicit (i.e. finite), and
describe the corresponding clauses in the infinitive as implicit (i.e. non-
finite).

2. The origin of the construction

When considered from a typological point of view, the construction that

Latin grammarians called ACCUSATIUUS CUM INFINITIUO (henceforth Acl)
is rarely found in other languages. 4 This kind of construction seems likely to
go back to Indo-European, for it is also attested in Hittite, Vedic Sanskrit,
Iranian, Ancient Greek and Osco-Umbrian, 5 but it was precisely in Latin
that the construction attained its greatest development and found a variety
of usages without equal in other Indo-European languages.
Reams of pages have been written to account for the origin of this con-
struction, but a definitive solution has not been found yet. As Pinkster (1990:
126 and 140) says, however, each and every one of the theories which have
been formulated, reasonable though it may seem, offers somewhat too sim-
plistic an explanation to prove convincing. At present the most accredited
theory, or at least the theory which seems the most trustworthy, posits that
Acl started in dependence on verbs taking a double accusative and then
spread to other areas. 6 This theory has been gradually elaborated in the last
three or four decades.
Thus the case of the subject of Latin object infinitive clauses is tradi-
tionally described as the accusative of the object of the verb in the superor-
dinate clause. According to Calboli (1962: 3), 7 as there was (5) so there ex-
isted also (6) in Latin.
 Latin object and subject infinitive clauses 343

(5) Doceo aliquem grammaticam

teachpREs.nro.i.sG somebody ACC grammar ACC
Ί teach somebody grammar.'

(6) Doceo aliquem sapere

teach PRESIND .i.sG somebody ACC know PRESINF
Ί teach somebody to know.'

In (5) two distinct accusative forms combined with doceo, while in (6)
aliquem was the object of doceo. Thus (7) can be explained in the light of
(6).

(7) Dicunt me uenire

Sa come
yPRES.IND.3.PL W c PRES.INF
'They say that I am coming.'

The pronominal form me in (7) was originally the object of dicunt and then
could be understood as a real subject of the infinitive form uenire. This, on
the other hand, was intrinsically a verbal noun but eventually came to be
equated with finite verb forms with their distinctions of tense and concord.
Bolkenstein (1976) rightly rejected the hypothesis, supported by Ernout-
Thomas (1953: 271), that Acl originated in structures where perception
verbs were involved, for these verbs always admit of two interpretations: the
one has a merely perceptive value, the other is connected with a perceptive-
cognitive purport. 8
Bolkenstein (1976) also excludes that Acl has structures with verbs like
admonere 'to advise' as its starting point. In point of fact, sentence (8) allows
of two interpretations, as (9) and (10) make these clear.

(8) Admonet eum loqui

advise PRES IND 3 SG he A cc speak PRFS INF
'He advises him to speak.'

(9) pro admonet eum [loqui],

(10) pro admonet [eum loqui].

In (9) eum depends directly on the main verb, which is not the case in (10).
On analogous grounds Miller (1974) denies that the origin of Acl can be
found in structures with a verb like docere, as was implicitly suggested by
344 Lucio Melazzo

Ernout-Thomas (1953: 271) and accepted and further developed by Calboli

(1962). Once again structures having verbs of the type of docere followed
by Acl may be in fact interpreted in two different ways. So, for example, in
(6) aliquem may be regarded as either the object of doceo or the subject of
sapere, and in the first of these two alternatives sapere must be taken to
have originally had a 'jussive' function similar to that of an imperative.
Sentences with command verbs (VERBA IUBENDI) as main predicates
also permit of a double interpretation. This is shown by (12) and (13) in
relation to (11).

(11) Iussi eum abire

commandpp.rND.i.so heACC depart PRES . mF

Ί commanded him to go away.'

(12) pro iussi eum [abire],

(13) pro iussi [eum abire].

Whether the structure of (11) is (12) or (13), however, the reference of pro
exercises some control, and the difference is that between direct and indi-
rect commands. Therefore Hahn (1950), Hofmann-Szantyr (1966: 3 5 3 -
354), Miller (1974), Wales (1982), Baldi (1983) and Wanner (1987: 69-70
and 304-305) consider Acl to have originated from sentences like ( l l ) . 9
Nevertheless, both syntactic and semantic reasons prevent such a hypothe-
sis from carrying much conviction: Acl could hardly have developed be-
ginning from sentences which were either susceptible of two structural
analyses connected with two diverse interpretations or containing a main
verb rather different in meaning from either a speech-act or a mental-state
verb. Hence it is not surprising that other proposals should have been made.
These, however, have been neither revolutionary nor more convincing.
Generally this type of explanation conjures up the image of a dog biting its
own tail: it circles round the question without managing to find any way
out that would allow it to set off along a tangent in search of new paths to
tread. Christol (1989) supposes Acl to have developed from a proleptic
construction of Proto-Indo-European: in such a construction an extracted,
i.e. proleptic, item is assigned Case in the position of the sentence where it
has landed. 10 Hettrich (1992: 222, note 4) regards Acl as a substitute for an
earlier construction with a double accusative or with an accusative form
and a predicative adverb, while Pinkster (1990: 126-130) suggests that Acl
should be distinguished from other constructions which similarly exhibit an
 Latin object and subject infinitive clauses 345

infinitive. Acl should be exclusively related to object infinitive clauses like

that in (7): in this sentence me and uenire both constitute one argument of
dicunt and fulfil the grammatical function of complement of this verb,
while in no circumstances can me refer to the addressee of the speech act
that dicunt alludes to.

3. The question of the case of the subject

In all likelihood the various attempts at an explanation of Acl in some dia-

chronic perspective have derived from a persisting shortage of any theoreti-
cal justification for the subject of the object infinitive clauses occurring in
the accusative. Dealing with Acl has been only a matter of giving a good
reason for the accusative instead of the nominative as the case of the sub-
ject of a verb in the infinitive. Many solutions have been put forward; all
are however open to criticism. I shall now discuss most of them.
Bolkenstein (1989: 32) assumes that Latin non-finite clauses may have
their own subject constituent, and that there is a rule which assigns the ac-
cusative case form to these subjects instead of the nominative case form,
which is assigned to subjects of finite verb forms. As is immediately obvious,
however, such a proposal eludes the problem rather than solve it. Baldi
(1983) states that the accusative is due to perceptive-cognitive reasons, the
subject of the infinitive clause occurring in the position of the complement
of the finite verb form. Although it was accepted by Kurzovä (1989: 26),
this hypothesis has at least one weak point. If (7) is taken into account, it is
hard to believe that there the accusative form me is governed by the verb
which the infinitive clause depends on, and Baldi does not give any reason
for this oddity; nor does he elucidate the question of how Acl came to be
extended to clauses including verbs of diverse groups. Analogously, albeit
on different assumptions, Christol (1989) views the assignment of the accu-
sative case to the subject of an object infinitive clause as the transfer of the
specific grammatical function of direct object from the subordinate clause
as a whole to the word naming who or what performs or undergoes the action
stated by the infinitive verb form.
Though not faultless, L a k o f f s (1968) work arouses interest in this ques-
tion among linguists acting within the framework of generative grammar.
Acl is initially accounted for by resorting to a rule called SUBJECT-TO-
OBJECT RAISING. So Saltarelli (1976) and Pepicello (1977 and 1980) have
the subject of the dependent clause rise and come to be the object of the
346 Lucio Melazzo

verb in the superordinate clause. Bolkenstein (1989) and Maraldi (1983),

however, demonstrate that such a rule has never existed and that the accu-
sative marks the subject of the infinitive clause. 11
The approach to the problem has changed since Chomsky (1981). As is
well known, English accusative-and-infinitive constructions like (14) get a
different explanation.

(14) I believe him to be a fool.

Object clauses like him to be a fool in (14) are in fact thought not to have a
left periphery, i.e. a CP-layer, and to be the complement of the upper V.
This is what is usually called ECM (EXCEPTIONAL CASE-MARKING), a no-
tion that Chomsky (1995) deals with again. As a notion, ECM is generally
accepted in generative grammar and continues to stimulate researchers to
further efforts to explain it. See e.g. Lopez's (2001) and Tanaka's (1999)
recent attempts. This hypothesis has been accepted and adapted to shed
light on Latin infinitive clauses even though Chomsky (1988: 140, note 25)
himself recognizes that the case of accusative subjects of infinitives in
Greek or Latin remains more problematic. Chomsky's theory finds an appli-
cation to Latin in Calboli (1983) and Maraldi (1983). According to Calboli
(1983), 12 (15) should have (16) as a structure.

(15) lube hunc abire hinc aliquo

command I M P 2.SG thisPR0N.M.ACc depart PRES . INF hence elsewhere
'Command this (man) to depart from here for (i.e. and go) some-
where else.'

(16) lube hunc [s PRO abire hinc aliquo].

Structure (16) is typical of sentences with an object-control predicate, and

thus, as the object of the main verb, hunc controls the PRO subject of the
infinitive complement. 1 3 On the other hand, (17) is different from (15).

(17) Dico [s te bonum esse]

sa UC
yPRES .IND.l.SG you A C C good ACC PRES.INF
Ί say that you are good.'

In (17) dico is not a control verb. So it can assign the objective case to the
pronoun te in as much as the IP of the esse-clause, not being dominated by
 Latin object and subject infinitive clauses 347

a CP, does not constitute a barrier to government. 14 This hypothesis clashes

with subject infinitive phrases like those in (2a) and (2b). In fact, the main
clauses of these sentences contain predicates which appear unfit for the as-
signment of the accusative case.15 In spite of this, Calboli (1983) proposes
that subject infinitive clauses must have sprung up when the accusative was
already widely used as the case of the subject of clauses with a verb in the
infinitive form. 16
Undoubtedly, some of the arguments that have been referred to take the
absence of a CP into account. Should such a projection be present in the
structure of the clause, it is supposed to constitute a barrier; otherwise the
subject of any complement clause, including those introduced by quod,
would receive its case from the V° of the matrix clause. Conversely,
Cecchetto-Oniga (2002) do not exclude the presence of a CP in Acl clauses
and conclude that these do not entail ECM. They assume that the subject of
infinitive clauses is assigned the accusative case by an empty affixal com-
plementizer which they term 0COMP· They claim that such a complemen-
tizer satisfies its affixal character by having the infinitive moved to its posi-
tion in LF.' 7 It might also be worth noting that Tantalou (2003) suggests
that Acl syntax in Ancient Greek should be described in the same way in as
much as the CP-status of infinitive clauses in the Acl construction of that
language cannot be questioned, and that in this miscellany Spyropoulos fol-
lows a similar line.

4. The presence of a left periphery in the Latin Acl construction

The first question to be answered concerns the presence of left periphery in

infinitive clauses. The case of Italian and Portuguese infinitives is well
known. As Rizzi (1982: 78 ff.) showed, sentences like those in (18) assure
the presence, in formal Italian, of subordinate clauses with an auxiliary in
the infinitive and, as appears in (18c), a subject in the nominative. 18

(18) a. La Corte ha deliberato esser I'imputato

the Court havePRES.IND.3.SG ruleP.PPLE bePRES.INF the defendant
'The Court has ruled the defendant to be innocent.'
innocente
innocent
348 Lucio Melazzo

b. *La Corte ha deliberato l 'imputato esser

the Court havepREs.jND.s.sc ™leP.PPLE the defendant bePRESUNF
innocente
innocent
'The Court has ruled the defendant to be innocent.'
c. La Corte ha deliberato esser tu innocente
the Court h a v e P R E S i [ N D . 3 . S G rulep.pp LE be PRES . INF you NOM innocent
'The Court has ruled you to be innocent.'

Further, from the contrast displayed in (19), which I have taken from Ra-
poso (1987: 87), it is clear that a similar situation is to be found in Portu-
guese:

(19) a. Eu penso / afirmo [terem os deputados

INOM think / maintainPRES J N ι SG havePRES.INF the N0M delegates N0M
D

trabalhado pouco ]
work PPPLE littleADV
Ί think /maintain that the delegates have worked little (i.e. done
little work).'
b. *Eu penso / afirmo [05 deputados terem
INOM think / maintainPRES.i.SG [theNOM delegates NOM haveINF PRES
trabalhado pouco],
workp.PPLE littleADV]
Ί think / maintain that the delegates have worked little.'

Nonetheless, Portuguese differs from Italian in that it has two infinitive

forms, one of which is inflected. As in (18a) and (18b), so in (19a) and
(19b), the auxiliary in the inflected form terem occurs to the left and to the
right respectively of the subject os deputados, its position being the deter-
minant of grammaticality.' 9
To account for cases like those of (18) in Italian and of (19) in Portu-
guese, a head-movement has been hypothesized and denominated AUX-TO-
COMP and INFL-TO-C respectively. As for Italian, the nominative would be
assigned to the subject of the infinitive clause by a special rule which Rizzi
(1982: 87) formalizes as follows:

(20) Assign Nominative Case to NP in the context Aux .

 Latin object and subject infinitive clauses 349

Similarly, Raposo (1987: 92 and 95) puts forward (21) as an explanation of

the Case assignment to the subject of the infinitive clause in Portuguese.

(21) a. In the absence of [+Tense], Infi (or Agr in Infi) is capable of as-
signing (the 20 ) nominative Case to a lexical subject only if it is it-
self specified for Case
b. penso [CP[C[C/T[I Agr],] [ IP os deputados [,· t, VP]]]].

(21b) shows in a precise manner the movement of the inflected infinitive

from I to CT, the head C + the operator T(ense). This movement allows
(Agr in) Infi to be specified for case.
Supposing the situation is as described with regard to the presence of a
CP-layer, the grammatical structure of object infinitive clauses will be con-
sidered susceptible of variation from one language to another, regardless of
their degree of kinship so that languages related to each other can be opposed
to each other or even to their own ancestor. Latin could in fact have infini-
tive clauses devoid of a CP-layer and could therefore differ in this respect
from Italian and Portuguese, which developed from it. In principle there is
no reason why this could not be the case; the facts, however, point to a dif-
ferent conclusion. Let me consider (22) and (23) for the sake of argument.

(22) Haec memini, et uictum

thisN.Acc.PL remember PF IND i SG and vanquishp.PPLE.M.ACC.SG
frustra contendere Thyrsin
in-vain strivePREsj^p Thyrsis ACC
'These things I remember, and that Thyrsis, vanquished, strove in
vain.' (VERG. Eel.7.69)

(23) Plurima quae

no-end-of-things SUPERL N .ACC.PL w hich R E L P R O N N A C C P L
invideant pure adparere tibi rem
prevent PRES SUB j 3 pl clearly appear PRES . !NF you DAT matter ACC
'No end of things which prevent the matter from appearing clearly to
you.' (HOR. 5. 1.2.100)

As instances of an OV language with a non-rigid order, 21 both these pas-

sages deserve the greatest attention in that they supply some decisive data.
Indeed, if (22) and (23) are regarded from the perspective of Kayne's (1994)
antisymmetry of syntax, then uictum frustra contendere and pure adparere
350 Lucio Melazzo

tibi, once they have been moved, must have ended up in some functional
position, and the place where they are found in their respective strings must
be in the structure of split-CP which has been proposed by Rizzi (1997). I
shall not go into details about this matter now. 2 2 1 shall instead quote (24) in
support of what I have just stated.

(24) Negat Piso scire se [...] quicquam

denyPRES.IND.3.sG Piso NOM know PRES . INF himself ACC anything N ACC SG
'Piso denies he knows (of) anything.' (CIC. Phil. 12.3.4)

Accordingly, (24) helps to show that object infinitive clauses in Latin can-
not be regarded as devoid of left periphery. In point of fact, scire has moved
upwards beyond its subject, which is the long anaphor se. The facts speak
for themselves, and (24) has not been chosen at random. It is an instance
taken from a prose work by such a writer as Cicero, who is esteemed an
authority of the utmost importance on Latin language and style. 23 Of course,
many passages could have been cited here from texts either in verse or in
prose, but an instance taken from prose writings has been preferred to con-
vince those who believe that poetry enjoys a great deal of freedom to
change the normal rules of language.
If I am on the right track and Latin object infinitive clauses also have a
left periphery, then credence should be given to L a k o f f s (1968: 5) view that
the structures of two constructions which can be esteemed equivalent to
each other are unlikely to be different. 24 So similar were infinitive clauses
and gwoJ-clauses that the latter was to supplant the former, with which it
had coexisted for a long time. Hence, to speak of infinitive clauses as pos-
sessing the structure of a CP as well does not in any way sound odd.

5. The structure of Latin infinitive clauses

A second question is that of the structure of the construction. To attempt to

settle the matter I propose first taking account of sentences like (3a) and
(3b). The theories of the origin of Acl would have run into some difficulties
with sentences like (3a) if they had been formed on the evidence of real
texts rather than on the basis of fictitious examples. Indeed in (3 a) the main
verb respondeo would have had to assign the accusative case to its own
complement, the first id, which is singular and neuter in gender, and to the
subjects of the infinitive clauses, te and the second id.25 All in all, a con-
 Latin object and subject infinitive clauses 351

ceivable way of solving the problem would be for (3a) to have a structure
like (25).

(25) CP

ego, id, IP

t'/ respondeoi VP

te dolorem... fuisse id
ferre moderate humanitatis tuae

potuisse
non commoveri

As can be easily seen in (25), the first id in (3a) is assumed to have landed
at a place in the highest CP-layer - possibly Spec,Foe - by moving from the
specifier position of a DP which has an &P as its complement. Alternatively,
it could be assumed that it occupies the head position in DP, and the matrix
DP is focussed, with the &P extraposed. The complement of this &P, whose
head is the second nec of (3a), is the CP represented by the string fuisse id
humanitatis tuae, while its specifier position is filled by another &P with
the first nec in (3a) as a head and two CPs occupying the other two posi-
tions: te dolorem... ferre moderate is found in the position of the specifier,
whereas potuisse non commoveri occurs in that of the complement.
352 Lucio Melazzo

It is now worth considering cases like (4c), which undoubtedly permits one
to infer (26) as its structure.

reprehend is,

solere^ me w ty dicas

ι* tw de m e ipso
gloriosius praedicare

When considered carefully, (26) indicates that hoc in (4c) has undergone a
m o v e m e n t parallel to that described for the id of (3a) in (25): the starting-
point is identical and the final destination in the CP-layer might also be the
same. Consistently with both the school grammarians' view mentioned at
the end of section 1 and the observation that the use of a non-finite verb in
a Latin A c l construction is always accompanied by a null complementizer,
a finite verb being only found with an overt complementizer, (25) and (26)
suggest that both Latin object clauses (that with the verb in the infinitive
and that introduced by quod) are base-generated in the complement posi-
tion of a D P with another DP as a specifier 2 6 while, as is shown in (25), the
possibility is not excluded that a &P is inserted into this complement posi-
tion of the former DP. 2 7 Odd though it might appear, my hypothesis that
Latin Acl constructions are CPs embedded within a DP the specifier of
which is filled by another DP is not groundless. After all, this proposal pro-
vides some syntactic foundation for the semantic intuition that Acl con-
 Latin object and subject infinitive clauses 353

structions and declarative gwod-clauses function as noun phrases: object

and subject are terms that usually refer to NPs, and NPs join with a D-head
to project it into a DP exactly as CPs are assumed to do in the constructions
here at issue. Notwithstanding this consideration, someone might find it
difficult to attribute the structures described in (25) and (26) to Latin, a lan-
guage bereft of articles. Latin, however, did have demonstrative adjectives
and pronouns, and demonstrative adjectives and pronouns have been held
to project a full structure in the specifier position of the nominal extended
projection (DP), while the article has been regarded as a functional head
(D). According to Giusti (2001: 159-168), in the process of change which
led from Latin to Romance languages it was a demonstrative adjective that
was reinterpreted as an element of D.28 So, what still needs to be specified
is that the demonstrative pronoun of the full structure in Spec,DP may be
either overt and move as in (3a) and (4c) or covert as in (la), (lb) and (4a).
In this perspective Greek sentences like (27) are also of great interest.

(27) to min ού pote elpeto thumöi \ tethndmen

theART.N.so he ACC not ever s u p p o s e ^ . , ^ ^ mind DAT diePF.INF
'Never did he suppose in his mind that he had died.'
(HOM .11. 17.404-405)

Note that, unlike Latin, Ancient Greek had the definite article, which origi-
nally was a demonstrative pronoun. In the neuter gender, this Greek form
could precede infinitive clauses. It is all grist to the mill from my perspec-
tive, for this further evidence is corroboration of the interpretation I propose.
Indeed, to in (27) has landed at a position in the higher CP-layer as id has
in (3a). In (3a) id is preceded by ego, the topicalized subject of respondeo;
so it can be assumed to occupy the lower Spec,Top of the higher CP-layer.
In (27) to is followed by the enclitic pronoun min in compliance with
Wackernagel's (1892) law. Being the subject of the infinitive clause with
tethndmen as a predicate, the accusative form min must consequently have
moved to its final place via some position in the lower CP-layer. 29 The syn-
tactic status of Wackernagel's Law should be clarified before reaching a
decision on this matter. 30 As to the position from which to in (27) started to
move, consider (28).
354 Lucio Melazzo

(28) Ho Tissaphernes proaisthomenos tä

theART.M.NOM.sG Tissaphernes N OM get-wind-of A oR.ppLE.M.NOM.sG theN.ACC.PL
autä taüta bouleuomenös
same N .Acc.PL that PR0 N.N.ACc.PL consider-the-possibility-ofPRES.PPLE.M A C C . P L
aposthenai pros Kyron ...
g o - o v e r A O R . I N F TOPREP Cyrus AC c
'Having got wind of (their) considering the possibility of going over
to Cyrus, Tissaphernes...' (Xen. An. 1.1.7)

Notice that, independently of the presence of an overt subject in the com-

plement clause aposthenai pros Kyron, tä autä taüta corresponds closely to
id in (3 a), while the three determiners which compose this Greek string can
be imagined as filling three different places in (29).

Hence, Ancient Greek allows us to consider (30) as a structure of infinitive

clauses.

(30) [ DP DP [D° CP]]

Here D°, combining with CP, also has a DP as its specifier. Like id in (3 a),
to in (27) might have been base generated in the DP filling the specifier
position in (30). 31 For the time being, however, I am not able to be any
more precise.
Further clarification of my hypothesis on the structure of Latin infinitive
clauses can be provided by (31)—(33), in that they give more details about
(22)-(24), which were cited in the previous section.
 Latin object and subject infinitive clauses 355

(31) CP

haec, IP

pro, meminit VP

t, t* &P

t, &P

et DP

CP

P R O victum, IP
frustra y contendere.
ty Thyrsin w t'z VP

tx t w t 2

(32) DP

plurima CP

quae, IP

t, invideanti VP

DP VP
Δ
t/ tk DP

CP

pure ; adparere v IP
tibi v
rem,,, tV t, VP

tvv t.v t y
356 Lucio Melazzo

(33) CP

negat, IP

Piso, t, VP

t, t, DP

pro CP

scire* IP

se* f , VP

ti t j quicquam

As (31) shows sufficiently, in (22) haec and uictum frustra contendere

Tyrsin combine with et, the head of &P. Haec has been raised to Spec,Top
in the higher split-CP. On the other hand, in the second coordinated mem-
ber, which has (30) as a structure, PRO uictum, frustra, and contendere have
moved, directly or by successive stages, from where they had been base-
generated, and have landed at three different positions in the lower split-CP.
Except perhaps in the case of PRO uictum, an adjunct that has the structure
of a small clause and once occupied a specifier position within the DP con-
stituting the external argument of contendere, each of these movements that
have been assumed is canonical and can be easily followed in the tree
sketched in (31).
If examined more closely, the part of (32) where the constitution of the
infinitive clause is presented indicates that pure and adparere in (23) have
eventually reached some positions in the split-CP, and the same applies to
tibi, the complement of adparere. However, both the pronominal form
immediately to the right of the verb in the infinitive and some contingent
nuances of meaning, which are of course a matter of interpretation, make it
likely that the positions which frustra and contendere in (22) have risen to
occupy do not correspond to those at which pure and adparere in (23) have
finally arrived.
As far as (24) is concerned, (33) displays a lower CP-layer which is, as it
were, less heavy than those manifested by (31) and (32), for only the infini-
tive scire has been raised to the CP-layer, while like Thyrsin in (22) and rem
in (23), se, the subject of the infinitive clause, does not move out of the IP-
layer. Moreover, (31)-(33) reveal that in (22)-(24) the determiner combining
 Latin object and subject infinitive clauses 357

with the CP of the infinitive clause to be projected into a DP is covert ex-

actly like the determiner which is the head of the projection in the specifier
position of this DP.
Now, however synchronic my analysis may be, some diachronic consid-
erations cannot be avoided. A series of syntactic correlations are generally
admitted to have existed in Proto-Indo-European. When applied to clauses
which were mutually connected, they were expressed by two different words:
one of these derived from the w/z-stem *kwo-/ *kwi- while in the other the
form */- (or previously *to-) could be recognized which constituted a base
for both prenominal and pronominal determiners. These correlations are
assumed to have changed historically as follows:

(34) kwo- ... *to- *kwo- ... */- -> *i- ... *kwo- 0 ... *kwo-
Phase 1 Phase 2 Phase 3 Phase 4

This is the pattern which Haudry (1973) and Lehmann (1974) have called
'diptyque normale' and 'correlative diptych' respectively. In this pattern, in
which the w/z-word originally preceded and eventually followed the deter-
miner, the correlation had a meaning which can be exemplified by expres-
sions like 'the man who/the thing which... this man!this thing...' corre-
sponding to the Latin 'qui!quod... is/id...'.32 The transition from the Indo-
European to the Latin phase has been convincingly stated by Haudry (1973),
who recognizes three different stages in the process of change. Here it is
worth noting that the w/2-form could be converted to a complementizer only
when, as described by Phase 3 in (34), the two members of the diptych had
already inverted their positions. 33 In all likelihood it can be maintained that
the structure of both the Latin constructions here at issue, that with the sub-
ject in the accusative and the verb in the infinitive, and that with quod and a
finite verb form, is directly related to Phase 3 or Phase 4 of (33), the neuter
pronoun, either overt or covert, and the infinitive clause or the quod-clause
being the two members of a correlative diptych.
As things stand, the structure of both the Latin object quod-clause or, as
will be said in section 7, the subject quod-clause and its or their counterpart
with covert C° and verb in the infinitive somewhat resembles that of Eng-
lish relative clauses which has been proposed by Kayne (1994) and is re-
produced here as (35).

(35) [Dp D° CP]

358 Lucio Melazzo

6. The reason why the subject of Latin infinitive clauses is in the

accusative

Although I am not yet able to present a thorough solution of the particular

case assignment characterizing the construction at issue, I am inclined to
believe that the accusative case of the subject of infinitive clauses is li-
censed by something in the CP-layer. This section will be devoted to the
exposition of the reasons that induce me to make such a hypothesis.
To begin with I think it useful to take account of cases like (36) and
(37), which display instances of what is traditionally called the historical
infinitive:

(36) Invidere omnes mihi, \ mordere clanculum; ego

envyPREs.iNF nom.pl Idat backbitepRgs.iNF on-the-sly ADV Inom
non flocci pendere
not strawGEN hold PRESJNF
'All envy me and backbite me on the sly, I do not care a straw.'
(TER. Eu. 410-411)

(37) Ubi minitari Artabanus [...] rector Syriae

ascow threatenPRES INF ArtabanusN0M governorNOM Syria<3EN
Creticus Silanus excitum custodia
CreticusNOM Silanus N0M summonP.ppLE.M.ACc.sG guardABL
circumdat
surroundpREs.iNDj.so
'As Artabanus began to threaten, Creticus Silanus, governor of Syria,
surrounded with the guard (i.e. placed under guard) him, whom he
had summoned.' (TAC. Ann. 2.4)

Undoubtedly, omnes and ego in (36) and Artabanus in (37) are in the
nominative. Both (36) and (37) exhibit verbs in the present infinitive where
the imperfect indicative form is expected. Displaying root sentences, (36)
does not have any overt C°, while ubi is found in the first part of (37),
which is an adjunct with a temporal meaning. On the other hand, as shown
by (38), a subject in the accusative occurs with an exclamatory infinitive.

(38) For as aedibus me eicil

outwards apartments^ IACC throw-out PRFS[NFPASS
'Thrown out of the apartments, am I!' (PL. As. 127)
 Latin object and subject infinitive clauses 359

These facts cannot be left out of consideration in any attempt to explain the
case form in which the subject of the infinitive is found.
If the 'diachronic' interpretations put on Latin Acl and discussed above
in section 2 are left aside and the analysis is not limited to the language of
ancient Rome, then it is clear that the explanations proposed so far by
scholars either accept E C M or bring up the TP or the C P or both of them.
As has been shown above, a CP cannot be disposed of in Acl; hence the
ECM-hypothesis does not hold good, and nothing remains except for me to
search for a solution involving the CP-layer as well. My suggestion profits
from the indisputable fact that a complementizer is connected with certain
properties of the predicate of a clause. Thus, for example, a complementizer
is chosen in accordance with the finite or non-finite character of an embed-
ded clause and, in the case of Latin finite embedded clauses, quod intro-
duces those with a verb in the indicative and ut is selected for those with a
subjunctive. Consequently, a complementizer can be thought to match the
inflectional features of the verb in the embedded clause. To some extent my
hypothesis is in line with those proposals which suggest a connection of the
temporal properties of a clause with the C-domain. According to the analyses
of Stowell (1982), Chomsky (2001), Martin (2001), the temporal properties
of 1° are selected by a tense feature [T] of C° and licensed through a covert
movement that ensures their matching. Furthermore, Rizzi's (1982) and
R a p o s o ' s (1987) explanations of the Italian and Portuguese infinitive
clauses that have been reported above in section 3 go in the same direction,
and Longobardi (1996) assumes that a covert operation of V-to-C raising
licenses the accusative case of the subject in English gerunds.
In the light of so close a connection between C P and IP I shall assume
that the bundle of features which substantiates FIN° in Split-CP includes a
specification for case. This comes into play when the case of the subject
cannot be licensed in the IP-layer in accordance with what is common
knowledge in generative grammar. When this condition exists, the subject
rises overtly or covertly to the higher projection to have its case checked.
The specific case-feature will be [+Acc] if the structure containing a verb
in the infinitive form is a complement clause, and it will be [ + N o m ] other-
wise. I assume that the infinitival CP is a complement to a D head only when
the infinitival clause functions as an argument of the matrix predicate. Hence
root and adjunct infinitival clauses are generated without a subsuming DP.
That is, the FIN° head can have either of the following two sets of features:
360 Lucio Melazzo

(39) FIN 0 (40) FIN°

+ Complement - Complement
+ Acc + Nom

Evidently, the accusative case of the infinitival subject in (31), (32), and
(33) is licensed by the feature specification in (39), whereas the nominative
case of the infinitival subject in (36), and (37) is licensed by (40). (38),
containing an exclamative root infinitive with an accusative subject, ap-
pears at first sight to represent a problem for (40). In fact, we could assume
that an exclamative clause is complement to an implicit speech act verb.
This assumption is supported by the fact that in many languages exclama-
tive clauses begin with the regular subordinating complementizer. Under
this proposal, the FIN° head of (38) is specified as in (39) 34 .
My explanation also receives confirmation from (41).

(41) Quid Alexandrum Pheraeum quo

what ACC Alexander ACC Pheraeus ACC which INTERR .ADJ.M.ABL.SG
animo vixisse arbitramurl
frame-of-mindABL livepp.^ thinkpREs.iND.i.PL
'What (do we think?) - In what frame of mind do we think that
Alexander Pheraeus lived?' (CIC. O f f . 2. 25)

(41) seems to represent a so-called partial wh-movement construction (see

e.g. Dayal 1994; and Horvath 1997), with the matrix wh-pronoun {quid)
functioning as a proleptic pronoun associated with the embedded wh-clause
Alexandrum Pheraeum quo animo vixisse. In the framework proposed in
this paper, quid and the infinitival wh-clause represent the specifier and the
complement of a [+wh] D, respectively. The FIN° head of the infinitival CP
is specified as in (39), assigning accusative case to the subject.

7. Entities must not be multiplied beyond necessity

Now is the time to tackle the problem of subject infinitive clauses, a type
exemplified by (2a) and (2b). From (3b) it is possible to infer that the struc-
ture of these clauses is not different from that of object infinitive clauses
given in (25). If I am on the right track, in fact, in (3b) the DP containing
illud has moved into Spec,IP, whereas the infinitival complement of D has
been extraposed. 35 Moreover, the reason why the subject me carries the
 Latin object and subject infinitive clauses 3 61

accusative case is the same as that which has been given above with respect
to object infinitive clauses. Further, (4d), which exhibits a subject quod-
clause like (4b), can be assumed to have a similar structure to that in (26)
once the quod-clause is admitted to have been moved upwards. Subject in-
finitive clauses also have a CP-layer. This is shown by (42).

(42) Illud[...] me movet, in tanta

thatpRoN.N.NOM Iacc disturb PRES [»ο 3 s g in PREP so-great F ABL SG
militum paucitate abesse tris cohortis
soldiers GEN shortage ABL be-absent PRES INF three ACC cohorts ACC
'That does disturb me - that, in so great a shortage of soldiers, three
cohorts are absent.' (CIC. Fam. 3.6.5)

Here abesse has in fact moved leftwards across the subject tris cohortis
(= tres cohortes) to arrive at a position, in the embedded CP-layer, lower
than that of in tanta militum paucitate, which is topicalized.
In this respect the main clause (2b) might also give further evidence. It
is generally assumed that a D selecting a clause subcategorizes a CP. As a
matter of fact, in such a null-subject language as Latin, a pro associated
with the infinitive clause is evidence of a DP projection above CP. In (2b)
this pro is specified not only as 3 rd person singular but also as neuter in
gender. This can be easily inferred from the form of the adjectival predicate
difficile, which is also neuter in gender and in the nominative case, for in
Latin copular constructions subject and predicate agree in case and, if the
predicate is not a noun, in gender as well. On the other hand, the perfect
participle nuntiatum is also neuter and is also in the nominative in (2a),
where the presence of a pro is to be supposed, and the same specifications
are exhibited by the gerundive form faciendum in (3b). Here, however, to
use traditional terms, an overt illud is to be found, which is a proleptic neuter
pronoun responsible for the epexegetical character of the subject infinitive
clause, and which has, in my opinion, the same syntactic origin as the first
id of (3 a).
Exactly like that of object infinitive clauses, the CP-layer of subject in-
finitive clauses has a FIN° distinguished by [+ Complement] 35 and [+ Acc],
This should suffice to explain the case of their subject. Once both object and
subject clauses are assumed to have one and the same structure, a solution
to a problem which scholars have long puzzled over seems to have been
found. Coherently, in subject infinitive clauses subjects take the accusative
case for the same reason why the subject of object infinitive clauses is in
the accusative.
362 Lucio Melazzo

I could consider my paper as concluded for the time being if the construction
traditionally called NOMINATIUUS CUM INFINITIUO (Ncl) were not to be ac-
counted for. Like (43), (44) displays a construction which evidently in-
volves a subject infinitive clause.

(43) Dicitur eo tempore matrem Pausaniae

say P REs.iND.PASS.3.SG that ADJ . RABL time ABL mother ACC Pausanias GEN
vixisse
livePF.INF
'It is said that Pausanias' mother lived in that period.'
(NEP. Paus. 5.3.1)
(44) Quem quidem [...] per sequi Caesar
w h O R E L .PRON.M.ACC in-truthPTCL pursue PRES INF Caesar N0 M
dicitur
sa
YPRES.IND.PASS.3.SG
'Whom in truth Caesar is said to pursue.' (CIC. Att. 7.23.1)

Unlike in (43), the subject of the infinitive is in the nominative in (44).

Here is a further example modelled on (44):

(45) Qui unus omnium iustissimus

one a
WhOREL.PRON.M.NOM PRON.M.NOM 'WoN.M.GEN.PL the-mOSt-juStNOM
fuisse traditur
bePF .INF s a yPRES.IND.PASS.3.SG
'Who is said to have been the one most just of all men.'
(CIC. Sest. 141.7)

In (44)-(45) the accusative case feature of the FIN° head of the infinitive
clause has not been activated. These sentences are presumably saved in the
following way: the subject of the infinitive clause is raised into the empty
specifier position of the DP subsuming the infinitival CP. When this DP is
moved into the Spec,IP position of the matrix clause, the nominative licen-
sed by the matrix inflection, generally realized on the pronoun in the speci-
fier of the DP, is now realized on the subject of the infinitive raised to this
position. Consider the structure of (44). Assuming that the subject of the
infinitive, Caesar, is in the specifier of the DP occupying the matrix
Spec,IP position, the remnant infinitive clause must have been removed
from under the DP into the matrix CP domain.
 Latin object and subject infinitive clauses 363

Naturally, it can happen also in the case of object infinitive clauses selected
by a transitive verb that the case feature of the FIN° head is not activated,
however, the subject of the infinitive can assume accusative case in the
empty specifier position of the object DP subsuming the infinitive clause.

8. Conclusion

I may have taken my readers along a tortuous road, but I believe I have had
both feet on the ground. Though certainly synchronic, the explanation I
have proposed involves consideration of diachronic features going very far
back in time. On the whole, I think that the implication of diachronic con-
siderations in synchronic analyses can also contribute to our understanding
of language and mind. Much still remains to be done. Firstly, my hypothe-
sis must be checked by taking into account the data systematically extracted
from a representative sample of Latin texts from different periods. Secondly,
the other constructions containing an infinitive must be accounted for.
Thirdly, a definite attempt should be made to obtain cross-linguistic gener-
alizations. As this is a weighty matter, I willingly leave the question open
for the time being, in the hope I may soon return to these matters with fur-
ther data and explicative solutions.
364 Lucio Melazzo

Notes

1. Whenever it could be done, I resorted to actual quotations from Latin texts and
I mentioned them in round brackets. Latin and Greek passages have been for
the most part quoted from the critical editions of both series of the Clarendon
Press (Oxon), but I also made use of Borzsäk (1992), Guillemin (1961), Lau-
rand (1965), Wuilleumier (1961). This will make clear which system of chap-
ter, paragraph, etc., division, or of pagination, is being followed. Latin and
Greek authors and works are printed in the abbreviated forms adopted in the
Oxford Latin and in the Oxford Greek Dictionary respectively.
In producing the glosses on the single words of the Latin and Greek exam-
ples I have presented as evidence I have made use of the following abbrevia-
tions: 1. = first person; 2. = second person; 3. = third person; ABL. = ablative;
ACC. = accusative; ADJ. = adjective; ADV. = adverb; AOR. = aorist; ART. = arti-
cle; COMPAR. = comparative; CONJ. = conjunction; DAT. = dative; F. = femi-
nine; FUT. = future; GDVE. = gerundive; GEN. = genitive; IMP. = imperative;
IMPF. = imperfect; IND. = indicative; INDEF. = indefinite; INF. = infinitive; IN-
TERR. = interrogative; M. = masculine; MD. = middle; N. = neuter; ΝΟΜ. =
nominative; p. = past; PASS. = passive; PF. = perfect; PL. = plural; PPLE. = par-
ticiple; PREP. = preposition; PRES. = present; PRON. = p r o n o u n ; PTCL. = parti-
cle; REL. = relative; SG. = singular; SUBJ. = subjunctive; SUPERL. = superlative.
So that readers who are not acquainted with Latin could easily identify the in-
finitive clauses and the quod-clauses found in the examples that are presented,
I thought it useful to have them printed in bold type. As far as possible, my
English translations aim to be reliable, but I am well aware that the differences
existing between English and Latin might not make things easy for some, even
most readers. These are kindly requested to read the glosses carefully. For my
part I assure them I have consulted English speakers about this matter, but I
must confess that I have not been so diligent and credulous as to look for the
Latin speakers (sic\) to whom my first anonymous reviewer referred.
2. Actually, quod was replaced by quid in Late Latin. As is found in Rohlfs
(1969: 188) the Italian clause-introducing word che derives from this Latin
word quid. In standard Italian che is a multifunctional complementizer and
thus corresponds to French que, which has the same etymological origin, to
English that and German daß. By reason of its characteristics of a generalized
complementizer, che can introduce more or less all the subordinate clauses, in-
cluding relative clauses where it can occur as either a subject or a direct object
in place of the pronoun. For a comprehensive discussion of the matter, cf.
Cinque (1988: 446 and 463-465).
3. See Cuzzolin (1994) for extensive discussion on this.
4. Lehmann (1988) mentions some parallel cases in Quechua.
5. See Miller (1974), Disterheft (1980) and Coleman (1985). It is also worth no-
ticing that, as Heftrich (1992: 223) states, this type of construction can be held
 Latin object and subject infinitive clauses 365

to have existed solely in Ancient Greek and Latin, provided the denomination
Acl is confined only to infinitive clauses depending on both mental-state and
speech-act verbs ( V E R B A SENTIENDI ET DICENDI).
6. The traditional meaning of the double accusative includes constructions corre-
sponding, for example, to that of (i).
(i) Caesar Aeduos frumentum flagitabat
CaesarNOM AeduiACC.PL cornACC demandIMPF.IND.3.S0
'Caesar demanded corn from Aedui.' (CAES. Gal. 1.16.1)
7. Calboli's paper also includes the previous history of the matter here at issue
with a useful bibliography.
8. See also Pinkster (1990: 196-197) and Maraldi (1980).
9. Of course, such an explanation is rejected by Bolkenstein (1976).
10. This appears to be anything but a new idea. See Cuzzolin (1994: 38, note 18).
11. Pillinger (1980) must be counted in as well, but see Comrie's (1981) reply to
his paper.
12. The same scholar asserts that an earlier infinitive clause would have been the
nominative+infinitive construction typical of ergative languages with SOV or-
der; the Acl construction would have developed once Latin had become a
nominative-accusative language.
13. I shall not treat the syntax of sentences containing a verb of command in this
paper, for I think they still need to be analysed thoroughly.
14. As to the structures of (15) and (17), see also Maraldi (1983).
15. On the other hand, a sentence like (3a) does display id as the object of the main
verb respondeo. This makes it unlikely that this verb governs other objects.
16. With reference to the accusative as the default case, see also Calboli (1996).
17. Being assumed to have an affix status, covert complementizers are thought to
be incorporated into the matrix verb (see Stowell 1981; Kayne 1984; Pesetsky
1996; Ormazabal 1995). As a local operation, incorporation is not allowed
when adjacency is not complied with (see Boskovic 1997; Martin 1996). On
that account Latin 0 C O M p differs from English 0tha, in some respects.
18. Apart from essere 'to be' and avere 'to have', also but not necessarily em-
ployed as auxiliaries, the other verbs that can occur in the infinitive in such
subordinate clauses are dovere 'to have to, must', potere 'can, to be able to',
trattarsi 'to be a matter of, to be about', esistere 'to exist', and spettare 'to be
the concern o f . For some first information on these Italian infinitive clauses
with overt subject, cf. Skytte and Salvi (1991: 527-529).
19. As (ii) displays, os deputados in (19a) is also in the nominative.
(ii) Sera dificil [eles aprovarem a proposta].
difficult [heNOM.P1. approve^.,», the proposal]
BeFUT.IND.3..SÜ

'It will be difficult for them to approve the proposal'.

Here in fact the pronoun eles is in the nominative form. Further, according to
Raposo (1987: 86), (iii) is ungrammatical but (iv) is not.
366 Lucio Melazzo

(iii) *Serä dificil [eles aprovar a proposta]

(iv) Sera dificil [PRO aprovar α proposta]
As the Portuguese scholar states, the ungrammaticality of (iv) derives from the
occurrence of the uninflected infinitive form aprovar. The same form is found
in (iv), however, and (iv) is grammatical because it exhibits a covert null-case
item PRO as the understood subject of the infinitive clause.
20. I have inserted the article the.
21. For word order in Latin see Elerick (1994) and Salvi (2001: 304-307).
22. But see the next section.
23. A sentence like (v) is an interesting case in point.
(v) Dionysius noster graviter queritur [...] a discipulis
DionysiusNOM ourNOM SG vehemently complainPR,,:sJND 3 SG from pupilsAB1
abesse tarn diu
be-awayPRES.INF so long
'Our friend Dionysius vehemently complains he has been away from his
pupils for so long.' (CIC. Att. 13.2b)
It is taken from the Clarendon Press (Oxon). Other editions insert se between
abesse and tam, which is entirely plausible as se may have dropped out through
haplology (abesse se > abesse). If this is the case, (v) is similar to (24). Other-
wise it shows that there might not be a great difference between a complement
clause with a subject in the accusative and a complement clause with a PRO
subject.
24. Lakoff refers to the difference she assumes there is between the structure of
(vi) and (vii), which is related to (viii).
(vi) Sentio quod venit
hearPRES iND ] SG thatC0NJ comePRE;s IND 3 SG
Ί hear that he is coming'
(vii) Sentio eum venire.
hearPR]£S IND iSG heACC comePRES [NP
Ί hear him come'
(viii) Sentio eum, sentio venire.
hearPRES IND 1SG heACC, hearPRES ,ND 1SG comePRES |NI.
Ί hear him, I hear (him) coming.'
According to traditional grammar, (vii) should derive from (viii) and thus
originate from two main clauses coordinated asyndetically.
25. See note 14. Allowance should be made for the consideration which has been
expressed in the second paragraph of section 3.
26. As will become clear in section 7, the same suggestion can be made for Latin
subject clauses.
27. &P is the notation of a coordination phrase. As (25) shows, another &P may
fill the specifier position of the former. As for coordination, I refer to Lanzetta-
Melazzo's paper in this miscellany.
 Latin object and subject infinitive clauses 367

28. Even though she does not express herself in exactly the same way, this is the
sense of Giusti's argument, which takes Renzi's (1997: 1 - 1 1 ) proposal into
account. See also Bianchi (1999: 225).
29. As will become plain in the next section, this personal pronoun did not need to
move from where I think it had been base-generated in order to check its accu-
sative case.
30. See note 4 of Lanzetta-Melazzo's paper in this miscellany.
31. If this is the case, then Homeric Greek may show signs of a development of the
demonstrative into the definite article similar to that noticeable in the transition
from Latin to Romance languages. As to does not seem to be the head which
the CP permits to project into a larger DP structure, it can be inferred that at
the age of the composition of the Homeric poems the demonstrative had not
yet completed its transition to being an article - which is what is traditionally
said, after all.
32. The pattern was also realized in Latin with pairs of forms like ' q u o ("in the
same proportion as")...eo ("by that degree")...', ' q u o m o d o ("in the manner in
w h i c h " ) . . . e o d e m modo ("in the same m a n n e r " ) . . . ' , ' q u o a d ("up to the time
that")...usque eo ("up to that point in t i m e " ) . . . ' , the wÄ-stem being as recog-
nizable in the first item of each pair as the /-stem in the second.
33. It is worth noticing that the ancestor of the English complementizer that and its
corresponding forms in the other languages of the Germanic group had *to- as
a base. This could be explained by assuming the original existence of the cor-
relative diptych.
34. This hypothesis vaguely recalls one of Harris's (1976) proposals.
35. In (3b), instead of id, illud occurs, which is the neuter form in the nominative
case of ille, ilia, illud 'that', another determiner with a demonstrative value
analogous to is, ea, id 'he, she, it'.

References

Baldi, Philip
1983 Speech Perception and Grammatical Rules in Latin. In Latin Linguis-
tics and Linguistic Theory, Proceedings of the 1st International Collo-
quium on Latin Linguistics. Amsterdam 1981, Harm Pinkster (ed.),
11-26. Amsterdam/Philadelphia: John Benjamins.
Bianchi, Valentina
1999 Consequences of Antisymmetry. Headed Relative Clauses. Berlin/
N e w York: Mouton de Gruyter.
Bolkenstein, Machtelt
1976 The Relation between Form and Meaning of Latin Subordinate
Clauses Governed by verba dicendi in Latin. Mnemosyne 29: 1 5 5 -
175,268-300.
368 Lucio Melazzo

Bolkenstein, Machtelt
1989 Parameters in the Expression of Embedded Predications. In Subordi-
nation and other Topics in Latin. Proceedings of the Third Colloquium
on Latin Linguistics, Bologna, 1-5 April 1985, Gualtiero Calboli
(ed.), 3-35. Amsterdam/Philadelphia: John Benjamins.
Borzsäk, Stefan
1992 Cornelii Taciti Ab Excessu Divi Augusti Libri I—VI. Stutgardiae et
Lipsiae: In /Edibus B. G. Teubneri.
Boskovii, Zeljko
1997 The Syntax of Non-finite Complementation. Cambridge, Mass.: The
MIT Press.
Calboli, Gualtiero
1962 Studi grammaticali. Bologna: Zanichelli.
1983 The development of Latin (Cases and Infinitive). In Latin Linguistics
and Linguistic Theory, Proceedings of the Γ1 International Collo-
quium on Latin Linguistics. Amsterdam 1981, Harm Pinkster (ed.),
41-57. Amsterdam/Philadelphia: John Benjamins.
1996 The accusative as a default case in Latin. Aspects of Latin. In Papers
from the seventh International Colloquium on Latin Linguistics, Jeru-
salem, April 1993, Hannah Rosen (ed.). Innsbruck: Sonderdruck.
Cecchetto, Carlo and Renato Oniga
2002 Consequences of the Analysis of Latin Infinitival Clauses for the
Theory of Case and Control. Lingue e linguaggio 1: 151-189.
Chomsky, Noam
1981 Lectures on Government and Binding. Dordrecht: Foris.
1988 Lectures on Government and Binding. Dordrecht/Providence: Foris.
1995 The Minimalist Program. Cambridge, Mass.: The MIT Press.
2001 Derivation by Phase. In Ken Hale: A Life in Language, Michael
Kenstowicz (ed.), 1-52. Cambridge Mass.: The MIT Press.
Christol, Alain
1989 Prolepse et syntaxe indo-europeenne. In Subordination and other
Topics in Latin. Proceedings of the Third Colloquium on Latin Lin-
guistics, Bologna, April 1985, Gualtiero Calboli (ed.), 65-89. Am-
sterdam/Philadelphia: John Benjamins.
Cinque, Guglielmo
1988 La frase relativa. In Grande grammatica italiana di consultazione. I.
La frase. I sintagmi nominale e preposizionale, Lorenzo Renzi (ed.),
443-503. Bologna: II Mulino.
1999 Adverbs and Functional Moods: A Cross-linguistic Perspective. New
York/Oxford: Oxford University Press.
Coleman, Robert
1985 The Indo-European Origins and Latin Development of the Accusative
with Infinitive Construction. In Syntaxe et Latin, Christian Touratier
(ed.), 307-342. Aix-en-Provence: Universite de Provence.
 Latin object and subject infinitive clauses 369

Comrie, Bernard
1981 The Theoretical Significance of the Latin Accusative and Infinitive:
A Reply to Pillinger. Journal of Linguistics 17: 345-349.
Cuzzolin, Pierluigi
1994 Sull'origine della costruzione dicere quod: aspetti sintattici e seman-
tic!. Firenze: La Nuova Italia Editrice.
Dayal, Veneeta
1994 Scope marking: cross linguistic variation in indirect dependency. In
Wh-Scope Marking, Ulrich Lutz, Gereon Müller and Arnim von
Stechow (eds.), 157-193. Amsterdam: John Benjamins.
Disterheft, Dorothy
1980 The Syntactic Development of the Infinitive in Indo-European.
Columbus, Ohio: Slavica.
Elerick, Charles
1994 How Latin Word Order Works. In Papers on Grammar, IV, Gualtiero
Calboli (ed.), 99-117. Bologna: CLUEB.
Ernout, Alfred and F r a n c i s Thomas
1953 Syntaxe Latine. 2 e edition revue et augmentee. Paris: Klincksieck.
Giusti, Giuliana
2001 The Birth of a Functional Category: from Latin ILLE to the Ro-
mance Article and Personal Pronoun. In Current Studies in Italian
Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque and
Francesco Salvi (eds.), 157-171. Amsterdam/London/New York/
Oxford/Paris/Shannon/Tokyo: Elsevier.
Guillemin, Anne Marie
1961 Cornelius Nepos. CEuvres. Paris: Societe d'Editions «Les Belles
Lettres».
Hahn, Adelaide
1950 Genesis of the Infinitive with Subject-Accusative. Transactions and
Proceedings of the American Philological Association 31:117-129.
Harris, Zellig S.
1976 Notes du cours de syntaxe. Paris: Editions du Seuil.
Haudry, Jean
1973 Parataxe, hypotaxe et correlation dans la phrase latine. Bulletin de la
Societe linguistique de Paris 68: 147-186.
Heftrich, Heinrich
1992 Die Entstehung des lateinischen und griechischen Acl. Rekonstruktion
und relative Chronologie. In Akten der VIII. Fachtagung der Indo-
germanischen Gesellschaft, Leiden, 31. August-4. September 1987,
Robert Beekes, Alexander Lubotsky and Jos Weitenberg (eds.), 221—
234. Innsbruck: Institut fur Sprachwissenschaft der Universität.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax und Stilistik. München: Beck.
370 Lucio Melazzo

Horvath, Julia
1997 The status of 'w/z-expletives' and the partial movement construction
in Hungarian. Natural Language and Linguistic Theory 15: 509-572.
Kayne, Richard S.
1984 Connectedness and Binary Branching. Dordrecht: Foris.
1994 The Antisymmetry of Syntax. Cambridge, MA, London: The MIT
Press.
Kurzovä, Helena
1989 Subordination in Latin. Eirene 26: 23-26.
Lakoff, Robin Τ.
1968 Abstract Syntax and Latin Complementation. Cambridge, Mass./
London: The MIT Press.
Laurand, Louis
1965 Ciceron. L 'amitie. Paris: Societe d'Editions «Les Belles Lettres».
Lehmann, Christian
1988 Towards a Typology of Clause Linkage. In Clause Combining in
Grammar and Discourse, John Haiman and Sandra A. Thompson
(eds.), 181-225. Amsterdam/Philadelphia: Benjamins.
Lehmann, Winfred Paul
1974 Proto-Indo-European Syntax. Austin: University of Texas Press.
Longobardi, Giuseppe
1996 The Syntax of N-raising: A Minimalist Theory. OTS Working Papers
in Linguistics, 96-105. Utrecht: Utrecht Institut of Linguistics.
Lopez, Louis
2001 On the (Non)complementarity of Θ-Theory and Checking Theory.
Linguistic Inquiry 32: 624-716.
Maraldi, Mirka
1980 The Complement Structure of Perception Verbs in Latin. In Papers
on Grammar, I, Gualtiero Calboli (ed.), 47-79. Bologna: CLUEB.
1983 New Approaches to Accusative Subjects: Case Theory vs. Raising.
In Latin Linguistics and Linguistic Theory. Proceedings of the Is' In-
ternational Colloquium on Latin Linguistics. Amsterdam 1981, Harm
Pinkster (ed.), 167-176. Amsterdam/Philadelphia: John Benjamins.
Martin, Roger
1996 A Minimalist Theory of PRO and Control. University of Connecticut:
Ph.D. dissertation.
2001 Null Case and The Distribution of PRO. Linguistic Inquiry 32: 141-
166.
Miller, Stephen G.
1974 On the History of the Infinitive Complementation in Latin and Greek.
Journal of Indo-European Studies 2: 223-246.
Ormazabal, Javier
1995 The Syntax of Complementation: On the Connection between Syntactic
Structure and Selection. Ph.D. dissertation, University of Connecticut.
 Latin object and subject infinitive clauses 371

Pepicello, William
1977 Raising in Latin. Lingua 42: 209-218.
1980 An integrated Approach to Diachronic Syntax: A Case Study from
Latin. General Linguistics 20: 71-94.
Pesetsky, David
1996 Zero Syntax. Cambridge, Mass.: The MIT Press.
Pillinger, Ο. Stephen
1980 The Accusative and Infinitive in Latin: A Refractory Complement
Clause. Journal of Linguistics 16: 5 5 - 8 3 .
Pinkster, Harm
1990 Latin Linguistics and Semantics. London: Routledge.
Raposo, Eduardo
1987 Case Theory and Infl-to-Comp: The Inflected Infinitive in European
Portuguese. Linguistic Inquiry 18: 85-109.
Rizzi, Luigi
1982 Issues in Italian Syntax. Dordrecht, Cinnaminson: Foris Publications.

Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar: A
Handbook in Generative Syntax, Liliane Haegeman (ed.), 281-337.
Dordrecht: Kluwer.
Rohlfs, Gerhard
1969 Grammatica storica della lingua italiana e dei suoi dialetti. Sintassi
e formazione delle parole. Torino: Einaudi.
Saltarelli, Mario
1976 Theoretical Implications in the Development of Accusativus cum In-
finitivo Constructions. In Current Studies in Romance Linguistics,
Marta Lujän and Fritz Hensey (eds.), 88-99. Washington: George-
town University Press.
Salvi, Giampaolo
2001 The two Sentence structures of Early Romance. In Current Studies in
Italian Syntax. Essays offered to Lorenzo Renzi, Guglielmo Cinque
and Giampaolo Salvi (eds.), 297-312. Amsterdam/London/New
York/Oxford/Paris/Shannon/Tokyo: Elsevier.
Skytte, Gunver and Giampaolo Salvi
1991 Frasi subordinate aH'infinito. In Grande grammatica italiana di con-
sultazione. 11. I sintagmi verbali, aggettivali, avverbiali. La subordi-
nazione, Lorenzo Renzi and Giampaolo Salvi (eds.), 4 8 3 - 4 8 5 , 4 9 7 -
569. Bologna: II Mulino.
Stowell, Tim
1981 Origins of Phrase Structure. Cambridge, Mass.: MIT dissertation.
1982 The Tense of Infinitives. Linguistic Inquiry 13: 561-570.
Tanaka, Hidezaku
1999 Raised Objects and Superiority. Linguistic Inquiry 30:317-325.
372 Lucio Melazzo

Tantalou, Niki
2003 Infinitives with Overt Subjects in Classical Greek. Studies in Greek
L ingu is tics 23: 358-365.
Wackernagel, Jakob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermani-
sche Forschungen 1: 353-436.
Wales, M. L.
1982 Another Look at the Latin Accusative and Infinitive. Lingua 56:
127-152.
Wanner, Dieter
1987 The Development of Romance Clitic Pronouns. From Latin to Old
Romance. Berlin/New York: Mouton de Gruyter.
Wuilleumier, Pierre
1961 Ciceron. Caton l'Ancien (de la vieillessej. Paris: Societe d'Editions
«Les Belles Lettres».
Latin word order in generative perspective:
An explanatory proposal within the sentence
domain*

Chiara Polo

The present paper explores the explicative advantages deriving from ana-
lyzing Latin word order under a generative-inspired perspective. Relying
on a minimal comparison between Latin and Italian morpho-syntactic struc-
tures, Latin is proven to be sensitive to the same set of derivational con-
straints which have been less controversially assumed for Italian under
pragmatically and / or structurally (homogeneously) marked conditions of
production, so much so that its syntax turns out to constitute no exception
to what is predicted by Universal Grammar in terms of a basic, unmarked
word order alternating with a number of derived, marked patterns. The
commonplace of traditional, functionalistic approaches which view Latin as
a free word-order language is thus demonstrated untenable and is conse-
quently abandoned in favor of a principled account of Latin formal codifi-
cation worked out under the Principles and Parameters Theory.

Introduction

The communis opinio about Latin word order is that it is arbitrary, idiosyn-
cratic, non-systematic and immune to any type of explanation: further, the
upholders of the traditional, functionalistic approach (Meillet 1903, and
Panhuis 1982, 1984, among the others) go so far as to draw a straightfor-
ward connection between the seeming 'randomness' of Latin word order
and the fully inflectional status of the language, as if the presence of a pro-
ductive system of case-specifications stood in a cause-effect relationship
with the structural anarchy detected on the surface and the linguist could be
exempted from looking for the ratio concealed underneath it. On this view,
in fact, the paradigmatic information encoded by overt morphology is taken
to play the role of unambiguously signalling the function of the case-marked
constituents: the latter would thus escape the rigid syntactic codification
374 ChiaraPolo

they are subject to in modern, uninflected languages, with the unavoidable

corollary that what looks to us as an unruly syntactic distribution to be ac-
counted for is no longer viewed as an explanandum, but is simply consid-
ered a self-evident, functionally-motivated phenomenon. 1
Yet, things being so, the main aim of the present paper is to readdress
the vexata quaestio of Latin syntactic patterning by showing how it can be
explained and even reduced to a principle of orderliness on condition that it
is analysed in the framework of generative theory and the concept of Prin-
ciples and Parameters involved in Universal Grammar (Chomsky 1981 a &
b, 1986). More precisely, the claim I want to set forth with Ostafin (1986)
is that, in spite of the different morphological type it belongs to, Latin pat-
terns with Italian and other deflected natural languages in sticking to the
logic of a basic, underlying, and unmarked order (which happens to coin-
cide with the string (S)OV), along with a number of exceptions to it that are
similarly motivated on pragmatic or stylistic grounds and seem to entail the
same type of (universal) displacement rules which are less controversially
and independently posited for such languages as Italian. In the spirit of
Lambrecht (1994), Molinelli (1986) and Salvi (2004), the application of a
multi-factorial approach to word order will be shown to help the task of
gaining an insight into the restrictive distributional parameters Latin syntax
is modelled upon. Crucially, comprising Information Structure, besides
morphology, syntax, and, even if less straightforwardly so, prosody among
the interacting modules of grammar leads us to revalue the direct bearing
discourse functions have on the formal shaping of the sentence, and to ap-
preciate the prominent role they play in the activation of licensing strategies
of derivation in specific, marked contexts - with dead languages like Latin
standing out as no exception, provided that the impossibility to access na-
tive speakers' intuitions about their prosodic and pragmatic competence be
opportunely compensated. 2
To carry out this task I have adopted a methodologically rather innova-
tive and highly informative strategy, consisting of a minimal confrontation
between Latin and Italian surface strings as allowed for on sentence level in
a bilingual version of Cena Trimalchionis by Petronius (chapters XXVI-
LXXVIII of the Satyr icon).3 The text has been selected on account of the
version of Latin it hands down to us, which happens to adhere as faithfully
as the artistic elaboration permitted to that sermo vulgaris or familiaris
which modern Romance languages (including Italian) are universally
thought to derive from. 4 This, in its turn, ensures that the empirical data-base
we can concentrate on fully meets the restrictive requirements imposed by a
 Latin word order in generative perspective 375

research in twofold perspective, synchronic and diachronic. The two lin-

guistic objects selected for comparison stand in a relationship of linguistic
and historical continuity; furthermore, the uniformity which underlies the
contextual conditions under which the speech acts are being produced and
the syntactic codification has been realized endows with scientific rigour
the generalizations we come up with about the basic likeliness of syntactic
behaviour of Latin and Italian, at the same time enabling us to view in the
right perspective the relatively few loca of syntactic distance existing be-
tween the two languages (which seem to be confined mainly to their basic
word order - (S)OV vs (S)VO) and to explain them under the Principles
and Parameters Theory in terms of variations in discrete parametric values. 5
On these premises, the newness of the methodology employed, along
with the significance of the results achieved, becomes fully appreciable in
relation to the systematic association of a statistically-grounded analysis of
syntactic organization with an exegetical reconstruction of the informa-
tional imports matching with every single sentence of the text. This has
made possible to preliminarily discriminate between marked and unmarked
patterns both in Italian and, most interestingly, in Latin - owing to the ho-
mogeneity of the communicative conditions framing the bilingual text -
and, only subsequently, to advance hypotheses about the strings invariably
delivering lato sensu neutral imports in the two languages, to finally access
the regularity entailed in the derivation of supposedly deviant, marked con-
structions in Italian as well as in Latin. The greatest merit this integrated
approach must be credited with consists in demonstrating that the seeming
exceptional, idiosyncratic orderings ascribed to Latin can be treated on a
par with Italian marked orders, in that they are punctually detected under
universally uniform conditions of markedness involving constituents posi-
tively-specified for such discourse-related or structurally-rooted features as
[+ Contrastive Focus], [+ Emphasis], [+ Topic], [+ Heavy], etc. which, in
their turn, can be held responsible for the activation of familiar constraints
like left-hand and right-hand focalizations and topicalizations, or structurally-
grounded heavy NP-Shifts and verb frontings. The apparent arbitrariness of
Latin word arrangements is thus unmasked as an optical illusion arising
from impressionistic, superficial interpretations of the data which treat all
word orders alike, with no further concern for the marked or unmarked
reading they systematically pair with and for their respective statistical rate
of occurrence. The conclusions pointed to are thus of great theoretical in-
terest because they yield sound, corroborative evidence in favour of the ex-
plicative adequacy generative theory can be credited with, so much so when
376 Chiara Polo

languages traditionally dismissed as counterevidence to the set of regulari-

ties sketched by Universal Grammar can be demonstrated to indirectly con-
firm its predictive potentialities about the properties of natural languages.
In view of the goals it pursues, the present paper is organized in the fol-
lowing way: section 1 is devoted to a systematic exposition of the statistical
information obtained through an excursus on Latin and Italian surface
strings. The objective it addresses by this stage is just to draw a rough dis-
tinction between high- and low-frequency orders, to subsequently advance
a tentative hypothesis about the pattern which qualifies as dominant, possi-
bly unmarked and basic in the two languages, and proceed to submit it to
sharper, more reliable tests. Section 2 is an introduction to the set of hypo-
theses which have inspired the present approach to Latin word order. Start-
ing from the assumption of Latin as being sensitive to the same derivational
machinery shaping Italian syntax (section 2), I move on to check with the
help of statistics what degree of explicative adequacy we can attain when
mapping Latin and Italian surface strings into marked, and, hence, derived,
and unmarked, and, as such, underlying word orders (Section 2.1). The rest
of part 2 (2.2 through 2.15) focuses on a brief, even if hopefully exhaustive,
overview of the invariable constraints deriving surface syntax: each of them
is introduced jointly with an analysis of the contextual and structural condi-
tions which frame its activation, while a comparative survey is centred on
examples selected both in Latin and in Italian as representatives of the dis-
placement rules previously described. Section 2 ends with an overall dis-
cussion and illustration of the explicative success a pragmatically- and
structurally-motivated interaction between the invariable triggers assumed
under UG can achieve (§ 2.15), provided that Petronian Latin be thought of
in terms of a basic order (S)OV and five generable options of word permu-
tation. In section 3 a shift is proposed from a synchronic to a diachronic
perspective in order to exploit the explanatory potentialities made available
by an empirically-justified hypothesis of "grammars in competition"
(Kroch 1989, Lightfoot 1999, Pintzuk 1991) within the Petronian text: such
a proposal will be proven to have the advantage of providing a natural ex-
planation for the residual percentage of Latin structures (S)VO (± 10%:
47/479) which turn out resistant to the assumed grammar (S)OV, being
systematically paired with neutral contexts of occurrence. Finally, section 4
draws the conclusions of the whole paper and outlines some lines for the
research of the future.
 Latin word order in generative perspective 377

1. A statistical approach to Latin and Italian word order:

A preliminary overview of high and low frequency patterns

Before moving on to the triggers and the displacement operations which

underlie seemingly unmotivated word order variations, I want to show how
the use of rigorous statistical tools can help to gain an overall, rough picture
of Latin syntax which is by itself sufficient to highlight the affinities it pre-
sents with respect to Italian and to proportionally reduce the idiosyncrasies
which have been traditionally ascribed to the former under functional ap-
proaches. Obviously, by this stage, the analysis is developed on a pre-theo-
retical level and it has a purely orienting function; no other goal is pursued
besides the sketching of some provisional generalizations about the domi-
nant, hopefully basic, word order, along with the array of the numerically
less representative surface strings, in tune with Derbyshire's (1977: 592)
prescriptions:
Statistical predominance may give some prima facie evidence about basic
order, but it is not in itself sufficient to establish the order beyond doubt,
since word order movements are very often conditioned by their discourse
environment. What is needed to establish the basic order is an informed hy-
pothesis about the grammar that will account for movements from that basic
order, both structural and stylistic movements.

Statistics remains however a valuable means to figure out the grammatical

skeleton of the language under scrutiny, even if complementary instruments
have to be used to arrive at the otherwise missing refinements.
The rest of the present section contains a systematic presentation of the
statistical information referring to Latin and Italian word orders as they
surface in the fifty-two chapters of Petronius' Cena Trimalchionis. The
four tables below encode the relative rates of occurrence with which the six
logically possible word order patterns involving main sentence constituents
are documented in the bilingual text under scrutiny. In greater detail, the
first two tables relate to the distribution of Subject, Verb, and direct objects
(S/V/O), while the last two focus on the relative ordering of obliquely case-
marked complements (Cs), corresponding to Prepositional Phrases (PPs) in
deflected languages like Italian, with respect to Subjects and Verbs
(S/V/C/PP). 6
378 ChiaraPolo

Table 1. Latin

Order All clauses Main clause Sub. clause

l.(S)VO 188 23.0% 132 25.5% 56 18.5%

2.(S)OV 578 70.0% 345 66.5% 233 77.0%
3.VSO 10 1.2% 10 2.0% 0 0
4.VOS 8 1.0% 7 1.3% 1 0.3%
5.0SV 28 3.5% 18 3.5% 10 3.2%
6.0VS 11 1.3% 7 1.3% 4 1.0%
Totals 823 100% 519 100% 304 100%

Table 2. Italian

Order All clauses Main clause Sub. clause

l.(S)VO 630 77% 382 74.0% 248 81.0%
2.(S)OV 144 17.5% 99 19.0% 45 15.0%
3.VSO 4 0.5% 3 1.0% 1 0.3%
4.VOS 12 1.0% 10 2.0% 2 0.6%
5.0SV 16 2.0% 11 2.0% 5 1.5%
6.0VS 17 2.0% 12 2.0% 5 1.5%
Totals 823 100% 517 100% 306 100%

Actually, the first interesting observational statement we can make on the

ground of the statistics illustrated above is that Latin and Italian converge
on allowing for a dominant word order, which happens to be different in
the two languages - coinciding respectively with (S)OV (or (S)C/PP V) in
Latin, and (S)VO (or (S)V PP) in Italian - but which, meaningfully enough,
presents comparable rates of occurrence in both of them, ranging from the
7 0 % of Latin to the 77% of Italian (or from 79% to 85%, when an
obliquely case-marked or a Prepositional Phrase complement shows up). In
both the languages, furthermore, there appears to be a string which comes
next in the frequency scale, (S)VO in Latin and (S)OV in Italian: the crucial
point to be made is that the percentages it matches with do not show rele-
vant discrepancies in Latin and Italian, as shown by the 2 3 % of the former
to be compared with the 17.5% of the latter (or as proven by the proportion
 Latin word order in generative perspective 379

Table 3. Latin

Order All clauses Main clause Sub. clause

l.(S)VPP/C 43 15% 27 17.0% 16 13.0%
2.(S)PP/CV 219 79% 121 77.0% 98 82.0%
3.VS PP/C 2 1.0% 1 1.0% 1 1.0%
4.V PP/C S 0 0 0 0 0 0
5.PP/C SV 12 4.0% 7 4.0% 5 4.0%
6.PP/C VS 2 1.0% 2 1.0% 0 0
Totals 278 100% 158 100% 120 100%

Table 4. Italian

Order All clauses Main clause Sub. clause

l.(S)V PP 236 85.0% 133 84.1% 103 86.0%
2.(S)PP/CV 26 9.0% 13 8.2% 13 11.0%
3.VS PP 0 0 0 0 0 0
4.V PP S 2 0.7% 2 1.2% 0 0
5.PP SV 4 1.4% 4 2.5% 0 0
6.PP VS 10 4.0% 6 3.7% 4 3.0%
Totals 278 100% 158 100% 120 100%

15%: 9% found when Cs or PPs are involved); crucially, the four remaining
word arrangements logically conceivable - VSO, V O S , O S V , and O V S -
display a recessive frequency rate which amounts in its whole to the 7% of
the total in Latin, not so different from the 5.5% found in Italian (or, once
again, the percentage becomes 6% in both languages when the argument is
represented by C or PP). N o matter how naive and superficial these obser-
vations may appear, they suffice to cast doubt upon the commonplace of
Latin as freely admitting any option of word arrangement: on the contrary,
an employ of statistics as rigorous as possible allows to intuitively ap-
preciate the clear-cut boundary holding between high- and low-frequency
orders and to shed light on the way percentages are distributed among
them, up to the final discovery of the meaningful similarities featuring Latin
and Italian syntactic behavior. Obviously, these conclusions simplistically
380 ChiaraPolo

override a number of complications and problems which nonetheless per-

tain to Latin syntax: among the latter, the phenomenon of syntactic break-
up is one of the points of greater typological distance between Latin and
Italian, possibly one of the hardest challenges an analytical approach to
Latin word order has to face to test its explicative power.7 The next step to
take consists accordingly in showing how the exceptional patterns can be
brought back to a large extent to the derivation strategies commonly as-
sumed by generative theory.

2. Latin word ordering in the light of the (universal) derivational

constraints assumed under generative theory: An introduction

As expected, moving from a purely statistical to a qualitative examination

of the data leads the way to a multi-faceted picture of Latin patterning,
which might at first appear hard to deal with, but which, in the end, does
also provide the key to a sharper comprehension of syntactic organization.
In other words, by targeting a second layer of analysis through an operation
of informational exegesis, the statistical data become a simple basis from
which more revealing information can be drawn, so much so because the
possibility of matching every word pattern with its context of occurrence
makes available a more reliable definition of the basic, underlying order(s),
along with the inventory of the presumably exceptional, marked patterns,
up to the more ambitious objective of substantiating the derivability of the
latter in Latin, under the universal displacement rules appealed to by gen-
erative theory. 8 In greater details, even if schematically, the logical steps
my research program progressed through are the following:

- Minimal assumption: Pretend that Latin does conform to the logic of a

neutral, basic word order, along with a wider range of marked, derived
strings (both of them to be tentatively identified on the ground of an op-
eration of informational exegesis);
- further, let's conform to the working hypothesis that the derivational ma-
chinery it answers to ranges over the same invariable constraints already
posited for other languages (according to the list already hinted at);
- then, basically conforming to the template proposed by Salvi (2004: 2.3)
for Latin sentence structure as illustrated in (1) below, let's see how
many word order patterns we are able to 'justify' and generate in their
context of occurrence by means of this hopefully universal bundle of
displacement rules, and how many, on the contrary, still remain in the
 Latin word order in generative perspective 3 81

number of the idiosyncratic counterexamples to the set of regular deri-

vational strategies invoked;
- finally, let us further examine this residual percentage of cases which
runs unaccounted for under our proposal with the aim of trying to re-
duce them to an articulated explanation under the Principles and Para-
meters Theory either in synchronic or in diachronic perspective.

As already hinted at, in order to be thoroughly understood in its explicative

potentialities, my proposal rests on a preliminary hypothesis about Latin
basic sentence structure. It is this undelying skeleton that every derivational
operation hypothesized will be interpreted on. As a matter of fact, the tree-
diagram quoted in (1) happens to coincide with a slightly modified version
of the template recently proposed by Salvi (2004: 2.3) to account for Latin
surface patterns under his restrictive theory of syntactic codification. It dif-
fers from Salvi's structure in that it instantiates a basic S-O-V order, as
allowed for under the pre-Kaynian framework I adopt, which is on the con-
trary disallowed for under Salvi's analysis inspired by Kayne.

quel0 SpecT" T'

LPeripl Χ" Τ Foe"

0/que SpecFoc" Foe'

X'7°lp FoiT^^Y

0/V/Jue Specl^ Γ

S/(J/X que
382 ChiaraPolo

The tree-diagram consists of the following projections:

- a Complementizer Phrase (CP) which admits two basic options: either
its specifier hosts a relative phrase {qui,...) or its head (C°) is filled by
a lexically realized or an empty complementizer (cum or 0 = ut)f
- proceeding from top to bottom, the Left Periphery is represented by a
recursive projection, namely T(opic) Phrase - TP - with its Specifier
hosting peripheral elements and its head being usually empty;
- a Functional Projection (FP = Focus Phrase) is postulated to provide a
suitable landing site for focalised constituents: the latter would occupy
its specifier position, while its head would remain empty. Whenever
focus phenomena do not involve single constituents, but rather concern
whole events (as in presentational sentences), or the illocutionary and
assertive strength of the sentence (as it happens respectively with jus-
sive and assertive or concessive clauses), the hypothesis set forth is that
the specifier position is occupied by an abstract operator (i.e. respec-
tively, an abstract locative in presentational sentences, a jussive, asser-
tive and concessive operator in the corresponding clauses, to end with a
temporal or causal operator in event clauses, in which case the ante-
position of the verb would serve to signal that the events described in
such clauses causally or temporally follow from those described in the
preceding sentence), while the Focus head harbours the verb, as re-
quired by the semantic activation of the former;
- a number of projections above AgrS" are thought of as necessary for
case-checking purposes. This holds true both under Salvi's (Kayne-
related) proposal that there is a universal phrase structure Specifier-
Head-Complement order - with the implication for languages display-
ing a surface SOV sequence that their dominant order be derived
through a default movement of base-generated, post-verbal constituents
to a preverbal position (SVO - * SOV) where they raise to receive /
check morphological case - and under my proposal, where SOV is the
underlying order, with the entailment that the raising of the object for
case-checking reasons may have no overt reflex on the linear sequenc-
ing of the constituents, as proven by the fact that it keeps showing up in
front of the verb (possibly placed in AgrS);
- a Functional Projection (GP) lower than AgrS (to whose head the verb
usually raises) is meant to host post-verbal focalised constituents in its
Specifier position;
- lower projections contain the positions in which sentence constituents
are supposed to be basically generated (verb, arguments...);
 Latin word order in generative perspective 3 83

- finally, right-dislocated, heavy and epexegetic constituents are located

in the right periphery (they always come after focalised, post-verbal
elements), together with the post-posed element of a discontinuous
constituent whenever its pre-verbal part is a focus. 1 0

The following two summarizing schemes can help understand better the
syntactic architecture just described:
Main clause: Left Periph. | Focus [SOXV] Focus | Right Periph.
Subordinate clause: que Left Periph. | Focus [SOXV] Focus | Right Periph.

As revealed by this brief structural survey, my objective is to push forward

the line of research already inaugurated by Ostafin (1986) and Salvi (2004)
in their pioneering work on Latin syntax. With them I take as a start as re-
strictive a theory as the one assumed under the Principles and Parameters
Theory and try to empirically illustrate that the apparently idiosyncratic
behavior of Latin is amenable to its central notion of syntactic movement
and transformational relationship between sentences (Harris 1960), with
general E c o n o m y conditions being at work to assure that the movement be
somehow licensed under pragmatically, stylistically, or structurally marked
settings. On the other hand, the notion of parameter which feeds the articu-
lation of U G enables us to handle the most macroscopic differences singled
out in synchrony with respect to Italian syntax; finally, the residual cases
(S)VO which go unjustified under the set of regular constraints assumed by
U G can be shown to be reducible to an explanation involving a stage of
grammars in competition.

2.1. A qualitative, comparative insight into Latin and Italian word ordering:
An overview in synchronic perspective

The first step I want to take is to comment upon the statistical data drawn
from a 'qualitative' overview of Latin and Italian word orders: as a matter
of fact, the statistics below open the path to a straightforward appreciation
of the surface strings which can be taken as basic, underlying in the two
languages - owing to the fact that, in the greatest majority of cases, they do
manifest a compatibility exclusively with pragmatically unmarked readings
of their constituents (while no pairing can be posited in the given context
with a Topical, Contrastive Focus, Emphatic reading of them), and no in-
dependent evidence arises for hypothesizing an association with other
structurally-driven marked strategies of derivation (such as Heavy-NP
Shifts, Verb fronting rules, etc.). 11
384 ChiaraPolo

TableS. Latin Table 6. Italian

Unmarked order Chap. 26-78 Unmarked order Chap. 26-78

1.(S)VO 47 10% 1.(S)VO 557 100%
2.(S)OV 432 90% 2.(S)OV 0 0
Totals 479 100% Totals 557 100%
1.(S)V PP/C 18 9% 1. (S)V PP 222 100%
2. (S) PP/C V 189 91% 2. (S) PP V 0 0
Totals 207 100% Totals 222 100%

Table 7. Latin

Word order type (S)VO (S)V PP/C (S)OV (S) PP/C V

1. 'Unmarked' 47 25% 18 42% 432 75% 189 86%
2. 'Marked' 141 75% 25 58% 146 25% 30 14%
Totals 188 100% 43 100% 578 100% 219 100%

Table 8. Italian

Word order type (S)VO (S)V PP (S)OV (S) PP V

1. 'Unmarked' 557 88.5% 222 94% 0 0 0 0
2. 'Marked' 73 11.5% 14 6% 144 100% 26 100%
Totals 630 100.0% 236 100% 144 100% 26 100%

As for tables 5 and 6, they register the word order types that (information-
ally or structurally) unmarked patterns are mapped into. What they unveil is
that Italian approaches the typologically ideal condition in which neutral
imports pair with one and the same string (S)VX in the 100% of cases,
while Latin slightly departs from this desirable condition because it seems
to admit two surface strings - (S)XV and (S)VX - to carry unmarked val-
ues, even if it favors the former over the latter by a significant proportion of
± 9 0 % to ±10%. This datum seems to provide a measure of our failure to
wholly penetrate the regularity we want to reconstruct for and to ascribe
also to Latin syntactic organization. The problem consists, in fact, in the
 Latin word order in generative perspective 385

anomalous option of Latin letting a presumably deviant order ((S)VX) to

carry neutral imports contra the minimal hypothesis that Latin tolerates
exceptions to the basic order only when particular discourse functions are
effected. However, synchronically, the 10% of neutral (S)VX strings might
be dispensable with as an experimental variable, ultimately amenable to the
different methodology we are compelled to use when dead languages are
under study. In other words, the impossibility to access grammaticality
judgments might interfere with the grasp we can obtain of the markedness
featuring anomalous, (S)VX patterns and of the rule which places the D O
in post-verbal position. Or, conversely, this residual, problematic percentage
of (S)VX strings could be modularized into a number of sub-cases, whose
origin might be traced back to sub-regularity principles; or, diachronically,
it might be the case that it mirrors a sociolinguistically marked situation of
grammars in competition according to two hypotheses which are going to
be explored in section 3. As to tables 7 and 8, they are meant to further en-
lighten the relative percentages with which the two numerically dominant
word orders (S)XV and (S)VX recur respectively under marked or unmarked
situational conditions: in Latin the percentages confirm that unmarked read-
ings override marked interpretations when (S)XV is at stake, while the pro-
portion is reversed when (S)VX comes to be considered; in Italian, on the
contrary, the picture is overturned, given that it is (S)VX to predominantly
select unmarked imports, unlike (S)XV which features a positive marking
for specific discourse-based traits of the pre-verbal constituent. The next
task we have to cope with consists at this point in describing and exempli-
fying the movement operations which, through their interaction, yield the
formal structuring of the so-called marked sentences in Latin on a par with
Italian.

2.2. An excursus on the invariable constraints taken to drive surface

derivations: An introduction

At this point, a brief outline seems to be in order of the map of movement

strategies - together with the pragmatic and structural conditions they match
with - which have been proposed as universale in the generative framework
on account of the wide array of natural languages which offer corroborative
evidence to their existence. They have been already widely studied with re-
spect to Italian formal syntax, while their application to Latin word pattern-
ing is still a scarcely explored field. 12 In the light of these considerations,
386 ChiaraPolo

the main goal of what follows consists precisely in showing how they can
be exploited to successfully deal with apparently unmotivated word switch-
ing in Latin: technically, they lie at the base of my choice to classify the
surface patterns I run into in the Latin text with justified, marked orders,
rather than with unjustified ones. The survey will progress from pragmati-
cally-driven constraints to structurally-motivated ones, conforming to the
list proposed below:

a. Right- and left-dislocations;

b. Lefthand and righthand focalizations;
c. Heavy-NP shifts;
d. V-to-C raisings;
e. Verb second-like movements;

f. Movement of the clitic verb forms of sum to the Wackernagel position.

2.3. Left-dislocations
As to the first two displacement rules (listed under a), they can be roughly
likened to each other on the ground of the uniform, pragmatically marked
value the moved constituent seems to bear, even if the way in which the
movement takes place may be different: what cuts across the two construc-
tion types is in fact the discourse-based feature [+ Topic] tagging the con-
stituent, which derives from the old, known, given, and presupposed infor-
mation it delivers. More precisely, as far as Left-dislocations are concerned,
the most substantial features of the construction can be summarised as fol-
lows: first, in such languages as Italian, the left-dislocated Topic signals its
syntactic relationship with the rest of the clause either through its introduc-
tory preposition, which is moved along with it (ex. 2b), or through the re-
sumptive clitic coreferential to it, which becomes obligatory whenever the
Topic coincides with the direct object (Clitic Left Dislocation - CLLD: ex.
3b), or through both of them (ex. 2b); on the contrary, in such languages as
Latin which do not possess a system of (resumptive) clitics, no such ele-
ments ever occur, with the consequence that a diagnostic tool for left-
dislocation processes happens to be unavailable. 13 Second, no uniqueness
requirement holds on structural Topic positions, given that a clause can
contain as many left-dislocated topics as its (topicalizable) arguments and
adjuncts are, so that, for example, a left-dislocated subject may co-occur
with other left-dislocated constituents which stand in-between it and the
verb (ex. 3c): 14
 Latin word order in generative perspective 387

(2) a. Maddalena regalera a Giacomo un orsacchiotto.

Maddalena will-give-3sG to Giacomo a teddybear
'Maddalena will give Giacomo a teddybear'.
b. A Giacomo. Maddalena (six) regalera un orsacchiotto.
to Giacomo, Maddalena (him) will-give-3sc a teddybear

(3) a. Giorgio compra i giornali alia stazione.

Giorgio buys-3sG the newspapers at-the railway station
b. I giornali, Giorgio U compra alia stazione.
the newspapers, Giorgio them buys-3sG at-the railway station
c. Giorgio, i giornali. li compra alia stazione.
Giorgio, the newspapers, them buys-3so at-the railway station
d. Giorgio, i giornali. nott U compra mai.15
Giorgio, the newspapers, not them buys-3sG never

As previously stated, the account of left dislocation structures we appeal to

involves movement, as first proposed in Cinque (1977: 397^411), where a
set of syntactic and interpretative phenomena are shown to be consistent
with a movement analysis of LD. This entails that both in, say, Latin and
Italian, OV sequences are supposed to be generable through the leftward
movement of the Ο whenever the contextual conditions allow for a Topic
reading of the constituent. 16 In line with Rizzi (1998: 112-158) and Zubi-
zarreta (1998: 99-158), I furthermore claim that the object endowed with a
discourse-based functional feature [+ Topic] is required to check it before
spell-out (because it is marked as [ - Interpretable]) through raising from its
base position within VP (where an empty category is left behind) to the
Specifier of a functional projection - commonly identified with a Topic
Phrase, TP (as the one represented in Salvi's template quoted under (1) and
already commented upon) - where it is checked against the corresponding
feature carried by the head T° (while the verb would move to its normal
AgrS position). In cases like these, the resulting structure OV is clearly
viewed as (pragmatically) motivated and derived. This implies that, on the
one hand, it is considered devoid of probative value for the identification of
OV with the basic order in Latin 17 - as we might be trickily led to think if
no attention were paid to the fact that the movement of the [+ Topic]-
marked object is in any case string-vacuous in Latin, unlike its Italian
equivalent, and the resulting string OV superficially coincides with the un-
388 ChiaraPolo

marked order - and, on the other hand, it substantiates my point that word
order variation entails the same dynamics and can be accounted for on
similar grounds in Italian and in Latin.

2.4. A comparative Latin-Italian exemplification of Left-dislocation

Let us briefly consider the following example, quoted from Chapter LXXIV
of Cena Trimalchionis:

S
(4) a. Sed hie, (qui in pergula natus est),
but this-SG.M.NOM, (who is born in the shed outside),
DO(T) V
aedes non somniatur.
mansion-PL.F.ACC not dreams-3so
'But if you are born in the shed outside, you don't aspire to the
mansion' CT, 74, [14] 18
S
b. Ma questa gente qui, (che nasce nelle catapecchie),
but this person here, (who is born in the slums),
DO(T) V
/ palazzi non se U sogna neanche
the mansions not them dream-3PL even
'But the person who is born in the slums don't even dream the
mansion'

Under the contextual conditions found here, the uniform, preverbal position
the DO aedes (or the Italian i palazzi) occupies can be given a pragmatic
explanation: the sentence is pronounced by Trimalchio after his wife Fortu-
nata has attacked him for his shameful behavior with a handsome slave of
his. The point the speaker wants to make is to convince Fortunata along
with his guests that her railing at him was extremely inopportune and un-
justified, because it did not suit her humble origins to give herself airs with
the person who rescued her from the slave-stand and made of a flute-girl
the respectable woman she now is. In the clause under scrutiny and in the
immediately preceding linguistic context, in particular, Trimalchio is re-
minding Fortunata her past wretchedness as opposed to her present welfare
 Latin word order in generative perspective 389

which probably caused the arrogant behaviour she has just displayed. The
crucial remark in his argumentation is that whoever is born in the shed out-
side (as she is) does not even dream of the mansion, while she possesses it
and should content herself with the wealth she has obtained, without stand-
ing up for further rights. Focusing now on the surface string at stake, the
DO aedes can be regarded here as a Topic, bearing old information some-
how inferable from the context, as indirectly substantiated by the fact that it
stands out as the second (easily guessable) item of a contrast which is
drawn between pergula and its opposite aedes, while the rest of the clause
can be qualified as the comment, being an open sentence predicated of the
topic which introduces new information, with the verb in the negative form
qualifying as an informational focus (non somniatur). Thus, the fact that
the context allows for a Topical reading of the DO makes possible to inter-
pret the preposed position the nominal constituent occupies in the clause as
derived through a left-dislocation process (possibly also in Latin, even if in
the latter a reliable test is missing), which finds in the presence in Italian of
a resumptive clitic coreferential with the topical element a further confir-
mation. 19

2.5. Right-dislocations

Turning now to right-dislocations, they can be considered symmetrical to

left-dislocations: pragmatically, no salient differences between the two phe-
nomena can be envisaged, because they both involve topics 'given' in the
universe of discourse or, alternatively, supposed to be present in the
hearer's mind at the time of utterance (deictically or anaphorically
'known': ex. 5a-e); the only discrepancy amounts to the fact that right-
dislocations can only apply to a theme which is assumed as given and, un-
like their leftward counterparts, cannot create a universe of discourse by
positing a theme, even if characterized as expected or given in the discourse
(ex. 6b vs. 6a), except when it occurs in abruptive contexts, and it rests on
the speaker's assumptions (ex. 5a,b):

(5) a. Porto domani, ildolce.

bring-ISG tomorrow, the cake
Ί will bring the cake tomorrow'

b. Lo porto domani, il dolce.

it bring-ISG tomorrow, the cake
390 Chiara Polo

c. GU parli subito pero, a Piva.

him speak-2sG at once however, to Piva
'However you will speak immediately to Piva'
d. Diglielo. a mamma, che la chiamo io.
tell-her-it, to mum, that her call-ISG I
'Tell mum that it is me who will call her'
e. Chi e che le ha fatte, tutte 'ste foto?20
who is that them has-3sG made, all these photographs?
'Who has made all these photos?'

(6) a. II dolce, lo porto io; il vino, h porti tu.

the cake, it bring-ISG I; the wine, it bring-2sG you
'The cake, I will bring it; the wine, you will bring it'
b. *Lo porto io, il dolce; lo porti tu, il vino.
it bring-lsG I, the cake; it bring-2sG you, the wine

The construction is marked by a set of peculiar traits, among which figure

the following ones: the right-dislocated Topic can be resumed by a clitic
coreferential to it, but, differently from left-dislocations, such an element is
always optional, even when the Topic coincides with the direct object {ex.
5a), to the effect that clear diagnostic tools are often absent (also in lan-
guages which dispose of a clitic system like Italian) to establish which con-
stituents undergo movement, and which remain in situ.2i The significance
such a displacement rule has for my analysis consists in the possibility of
analysing some of the SVO strings documented in Cena Trimalchionis as
pragmatically motivated deviations from the unmarked order SOV. This
turns out all the more interesting in view of the widely accepted thesis of
Latin as tolerating a free switching between object and verb - (S)OV vs.
(S)VO - as we move from its two chronological extremes to the period
from Plautus to the early Empire, in that the almost pure type the language
is supposed to instantiate respectively in the earliest, archaic period (OV),
and in late antiquity (VO), gives the way to a superficially inconsistent sys-
tem, manifesting frequent alternation OV / VO of little statistical interest. 22
Yet, to escape from this false picture of Latin syntactic unruliness in its
intermediate stages of development, we can show that a high percentage of
(S)VO orders can be derived through right-dislocation processes triggered
by the Topic reading of the object to the effect that this class of orders is
still consistent with the OV-grammar hypothesized for Latin. In structural
 Latin word order in generative perspective 391

terms, right-dislocation could be worked out under the pre-Kaynian model I

remain faithful to as involving a motivated rightward movement of the
[+Topic]-specified object to an A' position right-adjoined to the VP (Graffi
1994: 272). 2 3

2.6. A comparative Latin-Italian illustration of right-dislocations

Let us focus on the following examples:

V DO(T)
(7) a. Adcognosco, inquit, Cappadocem:...
recognize-lsG - he said - Cappadocian-SGM ACC:...
Ί recognize the Cappadocian, he said' CT, 69, [2]
V DO(T)
b. (Lo) riconosco, disse, il Cappadoce:...
(him) recognize-ISG, he said, the Cappadocian:...
Ί recognize the Cappadocian, he said'
The contextual circumstances under which Trimalchio utters the sentence
under investigation yield a compatibility with a Topic reading of the DO
constituent Cappadocem (or its Italian equivalent il Cappadoce)·. the state-
ment follows Scintilla's (i.e. Habinnas' wife) reaction to Habinnas' enthu-
siastic panegyric of his own slave Massa, whose versatility and talents he
has just listed one after the other. The point for Scintilla is to denounce
Habinnas' partiality in his portrayal of his slave, as proven by the fact that
he did not count his being a pimp among his trickery. The hint she makes at
Massa as a pimp causes Trimalchio to intervene with the observation re-
gistered here, where he reassures the jealous Scintilla by claiming that he -
as well as all those present - is able to recognize the Cappadocian (i.e.
Habinnas, coming from Asia as he did) from this detail he has in common
with his slave of being a pimp and a libidinous lover, which all the rest of the
people coming from this place had a reputation for. Obviously, he takes for
granted that the sum of information carried by the constituent Cappadocem
sounds as given, old, and presupposed in the set context, because it was
available and salient in the previous discourse, where Habinnas and his
slave played a role of protagonists, and the provenance of Habinnas from
Cappadocia (namely, his status of Cappadocian) has become a piece o f
shared information among his dinner guests (the very name Habinnas has
been frequently related to his Cappadocian origins). Trimalchio's utterance
392 Chiara Polo

(Lo) riconosco, il Cappadoce thus fits in with an analysis in terms of right-

dislocation of the given topic.

2.7. Lefthand Focalizations

All the constructions considered so far entail a [+ Topic]-marked constituent;

there are however other marked structures which instantiate the preposing
of an element specified for [+ (Contrastive) Focus]/[+ Emphasis] and which
are differently referred to as "Topicalisations" or "Contrastive Topicalisa-
tions" in the traditional terminology (adopted also in Cinque 1990 and
Benincä, Frison and Salvi 1989), along with Focus (Presupposition) struc-
tures (Rizzi 1998). To avoid terminological ambiguities, I will henceforth
employ the term "(Lefthand) Focalization" for Focus (Presupposition)
structures, to distinguish it from "Topicalisation" in evidence in Topic
(comment) structures. 24 Leftward Focalization affects focal elements which
introduce new information, bearing focal stress, and, in Italian at least, it is
limited to contrastive or emphatic focus - unlike other languages as Hun-
garian where it can also match with non-contrastive, exhaustive focus (i.e.
focus expressing exhaustive identification, as pointed out to me by Katalin
Kiss - personal communication) - while the open sentence expresses con-
textually given information (which the speaker assumes to be shared by the
hearer): 25

(8) a. ANCHE CARLO, dovremmo invitare.

ALSO CARLO, should-1 PL invite
'ALSO CARLO, we should invite'
b. CARLO, dovremmo invitare.
C A R L O , should-1 PL invite
' C A R L O , w e should invite'

c. [Dovremmo invitare (solo) Giorgio],

[we should invite (only) Giorgio] 26

In the examples listed above a contrast is in evidence with the context or

with the inferences suggested by the context (as expressed by the statement
in square brackets): a (contextual) presupposition is corrected, be it the be-
lief that we have to invite solo Giorgio (corrected with anche Carlo·, ex. 8a)
or the inference that we have to invite Giorgio (corrected with Carlo: ex.
 Latin word order in generative perspective 393

8b). Zubizarreta (1998: 6-7) provides a technical account of contrastive

focus by introducing the notion of context statement, defined as a preceding
statement in the discourse which creates the context for contrastive focus.
Contrastive focus is claimed to entail a twofold effect: on the one hand, it
negates the value assigned to a variable in the Assertion Structure of its
context statement (as manifested by the explicit or implicit negative tag
matching with contrastive focus); on the other hand, it introduces an alter-
native value for such a variable. The example she provides to illustrate her
point is the following (where the context statement is enclosed in square
brackets, and contrastive stress is graphically represented by capital letters):

(9) a. John is wearing a RED shirt today (not a blue shirt)

b. [John is wearing a blue shirt today]

The Assertion Structure of (9a) is viewed as a conjunction of two ordered

propositions - John is not wearing a blue shirt today and John is wearing a
red shirt today - resulting in a conjunction of two main assertions:

(10) a. A]: there is an j , such that John is wearing χ

b. A2: it is not the case that the χ (such that John is wearing x) =
a blue shirt & the χ (such that John is wearing x) = a red shirt

She further likens contrastive focus with emphasis, viewing them as two
aspects of one and the same phenomenon.27 As for the syntactic properties
of the construction, the preposed constituent must be accompanied by its
introductory preposition (in Italian), but it is never resumed by a ^referen-
tial pronoun, not even so in clitic-endowed languages like Italian:

(11) A TUO PADRE ho telefonato / *gli ho

TO YOUR FATHER have-ISG telephoned / *him have-lso
telefonato.
telephoned
'TO YOUR FATHER I have telephoned (*him)'

No more than a single constituent can be 'focused', and subjects, when

present, tend to be placed in post-verbal position (ex. 13a vs. 13b):28
394 Chiara Polo

(12) *E' FRANC ESC Α che MARIA ha aiutato.

*is-3sG FRANCESCA that MARIA has-3sa helped
*'It is FRANCESCA that MARIA has helped'

(13) a. IL DOLCE, fa, Maria.

THE CAKE, makes-3sG, Maria
'THE CAKE, Maria makes'
b. *IL DOLCE, Maria fa.
*THE CAKE, Maria makes-3sc

The pertinence of 'focalizations' to our analysis can be fully appreciated in

view of the pragmatic motivation we are trying to document for deviations
from the 'unmarked order' (SOV) in Latin, along with the parallel verifica-
tion of the comparable tolerance admitted by Italian to word order permuta-
tions when marked contextual conditions hold. As already assumed for Topic
phrases, (contrastive) focus or emphatic phrases are supposed to undergo
(Focus) movement, in that they bear a discourse-based functional feature
(be it [+ (Contrastive) Focus] or [+ Emphasis]) which needs checking by
raising to the Specifier of the appropriate functional category indicated as
Focus Phrase in the tree-diagram (1). The order (S)OV (or, alternatively,
OSV, OVS, along with other linear combinations, depending on the landing
site the focused constituent is moved to and the eventual interactions with
other displacement rules affecting different constituent types) thus at times
happens to have an informationally grounded status both in Latin and in
Italian.

2.8. An exemplification of left-hand focalizations

In the light of the properties listed above, a left-hand focalization can be

assumed underneath the derivation of the following example both in Italian
and in Latin, even if in the former it has an overt correlate in the switching
of the basic order VX into XV, while in the latter nothing similar can be
claimed to take place, given that the surface order apparently continues to
conform to the underlying pattern XV {i.e. the movement is string-
vacuous):
 Latin word order in generative perspective 395

P(CF) V
(14) a. Homines sumus, non dei.
men-PL.M.NOM are-lPL, not gods-PL.M.NOM
'We're human, not gods' CT, 75, [1]
P(CF) V
b. UOMINI siamo, non dei.
men are-lPL, not gods
'We're human, not gods'

The utterance is made by Habinnas at the beginning of chapter LXXV to

the aim of persuading Trimalchio to relent and have mercy of his wife
Fortunata after she rebuked him (Chapter LXXIV) for his randy behavior
with a handsome slave: the speaker tries to be as persuasive as he can, and
bases his pleading on the fact that we all make mistakes, precisely because
we are human, and not god-like creatures. Thus, as far as the statement under
examination is concerned, the contextual conditions of production deliver a
Contrastive Focus reading for the preverbal predicative complement of the
subject (P) homines, as further demonstrated by the function it carries out
of negating the value assigned to a variable in the Assertion Structure of its
context statement (as indicated by the explicit negative tag associated with
Contrastive Focus: non dei), and of introducing an alternative value for
such a variable, homines (Zubizarreta 1998: 6 - 7 ) . In turn, the possibility to
associate such a marked value with the Ρ complement seems to have a uni-
form bearing on the formal structuring of the clause across the two lan-
guages, with the fronted constituent ending up in the specifier of a structur-
ally higher Focus Projection as obtained through a 'Lefthand Focalization'
(derivational) process. 29

2.9. Righthand focalizations

In addition to a "lefthand focalization", we will assume with Rizzi (1998)

and Salvi (2004) the existence of a lower, "righthand focalization", charac-
terized by focal stress (contrastive, even if not necessarily so) on an (in
situ) element. Its existence is well-documented in Italian, as illustrated by
the following example taken from Rizzi (1998: 119), which makes it plau-
sible to tentatively hypothesize a lower, post-verbal, righthand (contrastive
or emphatic) focus position also in Latin, as exhaustively argued for by
Salvi (2004: 1.2.1.3):
396 ChiaraPolo

(15) Ho letto ILTUOLIBRO, (nonilsuo).

have-1 SG read THE YOUR BOOK, (not the his)
Ί have read Y O U R BOOK, (not his)'

From an explicative viewpoint, it is worth noticing that, while there is gen-

eral consensus among scholars about an analysis in terms of movement of
'Lefthand Focalizations', the question remains basically open and more
controversial as regards 'Righthand Focalizations', owing to the fact that
they can be interpreted either in terms of an in situ strategy of focalization
(as seems to be the case in Italian, which is supposed to be underlyingly
SVO, even if the movement of the focal object to the specifier of a Focus
Projection seems to be implied anyhow, although at LF - see Rizzi 1998:
119, and the bibliography here referred to) or on the ground of a remnant,
leftward movement of the rest of the clause to let the rightmost constituent
receive focal prominence. As for Latin, under the hypothesis I set forth of
an underlying (S)OV order, the derivational alternative to be posited for
Righthand Focalizations would be a leftward movement of the verb across
the focalized object - harbored, in its turn, in Spec,GP, the lower Focus
Projection represented in the tree-diagram (1) - to the head of the AgrS
Projection. 30 However, abstracting away for the time being from the uncer-
tainties involved by such a choice, the legitimacy of assuming 'right-hand
Focalisation' operations both in Italian and in Latin seems to be defendable,
to the effect that analogous derivational devices can be figured out to ac-
count for marked, (S)VO patterns across the two languages .

2.10. An exemplification of Righthand focalizations

Consistently, the hypothesis of a righthand focalization enables us to han-

dle cases of S V O order like the following, where the D O alios surfaces
clause-finally, in clear deviation from its standard, pre-verbal ordering: in
this case, the anomalous distribution of the object cannot be taken to pro-
vide substantive evidence in favour of a free switching between (S)OV and
(S)VO in Latin, given that it seems to be pragmatically motivated on the
base of the Contrastive Focus interpretation the object is exposed to in the
given context:
 Latin word order in generative perspective 397

S V DO(CF)
(16) a. (Bellumpomum), qui rideatur alios;...
(a fine specimen), wh0-3sG.N0M laughs-3sG others-PL.M.ACC
'he is a fine specimen to be laughing at others;...' CT, 57, [3]
S V DO(CF)
b. Beltomo, che deride gli ALTRI (non se stesso!);...
(a fine specimen), who laughs-3sG the others (not himself);...
'he is a fine specimen to be laughing at others, rather than at him-
self;...'

As a matter of fact, the sentence under investigation is part of a violent ver-

bal attack the speaker Ermerotes addresses to his interlocutor Ascyltus as a
consequence of the irreverent, disrespectful attitude the latter has displayed
to him. Actually, under similar contextual conditions, the Contrastive Focus
reading hypothesized for the DO constituent alios or gli altri becomes a
function of the sharp contrast the speaker draws between the impolite
Ascyltus, on the one hand, and all the dinner guests and their host Trimal-
chio, on the other: the definition of the former as a 'runaway fly-by-night,
not worth his own piss' plainly reveals the gap assumed between him and
the respectable people Trimalchio is surrounded by. The fact that such a
worm all the same dares to laugh at others instead of looking at himself and
his own wretchedness and pathetic ridiculousness seems to be underlined as
an inexplicable datum by Ermerotes' words in the sentence under scrutiny;
this makes plausible to assume a Contrastive Focus interpretation for alios
in the set context, as illustrated more transparently by the Italian example,
where the (implicit) negative tag non se stesso has been added to exemplify
how the dynamics involved by Contrastive Focus contexts are at work.
Consistently with the definition given by Zubizarreta (1998: 6 - 7 ) , contras-
tive focus has the twofold effect of negating the value assigned to a variable
in the Assertion Structure of its context statement and of introducing an
alternative value for such a variable, gli altri.

2.11. Heavy N P Shift

Further exceptions to a supposedly unmarked (S)OV order are represented

by objects standing out as 'heavy' or internally structured and complex
constituents; the widespread tendency is to collocate them post-verbally,
thus giving rise to (S)VO strings. The typology of constituents which are
398 Chiara Polo

frequently found displaced from their pre-verbal position comprises long

constituents (such as long lists, enumerations of direct and indirect object
NPs, quotations, long attributes and appositions), and objects which func-
tion as antecedents for a relative clause immediately following them. 31
Whenever an (S)VO string is licensed under similar conditions, it seems
legitimate to invoke a heavy NP shift, which results in the displacement of
'heavy constituents' toward the end of the clause. 32 The claim remains valid
both if we adhere to a more traditional formulation of Heavy NP Shift in
terms of 'a local, strictly bounded movement that shifts stressed NPs from
their base-generated position rightward to adjoin to the right periphery of
the clause, generating a syntactic variable structure' (Pintzuk 1991: 160),
and if we take side with Kayne (1994: 71-76) and his prohibition against
rightward adjunction and reinterpret it as involving leftward, not rightward,
(remnant) movement, as an instantiation of scrambling (as found in Ger-
man) to a higher specifier position past the "heavy NP":

(17) a John gave to Bill all his old linguistics books.

b. John gave [[to Bill]; [X° [[all his old linguistics books] [Y° [e]i...

The analysis worked out by Kayne (1994) for sentences like (17) is that the
PP to Bill originates in a small clause whose specifier position hosts all his
old linguistics books and, from there, it undergoes leftward movement in-
dependently of the verb past the object to a still higher specifier position
with the result that the surface string derived has the 'heavy constituent' in
clause final position. When applied to Latin, the assumption of a Heavy NP
Shift strategy provides a straightforward explanation for such instances of
( S ) V O H as the following, where the post-verbal D O ostrea pectinesque
crucially functions as antecedent for the (non-defining) relative clause fol-
lowing in apposition quae collecta puer lance circumtulif.

(18) (Constemati nos insolentia ebriorum intentavimus oculos in

proeliantes),
(taken aback at this outrageous behaviour of the drunken pair, we
stared at the two disputants),
V DO(N+H)
notavimusque ostrea pectinesque
noticed-and-LPL oysters-PL.N.ACC scallops-and-PL.M.ACC
e gastris labentia, quae
from pitcher-PL.N.ABL tumbled down, which- PL.N.ACC
 Latin word order in generative perspective 399

collecta puer lance circumtulit.

gathered-PL.N.ACC slave-sc.M.NOM dish-SG.F.ABL carried-round-3sG
'Taken aback at this outrageous behaviour of the drunken pair, we
stared at the two disputants, and we noticed oysters and scallops tum-
bling from the pitchers; a slave gathered them, and carried them
round in a dish'. CT, 70, [6]

2.12. Structural constraints affecting the verb: V-to-C in imperative clauses

A deviating, VO surface string may be derived not only when the internal
argument carries pragmatically or structurally marked values which trigger
displacement phenomena, but also when there are good reasons to assume
that it is the verb that undergoes movement: this seems to be the case for
verbs of imperative clauses which tend to surface clause-initially in Latin as
a consequence of the activation of a V-to-C constraint of a type already
well-attested in a number of SOV languages (thus deriving ViO sequences,
both when the imperative is in its proper form and when it is in the form of
a hortative subjunctive). A crucial piece of evidence for assuming a V-to-C
movement of the (imperative) verb comes from recent studies on Gothic
syntax (Ferraresi 1992 and Longobardi 1994), inspired by an observation al-
ready contained in Meillet (1908/09) in relation to a passage of the Gospel
by Mathew (8, 3). 33 The analogies detected between Gothic, on the one hand,
and Latin, on the other hand, seem to justify a uniform analysis for the in-
flected verb in imperative clauses, pointing to a V-to-C movement behind
the verb first pattern they converge on, as illustrated in the following Latin
sentence:

V DO
(19) Agite, inquit, scordalias de medio.
banish-2PL, (he said), quarrels-PL.F.ACC from here
'Come on, he said, that's enough wrangling' CT, 59, [1]

2.13. V2-like constraints in Latin

A number of verb first occurrences in main clauses is amenable to a Verb

Second constraint (V2) which seems to be at work in Latin in a limited ar-
ray of contexts; the latter can be equated with a main clause preceded by a
400 Chiara Polo

subordinate clause, which results in a [ sub XP] - V2... pattern. The tendency
to move the verb to the front position of a main clause is well documented
after an ablative absolute construction (as observed by Marouzeau 1938:
80) and a conditional or, even more frequently, a temporal subordinate
clause (Mobitz 1924: 120): this is upheld by Petronious' work, where the
construction recurs frequently after temporal, but also after conditional
subordinates (Kroll 1918: 117; Mobitz 1924: 121), to substantiate its wide-
spread employ in everyday speech. According to Bauer (1995: 96-97), the
rate of occurrence of this pattern increased during the course of Latin and
gradually spread to popular texts: in Cena Trimalchionis, the construction
is documented with a certain frequency, and the properties it manifests
seem to be reminiscent of verb second phenomena attested in Germanic
languages. 34 The trigger for the movement of the verb could be tentatively
taken to be similar in nature to the one Salvi (2004: 1.2.4) proposes for V]
clauses: 35 namely, it could be the case that the role played by an abstract
operator located in the specifier of a high functional projection deriving V)
phenomena is taken over here by temporal / conditional clauses which
would land in the specifier of the same projection to result in V2 effects: 36

V IO(N)
(20) (Quod si hoc fecerit), eripiat Norbano
(if it comes off), will-steal-3sG Norbanus-SG M.DAT

DO(N)
totum favorem.
all favour-sG.M.ACC
'If it comes off, he'll put Norbanus right out of the running'.
CT, 45, [10]

Things being so, cases of verb first in main clauses preceded by subordi-
nate clauses can be considered structurally motivated deviations from the
unmarked word order (S)OV in Latin and, as such, will be referred to the
old, OV Latin grammar. 37

2.14. Movement of the clitic verb forms of sum to the Wackernagel position

Word orders which evidence a displacement of the verb from the canonical,
sentence-final position to second sentence position can be also found when
the different forms of sum happen to be affected: as underlined by Salvi
 Latin word order in generative perspective 401

(2004: 1.2.5), at least some forms of sum frequently recur in second sen-
tence position immediately after the first element (word or phrase) of the
clause, both in main and in subordinate clauses (in the so-called Wackernagel
position, as defined by Adams 1994). In this typology of structures, the
deviation from an unmarked adjustment of sentence constituents can be
derived by appealing to an (independently well-documented) clitic move-
ment of the verb to a specific 'landing site': 38

(21) Nulla est a me umquam sententia

no-SG.F.NOM is-3so by m e - I S G . A B L ever judgment- SG.F.NOM

dicta in fratrem tuum

said on brother your-SG M ACC
Ί have never pronounced a judgment on your brother'
(Cie. Farn. V. 2. 9: from Salvi 2004: 1.2.5.)

2.15. Latin word orders: A summary

In summary, the advantages of generative theory over the traditional ap-

proach come plainly to the surface when the 'dogma' of Latin as a non-
configurational, free word-order language happens to be reconsidered: the
theoretic possibility of admitting a naturalis ordo along with an array of
deviations from it - trangressio - as already pointed to by Quintilian in his
Institutio Oratoria contributes to reducing Latin seeming superficial unruli-
ness to a limited set of hopefully universal constraints and invariable trig-
gers, whose derivational success crucially rests upon the assumption of an
underlying, hierarchical structure, in which every constituent has its ap-
pointed slot and a set of landing sites it can move to. On this view, once
(S)OV is assumed as the basic, unmarked order, the other five options of
word combination in the greatest majority of cases turn out derivable, as
made also intuitively clear by the markedness of the situational conditions
framing their occurrence. 39 This generalization applies exceptionlessly to
the string-types VSO, VOS, OSV, OVS - which most frequently involve
the simultaneous application of two displacement constraints, as licensed
by the context - and, even if more problematically so, also to the string
(S)VO, although 10% of its occurrences (to be confronted with the remain-
ing 90%) remains basically unaccounted for under our proposal, and conse-
quently calls for further investigation. However, before attempting an ex-
planation for this residual percentage of seemingly unruly (S)VO patterns, I
402 Chiara Polo

would like to spend just two words on the way VSO, VOS, OS V, and OVS
orders are derivable, using to this purpose some meaningful examples. As
to the former two patterns VSO and VOS , they converge on a verb-first
organization, which might be interpreted in terms of the activation of a verb
fronting rule, jointly or not to the post-position of the subject (VOS), which
happens to be motivated on the ground of its marked (rhematic or heavy)
status - as already assumed by Ostafin (1986: 159).40 As far as this verb-
first pattern is concerned, on the other hand, according to Salvi (2004:
1.2.4.), it is frequently found with jussive and concessive sentences, along
with clauses with assertive value, clauses instantiating a contrast, presenta-
tional clauses, to which we add narrative contexts recording a crucial pas-
sage. Technically speaking, the particular typology of clauses featuring this
construction-type might be characterized in terms of the presence of an
abstract operator in the specifier of a functional projection of a certain type -
possibly, FocP as argued for by Salvi (2004: 2.1), whose semantic activation
would follow only from the presence of lexical material in the position of
the head of the same projection. On this view, the raising of the verb could
be regarded as a possible solution for the recovery of the semantic content
of the operator:

V S
(22) Convertit ad hanc scaenam Trimalchio
turned-3sG to this Scene-SG.F.ACC Trimalchio-SG.M.NOM
DO(N)
vultum et: "Amici"...
face-SG.M.ACC and: "Friends"...
'Trimalchio surveyed this tableau, and said: "My friends",...'
CT, 33, [5]
V S DO(N+H)
(23) Accipimus nos cochlearia non minus
p i c k e d - u p - l P L we-lPL.NOM spoons-PL.N.ACC n o t less
selibras pendentia
h a l f pound-PL.F.ACC weighing-PL.N.ACC
(ovaque ex farina pingui figurata pertundimus)
(and assaulted the eggs which were made of flour baked in oil)
'We picked up our spoons weighing not less than half a pound, and as-
saulted the eggs which were made of flour baked in oil' CT, 33, [6]
 Latin word order in generative perspective 403

Both the VSO surface strings illustrated above are attested in Chapter
XXXIII and, crucially, occur one after the other in the same paragraph:
they seem to carry out the function of marking the return of Trimalchio on
the scene, after the short interval dedicated to finishing his game. The
'event' seems to be worth an ironic underlining, which is entrusted to a
syntactically marked pattern with a fronted verb to highlight that Trimal-
chio has resumed his role of unrivalled director of the action and a remark-
able shift has occurred in the narrative. (Under this hypothesis, then, an
abstract, narrative operator - with the meaning of 'Then, At that point' -
occurring in the specifier of the Focus Phrase could be taken to underlie the
raising of the verb to Foc° to effect its semantic content activation). As a
matter of fact, it is on Trimalchio and on his action that the focus is placed,
when he surveys the tableau newly brought in and suggests to his guests to
test whether the eggs distributed among them are soft enough to eat; in
obeisance to his orders, the protagonists pick up their spoons and assault
the eggs and their action is given the same prominence as that performed by
Trimalchio which immediately precedes and to which it is added - as in a
list - through the same syntactically marked format VSO, intended to have
the pressing rhythm of the actions orchestrated by the master of the house
suitably emphasized.

V DO(T) S(Foc+H)
(24) a. Tegebant asellum duae lances (in quarum
covered-3PL donkey-SG.M.ACC t w o dishes-PL.F.NOM, ( o n t h e r i m s
marginibus nomen Trimalchionis inscriptum erat et argenti pondus)
of which were engraved Trimalchio's name and their weight in
silver)
'Covering the donkey were two dishes, on the rims of which were
engraved Trimalchio's name and their weight in silver'
CT, 31, [10]
V DO(T) S(Foc+H)
b. (Lo) Coprivano, I 'asinello, due piatti, (sui cui bordi erano
(it) covered-3PL, the donkey, two dishes, (on the rims of which
incisi i nomi di Trimalcione e la caratura dell 'argento),
were engraved Trimalchio's name and their weight in silver).
'Covering the donkey were two dishes, on the rims of which were
engraved Trimalchio's name and their weight in silver'
404 Chiara Polo

Turning now to the VOS string quoted above, it is documented in Chapter

XXXI, when a most elegant hors d 'ceuvre has been brought in, and the nar-
rator has provided a first, rough description of it as consisting of a donkey
made of Corinthian bronze, bearing a double pannier which contains white
olives on one side, and black on the other. The narrator Encolpius then con-
tinues his description of the entree dish, providing further details about it:
he explains, for example, that the donkey is covered by two dishes, going
on with their portrayal. By now, however, the referent denoted by asellum
is already present in the universe of discourse, and is consistently treated as
a topic by the speaker, because it expresses old information, available and
salient in the previous discourse: its pragmatic value could provide an ex-
planation for its post-verbal placement in terms of a right-dislocation (as
made explicit in the Italian translation by the presence of the anticipatory
clitic lo coreferential with the right-dislocated constituent I 'asinello). As to
the clause-final, deviant placement of the subject, it could be accounted for
by taking into consideration the new information it introduces and the
'heaviness' it manifests 41 , both of them qualifying as licensing factors for
its clause-final ordering (possibly to be worked out in terms of a heavy NP-
Shift). In this case, then, the verb first string VOS would be derived
through a double (rightward) movement of the Topic object and the heavy
subject.
Turning now to OSV and OVS, a brief overview of the contextual con-
ditions under which these structures recur confirms their markedness,
which can be figured out in terms of the specific values carried by the ob-
jects - endowed either with a Topical value or a Contrastive Focus or Em-
phatic reading, which argues for a left-dislocation or a left-hand focaliza-
tion process underneath their anomalous front positioning - co-occurring or
not with a post-posed, marked, Rhematic subject, expressing new informa-
tion, unavailable in the preceding linguistic context, as shown by the fact
that it often introduces a new character into the universe of discourse. 42
Alternatively, as pointed out by Ostafin (1986: 154-157), it can also read
as a Contrastive Focus constituent, even if it usually rejects a Topic inter-
pretation (under the latter pragmatic situation, a canonical S...V ordering is
preferred over its marked alternant ...VS) or, finally, it can also consist of a
heavy, complex constituent:
 Latin word order in generative perspective 405

P(T) Vinf S(H)

(25) a. nec magnum esse peccatum suum propter quod
nor Severe-SG.N.ACC (to) be fault his-SG.M.ACC for w h i c h
periclitaretur;...
was-in-danger-3sG;...
'The cause of his predicament, he said, was a mere peccadillo:...'
CT, 30, [7]
P(T) V S(H)
b. ne grave era la sua infrazione per la quale ora si trovava
nor severe was the his fault for the which now was
nel pericolo di essere punito
in danger of being punished
'The cause of the predicament for which he was now at risk of
being punished was not so severe'

The PVS string occurs in Chapter XXX: it refers to the moment when a
slave stripped for flogging grovels at the protagonists' feet, and proceeds to
implore them to rescue him from punishment. His words are evoked by the
narrator through the indirect speech technique and are aimed to convince
his interlocutors that the fault for which he is being punished is not that
serious. A quick glance at the formal shaping of the clause set in its context
unveils the structural complexity of the 'late' subject peccatum suum propter
quod periclitaretur, to the effect that its postposed position turns out to be
justifiable in structural terms.
Going back to the string (S)VO, we have already seen how the tradi-
tional claim of a free variation between OV and VO based exclusively on
their occurrence rates does not find any empirical support, given that the
greatest majority of deviant, (S)VO orders occur under pragmatically or
contextually marked environments and are consequently liable of being
derived from a basic (S)OV order through one of the following licensing
strategies:

- Right-dislocations and Right-hand focalizations, triggered by the

positive specification carried by the displaced object for such discourse-
related features as [+ Topic] and [+ Contrastive Focus] / [+ Emphasis]
((S)VO T /CF/E);
- Phenomena of syntactic discontinuity affecting the DO constituent, split
by the intervening lexical verb between its adjectival modifier and its
nominal head (DO SA -V-DO SN ) to instantiate a mise en relief- Marouzeau
406 Chiara Polo

(1922) - of the preposed, focused, adjective, as carried along by its dis-

junction from the agreeing, post-verbal nominal component; 43
- Heavy NP-Shifts, promoting the late placement of intrinsically complex
objects ((S)VOH);
- V-to-C movements, to be found without exceptions in imperative
clauses (VO!), and, possibly, verb-seconding phenomena, most fre-
quently detected in main clauses immediately preceded by temporal or
conditional subordinate clauses [ s u b XP]-V 2 -0, as already argued for in
Bauer (1995: 95), Kroll(1918: 117), and Mobitz (1924: 121);
- Movement to the Wackernagel position of the clitic forms of the verb
sum.
Notwithstanding the explanatory success granted for by the derivational
apparatus just surveyed, 10% of neutral patterns calls our attention because
structured into an (S)VO order contra our expectations. The goal of the next
section is to discuss some proposals meant to shed light on the syntactic
anarchy they seem to manifest.

3. Evidence for grammars in competition?

A look at the diachrony of Latin

At this point, once established that Petronian Latin proves consistent with
an (S)OV grammar at least as far as the domain of sentence syntax is con-
cerned, the question arises as to the residual percentage of cases (10%)
which stand out as counterevidence to the typologically ideal situation em-
bodied, for example, by Italian, in which one and only one surface pattern
is permitted to occur under informationally and structurally neutral envi-
ronments. As already hinted at, this small set of 'idiosyncratic patterns' can
be handled under a twofold proposal:

I. 34% of it (16/47) is liable of being modularised into a number of sub-

classes which answer to synchronic sub-regularity principles, among
which the following two seem to play a crucial role: 44

a. the search for varietas in a sample of artistic prose which, although

mimetically close to a familiar code (Zamboni 2000: 45), remains al-
together exposed to the influence of literary precepts and might have
been affected by chiastic rules of constituent distribution (VO, OV):
 Latin word order in generative perspective 407

S V
(26) Nam puer quidem ... circumibat iam dudum
as for slave-SG.M.NOM was-going-around-3SG already

DO(N) DO(N) V
pedes nostros et missionem rogabat
feet-our-PL.M.ACC and mercy-SG.F.ACC p l e a d e d - f o r - 3 s G
'As for the slave who had tumbled, he was already doing the
rounds of our feet, and was pleading for mercy'. CT, 54, [3]

b. the idiomatic nature of the expressions exhibiting an anomalous se-

quencing VO which, as already observed by Salvi (2004: 1.2.1.4),
would grant for frequent permutations in the relative order of verb
and internal argument (OV and VO), as made possible by the strict
syntactic bond existing between them; 45
II. the 66% (31/47) of them might function as a litmus-paper for a sociolin-
guistically marked situation o f ' g r a m m a r s in competition' (Kroch 1989;
Lightfoot 1999; Pintzuk 1991) in which the old, still dominant grammar
OV reconstructed for Petronian Latin coexists side by side with an in-
novative, Romance-like one VO which was intruding into a once-stable
system up to its complete overthrow. 46

The possibility to pursue a second, diachronic line of research besides the

synchronic one explored so far affords new chances to penetrate the ratio
underlying superficially idiosyncratic and asystematic phenomena: in the
case in point, there are good empirical reasons to believe that this small
number of apparently arbitrary orders of Latin relates to an innovating
grammar V O which crucially unfolds some of its distinctive properties pre-
cisely in this late Latin text. 47 On the other hand, a thorough examination of
these 47 examples helps unmasking the inconsistencies of the predictions
yielded under functionally-oriented proposals of analysis. Contra Meillet,
the trigger for the establishment of a rigid typology VO seems not to be
identifiable with the loss of accusative morphology on the object - as dem-
onstrated by the fact that the accusative marker shows up in all the 47 ex-
amples which turn out problematic to our eyes (iam embasicoetas
praeferebat facem,.. CT, 26, [1]; quemadmodum Cassandra occidit flllos
suos,... CT, 52, [l]), 48 and, hence, the process cannot be read as a func-
tional replacement of an overt case-marker with a rigidly codified order
VO. To conclude, the additional dimension of analysis just hinted at seems
to enable us to catch a glimpse of the unexpected regularities which lie
408 Chiara Polo

concealed underneath superficially inexplicable phenomena: in this sense,

the Saussurrean conception (1916) of synchrony and diachrony as two dis-
tinct, but also interacting and mutually enlightening aspects of linguistic
analyses lets its everlasting modernity come plainly to the surface and be
thoroughly appreciated in its heuristic potentialities.

4. Conclusions and perspectives for the research of the future

In sum, the application of an integrated methodology of investigation -

quantitative and qualitative - to a comparative study of Latin and Italian
syntax has resulted in undermining the explicative and predictive power of
functionally-driven, simplistic views of languages as dichotomically split
between inflectional and free word order versus morphologically impover-
ished and syntactically rigidly codified ones. The possibility to use more
sophisticated tools of investigation, accessing also the informational im-
ports matching with surface strings, led us to reconstruct the regularity en-
tailed also in the syntactic organization of fully inflectional languages like
Latin; at the same time, it has contributed to unfold the manifold similari-
ties holding between Latin and Italian syntactic shaping, to start from the
existence of a basic, underlying, statistically dominant order ((S)OV vs
(S)VO), along with a number of pragmatically or structurally motivated
deviations, which prove nonetheless derivable through the well-assessed set
of displacement rules assumed under the generative framework. The loca of
irreducible variation between the two genealogically related languages have
been thus proven to be accountable under the Principles and Parameters
Theory in terms of differences in discrete parametric values (concerning
mainly the setting of the head-complement parameter, as allowed for in a
pre-Kaynian perspective). Furthermore, the conjunction of a synchronic
with a diachronic line of investigation has enabled us to recognize an as-
cending, innovating grammar VO competing with the older OV in the
Petronian text, thus yielding an interpretative key for that residual percent-
age of Latin structures (S)VO (10%) which apparently resisted our expla-
natory efforts. The traditional claims which relate the reanalysis OV VO
to the loss of accusative-marking on the postposed object have been fur-
thermore proven to be devoid of empirical support.
The fascinating horizons which are left for future research range over
the widening of the empirical data-base to comprise the greatest number of
texts dating back to the chronological hiatus in-between the Satyr icon and
 Latin word order in generative perspective 409

the e m e r g e n c e of the R o m a n c e paradigm; it seems of crucial importance to

have access to texts belonging to different genres, going f r o m inscriptions,
semi-literary sources like the letters of the soldier Claudius Terenzianus
(2 nd century A D ) to continue with innovating and linguistically less con-
trolled testimonies like Peregrinatio (Itinerarium) ad loca sancta of the
th
Spanish nun Egeria of the end of the 4 century in order to further refine
our generalizations and arrive at more stable conclusions. 4 9

Acknowledgements

I would like to a c k n o w l e d g e my debt to Katalin Kiss and G i a m p a o l o Salvi

for their useful comments, advice and suggestions, which greatly contributed
to the i m p r o v e m e n t of the final version of my paper.

Notes

* In the present paper the following abbreviations and symbols have been adopted:
ABL = ablative; ACC = accusative; C = (obliquely case-marked) complement/
complementizer; CF = contrastive focus; cl = clitic; CT = Cena Trimalchionis;
D = discontinuous; DAT = dative; DO = direct object; Ε = emphatic; f. =
feminine; FOC = focus; GEN = genitive; Η = (structurally) heavy; INST = in-
strumental; IO = indirect object; LOC = locative; m. = masculine; n. = neuter;
Ν = (informationally) neutral / unmarked; NOM = nominative; NP = noun
phrase; Ο = object; Ρ = predicative complement (of the S); pi. = plural; PP =
prepositional phrase; S = subject; sg. = singular; Τ = topic; ν = auxiliary verb;
V = verb; Vpp = past participle verb; VP = verbal phrase.
1. These hypotheses find a consistent exposition in the following passage by
Meillet (1903 [1964]: 439): the author is here mainly concerned with the syn-
tactic effects brought about diachronically by the morphological levelling many
a natural language underwent, but he also hints at the typological intertwine-
ments between morphology and syntax supposedly showing up in the syn-
chrony of the Latin language:
Les transformations du type morphologique s'accompagnent de transformations
paralleles de la phrase. La phrase indo-europeenne se composait de mots auto-
nomes, dont chacun suffisait ä exprimer un sens complet et la fonction remplie.
Quand les n o m s ont reduit, puis perdu la declinaison, la fonction des noms dans
la phrase a ete indiquee par deux precedes nouveaux: 1° Un ordre de mots ä
410 ChiaraPolo

valeur g r a m m a t i c a l . En frangais ou en anglais, la place du nom suffit en gene-

ral ä indiquer la fonction: le pere aime le fits indique par l'ordre ce que le latin
indiquait par la flexion: pater fllium amat, filium pater amat, amat filium pater,
etc, et l'ordre est la seule marque de la difference de valeur grammatical. 2°
Des mots accessoires. (The boldface italics are mine).
I would further point out that in this paper I am primarily interested in tradi-
tional, functionally-oriented positions a la Meillet which have exerted a perva-
sive influence on the literature available on Latin word order up to now. I will-
ingly leave aside alternative, traditional views of Latin syntax as the ones ex-
pounded in Hofmann and Szantyr's (1965) and Kühner and Stegmann's (1955)
handbooks (or even in traditional grammarians' descriptions like Quintilian's)
which, on the contrary, tend to converge on acknowledging the existence of
some distributional tendencies in Latin as well, although their intuitions are
expressed through a traditional and unformalized terminology. I am aware, for
instance, that the observational statement that Latin verbs normally occur
clause-finally is not missing from their descriptions, not unlikely from the real-
ization that deviating word orders deliver special pragmatic or semantic values.
Yet, I deliberately decided to focus on functionalistic conceptions of Latin
word order by reason of the important role they continue to play in contemporary
debates on the topic; the verification of their validity appeared to me a suitable
and stimulating starting point for the development of my analyses.
2. Marouzeau (1948: 155) shows a similar concern for pragmatic principles pre-
sumably underlying Latin word ordering. His intuitions stand out for their
modernity, as exemplified by the following quote, where greater attention is
drawn to the semantic discrepancies delivered by different linear combinations:
"Le latin offre un bon exemple de la liberte de la construction: liberte de choix,
non d'indifference".
3. Actually, the confrontation of parallel texts is a well-assessed practice in other
domains of research, even if it has been scarcely explored with reference to
Latin; in this sense, the methodology employed here can be defined as 'innova-
tive', not because it be absolutely new in itself, but because it happens to be at
one of its first applications to the study of Latin word patterns (on this point, I
would like to thank Salvi for reminding me of Alisova's (1972) attempt to ex-
ploit the same, textual-comparative strategy for the conjoined analysis of Ital-
ian and French word orders).
4. The legitimacy of considering Cena Trimalchionis as a suitable source of spo-
ken Latin has been widely debated among scholars and still remains a basically
unsettled question. Two apparently irreconcilable positions have been worked
out, depending on whether the language recorded by the text is supposed to
adhere faithfully to the standardized patterns of the literary, classical norm, or
whether, on the contrary, it is thought of as a mimetically reliable reflex of the
speeches of lower social classes, even if it remains undeniably the artefact of a
 Latin word order in generative perspective 411

culturally refined author like Petronius. In this scenario, Adams (1976: 83-94)
strenuously defends the former view, as shown by his claim that Cena Trimal-
chionis is an untrustworthy basis to reconstruct spoken patterns ('In word order
Cena Trimalchionis is unrepresentative of genuine Vulgar Latin ... being an
artificial concoction by a man of learning'), unlike the approximately contem-
porary letters of the soldier Claudius Terentianus which, on the contrary, are
taken to give 'a glimpse of semi-literate Latin'. Hinojo (1985: 245-246) and
Zamboni (2000: 4 4 ^ 5 ) , on the other hand, tend rather to cautiously credit the
text with a documentary value for the study of Vulgar Latin features ("... con
las debidas limitaciones, <la obra> puede considerarse como uno de los textos
caracteristicos del latin vulgar"; "Un documento citatissimo in questa prospettiva
come il Satyricon di Petronio rivela ... la presenza dell'orale nello scritto (un
orale di natura mimetica, ricostruito e non oggettivamente registrato come
farebbe oggi un ricercatore), in altre parole elementi e parametri deWimme-
diato comunicativo"). Given these alternative perspectives, I take side with
Hinojo and Zamboni and consistently regard the text under scrutiny as one of
the few choices I could make to have an insight into the familiar and colloquial
registers mastered by 1st century Latin speakers.
5. The Principles and Parameters Theory captures the intuition that the interlin-
guistic variation showing up on the surface both synchronically and diachroni-
cally can be uniformly characterized as a by-product of the specific articulation
of our language faculty (roughly coinciding with Universal Grammar - UG),
which is supposed to consist of invariable principles, along with a set of open
parameters. The latter would be open choices, ideally allowing for binary op-
tions, to be set by language learners under exposure to their primary linguistic
corpus; the value parameters are given is thus a function of the triggering expe-
rience and can accordingly vary not only from language to language, but also
from generation to generation within one and the same linguistic community.
In a pre-Kaynian framework as the one tentatively adopted here, the head-
complement parameter is one of the open choices specified by UG which chil-
dren have to fix as complement-head or head-complement depending on the
linguistic environment they are surrounded by. Hence, under our proposal, the
discrepancies detected between Latin and Italian syntax are ultimately amena-
ble to a different setting of the head-complement parameter (fixed respectively
as complement-head in Latin and head-complement in Italian). On the con-
trary, under Kayne's (1994: 5-6, 35-36) antisymmetric view of syntax, there is
only one universal order underneath surface variation - Specifier-Head-
Complement (coinciding with an S-V-0 linearization borne out by the Linear
Correspondence Axiom for all language types), with the implication that the
order (S)OV which we tentatively characterize as basic and unmarked in Latin
is held to be derived out of necessity, and the clear opposition (S)OV vs (S)VO
displayed by Latin and Italian syntaxes needs to be worked out in other terms.
412 ChiaraPolo

6. It seems finally worth pointing out that the method employed to classify the
documented strings conforms to the parameters made available by modern ty-
pological studies; it willingly adheres to the idealizations and simplifications
which are commonly adopted under this field of studies to let broad linguistic
generalizations come to the surface. However, there is a number of problems 1
had to face in my comparative survey of Latin and Italian word patterns which
I think necessary to mention very quickly so as to provide a glimpse of the
cataloguing procedure I stuck to, along with its limits and the possible ways of
compensating for them. The first difficulty I was confronted with has to do
with cases where one of the nominal constituents - most frequently the DO -
evidences in Latin a syntactic discontinuity between its adjectival modifier and
the nominal head it refers to, owing to the presence in-between of some lexical
material - most frequently the verb (DO AP - V - DO N P ) - to the effect that part
of it surfaces in front of the intervening verb, while the rest comes after it:
(i) et ipse capaciorem poposcit scyphum, ...
as well he-SG.M.NOM larger-SG.M.ACC demanded-3SG cup-SG.M.ACC
'Trimalchio too demanded a larger cup,...' CT, 65, [8]
As a matter of fact, the split status of the DO makes it difficult to decide
whether the structure should be classified with OV- or, rather, with VO-strings.
In these cases, I conformed to Ostafin's practice to take the position of the
nominal head as indicative of the ordering of the discontinuous constituent {i. e.
D O a p - V - DO N P = V-DO), as made plausible by the emphasis usually placed
on the fronted, adjectival head (often in the comparative degree), which might
argue for a strategy of mise en relief par disjunction - in Marouzeau's (1922)
terms - affecting the modifier in isolation, with the enhanced effect of turning it
into a marked constituent and, as such, into an unreliable indicator of the basic
position of the constituent. Anyway, even if my choice should sound arbitrary
or questionable, the paucity of the examples exhibiting this 'anomaly' (3.5%
of the totals: 42/1181, with just 14 affecting DOs - equal to the 1.1%, as
shown in the table below) ensures that the final statistical amount recorded for
each word order pattern is not dramatically altered by the decision 1 was forced
to take.

Tab. 1: Discontinuity

Discontinuity S DO IO Ρ PP C Gen N°/Gen All constituents

1. [-D] 227 427 21 102 140 37 95 90 1139 96.5%
2. [+D] 6 14 0 5 4 4 2 7 42 3.5%
Totals 233 441 21 107 144 41 97 97 1181 100%

For the time being, however, I leave the question willingly open, given that its
thorough discussion goes beyond the aims of the present paper; I refer to note
 Latin word order in generative perspective 413

43 for some hypotheses on the derivational machinery involved in these dis-

continuity types. Secondly, I had to deal with analogous problems when the
verb consists of a periphrastic form, made up of an auxiliary verb, along with a
past participle, showing up in two different positions with respect to the other
constituents analysed (S and DO) or, more simply, occurring in the two possi-
ble sequences with respect to each other (i.e. V pp v and vV pp ): in the former
case, I preferentially relied on the position of the lexical verb to make out
which typology of surface pattern is involved, on account of the clitic nature
frequently attributable to the auxiliary, which, in its turn, argues for its likely
movement to a Wackernagel position and transforms it into an untrustworthy
marker for the (underlying, hopefully, unmarked) position of the verb; in the
latter case, on the other hand, no such problem seems to arise, given that the
two verb forms stand in the same position with respect to the other sentence
constituents and no choice between them has to be made, even if their collapse
into a unique symbol V entails that a significant element of typological consis-
tency or inconsistency with respect to a presumable Latin type OV is missed
(see Greenberg's (1963: 83-85) universals 13, 15, 16). As previously stated,
however, I leave willingly aside the discussion of all these open questions,
given that I am here primarily interested in drawing a rough outline of Latin
syntax on sentence level; subtler problems are left for investigation to later
sections, where higher degrees of explicative adequacy will be in order.
7. This aspect of Latin syntax would require a separate chapter; for this reason, I
cannot deal with it in this paper. For a discussion of syntactic break-up phe-
nomena, I refer to Polo (2004), Chapter 4, and Polo (forthcoming); see also the
hints contained in note 6 and 43.
8. I would like to thank Salvi (personal communication) for warning me against
the risk of straightforwardly equating analogous (marked) pragmatic functions
with one and the same syntactic constructions and derivational strategies
across the two languages, as made advisable by the lack of a biunivocal corre-
spondence between them: yet, the point I want to make is that the possibility
for a match between similar pragmatic values and comparable syntactic struc-
tures cannot be a priori excluded, even if it needs demonstrating cautiously. In
turn, the necessity for its careful checking does not deprive such a methodo-
logical assumption of its value as a simple working hypothesis: it is precisely
in this perspective that it is proposed and exploited here, while further verifica-
tions are being pursued and are left for future research.
9. As stated by Salvi (2004: 2.3), it may also be the case that the complementizer
is realized (1) after a peripheral element (in T°), (2) after a focused constituent
(possibly in F°), and (3) after an element inside the sentence (within I"): que
Left Periph. que Focus que [i··... que V],
10. Another point of departure of my proposal from Salvi's concerns right dislo-
cated elements; since Salvi adopts Kayne's approach, no independent position
414 ChiaraPolo

is postulated for them, contra what I admit working under a pre-Kaynian

framework; furthermore, unlike Kayne's theory, no general ban holds in my
analysis against rightward movements and right-hand adjunctions (on this
point, see also: Larson 1988, 1990).
11. As far as tables 5 and 6 are concerned, I would underline that the discrepancies
holding in the total number of unmarked orders between Latin and Italian are
imputable to the fact that unmarked imports do not always happen to be
mapped into syntactically comparable structures in the two languages, to the
effect that, occasionally, one and the same informational import is delivered
through a transitive construction in Latin, but an intransitive one in Italian. In
these cases, given that our attention is willingly restricted to transitive, S/V/O
strings, the choice I felt advisable to make was to exclude these structures from
my computations, because of the lack of the homogeneous conditions essential
to a minimal confrontation.
12. This holds true for the still unverified hypothesis - which sounds reasonable in
principle - that Latin and Italian syntactic constructions be overlapping, at least
as far as some well-circumscribed empirical sub-domains are concerned.
13. This fact, however, does not hinder us from tentatively assuming left-disloca-
tions in Latin on a par with what has been independently documented for a
language like Italian; furthermore, the well-assessed status of this linguistic
phenomenon in Italian explains why all the examples quoted in this section are
drawn from the copious literature available on this topic in Italian.
14. From an intonational viewpoint, left-dislocated sentences may not differ from
their unmarked counterparts, even if a 'comma intonation' can be used to set
the preposed element(s) off from each other and from the rest of the clause. As
far as intonational considerations are concerned, we know that, when applied
to Latin, they lose importance, owing to the written nature of the material we
are forced to work on (on this point, see Salvi 2004: 1, where he states that
"Gli argomenti intonativi possono essere praticamente esclusi, data la natura
scritta del nostro materiale"). Prosody happens thus to play a minor role in the
investigation of the sentence structure of a dead language; however, this fact
should not undermine the validity of our line of research, so much so because
we cannot a priori exclude that the prosodic observations made for Italian Left-
dislocated structures hold minimally true also for Latin.
15. All the examples quoted are taken from: Benincä, Frison and Salvi (1989: 120;
130-131). For a detailed account of Left-dislocation phenomena, I further ma-
ke reference to: Berretta (1994: 79-83); Cinque (1977: 397-411), and Rizzi
(1998: 112-131).
16. Even if Left-dislocated structures did not rest upon a movement operation, as
has been frequently argued for (Chomsky 1977: "CLLDed topics are licensed
by 'rules of predication' that require that the topic be 'base-generated' as sister
of the XP that is predicated of it"; see also Duarte 1988), nothing would
 Latin word order in generative perspective 415

change from our perspective. In fact, to our purposes it is sufficient to show

that, under the pragmatically marked conditions associated with left-dislocation
contexts, OV sequences turn out to be convergent both in Latin and in Italian.
17. Because these OV-types are classified with marked strings.
18. The English translation of the Petronian examples is based on Walsh (1996);
the indications provided, namely CT, 74, 14, standing for Cena Trimalchionis,
chapter 74, 14th section of the chapter, are referred to the edition translated by
Aragosti (1999).
19. The interpretation of the clause quoted in (4) seems to raise an interesting
problem: as signalled to me by Salvi and discussed by him in the present vol-
ume, translators do not unanimously converge on assigning a topic value to the
object aedes. More precisely, Aragosti's (1999) choice to view it as a Topic
contrasts with a general preference Petronious' translators show for regarding
it as part of the rheme (Non si sogna certo un palazzo). This fact draws atten-
tion to a risk we can bump into when working on dead languages, namely the
danger of hazardously reconstructing the author's communicative intentions,
ending up with a number of different interpretative solutions, as the case under
scrutiny seems to instantiate. However, from my viewpoint, the example just
commented upon proves useful to show that, when the context does not cate-
gorically exclude a marked reading of the object, Latin and Italian can be sup-
posed to converge on activating the same derivational strategies which, in turn,
lead the way to an OV-structuring. This conclusion would not be falsified if
the object aedes should happen to match with a non-topical import. In this ca-
se, what we would have to document is that Italian departs from the (basic)
OV-organization of Latin for adhering to an underlying VO-order, as expected
if no marked licensing strategy could be posited to derive the anomalous order-
ing of the object in front of the verb. Nothing incompatible with our assump-
tions seems thus to be evidenced; at worst, we would have to acknowledge that
the example submitted to analysis is not the most suitable for representing left-
dislocation structures.
20. Examples (5a-c), along with (6a,b) are quoted from Benincä, Frison and Salvi
(1989: 146, 148); examples (5d,e), on the contrary, are taken from Berretta
(1994: 83). The first - (5d) - illustrates a case of deixis to an element which is
known and salient {mamma, referring to 'your mother'); the second - (5e) - is
a case of deixis stricto sensu under the hypothesis that the speaker pronounces
it with the photographs under his eyes.
21. On the contrary, as observed in Benincä, Frison and Salvi (1989: 147), whenever
resumptive clitics are there, we are allowed to assume that (right-dislocated)
elements surface displaced from their original positioning, both on analogy
with left-dislocations, and on account of the agrammaticality of the following
clauses involving a unique tonal unit (/. e. no pause):
416 ChiaraPolo

(i) a. *L'ho comprato un bei libro.

* it have-1SG bought a nice book
b. *Lo inviteremo Giorgio.
*him will-invite-1 PL George
Turning to prosodic features, right-dislocations are said to frequently involve an
intonation pause before the dislocated constituent, which forms an independent
prosodic phrase from the rest of the clause, and, as pointed out by Zubizarreta
(1998: 154), bears its own nuclear pitch accent, while a certain emphasis is
placed on the verb or on another constituent.
22. See, in particular, Adams for a critical approach to the question (1976: 72, 77).
23. Under Kayne' framework, on the contrary, right-dislocations would be derived
from an universal SVO order through remnant leftward movement of the rest
of the clause. In this respect, however, the more traditional framework does not
differ from Kayne's proposal as to the empirical predictions it makes, with the
entailment that the label "rightward movement" can be used with no further
specification for right-dislocated structures, as allowed for by its translatability
into both the frameworks.
24. Cinque (1990) uses the term "Topicalisation" to refer to the English construc-
tions (i) and (ii), along with the Italian (iii):
(i) Your book, you should give t to Paul (not to Bill)
(ii) YOUR BOOK you should give t to Paul (not mine)
(iii) IL TUO LIBRO ho letto (non il suo)
THE YOUR BOOK have-lSG read (not the his)
Rizzi (1998: 117, 121), on the contrary, prefers to avoid the term "Topicalisa-
tion", to replace it with Topic (Comment) structures for (i), and (iv) below, and
Focus (Presupposition) structures for (ii) and (iii).
(iv) II tuo libro. lo ho letto.
the your book, it have-lSG read
'Your book, I have already read'
In our practice, structures like (iv) are indicated as Left-dislocations (involving
a preposed topic constituent), while constructions like (ii) and (iii) are pointed
to as (lefthand) focalizations.
25. In Latin, on the other hand, the exact interpretative import of focus construc-
tions has not been satisfactorily studied and ascertained yet: however, scholars
tend to believe that it has a broader scope of employ if compared to Italian, in
that it can be non-contrastive, besides contrastive.
26. The examples listed above are taken from Benincä, Frison and Salvi (1989:
135).
27. Contrastive focus and emphasis are both supposed to 'make a statement about
the truth or correctness of the assertion introduced by their context statement',
but while contrastive focus negates certain aspects of the assertion, emphasis
 Latin word order in generative perspective 417

may respectively negate or reassert the assertion introduced by its context

statement.
(7) NOBODY lied to me.
'there is no x, such that χ lied to me' (Zubizarreta 1998: 7)
[someone lied to you]
To our purposes, however, the important point to be made as regards emphasis
is that it is supposed to have the same syntactic reflexes on the formal shaping
of the clause as contrastive focus.
28. For a complete discussion on this point, see Rizzi (1998: 112-158). Further-
more, I make reference to Zubizarreta (1998: 17-18), where she distinguishes -
with Chomsky (1971) and Jackendoff (1974) - phrasal prominence related to
contrastive focus or emphasis from phrasal prominence related to noncontrastive
focus. Actually the different sources generating them (the Emphatic/Contrastive
Stress Rule and the Nuclear Stress Rule, respectively) underlies the discrepant
distributional properties associated with the two typologies of constituents. In-
tonationally, the focused element is pronounced emphatically, as signalled by a
raise in intensity, a raise in pitch, and its detachment from the rest of the clause
to create an independent tonal unit.
29. Something very similar holds true for the following example drawn from
Chapter XLVIII, where the fronted, clause-initial position the DO constituent
occupies both in Latin and in Italian seems to be imputable to the emphatic re-
ading it pairs with in the given context:
DO(E) V
(8) a. tres bybliothecas habeo, unam Graecam, alteram
t h r e e libraries-PL.F.ACC have-ISG, o n e Greek-SG.F.ACC, o t h e r
Latinam...
Latin-SG.F.ACC
' I ' v e got three libraries, one Greek and one Latin' CT, 48, [4]
DO(E) V
b. ben TRE biblioteche ho, una di greco,
nothing-less-than three libraries have-ISG, one of Greek,
I'altra di Latino.
the other of Latin
' I ' v e got three libraries, one of Greek and the other of Latin'
Actually, the emphatic value associated with tres bybliothecas seems not to be
in question here, given that it fits perfectly Trimalchio's intention to emphasise
the 'huge' number of libraries he has at his disposal as an irrefutable piece of
evidence for his self-taught, 'broad' culture: the emphatic function of the ex-
pression is made here even more apparent by the comic effect it carries along
whereby the fronted, emphatic numeral tres is not accompanied by the logi-
cally expected specification of which libraries the speaker effectively owes
418 ChiaraPolo

(Trimalchio makes reference only to a Greek and a Latin library, while he is

incapable of imagining anything besides Greek and Latin to provide the content
for the third). The compatibility of the DO with such an informational content
grants for a uniform syntactic shaping of the sentence across the two lan-
guages, as carried along by a Left-hand focalization process. As far as the ex-
ample under scrutiny is concerned, it might be objected that standard Northern
Italian would have preferred rendering emphasis through a more natural VO
organization - Ho ben tre biblioteche - in place of the OV-construction type
resorted to in 8b: yet, the fact that the structure instantiating the fronting of the
object turns out grammatical also in Italian (with some dialectal variation be-
tween Northern and Southern varieties, which do manifest different, increasing
degrees of tolerance to the anteposition of the emphatic object) is sufficient to
prove my point that marked communicative imports can have homogeneous
syntactic reflexes across morpho-typologically distinct languages.
30. Under this hypothesis, the focalised object hosted in a post-verbal, Spec,GP
position might be transmitted case from an empty expletive located in a pre-
verbal, case-checking position: Salvi (2004: 2.2).
31. These descriptive generalizations are in keeping with what was observed by
Salvi (2004: 1.2.1.1.), Schneider (1912: 90) and Bauer (1995: 98-101).
32. Benincä, Frison and Salvi (1989: 121) address the question of the "heaviness"
or complexity of constituents and observe that a heavy constituent tends to be
moved towards the end of the clause: this is claimed to be due both to rhythmic
reasons, and to the fact that the final part of the clause is the one where the
most salient information is normally harboured, other conditions being equal.
Under these premises, the late placement of heavy constituents can be viewed
as a function of the high informative content they normally carry.
33. In the passage at stake, a remarkable alternation shows up between the two
orders adjectival predicate - verb and verb - adjectival predicate:
(9) a. hrain warth (M 8, 3)
pure became
έκαθαρίσθη
b. wairth hrains (M 8, 3)
become pure!
καθαρίσθητι
In the two examples above, the order of the constituents is reversed depending
on the clause type: it sticks to the statistically dominant XV kind in the first,
declarative clause, while it changes into VX (!) with the inflected verb in first
sentence position in the second, imperative clause. (Longobardi (1994: 367)
interestingly notices that, in the case under examination, the Greek original
cannot have exerted an influence on the Gothic order, because Greek resorts to
a synthetic construction which does not distinguish between the verb and the
nominal predicate). Ferraresi (1992) shows that the alternation between an in-
 Latin word order in generative perspective 419

ternal and a fronted position of the verb is systematic and matches with a de-
clarative or imperative interpretation of the clause; her examination of the rela-
tive ordering of the verb with respect to locative adverbs confirms that the in-
flected verb obligatorily moves across the adverbial material in imperative
clauses.
34. For a detailed analysis of V2 constructions in Germanic languages, I refer to
Tomaselli (1990). As far as this topic is concerned, further research is needed
to arrive at more reliable conclusions; in particular, by exploiting the kind of
evidence used by Salvi in his contribution to the present volume, it would be
necessary to establish with a higher degree of precision which is the structural
position occupied by the constituent preceding the verb, in order to exclude
that it occurs in positions meant for left-dislocated elements (this could be as-
sessed by observing the distribution of weak words, given that, if the preverbal
phrase should happen to be left-dislocated, weak words would follow the verb).
However, such a verification transcends the scope of the present article, on ac-
count of the fact that the empirical corpus I relied on for investigations (i.e.
Cena Trimalchionis) seems to be lacking in the kind of examples (exhibiting
weak pronouns); the inescapable conclusion is that the identification of Ger-
manic V2 constructions with this set of Latin structures remains a hypothesis
in need of sounder proving.
35. For a more detailed description of Salvi's (2004) proposal, see section 2.15.
36. Under this hypothesis, however, the landing site for the movement of the verb
would be the head of the Focus Phrase, on analogy with what has been sup-
posed for VI constructions, in which the abstract operator in need of semantic
activation through lexical support by the verb was taken to occur in Spec,FocP
(in the case of these V2-like structures, the temporal / conditional clause would
similarly occur in Spec,FocP and act as a trigger for verb movement - I thank
Katalin Kiss for hinting at the possibility of extending the trigger assumed for
VI clauses to V2 instances). Things being so, we would not be authorised to
speak of German-like V2 phenomena, because the structural positions in-
volved would not be the same.
37. Similarly, an analysis in terms of verb second is corroborated by indirect inter-
rogative contexts which frequently match with the typical properties of verb
second phenomena.
38. The landing site involved can be tentatively identified with the C° head or, in
the case illustrated in (21), with the Foc° head, given that the fronted constitu-
ent nulla reads as an emphatic, focalised adjective, disjoined from the noun
head sententia it agrees with, and the following, raised verb est can be consis-
tently supposed to occur in Foc°. A plausible hypothesis we can further ad-
vance is that the verb second constraint of Germanic languages represents a
generalization to all finite verbs of a movement to the Wackernagel position
which was originally reserved for a limited class of verbs. This thesis, however,
420 Chiara Polo

has been already proposed by Wackernagel (1892) for Germanic V2, by Thur-
neysen (1892) for Old French, and criticized as untenable by Salvi (2003).
39. As a matter of fact, semantic evidence remains the main source of information
on which any hypothesis on Latin sentence structure can be grounded. In the
absence of grammaticality and intonational judgements, in fact, our linguistic
arguments can be based on the scanty syntactical evidence we can gather about
Latin, along with (and more heavily so) indirect pragmatic evidence.
40. Under a Kaynian approach, VOS orders could be derived by assuming the
raising of both the light constituents V and O, with the heavy S remaining in
place. Similarly, under Zubizarreta's (1998: 135-138) proposal, VOS struc-
tures could be taken to derive as in Italian from an underlying, universal SVO
order through the leftward adjunction of VO (TP) to a Focus Phrase to ensure
that the right-hand subject ends up provided with the narrow focus interpreta-
tion she attributes to it, and the conflict which would otherwise arise between
the NSR (Nuclear Stress Rule) and the FPR (Focus Prominence Rule) is suc-
cessfully handled.
41. It actually plays the role of antecedent for the non-restrictive relative clause
following in apposition.
42. According to Berretta (1994: 80) among the others, there is a correlation holding
interlinguistically between the subject and the Theme (or, to state it in Givon's
(1983) terms, the subject is a grammaticalized topic; the object, on the con-
trary, coincides with the Rheme). When this equation is not borne out, and the
subject is rhematic and / or the object is thematic, marked word orders as the
ones being reviewed arise.
43. Besides what has been already said on low-level discontinuities in note 6, I
would add some speculative remarks on the way these constructions can be de-
rived under Kayne's (1994), van Riemsdijk's (1989) and my analysis. Possibly,
the simplest way to account for them can be obtained under Kayne's frame-
work, on condition that a small amendment to it be carried out; thus, on the as-
sumption that VO is the universal order, it should be possible to generate OV
not only through the movement of the object as a whole for feature-checking
reasons (to the specifier of an appropriate functional projection), but also
through the displacement of a subset of it {i.e. the adjective agreeing with the
nominal head) capable of checking case in its (pre-verbal) position, and to
transmit it via agreement to the nominal head left behind, in post-verbal posi-
tion. Obviously, the second option is chosen only when a sub-part of the con-
stituent - namely, the adjectival component - needs focusing, and its fronting
matching with its parallel separation from the rest of the constituent is the
marked strategy made available by the language to carry out a mise en relief
function, using Marouzeau's (1922) terminology. A derivational alternative
which seems to stand out as a further refinement of Kayne's analysis is ame-
nable to van Riemsdijk's (1989) explanation of similar German constructions,
 Latin word order in generative perspective 421

which could find an interesting application to Latin discontinuous noun

phrases. In his proposal, in fact, two noun phrases are to be posited in these
structure types, under the assumption that the adjective phrase is allowed to
leave its noun phrase (possibly because Latin has no article blocking the
movement) and to regenerate an empty nominal head in the higher position it
has landed in. Finally, under the pre-K.aynian framework 1 tentatively adhere
to, the question of how to derive discontinuous patterns from an underlying
(S)OV order seems to be more controversial, given that it entails either the
movement of the verb to a position in-between the adjective and the noun con-
stituent aimed to endow the former with some prominence - springing from its
disjunction from the head it depends upon, or a rightward movement of the
nominal head to 'benefit' the preverbal, disjoined adjective and let it receive
the emphasis it is worth of under the exceptional conditions instantiated here.
Owing to its complexity, the problem remains basically open, being worth a
separate research project (Polo: forthcoming).
44. As far as the factors appealed to in 1.1 and 1.2 are concerned, an observation
seems to be in order: the fact that they have been paired under the same list
does not derive from any similarity of nature holding between them, but has
been simply dictated by expository exigencies of conciseness. Actually, the
two strategies fall under two different classes, having to do respectively with a
stylistic device of the literary language (1.1) and a grammatical rule feeding in-
version in periphrastic / restructured constructions (1.2: laudatus est 'praised
was' / est laudatus 'was praised' - Salvi 2004: 1.2.1.4) reminding of parallel
Germanic structures (singen soll 'sing must' / soll singen 'must sing'). I am
perfectly aware that explanations based upon stylistic procedures like chiastic
configurations may fall short of the explanatory adequacy target we have tried
to cope with so far; on the other hand, however, we cannot a priori exclude that
these devices exerted a certain influence on the syntactic organization of the
text, so much so in consideration of the undeniable artistic elaboration it has
passed through.
45. This is frequently found, for example, with the idiom agere gratias which can
surface both with the innovating order V O and with the reversed, ancient one
OV.
46. In reality, as already referred to, a third explanation can be envisaged for the
(few) exceptions to the basic order assumed for Latin; we might think of them
as the by-product of an experimental variable, ultimately imputable to the im-
possibility to access native speakers' judgements, and, as a consequence, to our
imperfect grasp of the rule deciding for the post-verbal placement of the DO.
47. A m o n g the latter, a special mention must be reserved for a quantitatively ap-
parent tendency to postpone to the verb non-prototypical patients, which de-
note referents ranked higher in the animacy, humanity and definiteness scale,
running against a principle of typological consistency (Comrie 1981), as already
422 Chiara Polo

pointed out in Magni (2000) in relation to Pompeian inscriptions. In the Petro-

nian text, in particular, 85% (17/20) of the post-verbal objects attested in un-
embedded contexts and otherwise ruled out in this position designates non-
prototypical referents and, among these, 70% (14/20) is [+ definite]-marked,
with the entailment that the defmiteness feature starts becoming a criterion for
the postposition.
48. In reality, the data I run into differ from those gathered by Magni (2000) in her
study devoted to the Pompeian inscriptions; actually, the picture she came up
with points to a 50% (19/38) of post-verbal objects unmarked for accusative
case, as opposed to the 23% (9/38) which surfaces in pre-verbal position under
the same inflectional conditions. This result clearly contrasts with the excep-
tionless situation documented in Cena Trimalchionis, where late objects pattern
with pre-verbal objects in that they never occur deflected. As far as this incon-
sistency is concerned, however, it might be the case that the different typology
of the written sources taken into account (respectively epigraphic vs literary)
played a role in the morphological distance between two almost contemporary
documents (given that the former dates back to a period in-between 62 and 79
A.D. and the latter to the 1st century AD). These questions call for the extension
of the empirical data-base to a greater variety of Latin texts, picked up from
different genres, to arrive at more conclusive and safer generalizations.
49. A research meant to attain the objectives just sketched is already being carried
out.

References

Adams, James Noel

1976 A typological approach to Latin word order. Indogermanische For-
schungen 81: 70-99.
1994 Wackernagel's law and the placement of the copula Esse in classical
Latin. Cambridge: The Cambridge Philological Society.
Alisova, Tatiana Borisovna
1972 Strutture semantiche e sintattiche della proposizione semplice in
italiano. Firenze: Sansoni.
Aragosti, Andrea
1999 Satyricon, Introduzione, traduzione e note di Andrea Aragosti. Milano:
Biblioteca Universale Rizzoli.
Bauer, Brigitte
1995 The emergence and development of SVO patterning in Latin and
French. Diachronic andpsycholinguistic perspectives. Oxford: Oxford
University Press.
 Latin word order in generative perspective 423

Benincä, Paola (ed.)

1989 Dialect variation and the theory of grammar. Dordrecht: Foris.
Benincä, Paola, Lorenza Frison and Giampaolo Salvi
1989 L'ordine degli element! della frase e le costruzioni marcate. In Grande
grammatica italiana di consultazione. La frase. Vol. 1, Lorenzo Renzi
(ed.), 115-226. Bologna: II Mulino.
Benincä, Paola and Salvi Giampaolo (eds.)
1998 Romance syntax. Budapest: L. Eötvös University.
Berretta, Monica
1994 Ordini marcati dei costituenti di frase in italiano. La frase scissa. Vox
Romanica 53, 8: 79-105.
Butler, Η. E. (ed.)
1980 The 'Institutio Oratoria' of Quintilian, with an English translation by
Η. E. Butler. Cambridge, Mass.: Harvard University Press.
Chomsky, N o a m
1971 Deep structure, surface structure and semantic interpretation. In Se-
mantics: an interdisciplinary reader in philosophy, linguistics and
psychology, Danny D. Steinberg and Leon A. Jakobovits (eds.), 183—
216. Cambridge: Cambridge University Press.
1977 Essays on form and interpretation. Amsterdam: Elsevier North-
Holland.
1981 a Lectures on government and binding. Dordrecht: Foris.
1981b Principles and parameters in syntactic theory. In Explanation in lin-
guistic theory: the logical problem of language acquisition, Norbert
Hornstein and David Lightfoot (eds.), 32-75. New York: Longman
Inc.
1986 Knowledge of language. Its nature, origin, and use. New York:
Praeger.
Cinque, Guglielmo
1977 The movement nature of left-dislocation. Linguistic Inquiry 8 (2):
397—411.
1990 Types of A'-dependences. Cambridge, Mass.: The M I T Press.
Comrie, Bernard
1981 Language universals and linguistic typology. Syntax and morphol-
ogy. Oxford: Basil Blackwell.
Derbyshire, Desmond C.
1977 Word order universals and the existence of OVS languages. Linguistic
Inquiry 8 (3): 590-597.
Duarte, I.
1988 Α costrufäo de topicalizaijäo na gramätica do Portugues: regencia,
ligagäo e condigoes sobre movimento. PhD dissertation, University
of Lisbon.
424 Chiara Polo

Ferraresi, Gisella
1992 Die Stellung des gotischen Verbs im Licht eines Vergleichs mit dem
Althochdeutschen. Unpublished dissertation, University of Venice.
Givön, Talmy
1983 Topic continuity in discourse: quantitative cross-language studies.
Amsterdam Philadelphia: Benjamins.
Graffi, Giorgio
1994 Sintassi. Bologna: II Mulino.
Greenberg, Joseph H.
1963 Some universals of grammar with particular reference to the order of
meaningful elements. In Joseph H. Greenberg (ed.), 7 3 - 1 1 3 .
Greenberg, Joseph H. (ed.)
1963 Universals of language. Cambridge, Mass.: The M I T Press.
Harris, Zellig S.
1960 Structural linguistics. Chicago: University of Chicago Press.
Hinojo, Gregorio
1985 Del orden de palabras en el Satiricon. In Symbolae Lvdovico Mitx-
elena Septvagenario Oblatae, Victoriaco Vasconvm. Jose L. Melena
(ed.), 245-254.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax and Stilistik. München: Beck.
Hornstein, Norbert and David Lightfoot (eds.)
1981 Explanation in linguistic theory: the logical problem of language
acquisition. New York: Longman Inc.
Jackendoff, Ray S.
1974 Semantic interpretation in generative grammar. Cambridge, Mass.:
The M I T Press.
Kayne, Richard S.
1994 The antisymmetry of syntax. Cambridge, Mass.: The M I T Press.
Kroch, Anthony
1989 Reflexes of grammar in patterns of language change. Journal of
Language Variation and Change 1: 199-244.
Kroll, Wilhelm
1918 Anfangsstellung des Verbums im Lateinischen, Glotta 9: 112-123.
Kühner, Raphael and Carl Stegmann
1955 Ausführliche Grammatik der Lateinischen Sprache. II. Satzlehre.
Leverkusen: Gottschalk.
Lambrecht, Knud
1994 Information structure and sentence form. Topic, focus, and the men-
tal representations of discourse referents. Cambridge: Cambridge
University Press.
Larson, Richard K.
1988 On the double object construction. Linguistic Inquiry 19: 335-391.
 Latin word order in generative perspective 425

Larson, Richard K.
1990 Double objects revisited: reply to Jackendoff. Linguistic Inquiry 21:
589-632.
Lightfoot, David
1999 The development of language. Acquisition, change, and evolution.
Oxford: Blackwell.
Longobardi, Giuseppe
1994 La posizione del verbo gotico e la sintassi comparata dei comple-
mentatori germanici: alcune riflessioni preliminari. Miscellanea di
studi linguistici in onore di Walter Belardi: 353-373.
Magni, Elisabetta
2000 L'ordine delle parole nel latino pompeiano: sulle tracce di una de-
riva. Archivio Glottologico Italiano LXXXV-I: 3-37.
Marouzeau, Jules
1922 L'ordre des wots dans la phrase latine. I. Les groupes nominaux.
Paris: Champion Editeur.
1938 L'ordre des mots dans la phrase latine. II. Le verbe. Paris: Les
Belles Lettres.
1948 Quelques vues sur l'ordre des mots en latin. Lingua I: 155-161.
Meillet, Antoine
1903 Introduction α Γ etude comparative des langues indo-europeennes.
Paris, Librairie Hachette. [Reprint 1964. Alabama: University of
Alabama Press],
1908/09 Notes sur quelques faits gotiques. MSLP 15: 73-103.
Melena, Jose L. (ed.)
1985 Symbolae Lvdovico Mitxelena Septvagenario Oblatae, Victoriaco
Vasconvm.
Mobitz, Ο.
1924 Die Stellung des Verbums in den Schriften des Apuleius. Glotta 13:
116-126.
Molinelli, Piera
1986 L'ordine delle parole in latino: studi recenti. Lingua e Stile 21-4:
485^198. Bologna: II Mulino.
Ostafin, David Mark
1986 Studies in Latin word order: a transformational approach. PhD disser-
tation, The University of Connecticut, Storrs.
Panhuis, Dirk
1982 The communicative perspective in the sentence: A study of Latin word
order. Amsterdam: Benjamins.
1984 Is Latin an SOV language? A diachronic perspective. Indogerma-
nische Forschungen. Zeitschrift für Indogermanistik und Allgemeine
Sprachwissenschaft 89: 140-159. Berlin, Germany.
426 Chiara Polo

Pintzuk, Susan
1991 Phrase structures in competition: Variation and change in Old English
word order. PhD dissertation, University of Pennsylvania, Phila-
delphia.
Polo, Chiara
2004 Latin, Italian and Slovene between morphology and syntax. Padua:
Unipress.
forthc. An analysis of low-level discontinuities in Latin. Unpublished Ms.,
University of Padua.
Quintilianus, Marcus Fabius
1980 See under: Η. E. Butler.
Renzi, Lorenzo (ed.)
1989 Grande grammatica italiana di consultazione. La frase. Vol. 1.
Bologna: II Mulino.
Riemsdijk, Henk van
1989 Movement and regeneration. In Paola Benincä (ed.).
Rizzi, Luigi
1998 The fine structure of the left periphery. In Romance syntax, Paola
Benincä and Salvi Giampaolo (eds.), 112-158. Budapest: L. Eötvös
University.
Salvi, Giampaolo
2003 Teoria sintattica e spiegazione diacronica. In Syntaxtheorien: Modelle,
Methoden, Motive, Elisabeth Stark and Ulrich Wandruszka (eds.).
Tuebingen: Narr.
2004 La formazione della struttura di frase romanza. Ordine delle parole e
clitici dal latino alle lingue romanze antiche. Tuebingen: Niemeyer.
Saussure, Ferdinand de
1916 Cours de linguistique generale: publie par Charles Bally et Albert
Sechehaye avec la collaboration de Albert Riedlinger. Paris: Payot.
Schneider, N.
1912 De verbi in lingua latina collocatione. Münster: Monasterii Guest-
falorum.
Stark, Elisabeth and Ulrich Wandruszka (eds.)
2003 Syntaxtheorien: Modelle, Methoden, Motive. Tübingen: Narr.
Steinberg, Danny D. and Leon A. Jakobovits (eds.)
1971 Semantics: an interdisciplinary reader in philosophy, linguistics and
psychology. Cambridge: Cambridge University Press.
Thurneysen, Rudolf
1982 Zur Stellung des Verbums im Altfranzösischen. Zeitschrift für roma-
nische Philologie 16: 289-307.
Tomaselli, Alessandra
1990 La sintassi del verbo finito nelle lingue germaniche. Padova: Unipress.
 Latin word order in generative perspective 427

Wackernagel, Jacob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermani-
sche Forschungen 1: 333-436 (= Wackernagel 1955: 1-104).
Walsh, P. G.
1996 The Satyricon, Translated with introduction and explanatory notes by
P. G. Walsh. Oxford: Clarendon Press.
Zamboni, Alberto
2000 Alle origini dell'italiano. Dinamiche e tipologie della transizione dal
latino. Roma: Carocci.
Zubizarreta, Maria Luisa
1998 Prosody, focus, and word order. Cambridge, Mass.: The MIT Press.
Some firm points on Latin word order:
The left periphery

Giampaolo Salvi

The aim of this chapter is to show how structural and in particular genera-
tive linguistic reasoning can be applied to the study of Latin word order.
Although the possibilities of this research are severly limited by the kind of
data we have at our disposal (sect. 1), some more or less firm conclusions
may be reached: we examine here the left periphery (i.e. the initial section)
of the main clauses (sect. 2) and of the embedded clauses (sect. 3).
The material of this chapter is drawn from ch. 2 and 4 of Salvi (2004)
(written in 1998-99), to which we refer for further details and bibliography.
The analysis presented here is, however, somewhat different and, we hope,
improves on our previous one.
The theoretical frame utilized is a simplified version of Government and
Binding Theory (for which see Graffi 1994).'

1. Methodological introduction

Syntactic evidence that can enable us to reconstruct the sentence structure

of Classical Latin is scanty. As with other dead languages, we do not have
at our disposal native speakers who could supply us with positive or nega-
tive evidence (e.g. about the correct interpretation of a sentence or the
(a)grammaticality of a construction), and we do not have direct prosodic
information which could assist us in the reconstruction of the syntactic
domains of the sentence. But in the case of Latin the situation is more seri-
ous. In order to clarify the nature of the problems involved, let's begin with
an example taken from Medieval Romance, which may also be considered
a dead language (group) as many aspects of its grammar were very differ-
ent from the corresponding ones in its modern varieties.
430 Giampaolo Salvi

1.1. The determination of syntactic positions...

Scholars of Medieval Romance languages have been aware (since Meyer-

Lübke 1897) that one can distinguish no less than two syntactic positions
before the finite verb. This intuition was systematically developed for Old
French by Skärup (1975) in the frame of the positional syntax model of
Diderichsen (1966), and further by generative linguists during the last twenty
years (see Benincä in press). These two positions, one immediately before
the finite verb (P2) and another one before that (Pi), are not always syntac-
tically distinct, but in the case of definite Direct Objects the distinction is
clearly marked: a DO in P2 a) is in complementary distribution with the
Subject, which must appear in postverbal position ( l a ) or in Pi (lb), and b)
is never doubled by a clitic pronoun (lab); in contrast, a definite DO in Pi
is compatible with a preverbal subject and is always doubled (2): 2

(1) a. (Old French) Γ acointance de nous deus (DO)/

the friendship of us both
desirroie (V) jou (S) a avoir
wished I to have
Ί wished us to be friends' (Chrestien de Troies, Perceval 4491)
b. cascuns (S) le pape (DO) encosa (V)
everyone the pope blamed
'Everyone blamed the pope'
(Adam de la Halle, Le Jeu de la Feuillee 435)

(2) Ceste bataille (DO), sire, je (S) la (resumptive clitic) demant

this fight sir I it demand
Ί demand this fight, sir' {Le couronnement de Louis 2446)

Pi and P2 may also be kept distinct by the position of clitic pronouns: if the
preverbal constituent is in Pi, clitics are postverbal (3a); if it is in P 2 , they
are preverbal (3b) - we exemplify with a preverbal DO, so one can notice
that the position of clitics and the presence of a resumptive pronoun both
depend on the position occupied by the preverbal DO: if it is in Pi, we have
a resumptive clitic pronoun, and this is postverbal (3a); if it is in P 2 , we
have no resumptive pronoun, and the other clitics are preverbal (3b): 3
 Some firm points on Latin word order: The left periphery 431

(3) a. Nos qatre dex (DO) ont (V) ies (resumptive clitic) il (S) retenu?
our four gods have them they got
'Have they got our four gods?' (La chanson d'Aspremont 8168)

b. Le cheval (DO) vos (clitic) garderai (V) gie (S)

the horse to-you will-watch I
Ί will watch over your horse' (Chrestien de Troies, Perceval 6730)

1.2. .. .and its application to Latin

In Latin, evidence of this type is much more difficult to find and to evaluate.
Firstly, the use of resumptive pronouns is not very frequent: since zero-
anaphora is also the unmarked type of anaphora between sentences, it is
completely dominant within sentences, where overt anaphora is extremely
rare. We can find some examples, but we cannot automatically draw from
them the same conclusions as from the Romance ones discussed above.
Consider (4a), where the pronoun is doubles the Subject tuus... dolor, we
could think that the Subject occupies Pi because it is doubled, and that the
predicative AP humanus occupies P2 because the resumptive (and pre-
sumably atonic) pronoun immediately follows it (as in (2) and (3b) the
resumptive clitic immediately follows the constituent in P 2 ). But consider
now (4b), where the pronoun ea doubles the Subject urbana plebes: the
Subject is doubled, so it must be in Pi; but why does the resumptive pro-
noun immediately follow the constituent in Pi, in contrast to all our previous
examples (cf. again (2) and (3b))? But if the Subject occupied P 2 , why do
we have a resumptive pronoun (cf. (1) and (3b))? 4

(4) a. tuus autem dolor (S) humanus (Pred) is (resumptive pronoun)

your(N) but grief(N) human it(N)
quidem, sed magno opere moderandus
indeed (is) but strongly to-be-moderated
'But your grief, though it is a kindly weakness, should be kept well
in check' (Cie. Att. XII. 10)

b. sed urbana plebes (S), ea (resumptive pronoun)

but urban(N) populace(N) it(N)
vero praeceps erat de multis causis
in-truth precipitate was for many(Abl) reasons(Abl)
'But the city populace in particular acted with desperation for
many reasons' (Sail. Cat. 37.4)
432 Giampaolo Salvi

These contradictory data can be elegantly explained if one assumes that the
resumptive pronouns in (4a) and (4b) differ in stress and consequently, per-
haps, in grammatical category and syntactic properties: the is of (4a) would
indeed be atonic and, as a weak pronoun, could not immediately follow Pi,
while ea in (4b) (the feminine form of is) would be stressed and of category
NP, so it could occupy P2 and immediately follow Pi. In this way, we can
mantain that, in Latin too, there exists a syntactic position P) with its own
distinguishing properties: a) it cannot be immediately followed by an
atonic pronoun, and b) only constituents in Pi may be doubled. 5

1.3. Strong and weak use of Latin pronouns

This hypothesis would work perfectly if we had some means to distinguish

when a pronoun is atonic, and when it is stressed. But the problem is that in
Latin there is no morphological distinction between the (supposed) atonic
pronouns and the (supposed) stressed ones. As a consequence, we cannot
automatically use the position of the pronouns as a device in the individua-
tion of different syntactic positions, as is possible in Romance, because,
even if they exist, atonic pronouns are not recognizable by themselves.
Notice that we have no problem in demonstrating that stressed forms
existed: in (4b) ea must be stressed because vera is an enclitic word which
needs a preceding stressed word as a host. But we have no independent
direct evidence that atonic forms with their own syntactic properties also
existed.
Some indirect evidence, though, does exist. Other languages with only
one morphological type of personal pronoun admit two uses of them (e.g.
modern German; cf. Cardinaletti 1992, Lenerz 1992): they may be stressed
or unstressed, and this prosodic difference is accompanied by a different
syntactic behaviour (e.g. possible positions) and a different pragmatic use.
The existence of two prosodically different uses of Latin personal pro-
nouns is demonstrated by the very existence of two different Romance
outcomes of the same Latin personal pronoun, where the difference in the
phonetic evolution can be explained on the basis of the stressed or un-
stressed nature of the form: so, e.g., from Latin me we have in Old French
free mei/moi and clitic me, with the normal evolution of Latin stressed e to
the diphthong ei/oi and of unstressed e to [s] (Schwan and Behrens 1932:
13).
 Some firm points on Latin word order: The left periphery 433

Latin pronouns could serve different pragmatic uses: they could have
the textual function of focus, of contrastive topic, of new topic, or they
could be simply anaphoric. These differences in pragmatic function could
well be associated with the prosodic difference just noted - just as in mod-
ern R o m a n c e languages, free pronouns and clitics have different pragmatic
functions (or semantics, as in Calabrese's (1980) analysis, where clitics
have expected referents and free pronouns have unexpected referents).
As for the different syntactic use of Latin personal pronouns, we will as-
sume that differences in stress, in meaning and pragmatic function, and in
syntactic behaviour go hand in hand, and that they distinguish two principal
uses of Latin personal pronouns: strong (use of) pronouns and weak (use
of) pronouns. Strong pronouns were stressed and had the function of focus,
contrastive topic or new topic (unexpected referent); weak pronouns were
unstressed and had a purely anaphoric function (expected referent). Syntac-
tically, strong pronouns could occupy the same positions as nominal NPs,
but weak pronouns were limited to a number of readily definable positions
- as in Old Romance, those described by a form of Wackernagel's Law:
they were enclitic to the first constituent of a specific syntactic domain.
This hypothesis renders the explanation offered above of the contrast in
(4) a plausible one (mainly in the light of the R o m a n c e facts), but notice
that the explanation is not supported by cogent internal evidence as is the
minimal pair in (3) for Medieval Romance. Certainly, if we denied that is
in (4a) and ea in (4b) are of different syntactic categories, we would lose
an interesting explanation of the facts in (4) and a fundamental device for
the analysis of Latin sentence structure. But when we use the position of
Latin pronouns in syntactic argumentation, we must keep in mind that we
are using a hypothesis, so our results on sentence structure below will be,
so to speak, "second degree hypotheses" about Latin syntax.
In order to once again illustrate the hypothetical degree of the analysis:
one can identify the weak forms of the personal pronouns assuming (on the
basis of the pairing form-semantics) that all pronouns that do not function
as topic or focus are purely anaphoric, in other words bona fide weak
forms. This procedure gives relatively good results: the great majority of
the supposedly weak pronous behave in conformity to Wackernagel's Law
(Salvi 1996; see also the discussion in Adams 1994a). But to get this result,
we must first and foremost exclude all the personal pronouns which stay in
sentence initial position without previously questioning their seman-
tic/pragmatic status: the question does not even arise as to whether they are
weak or strong; the fact that they are sentence initial means they can only
434 Giampaolo Salvi

be strong (since, being enclitic, a weak pronoun cannot stay in sentence

initial position).

1.4. Pragmatic arguments and their problems

Besides this syntactic fact, the arguments one can use in the study of Latin
sentence structure are mainly based on the possible pairings of distributions
and pragmatic values: we may assume that marked pragmatic values are
paired with marked word orders and, if we observe the systematic recur-
rence of a given word order with a given pragmatic value, we may suppose
that the observed word order makes use of a special structural position. But
this is not without its problems.
Notice first that in the absence of other structural information, this con-
clusion may be hazardous: nothing assures us that one pragmatic function
is tied to only one structural position or that one structural position is tied
to only one pragmatic function. We may show this with the two syntactic
positions of Medieval R o m a n c e we have identified above on purely syntac-
tic grounds: it is widely acknowledged that constituents with thematic (i.e.,
non-rhematic) value may occur both in Pi (5a) and in P2 (5b) (= (3b))
without any difference in meaning (Salvi 1993) and that constituents occur-
ring in P2 may be not only themes (5b), but also foci (5c): 5

(5) a. (Old French) Ceste bataille (Pi/Theme), veirement la ferum

this battle indeed it we-will-do
' W e will fight this battle, indeed' (Chanson de Roland 882)

b. Le cheval (P2/Theme) vos garderai gie tant com je

the horse to-you will-watch I as-far-as I
le porrai tenir
it can keep
Ί will watch over your horse as long as I can keep it'
(Chrestien de Troies, Perceval 6730)

c. Un molt presc'ieus saintuaire (P2/F0CUS) li a I 'en

a very precious reliquary to-him has one
maintenant fors trait
immediately out taken
'They took out for him a very precious reliquary'
(Chrestien de Troies, Perceval 6195)
 Some firm points on Latin word order: The left periphery 435

Notice that the possible occurrence of constituents with different pragmatic

functions in the same structural position may suggest that the analysis must
be refined, and that there are in fact two positions. Indeed, on a closer ex-
amination of the data, P2 proves to contain at least two syntactic positions,
a theme position (P2A) and a focus position (P2BX in this order, although
ordinarily only one may be realized. One case where the two structural
positions are both occupied by overt constituents is the following: 7

(6) vostre terre, qui desfandra quant Ii rois Artus i vendra...?

your land who will-defend when the king Arthur there will-come
'Who will defend your land when king Arthur comes here?'
(Chrestien de Troies, Yvain 1619)

In the light of our criteria, the theme constituent vostre terre must be in P2
because there is no resumptive pronoun; in (6), therefore, a theme and a
focus co-occur in P2 (vostre terre is in P2A, qui in P2B) - this is only possi-
ble if the focus is a Wh-phrase. 8
From the examples above we may conclude that it is a good policy to
postulate different syntactic positions for constituents with different prag-
matic functions (following E. Kiss 1987, we may think that this is all the
more probable for a language which, thanks to its case system, need not
use syntactic structure in order to mark syntactic functions); but we must
bear in mind that providing syntactic evidence in their support may be very
difficult, if not impossible, in the case of dead languages.
As for the converse problem, if a constituent with a given pragmatic
function may occupy different syntactic positions, the use of pragmatic
function as a criterion for syntactic structure may be only suggestive in the
absence of other criteria.

1.5. Further problems

But there is another problem with the use of pragmatic evidence: individu-
ating the pragmatic values expressed in a sentence implies the perfect re-
construction of the author's communicative intentions, a task which is not
always easy. Even if we assume that the principles which organize the dis-
course are the same universally, there are always cases where two or more
different ways of arranging what we want to say are equally suitable to our
communicative needs. We may see this if we examine different translations
436 Giampaolo Salvi

of the same text: in the sentence reported in (7) the constituent aedes has
been interpreted as a marked topic by the Italian translator (a) (cited and
discussed by Polo 2002: 168) and as part of the rheme by the translator (b):

(7) aedes non somniatur

mansion(Acc) not he-dreams
'[he doesn't] aspire to the mansion' (Petr. Sat. 74.14)
a. ipalazzi non se li sogna nemmeno (A. Aragosti)
b. non si sogna certo un palazzo (G. Reverdito)

Both solutions are defensible on pragmatic grounds, but neither has syntac-
tic evidence that may support it.

1.6. Conclusion

In spite of all this, there is some general agreement between Latin scholars
about the main lines of interpretation of word order phenomena as mani-
festing pragmatic functions - see the standard descriptions in Kühner and
Stegmann (1955: sect. 245-249), Hofmann and Szantyr (1965: sect. 397-
410) and Pinkster (1988: ch. 9).9 So, in our analysis of Latin sentence
structure, we too will make use of these facts, in addition to those concern-
ing the position of weak pronouns.
This rather long introduction was intended to stress the limits of the
evidence available in the study of Classical Latin sentence structure and of
the conclusions we may draw from this evidence. With all this in mind, we
may pass to the analysis.

2. The left periphery in main clauses

2.1. The Topic position

A left peripheral structural position similar to the Romance Pi position can

be identified with the help of a) the position of the weak pronouns, b) the
pragmatic function of the constituents that occupy that structural position,
and c) a special marker for that pragmatic function:
 Some firm points on Latin word order: The left periphery 437

a) on the basis of the position of the enclitic weak pronouns (in non-italics)
we may separate the section of the sentence preceding the host constitu-
ent of the pronoun (marked in the following examples with a vertical
bar);
b) the constituents standing in this section of the sentence pragmatically
function as the Frame (often an adverbial clause, as in (9)) or the Topic
of the sentence; and
c) they may be introduced by the preposition de, meaning 'as for' (8):

(8) a. De Aufidiano nomine \ nihil

about relative-to-Aufidius(Abl) debt(Abl) nothing(Acc)
te hortor
you(Acc) I-urge-on
'In the matter of Aufidius's debt, I put no pressure upon you'
(Cie. Fam. XVI. 19)
b. Iis autem de rebus quas
these(Abl) however about things(Abl) which(Acc)
coniectura consequi possumus \ non mihi sumo
conjecture(Abl) grasp we-can not myself(D) I-assume
ut plus ipse prospiciam quam...
that more(Acc) myself(N) perceive-beforehand than
'As to those matters, however, which can only be the subject of
conjecture, 1 do not presume to claim a deeper insight into the fu-
ture than...' (Cie. Fam. VI.5.2)

(9) a. si proficiscerer ad bellum, \ periculum te meum

if I-left to war(Acc) danger(N) you(Acc) my(N)
commovebat
touched
'if I set out for the scene of war, you were appalled at the thought
of my danger' (Cie. Fam. VII.3.1)
b. ad eorum voluntatem mihi conciliandam \ maximo
to their goodwill(Acc) 1(D) to-be-winned greatest(D)
te mihi usui fore video
you(Acc) 1(D) utility(D) to-be I-see
Ί can see that you will be of the greatest assistance to me in win-
ning their goodwill' (Cie. Att. 1.2.2)
438 Giampaolo Salvi

c. quae, quia tibi sunt propiora quam nobis,

which(N) because you(D) are more-nearly than we(D)
tua me causa magis movent quam mea
your(Abl) I(Acc) interest(Abl) more touch than mine(Abl)
'and [lit.: which] as they touch you more nearly than they do me,
they cause me more agitation on your account than my own'
(Cie. Farn. XII. 17.1)

(10) a. Itaque illud alterum quod dixi litter arum

therefore that(N) other(N) which I-said letter(G)
genus I cottidie mihi, ut spero, fiet proclivius, ut
kind(N) daily 1(D) as I-hope will-become easier so-that
etiam polliceri possim
even promise I-can
'Therefore that second kind of letter-writing which I mentioned
will daily become easier for me, so that I can even make promises'
(Cie. Farn. VI.10b.3)
b. Nam et consilium tuum, vel casus
for on-the-one-hand decision(N) your(N) or misfortune(N)
potius, diutius in armis civilibus commorandi \ semper
rather too-long in war(Abl) civil(Abl) of-remaining always
mihi magno dolori fuit
1(D) great(D) grief(D) was
'For I have been much troubled by your policy, or rather your mis-
fortune, in remaining too long in civil war' (Cie. Fam. VI.lOa.1)

2.2. The Focus position

A further structural position can be identified with the help of a) the posi-
tion of the weak pronouns, b) the pragmatic function of the constituents
that occupy that structural position, and c) their category:
a) the constituent in this position immediately precedes the weak pronouns
(in non-italics);
b) it functions pragmatically as a Focus; and
c) it may contain a Wh-word (11) or a quantifier (12):
 Some firm points on Latin word order: The left periphery 439

(11) a. quid me ista res consolatur in tantis

what(Acc) I(Acc) that(N) thing(N) comforts in so-great(Abl)
tenebris et quasi parietinis rei publicae?
obscurity(Abl) and as-it-were ruined-walls(Abl) Republic(G)
'What consolation is that to me, amid this oppressive gloom, and
what I may call the crumbling walls of the Republic?'
(Cie. Fam. IV.3.2)
b. quo me in silvam venatum vocas?
where I(Acc) into wood(Acc) to-hunt you-summon
'Whither dost thou summon me a-hunting in the woods?'
(PI. Men. 836)

(12) a. nihil te omnino fefellit

nothing(N) you(Acc) absolutely escaped-the-notice-of
'nothing whatever escaped your notice' (Cie. Fam. IX.2.2)

b. Nimia mira mihi quidem hodie exorta

too-much(N) strange-things(N) 1(D) indeed today arose
sunt miris modis
strange(Abl) ways(Abl)
'Well, well, how strangely strange things have happened to me to-
day!' (PI. Men. 1039)

(13) a. ita se cum multis conligavit

so-much himself(Acc) with many(Abl) he-tied
'so inextricably has he tied himself up with his multitude of
counsellors' (Cie. Fam. IX. 17.2)

b. Grata mihi vehementer est memoria nostri

pleasant 1(D) extremely is remembrance(N) we(G)
tua quam significasti litteris
your(N) which(acc) you-indicated letter(Abl)
'Your remembrance of me, as indicated by your letter, gives me
extreme pleasure' (Cie. Fam. XII. 17.1)

Notice that the fact that a constituent stands before the weak pronoun is not
in itself proof that that constituent stands in the structural position identified
in this section, since weak pronouns also attach to constituents which are
not focalized:
440 Giampaolo Salvi

(14) a. Caninius noster me tuis verbis admonuit ut...

C.(N) our(N) I(Acc) your(Abl) words(Abl) reminded that
O u r friend Caninius gave me your message, reminding me t o . . . '
(Cie. Fam. IX.6.1)
Haec tibi antea non rescripsi, non quo...
these-things(Acc) you(D) before not I-wrote not in-order-that
'If I did not send you this reply before, it is not that...'
(Cic.Fam. IX. 17.3)
Ex Syria nobis tumultuosiora quaedam
from Syria(Abl) we(D) rather-turbulent(N) certain-things(N)
nuntiata sunt
were-reported
' W e have reports from Syria of some rather serious disturbances
there' (Cie. Farn. XII. 17.1)
ante te certiorem faciam, ut...
before you(Acc) I-will-inform that
Ί shall [...] give you notice beforehand, so that...'
(Cie. Fam. IX.5.3)

But while non-focussed constituents may also occupy other positions in the
sentence (e.g. ista res in (11a) appears after the weak pronoun), standing
before the weak pronouns is the only possibility for Wh-phrases. Although
the situation with other focussed constituents is not so clear-cut, we may
assume that there is a special structural position for Wh-phrases and other
focussed constituents at the beginning of the sentence (but after the posi-
tion identified in sect. 2.1). Frequently, only a part of the focussed con-
stituent appears in this position, the rest of the phrase standing inside the
sentence (15) or in a right-adjoined position (16) - a situation that is not
usual with non-focussed constituents: 10

(15) Magnam haec res Caesari difficultatem ad

great(Acc) this(N) thing(N) Caesar(D) difficulty(Acc) to
consilium capiendum adferebat
decision(Acc) to-be-taken took
'This action of Vercingetorix caused Caesar great difficulty in forming
his plan of campaign' (Caes. Bell. Gall. VII. 10)
 Some firm points on Latin word order: The left periphery 441

(16) Q u o d te urget scelus...?

what(N) you(Acc) presses curse(N)
' W h a t possesses y o u . . . ? ' (PI. Men. 322)

Notice that if in (15) magnam were restitued into its normal position pre-
ceding difficultatem, the sentence would have the unmarked word order S
(haec res)-IO (Caesari)-DO ( m a g n a m difficultatem)-.. .-V ( a d f e r e b a t ) : we
may conclude from this, too, that magnam moved to occupy a special
structural position preceding that of the subject.

2.3. Verb in initial position

The verb, too, which usually stands in sentence final position (cf. e.g. (15)),
may appear in sentence initial position, in which case weak pronouns im-
mediately follow it (sentence connectives such as enim in (17a) are always
put after the first word of the sentence, so they do not count in the compu-
tation of structural positions):

(17) a. erit enim nobis hones tins [...] videri venisse

it-will-be for we(D) more-creditable to-seem to-have-come
in ilia loca ploratum potius quam natatum
into those(Acc) places(Acc) to-grieve rather than to-swim
' f o r it will be more creditable to us [...] to be thought to have
visited those districts to indulge in sorrow rather than in sea-
bathing' (Cie. Fam. IX.2.5)

b. Miscuerat se legatis Musonius Rufus

had-associated himself(Acc) delagates(D) M.(N) R.(N)
equestris ordinis, Studium philosophiae et
equestrian(G) order(G) study(Acc) philosophy(G) and
placita Stoicorum aemulatus
opinions(Acc) Stoics(G) who-has-studied
' M u s o n i u s R u f u s had joined these delegates. He was a member of
the equestrian order, a man devoted to the study of philosophy and
in particular to the Stoic doctrine' (Tac. Hist. III.81.1)

It is worth noting that the anteposition of the verb never co-occurs with the
anteposition of a focussed constituent. Examples such as the following are
442 Giampaolo Salvi

no exceptions, for in (18) the verb form has not been preposed as in (17),
but it occupies the position of weak forms (Adams 1994b):

(18) An quid est homini Salute melius?

interr. particle what(N) is man(D) Salvation(Abl) better
'Eh? What's better for a man than Salvation?' (PI. Asin. 717)

In other words, the word order focus-verb is only possible with forms of
sum, which, when weak, form a cluster with other weak forms in Wacker-
nagel's position, as we can see in (19):"

(19) a. haec enim ornamenta sunt tibi etiam cum

these(N) for distinctions(N) are you(D) also with
aliis communia
others(Abl) common
'for these are distinctions which others can claim as well as your-
self (Cie. Fam. VI.5.3)
b. non sum ego propter nimiam fortasse
not am I(N) because-of excessive(Acc) perhaps
constantiae cupiditatem adductus ad causam
consistency(G) desire(Acc) been-brought to side(Acc)
Ά perhaps unreasonable regard for consistency prevented me
from going over to his side' (Cie. Pis. 79)

Cases such as (18) not being an exception, we may conclude that, since
focussed constituents and preposed verbs are in complementary distribu-
tion, they occupy the same structural position. But in current generative
theory, where these two types of elements belong to different categories,
this latter conclusion is not directly derivable: focussed constituents are
phrases, while verbs are heads, so they cannot occupy the same syntactic
position. Another, more abstract explanation must be found.

2.4. The architecture of syntactic structure...

Situations of this type are not uncommon: in many languages Wh-phrases

and complementizers are in complementary distribution, as in the following
Italian examples:
 Some firm points on Latin word order: The left periphery 443

(20) a. (Italian) Mi ha detto quale problema devi risolvere

1(D) he-has told which problem you-must solve
' H e has told me which problem you must solve'

b. Mi ha detto che devi risolvere un problema

1(D) he-has told that you-must solve a problem
' H e has told m e that you must solve a problem'

c. *Mi ha detto quale problema che / che quale problema

1(D) he-has told which problem that that which problem
devi risolvere
you-must solve

The theory assumes that in these cases we have two distinct structural posi-
tions and that in Italian only one of these may be occupied by phonetically
realized elements. The structure underlying (20ab) is therefore (21a), an
instance of the more general structure (21b) on which every wider syntactic
structure is based (cfr. Chomsky 1986; C = complementizer, Spec = speci-
fier, C o m p l = complement):

(21) a. [c-SpecC" [ c C [...]]]

b. b c S p e c X " [χ'Χ Compl]]

The Wh-phrase quale problema of (20a) occupies the specifier position

where only phrases stand, while the complementizer che of (20b) occupies
the head position C, as represented in (22):

(22) a. [c"[sPecC" = N" quale problema] [ c [ c 0 ] [•••]]]

b. [ C "[specC"0] [c-[c che] [...]]]

The correctness of this hypothesis is confirmed by the fact that there are
languages where both positions may be occupied, as in substandard Italian
(23a), to which we may assign the (partial) structure (23b):

(23) a. Non so a chi che pensa

not I-know to who that he-thinks
Ί d o n ' t know about whom he is thinking'

b. [c"[specc = P" a chi] [ c [ c che] [...]]]

444 Giampaolo Salvi

2.5. .. .and its application to Latin word order

To return to the Latin sentences (11)—(13) vs. (17), we can explain this
particular complementary distribution of focussed constituent and verb
along the same lines. In particular, following Brody's (1990) analysis of
Hungarian sentence structure, we may assume that Latin had a special
functional projection F(ocus) into whose specifier focussed constituents
were moved, in which case the head position of the projection remained
empty:

(24) [HspecF-X"] [ F ' [ F 0 ] [l"...V]]]

When the verb was preposed, it occupied the head position of the func-
tional projection, while the specifier position remained empty:

(25) [F"[specF"0] H f V ] [,"-..]]]

(Since verb-preposing always had a special semantic import in Latin, we

may assume that the specifier position was occupied by one of a series of
abstract operator(-phrase)s with different possible values: jussive, conces-
sive, assertive, etc. - we will not go into this matter here, but see Salvi
2004: ch. 2).
Notice that this analysis, although more complex, predicts that there
may exist varieties where both positions, the specifier and the head, are
occupied by phonetically realized elements, as in the Italian case studied
above. In this specific case, the variety in question is a diachronical variety
of Classical Latin, Late Latin, where examples such as the one in (26) were
possible:

(26) quid administrabat tunc Silvanus in clero?

what(Acc) managed then S.(N) in community(Abl)
'What did Silvanus manage at that time in the community?'
0Gesta 192.22)

In (26) a focussed constituent (quid) and the verb (administrabat) co-occur

in sentence initial position, an innovation with respect to Classical Latin
which would eventually give rise to the V2 word order of Early Romance.
 Some firm points on Latin word order: The left periphery 445

2.6. Conclusion

We can thus conclude that in the left periphery of the Latin sentence struc-
ture we must minimally distinguish two positions: one first position for
topic-like constituents, with the properties examined in 2.1, and one second
position for focus-like constituents, with the properties examined in 2.2.
More accurately, since we may have more than one topic-like constituent at
the beginning of the sentence (as in (27), where the constituents are num-
bered), we must assume that the first position is a recursive one:

(27) a. (i) Ego si te intellegerem plus conatum esse

I(N) if you(Acc) realized more to-have-attempted
suscipere reipublicae causa muneris quam
to-undertake state(G) for-the-sake-of task(G) than
quantum praestare potuisses, tamen, (ii) quibuscumque
how-much(Acc) accomplish you-could still whatever(Abl)
rebus possem, \ ad earn condicionem te
things(Abl) I-could to that(Acc) condition(Acc) you(Acc)
vivendi quae daretur, quaecumque esset, hortarer
of-living which(N) were-offered whatever it-were I-should-urge
'If I believed that you had attempted in the interests of the state to
undertake a task beyond your power to accomplish, 1 should still
urge you, to the best of my ability, to accept such terms of life as
were offered you, and were available' (Cie. Farn. VI.22.2)

b. (i) Nunc, (ii) ut ad te antea scripsi, (iii) si ad nos

now as to you(Acc) beforehand I-wrote if to we(Acc)
veneris, \ consilium totius rei capiemus
you-will-have-come advice(Acc) whole(G) matter(G) we-will-take
'Now, as I said in my last letter, if you will come I can take your
advice on the whole matter' (Cie. Att. III.2)

Moreover, if we accept that syntactic structures are built of substructures of

the form represented in (21b), we must assume that the two phrase-
positions we have identified are paired with a head position; whilst the
head position of the Top(ic) projection is always empty, that of the F(ocus)
projection may be occupied by the verb, as discussed above. We have thus
the following structure (where Top" is recursive):

( 2 8 ) [Top-tspecTop-X"] [ τ ο ρ ' [ τ ο ρ 0 ] H s p e c H X " ) ] H F ( V ) ] [,•'...]]]]]

446 Giampaolo Salvi

3. The left periphery in embedded clauses

3.1. Basic facts and generalizations

Subordinate clauses are normally introduced by a complementizer (29) or a

Wh-phrase, interrogative (30a) or relative (30b); in certain cases the com-
plementizer ut may be omitted (31):

(29) a. ut te hinc abducat

so-that you(Acc) from-here she-lead-away
'so that she may bring you in' (PI. Men. 332)
b. nisi me ad has exercitationes rettulissem
if-not myself(Acc) to these(Acc) exercises(Acc) I-had-taken-again
'had I not taken to this form of exercise again' (Cie. Fam. IX. 18.3)
c. cum me hospitio recipias
when I(Acc) guest-accomodation(D) you-receive
'in receiving me as a guest' (Cie. Fam. IX. 16.7)

(30) a. quo me coniectura ducat

where I(Acc) conjecture(N) leads
'where my conjectures lead me' (Cie. Fam. IX.2.4)
b. qui se domo non commoverunt
who themselves(Acc) home(Abl) not stirred
'who never stirred from home' (Cie. Fam. IX.5.2)

(31) Tu ad me velim litteras crebrius mittas

you(N) to I(Acc) I-would-want letter(Acc) more-frequently send
Ί should be glad if you would write to me more frequently'
(Cie. Fam. V.6.3)

The introductory element of the embedded clause may be preceded by a

constituent which pragmatically functions as a topic (32):

(32) a. Ego quanti te faciam semperque fecerim

I(N) how-much(G) you(Acc) esteem and-always have-esteemed
'How highly I esteem and always have esteemed you'
(Cic.Fam. VI. 10.1)
 Some firm points on Latin word order: The left periphery 447

b. Haec quamquam nihilo meliora sunt nunc atque etiam

these-things(N) although no better are now and even
multo desperatiora
much more-hopeless
'Although the present situation shows no improvement whatever
and is even far more hopeless than it was' (Cie. Fam. VI.22.1)

But constituents with the function of topic/frame may also follow the in-
troductory element of the subordinate clause (33); moreover, the introduc-
tory element may be followed by a focus element (34); and topic/frame and
focus constituents may appear together (35) (=(9c)):

(33) a. ut, si res a nobis abisset, liberti nostri essent...

that if estate(N) from we(Abl) passed freedmen our they-were
'that if my estate passed out of my hands they were to be my
freedmen...' (Cie. Fam. XIV.4.4)

b. qui, cum tu a me rebus

who(N) though you(N) by I(Abl) things(Abl)
amplissimis atque honorificentissimis ornatus esses,
most-numerous(Abl) and most-honourable(Abl) were-honoured
timuissent ne...
were-afraid that
'who, though I had complimented you by giving you the most
handsome and honourable commissions, had shown themselves
afraid o f . . . ' (Cie. Fam. V.2.1)

(34) a. ut ad eas te refer as art is

that to those(Acc) yourself(Acc) take-back accomplishments(Acc)
quibus...
which(D)
'to return to those accomplishments to which...' (Cie. Fam. IV.3.3)

b. qui nulla sibi subsidia ad omnis vitae

who(N) no(Acc) themselves(D) resources(Acc) to all(Acc) life(G)
status paraverunt
circumstances(Acc) provided
' w h o have provided themselves with no resources against any
vicissitudes of existence' (Cie. Fam. IX.6.4)
448 Giampaolo Salvi

(35) quae, quia tibi sunt propiora quam nobis, j

which(N) because you(D) are more-nearly than we(D)
tua me causa magis movent quam mea
your(Abl) I(Acc) interest(Abl) more touch than mine(Abl)
'and [lit.: which] as they touch you more nearly than they do me, they
cause me more agitation on your account than my own'
(Cie. Farn. XII. 17.1)

The examples above show that of the possible permutations of topic/frame

constituents (T), focus constituents (F) and introductory elements of the
embedded clause (C), only two are permitted (for apparent exceptions see
3.5 below):

CTF: ex. (35) (cf. (33), too)

TCF: ex. (32b)

The permissible permutations are even fewer if we consider the different

types of introductory elements separately: only complementizers (c) are
found in both permutations, as one can see in (33a) and (32b). Relative
Wh-phrases (r) admit only permutation CTF (ex. (33b), (35)), while inter-
rogative Wh-phrases (/) do not co-occur with focussed constituents, so the
only possible variation is TC (ex. (32a)). We therefore have the following
variations:

cTF / TcF // rTF // Ti

From these generalizations we may conclude that:

a) interrogative Wh-phrases and focussed constituents occupy the same
structural position (being in complementary distribution), i.e. SpecF";
b) the order of topic/frame and focussed constituents is the same as in main
clauses, i.e. Top"-F";
c) relative Wh-phrases occupy a position which precedes that of topic/
frame constituents (Top").

The position of the complementizer cannot however be established univo-

cally, for it can both precede and follow the topic/frame constituents.
 Some firm points on Latin word order: The left periphery 449

3.2. A structural hypothesis...

This situation is reminiscent of what we find in some well-studied modern

languages, e.g. Italian: left-dislocated constituents follow relative Wh-
phrases (36), but precede interrogative Wh-phrases (37); they may both
follow and precede the complementizer: they follow it in the case of the
finite complementizer che (38a), and they precede it in the case of the in-
finitive complementizer di (38b):

(36) Laragazza a cui questo libro nonlo regalerei mai...

the girl to whom this book not it(Acc) I-would-give ever
'The girl I would never give this book t o . . . '

(37) Non so questo libro chi lo potrebbe leggere

not I-know this book who it(Acc) could read
Ί wonder who could read this book'

(38) a. Credo che questo libro / ' ho gia letto

I-believe that this book it(Acc) I-have already read
Ί believe I have already read this book'
b. Credo questo libro di averlo gia letto
I-believe this book to have-it(Acc) already read
'id:

Since the left-dislocated constituents may both precede and follow the ele-
ments introducing the embedded clause, a univocal serialization may not
be established if the subordinators are placed in only one position. Rizzi
(1997) resolves this problem by postulating that the functional projection
responsible for subordination is composed of different layers: the specifier
of each layer can host only a specific type of constituent (relative Wh-phrase,
topic/frame constituent, interrogative Wh-phrase), while the complementizer
may appear in each one of the heads of the different layers (modulo other
independent principles of grammar which regulate the grammatical combi-
nations). If we assume the existence of three layers (Ci", Top" and C 2 "), we
get the structure (39) and the analyses in (40) for the sentences in (36)—(38):

(39) [cr'SpecCi" [crCi [ To p"SpecTop" [Τθρ·Τορ [C2»SpecC2" [ C 2 'C 2 [,<·...]]]]]]]

450 Giampaolo Salvi

(40) a. [crtspecCv>a cui] [ c r [ c i 0 ] [Top"[sPecToP-questo libro]

[τορ'[τορ0] [C2"· ··]]]]] (=(36))
b· [Top"[SpecTop"?we5to libro] [τορ'[τορ0] [c2"[specC2"C/zz]
[c2'[c20] [r··•]]]]] (=(37))
C. [ci" [cr[ciChe] [Top"[specTop"questo libro] [το Ρ '[το Ρ 0]
[c2"· · ·]]]]] (= (38a))
d. [Top"[SpecTop'^We5to libro] [τορ'[τορ0] [c2"[sPecC2"0]
[ c r M t i - · ••]]]]] (= (38b))

3.3. ... and its application to Latin

A parallel solution may be envisaged for Latin (with F" instead of C2"): cf.
(41), an extension of structure (28), and the analyses for (32)—(35) in (42):

(41) [ c ·.SpecC" [c'C [ To p"SpecTop" [ T o p T o p [ρ-SpecF" [ f F [ R ...]]]]]]]

(42) a. [Top"[specTop"ego] [τ ορ '[τορ0] [F"[s P ecF-quanti] [F [F0] [r···]]]]]

(= (32a))
b. [Top"[specTop"^ec] [jop'[jopquamquam] [F"···]]] (= (32b))
C. [c"[c'[c«f] [Top"[specTop"^' res...] [τορ'[τορ0] [r···]]]]] (= (33a))
d. [c"[specc"9"i] [c'[c0] [Top"[SpecTop"CMW til... [τορ'[τορ0] [γ· •·]]]]]
(= (33b))
e. [ c " [ c ' M [HspecF-adeas] H F 0 ] [Γ·•·]]]]] (= (34a))
f. [c-[specc"9«/] [ c [ c 0 ] [F"[specF-nulla] [ F [ F 0 ] [ r . . . ] ] ] ] ] (= (34b))
g- [c"'[specc"<?wae] [c'[c0] [Top"[sPecToP-quia tibi...] [τορ'[το ρ 0]
H s p e c F M [F'[F0] [R...]]]]]]] (=(35))
Much of the analysis above is certainly stipulative (e.g. the choice of where
to put the complementizer in (42b,e)), and a hypothesis must be worked out
which rules out the ungrammatical combinations (e.g. the co-occurrence of
a wh-phrase and a complementizer). But in spite of this, our hypothesis
accounts for some generalizations of Latin word order in embedded sen-
tences, capturing some important restrictions on possible combinations and
showing how Latin adapts otherwise universal structures to its grammar.
 Some firm points on Latin word order: The left periphery 451

3.4. The position of weak pronouns

In the preceding discussion we did not consider the position of weak ele-
ments, a type of evidence which played an important role in our discussion
of word order in main clauses. In sect. 2 we saw that the domain inside
which weak elements occur exclude the sentence initial topic/frame con-
stituents (i.e. Top"): weak pronouns appear after the first constituent of F"
(focussed constituent (11)-(13) or preposed verb (17)) and, if F" is not
present, after the first constituent of I" (as in (14)). We therefore expect
them to appear in similar positions in embedded clauses, too.
This expectation is only partially fulfilled. When SpecF" is present, we
find the weak pronouns after the focussed constituent, as in (30a), (32a),
(34) and (35); also the examples in (29) would conform to our expectations
if we assumed that there the complementizer occupies the head F. But the
type represented by (30b), where the weak pronoun follows a relative Wh-
phrase, is not expected, as the host of the weak pronoun is outside the F"
domain (the relative Wh-phrase being in SpecC"). Likewise in the follow-
ing examples the weak pronouns precede the focussed constituent and are
hosted by a complementizer outside the domain of F" (43a) or by a relative
Wh-phrase in SpecC" (43b):

(43) a. quamquam me non ratio solum (Focus) consolatur...,

although I(Acc) not reasoning(N) only comforts
sed etiam...
but also
'although I am comforted not only by rational reflection..., but
also...' (Cie. Fam. VII.28.3)

b. qui mihi soli (Focus) videntur vim virtutis

who(N) 1(D) alone(N) seem principle(Acc) virtue(G)
tenere
to-grasp
'who alone appear to me to grasp the true meaning of virtue'
(Cie. Fam. IX. 16.5)

We must thus assume that in embedded clauses the domain for the place-
ment of weak pronouns may be extended from F" to C". This is confirmed
by examples such as the following, where the pronoun eum precedes a
topic/frame constituent, supposedly in SpecTop":
452 Giampaolo Salvi

(44) ut eum, si velint (Topic), describant ad teque mittant

that it(Acc) if they-want they-copy to you(Acc)-and they-send
'(I will tell your people,) if they are so inclined, to copy it out, and
send it to you' (Cie. Fam. XII. 17.2)

This conclusion perhaps finds its rationale in the fact that the embedded
clause may be considered as a whole (as C") or in its core part (as F"): both
domains may serve as the placement domain for weak pronouns (but not
the intermediate domain Top", either in embedded or in main clauses).

3.5. Unsolved problems

In the light of the analysis sketched above, the treatment of the following
examples is not straightforward: 12

(45) a. hue enim ut venirem superior longius

here indeed in-order-that I-arrived earlier(N) further
quam volui fluxit oratio
than I-wanted ran words(N)
'And indeed it was to arrive at just this point that my earlier re-
marks have run on further than I intended' (Cie. Fam. IX.6.4)

b. Mortem mihi cur consciscerem causa non visa est,

death(Acc) myself(D) why I-inflicted reason(N) not was-seen
cur optarem multae causae
why I-desired many(N) reasons(N)
'Why I should contrive my own death there seemed no reason;
why I should pray for it there were many' (Cie. Fam. VII.3.4)
c. Curio ad focum sedenti magnum auri pondus
Curius(D) near fire(Acc) sitting(D) great(Acc) gold(G) mass(Acc)
Samnites cum attulissent, repudiati sunt
Samnites(N) when had-brought they-were-rejected
'When the Samnites had brought him a great mass of gold as he
sat before the fire, [Curius] declined their gift with scorn'
(Cie. De sen. 56)

In (45a), where a focussed constituent {hue) precedes the complementizer

ut, we might assume that ut occupies the head of the projection F" in whose
 Some firm points on Latin word order: The left periphery 453

specifier the focussed constituent is placed. In (45b) the weak pronoun mihi
is placed between the topic/frame constituent mortem and the interrogative
Wh-phrase cur. the placement domain for weak pronouns in embedded
clauses could perhaps be extended to Top", besides C " (contrary to our
hypothesis in 3.4). But ex. (45c) is resistant to every kind of adjustment of
our general approach because the complementizer cum apparently occupies
a position too low in the structure: the constituent Curio ad focum sedenti
may well be a topic (in SpecTop"), magnum auri pondus is probably fo-
cussed (it is anyway rhematic and stands therefore in SpecF" or in I"),
Samnites, too, is rhematic (in I"), so in this clause the complementizer cum
seems to be placed in the interior of I". With a more c o m m o n word order
cum would have been in C (as in (33a)/(42c): cum Curio ad focum sedenti
magnum auri pondus Samnites attulissent) or in Top (as in (32b)/(42b):
Curio ad focum sedenti cum magnum auri pondus Samnites attulissent).
Looking at this example from a different point of view, we may notice
that in (45c) the subordinator is strikingly placed before the verb at the end
of the clause, and this analysis might also be extended to (45ab), although
these clauses are too short for us to draw firm conclusions from them. This
construction is possible not only with complementizers (45ac), but also
with interrogative (45b) and relative (46) Wh-words: 1 3

(46) a. Domi servi qui sunt

at-home servants who(N) are
' w h o are servants in the house' (PI. As. 237)
b. Patronus qui ν ob is fuit futurus
patron who(N) you(D) would-have-been
' w h o would have freed you and been your patron' (PI. As. 621)

Examples (46), from Early Latin, are all the more striking because, other-
wise, relative Wh-phrases are always the first constituent in embedded
clauses. In all probability we have here an alternative form of the embed-
ded clause, a relic of the archaic age where the movement of constituents to
the front of an embedded clause was freer and, unlike Classical Latin, was
also possible in relative clauses (Watkins 1993: 9.3.1).
454 Giampaolo Salvi

Notes

1. Unless otherwise indicated, the examples in this chapter were collected by us.
The non-literal translations of Latin examples are those of Loeb's Classical
Library. Abbreviations used in the glosses: Abl(ative), Acc(usative), D(ative),
G(enitive), N(ominative). 1 am grateful to Katalin E. Kiss and to Alessandro
Parenti for helpful discussion.
2. Ex. ( 1 H 2 ) from Skärup (1975: 192, 431, 440).
3. Ex. (3) from Skärup (1975: 372, 191).
4. Ex. (4) from Fraenkel (1933 [1964]: 352 [127]) and Jones (1991: 86).
5. Doubling may make use of a stressed pronoun too, a possibility we have not
stated explicitly above.
6. Ex. (5c) from Skärup (1975: 195).
7. Ex. (6) from Skärup (1975: 438).
8. This would be, I suppose, Benincä's (in press) analysis, which develops
Rizzi's (1997) approach, although she does not cite examples of the type of
(6). Alternative analyses are possible, but we do not develop on this problem
here, as it does not concern the main topic of this study.
9. A note of caution is in order here: the pairing of linear word order arrange-
ments and pragmatic functions is traditional in the literature on Latin syntax
(and in studies written in the frame of the functional approach of the Prague
School), but these studies do not in general try to give a structural explanation
for the complex interaction of pragmatic and syntactic facts as happens in the
generative approach and in the present study. For a critique of the traditional
approach to diachronic explanation in syntax and an evaluation of the function
of syntactic theory in the diachronic domain cf. Salvi (2003).
10. Ex. (15) from Ostafin (1986: 6.2).
11. Ex. (19b) from Adams (1994b: 39).
12. Ex. (45c) from Jones (1991: 84).
13. Ex. (46) from Somers (1991).

References

Adams, James N.
1994a Wackernagel's Law and the Position of Unstressed Personal Pronouns
in Classical Latin. Transactions of the Philological Society 92: 103—
178.
1994 b Wackernagel's Law and the Placement of the Copula Esse in Classi-
cal Latin. Cambridge: The Cambridge Philological Society.
Benincä, Paola
In press On the functional structure of the left periphery: evidence from
medieval Romance. Probus.
 Some firm points on Latin word order: The left periphery 455

Brody, Michael
1990 Remarks on the Order of Elements in the Hungarian Focus Field. In
Approaches to Hungarian, Vol. 3: Structures and Arguments, Istvän
Kenesei (ed.), 95-121. Szeged: JATE.
Calabrese, Andrea
1980 Sui pronomi atoni e tonici delPitaliano. Rivista di Grammatica
Generativa 5: 65-116.
Cardinaletti, A n n a
1992 On Cliticization in Germanic Languages. Rivista di Grammatica
Generativa 17: 65-99.
Chomsky, N o a m
1986 Barriers. Cambridge, Mass.: The MIT Press.
Diderichsen, Paul
1966 Logische und topische Gliederung des germanischen Satzes. In
Helhed og Struktur, 52-63. Kebenhavn: Akademisk Forlag.
Fraenkel, Eduard
1933 Kolon und Satz: Beobachtungen zur Gliederung des antiken Satzes
II. Nachrichten der Göttinger Gesellschaft der Wissenschaften. Phi-
losophisch-historische Klasse 319-354.
[1964] [reprinted in: Kleine Beiträge zur klassischen Philologie. Vol. I.
Roma: Edizioni di Storia e Letteratura, 93-130.]
Giacalone Ramat, Anna and Paolo Ramat (eds.)
1993 Le lingue indoeuropee. Bologna: il Mulino.
Graffi, Giorgio
1994 Sintassi. Bologna: il Mulino.
Haegeman, Liliane (ed.)
1997 Elements of Grammar. Dordrecht: Kluwer.
Hofmann, Johann Baptist and Anton Szantyr
1965 Lateinische Syntax und Stilistik. München: Beck.
Jones, F.
1991 Subject, Topic, Given and Salient: Sentence-Beginnings in Latin.
Proceedings of the Cambridge Philological Society 37: 81-105.
E. Kiss, Katalin
1987 Configurationality in Hungarian. Budapest: Akademiai kiado.
Kenesei, Istvän (ed.)
1990 Approaches to Hungarian. Vol. 3: Structures and Arguments.
Szeged: JATE.
Kühner, Raphael and Carl Stegmann
1955 Ausführliche Grammatik der Lateinischen Sprache. II: Satzlehre.
Leverkusen: Gottschalk.
Lenerz, Jürgen
1992 Zur Syntax der Pronomina im Deutschen. Sprache und Pragmatik
1-54.
456 Giampaolo Salvi

Meyer-Lübke, Wilhelm
1897 Zur Stellung der tonlosen Objektspronomina. Zeitschrift für ro-
manischen Philologie 21:313-334.
Ostafm, David Mark
1986 Studies in Latin Word Order: A Transformational Approach. Unpub-
lished Ph.D. dissertation, University of Connecticut, Storrs.
Pinkster, Harm
1988 Lateinische Syntax und Semantik. Tübingen: Francke.
Polo, Chiara
2002 Word Order in Latin, Italian and Slovene between Morphology and
Syntax. Unpublished Ph.D. dissertation, University of Padua.
Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar,
Liliane Haegeman (ed.), 281-337. Dordrecht: Kluwer.
Salvi, Giampaolo
1993 Ordine delle parole e struttura della frase nelle lingue romanze an-
tiche. In Grammatikographie der romanischen Sprachen, Christian
Schmitt (ed.), 455^177. Bonn: Romanistischer Verlag.
1996 From Latin Weak Pronouns to Romance Clitics. Budapest: MTA
Nyelvtudomänyi Intezete.
2003 Teoria sintattica e spiegazione diacronica. In Syntaxtheorien. Modelle,
Methoden, Motive, Elisabeth Stark and Ulrich Wandruszka (eds.),
291-309. Tübingen: Narr.
2004 La formazione della struttura di frase romanza. Ordine delle parole
e clitici dal latino alle lingue romanze antiche. Tübingen: Niemeyer.
Schmitt, Christian (ed.)
1993 Grammatikographie der romanischen Sprachen. Bonn: Romanis-
tischer Verlag.
Schwan, Edmond and Dietrich Behrens
1932 Grammaire de l'ancien frangais. Leipzig.
Skärup, Povl
1975 Les premieres zones de la proposition en ancien franqais. Essai de
syntaxe de position. Kobenhavn: Akademisk Forlag.
Somers, Maartje H.
1991 Theme and Topic: The Discourse Function of Initial Contituents in
Latin. [Paper presented at the VI th International Colloquium on Latin
Linguistics, Budapest.]
Stark, Elisabeth and Ulrich Wandruszka (eds.)
2003 Syntaxtheorien. Modelle, Methoden, Motive. Tübingen: Narr.
Watkins, Calvert
1993 II proto-indoeuropeo. In Le lingue indoeuropee, Anna Giacalone
Ramat and Paolo Ramat (eds.), 45-93. Bologna: il Mulino.
Classical Sanskrit, "wild trees", and
the properties of free word order languages

Brendan Gillon and Benjamin Shaer

In this study, we investigate the word order of Classical Sanskrit, making use of
two searchable corpora and taking as our theoretical point of departure Staal's
(1967) "wild tree" proposal for the language. We show that this proposal, accord-
ing to which the daughters of a phrase are unordered in the syntax, can capture
much, but not all, of the attested word order variation in the language. We also
show, however, that these additional word orders can still be captured on a "wild
tree" analysis that (i) incorporates a "flatter" treatment of sentential structure, in
which the verb does not project a VP; and (ii) recognizes that long-distance de-
pendencies make a contribution to word order freedom distinct from that related to
"wild trees" themselves.

The linguistic study of Sanskrit, Ancient Greek, and Latin, it might go

without saying, has a very long history, and yet key aspects of these lan-
guages remain poorly understood. One aspect in particular is their word
order. The minor role that this has played in the study of these languages
seems to have some obvious sources, though this study's lack of theoretical
sophistication is not one of them. Indeed, the grammar of Ancient Greek
produced by Dionysius of Thrax (2nd century BCE) and the Latin gram-
mars based on it, produced by Donatus and Priscian (4th and 5th centuries
CE, respectively), already made use of such concepts as parts of speech,
case, and government, and accordingly provided deep insights into the
workings of language. And the Astädhyäyl [literally 'of eight chapters'],
written or compiled by Pänini in the 4th century BCE, is acknowledged not
only as the earliest extant grammar and the source of such concepts as the-
matic roles, but as the earliest generative grammar, providing a procedure
for constructing an infinite set of well-formed Sanskrit sentences from a
basic set of lexical units. What is nevertheless at most implicit in Pänini's
grammar and lacking altogether in the Latin and Greek grammars is a no-
tion of phrase structure, which in modern generative research has been
basic to the study of word order patterns both within and across languages.
458 Brendan Gillon and Benjamin Shaer

In this chapter, we shall be exploring this understudied aspect of Classical

Sanskrit, describing its phrase structure and showing how this can be re-
lated to the phenomenon of "displaced" phrases, long recognized as "an
irreducible fact of human language" (Chomsky 1995: 403). As such, this
study (which develops research reported in Gillon 1996, 2004) stands
alongside other contemporary generative studies of Sanskrit, including
Schäufele (1990) and Hale (1996), which have also considered these issues.
What, however, distinguishes this study from others is its use of two
searchable corpora, assembled and hand-tagged by the first author. These
corpora consist, respectively, of the prose examples given in Apte's (1885
[I960]) grammar of Sanskrit, which are drawn from Classical Sanskrit lit-
erature; and almost 1500 prose sentences from the "self-commentary" of
the Indian philosopher DharmakTrti (7th century CE) on one chapter of his
verse work Pramänavärttika. Since the basic generalizations about Sanskrit
word order that we offer below derive from a systematic examination of
these corpora, we can vouch for their robustness, however they are ulti-
mately analysed.
As regards the theoretical point of departure for our investigation, this is
Staal (1967), the first generative linguistic study of word order in the lan-
guage. What makes this study an intriguing one is its characterization of
Sanskrit phrase structure in terms of "wild trees", in which the daughters of
a given node constitute a set of unordered nodes. This means not only that
"wild trees" significantly underdetermine the linear ordering of a sentence's
terminal nodes but also that a sentence simply has no "basic" word order.
Such a perspective on displacement is, of course, rather different from that
of recent Minimalist work, which derives a great number of attested word
orders through movement operations. It is likewise different from the per-
spective adopted by many generative studies of free word order languages
over the years, including those in this volume, in not taking constituents to
be rearranged through "scrambling" operations - a common feature of such
work ever since Ross (1967) proposed them for Latin word order patterns.1
At the same time, Staal is hardly the only, or even the first, scholar to have
recognized the utility of distinguishing between hierarchical and linear
ordering of syntactic elements. Such an idea, as Gazdar et al. (1985: 47, 55)
have noted, can already be found in Curry (1961), and has been highly in-
fluential in the development of contemporary phrase structure grammars
(see e.g. Kathol 2000: 23-46 for discussion). And, of course, it is also a
prominent feature of Minimalist research, where efforts to remove linear
ordering from the description of phrase structure and to make this "part of
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 459

the phonological component" rather than the syntax itself (Chomsky 1995:
413) strongly echo Staal's concerns, despite otherwise great differences
between these approaches.
More to the point, a "wild tree" analysis permits a very natural descrip-
tion of many of the word orders attested in Classical Sanskrit. Yet, as Staal
(1967: 34) himself acknowledges, such an analysis by itself cannot capture
the full range of word orders attested even for simple transitive sentences in
Classical Sanskrit - in particular, capturing neither those due to long-distance
dependencies nor those due to verb fronting. What we shall show, however,
is that the basic "wild tree" claim can be readily extended to account for the
latter orders and further supplemented to account for the former. These
revisions preserve Staal's insight that the ordering of sister nodes within a
Classical Sanskrit phrase is essentially free, but (following his own sugges-
tions) admit into the language operations that underwrite dependencies
between phrases "longer-distance" than those attributable to "wild trees"
themselves.
Of course, a "wild tree" approach to Classical Sanskrit can also be seen
to have more general consequences for the analysis of word order permuta-
tions across languages. While it is true that at least some of the word order
patterns that this approach seeks to capture are particular to languages with
substantial word order freedom, others, such as that involving verb front-
ing, are found in a much greater range of languages. Thus, the proposal to
capture this and other patterns without invoking a "basic" structure repre-
sents a distinct mechanism for deriving such patterns. The question that
arises, then, is whether this mechanism should be understood as coexisting
in Universal Grammar with others that derive the same ordering effects or
as a theoretical alternative to approaches that recognize little or no variation
in base word orders and make prolific use of movement to derive attested
word orders. Sharpening our own interest in this question are two different
observations about word order. One, which pertains specifically to Sanskrit,
is that certain preferred or typical word orders have commonly emerged
from the examination of Sanskrit texts. The other, not unrelated, pertains to
the substantial body of research on "information packaging" (see e.g. Vall-
duvi 1993 for a survey). This research has revealed that truth-conditionally
equivalent sentences with different word orders may have substantially
different information-packaging properties and thus not be "interpretatively
equivalent in absolute terms" (Vallduvi 1993: 2). These considerations raise
the additional question of how one captures the fact of more "marked"
word orders in Classical Sanskrit and other languages, which correlate
460 Brendan Gillon and Benjamin Shaer

strongly with particular information-packaging functions, without assuming

a "basic" word order in these languages.
Our answer to the first question, consistent with Staal (1967) and many
others since, is that "wild tree" phrase structures are best understood as
representing a theoretical alternative to fixed ones. This is because once we
distinguish hierarchical from precedence relations between syntactic ele-
ments, it is natural to take wild trees to underlie languages with both greater
and lesser word order freedom, with restrictions on word orders represent-
ing further (language-particular) specifications of the relations between
sister elements. As for the second question, we follow many other research-
ers in taking information packaging to be a "modular" matter, involving
complex interactions between lexical, phrase-structural, linear, and pro-
sodic properties of sentences; and thus do not see the fact of information-
packaging function alone as evidence in favour of deriving less common
word orders from more "basic" ones. These answers reflect our contention
that a "wild tree" approach is as capable as a "base word order" approach
both of indicating an association of particular word orders with particular
information-packaging functions and of describing languages with greater
and lesser word order freedom.
In what follows, we describe certain word order patterns in our Sanskrit
corpora - patterns we believe are basic to the analysis of Sanskrit word
order - assuming familiarity with contemporary syntactic theory but none
with Sanskrit itself. Our goal at this stage is to sketch a simple and, as much
as possible, theory-neutral account of these patterns, 2 rather than develop-
ing a fully explicit account, which awaits more extensive collection and
investigation of Sanskrit data. We nevertheless hope to show that a "wild
tree" approach to Sanskrit and other free word order languages, both living
and dead, is a fruitful one.
The rest of this chapter is organized as follows. In §1, we review some
of the basic features of Classical Sanskrit. In §2, we describe the basic phe-
nomenon of free word order and traditional Indian and Western views of
the role of word order in the grammar of Sanskrit. In §3, we outline Staal's
"wild tree" analysis of Classical Sanskrit phrase structure and in §4 address
three consequences of this analysis, related to the grammatical significance
of word order permutations, the position of heads in phrases, and word
orders not predicted by the "wild tree" analysis, respectively. As regards
the third, we show how some of these additional word orders can be de-
scribed in terms of a sentence structure for Classical Sanskrit that includes
the object as an immediate constituent. In §5, we address other unpredicted
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 461

word orders, showing how these can be described in terms of long-distance

dependencies whose effects on word order are distinct from those attribut-
able to "wild tree" phrase structure; and also point out some remaining
word order puzzles. Finally, in §6, we offer a summary and some conclu-
sions, including prospects for future research.

1. Some basic features of Sanskrit

To proceed with our investigation, let us review certain aspects of the

grammar of Sanskrit that will be figuring prominently in our discussion.
Sanskrit is an Indo-European language, its earliest extant form being Vedic
Sanskrit (the Sanskrit of the Vedas, the sacred texts of Hinduism), spoken
in northwest India from as early as 1800 BCE. Classical Sanskrit, which
developed from Vedic Sanskrit, appeared as a spoken language around the
fifth century BCE, flourishing as the language of the educated until the end
of the first millennium and as the language of the Hindu literati until the
last century. As it happens, the language has been described by some
authors as "living" rather than "dead", given its continued "religious, edu-
cational, cultural, and political influence in South Asia... and among the
South Asian diaspora all over the world" (Killingley and Killingley 1995: 1).
Sanskrit provides evidence for the occurrence of the four basic lexical
categories widely attested across languages: namely, nouns (N), verbs (V),
prepositions (P), and adjectives (A). 3 Commonly classified as adjectives is
a set of words that includes deictic forms such as tad 'that', relative and
interrogative forms such as yad 'which' and kim 'which?' or 'what?', and
certain partitive forms such as sarva 'all', eka 'one', and anya 'other' (see
e.g. Killingley and Killingley 1995: 27-28). It is worth noting that the mean-
ings of these forms and certain aspects of their grammatical behaviour,
such as their occurrence before other adjectives, might suggest a determiner
analysis for them. However, since they have forms for each gender and
may either stand alone or modify a noun (in which case they agree with the
noun in case, number, and gender), just like other adjectives in Sanskrit, it
is equally natural to treat them as adjectives, and we shall be doing so in
what follows.
Like many members of the Indo-European and other language families,
Sanskrit has a system of inflection for verbs, nouns, pronouns, and adjec-
tives. Verbs are inflected for tense, mood, voice, number, and person, al-
though in copula constructions, the copula form asti ' b e ' is commonly
462 Brendan Gillon and Benjamin Shaer

omitted altogether. Nouns and adjectives are inflected for gender (mascu-
line, feminine, or neuter), number (singular, dual, or plural), and case
(nominative, accusative, instrumental, dative, ablative, genitive, locative, or
vocative), with adjectives agreeing with the nouns that they modify. (Al-
though we take the grammatical function of these cases to be largely famil-
iar, it is worth noting that instrumental case, beyond indicating the means
by which or instrument with which an action is performed, also indicates,
for example, the person with whom an action is performed, i.e. a comitative
use, and the agent of the action in passive constructions.) The noun repre-
senting a clause's subject, though it need not appear (making this a "pro-
drop" language), must agree in number with the clause's main verb when it
does. Personal pronouns are inflected for case and number. Some of these
features of the language are illustrated in the following examples: 4

(1) a. vyädhah gajam pasyati

hunter-NOM elephant-ACC see-3SG.NONPAST
'The hunter sees the elephant.'
b. rämasya grantham pasyämi eva
Räma-GEN book-ACC see-lsg-NONPAST indeed
Ί do see Rama's book.'
(based on Killingley and Killingley 1995: 47, (4); 42, (lc)) 5

As regards word order, this is widely recognized to be substantially freer

than that of, say, English, 6 although this matter has received relatively little
attention. We consider Classical Sanskrit word order in the next section.

2. Word order in Sanskrit: Some basic issues

At the heart of the question of word order in Classical Sanskrit is the frequent
observation that a sentence like that in (2) can be equivalently expressed by
any of the five other possible linear orderings of its three words:

(2) Rämah apasyat Govindam

Räma-NOM saw Govinda-ACC
'Räma saw Govinda.'

From the standpoint of contemporary generative grammar, however, the

remarkable word order freedom displayed in this sentence tells us little
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 463

about the extent of such freedom permitted by the language more generally.
One way to highlight this is with a craftily chosen sentence from English,
which would - contrary to the intuitions of native speakers and substantial
linguistic research - suggest that it, too, tolerated considerable word order
freedom with little or no effect on meaning:

(3) a. Timmy has grown.

b. Grown Timmy has.
c. Has Timmy grown!

What is missing here is a recognition of one of the major insights of con-

temporary linguistic theory: namely, that the freedom with which words
may occupy different positions in sentences is related to the way in which
they are organized into phrases, these phrases having their own internal
structure and behaving as units with respect to such processes as extraction,
ellipsis, and coordination. As it happens, previous research on Vedic as
well as Classical Sanskrit has found clear signs of words behaving as con-
stituents of phrases - in particular, displacement involving phrases rather
than just heads (see e.g. Schäufele 1990: 96-102). This suggests a picture
of Sanskrit as having far greater restrictions on word order than Warlpiri,
for example, where any permutation of the words in a sentence like (4) is
acceptable so long as it respects the language's condition that the auxiliary
element be in second position (Donohue and Sag 1999: §1):

(4) Kuru ma purra-nja-rla nga-rnu ngajulu-rhu

meat 1SG.PAST c o o k - I N F - P R I O R C O M P eat-PAST l s g - E R G
' H a v i n g cooked (it) I ate (the) meat.' (Laughren 1989: 322)

So the question, then, is what the phrase structure of Classical Sanskrit is

and what restrictions the grammar of the language imposes on the linear
ordering of words.

2.1. Traditional Indian and Western views

A good place to start in looking for an answer to this question might be a

brief review of the t w o basic positions on Sanskrit word order that San-
skritists have held. Following Staal (1967) and others, we can describe
these as traditional Indian and Western views, respectively, although keep-
ing in mind that both still figure in the current study of Sanskrit.
464 Brendan Gillon and Benjamin Shaer

According to Staal (1967: 49), the almost unanimous view of traditional

Indian scholars is that Sanskrit word order is free. As he explains, "the San-
skrit grammarians assumed... not that such rules are beyond the scope of
grammar, but that there are no such rules" (Staal 1967: 36). Significantly,
this view is also reflected in the work of Apte (1885 [I960]), an influential
modern Sanskritist, for whom word order is not "a material point for con-
sideration" (1885 [I960]: §399). It is likewise found in the work of con-
temporary scholars such as Lazzeroni, who notes that "the order of con-
stituents within the sentence does not have grammatical function and is thus
largely free" (1998: 122).
In contrast, most descriptions of Sanskrit word order by European
scholars, including Delbrück (1878), Speyer (1896), Canedo (1937), and
Gonda (1952), take the language to have a basic word order, described
variously as "traditional", "habitual", or "preponderant". The core of this
basic word order is a sentence-initial subject NP and a sentence-final verb,
with direct and indirect object NPs positioned between them and head-final
patterns for NPs and PPs (Staal 1967: 51-59). Departures from this order
serve to highlight elements displaced to the beginning or end of the sen-
tence or are related to the specialized functions of various sentence types.
For example, Gonda (1952) (cited in Staal 1967: 58) takes sentence-initial
occurrences of the verb to have many different motivations, including an
association with (i) sentence types such as imperatives, questions, and ex-
clamatives; (ii) the expression of the future or the intention of the speaker;
(iii) the occurrence of the copula verb asti ' b e ' or of a verb of knowing,
often in the first person; (iv) a subordinate clause; (v) the continuation of a
narrative; and finally, (vi) special emphasis of the verb. Medial occurrences
of the verb are likewise accounted for in terms of special considerations,
such as the displacement of other elements to the end of the sentence "in
order to facilitate transition to the next sentence".
Of course, the Western approach to Sanskrit just sketched, with its pos-
iting of a basic word order in the language and its appeal to special consid-
erations for the derivation of other orders, can be seen as consistent with
contemporary research such as that of Hale (1996), who posits a base order
for the language and seeks to derive other attested orders by movement. In
contrast, Staal's (1967) approach to Sanskrit was influenced by the tradi-
tional Indian view of word order - although, in fact, he saw this and the
Western view as easily reconcilable. We shall have more to say about his
attempt to reconcile these views in §4.1 below. Before doing so, however,
let us turn to the details of Staal's own analysis.
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 465

3. Staal(1967)

As just noted, Staal (1967) adopted in his own analysis of Classical Sanskrit
the traditional Indian view that the language's word order had essentially
no grammatical significance. Central to this analysis was his rejection of
what was long a standard assumption in generative grammar: namely, "that
order is not only a feature of the terminal strings and finally derived sen-
tences... but also figures in the deep structures underlying t h e m " (Staal
1967: 7-8). 7 Appealing to Pänini's concept of the karaka relations —
equivalent to current notions of thematic relations or valences - Staal took
such relations rather than word order to be grammatically basic, asking why
it was necessary to "preserve order in the deep structure" instead of intro-
ducing rules of the base that imposed no such order and left this to be de-
termined by other rules.

3.1. The "wild tree" analysis

Staal's attempt to specify such rules forms the basis of his "wild tree"
analysis of Sanskrit word order. What he proposed was that the set of linear
orderings of words in a clause is precisely the set obtained from the free
transposition of all of the sister constituents in the clause, where such trans-
position induces no tangling. Staal referred to phrase markers that under-
specified linear order in this way "wild trees" (Staal 1967: 15). The phrase
structure rules that he proposed for generating such trees looked like those
in (3), where heads and phrases form unordered sets, as indicated by brace
brackets: 8

(3) a. S - {NP, VP}

b. V P - > {NP, V}
9
c. N P - » {N, (S)} (based on Staal 1967: 65)

When applied to a Classical Sanskrit sentence like that in (2) above, these
rules result in the following four ordering possibilities (where " N P S " indi-
cates the nominative-marked subject and " N P 0 " the accusative-marked ob-
ject): 1 0
466 Brendan Gillon and Benjamin Shaer

(4) a. N P - V P permutations
i. S NPS-V-NP0
[HPS R ä m a h ] [Vp apasyat [Np0 Govindam]]
ii. S - > N P s - N P 0 - V
[HPS R ä m a h ] [Vp [NP0 Govindam] a p a s y a t ]
b. V P - N P permutations
i. S V-NPo-NPs
[vp a p a s y a t [NP0 Govindam ]] [ NPs R ä m a h ]
ii. S —* N P 0 - V - N P S
[VP [NPO Govindam apasyat ] [ NPs Rämah]

Significantly, all of the patterns in (4) are attested in the Apte corpus de-
scribed a b o v e , " if we include among VPs non-verbal predicate expressions
with no overt copula. As Gillon (2004: § § 4 . 2 - 4 . 3 ) shows, this corpus re-
veals (i) non-copular VPs preceding and following subject N P s ; (ii) VPs
with overt copula f o r m s that express existence preceding and following
subject N P s ; and (iii) predicates with no overt copula preceding and follow-
ing the subject noun phrase. These are illustrated in (5), (6), and (7), respec-
tively: 12

(5) a. [NPs iti-ete me samkalpäh] [vp präduräsan]

these of me thoughts sprang up
'These thoughts of mine sprang up.' (SG 14.1.6)

b. [vp smartavyah] tu [NPs a y a m j a n a h ]

is to be remembered but this person
'But let this person (i.e. me) be remembered.' (SG 14.1.4)

(6) a. [ NPs sarvam eva etat] [Vp na asti ]

all only this not is
'Absolutely all this does not exist.' (SG 12.1.6) 13

b. [vp na asti] [ NPs jTvität anyat abhimatataram iha jagati

not is life other more liked here in world
sarva-jantünäm]
all-creatures
'There is nothing more liked by all creatures in this world other
than life.' (SG 8.1.4)
 Classical Sanskrit, "wild trees", and the properties of FWO languages 467

(7) a. [Nps iti-esah] [Vp vah dharmah]

this your duty
'This is your duty.' (SG 14.2.1)

b. [vp capalah] [ NP ayam ba.tuh]

fickle this fellow
'This fellow is fickle.' (SG 14.2.1)

The only exceptions to this pattern, as suggested by an examination of the

Pramänavärttika (henceforth PV) corpus, 14 are sentences with an overt
copula of predication, in which the subject NP never follows the VP:

(8) a. na [NPs anyah hetuh] [Vp gamakah asti]

not other ground indicative is
'No other ground is indicative.' (PV 8.12)

b. na [vp asti gamakah] [NPs anyah hetuh]: unattested

not is indicative other ground
'No other ground is indicative.'

Note that this restriction may simply be an artefact of the relatively small
size of the corpus, and can be verified only through the investigation of
additional data. (Another possibility, however, is that this gap represents
something more systematic, which is related to the status of VP in the
grammar of Sanskrit. We shall be sketching this possibility in §4.3 below.)

4. Some issues raised by the "wild tree" claim

Despite the virtues of Staal's analysis of Sanskrit, as shown above, it does

have various consequences for the study of word order that seem difficult
to square with the claims of more recent research. In addition, and rather
more damagingly, the analysis cannot capture the full range of word order
patterns attested for Classical Sanskrit. We shall be taking up these matters
briefly in the following sections. What we shall be suggesting is that certain
more contentious aspects of Staal's analysis can be reconciled with more
standard claims after all, while others reflect open questions, to which an-
swers consistent with a "wild tree" approach are at least as plausible as
those couched in terms of "basic" word orders. Finally, as regards the more
straightforward empirical problems with Staal's analysis, these turn out to
468 Brendan Gillon and Benjamin Shaer

be amenable to solutions, sketched below, broadly consistent with the "wild

tree" claim.

4.1. The grammatical significance of word order permutations

One aspect of Staal's "wild tree" analysis that seems squarely at odds with
much recent research is his claim that word orders in Sanskrit (and many
other languages) "have no grammatical significance" (Staal 1967: 59).
Such a claim is a contentious one, since it appears to deny the relation
commonly observed between word order and the information-packaging
properties of sentences. As it happens, Staal (1967: 6 3 - 6 4 ) does not see his
claim as incompatible with the possibility of "two Sanskrit sentences...
[differing only] in the arrangement of the words" but not being inter-
changeable, the sentences being understood to give rise to "a difference in
meaning" or not to "fit into larger contexts equally well." This is because
such differences in word order, though clearly relevant to interpretation
broadly construed, are for Staal best treated in a theory of performance
rather than competence. N o w , such a conclusion might be consistent with
views of the competence/performance distinction current at the time that
Staal was writing (see e.g. Chomsky 1965: 11 for some remarks on this).
However, the detailed investigation of information packaging conducted
since then (see e.g. Birner and Ward 1998; Lambrecht 1994; Prince 1981;
Vallduvi 1993) strongly suggests that at least some reflex of speakers'
knowledge of such properties is very much part of their linguistic compe-
tence and not plausibly relegated to performance after all.
W e might still be able to reconcile these diverging views of word order
by observing that on a "modular" view of linguistic competence, the inclu-
sion of information-packaging properties of sentences within the grammar
should have little bearing on the "wild tree" claim in any case. This is be-
cause the determination of possible word orders by the syntax can, on this
view, be seen as orthogonal to the information-packaging functions to
which different word orders are put (granting the possibility of certain word
orders correlating directly with certain information-packaging functions).
Moreover, the strategy of deriving less common word orders with special-
ized information-packaging functions from more basic syntactic structures,
though widely employed in describing the relation between syntactic and
information structure, seems to us considerably less plausible than ones
consistent with the "wild tree" claim. 15 One strategy of the latter kind would
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 469

be to indicate a special status for a particular permutation of a wild tree

through feature annotations on the relevant node in a phrase, reminiscent of
Pollard and Sag's (e.g. 1994: 43) treatment of subject-auxiliary inversion
structures, in which the sentence-initial occurrence of an auxiliary verb is
specified as [+ INV]. Thus, notwithstanding the difficulty of determining the
respective information-packaging properties of sentences with different
word orders on the basis of corpus evidence alone, it does seem possible to
incorporate such findings into a description of Classical Sanskrit without
abandoning the idea that the order of the daughters in a given phrase in the
language is essentially free.
Another consideration favouring the view of syntactic and information-
packaging systems as interacting but autonomous is that word order varia-
tions observed in other languages with substantial word order freedom are
not plausibly explained in terms of differences in information-packaging
properties alone. For example, Schachter and Otanes (1972; cited in
Kroeger 1993: 110-111) note about Tagalog that "stylistic permutations" in
word order obey ordering principles related to the position of the Actor
phrase with respect to other arguments, the nominative-case-marked N P
with respect to other arguments, and the position of "heavier" NPs with
respect to "lighter" ones. A similar confluence of factors is also recognized
to underwrite the word order patterns of German (see e.g. Fanselow, to
appear for discussion). Such observations provide compelling reasons to
reject the conclusion that the information-packaging role of particular word
orders in Classical Sanskrit must be seen as bearing directly on the deriva-
tion of these different word orders in the syntax and, more to the point, that
there must be some "basic" order of constituents in a Classical Sanskrit
sentence.

4.2. "Wild trees" and head position

A second and, for some, again undesirable consequence of the "wild tree"
analysis of Classical Sanskrit is that it offers no characterization of phrases
in the language in terms of the position of their heads relative to their other
daughters, even though such a characterization has been widely seen as
basic to the description of phrase structure across languages. In fact, though
we have not yet subjected this matter to detailed scrutiny, what we have
found so far gives us little reason to include a specification of head position
in the syntactic description of Classical Sanskrit. For example, such a speci-
470 Brendan Gillon and Benjamin Shaer

flcation is not supported by Schäufele's (1990: 6 1 - 6 3 ) observation that

Sanskrit throughout its history had no more than a general tendency for
head-final phrases, with the Vedic Sanskrit corpus displaying numerous
exceptions to this tendency in all categories. 16 ' 17 Nor is it supported by our
own examination of the Apte and PV corpora, where we have found what
are plausibly analysed as right- and left-headed VPs, 18 PPs, and NPs, as
illustrated in (9), (10), and (11), respectively:

(9) a. na [MPs anyah hetuh] [Vp [AP gamakahj asti]]

not other ground indicative is
' N o other ground is indicative.' (= (8a))

b. [VP vijayetam [NP Rama-Laksmanau ]]

may overcome Rama-Laksmanau-ACC
[NP Kumbhakarna-Meghanadau ]
Kumbhakarna-Meghanadau-NOM
'May Kumbhakarna and Meghanada overcome Rama and
Laksman.' (SG29.1.6) 1 9

(10) a. ... [PP [NP äsrayam] antarena]

substratum without
' . . . without a substratum'

b. ... [ppvinä [NP an-upalabdhyä ]]

without non-apprehension
' . . . without (the mark of) non-apprehension' (PVS 4.17)

(11) a. [NP [ a p mahän] kalakalah]]

great uproar
'great uproar' (SG 27.1.11)

b. [NP sütän [etän mätr-hinän ]]

children these mother-deprived
'these children, deprived of a mother' (SG 13.2.2)

As it happens, such corpus findings are broadly consistent with some key
results of Minimalist research, including its dispensing with primitive dif-
ferences in the direction of case and theta role assignment in deriving cross-
linguistic word order differences and its analysis of certain word orders,
such as those associated with PP extraposition, in terms of processes "in the
phonological component" (Chomsky 1995: 426; see also Chomsky 2005)
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 471

rather than the syntax. Taken together, such conceptual and empirical con-
siderations point to a treatment of head position in Classical Sanskrit
phrases not as reflecting any basic syntactic properties of the language, but
rather as determined largely by non-syntactic factors, including those of the
kind enumerated by Gonda (1952), as described in §2.1. Although these
remarks are obviously preliminary, they do at least suggest that the head-
final tendency observed of Sanskrit phrases may be better captured in terms
consistent with a "wild tree" analysis than in terms of extraction operations
that apply to "basically" head-final phrases or of some bifurcation between
" u n m a r k e d " and " m a r k e d " head positions, respectively related to the "core"
and the "periphery" of the grammar.

4.3. Word orders beyond the "wild tree" claim

A final consequence of Staal's "wild tree" analysis of Classical Sanskrit

word order - and in fact the most serious one, because it reveals a clear
empirical problem for the analysis - pertains to the word orders that it does
and does not predict to be available. What is clear from the basic form of
Staal's analysis is that it generates only four linear orderings for sentences
containing a subject, a verb, and an object, as given in (4), and crucially
does not predict the availability of two additional orderings. The orderings
in question, as shown in (12), are those whereby the subject N P intervenes
between the two daughters of the VP, destroying the continuity of the V P
node:

(12) a. [ NPo Govindam] [ NPs Rämah] apasyat

b. a p a s y a t [ NPs R ä m a h ] [ NPo Govindam]

In fact, such discontinuities are familiar to scholars of Classical Sanskrit

and attested in the Apte and PV corpora. For example, the sentence in (13a)
has the same structure as that in (12a), and the sentence in (13b) (where, in
contrast to the h o m o p h o n o u s string in (9b), the first N P represents the sub-
ject and the second the object) has the same structure as that in (12b):

(13) a. [karana-guna-vaktu-kämate hi ] vacanam

organ-quality-speaking-desire-ACC alone speech-NOM
anumäpayet
should imply
'Speech should imply both qualities in the organ of speech and a
desire to speak alone.' (PVS 10.1)
472 Brendan Gillon and Benjamin Shaer

b. vijayetam [MP Rama-Laksmanau ]

may overcome Rama-Laksmanau-NOM
[NP Kumbhakarna-Meghanadau ]
Kumbhakarna-Meghanadau-ACC
'May Rama and Laksman overcome Kumbhakarna and Meghanada.'
(SG 29.1.6)

Now, Staal was not unaware of this problem, and observed that while "the
points dominated by a single node" in a wild tree can "be arranged in any
desired order", "rearrangements are required" to derive additional orders
"which go beyond the confines of single constituents or elements domi-
nated by a single node or category symbol" (Staal 1967: 34). In other words,
Staal clearly recognized the existence of dependencies between pairs of
phrases that were "longer-distance" than could be captured on his "wild
tree" analysis, even though he did not actually pursue this point in his
study. 20
Since there is, in fact, ample evidence for a range of "longer-distance"
dependencies in Classical Sanskrit, which (as we shall be suggesting in §5)
can be handled in terms of extraction, it would seem straightforward to treat
sentences like those in (12)—(13) in similar terms. Accordingly, the word
order exemplified in (12a)-(13a) would arise from leftward extraction of
the object N P out of a sentence-final VP; and that exemplified in (12b)—
(13b) from rightward extraction of the object NP out of a sentence-initial
VP.
Of course, while an appeal to extraction here does serve to account for
the observed word orders, it does so at a price: namely, that it robs the
"wild tree" analysis of some of its cogency. This is because such as analy-
sis seeks to offer an alternative to movement-based approaches of the word
order facts outlined in (4), so that the case for recognizing "wild trees" in
the first place is somewhat undermined by a recourse to movement in ac-
counting for similar word order facts. It is also because the recourse to
movement here introduces an asymmetry between these word orders and
those given in (4), which is at odds with the original observation that all of
the permutations in question are syntactically and semantically equivalent.
As it happens, an alternative to movement exists for these cases which is
very much in the spirit of Staal's (1967) basic "wild tree" analysis. This is
to take the Sanskrit sentence to be even flatter than Staal claimed, and in
particular to take the verb not to project to the phrasal level - making it, its
sisters, and the subject N P all immediate constituents of the sentence. Such
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 473

a proposal is hardly a new one, and figures in accounts of Warlpiri such as

Bresnan (1982) (see e.g. Baker 2001: 4 0 9 - 4 1 0 for s o m e review). M o r e
recently, it has been offered, for example, for T a g a l o g by Kroeger (1993:
1 0 9 - 1 6 6 ) and closely follows the proposal of Pollard and Sag (1994: 40)
for a variety of languages with freer word order, which involves a general
phrase structure s c h e m a "in which all complements (including the subject)
are realized as sisters of the lexical head". T h e sentence structure in ques-
tion, then, is the one shown in (14), where w e m a k e use of Staal's "wild
tree" notation:

(14) S {V, NPS, NPo}

Interestingly, this structure is not only consistent with the word orders in
(12)—(13) - which are accordingly treated not in terms of discontinuous
V P s but simply as t w o available linear orderings of the immediate constitu-
ents of a " f l a t " sentence - but also offers a solution to certain other word
order puzzles. O n e is related to our failure so far to uncover sentences in
the Apte and PV corpora in which the copula and a predicate adjective pre-
cede the subject, as illustrated in (8b) and repeated in (15a) below; and
more generally to uncover sentences in which the verb and its sisters be-
have as a constituent. If we reanalyse (15a) as in (15b), then the copula and
the predicate adjective simply do not form a constituent and thus would not
be expected to behave as one:

(15) a. na [ypasti gamakah] [ NPs any ah hetuh]: unattested

not is indicative other ground
' N o other ground is indicative.'

b. na [V asti] [AP gamakah] [ NPs anyah hetuh]: unattested

not is indicative other ground

O f course, this "no V P " analysis would not actually rule out such a struc-
ture, given the "wild tree" claim. At this stage, this is what we want, since a
more extensive corpus search might still reveal such a structure. However,
the claim that there is no V P constituent would arguably m a k e the infre-
quency of a sentence like that in (15) less surprising, particularly given the
possibility of other ordering factors conspiring to m a k e such a sentence
infrequent or even unavailable.
Another word order puzzle that a "no V P " analysis would help to explain
is the anomaly reported by Gillon (1996) in the restrictions on extraction
474 Brendan Gillon and Benjamin Shaer

from an assumed VP relative to the restrictions on extraction from other

phrases. What Gillon argues (and as we shall be considering in more detail
in §5) is that putative cases of extraction target both phrases that are the
immediate constituents of the subject N P and the VP and those that are the
immediate constituents of the sisters of the verb. This is illustrated in (16)
(Gillon's (24)), where an instrumental-marked adjective phrase has, by hy-
pothesis, moved leftward out of an instrumental N P dominated by the VP:

(16) yathä [Ap itah-mukha-ägataih api

as this-heading-coming-INST even
[NPs mahän kalakalah] [Vp srutah [ NP Ι asmabhih]]
great-NOM uproar-NOM heard by us-INST
'As a great uproar was heard by us, even (while) coming in this direc-
tion.' (SG 27.1.11)

If there is, in fact, no VP node here, then this AP can be analysed as having
simply been extracted from the instrumental NP that is an immediate con-
stituent of the sentence.
Admittedly, the claim that Classical Sanskrit had no VP is a strong one,
with far-ranging implications for its grammar - predicting, in particular, the
absence of a range of subject/object asymmetries that have provided sup-
port for VP structure. These asymmetries include ones related to referential
and quantificational dependencies between subjects and objects and the
possibility of pro-forms and of ellipsis processes that include the verb and
object but exclude the subject (see e.g. Baker 2001: 407-418). So far, we
have uncovered no compelling evidence for such asymmetries in our cor-
pora. We have, however, discovered the following sentence, with an object
reflexive occurring to the left of the verb:

(17) [NP [AP bhartr-gatayä ] cintayä ][Npätmänam a p i ] na

husband-turned-INST thought-INST herself-ACC even not
[NPSesä ] [ν vibhavati ] [ NP [kim punar] agantukam]
she aware how still guest-ACC
'With her thought turned toward her husband, even herself she was
not aware of, let alone her guest.' (SG 23.1.3)

Such a sentence can easily be handled on an analysis in which the anaphor

has been moved out of the VP and over the subject. However, it is also
compatible with a "wild tree" analysis of Classical Sanskrit in which there
 Classical Sanskrit, "wild trees", and the properties of FWO languages 475

is no VP, particularly if we adopt a non-configurational binding theory of

the kind proposed by Pollard and Sag (1994: 248-266), in which an anaphor
need not be c-commanded by the NP with which it is coindexed, but must
have a grammatical function more oblique than that of the latter NP. What
such data certainly make clear is that a good deal remains to be done to find
robust evidence for this "no VP" proposal; and also that what we do find
may ultimately lead us back to the conclusion that Classical Sanskrit did
have a VP. At this stage, however, an analysis of the language as lacking a
VP does seem compelling.
It is worth pointing out, though, that another, subtler difficulty exists for
this "no V P " analysis in the context of our recognition of long-distance
dependencies in Classical Sanskrit, to be described in §5 below. This is that
the proposal to broaden the "wild tree" analysis by including subject NPs
among the constituents that may permute with the verb and its sisters inevi-
tably creates a tension between the analysis of word orders attributable to
such free arrangement of phrasal daughters and those attributable to longer-
distance dependencies. What we shall be arguing, however, is that a dis-
tinction between word order permutations due to a "wild tree" phrasal ar-
chitecture and those due to longer-distance dependencies is one that the
grammar of Classical Sanskrit might very well have made; so that the ten-
sion in question, while inevitable, might nevertheless be minimal.
To sum up this section: we have explored three consequences of Staal's
(1967) "wild tree" analysis, which were related to (i) the different informa-
tion-packaging properties commonly associated with different word orders;
(ii) the position of heads in their phrases, which research on word order has
commonly taken to be part of a language's basic grammatical specification;
and finally (iii) word orders not predicted by the basic "wild tree" analysis.
What we found was that these consequences of the analysis, though all
apparently undesirable, could either be reconciled with current research
after all or be handled in ways broadly consistent with the "wild tree"
claim.

5. Long-distance dependencies in Sanskrit

As a final demonstration of the fruitfulness of Staal's approach, we outline

what we have found in the Apte and PV corpora regarding the basic possi-
bilities for long-distance dependencies in Classical Sanskrit - understood
here as those that extend beyond the phrase of which the elements in the
476 Brendan Gillon and Benjamin Shaer

dependency relation are the putative daughters. Although we take the ulti-
mate analysis of these dependencies to be an open question, our findings
nevertheless suggest that a useful distinction can be drawn between the free
transposition of a phrase's daughters at the heart of the "wild tree" analysis
and word order permutations that are the result of other processes, thus
providing strong indirect support for "wild trees" themselves.
The long-distance dependencies that we have observed in the Apte and
PV corpora, and the constraint that appears to be operating on them, can
best be illustrated with constructed examples like those in (18). In this set,
(18a) represents a sentence with no such dependencies, while (18b) repre-
sents one with a licit dependency and (18c), which is unattested and thus by
hypothesis ungrammatical, represents one with an illicit dependency:

(18) a. [NP purusah [ AP [ NP rämät ] uccatarah ]]

man-NOM Räma-ABL taller-NOM
[NP vanam ] gacchati
forest-ACC is going
Ά man taller than R ä m a is going to the forest.'

b. [ NP purusah ι ] [ NP vanam ] gacchati

man-NOM forest-ACC is going
[AP [NP rämät ] uccatarah ];
R ä m a - A B L taller-NOM

c. [NP purusah [AP uccatarah Ι ]]

man-NOM taller-NOM
[NP vanam ] gacchati [Np rämät ] Ι: unattested
forest-ACC is going Räma-ABL

What these examples suggest, then, is that such dependencies in Classical

Sanskrit are possible but highly restricted in their occurrence, and that this
restriction can be described in terms of a constraint on extraction. 21 More
specifically, the extraction of a phrasal daughter of the subject N P , as in
(18b), is possible, whereas the extraction of a phrase that is the daughter of
this daughter, as in (18c), is not. The generalization that emerges from ex-
amination of the large number of relevant sentences in the Apte and PV
corpora is that Classical Sanskrit permits the extraction of only (i) immedi-
ate constituents of the sentence; and (ii) daughters of these constituents. In
addition, if a daughter constituent is extracted, then all of its own daughters
are extracted. Accordingly, extraction of a phrase that crosses any pair of
 Classical Sanskrit, "wild trees", and the properties of FWO languages 477

phrases in the language gives rise to a kind of island effect - albeit an un-
familiar-looking one, a point to which we shall be returning below.
Now, despite the robustness of the dependency patterns just described,
which we shall be describing in more detail below, various considerations
make it unclear that they should, in fact, be assimilated to standard extrac-
tion processes. The most obvious one is the very restricted nature of these
dependencies in Classical Sanskrit: unlike extraction in English and other
languages in which such processes have been studied extensively, the de-
pendencies in Classical Sanskrit are always clause-bound whether the dis-
placed element occurs on the left or the right edge of the clause. 22 This
makes the extraction in question a peculiarly local process, and thus more
like the clause-bound phenomenon of extraposition - which, as noted
above, has been seen by Chomsky (e.g. 1995, 2005) and others as not even
part of the syntax proper - than like unbounded extraction processes such
as wh- and N P movement. Adding to the doubts about a treatment of these
dependencies in terms of extraction are the island effects that we have of-
fered in support of such a treatment. Notwithstanding the unfamiliar form
of these island effects (which arguably makes the connection to well-
studied extraction processes even more tenuous), there has been a growing
recognition that even familiar island effects are more likely the result of
processing factors rather than the violation of purely syntactic constraints
(see e.g. Kluender 1998). Thus, while it may well be that the hypothesized
island effects in Classical Sanskrit, as described above, have the same
source as more familiar ones, the source in question is not likely to be
purely syntactic.
Of course, even setting aside the possibility that extraposition does not
represent true extraction, it must still be noted that extraction processes are
not the only grammatical means to establish long-distance dependencies,
nor therefore the only possibility for describing patterns like that in (18).
One intriguing alternative, for example, is that these dependencies are es-
sentially anaphoric ones not established through extraction. 23 On such an
analysis, the two phrases in a dependency could, syntactically speaking,
have simply an appositive or correlative relation, with case-marking estab-
lishing the anaphoric connection between them. As applied to (18b), this
would suggest a literal paraphrase along the lines of Ά man is going to the
forest, he taller than Rama'. While this possibility is indeed an intriguing
one, we have not yet subjected it to sufficient scrutiny to determine how
well it generalizes to all of the long-distance dependencies that we have
uncovered in the Apte and PV corpora - many of which still look, to our
478 Brendan Gillon and Benjamin Shaer

minds, strikingly like extraction. Accordingly, in what follows we shall

continue to describe these dependency patterns in terms of extraction, recog-
nizing that much work still needs to be done to determine whether this is
the best analysis for them.

5.1. Leftward extraction

Let us first turn to cases of leftward extraction. What we can report here is
that these occur in all types of clauses, including main, relative, infinitival,
participial, gerundial, sentential subject, and absolutive clauses (see Gillon
1996 for descriptions and illustrations of these various clause types).
For example, we find phrases extracted leftward from main clause sub-
ject and verb complement N P s in sentences like those in (19a) and (19b),
respectively:

(19) a. [Ap te ]i [np ete [karya-svabhäva-anupalabdhi-laksanah]]

t h e - N O M those-NOM effect-indiv'l-non-apprehension-NOM
[np ι trayah hetavah]
three-NOM reasons-NOM
' T h e three reasons are those of effect, of individual, and of non-
apprehension.' (PV2.14)

b. yathä [ A P itah-mukha-ägataih api], [ NPs mahän kalakalah ]

as this-heading-coming-INST even great-NOM uproar-NOM
srutah [ N p ι asmabhih ]
heard by us-INST
' A s a great uproar was heard by us, even (while) coming in this
direction.' (SG 27.1.11)

W e likewise find such dependencies, among other places, in relative clauses

like that in (20a), where a genitive-marked relative pronoun is extracted
from the subject N P ; and in participial clauses like that in (20b), where a
genitive-marked N P is extracted from the object N P (subordinate clauses in
these examples are indicated by "S"'):
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 479

(20) a. [ s [s' [Npyasya]i [NP a-darsana-mätrena]

whose-GEN non-observation-mere-INST
[NPS ι vyatirekah] pradarsyate]s' [NP tasya samsaya-hetutvät]
exclusion-NOM is shown its-GEN doubt-grounds-ABL
[AP sesavat] [NPSTAT ] udahrtamjs
inadequate-NOM it-NOM is called
'That whose exclusion is shown through mere non-observation is
called inadequate, because it is grounds for doubt.' (PV 10.19)
b. [s [NPS U p s a h api dusta-äsayah ] . . . [S'[NP jala-caranasya]i
the P R T evil-intentioned-NOM water-goers-GEN
[NP mithyä-vartä-samdesakaih] [Np Ι manämsi ]
false-report-tidings-INST minds-ACC
ranjayan]]s' nityam iva [NP ähära- vrttim ] akarot]s
appeasing always as food-making-ACC made
' T h e ill-intentioned [crane]... appeasing the minds of the fish with
false reports and tidings, made his dinner as usual.' (SG 15.1.1)

5.2. Rightward extraction

Rightward extraction of phrases is also widely attested in our corpora. For

example, w e find cases like (21a), in which a nominative-marked AP is
extracted from the subject N P ; and (21b), in which a nominative-marked
A P containing a locative N P is extracted from a predicate N P :

(21) a. api ca [NPS [AP artha-antara-nimittah hi ] dharmah Ι ]

moreover thing-other-cause-NOM P R T property-NOM
syät [NP anyah eva ] [Ap sah ]ι
must be different-NOM completely the-NOM
'Moreover, the property whose cause is different from a thing (in
which it inheres) must be different (from it).' (PV 20.18)

b. [NP a-pravrttih [NP pramänänäm ] Ι ]

non-adversion-NOM epistemic cognition-GEN
[NPSanupalabdhih ] [ A P a-pravrtti-phalä [NP a-sati ] ]]
non-apprehension-NOM non-acting-result-NOM non-pres-LOC
'Non-apprehension is the non-adversion of an epistemic cognitive
means, which results in one's not acting on what is not present.'
(PV 4.5)
480 Brendan Gillon and Benjamin Shaer

The cases of leftward and rightward extraction given here are among the
many that we have found in the Apte and PV corpora, including certain
more complex ones that await further analysis. We take these examples to
provide solid evidence that such processes did exist in the grammar of
Classical Sanskrit, and understand these processes to have operated on the
phrases in "wild trees". One question that naturally arises from our recogni-
tion of extraction in the language is how this was related to the language's
"wild tree" architecture. At this stage we can offer only a tentative answer
to this question, although one that we consider promising, given the paral-
lelism displayed by leftward and rightward extraction. This is that extracted
phrases were similarly unordered with respect to the rest of the sentence.
What Classical Sanskrit might have had, in other words, was wild trees all
the way up.

5.3. Some remaining puzzles

Of course, there are many other questions that our discussion of extraction
in Classical Sanskrit has so far left unanswered. Among these is the status
of various elements found at the periphery of the clause, which, as Gillon
(1996) observes, include such elements as interjections, vocatives, clausal
conjunctions, and sentential adverbials. Some of these peripheral elements
are illustrated in (22):

(22) a. Conjunctions (left edge):

iti [AP iyam]I [NP prayoga-bhedät ] [ AP asta-dhä]
so this-NOM application-difference-ABL eight-fold-NOM
[MPS Ι anupalabdhih]
non-apprehension-NOM
'So, this (sort of) non-apprehension is eight-fold, due to (its) dif-
ferent applications.' (PV6.18)

b. Conjunctions (right edge):

[ NP kärunikasya api]i [ AP nih-phalah ]
compassionate-GEN even pointless-NOM
[NPS Ι ärambhah ] [ NP a-vipäryasät 1 iti cet
undertaking-NOM non-delusion-ABL if
'Suppose it is argued that even a compassionate person's under-
taking is pointless because of his non-delusion.' (PV 9.16)
 Classical Sanskrit, "wild trees", and the properties of FWO languages 481

c. Adverbials:
[NP sarvatra ][ N P audarikasya],
in every case-LOC glutton-GEN
[ [ N P a b h y a v a h ä r y a m eva] [nps ι visayah]]
food only object
'In every case, a glutton's object is only food.' (SG 1.1.2)

d. Vocatives:
[isipsakhe] [Ap sä eva ]! [ AP dhänyä ]
friend-VOC the-NOM just lucky-NOM
[NPS Ι gänikä- därikä ]
prostitute-daughter-NOM
[s' [NP yarn] evam [ NPs bhavan-manas ] abhinivisate]s
to w h o m - A C C in this way y o u - m i n d - N O M is devoted
'Friend, the very same prostitute's daughter to whom your mind is
devoted in this way is the lucky one.' (SG 30.1.3)

Given our distinction between word order permutations due to extraction and
those due to "wild trees", the question of what positions such elements occu-
py turns out to be a relevant one, as we shall be suggesting in what follows.
Inspection of the examples in (22) reveals that most of these elements
appear at the left periphery of the clause but are not confined to this periph-
ery, given that we also find cases of right-peripheral conjunctions like that
in (22b). Moreover, since in their left-peripheral occurrence these elements
precede extracted elements, we might take them to occupy positions be-
yond the landing sites of extraction, which are themselves standardly as-
sumed to be beyond S.
Yet, the question arises whether these peripheral elements actually do
mark sentence edges, 2 4 given cases like the following ones, in which
phrases hypothesized to be extracted occur to the left of these elements:

(23) a. [ A p iyam ]I [CNJ c a ] [ NP hetu-a-siddhyä ] [eva]

this and cause-non-establishment-INST only
[ NPs ι tat-viruddha-siddhih ] [ADVpräk eva]
it-contradictory-establishment-NOM above just
nirdistä
pointed out
' A n d pointed out just above by the non-establishment of a cause
was the establishment of what is contradictory with it.'
(PV 6.17)
482 Brendan Gillon and Benjamin Shaer

b. [ap tena]i [ c N j c a ] [ NP pramänena ]

that-INST and epistemic cognition-INST
[NPS [NP sädhya-dharmasya ] tat-mätra-anubandhah ]
provable-property-GEN it-mere-dependence-NOM
khyäpyate
is made known
'And the dependence of the property to be proved merely on it is
made known by that epistemic cognition.' (PV 18.1)

c. [s [ s ' U p y a h ] i [cNjtarhi] [wpsamagrena hetunä ]

which then complete-INST cause-INST
[NPS Ι kärya-utpädah ] anumiyate] S '
effect-arising-NOM is inferred
[NPS sah] [ ADV katham] [Nptri-vidhe hetau] antarbhavati] s
it how 3-fold-LOC reason-LOC is included
'How, then, is the arising of an effect which is inferred through (its)
complete (set o f ) causes included within the threefold reason?'
(PV 6.24)

Whatever such elements do mark, it would be a boon to our distinction

between freely arranged and extracted phrases if at least some of these ele-
ments turned out to be reliable boundaries between these two kinds of
phrases, and occupied positions consistent with the hypothesis that freely
arranged phrases had positions lower in a tree than extracted ones.
An important first step in solving the puzzle of these peripheral ele-
ments is to note that their distribution is far from uniform. For example,
vocative expressions and interjections across languages are often described
as "extrasentential" - that is, as not "obviously constituents of [the] sen-
tences" with which they are associated (McCawley 1998: 750; see also
Espinal 1 9 9 1 ) / 3 Since, on this view, these elements would simply be out-
side of the sentence, their occurrence at the sentence's left or right edge
would seem to indicate the sentence's boundaries.
Yet, various complications arise in using vocative expressions to detect
sentence edges. One is that these expressions actually have a very free dis-
tribution, as McCawley (1998: 751), for example, has pointed out. This is
illustrated by the English examples in (24), although the same point holds
cross-linguistically : 26
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 483

(24) a. John, I ' m afraid that this time you're mistaken.

b. I ' m afraid, John, that this time you're mistaken.
c. I ' m afraid that this time, John, you're mistaken.
d. I ' m afraid that this time y o u ' r e mistaken, John.

The examples in (24) clearly indicate that vocatives can interrupt the sen-
tence at various points, and thus that any material to their left (other than
interjections or other extrasentential expressions) might still be contained in
the sentence. This indicates, in turn, that vocatives can reliably detect sen-
tence boundaries only when no material occurs to their left.
But here another complication arises, this one related to clausal conjunc-
tions, another in the set of peripheral expressions described above. The
complication here is that vocatives may occur to the left of clausal conjunc-
tions l i k e y a d i ' i f , as shown in (25):

(25) deva yadi...

lord-VOC if
'Lord, i f . . . ' (SG 18.1.3)

This complication, however, can be readily handled in terms of the standard

observation (e.g. Killingley and Killingley 1995: 40^11) that conjunctions
in Sanskrit divide into different distributional classes. Thus, while conjunc-
tions like iti ' s o ' , as given in (22a), might be treated as "extrasentential", on a
par with vocatives, others like ca 'and', as given in (23a-b), are recognized
as enclitics that occur after the first word in the clause. The enclitic status
of this conjunction is also demonstrated in (26), where it follows a sentence-
initial verb. Recall that the occurrence of the verb here is, by hypothesis,
the result of the free arrangement of sentence constituents rather than of
verb movement, consistent with an enclitic status for the conjunction:

(26) bhavati [CNJ ca ] [NPs dhümah ] [ PP agnim antarena]

is and smoke-NOM fire-ACC without
tat na [ A P tat-hetuh] syät
then not it-cause would be
'Yet if smoke arises without fire, then (smoke) would not have it (i.e.
fire) for its cause.' (PV 22.9)

What is still left to determine, though, is the status of conjunctions that may
occur non-initially but are not known to be clitics. These include not only
484 Brendan Gillon and Benjamin Shaer

yadi, as just noted, but also tar hi 'then', as given in (23c). Given the latter's
apparent function of relating one sentence to another in a discourse, it
seems reasonable to assign it an "extrasentential" status and take its non-
initial occurrence to be explainable in much the same way that such an
occurrence of vocatives is. (Note, though, that the possibility of its being
more tightly integrated into the sentence cannot be ruled out, particularly
given analogous elements in German, like dann 'then', that are clearly inte-
grated into sentence structure.) 27 As regards yadi, however, its meaning in-
dicates a function as a clause-introducer, just like its counterparts in English,
German, and other languages, and thus a syntactic status as an element
outside S but still within the sentence periphery. 28 Such a status is further
supported by its occurrence to the right of another peripheral element, the
adverb tat 'then', as in (27):

(27) tat yadi...

then if
'Then, i f . . . ' (SG 18.1.4)

Of course, a great deal of work remains to be done to sort out the analysis
of conjunctions and other peripheral elements. Yet, the importance of doing
so is clear, given the role of such elements in determining sentence bounda-
ries and in helping to distinguish between extracted and freely arranged
phrases. Among the many questions left unanswered here is whether we
can, in fact, isolate a class of sentence-internal conjunctions in Classical
Sanskrit, which cannot occur higher than extracted elements; and, relevant
to the first question, whether the language permits multiple extraction of
phrases. Interestingly, the latter question, in turn, is closely related to the
question of whether Classical Sanskrit had a VP, as discussed in §4.3. That
these three questions are indeed interrelated becomes clear if we once again
consider the sentence given in (23c), which is repeated below:

(28) [ s [s> [ A pyah]I [CNJ tarhi] [ NP samagrena hetunä ]

which then complete-INST cause-INST
[ NPs Ι kärya-utpädah ] anum!yate] S ' [NPS sah] [ADV katham]
effect-arising-NOM is inferred it how
[NP tri-vidhe hetau ] antarbhavati] s
three-fold-LOC reason-LOC is included
'How, then, is the arising of an effect which is inferred through (its)
complete (set of) causes included within the threefold reason?'
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 485

Note that on an analysis of Classical Sanskrit in which it contains a VP,

such a sentence would involve two operations of leftward extraction, one
moving the AP yah out of the subject NP, as indicated, and another moving
the NP samagrena hetunä out of the VP whose head is the verb anurmyate.
In contrast, on a "no VP" analysis of the language, only the AP would be
extracted. Significantly, the latter possibility would derive additional sup-
port if we could show that conjunctions like tarhi actually indicated a
boundary for base-generated elements, since elements to their left could
then be seen as moved there and elements to their right as occupying a po-
sition related to the free arrangement of sentence elements.
These considerations also have consequences for the possibility of mul-
tiple fronting in the language. This is because the only purported cases of
such fronting that we have found so far are those like (28), in which the
innermost of two purportedly fronted elements would be analysed as ex-
tracted only if we posit a VP in Classical Sanskrit. If the language had no
VP, its position would result simply from the free arrangement of wild tree
daughters. Of course, the foregoing discussion of "edge issues" in Classical
Sanskrit has been rather cursory and highly speculative. Nevertheless, it
shows that these issues are important ones for the analysis of the language's
sentence structure, and that a more detailed investigation of them is clearly
called for.

6. Conclusion

In this chapter, we have investigated the question of word order freedom in

Classical Sanskrit, making extensive use of two searchable corpora. Building
on the work of Staal (1967), we showed that Staal's idea of a "wild tree", in
which the daughters of a given node constitute a set of unordered nodes,
captured a range of attested word orders without recourse to the assumption
that these word orders were derived from a more basic one. Moreover, em-
pirical problems that we discovered with the original "wild tree" analysis
turned out to have interesting solutions that we intend to explore in future
research. One involved a "flat" sentence structure for Classical Sanskrit,
consistent with a good deal of research into languages with substantial
word order freedom. This suggestion preserved Staal's idea that word order
variation need not involve "basic" and "derived" sentence structures, and
extended this idea to account for two linearization patterns that Staal's
original proposal did not capture. Another, which similarly extended ideas
486 Brendan Gillon and Benjamin Shaer

already found in Staal's original study, involved deriving certain phrase-

structure discontinuities from extraction processes broadly similar to those
attested in other languages.
Despite these very promising results, a great deal of work remains to be
done, as we have emphasized at various points in the discussion. This work
involves not only verifying the basic claims about word order that we made
above and extending their empirical coverage, but also offering a more
explicit account of the phrase structure of Classical Sanskrit itself. Inevita-
bly, progress on all of these fronts will be slow, given the considerable
hurdles created by our reliance on textual evidence and lack of access to
native speakers for the testing of our hypotheses.
The Apte corpus in particular highlights the difficulties of conducting
research on word order based only on textual sources. Despite the enor-
mous usefulness of this corpus, it is not without difficulties for syntactic
research. For example, although the sentences that constitute this corpus -
with which Apte illustrated traditional Indian wisdom regarding the usage
of particles and of verbal and nominal forms - are drawn from a rich array
of sources, these sources represent the best of Classical Sanskrit literature
and thus not the Classical Sanskrit corpus as a whole. In addition, although
Apte chose these sentences to illustrate points of morphology and particle
usage and not to defend any particular view of Classical Sanskrit word
order, this alone cannot ensure that they are broadly representative of the
language's word order patterns or that they include the structures required
to confirm or disconfirm particular hypotheses.
What this means, then, is that a broader range of texts needs to be inves-
tigated to put our claims about Classical Sanskrit to the test. This is without
doubt a truly daunting task. Nevertheless, the recent proliferation of digit-
ized versions of Classical Sanskrit texts gives us some hope that the task is
a tractable one - in particular, because of the possibility of building tree
banks of Classical Sanskrit sentences, files of syntactically and morpho-
logically tagged sentences, which can be searched. We believe that the
increased availability and use of such databases have the ability to radically
transform the study of Classical Sanskrit and other dead languages.
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 487

Notes

* This study grows out of the first author's research on Classical Sanskrit, as
reported, in particular, in Gillon (1996, 2004), which the second author has fol-
lowed closely over many years. The first author wishes to express his gratitude
to his teacher and friend Sri Venugopalan for assistance in the research on
which the study is based; and to the late Pandit Sri T. S. Snniväsa Sästri for
helpful advice on the analysis of various sentences. The second author wishes
to thank Philippa Cook, Werner Frey, Ewald Lang, and Luis Lopez for helpful
discussion. Both of us wish to thank Steven Schäufele for very graciously
sending us a copy of Schäufele (1990), Ruth Kempson for detailed comments
on an earlier version of this chapter, and Katalin E. Kiss for her considerable
efforts to put this volume together.
1. Interestingly, Ross's original proposal took this operation to be part of a "sty-
listic component" rather than of the transformational component of the gram-
mar (on this, see Thräinsson 2001: 154-255). As such, recent work on such
operations that conceives of them as part of the PF component brings them
more in line with Ross's original conception and with the ideas presented here.
2. This is perhaps easier said than done; and inevitably our sketch lends itself
more easily to constraint-based than derivation-based formalization.
3. Although adjectives and nouns in Sanskrit are sometimes thought to be non-
distinct categories, Dash and Gillon (1995: 291-294) present some reasons for
distinguishing them.
4. We use the following abbreviations in the glosses of this and other examples:
(word classes) ADV: adverb, CNJ: conjunction, INJ: interjection, PRT: particle;
(cases) ABL: ablative, ACC: accusative, DAT: dative, ERG: ergative; GEN:
genitive, INST: instrumental, LOC: locative, NOM: nominative, VOC: vocative;
(numbers) SG: singular, DU: dual, PL: plural; (others) INF: infinitive; PRIOR-
COMP: prior completion; PRT: particle.
5. In these and other examples given in the text, we suppress sandhi.
6. A difference is generally recognized between the word order variation permis-
sible in Sanskrit poetry and that permissible in Sanskrit prose. Since we take
the former, on the one hand, to permit greater freedom than prose allows and,
on the other, to be subject to artificial constraints imposed by metrical con-
siderations, we shall restrict our discussion of word order to that of sentences
in prose texts.
7. Of course, as already noted in the text, such an assumption is rejected in Mini-
malist research, as indicated by Chomsky's (1995: 413) remark that "[t]here is
no clear evidence that order plays a role at LF or the computation from Ν to
LF."
8. Note that Staal's (1967) rules make use of only zero and maximal levels of
phrase structure, consistent with assumptions current at the time he was
writing - and, as it happens, with Chomsky's (e.g. 1995: 396) hypothesis that
488 Brendan Gillon and Benjamin Shaer

only minimal and maximal but not intermediate projections are visible "at the
interface and for computation". Because, like Schaufele (1990: 116), we have
found no language-internal evidence that leads us to abandon this assumption,
we shall make use of it throughout our discussion.
9. Staal's original rules also make reference to case and number morphology,
which we omit here.
10. In fact, Staal offers no explicit statement for deriving word orders in Classical
Sanskrit such that they are free up to the free transposition of sisters within a
constituent. Such a statement is, however, offered by Gillon (1996, 2004), who
conjectures that the words in a given sentence may be linearized in any way
whatsoever, modulo the constraint that he formulates as in (i), which he dubs
"Staal's Constraint":
(i) Let Τ be a labelled, order-free phrase marker associated with a sentence S;
let rij and nj be distinct sister nodes in T. Then, if any word in S dominated
by nj precedes any word in S dominated by nj, then every word in S domi-
nated by nj precedes every word in S dominated by η(.
Given this constraint, constituents will be continuous and word order will be
free among the sisters of a given constituent. Gillon's conjecture has subse-
quently been proven by Nielsen (2004).
As it happens, Staal (1967: 34, 79) recognizes the necessity of additional
extraction rules that create discontinuous constituents from these continuous
ones, in order to capture additional word order patterns attested in Classical
Sanskrit. On this, see §4.3 below. However, this constraint on linearization ap-
pears to apply equally well to the output of the syntax in which such extraction
rules also figure.
11. Short references to this corpus are indicated with "SG" (abbreviating the title of
Apte's work, Student guide to Sanskrit composition), followed by the relevant
chapter, exercise set, and example sentence.
12. Note that VP and NP are separated by the conjunction tu in (5b), which occurs
in second position in the sentence. The question of what structural position
such elements occupy is taken up in §5.3.
13. We are assuming for present purposes that the negative element na 'not' is
dominated by VP. That this analysis is not the only one available will emerge
in §4.3.
14. Short references to this corpus are indicated with "PV", followed by the rele-
vant line and page of this work as edited by Gnoli (1960).
15. Staal (1967: 56) was already expressing scepticism of rules that apparently
served "to derive less frequent expressions from more frequent ones".
16. Note that compounds in Classical Sanskrit are always head-final (see e.g. Gillon
1995; Killingley and Killingley 1995: 42-46). However, because word-level
and phrase-level operations do not obviously make reference to the same set of
principles (pace Marantz 1997 and others; see Smirniotopoulos and Joseph
 Classical Sanskrit, "wild trees", and the properties of FWO languages 489

1998 for one recent defence of the syntax/morphology distinction), we do not

take this fact about compounding to shed much light on the matter of head po-
sition in Classical Sanskrit phrases.
17. Such headedness patterns are by no means unique to Sanskrit. For example,
Korean, as Ruth Kempson (personal communication) informs us, seems to dis-
play a similar head-final tendency that is overridden by speakers' frequent
placement of expressions after the head in conversation.
18. Although we shall be offering another possibility for the syntax of verbs in §4.3.
19. Note that this sentence is ambiguous between the reading given in the text -
consistent with the case-marking indicated in the glosses and ordering possibil-
ity spelled out in (4bi) - and the reading 'May Kumbhakarna and Meghanada
overcome Rama and Laksman'. The availability of the latter reading follows
from the identical form of the nominative dual and accusative dual endings and
an additional linear ordering possibility for simple transitive sentences, as de-
scribed in §4.3 and exemplified in (13b).
20. Staal is rather explicit on this point, emphasizing that "the system here advo-
cated does not intend to explain free word order by replacing 'trimmed' trees
by stratified 'wild trees'; this attempt would fail, e.g. for Sanskrit", since
"words may move beyond the boundaries of substrings dominated by particu-
lar constituents... This failure itself is one of the arguments which favours the
assumption that word order cannot be described in terms of trees, whether
trimmed or wild. Therefore, the conclusion is drawn that word order does not
belong to the deep structure, but must be relegated to the surface structure"
(Staal 1967: 79, n. 6).
21. Note that the treatment of extraction most compatible with Staal's and our
rejection of a "basic" word order for Sanskrit is one that likewise rejects the
postulation of an earlier structure from which the extraction structure is de-
rived and captures extraction in representational terms.
22. Schäufele (1990: 95) reports the same restriction for Vedic Sanskrit.
23. Our sketch of this "non-extraction" possibility in the text was inspired by com-
ments from Ruth Kempson, who urged us to explore this possibility for (18)
and other sentences.
24. This problem as it pertains to Vedic Sanskrit is discussed e.g. by Schäufele
(1990: 69-71).
25. Schäufele (1990:60) offers such a characterization of Vedic Sanskrit vocatives,
which "[do] not 'count' as a member of the clause" and "may be freely gener-
ated within any phrasal constituent".
26. This is supported by a search of the Apte corpus, which revealed that sentence-
initial occurrences of vocatives were the most common, but that these forms
also occurred after interjections (SG 7.1.6, 10.1.5, 10.1.8, 11.1.7, 14.1.9,
14.2.5, 27.1.5), conjunctions (SG 27.2.2), and subjects (SG 14.1.5, 17.1.4), as
well as clause-finally (SG 11.1.3, 14.1.5).
490 Brendan Gillon and Benjamin Shaer

27. Thanks to Ewald Lang for his timely reminder of "integrated" and "non-
integrated" possibilities for peripheral elements in German.
28. Note that we have resisted labelling such elements in Sanskrit as complemen-
tizers, since it is unclear whether they have the same formal characteristics as
counterparts in other languages that are identified as such. On this matter as it
pertains to Vedic Sanskrit, see esp. Schäufele (1990: 146).

References

Apte, Vaman Shivaram

1885 The student's guide to Sanskrit composition: A treatise on Sanskrit
syntax for use of schools and colleges. Poona, India: Lokasamgraha
Press. [24th edition, I960.]
Baker, Mark C.
2001 The natures of nonconfigurationality. In Handbook of contemporary
syntactic theory, Mark Baltin and Chris Collins (eds.), 407^138.
Maiden, Mass./Oxford: Blackwell.
Baltin, Mark and Chris Collins (eds.)
2001 Handbook of contemporary syntactic theory. Maiden, Mass./Oxford:
Blackwell.
Birner, Betty and Gregory Ward
1998 Information status and noncanonical word order in English. Am-
sterdam: Benjamins.
Bresnan, Joan
1982 Control and complementation. In The Mental Representation of
Grammatical Relations, Joan Bresnan (ed.), 282-390. Cambridge,
Mass.: MIT Press.
Canedo, Jos
1937 Zur Wort- und Satzstellung in der alt- und mittelindischen Prosa.
Göttingen: Vanderhoeck & Ruprecht.
Chomsky, Noam
1965 Aspects of the theory of syntax. Cambridge, Mass.: MIT Press.
1995 Bare phrase structure. In Government and Binding Theory and the
Minimalist Program, Gert Webelhuth (ed.). Oxford/Cambridge,
Mass.: Blackwell.
2005 On phases. Unpublished Ms., MIT, Cambridge, Mass.
Curry, Haskell
1961 Some logical aspects of grammatical structure. In Structure of lan-
guage and its grammatical aspects: Proceedings of the 12,h sympo-
sium in applied mathematics, American Mathematical Society, 5 6 -
68.
 Classical Sanskrit, "wild trees", and the properties of FWO languages 491

Dash, Siniruddha and Brendan Gillon

1995 Adjectives in Sanskrit. Adyar Library Bulletin 1995: 285-294.
Delbrück, Berthold
1878 Die altindische Wortfolge im ζ; at a ρ at h a - Β r äh a m a η a dargestellt.
Syntaktische Forschungen 3. Halle: Buchhandlung des Waisenhauses.
Donohue, Cathryn and Ivan A. Sag
1999 Domains in Warlpiri. Unpublished Ms., Stanford University.
Espinal, M. Teresa
1991 The representation of disjunct constituents. Language 67: 726-762.
Fanselow, Gisbert
forthc. Free constituent order: A Minimalist interface account. Folia Lin-
guistica.
Gazdar, Gerald, Ewan Klein, Geoffrey Pullum and Ivan Sag
1985 Generalized phrase structure grammar. Oxford: Basil Blackwell.
Gillon, Brendan
1995 The autonomy of word formation: Evidence from Classical Sanskrit.
Indian Linguistics 56: 17-52.
1996 Word order in Classical Sanskrit. Indian Linguistics 57: 1-35.
2004 Subject predicate order in Classical Sanskrit. In Language and
grammar: Studies in mathematical linguistics and natural language,
Philip Scott, Claudia Casadio and Robert Seely (eds.). Stanford: CSLI.
Gnoli, Raniero (ed.)
1960 The Pramänavärttikam of DharmakTrti: The first chapter with the
autocommentary. Text and critical notes. Serie Orientale Roma 23.
Rome: Istituto Italiano per II Medio ed Estremo Orient.
Gonda, J.
1952 Remarques sur la place du verbe dans la phrase active et moyenne
en languge sanscrite. Utrecht: Oosthoek.
Hale, Mark
1996 Deriving Wackernagel's Law: Prosodic and syntactic factors deter-
mining clitic placement in the language of the Rigveda. In Ap-
proaching second: Second position clitics and related phenomena,
Aaron L. Halpern and Arnold M. Zwicky (eds.). Stanford: CSLI
Publications.
Kathol, Andreas
2000 Linear syntax. Oxford/New York: Oxford University Press.
Killingley, Siew-Yue and Dermot Killingley
1995 Sanskrit. Munich: Lincom Europa.
Kluender, Robert
1998 On the distinction between strong and weak islands: A processing
perspective. In The limits of syntax, Peter Culicover and Louise
McNally (eds.), Syntax and Semantics 29: 241-279. New York: Aca-
demic Press.
492 Brendan Gillon and Benjamin Shaer

Kroeger, Paul R.
1993 Phrase structure and grammatical relations in Tagalog. Stanford:
CLSI.
Lambrecht, Knud
1994 Information structure. Cambridge: Cambridge University Press.
Laughren, Mary
1989 The configurationality parameter and Warlpiri. In Configurationality:
The typology of asymmetries, Laszlo Maracz and Pieter Muysken
(eds.), 319-353. Dordrecht: Foris.
Lazzeroni, Romano
1998 Sanskrit. In The Indo-European languages, Anna Giacalone Ramat
and Paolo Ramat (eds.), 98-124. London/New York: Routledge.
Marantz, Alec
1997 No escape from syntax: Don't try morphological analysis in the
privacy of your own lexicon. In Proceedings of the 21s' annual Penn
Linguistics Colloquium, Alexis Dimitriadis et al. (eds.), Working
Papers in Linguistics 4 (2): 201-225. Philadelphia: Department of
Linguistics, University of Pennsylvania.
McCawley, James D.
1998 The syntactic phenomena of English. 2 nd edition. Chicago/London:
University of Chicago Press.
Nielsen, Janet
2004 A proof and discussion of Staal's theorem. Honours thesis, Depart-
ments of Linguistics and Mathematics, McGill University, Montreal.
Pollard, Carl and Ivan A. Sag
1994 Head-driven phrase structure grammar. Chicago: University of
Chicago Press.
Prince, Ellen
1981 Toward a taxonomy of given-new information. In Radical pragmatics,
Peter Cole (ed.), 223-56. New York: Academic Press.
Ross, John R.
1967 Constraints on variables in syntax. Unpulished doctoral dissertation,
MIT, Cambridge, Mass.
Schachter, Paul and Fe Τ. Otanes
1972 Tagalog reference grammar. Berkeley, CA: University of California.
Schäufele, Steven W.
1990 Free word-order syntax: The challenge from Vedic Sanskrit to con-
temporary formal syntactic theory. Doctoral dissertation, University
of Illinois, Urbana-Champaign.
Smirniotopoulos, Jane C. and Brian D. Joseph
1998 Syntax versus the lexicon: Incorporation and compounding in Modern
Greek. Journal of Linguistics 34: 447-^488.
 Classical Sanskrit, "wild trees ", and the properties of FWO languages 493

Speyer, J. S.
1896 Vedische und Sanskrit-Syntax. (Grundriss der Indo-Arischen Philo-
logie und Altertumskunde 1 (6).) Strassburg: Trübner.
Staal, J. F.
1967 Word order in Sanskrit and Universal Grammar. (Foundations of
Language, supplementary series, 5.) Dordrecht: Reidel.
Thräinsson, Höskuldur
2001 Object shift and scrambling. In Handbook of contemporary syntactic
theory, Mark Baltin and Chris Collins (eds.), 148-202. Maiden,
Mass./Oxford: Blackwell.
Vallduvi, Enric
1993 Information packaging: A survey. Research Paper HCRC/RP-44.
Edinburgh: University of Edinburgh.
A particular coordination structure
of Indo-European flavour

Emanuele Lanzetta and Lucio Melazzo

1. Preamble

Coordination is traditionally defined as a symmetrical relation between two

or more forms that have equivalent roles (see Matthews 1993: 100). As
regards the structure, such a relation is considered to be made up of con-
juncts viewed as XPs sisters to one another, for they are dominated by one
common maximal projection. According to Munn (1992: 18-21) a structure
like this cannot account for some facts affecting coordinated constituents. A
comparison between the two alternatives of (1), which we have taken from
Munn (1992: 19), proves to be especially interesting:

(1) a. John left. And he didn't even say goodbye,

b. *John left and. He didn't even say goodbye.

As Ross (1967: 90-91) has already observed, (1) shows that the conjunction
forms a constituent with the second conjunct and definitely not with the
first. Moreover, a symmetrical structure cannot explain why an R-expression
in the first conjunct may be co-referent with a pronoun in the second and
not vice versa, as is revealed by (2), which we have again taken from Munn
(1992: 20):

(2) a. John'sj dog and hei went for a walk,

b. *Hej and John's; dog went for a walk.

If (2a) and (2b) were to admit of a symmetrical structure as their source

John and he would c-command each other in both cases, and consequently
there would be a standard violation of Principle C of the Binding Theory by
reason of referential John's being bound by he. Thence Munn (1992) pro-
poses that coordinated constituents should originate from an asymmetrical
structure like that of (3), in which the conjunction is a functional head. He
496 Emanuele Lanzetta and Lucio Melazzo

calls the maximal projection of this head a Boolean phrase (BP), the term
Boolean standing for any type of logical operator. In (3) the entire BP,
whose head B° has the second conjunct, the YP, as its complement, is as-
sumed to be adjoined to the right of the first conjunct, the XP:

(3) [XP[XP][BPB°[YP]]]

Like Munn, Kayne (1994) imagines a structure in which the coordinating

conjunction is the head of a &P, a phrase having the second conjunct as its
complement. However, in a manner consistent with the assumptions of the
theory of the antisymmetry of syntax he puts forward (4) as a modification
of (3). Owing to the Linear Correspondence Axiom, which excludes any
right-adjunction, he supposes that the first conjunct occupies the specifier
position of the &P:

(4) UP [XP] UP &° [YP]]]

We shall adopt (4) here in an attempt to account for certain phenomena of

coordination in some ancient Indo-European languages that have not yet
been given an exhaustive explanation. The paper is organized as follows: in
section 2 structure (4) is applied to the study of coordination by an enclitic
conjunction and the phenomenon of the so-called Indo-Iranian inverse ca is
presented; 1 in section 3 an analysis of this phenomenon is attempted, and
then the same analysis is extended to cases of asyndetic coordination in the
Indo-Iranian languages and Hittite; in section 4 some concluding remarks
are made.

2. Coordination by an enclitic conjunction and inverse ca

The antisymmetric approach to syntax permits a straightforward explana-

tion of the type of coordination existing in those ancient Indo-European
languages in which an enclitic coordinating conjunction deriving from
Indo-European *kwe is found. Let us consider the Latin expression (5):

(5) terra marique

land-ABL sea-ABL-and-CONJ
'by land and by sea'.
 A particular coordination structure of Indo-European flavour 497

Kayne (1994: 143, note 3) suggests structure (6) as a source from which
cases like (5) are obtained:

(6) [&p [XP terra] [ & P que [ YP mari]]]

In his view the head of the YP rises leftward to the higher head position,
where it is left-adjoined to que·}

(7) [&P [XP terra] [ & P mari w + que [ YP [Y° t w ] ]]]

If coordination is realized by two items of que, K a y n e ' s theory predicts that

the head of each conjunct will be moved out of its phrase and left-adjoined
to the enclitic conjunction immediately above, as (8) shows:

(8) romulae genti date

of-Romulus-DAT people-DAT give-PRES-IMP-2PL
remque prolemque
wealth-ACC-and-CONJ issue-ACC-and-CONJ
'give the people of Romulus both wealth and issue'. (Hor., C. Sec. 47)

With regard to (8), however, two structural alternatives, (9) and (10), are
possible:
498 Emanuele Lanzetta and Lucio Melazzo

In (9) the &P 2 is complement of the &Pi, whereas in (10) the &Pi occupies
the Spec,&P2 position. W e shall soon see the reasons why the structure
illustrated in (10) is preferable to that presented in (9).
As expected, K a y n e ' s analysis of the enclitic conjunction is perfectly
appropriate for coordination by both a single and a double ca in the Indo-
Iranian languages, (j)a in Hittite 3 , te in Ancient Greek:

(11) a. yqm da mainiiü

which-ACC apportion-AOR-ING-2SG spirit-INSTR
ädräcä, [... ] xsnütam
fire-INSTR-and-CONJ satisfaction-ACC
'the satisfaction which you apportion by means of (your) spirit and
(your) fire'. (Y31.3aa')
D D
b. G[uls]as MAH=ja pärsijami
god-Fates-DAT god-GREAT-and-CONJ break-PRES-lSG
'(I) break (the bread and other) for the Fates and the Great Deities'.
(Wisurijanza Vs. 21)

c. patir andrön te theön te

father-NOM men-GEN and-CONJ gods-GEN and-CONJ
'father of both men and the gods'.

All the cases of coordination in (11) can be explained on the assumption

that the head of the second conjunct (cf. (12a)), or each of the conjuncts (cf.
(12b) which follows (10)) are left-adjoined to the immediately higher
to which they are enclitic:

(12) a. [ & ρ [ χ ρ Χ ° ] [ & p Y 0

+& 0
[YPV]]]

b. [ & P 2 UP, X° + [XP t x .]] [ & P 2 Y ° + & ° 2 [VP t Y °]]]

 A particular coordination structure of Indo-European flavour 499

Let us take another example from Ancient Greek to note that the analysis
we have been proposing has the advantage of explaining without any further
stipulations such coordination phenomena as (13), in which the enclitic
conjunction te and the stressed kai are adjacent:

(13) apö kratos te kal om δ η

from-PREP head-GEN and-CONJ and-CONJ shoulders-GEN

'from (his) head and shoulders'. (II. XVII, 205)

The starting structure of (13) might be in fact (14):

(14) [&P2 [&P, te [Xp kratos ]] [&p2 kai [YP ömön]]]

(13) might derive from (14) by left-adjunction of the head of XP to te. Such
an operation also brings about the adjacency of the enclitic conjunction to
the stressed one in the linear order of the sentence. Evidence for the cor-
rectness of the derivation of (13) from (14) seems to be provided by the
following Homeric verse:

(15) opsomenai te gheronth' hälion kai

visit-FUT-P.PLE and-CONJ old-man-ACC marine-ACC and-CONJ
dömata patros
halls-ACC father-GEN
'to visit the Old Man of the Sea and the halls of (our) father'.
(II. XVIII, 141)

Here the first conjunct remains in its original position of complement of te,
and in PF this conjunction leans on the participle opsomenai4.
Among Indo-Iranian languages Vedic Sanskrit and Avestan exhibit a
peculiar type of coordination by the enclitic form ca. In this usage such an
enclitic conjunction is called inverse ca , for in the pronunciation it rests on
the first conjunct rather than on the second 6 :

(16) havyani vocati devebhyas ca

oblations-ACC announce-PR£S-SUBJ-3SG gods-DAT-and-CONJ
pitribhya
fathers-DAT
'let him (seil. Agni) announce the oblations to the gods and the
fathers'. (RV, X 16.1 l c - d )
500 Emanuele Lanzetta and Lucio Melazzo

The idiosyncratic character of this kind of coordination has not slipped the
attention of several scholars, who have dealt with the question either by
referring to positional problems connected with Wackernagel's Law (see
Klein (1985: 193), who, however, argues that there are three other ways of
explaining this phenomenon), or alternatively by discovering a likely bidi-
rectional property of the enclitic conjunction (Melazzo 1997) 7 . As regards
the origin of inverse ca, any diachronic investigation which tends either to
establish the precedence of the non-inverse over the inverse use of ca or
describe the latter as secondary compared with the former may be left out
of consideration. However, Klein's (1985: 191-192) hypothesis arouses
some interest. According to this scholar, the use of inverse ca might have
arisen out of coordination structures with a double ca, probably as a result
of the deletion of the enclitic conjunction leaning on the second conjunct
when uttered. Such a hypothesis may in truth constitute an important start-
ing-point for an attempt to determine the precise structural configuration of
coordination by inverse ca from the antisymmetric perspective we have
adopted. In point of fact, such a phenomenon as inverse ca could create
some problems for an approach to coordination from an antisymmetric per-
spective on syntax. Let us assume that in a coordination structure the con-
junction is the head of a phrase of which the positions of specifier and
complement are occupied by the first and the second conjunct respectively.
The representation of this phrase, whose head is a non-inverse ca, will
therefore be the following:

(17) &P

XP &P

YP

ca

Starting from (17) coordination by non-inverse ca will be realized through

the adjunction of the head of the YP to the head of the &P:
 A particular coordination structure of Indo-European flavour 501

(18) &P

XP &P

YP

ty°

N o w (18) reproduces the structural representation (12a) fairly faithfully. If

coordination by inverse ca is assumed to be a direct result of some deriva-
tional mechanism starting from the structure in (18), then it will be evident
that such a mechanism may be nothing but the movement of a couple
formed by the second conjunct and ca, and that this movement is a head-
movement. In accordance with the basic principles of the antisymmetric
theory, a non-terminal node, while moving, has to pull along with itself its
specifier if this is projected. It seems, however, that a head-movement
should be excluded. The following citation is a case worth examining:

(19) dyaus ca yäm prthivi vävridhate

heaven-NOM-and-CONJ whom-ACC earth-NOM strengthen-IMPF-IND-3SG
'he (seil. Agni) w h o m (both) heaven and earth strengthened'.
(RV, VII 7 . 5 c )

In (19) dyaiis e prthivi are coordinated by an inverse ca. Besides, dyaiis and
ca happen to precede the relative pronoun yäm. The occurrence of an ele-
ment on the left of either a relative pronoun or a complementizer is usually
not an isolated phenomenon in Vedic Sanskrit or in other Indo-European
languages. As for Vedic Sanskrit, Hale (1987: 41) puts the fact at issue
down to a topicalization movement to the initial position of the sentence.
Rizzi (1997) has recently described topicalization as a syntactic operation
which moves a phrase endowed with a [topic] feature into the specifier po-
sition of a projection of the left periphery of the sentence, whose head is
provided with the same feature so as to realize a specifier-head agreement
configuration (see also Bianchi 1999: 183). Should this approach be ap-
plied to (19), the sequence [dyaiis ca\ will structurally represent a non-
terminal node. It is therefore clear that the movement of a head cannot de-
termine the correct configuration of the constituents in (19). 8 As a further
502 Emanuele Lanzetta and Lucio Melazzo

consequence this observation makes it unlikely that the order of the con-
juncts dyaüs and prthivi originates from such a structure as (18). Hence
coordination by inverse ca cannot derive through a syntactic operation from
a structure where coordination is expressed by non-inverse ca. If inverse ca
is base-generated, we must ask ourselves what structure could be assigned
to the string devebhyas ca pitribhya in (16). Most importantly, it must be
excluded that the conjunction ca dominates both the constituents; otherwise
there would be room for a symmetric structure, the possibility of which is
not allowed on theoretical grounds. We prefer to remain within the anti-
symmetric framework and hypothesize a structure with two heads project-
ing two separate &Ps. Naturally, one of these heads is ca and the other will
be null. This means that coordination by inverse ca will involve both the
structural alternatives represented in (9)—(10), except that the second con-
junction is not phonetically realized:

(20) a. [ &Pl ca [ &?2 [ XP devebhyas] [&p2 [ YP pitribhya]]]]

b. [&p2 Up, ca [Xp devebhyas]] [&p2 [γρ pitribhya]]]

Now that we have seen that the first conjunct of a structure containing an
inverse ca can be topicalized and that a topicalized element always has to
be a phrase, it is only the representation in (20b) that reproduces a coordi-
nating phrase with inverse ca correctly. Indeed since it stands in Spec,&P 2 ,
only in (20b) can the &Pi still move after devebhyas has been left-adjoined
to ca. Conversely in (20a), the &P| with head ca has the &P 2 as its com-
plement: in this state of things the &P| cannot move as a maximal projec-
tion without the pied piping of its complement. If we are on the right track,
(21), which exhibits the adjunction of the head of the XP to ca, describes an
immediate transition from the structure in (20b).
 A particular coordination structure of Indo-European flavour 503

3. The hypothesis of adjunction to &°

The correctness of (21) remains to be settled in that the grounds for justify-
ing the empty head of the &P2 still need to be given. Klein (1985) suggests
that in the instances of coordination that we have regarded as being ob-
tained from such a structure as (21) there should be the deletion (of course
at PF) of ca which is the head of &P2, but this cannot be an adequate expla-
nation because no reasons for the need for such a deletion are adduced by
that scholar. On the other hand, the existence of coordination by non-
inverse ca points to the absolute arbitrariness of any deletion phenomena in
the inverse ca construction. Let us briefly consider the well-known asym-
metry in (22) taken from Kayne (1994: 57) which exemplifies the coordina-
tion of more than two conjuncts:

(22) a. I saw John, Bill and Sam.

b. *I saw John and Bill, Sam.

As stated by Kayne (1994: 57) and shown by (23), in (22a) a subpart [Bill
[and Sam]], with and as the head, has to be taken into account. To this John
can be adjoined at the left, licensed by another head, which in English may
not be overt:

(23) [John [X° [Bill [and Sam]]]].

There could also be the same covert head in (22b), and (24) ought to repro-
duce its structure:

(24) *[John [and [Bill [X° Sam]]]].

Kayne (1994: 58) ascribes the grammatical ity of (22a) and (23) to the LF
raising of and, which licenses the phonetically unrealized X°, the ungram-
maticality of (22b) and (24) being due to the non-existence of a parallel LF
lowering of the overt conjunction capable of licensing the covert X°. If we
were to apply the same line of reasoning to the inverse ca coordination that
we have described in (21), we should paradoxically conclude that such a
construction is ungrammatical. This conclusion clashes, however, with the
evidence of the data which testify that such a coordination pattern of the
type in (21) belonged to the grammatical competence of the Vedic Sanskrit
and Avestan speakers.
504 Emanuele Lanzetta and Lucio Melazzo

It is now the time to put forward our solution to the question of how the
phonetically unrealized head is licensed in such cases as (21). As (25) shows
clearly, our solution consists in assuming that the head of the second con-
junct moves leftward in order to be left-adjoined to the covert head of
9
the coordinate phrase:

(25)

devebhyas x ca tx pitribhya y

Kayne (p.c.) notes that if we hypothesize a structure like (25), then we

should be able to mention some languages in which coordination could be
normally realized through empty heads, so that sentences like (26) would
be admissible:

(26) *I saw John, Mary.

In point of fact, as (27) makes clear, the Indo-Iranian and other Indo-
European languages repeatedly present constituents coordinated without
recourse to any conjunction:

(27) a. vS.nS hazaosaijho, vTspäijhö daidiiäi [·••]/ as am

V9.nä-PREV agreeing-NOM all-NOM propitiate-INF order-ACC
vohü manayhä, ιιχδΰ yäis
divine-INSTR thought-INSTR statement-INSTR which ! N S T R
ärmaitis / yazamnarjhö
Ärmaiti N oM worshippersNOM
'all the worshippers agree to propitiate the Order by means of the
divine thought (and) the statement with which Ärmaiti (is)'.
(Y 5 1 . 2 0 a a ' b b ' c )
 A particular coordination structure of Indo-European flavour 505

b. haumaiy ima xsagam fräbara [...]

he-NOM-I-GEN this-ACC kingdom-ACC bestow-IMPF-3SG
taya uvaspam umartiyam
which A cc possessed-of-good-horses-ACc possessed-of-good-men-ACC
'he bestowed on me this kingdom possessed of good horses (and)
possessed of good m e n ' . (DSf 10-12)
mv
c. n=as Halpa pait nu
and-CONJ-he-NOM city-Alep-DIRECT go-PAST-3SG and-CONJ
UKU URV
Halpan harnikta nu Halpas
city-Alep-ACC destroy-PAST-3SG and-CONJ city-Alep-GEN
NAM.RAmeS assu=set imu
Hattusi
PRISONERS OF WAR goods-ACC-their-ACC city-Hattusa-LOC
üdas
take-PAST-3SG
'and he went to Alep and destroyed Alep and took to Hattusas the
prisoners of war of Alep (and) their goods'. (Telipinus, IX 28)

d. uti sies volens propitius

so-that-CONJ be-PRES-SUBJ-2SG willing-NOM propitious-NOM
'so that (you) may be willing (and) propitious' (Cato, Agr. CL, 4)

The sentences in (27), where the conjuncts are in bold type, are taken from
Old Avestan, Old Persian 1 0 , Hittite and Latin texts. The type of coordination
illustrated by (27) can also involve more than two conjuncts. (28) is a good
Old Avestan example of this possibility: 11

(28) ärmatöis nä spantö, huuö cistl

Ärmaiti-GEN man-NOM prosperous-NOM he-NOM insight-INSTR
uxöäis siiaoSanä
statements-INSTR action-INSTR
'he (is) the prosperous man by the insight, by the statements (and) by
the (ritual) action (coming) from Ärmaiti'. (Y 51.21aa')

On the other hand, sentences in which the inverse ca coordinates three

members are not lacking in Old Avestan (see Kellens-Pirart 1990: 155) 12 ,
and (29) provides evidence for the truth of this statement:
506 Emanuele Lanzetta and Lucio Melazzo

(29) data vä anwrdtasca, utaiiüitT

given-NOM you-GEN immortality-NOM-and-CONJ two-youths-NOM
hauruuatäs draonö.
integrity-NOM portion-NOM
'immortality, two youths and integrity have been given by you as a
portion'. (Y33.8cc')

Hittite allows in (30) the asyndetic coordination of verb forms:

UZU UZU
(30) ser=a=ssan MG.GIG SA
ser-ADV-and-CONJ-ssarc-PARTC meat-liver meat-heart
kuirzi dai
CUt-PRES-3SG put-PRES-3SG
'and (he) cuts the liver (and) the heart (and) puts (them) here'.
(Wisurijanza Rs. 7 - 8 )

Here it is not immediately clear which phrasal categories are involved in the
coordination. Two VPs or at most two IPs might come into play. Naturally,
the assumption must be made that the complement in the second conjunct is
deleted because of its identity with the complement in the first conjunct.
Nothing in (30) allows one to hypothesize an instance of the coordination
of full sentences. However, cases like the following in which the verb is
left-adjoined to a sentence connective are to be found in Hittite:

(31) anda=kan halinas tessummius tarlipit

into-ADV-few-PARTC clay-GEN tessummius-ACC tarlipit-lNSTR
suwamus [...] petumini tarueni=ma=at
vessels-ACC bring-PRES-lPL call-PRES-lPL-but-CONJ-it-ACC
eshar
blood-ACC
' w e bring in tessummi-vessels of clay, filled with tarlipcr, we call it,
however, blood'.
(StBoT 8 I 2 6 - 2 7 = StBoT 25 no. 3, p. 6; Boley 1985: 230)

In (31) the verb tarueni is clearly attached to the left of the adversative en-
clitic conjunction ma. It must be considered that the unmarked word order
of the Hittite sentence requires the verb to be in final position and this con-
dition on the sequence of constituents is fairly strict in that language. A
similar instance is also to be found in archaic Latin:
 A particular coordination structure of Indo-European flavour 507

(32) utique [...] pastores ipecucuque

so-that-CONJ-and-CONJ shepherds-ACC herds-ACC-and-CONJ
salva servassis duisque
healthy-ACC keep-PRES-SUBJ-2SG give-PRES-SUBJ-2SG-and-CONJ
bonam salutem val iej tudinemque
good-ACC health-ACC prosperity-ACC-and-CONJ
'and so that (you) may keep the shepherds and the herds healthy and
give good health and prosperity'. (Cato, Agr. CL, 9 - 1 1 )

In (32) the verb duis is also adjoined to the left of que functioning as a
clause conjunction. Similarly, in archaic Latin the verb is subject to the
same rules of position that apply to the Hittite verb. A s far as the asyndetic
coordination of sentences is concerned, (33) proves even more revealing:

(33) n=as eszi=pät

and-CONJ-she-NOM remain-seated-PRES-3SG-pd/-PARTC
natt=as=apa aräi
not-NEG-she-NOM-apa-PARTC get-up-PRES-3SG
'she remains seated, she does not get up'.
(KBo XIX 163 IV 6; Watkins 1994: 304)

Here the negative item natt(a) occurs at the beginning of the second coor-
dinated sentence. Interestingly, the same sequence is previously found in
the same text with an important variation:

(34) n=as eszi=pat

and-CONJ-she-NOM remain-seated-PRES-3SG-pa/-PARTC
natta=ku(w)=as=apa aräi
not-NEG-and-CONJ-she-NOM-apa-PARTC get-up-PRES-3 SG
'she remains seated, and she does not get u p ' .
(KBo XIX 163 II 3 3 ' - 3 4 ' ; Watkins 1994: ibidem)

What marks the difference between (33) and (34) is the occurrence in the
latter of the enclitic particle ku, which Watkins (1994) has shown to be the
Hittite continuation of the Indo-European enclitic *kwe. A fair inference can
be drawn f r o m this evidence: in (33) natt(a) performs the identical process
by which its twin in (34) is left-attached to the conjunction ku, but the head
at which it has landed, is phonetically unrealized. 1 3
508 Emanuele Lanzetta and Lucio Melazzo

Something similar to what has been noted in Hittite is found in Old Persian
when the negative form naiy is involved. Thus, in (35) the second negative
seems to play the same role as e.g. Latin neque:

(35) yaSä haya tauviyä tayam skauditn

so-that-CONJ who-NOM stronger-NOM who-ACC weak-ACC
naiy jatiy naiy vimardatiy
not-NEG smite-PRES-IND-3 SG not-NEG destroy-PRES-IND-3SG
'so that the stronger does not smite or destroy the weak'. (DSe 3 9 - 4 1 )

There is not enough evidence here to prove this. However, the comparison
between (36a) and (36b) will arouse greater interest:

(36) a. yadiy imam hadugäm naiy apagaudayähy

if-CONJ this-ACC record-ACC not-NEG conceal-PRES-SUBJ-2SG
kärahyä 9ähy
people-GEN tell-PRES-SUBJ-2SG
'if you will not conceal this record (and, i.e. but) tell (it) to the
people'. (DB IV 5 4 - 5 5 )

b. yadiy imäm hadugäm apagaudayähy naiy

if-CONJ this-ACC record-ACC conceal-PRES-SUBJ-2SG not-NEG
&ähy kärahyä
tell-PRES-SUBJ-2SG people-GEN
'if you will conceal this record (and) not tell (it) to the people'.
(DB IV 5 7 - 5 8 )

In (36a) the asyndetic coordination concerns in all likelihood two IPs. In

(36b) the two conjuncts are also coordinated asyndetically, but the negative
naiy occurs at the beginning of the second one and the verb Sähy precedes
the indirect object kärahyä, whereas the latter comes before the former in
(36a). The peculiarity of the order of the elements in (36b) needs to be
stated more clearly. As can be easily observed in the first conjunct in (36a),
in Old Persian the negative precedes the verb placed in final position in the
clause, but it follows the arguments of this (see also Hale 1988). This im-
plies that in (36b) both the negative naiy and the verb Sähy have been
moved from their original positions, which can be further guessed by a
comparison of (36a) to (36b). It can therefore be assumed as a hypothesis
that naiy in (36b) has been moved into the empty head of the & P within
which the two IPs joined together are internally positioned. 1 4
 A particular coordination structure of Indo-European flavour 509

4. A few concluding remarks

Since its beginnings formal syntax has concentrated essentially on syn-

chronic generalizations, thus neglecting most typical questions of the dia-
chronic study of language. Such a state of affairs has only lately begun to
change now that interest in and work on historical syntax have increased
(see among others Roberts 1993; van Kemenade-Vincent 1997; Lightfoot
1999). There again, the imperialism of syntax has been peeping up through
the lines, so to speak (see, for example, Longobardi (2001), where the view
is convincingly expressed that formal syntax may sometimes enhance our
insight into such a traditional domain as etymological research). Our paper
is meant to be an attempt to deal with hystorical syntax, although it is not
the sole purpose. We dare not say that we have succeeded in our endeavour,
but we would like our readers to have the basic principles of our work
clearly stated. First, using Chomsky's (1964) classical approach as a model,
we mantain that observational adequacy is to be equated, as it were, to a
corpus linguistics of which the data must be checked with the help of phi-
lology, while descriptive adequacy and explanatory adequacy equal de-
scriptive synchronic linguistics and general linguistic theory respectively.
Secondly, it is precisely the acceptance of a set of reasoned ideas which
also impose strong constraints on the types of structures and operations
found in natural language grammars that has allowed us to tackle the study
of a significant, albeit minimal, fragment of the formal organization of
some languages, either ancient or dead, which have been traditionally con-
sidered so impassable an area of disorder and arbitrariness that traversing it
was more or less viewed as tantamount to crossing a minefield. Thirdly,
since the languages we have been dealing with are cognate with one an-
other, we have made allowance for some requirements of historical-
comparative linguistics.
In some of the languages we have been talking about coordination was
also realized by an enclitic conjunction whose various forms derived from a
common ancestor. We have therefore remained inside the classical paradigm
of historical-comparative linguistics. This type of linguistic research started
to draw early taxonomic conclusions from the pre-theoretical recognition of
the resemblance of lexical items in form and meaning. It was supported in
the process by the recognition of exact and regular sound correspondences,
and eventually received its salient characterization from Saussurean arbitrari-
ness. On this basis historical-comparative linguistics achieves the identifica-
tion of philogenetic relatedness among languages by comparing vocabularies
510 Emanuele Lanzetta and Lucio Melazzo

in the broadest sense of the term: the similarity of lexemes and m o r p h e m e s

(roots, affixes or inflections) that correspond strikingly to one another in
different languages is adduced as a decisive proof of philogenetic kinship.
In our case cognate f o r m s deriving from *kwe in some Indo-European
languages have turned out to have the same syntactic usage. Yet, when the
Hittite f o r m (j)a or asyndetic coordination were considered w e had already
gone beyond the boundaries of the classical paradigm of historical-com-
parative linguistics. A s a matter of fact, w e have been c o m p a r i n g rules of
grammar, n a m e l y syntactic rules, which appear less arbitrarily variable
from language to language. Various problems regarding some new system of
classification h a v e confronted us. Based on higher distinctive characteristics
which are likely to have existed further back in prehistory, this classification
needs a new paradigm. 1 5 N e w horizons are opening out before us all. Science
must go ahead. That is not a matter of opinion.

Notes

* The two authors worked jointly on the whole content of this article; however,
as far as legal requirements are concerned, E. Lanzetta takes official responsi-
bility for sections 1 and 3, and L. Melazzo for the remaining parts.
As regards the evidence that the authors took into account, the following
critical editions were used: Klingner (1970) and Mazzarino (1982), Kellens
and Pirart (1989), Carruba (1966) and Hoffmann (1984), Monro and Allen
(1902-1920 3 ), Aufrecht (1968) and Kent (1953).
In the examples taken from Carruba (1966) and Hoffmann (1984) we have
used continuous transcriptions instead of syllabic transcriptions, as used by the
authors.
The anonymous reviewer raises the objection that the examples we cited
were taken from poetic texts. Apart from the fact that this is not true, we would
like to state that although poetry may exhibit some unusual constructions,
these never actually break the rules of a given language. There may be differ-
ences of register between prose and poetry, but not of syntax. Though poetry
may contain constructions no longer in current usage, these nevertheless be-
long to the syntax of that language. In any case, our examples exhibit rules that
are valid in both prose and poetic texts.
1. The graphic form ca transliterates the (Vedic) Sanskrit conjunction. The Iranian
equivalent is now transcribed cä with a long vowel due to an Iranian ortho-
graphic rule. For the sake of brevity we shall write ca for both the Vedic San-
skrit and Iranian occurrences of this enclitic conjunction.
 A particular coordination structure of Indo-European flavour 511

2. It must be noted that the same explanation also holds good for disjunctive co-
ordination in those Indo-European languages which possess an enclitic dis-
junctive conjunction going back to the reconstructed form *ve.
The phrase category of nominal conjuncts is not specified any further here,
for we want to concentrate our attention on coordination structure. Only in (iv)
of note 8 will it be necessary to make use of precise structural hypotheses in-
volving the category of determiner to account for the peculiar order of the con-
stituents of the sentence. Thus in Indo-European languages the Ν Ρ can be
found in a complex structure in which there are various functional projections
dominated by a DP. We do not intend to assume this as a postulate on the basis
of the analyses conducted in modern languages, so we shall discuss the ques-
tion in a forthcoming paper.
3. The Hittite conjunction ij)a displays the same syntactic behaviour as the enclitic
conjunctions derived from *k" e in several languages having Indo-European as
their common ancestor, but it cannot be assimilated to them as far as etymology
is concerned. It is perhaps worth noticing that in ( l i b ) the conjunct to which
the enclitic form is attached is graphically represented by a Sumerogram.
4. In (15) the occurrence of te in second position might seem to be due to Wacker-
nagel's (1892) Law. Since it must be supposed that te has not undergone any
movement operations, the main question to solve is that of establishing the po-
sition of the participle opsomenai. If opsomenai has remained in situ since the
original formation of the sentence, then te attaches itself to this participial form
undoubtedly on account of phonological and/or prosodic reasons. If, however,
opsomenai is believed to have previously occupied the final position of the
sentence as expected in an SOV language like Ancient Greek, then the initial
position of this participle in the sentence has to be explained by assuming a
movement operation. Yet a left-adjunction of opsomenai to te has to be ex-
cluded, for such a head movement would be hindered by the presence of kai,
the tonic conjunction. Such being the case, the cliticization of te to opsomenai
will take place once all the syntactic operations have been performed, and con-
sequently it will again depend on phonological and/or prosodic reasons.
5. No sure case of inverse ca is testified in Old Persian texts, but see Schmitt
(1963).
6. The following citation from a Hittite text must be regarded as an interesting
case of multiple coordination which could contain a construction parallel to
Indo-Iranian inverse ca:
LÜMES
(i) DUMUme5-SU SESme5-SU gaenas=ses=sa
CHILDREN-HIS BROTHERS-HIS MEN-relatives-NOM-his-NOM-and-CONJ
LlJMEShassannas=sas Ü ERINmeS-SU
MEN-kin-GEN-his-GEN AND TROOP-HIS
"his children, his brothers and his relatives and the men of his kin (and)
his troop". (Telipinus I 3)
512 Emanuele Lanzetta and Lucio Melazzo

In ( i ) t h e l o g o g r a m s D U M U M E S - S U and S E S M E Ä - S U are c o o r d i n a t e d a s y n d e t i -
cally. They are followed by the conjunct LUMES gaenas=ses, to which the enclitic
conjunction (j)a in the assimilated form sa is attached. Finally, the conjuncts
LU hassannas=sas and ERIN MES -SU are joined by a conjunction denoted by
the logogram Ü. This sign is employed in Akkadian script to write the coordi-
nating conjunction, which is not enclitic in that language. A by no means trivial
question now arises. If the scribe is thought to have utilized the Akkadian word
with the intention of making use of a non-enclitic conjunction, then the type of
coordination in (i) merits consideration only on account of this unusual employ-
ment of a foreign form. Such being the case, it should be established whether
(i) would have a structure compatible with the antisymmetric analysis of coor-
dination. This does not seem to be the actual state of the matter. In point of fact,
providing the first two conjuncts are left out of consideration, as their structural
representation can be viewed as irrelevant at this moment, the remaining three
conjuncts should be imagined as having a structure of the following type:
0 0 Up, -JA [&p2 [χρ LU MES gaenas=ses] [ &P2 [ &Pl [ YP HJ MES hassannas=sas]
UP, Ü [zp ERIN m e S -SU]]]]]]
In (ii), where the adjunction of LU MES gaenas=ses to (j)a has been omitted, &P,
has the &P 2 as its complement. The head of this &P 2 has no phonetic content,
while Ü is the head of the &P 3 complement of the &P 2 . Now the question is
whether the covert head of the &P 2 can be licensed or not. For the time being,
there is room for doubt. As a matter of fact, (iii) shows that an analogous struc-
ture would be ungrammatical in languages like e.g. English (as to cases similar
to (ii) in Old Avestan, see Lanzetta 2 0 0 3 : section 2 . 0 ) :
(iii) *I saw both Mary, John and Luke.
If, however, on the basis of a perfectly plausible assumption the Akkado-
gramm Ü is held to hide the Hittite enclitic conjunction, then (i) will turn out
to contain an instance of inverse ca and therefore prove the existence of this
coordinate structure in that language. It is quite right to emphasize that on no
account can the Akkadogram Ü be the equivalent of the non-enclitic connective
nu, for this form occurs only at the beginning of the Hittite sentence.
7. In a more or less direct manner the phenomenon of inverse ca has been taken
into consideration by Zwolanek (1970: 80) and Dunkel (1982: notes 2, 38, 48).
Dressier (1965: 7 6 - 7 7 ) argued that the Greek and Latin enclitic conjunction
was not used in a way similar to that of the cognate Indo-Iranian conjunction.
8. The impossibility of a head movement operation shifting a non verbal element
to the left periphery of the sentence might seem to be belied by the following
quotation:
(iv) püro yäd asya sampinäk
fortresses-ACC when-CONJ he-GEN destroy-AOR-2SG
'when you destroyed his fortresses.' (RV, IV 30.13d)
 A particular coordination structure of Indo-European flavour 513

The syntactic relationship between püro and asya could cause one to think that
the latter has remained in the position of either the specifier or the complement
of the phrase with the former as a head. No matter which of these two alterna-
tives is the right one, the operation by which püro has been shifted should be
depicted as a head movement, for a maximal projection cannot move and leave
its specifier or its complement (or both of them) behind. The sheer difficulty of
puzzling out the answer to the question that (iv) raises depends in truth on our
poor knowledge of the constituency relationship within ancient languages, es-
pecially in consideration of the diffusion of such a phenomenon as hyperbaton
in them (see Devine-Stephens (2000) for some careful examination of hyper-
baton in Ancient Greek). A way of regarding (iv) as an instance of topicaliza-
tion of a maximal projection consists in assuming an extended nominal structure
and a DP node in which the N P with püro as a head occupies the complement
position. Under these conditions, following Giusti (2001: 164-166), the demon-
strative pronoun asya can be generated in Spec,NP or in a specifier position
within that extended structure and then move into Spec,DP. If we are on the
right track, the topicalization of püro in (iv) implies either an N P or some func-
tional projection dominating the NP and dominated by the DP, and not a head.
Such a hypothesis is corroborated by Cinque's (1995: 298) serialization of
APs. As stated by this Italian scholar, the modifiers of an NP are arranged in a
set order, each of them filling the specifier position of different functional pro-
jections placed between the DP and the Ν P. As for nouns denoting objects, the
first modifier immediately to their left turns out to be a possessive. In (iv) the
genitive form asya corresponds to a possessive, and thus can be held to remain
in the specifier position of that FP between the DP and the N P that is adjacent
to the latter.
9. If so, we could think of a kind of mechanism of specifier-head agreement of
the type proposed by Kayne (1994: 143, note 2) to account for those cases in
which the &P has a distributive value.
10. (27b) can be compared to (v), where nearly the same terms occur and are
joined by ca:
(v) baga vazraka Auramazdä [...] haya uvaspa
god-NOM great-NOM Auramazda-NOM who-NOM good-horses-ACC
uraSäcä kunautiy
good-chariots-ACC-and-CONJ make-PRES-3SG
"Ahuramazdä (is) the great god [...] who makes good horses and good
chariots". (DSs 1 - 6 )
It must be noted that, even though the passage quoted in (v) is seriously dam-
aged, the string uradäcä can be easily read (see Kent 1953: 146). Nevertheless,
Lecoq (1997: 241) rounds off the passage in a different manner, spotting a
three membered coordination very similar to that of (vii) in note 11, except for
the order of the conjuncts.
514 Emanuele Lanzetta and Lucio Melazzo

11. A comparison can be made between (28) and the following sentence (vi),
where a three-membered coordination has ca as its end:
(vi) höi scantü manarjhä, uxöäis
they-NOM devote-themselves-AOR-iMP-3PL thought-iNSTR statements-iNSTR
siiaoSanäiscä xsnüm mazda
actions-lNSTR-and-CONJ choice-ACC Mazda-GEN
"let them devote themselves with thought, with statements and with (ritual)
actions to the choice of Mazda". (Y 53.2aa'b)
As can be easily seen, the last two conjuncts in (vi) would be identical to those
in (28), but for the negligible difference in number displayed by the noun
siiao9ana-. In both instances this noun happens to occur in the instrumental
Case but the number is singular in (28) and plural in (vi).
An Old Persian instance of three-membered asyndetic coordination similar
in meaning to (27b) and (v) of note 10 might be the following:
(vii) hausaiy xsagam fräbara [... ] taya
he-NOM-he-GEN kingdom-ACC bestow-iMPF.iND.3sG which-ACC
uraSaram uvaspam
possessed-of-good-charioteers-ACC possessed-of-good-horses-ACC
umartiyam
possessed-of-good-men-ACC
"he bestowed upon him this kingdom [...] possessed of good charioteers,
possessed of good horses (and) possessed of good men". (DSp 2 - 3 )
However, the text of (vii) is the result of some conjecture and supplementation.
The form ura&aram seems certainly readable to Kent (1953: 146), whereas
Lecoq (1997: 240) reads and interprets (vii) on the basis of his interpretation of
(v) of note 10.
12. Humbach (1991: 94, part I) considers the inverse ca of (29) to coordinate only
amaratas and hauruuatäs, for he regards utaiiuitT as a sociative instrumental
(see Humbach 1991:97; 112, parti).
13. The following sentence deserves some discussion:
(viii) brähmä ca giro dadhiri
p r a y e r s - N O M - a n d - C O N J songs-NOM b e - c o n c e n t r a t e d - P E R F - 3 P L
säm as min
together-ADV him-LOC
"the prayers and the songs are concentrated in him". (RV, VI 38.3c)
Klein (1985: 176) regards the use of ca in (viii) as potentially ambiguous. He
thinks that its interpretation remains uncertain, an inverse ca being recognizable
as well as a ca which coordinates (viii) to the previous sentence in the text. Let
us suppose that ca joins two sentences together. If so, the constituents brähmä
and giro would be coordinated asyndetically with the same grammatical pattern
that we already know. Moreover, the position of ca in (viii) could contradict
 A particular coordination structure of Indo-European flavour 515

H a l e ' s (1987) view that when joining sentences ca and va (the latter is the dis-
junctive conjunction) are subject to Wackernagel's Law and reach their position
in the string only after the process of topicalization of a constituent has taken
place. A s a matter of fact, on no account do there seem to be the discourse
conditions that would allow one to suppose brähmä to have been topicalized. It
m u s t also be noted that (15) does not support H a l e ' s theory in any way because
the enclitic conjunction te definitely does not join sentences there.
14. In Old Avestan negative constituents are coordinated by nöit ... naedä. The
spelling naedä results f r o m a process of adjunction of the negative nöit to the
enclitic *u, w h o s e existence in the language of the sacred writings of Zoroas-
trianism can be inferred f r o m Vedic Sanskrit. The following passage appears
to prove that the adjunction of the negative to is operative even if this head
is phonetically unrealised:

(ix) nöit nä mana, nöit sSnghä nöit

not-NEG nö-PARTC thoughts-NOM not-NEG pronouncements-NOM not-NEG
xratauuö / naedä varanä, nöit uxSä naedä
intellects-NOM nor-NEG choices-NOM not-NEG statements-NOM nor-NEG
siiaodanä / nöit daena, nöit uruuqnö
actions-NOM not-NEG religious-views-NOM not-NEG souls-NOM
hacinte
follow-PRES-IND-3PL-M
"neither the thoughts nor the pronouncements, nor the intellects, nor the
choices, nor the statements, nor the (ritual) actions, nor the souls follow
(it)". (Y 4 5 . 2 c c ' d d ' e e ' )

Here the alternation of naedä and nöit, i.e. of coordination by the conjunction
*u and the asyndetic coordination, m a y hint at the m o v e m e n t of nöit to an
empty
15. A first attempt to tackle this problem has been very recently m a d e by Guardiano-
Longobardi (2003).

References

Aufrecht, W. T h e o d o r
1968 Die Hymnen des Rigveda. 2 Teile. Wiesbaden: Otto Harrassowitz.
Bianchi, Valentina
1999 Consequences of Antisymmetry. Headed Relative Clauses. Berlin/
N e w York: Mouton de Gruyter.
Boley, Jacqueline
1985 Hittite and Indo-European Place Word Syntax. Sprache 3 1 : 2 2 9 - 2 4 1 .
516 Emanuele Lanzetta and Lucio Melazzo

Carruba, Onofrio
1966 Das Beschwörungsritual für die Göttin Wisurijanza. Wiesbaden:
Otto Harrassowitz.
Chomsky, Noam
1964 Current Issues in Linguistic Theory. In The Structure of Language, J.
Fodor and J. Katz (eds.), 50-118. Englewood Cliffs, New Jersey:
Prentice Hall.
Cinque, Guglielmo
1995 Italian Syntax and Universal Grammar. Cambridge: Cambridge Uni-
versity Press.
Devine, A. M. and Laurence D. Stephens
2000 Discontinuous Syntax. Hyperbaton in Greek. New York/Oxford:
Oxford University Press.
Dressier, Wolfgang
1965 Inverses te, inverses -que. Glotta 43: 76-78.
Dunkel, George
1982 The Original Syntax of Conjunctive *-k w e' Sprache 28: 129-143.
Fodor Jerry A. and Jerrold J. Katz (eds.)
1964 The Structure of Language. Englewood Cliffs, New Jersey: Prentice
Hall.
Giusti, Giuliana
2001 The Birth of a Functional Category: From Latin ILLE to the Ro-
mance Article and Personal Pronoun. In Current Studies in Italian
Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque and
Giampaolo Salvi (eds.), 157-171. Amsterdam/London/New York/
Oxford / Paris / Shannon / Tokyo: Else vier.
Guardiano, Cristina and Giuseppe Longobardi
2003 Parametric Syntax as a Source of Historical-Comparative Generali-
zation. Unpublished Ms., Universities of Pisa and Trieste and CNRS
Paris.
Hale, Mark
1987 Notes on Wackernagel's Law in the Language of Rigveda. In Studies
in Memory of Warren Cowgill (1929-1985), Calvert Watkins (ed.),
38-50. Berlin/New York: Mouton de Gruyter.
1988 Old Persian Word Order. Indo-lranian Journal 31: 27-40.
Hoffmann, Inge
1984 Der Erlaß Telipinus. Heidelberg: Carl Winter Universitätsverlag.
Humbach, Helmut
1991 The Gäthäs of Zarathushtra parts 1-2. Heidelberg: Carl Winter Uni-
versitätsverlag.
Kayne, Richard S.
1994 The Antisymmetry of Syntax. Cambridge, Mass.: The MIT Press.
 A particular coordination structure of Indo-European flavour 517

Kellens, Jean and Eric Pirart

1989 Les textes vieil-avestiques vol. I. Introduction, texte et traduction.
Wiesbaden: Reichert.
1990 Les textes vieil-avestiques vol. II. Repertoires grammaticaux et
lexique. Wiesbaden: Reichert.
Kent, Roland
1953 Old Persian Grammar, texts, lexicon 2nd ed. New Haven: American
Oriental Society.
Klein, Jared S.
1985 Toward a Discourse Grammar of the Rigveda, vol. I: Coordinate
Conjunction. Heidelberg: Carl Winter Universitätsverlag.
Klingner, Friedrich (ed.)
1970 Q. Horati Flacci opera. Leipzig: B. G. Teubner Verlagsgesellschaft.
[= 1959 3 ].
Lanzetta, Emanuele
2003 La lingua degli Achemenidi. Analisi sintattica e analisi culturale.
Unpublished doctoral dissertation, University of Palermo.
Lecoq, Pierre
1997 Les inscriptions de la Perse achemenide traduit du vieux perse, de
l'elamite, du babylonien et de l'aramien. Paris: Gallimard.
Lightfoot, David
1999 The Development of Language: Acquisition, Change and Evolution.
Cambridge, Mass./Oxford, UK: Blackwell.
Longobardi, Giuseppe
2001 Formal Syntax, Diachronic Minimalism and Etymology: the History
of French chez. Linguistic Inquiry 32: 275-302.
Matthews, Peter
1993 Central Concepts of Syntax. In Syntax. An International Handbook of
Contemporary Research, Vol. 1, Joachim Jacobs, Arnim von Stechow,
Wolfgang Sternefeld and Theo Vennemann (eds.), 89-117. Berlin/
New York: Mouton de Gruyter.
Mazzarino, Antonio (ed.)
1982 M. Porci Catonis De Agri Cultura. Leipzig: B. G. Teubner Verlags-
gesellschaft.
Melazzo, Lucio
1997 II cosiddetto ca inverso nel Rigveda. In Scribthair a ainm n-ogaim.
Scritti in memoria di Enrico Campanile, Riccardo Ambrosiani, Maria
Patrizia Bologna, Filippo Motta and Chatia Orlandi (eds.), 635-645.
Pisa: Pacini.
Monro, David Binning and Thomas W. Allen (eds)
1908-20 3 Homeri Opera. Tomus II Iliadis XIII-XXIV. Oxford: Clarendon
Press.
518 Emanuele Lanzetta and Lucio Melazzo

Munn, Alan
1992 A Null Operator Analysis of ATB Gaps. The Linguistic Review 9:
1-26.
Rizzi, Luigi
1997 The Fine Structure of the Left Periphery. In Elements of Grammar.
A Handbook of Generative Syntax, Liliane Haegeman (ed.), 281-337.
Dordrecht: Kluwer.
Roberts, Ian
1993 Verbs and Diachronic Syntax: a Comparative History of English and
French. Dordrecht: Kluwer.
Ross, John R.
1967 Constraints on Variables in Syntax. Cambridge, Mass.: M I T disser-
tation.
Schmitt, Rüdiger
1963 Ein altpersisches 'ghostword' und das sog. 'inverse ca'. Orientalia
32: 4 3 7 - 4 4 8 .
van Kemenade, Ans and Nigel Vincent
1997 Parameters of Morphosyntactic Change. Cambridge: Cambridge
University Press.
Wackernagel, Jacob
1892 Über ein Gesetz der indogermanischen Wortstellung. Indoger-
manische Forschungen 1: 333^136.
[1969 2 ] [Reprinted in Wackernagel, Jacob, Kleine Schriften. 2 Bände. Göt-
tingen: Vandenhoeck & Ruprecht, 1969 2 , 1. Band: 1-104],
Watkins, Calvert
1994 Indo-European *-k w e 'and' in Hittite. In C. Watkins: Selected Writ-
ings, Lisi Oliver (ed.), 300-306. Innsbruck: Wolfgang Meid.
Zwolanek, Renee
1970 'väyav indrasca'. Studien zu Aurufungsformen im Vedischen, Awes-
tischen und Griechischen. Münchener Studien zur Sprachwissen-
schaft 5: 51-64.
Index

accusativus cum infinitivo, accusative, bridging contexts, 265, 270, 271, 272
with infinitive, Acl, 4, 7, 8, 301, causative, causativisation, 193, 194,
342-367 195, 196, 198, 199, 201,214, 222,
active-stative language, 22, 24,279-284 223, 224, 225, 226, 227, 228, 231,
adjunction, 61, 81, 82, 92, 130, 153-154, 232,287
398,420,499-503, 515 causee, 194
agreement, 16, 17, 18,31-33,44-65,69, cislocative, 167, 183
81,87, 88,90,91,92, 118-122, 147, Classical Greek (see also Ancient
156, 239-241, 244, 245, 248-261, Greek), 5, 6, 7, 11, 265-293, 295-
278, 292, 296, 322, 323, 327, 329, 337
332,436 cleft, 15, 16, 105-133, 185
Akkadian, 2, 3, 5, 6, 20, 21, 22, 161, clitic, enclitic, 4, 5, 14, 16, 17, 20, 30,
189-236, 512 46, 54, 58, 68, 70, 72, 73, 74, 75, 87,
Aktionsart, 33, 90, 270, 277, 278, 279, 91,93, 120, 129, 132, 163, 169, 171,
281,282, 287, 288 172, 173, 177, 180, 183, 184, 185,
Ancient Greek (see also Classical 229, 256, 257, 260, 353, 386, 389,
Greek), 4, 8, 19, 24, 28, 29, 3 0 , 2 6 5 - 390, 400, 401, 404, 406, 409, 413,
293, 295-337, 342, 347, 353, 354, 430, 431, 432-434, 437, 483, 491,
365,457, 498,499, 511, 513 496, 497, 498, 499, 500, 506, 507,
anticipatory genitive, 9, 161, 162, 168, 509, 510, 511, 512, 515
179, 180, 182, 183 complementizer, 7, 8, 15, 16, 17, 47, 55,
antisymmetry of syntax, 349,424, 496 56, 57, 58, 59, 75, 76, 77, 78, 80, 88,
apophony, 225, 226, 227, 270, 277 89, 92, 126, 127, 243, 259, 304, 309,
Arabic, 22, 44, 45, 92, 105, 111, 114, 317,319,320, 321,331,347,352,
128, 129, 189, 191,205,214,221, 357, 359, 360, 364, 367, 382, 409,
228, 230, 231, 233, 234, 239, 241, 413, 443, 446, 448, 449, 450, 451,
246, 248, 250, 252, 259, 260, 261 452, 501
article, 4, 25, 140, 141, 143, 155, 353, concord, 139, 147, 155, 272, 282, 343
366, 367, 421 converter, 154
definite, 10, 115, 116, 141, 142, 143, circumstantial, 141, 155
145, 147, 148, 149, 150, 151,300, relative, 141
307,353,367 Coptic, Coptic Egyptian, 6, 15,29, 105-
indefinite, 10, 141, 143, 147, 150, 134, 139-159
151, 155 copula, 5, 16, 26, 116, 117, 118, 119,
possessive, 141, 142, 143, 150, 151 120, 121, 122, 127, 129, 130, 132,
asyndetic coordination, 496, 506, 507, 177, 179, 184, 185,422, 461,464,
508, 510, 514, 515 466, 467, 473
520 Index

corpus, Classical Sanskrit, 486 extraction, 55, 57, 63, 64, 91, 463, 471,
defmiteness, 22, 92, 110, 129, 141, 143, 472, 473,476, 477, 478, 479, 480,
147, 148, 151, 154, 241,421 481,484, 485,486,488, 489
demonstrative, 10, 89,115,116,129,141, focus movement, focussing,
142, 145, 150, 154,353,367,513 focalization, 12, 15, 105, 113, 123,
determiner, 9, 10, 11, 115, 140, 141, 125, 127, 269, 300, 392-397, 420
143, 146, 147, 148, 150, 156,356, focus, focussed, 12, 14, 15, 16, 17, 27,
357,367,461,511 32, 54, 68, 69, 72, 74, 84, 85, 87,
determiner phrase, 147 105-134, 185, 210, 351, 377, 382,
diachrony, 284, 406, 408 383, 389, 391, 392-397, 403, 406,
discontinuity, 405, 412, 413 410, 413,416, 417, 420, 424, 427,
distributional context, 268, 271, 272 433, 435,440, 441, 442, 444, 447,
double accusative, 342, 344, 365 448, 451,452
D-stem, 20,21, 189-236 contrastive, 13, 50,67, 107, 108, 109,
dual subcategorization, 22, 265, 267, 111,393,397,409,416,417
280, 281,284 formal features,
embedded clause, 15, 76,77,78, 80, 126, interpretable, 65, 81, 139, 151, 212,
243, 248, 259, 298, 317, 323, 359, 219, 220,265,280, 282, 283, 323
429,446, 448, 449, 451, 452,453 uninterpretable, 16, 51, 65, 81, 84,
enclitic conjunction, 496,497,498,499, 139, 146, 147, 148, 150,265,280,
500, 506, 5 0 9 , 5 1 0 , 5 1 1 , 5 1 2 , 5 1 5 284,287
event, (sub-)event, 6, 13, 16, 21,28, 31, frame, 14, 35, 360, 376, 429, 430, 447,
32, 34, 36, 37, 52, 54, 59, 65, 87, 148, 448,449,451,453,454
177, 191, 198, 200,208,209,212, French, Old, 420, 430, 432, 434
215, 218, 219, 220, 226, 227, 231, gemination, geminate, 190, 215, 223,
232, 236, 2 9 8 , 3 1 1 , 3 1 8 , 3 3 1 , 3 8 2 , 224, 225,226,235
403 gender, 15, 16, 17, 44, 45, 47, 54, 55,
eventive, 17, 18, 32, 33, 34, 43, 53, 71, 57, 87, 89, 91, 92, 118, 119, 120,
75, 77, 88 121, 122, 139, 140, 141, 146, 147,
eventual plural, 191,214,219,220,223, 148, 150, 161, 162, 167, 168, 169,
224, 225, 226 173, 177, 183,229, 278,350,353,
Exceptional Case Marking, ECM, 4, 7, 361,461,462
17, 46, 47, 70, 71, 82, 83, 303, 305, genitive case, 9, 10, 148, 149, 153,
306, 307, 309, 317, 332, 333, 346, 163, 169, 181, 182, 183, 266, 267,
347, 359 268, 272, 273, 275, 276, 282, 286
exclamatory infinitive, 358 G-stem, 21, 22, 189-234
Extended Projection Principle, EPP, Hebrew, 5, 6, 16, 18, 22, 25, 26, 44,
16, 18, 31, 64, 65, 69, 71, 72, 74, 76, 49, 54, 62, 65, 85, 92, 94, 95, 96,
81, 84, 86, 87, 88,239,240,251,252, 102, 129, 132, 143, 148, 153, 158,
253, 258, 259, 260, 335 189, 191, 205, 214, 221, 228, 230,
external argument, 16, 21, 60, 196, 197, 231,232,234, 236, 239-264
198,200, 201, 208, 209, 214, 225, historical infinitive, 358
229, 232, 233, 279, 281, 306, 356
 Index 521

Hittite, 5, 19, 342, 496, 498, 505, 506, NumP, 11,21, 138, 146, 147, 149, 150,
507, 5 0 8 , 5 1 0 , 5 1 1 , 5 1 2 , 5 1 5 , 5 1 8 151, 153
Homeric Greek, 265, 277,279, 282, 367 object shift, 61, 66, 67, 68, 70, 73, 82
incorporation, 44, 46, 47, 55, 56, 58, Old Avestan, 505, 512, 515
90,91, 179,282,317,318,365 Old Persian, 505, 508, 511, 514, 516,
information 517
packaging, 459, 460, 468 Older Egyptian, 6, 16, 1 7 , 2 8 , 2 9 , 3 1 -
structure, 17, 29, 69, 72, 76, 84, 107, 101
161, 162,468 Osco-Umbrian, 342
intensive stems, 191 overlap, 269, 270
inverse ca, 496,499, 500, 501, 502, 503, passive, passivization, 4, 54, 57, 77, 78,
505,511,512,514,518 79, 80, 89, 90, 93, 189, 206, 208, 231,
Iranian, 342, 510 233, 297, 305, 306, 313, 331, 364,
iterative stems, 190, 213, 222, 232, 234 462
Latin, 4, 6 , 7 , 8, 11, 12, 13, 14, 15, 19, phrase structure, 28, 134, 140, 158, 382,
22, 24, 25, 26, 27, 28, 29, 30, 278, 457, 458, 460, 461, 463, 465, 469,
287,290,316,317,319,331,333, 473,486,487,490, 491,492
334, 335, 339-372, 373^126, 4 2 9 - Classical Sanskrit, 460
456, 457, 458, 496, 505, 506, 507, pied piping, 502
508,512,516 polysemy, polysemous, 49, 265, 267,
left dislocation, 387 269, 270, 286
left periphery, 9, 13, 15, 29, 88, 107, possession
108, 124, 128, 171,336,346,347, external, 9, 10, 161, 162, 168, 172,
350, 426, 429, 436, 445, 446, 454, 179, 180, 181, 183
481, 501, 512 inalienable, 9, 10, 179, 181
lexical root, 65, 285 internal, 181
lexicalization, 269, 270, 286 predication of, 152
LF lowering, 503 possession, possessed noun, 9, 10, 11,
LF raising, 503 32, 37, 53, 138, 141, 144, 146, 148,
light verb, 60, 278 149, 150, 151, 152, 153, 154, 156,
Linear Correspondence Axiom, 411,496 161, 172, 175, 177, 178, 180, 185,
nominative-accusative language, 24, 219
282,284,288, 365 possessive
nominativus cum infinitivo, construction, 9, 10, 145, 159, 179
nominative with infinitive, Ncl, expression, 142
301,302,362 morpheme, 149
null expletive, 50, 54, 92, 240, 241 possessor, 9, 10, 11, 37, 138, 141, 142,
number, 17, 44, 45, 49, 54, 55, 57, 87, 148, 149, 150, 151, 153, 154, 155,
91, 112, 118, 119, 120, 121, 122, 139, 156, 161, 162, 163, 168, 169, 171,
140, 141, 146, 147, 148, 150, 162, 172, 173, 175, 179, 180, 181, 184,
167, 168, 169, 173, 177, 178,210, 185
212, 214, 230, 253, 278, 461, 462, possessor, pronominal, 142, 144, 145,
488,514 151, 156
522 Index

PossP, 10, 11, 149, 150, 151 root, 8, 21, 22, 23, 32, 37, 43, 51, 59, 60,
predication, 89, 90, 109, 122, 123, 152, 65,69,73,78, 81, 82, 87,90, 91, 189,
221,414, 467 190, 202, 206, 224, 225, 229, 230,
pro-drop, 7, 46, 47, 48, 49, 50, 83, 88, 232, 235, 262, 265, 266, 267, 270,
260,462 271, 272, 277, 278, 284, 287, 358,
pronominal object shift, 66, 68, 82 359,360
pronoun, 8,9, 14, 15, 25,44,46,47, 50, Indo-European, 267
54, 55, 56, 58, 61, 67, 68, 70, 71, 75, Sanskrit, 5, 6, 11, 12, 26, 277, 292,457,
8 2 , 9 1 , 9 2 , 9 3 , 106, 107, 109, 110, 458, 459, 460, 461, 462, 463, 464,
111, 112, 114, 115, 116, 118, 120, 465, 467, 468, 469, 471, 472, 475,
121, 122, 129, 131, 143, 154, 155, 476, 477, 480, 483, 484, 485, 486,
156, 167, 171, 183,229,285,346, 487, 488, 489, 490, 491, 492, 493,
353, 357, 360, 361, 362, 364, 365, 501,510
367, 393, 430, 431, 432, 437, 451, Classical, 11,12, 457,458, 459,460,
478,495, 501, 513 461, 462, 463,465,467,469,471,
atonic, 432 472, 474, 475, 476, 477, 480, 484,
clitic, 4, 14, 93, 430 485,486,487,488,491
resumptive, 9, 15, 29, 56, 57, 58, 72, Vedic, 19, 29, 342, 461, 470, 489,
113, 122, 126, 127, 131, 155, 171, 490, 492, 499, 501, 503, 510, 515
430, 4 3 1 , 4 3 2 , 4 3 5 scrambling (see also word order, free),
stressed, 106, 454 398, 458,493
strong, 15, 433 Semitic, 6, 20, 26, 44, 94, 96, 102, 161,
weak, 14, 15, 29, 93, 419, 432, 433, 189, 1 9 2 , 2 0 5 , 2 1 8 , 2 2 7 , 2 2 8 , 2 2 9 ,
434, 436, 437, 438, 439, 440, 441, 231,233,235,236,239,256
451,452,453 situation aspect, 33
Proto-Indo-European, 265, 268, 277, small clause, 15, 16, 90, 107, 108, 122,
279, 286, 344, 357, 370 123, 124, 128, 260, 336, 356, 398
psych-verbs, 203, 204 S-stem, 189, 190, 191, 192, 193, 194,
quantifier, 70, 85, 256, 264, 438 195, 196, 198, 199, 200, 201, 223,
Quechua, 364 225, 226, 227, 228, 229, 231, 232,
relative clause, 9, 15, 16, 26, 27, 52, 55, 233,235
57,58, 105, 106, 107, 109, 111, 112, state, stative, 1, 16, 17, 22, 23, 32-52,
113, 114, 115, 116, 117, 118, 121, 75, 80-84, 88, 144, 153, 189, 203,
122, 123, 126, 127, 128, 129, 130, 204, 208, 209, 210, 215, 216, 217,
131, 134, 141, 152, 155, 169, 171, 219, 220, 221, 225, 231, 233, 268,
357, 364, 398, 420, 453,478 269, 270, 272, 273, 279, 281, 285,
Romance, 4, 13, 14, 25, 29, 102, 157, 286, 287
239,252, 259, 342, 353, 367, 369, structural case, 6, 23, 47, 144, 152, 273,
371, 372, 374, 409,423,426,431, 274, 283, 288, 304
432,433,436, 454, 456,516 subject-to-object raising, 345
Early, 371, 444 subordinate clause, 58,76,345,347,364,
Medieval, 13, 14, 429,430,433, 434 365,400, 401, 406,447,464, 478
 Index 523

Sumerian, 2, 3, 5, 6, 9, 10, 26,27, 30, universal constraints, 401

161, 162, 163, 165, 166, 167, 168, ν (little v), 196, 197, 198, 211, 212, 214,
170, 172, 175, 177, 179, 181, 182, 232, 234, 278, 280, 282, 287, 332
183, 184, 185, 186, 187, 188 valence, valency, 35, 59, 190, 194,
cases, 165, 166, 183 195, 196, 197, 199,204,205,225,
nominal template, 163 227, 228, 230, 266
noun phrase, 163, 165 Vedas, 461
verbal template, 167 verb phrase, 1 6 , 2 1 , 6 4 , 2 1 2
SVO, 13, 16, 17, 18, 22, 31, 32, 33, 59, verb second, 76, 400, 419
60,61,62,81,82, 88,91,92,242, verbal plural, 5 1 , 1 9 1 , 2 0 5 , 2 1 1 , 2 1 2 ,
243, 252, 258, 382, 390, 396,416, 213, 214, 218, 219, 222, 223, 225,
420, 422 227, 228, 230, 232, 233
synchrony, 383, 408, 409 vocative, 116, 117, 462, 482, 487
T(ense), 8, 16, 18, 64,65, 69, 71, 72, 75, Vedic Sanskrit, 489
76, 82, 100, 168, 273, 277, 283, 288, VP-internal subject, 92
297, 309 VSO, 16, 17, 18, 25, 31, 32, 53, 54, 59,
Tagalog, 469, 473, 492 60, 61, 62, 64, 65, 66, 68, 69, 71, 73,
template, 59, 91, 230, 232, 235, 380, 77, 87, 88, 92, 9 3 , 9 5 , 9 6 , 102,239,
381,387 240, 241, 246, 248, 251, 252, 253,
template morphology, templatic 260,264,378,379,401,403
morphology, 235 v- VP configuration, 191, 196, 197, 198,
topic, 10, 13, 14, 17, 32, 68, 69, 84, 87, 201,208,211,214,223,234
109, 124, 125, 168, 170, 171, 180, Warlpiri, 2 6 , 4 6 3 , 4 7 3 , 4 9 1 , 4 9 2
181,389,392,404,414,415,416, wh-word/phrase, 14, 113, 357, 450
420, 433, 436, 446, 447, 448, 449, wild tree, 11, 1 2 , 4 5 7 , 4 5 8 , 4 5 9 , 4 6 0 ,
451,453, 501 465, 467, 468, 469, 471, 472, 473,
new, 14, 110, 170, 171,433 474, 475, 476, 480, 481, 485, 489
topicalization, 12, 286, 501, 513, 515 word order
transition, 21, 22, 215, 216, 217, 218, basic, 13, 375, 380, 464
219, 220,221,223 Classical Sanskrit (see also
transitive verbs, 205, 208, 212, 213, discontinuity, Classical Sanskrit),
214,222,227, 231 462,471,486
transitivity alternation, 191, 192, 202 free, 11, 408, 457, 458, 460, 489
tree, wild - see wild tree Latin, 12, 13, 22, 28, 373, 376, 380,
typological change, 272, 284 401, 410,422,425, 429,444,450,
typology, 23, 25, 31, 33, 90, 279, 283, 458
288, 397,401, 402, 407, 413, 422, marked, 67, 420, 434
423,492 unmarked, 13, 17, 373, 400, 441,
unaccusative verbs, 21, 32, 35, 41, 43, 506
87,214, 228,231 variation, 4, 6, 12, 16, 138, 377, 388,
unergative verbs, 21,38, 39, 191, 195, 457, 4 6 9 , 4 8 5 , 4 8 7
199, 220 φ-features, 81, 278
List of contributors

Annamaria Bartolotta Katalin E. Kiss

Dipartimento di Scienze Filologiche Institute for Linguistics -
e Linguistiche - Universitä degli Hungarian Academy of Sciences
Studi di Palermo, Italy Budapest, Hungary
bartolotta.marina@lettere.unipa.it ekiss@nvtud.hu

Edit Doron Emanueie Lanzetta

Hebrew University of Jerusalem, Dipartimento di Scienze Filologiche
Israel e Linguistiche - Universitä degli
edit@vms.huii.ac.il Studi di Palermo, Italy
mane 1 inotheta@yahoo. it
Barbara Egedi
Department of Egyptology - Aniko Liptäk
Eötvös Loränd University Budapest, Center for Linguistics - University
Hungary of Leiden, The Netherlands
egedi@hotmail.com a.liptak@let.leidenuniv.nl

Brendan S. Gillon Lucio Melazzo

Department of Linguistics - Dipartimento di Scienze Filologiche
McGill University Montreal, Canada e Linguistiche - Universitä degli
brendan.gillon@mcgill.ca Studi di Palermo, Italy
melazzo.lucio@lettere.unipa.it
Christian Huber
Phonogrammarchiv - Chiara Polo
Austrian Academy of Sciences Department of Linguistics -
Wien, Austria University of Padua, Italy
christian.huber@oeaw.ac.at chiara.polo@unipd.it

Lisa Lai-Shen Cheng Chris H. Reintges

Center for Linguistics - University Center for Linguistics - University
of Leiden, The Netherlands of Leiden, The Netherlands
ll.cheng@let.leidenuniv.nl c.h.reintges@let.leidenuniv.nl
526 List of contributors

Giampaolo Salvi Vassilios Spyropoulos

Department of Italian Language and Department of Mediterranean
Literature, Eötvös Lorand University Studies, University of the Aegean
Budapest, Hungary Rhodes, Greece
gps@ludens.elte.hu spiropoulos@rhodes.aegean.gr

Benjamin Shaer Gabor Zolyomi

Zentrum für Allgemeine Department of Assyrology and
Sprachwissenschaft Hebrew Studies - Eötvös Lorand
Berlin, Germany University Budapest, Hungary
shaer@zas.gwz-berlin.de gzolyomi@ludens.elte.hu