CASE IN STANDARD ARABIC: A DEPENDENT CASE APPROACH

AMER AHMED

A DISSERTATION SUBMITTED TO
THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

GRADUATE PROGRAM IN LINGUISTICS AND APPLIED LINGUISTICS

YORK UNIVERSITY
TORONTO, ONTARIO

August 2016

© AMER AHMED, 2016

ABSTRACT

This dissertation is concerned with how structural and non-structural cases are assigned

in the variety of Arabic known in the literature as Standard Arabic (SA). Taking a Minimalist

perspective, this dissertation shows that the available generative accounts of case in SA are

problematic either theoretically or empirically. It is argued that these problems can be overcome

using the hybrid dependent case theory of Baker (2015).

This theory makes a distinction between two types of phases. The first is the hard phase,

which disallows the materials inside from being accessed by higher phases. The second is the

soft phase, which allows the materials inside it to be accessed by higher phases.

The results of this dissertation indicate that in SA (a) the CP is a hard phase in that noun

phrases inside this phase are inaccessible to higher phases for the purpose of case assignment. In

contrast, vP is argued to be a soft phase in that the noun phrases inside this phase are still

accessible to higher phases for the purposes of case assignment (b) the DP, and the PP are also

argued to be hard phases in SA, (c) case assignment in SA follows a hierarchy such that lexical

case applies before the dependent case, the dependent case applies before the Agree-based case

assignment, the Agree-based case assignment applies before the unmarked/default case

assignment, (d) case assignment in SA is determined by a parameter, which allows the dependent

case assignment to apply to a noun phrase if it is c-commanded by another noun phrase in the

same Spell-Out domain (TP or VP), (e) the rules of dependent case assignment require that the

NPs involved have distinct referential indices.

The major conclusion of the dissertation is that the functional head v in SA is a soft phase

head, due to its deficient ɸ-specification. That is why it is incapable of establishing an Agree

ii
relation with the object and assigning the structural accusative case to it. The structural

accusative case on the object is, therefore, always the result of the dependent case mechanism.

iii
ACKNOWLEDGEMENTS

There are a number of people who contributed in one way or another to this thesis. First, I would
like to thank my thesis supervisor, Dr. Gabriela Alboiu of York University for reading previous
versions of this manuscript and providing feedback. Her contribution to the ideas expressed in
this thesis is much appreciated.
Second, I would like to thank Dr. Ruth King of York University for her input on a
previous version of this manuscript. I would also like to thank Dr. Philipp Angermeyer and Dr.
Amila Buturovic of York University for their feedback on an earlier version of this thesis.
Third, a very special thank you should go to Dr. Arsalan Kahnemuyipour of the
University of Toronto for his meticulous comments on earlier versions of this thesis. His
insightful comments have greatly helped improve the quality of this thesis.
Fourth, a very big thank you is due to Dr. Youssef Haddad of the University of Florida
for agreeing to be the external examiner of my PhD thesis committee and for his insightful
comments on an earlier version of this thesis.
Fifth, Dr. Usama Soltan of Middlebury College was always there when I had questions
about his PhD work on Standard Arabic syntax. Never once did he fail to respond to my
questions. He is by all standards a role model for all scholars in the field of academia.
Sixth, Dr. Dinha Tobiya Gorgis of the Defence Language Institue has always been there
fore me both when I needed answers to some of my questions about Standard Arabic and when I
needed emotional support. He was the first person who started me thinking about linguistic
issues. He is truly a Godfather, a dedicated defender of science, an excellent educator and teacher
and a great moral figure in my life.
Seventh, Dr. Rashid Al-Balushi of Sultan Qabuus University kindly responded to my
questions about his PhD work on the syntax of case in Standard Arabic. Another thank you for
Dr. Fassi Fehri of Mohammed V University for kindly responding to one of my e-mails.
Eighth, a word of thanks should also go to Dr. Elizabeth Ritter of the University of
Calgary, Dr. Elizabeth Cowper and Dr. Diane Massam of the University of Toronto and Dr.
Martha McGinnes of the University of Victoria, Matthew A. Tucker of NYU Abu Dhabi
University, and Dr. Tue Trinh of the University of Wisconsin Milwaukee for their valuable input
on some of the ideas expressed in this thesis, which were presented in a number of linguistic
conferences both in Canada and in the States.
I would like to end with a very big thank you to Noam Chomsky. It was through some of
his words about science that I started seriously entertaining the idea of doing a PhD in theoretical
Linguistics. His words, “Science is tentative, explorative, questioning, and largely learned by
doing.” (Chomsky Rationality/science: 91), have an everlasting impact on me.

iv
TABLE OF CONTENTS
Abstract ........................................................................................................................................... ii
Acknowledgment .......................................................................................................................... iv
Table of contents ..............................................................................................................................v
List of Tables ............................................................................................................................... viii
Abbreviations used in glosses ....................................................................................................... ix
Chapter 1: Introduction ...................................................................................................................1

Chapter 2: Case in generative grammar…………………………………………………………...6

2.1 Introduction…………………………………………………………………………………… 6
2.2 On binding (Chomsky 1980)………………………………………………………………... 6
2.3 The Government and Binding (GB) approach to case……………………………………… 10
2.4 Case in Early Minimalism…………………………………………………………………... 14
2.5 Case in later Minimalism……………………………………………………………………. 16
2.6 Refinements to case theory…………………………………………………………………. 19
2.6.1 On two types of nonstructural cases (Woolford 2006)………………………………….. 19
2.6.2 On default case (Schütze 2001)…………………………………………………………. 24
2.7 Summary……………………………………………………………………………………..27

Chapter 3: Introduction to SA and previous accounts of case…………………………………...28

3.1 Introduction…………………………………………………………………………………..28
3.2 The syntax of SA: A brief overview…………………………………………………………28
3.3 Case in Medieval Arabic grammar…………………………………………………………..39
3.4 Previous generative accounts of case in SA…………………………………………………47
3.4.1 Agree-based accounts of case in SA……………………………………………………….48
3.4.2 A mixed approach to case (Fassi Fehri 1993)……………………………………………...79
3.5 The proposed account………………………………………………………………………..86
3.6 Summary……………………………………………………………………………………..88

Chapter 4: Alternate accounts of case with a focus on Baker’s dependent case theory and
relavant theoretical assumptions…………………………………………………………………89
4.1 Introduction…………………………………………………………………………………..89
4.2 Alternative accounts of case: The Case Tier theory…………………………………………89
4.2.1 Problems with the Case Tier theory………………………………………………………100
4.3 The proposed account of case for SA………………………………………………………105
4.3.1 The original version of dependent case (Marantz 1991)…………………………………105
4.3.1.1 Problems with the original version of dependent case theory………………………….114
4.3.2 The updated version of dependent case (Baker 2015)……………………………………116
4.3.3 Theoretical assumptions…………………………………………………………………..131
4.3.4 The rules of dependent case assignment………………………………………………….135
4.3.5 The preverbal position is a subject position………………………………………………138
4.3.6 Nominative is the unmarked/default case in SA………………………………………….145
4.4 Summary…………………………………………………………………………………....146

v
Chapter 5: Case assignment of core arguments in Standard Arabic…………………………… 147

5.1 Introduction………………………………………………………………………………… 147

5.2 vP is a soft phase in SA……………………………………………………………………..147
5.3 Case assignment in simple transitive sentences……………………………………………. 150
5.4 Case assignment in simple intransitive sentences………………………………………….. 151
5.5 Case assignment in sentences with double object constructions…………………………... 154
5.6 Case assignment in sentences introduced by the indicative (emphatic or neutral)
Complementizer……………………………………………………………………………….. 162
5.7 Case assignment in sentences with backgrounding encliticized topics……………………. 168
5.8 Case assignment in sentences with existential constructions……………………………… 169
5.9 Case assignment in sentences with believe-type predicates……………………………….. 176
5.10 Case assignment in sentences with control structures……………………………………. 185
5.11 Case assignment in sentences with adjectival/nominal predicates……………………….. 193
5.12 Case assignment in passive sentences……………………………………………………. 200
5.13 Case assignment in imperative sentences………………………………………………… 202
5.14 Case assignment in sentences with participials…………………………………………... 203
5.15 Case assignment in the PP domain……………………………………………………….. 211
5.16 Summary…………………………………………………………………………………..222

Chapter 6: Case assignment of non-arguments in SA………………………………………….224

6.1 Introduction………………………………………………………………………………... 224
6.2 The case assignment of adverbial NPs……………………………………………………. 224
6.3 Accusative case assignment of some other case-marked NPs……………………………...232
6.4 Case assignment of NPs in the left periphery of the clause……………………………….. 238
6.5 Summary…………………………………………………………………………………... 254

Chapter 7: Case assignment in the DP domain………………………………………………… 255

7.1 Introduction………………………………………………………………………………… 255
7.2 Case assignment in simple DPs……………………………………………………………. 255
7.3 Case assignment in derived nominals……………………………………………………… 265
7.4 Testing Baker’s dependent case theory against the case tier theory and the Agree-based
theory of case: Evidence from strcutures with event-denoting nominals………………..275
7.5 Case assignment in structures with pronominal possessives………………………………. 278
7.6 Summary…………………………………………………………………………………… 280
Chapter 8: Conclusions………………………………………………………………………… 282
8.1 Conclusions..…………….………………………………………………………………….282
8.2 Future research……………………………………………………………………………...288
References……………………………………………………………………………………… 289

vi
LIST OF TABLES

Table 1. The categories of case triggers in SA………………………………………….. 285

vii
ABBREVIATIONS USED IN GLOSSES

ABS Absolutive
ACC Accusative

DU Dual
COMP complementizer
CAUS Causative
CONJ conjunction
DAT dative
DEF Definite
EMPH Emphatic
ERG Ergative
F feminine
GEN genitive
IMP imperative
IMPF Imperfective
INDEF indefinite
INDIC Indicative
INF Infinitive
INSTR instrumental
JUSS Jussive
M masculine
NEG negative particle
NEUT Neutral
NOM nominative
PART Partitive
PASS Passive
PL plural
PRT particle
PTPL participle
Q question particle
SG singular
SUBJ subjunctive

viii
Case theory in Standard Arabic: A dependent case approach

Chapter One: Introduction

In this thesis, I address the issue of syntactic (structural and non-structural) as well as non-

syntactic (default and semantic) case in a variant usually known in the literature on Arabic as

Standard Arabic (SA).1 SA is a nominative-accusaive language, where the subject of transitives

and the subject of intransitives pattern together in contrast to the object. This alignment contrasts,

with other languages, ergative-absolutive languages, where the object of transitives and the

subject of intransitives pattern together in contrast to the subject of transitives. SA is the

language used exclusively as the medium of writing and orally in all forms of mass media in the

Arab World although it is mire coomon nowaydays to use a form of educated Spoken Arabic in

some programs (e.g. political shows or programs discussing social issues). Strictly speaking,

there are no native speakers of SA, as this is not the language that children in the Arab World

acquire from birth, but everyone is exposed to it from a very young age. However, Fassi Fehri

(1999: 21) argues that while SA is not a first language for the speakers of the modern variants of

Arabic, it is not a second language either, since the competence that speakers of the modern

variants of Arabic have of their particular variant of Arabic actually forms part of the

competence they develop later in SA. That is why, Fassi Fehri claims that SA is, for the speakers

of the modern varaints of Arabic, neither a first language nor a second language, but somewhere

in between. The main reason for choosing this dialect of Arabic instead of others is because this

variety of Arabic is the only one with morphological case. Throughout the thesis, I use my own

grammaticality judgments as an educated speaker of SA, as well as other sentences taken from

1
Througout the thesis, I use the lower case term case to cover both kinds of cases, syntactic and non-syntactic.
1
different written resources. A number of examples offered in the thesis are also taken from what

is sometimes known as Classical Arabic (CA) or Quranic Arabic, i.e. the Arabic variety that was

used between the 7th century and the 9th century, the time that spans the Ummayad and Abbasid

caliphates. This variety of Arabic was used by Medieval Arab tribes. These examples are also

documented in books that were written between the 8th century and the 13th century. I use these

examples because the syntactic structures they exhibit are still part of the syntax of SA.

According to Bateson (1967, as cited in Bin Muqbil 2006: 15), “M[odern]SA is a descendant of

CA and retains the basic syntactic, morphological, and phonological systems.” In terms of the

differences between the two, Bateson (1967, as cited in Bin Muqbil 2006: 15), states that “MSA

only uses a subset of the possible syntactic structures available in CA as well as a substantially

reduced lexicon.”

This thesis is written in the framework of the Minimalist Program (Chomsky 1995 and

subsequent work) as well as the dependent case theory of Baker (2015). The thesis addresses the

issue of case assignment in SA. In particular, the thesis attempts to answer the following

questions: (a) How are structural and non-structural cases assigned in this language? (b) What

theory can best account for the facts of case assignment in this language? (c) Do the available

Agree-based accounts of case actually explain the case facts of SA properly? Do accounts such

as that of Fassi Fehri (1993), which is grounded in the case tier theory of Yip, Maling and

Jackendoff (1987) fare better in explaining all the case facts of SA?

The major contribution of this thesis is to show that all the available accounts of case in

SA are problematic either theoretically or empirically. Specifically, the thesis shows the

following: (a) the available Agree-based accounts of case are problematic because they make the

wrong predictions when it comes to case assignment in DPs, where the head D is occupied by a

2
process nominal.2 In these strcutures, the theme object would be predicted to always bear the

structural accusative case, contrary to the facts, (b) Fassi Fehri (1993) offers a much better

account of case than the Agree-based account of case. However, his account can hardly be

incorporated into the case tier theory, despite his attempt to do so, (c) Fassi Fehri’s (1993) insight

can best be accommodated in the dependent case theory of Baker (2015), (d) contrary to the

claims of all the available literature on SA, v in SA is incapable of assigning the structural

accusative case because it is impoverished for its ɸ-features being specificed for a gender feature

only, (e) structural accusative case in SA is the result of the dependent case assignment rather

than the result of an Agree relation between v and the object, (f) case-assigning Agree does take

place but only when the dependent case fails to apply, (g) there is a hierarchy of case assignment

in SA such that the lexical case applies before everything else, the dependent case applies when

no lexical case does, the Agree-based case applies when the dependent case fails to apply, and

the unmarked/default case applies when nothing else applies, (g) case assignment in SA is

determined by one of the parameters proposed in Baker (2015). The parameter states that XP is

assigned the dependent accusative case if it is c-commanded by YP in any spell-out domain be it

TP or VP, (h) vP is a soft phase in the sense of Baker (2015: 149) such that “the contents of its

VP complement undergo Spell-Out (i.e. they may get their case features fixed) but they remain

active in the derivation”, (i) aside from vP, all other phases, namely CP, DP, and PP are shown to

be hard phases in the sense of Baker (2015) such that case assignment is determined inside these

phases alone, and the NPs inside these phases are inaccessible to subsequent syntactic

derivations.

2
Throughout the thesis, I use the terms NP and DP interchangeably to refer to the noun phrase. This choice is
motivated by the fact that there is inconsistency in the laiterature as to the lables, and using both would allow us to
maintain the original.
3
The thesis cosists of eight chapters. This chapter (i.e. chapter one) is an introduction.

Chapter two traces the development of case in Chomsky’s thinking from the early days of the

Government and Binding theory to the recent theoretical positions taken in what has come to be

known as the Minimalist Program. This chapter also covers two important contributions to the

generative understanding of case. The first such contribution is that of Woolford (2006), where

the argument is made that there are two types of non-structural cases. These are lexical case,

which is purely idiosyncratic case assigned by certain verbs and prepositions, and it is assigned

strictly inside VP, and inherent case, which is thematically linked, and it is assigned strictly

outside VP but inside vP. Another major contribution to the generative understanding of case is

that of Schütze (2001), who argues that UG (=universal grammar) must allow for a case, which is

provided when all other means of case assignment fail or are simply unavailable. Chapter three

provides an overview of SA. It also discusses the previous generative accounts of case in SA and

the problems that they face. Chapter four introduces alternative accounts of case, specifically the

case tier theory of Yip, Maling and jackendoff (1987), and the original version of the dependent

case theory as developed in Marantz (1991). The chapter discusses the problems that these two

accounts face. The chapter then introduces the updated version of the dependent case theory, as

developed in Baker (2015), which forms the theoretical framework adopted in this thesis. The

chapter also spells out other theoretical assumptions that are adopted in this thesis. Chapter five

is a full exposition of how the dependent case theory of Baker (2015) can account for the cases

of the core arguments in various structures of SA. Chapter six is an anaylsis of case assignment

in structures with non-arguments (i.e. adverbial NPs and NPs in the left-periphery of the clause).

Chapter seven shows how case assignment in the DP domain can be accounted for using Baker’s

dependent case theory. This chapter also demonstrates how the adopted theoretical framework

4
fares better than the Agree-based theory of case (Chomsky 2000, 2001) and the case tier theory

(Yip, Maling and Jackendoff 1987) in this respect. Chapter eight is a conclusion and discussion

with suggestions for further research.

5
Chapter Two

Case in generative grammar

2.1 Introduction

In this chapter, the history of case theory in the generative literature is discussed. In

section 2.2, the first formulation of case theory, which is developed in Chomsky (1980) is

discussed. In section 2.3, the Government and Binding approach to case developed in Chomsky

(1986) is introduced. In section 2.4, the early Minimalist approach to case developed in

Chomsky and Lasnik (1993) and Chomsky (1995) is discussed. In section 2.5, the later

Minimalist approach to case as developed in Chomsky (2000, 2001, 2005, 2007, 2008) is

introduced. Section 2.6 discusses other Minimalist approaches to case. In section 2.6.1,

Woolford’s (2006) account of two types of nonstructural case is introduced. In section 2.6.2

Schütze’s (2001) approach to default case is introduced.

2.2 On binding (Chomsky 1980)

The first formulation of case theory appears in Chomsky (1980). It is an attempt to revise

a former filter suggested in Chomsky and Lasnik (1977). The filter is stated in (1) below:

(1) “*[α NP to VP], unless α is adjacent to and in the domain of Verb or for ([-N])” (Chomsky

1980: 19).

The filter in (1) is meant to capture the ungrammaticality of (2a-b) and the condition in

the filter is meant to capture the grammatical status of (2c-e):

6
(2) a. *[John to come] is nice.

b. *It is nice [John to come].

c. For [John to come] is nice.

d. It is nice for [John to come].

e. I would like [John to come].

In a letter addressed to Chomsky and Lasnik, Vergnaud (1977) offers the formulation of

case in (3) and (4):

(3) “English has three cases: Subject Case, Genitive Case, “Governed Case” (“the case of

complements of verbs and of prepositions”).

(4) “The restrictions on subjects of infinitvals can follow from a general filter limiting the

distribution of NPs in the canonical Case.” (Vergnaud 1977, as cited in Lasnik 2008: 18).

Using Vergnaud’s insight, Chomsky (1980: 24) formulates his theory of case as follows:

“Suppose we think of Case as an abstract marking associated with certain constructions, a

property that rarely has phonetic effects in English but must be assigned to every lexical NP”. He

then offers the following general principles of Case Assignment:

(5) a. “NP is oblique when governed by P and certain marked verbs;

b. NP is objective when governed by V;

7
c. NP is nominative when governed by Tense.” (Chomsky 1980: 25). Given that the

above principles make reference to the structural relation of government, the latter is defined as

follows:

(6) “α is governed by β if α is c-commanded by β and no major category or major category

boundary appears between α and β” (Chomsky 1980: 25).

Given that all lexical NPs must have case, Chomsky offers the following filter:

(7) “*N, where N has no Case” (Chomsky 1980: 25).

Given the principles in (5) and the filter in (7), Chomsky can now account for the

examples in (2). The examples in (2a-b) are ill-formed because the lexical NP, John does not get

case. In (2a), the NP does not get case because it is the subject of an infinitival clause, and the

latter does not assign case because it is Tenseless. In (2b), the NP does not get case in its

infinitval clause, nor does it get case from the adjectival predicate, given that adjectives are not

case assigners. In contrast, the sentences in (2c-e) are well-formed. In (2c-d), the NP gets

objective case by (5a); in (2e), the NP gets objective case by (5b).

The account can also easily handle cases of control, as shown in (8)

(8) a. John wants [PRO to work hard].

b. John tries [PRO to work hard].

The sentences in (8) are well-formed because the NP, PRO is not subject to the Case Filter in (7).

In other words, PRO is not assigned case because it is a null rather than a lexical NP.

8
Chomsky (1980: 24) states that case assignment is clause-bound. Given this, he handles

cases of non-control verbs like believe as being marked in the sense of having a [+F] feature,

which enables them to assign objective case across a clausal boundary. This way, the examples

in (9) are accounted for in this approach to case:

(9) a. I believe [John to be a fool].

b. John is believed [t to be a fool].

c. who do you believe [t to be a fool].

The NP, John in (9a) is assigned objective case from the verb believe across a clausal boundary

because the verb is marked for a [+F] feature. In (9b), the NP is not assigned case in the

infinitival position and cannot get objective case from the verb, believe, given that the verb is

passivized, hence loses its ability to assign case. Therefore, the NP raises to receive nominative

case from the finite/Tensed clause. In (9c), the NP, who is assigned objective case in the

infinitival position from the verb, believe before it raises to the finite clause for focus purposes.

As for why structures such as (10) are grammatical, Chomsky claims that the

complementizer for may or may not be assigned the feature [+P(reposition)]. If it is assigned this

feature, it assigns an oblique case and is undeletable. If it is not assigned a [+P] feature it does

not assign case and is deletable. This is shown in (10):

(10) [S’[COMP for] [S NP to VP]].

a. For [John to do well on the exam] is nice.

9
b. I want [S’ COMP (for[+P]) [S John to do well on the exam].

In the example in (10a), the complementizer, for is assigned a [+P] feature; therefore, it assigns

an oblique case to the NP, John. In (10b), the complementizer is assigned a [+P] feature;

therefore, it assigns an oblique case to the NP, John. In this case, the complementizer is

undeletable. If the complementizer is not assigned a [+P] feature, it does not assign case, and is

deletable, and the NP receives objective case from the non-control verb, the latter has a [+F]

feature, which enables it to assign objective case across a clausal boundary.

2.3 The Government and Binding (GB) approach to case

In the Government and Binding (GB) approach, Chomsky (1981: 175-176, citing Freidin

and Lasnik (1979b) notes that not only lexical NPs seem to require case, but even variables. This

is illustrated in (11):

(11) a. *the man [S’ that [S you tried [S’ [S t to win]]]]

b. *the man [S’ that [S I wonder [S’ what [S t to see]]]] (Chomsky 1981, ex. 19: 176)

Based on examples such as those in (11), Chomsky argues that the Case Filter cannot be a PF

filter. For Chomsky, a PF case filter will not rule out the examples in (11) on the grounds that

the traces are not phonologically overt, and therefore, they do not need to acquire case. However,
10
the examples in (11) are ungrammatical, and the only reason to rule them out, according to

Chomsky, is because the variables (traces of relativization), t lack case.

At the same time, Chomsky notes that other structures show that the Case Filter cannot be

applying at LF. One such structure is the example in (12):

(12) *John tried [everyone to leave] (Chomsky 1981, ex. 20: 176)

Chomsky further notes that the LF representation of (12) is as in (13):

(13) John tried, for all x, x to leave (Chomsky 1981, ex. 21: 176).

Given that (12) does not get case in the surface structure, it should be saved if the Case Filter

applied at LF, as shown in (13). However, (12) is clearly not saved at LF, given that the sentence

is ungrammatical. It follows, Chomsky argues, that the Case Filter does not apply at LF.

To solve these problems, Chomsky (1981: 176) links case assignment with theta-role

assignment in the following manner: “Let us assume that elements of the form [α β] are

“invisible” to rules of the LF-component unless β contains some [case] feature. Thus, PRO is

visible as is Case-marked trace, but [NP e] is invisible when it contains no Case”. He then points

to the fact that each NP has a grammatical function, GF1 at the surface structure as well as a

function chain (GF1, …GFn), which represents the NP’s derivational history. He also uses the

11
distinction between A-positions (argument positions) and A’-positions (non-argument positions)

to formulate the condition in (14) on theta-role assignment:

(14) “Suppose that α has the A-function chain (GF1, …GFn). Then the chain is assigned a θ-role

only if α has [case] features” (Chomsky 1981, ex. 33: 179). To illustrate how the condition

works, Chomsky provides the examples in (15)

(15) Johni found a book [whichj he wanted [PROi to give tj to Mary]]]

He points out that in (15), which has case features, but it is in an A’-position. But, the trace tj is

an element of an A-function, and that it is this that enables the assignment of θ-role to the A-

function chain <which, tj> via the tail of the chain tj. The example in (15) contrasts with the

example in (16):

(16) *Johni found a book [whichj he wanted it to seem [tj to please Mary]]]]

The example in (16), Chomsky argues, is ungrammatical because the trace tj is an element of an

A-function; yet, it does not receive case because neither the tail tj nor the head which of the chain

<which, tj> is in an A-position, where they can receive case.

12
In Chomsky (1986a: 193), case theory is revised so that all lexical categories (V, N, A, P)

are case assigners. A distinction is then drawn between structural case, which is assigned by V

and INFL/AGR to an argument they govern and is assigned independently of theta-marking, and

inherent case, which is assigned by prepositions (oblique cases), nouns and adjectives (genitive

case) to their arguments only if the head also θ-mark their arguments, given that inherent case

applies at D-structure.

In this revised model, therefore, “a noun phrase can receive a θ-role only if it is in a

position to which case is assigned or is linked to such a position” (Chomsky 1986a: 94). The ‘or

clause’ in the quotation covers cases of expletive-argument pair, where the expletive is in a case

position and it transfers case to the associated argument, as in (17):

(17) There is a man in the room (Chomsky 1986a: 78).

In Chomsky (1986a: 135), the proposal is made that θ-roles and case are properties of

chains. Based on this, Chomsky (1986a: 135) reformulates the definitions for case and the theta

criterion, as in (18) and (19):

(18) “A CHAIN is Case-marked if it contains exactly one Case-marked position; a position in

a Case-marked CHAIN is visible for θ-marking.”

(19) “A CHAIN has at most one θ-position; a θ-position is visible in its maximal CHAIN.”

13
Given the formulation of case and the theta criterion in (18) and (19), Chomsky (1986a: 136)

points out that a chain must have exactly one θ-position and exactly one Case-position. Based on

this, Chomsky (1986a: 137) formulates the condition in (20):

(20) “ If C = (α1, …, αn) is a maximal CHAIN, then αn occupies the unique θ-position and α1 its

unique Case-marked position.”

2.4 Case in Early Minimalism

To solve the problem of the visibility of PRO for the LF component, Chomsky and

Lasnik (1993: 561) argue that PRO does have Case, which is null and which is licensed by

nonfinite INFL.

Another proposal of the early stages of Minimalism (Chomsky 1993, Chomsky and

Lasnik 1993, as cited in Chomsky 1995) is that all lexical items enter the derivation for syntactic

computation fully specified for their features (phonological, morpho-syntactic and semantic).

This is held to be the case for both lexical as well as functional categories. In this model, features

of lexical items are of two types, [+Interpretable] and [-Interpretable]. Syntactic derivation is

therefore driven by the need for lexical items with [-Interpretable] features to check their features

against lexical items with matching but [+Interpretable] features. For example, the categorial

feature N and the ɸ-features are [+Interpretable] on nouns, but [-Interpretable] on V and T.

Therefore, V and T have to probe for nouns in order to check their [-Interpretable] N and ɸ-

features. The syntactic relation that allows such probing is understood to be c-command.

14
Specifically, the probe must c-command the goal in order for checking to take place. In contrast,

nouns have a [-Interpretable] case feature that requires checking against a V and T with the

instructions [assign ACC] and [assign NOM] respectively. Note especially that unlike ɸ-features,

case is a special feature in that it is [-Interpretable] on both the probe and the goal. Thus, even

though V has the instruction [assign ACC] and T has the instruction [assign NOM], neither of

them has a positive value for case. It follows that in this model, case has a unique status. Thus,

Chomsky (1995: 278-279) describes this state of affairs by stating that “[c]ase differs from ɸ-

features in that it is always –Interpretable, for both terms of the checking relation. Case is

therefore the formal feature par excellence…” In this model, [-Interpretable] features have to

delete and become invisible for further interface LF operations once checked. They, however,

remain visible for the PF interface level since they have PF effects (Chomsky 1995: 279).

In sum, the early Minimalist model assumes that what enters the syntactic derivation are

full-fledged lexical items.The architecture of this model can be schematized as in (21) below:

(21)
Lexicon
↓

Syntax
↓
Spell-Out
3
PF LF

The architecture in (21) shows that lexical items come from the lexicon, which feeds the

syntactic derivation. The syntactic component diverges at the point of Spell-Out and provides
15
input to two interface levels, the PF (informally the phonological interface level) and LF

(informally the semantic interface level). Crucially, in this model, NPs enter the derivation fully

specified for their case features (nominative, accusative), and all that they need is to check these

cases against matching case features on functional heads, particulary T for nominative case and

v* for accusative case.

2.5 Case in later Minimalism

In Chomsky (2000), case is an uninterpretable feature on a goal (i.e. DP), which is

checked and valued only against a probe (v or T). The case feature of the goal will be given a

value depending on the probe, nominative if the probe is finite T, accusative if the probe is v, and

null if the probe is control infinitival (Chomsky 2000: 122-126). This is illustrated in the

following examples:

(22) [TP John T[+FINITE, uCase: NOM] [vP <John[uCase:__]> v[uCase: ACC] likes Mary[uCase:___]]].

(23) John tried to [CP PRO[uCase:___] T[-FINITE, uCase: NULL] to [vP <PRO> win the game]]

In (22), the subject DP John has an uninterpretable and unvalued case feature. This feature is

checked and valued against the uniniterpretable but valued nominative case feature of finite T.

Similarly, the object Mary has uninterpetable and unvalued case feature. This feature is checked

and valued against the uninterpretable but valued accusative case feature of the functional head

v*. In (23), the subject of the embedded CP is PRO, which has an uninterpteable case feature.

This feature is checked and valued against the uninterpretable but valued null case feature of the

non-finite T.

16
The uninterpretable case feature of the goal will only be checked and valued against a probe with

the full complement of ɸ-features. Thus, if the probe is v, v is defective lacking the feature

[person], as in participial and adjectival probes, and the goal will delete the ɸ-features of the

probe, but the probe will not be able to delete the uninterpretable case feature of the goal. In this

scenario, the goal remains active and is accessed by higher probes. The goal will have to move

towards the higher probe in order to check its uninterpretable case feature. Similarly, if the probe

is a defective T, as in control infinitivals, the probe is defective in that it has only a [person]

feature. The result of Agree is that the goal will delete the uninterpretable [person] feature of the

probe, but the probe will not be able to delete the uninterpretable case feature of the goal.

Therefore, the goal remains active and is accessed by higher probes. The goal will have to move

to the higher probe in order to have its uninterpretable case feature deleted. In the case of a weak

expletive, its ɸ-feature is uninterpretable. This renders this feature unnecessary for the expletive;

therefore, a case feature is also rendered unnecessary for the expletive (Chomsky 2000: 122-

126).

Thus, case in this model is an off-shoot of ɸ-feature checking and valuation. If a probe v

has the full-set of ɸ-features, it can agree with a goal (i.e. object DP) in ɸ-features, and the result

is that the goal has interpretable ɸ-features, which check and value the uninterpretable ɸ-features

of the head v. In return, the probe v can check and value the uninterpretable case feature of the

goal, object DP. All core functional categories in this model (C, T and v) may have ɸ-features.

These features are obligatory for v and T but not so for C (Chomsky 2000: 102). Chomsky

(2000: 121), posits that the uninterpretable features of functional heads are the sole driving force

of movement in natural language grammar. This is in contrast to the GB model, where case is

conceived as the driving force of movement.

17
In Chomsky (2001), the idea that v*P (i.e. a vP, which c-selects an external argument)

and CP are strong phases is introduced. Crucially, phasal v* assigns accusative case. These are

distinguished from TP and vP (i.e a vP which lacks an external argument as in passives and

unaccusatives, and which does not assign accusative case). , which are weak phases.

Chomsky (2005) makes the following proposals:

(a) The head of a lexical item is the only element capapble of acting as a probe, as it is the only

element visible for further syntactic computations (Chomsky 2005: 14).

(b) There are probably three phases, CP, vP with the full argument structure, and DP (Chomsky

2005: 17-18).

Chomsky (2007) makes the following claims:

(a) C and v* are phasal heads which bear inherent uninterpretable functional features which they

transfer to their labels (i.e. immediately selected heads). The label of C is T, and the label of v*

is V. The labels probe for a matching goal. For V, the matching goal is the uninterpretable case

feature of the object DP; for T, the matching goal is either the subject DP of v*P or the object of

V if V is passive or unaccusative. The uninterpretable features of the label receive the values of

the goal and the uninterpretable feature of the goal is assigned based on the properties of the

phase head, nominative if the label is T and the phasal head is C and accusative if the label is V

and the phasal head is v*. In case there are several goals, they all receive the same values in the

same manner the probe does (Chomsky 2007: 22).

(b) Unlike A-movement, A’-movement is successive-cyclic phase by phase (Chomsky 2007: 24-

25).

Chomsky (2008) makes the following claims:

18
(a) TP only has phase-like properties when it is selected by C (Chomsky 2008: 144).

(b) Only phasal heads trigger operations (such as IM), and even raising to Spec, TP is thus

possible because the latter inherits features from the phasal head, C (Chomsky 2008: 144).

2.6 Refinements to case theory

In addition to the Minimalist approaches to case developed in Chomsky and Lasnik

(1993), Chomsky (1995, 2000, 2001, 2005, 2007, 2008), the theory of case has been refined in

two other works. In the next two subsections, I discuss these two refinements.

2.6.1 On two types of nonstructural cases (Woolford 2006)

In Chomsky (1981, 1986), the only division is between structural cases (those that are

assigned in certain structural configurations) and nonstructural cases (those that are related to

certain thematic roles). Woolford (2006) argues that nonstructural cases are of two types, namely

lexical case and inherent case. For Woolford, lexical case is a purely idiosyncratic case that has

the two properties in (24):

(24)

(a) It is strictly assigned inside the VP domain by particular lexical heads such as certain Vs and

certain Ps.

(b) It is assigned only to internal arguments.

19
As examples of purely idiosyncratic cases on internal arguments that are grammatical subjects,

Woolford provides the following examples (Woolford 2006, exs. 7a-c: 114) from Icelandic:

(25) a. Bátnum hvolfdi.

boat-DAT capsized

‘The boat capsized

b. Bátnum rak á land

boat-ACC drifted to shore

‘The boat drifted to the shore.’

c. Jóns nýtur ekki lengur við.

John-GEN enjoys not longer at

‘John is no longer available.’

The examples in (25) show that none of the grammatical subjects receive a structural case,

namely the nominative case based on their grammatical position. Instead, these subjects receive a

purely idiosyncratic lexical case assigned to them by the particular verbs given inside the VP

domain before they raise to the TP domain. Semantically, the common feature that these subjects

share is that they are internal arguments of the verb.

20
As for the other type of nonstructural cases, namely the inherent case, Woolford argues that this

type has the two properties in (26):

(26)

(a) It is strictly assigned outside the VP domain but inside the vP phase.3

(b) It is only assigned to external arguments (e.g. goals, agents, cause, etc.). This case is licensed

only by little/light v heads.

As examples of nonstructural inherent case, Woolford offers the following pairs (examples 16-17

adapted from Wooford 2006: 118):

(27) a. þeir skiluðu Maríu bókinni.

they returned Mary-DAT book-the-DAT

‘They returned the book to Mary.’ (Icelnadic)

b. Maríu var skilað þessari bók.

Mary-DAT was returned this book-DAT

3
Woolfrod assumes that goal arguments are assigned the inherent case by a functional head v, which intermediates
VP and a higher v, which is the head of the phase vP.
21
‘The book was given to Mary.’ (Icelandic)

(28) a. Sie hilft ihm.

she helped him-DAT

‘She helped him.’ (German)

b. Ihm wird geholen.

he-DAT is helped

‘He is being helped.’ (German)

The examples in (27) and (28) are cases of inherent case given that they are preserved under A-

movement. Woolford (2006) shows that unlike lexical cases, which are purely idiosyncratic and

unpredictable, inherent cases are regular and predictable. Thus, all external arguments in

Icelandic receive the dative case when they do not receive the structural nominative case.

Woolford (2006) provides some diagnostic tests that can be used to separate structural

from nonstructural cases, and argues that once certain interfering factors are controlled for, one

can see that although inherent cases are thematically connected, inherent case is always

associated with external arguments as broadly construed rather than with the specific thematic

role of agents. For example, one of the most widely used diagnostic tests to distinguish syntactic

structural cases from syntactic non-structural cases is case preservation under A-movement.

However, Woolford argues that this diagnostic test might give misleading results when there are

some interfering factors. One such factor that Woolford mentions is the fact that in langugaes

22
such as Japanese, dative cases are prohibited in intransitive constructions. According to

Woolford, if one were to apply the test of case preservation under A-movement to the Japanese

pair in (29), one might falsely conclude that datives are structural cases in Japanese:

(29) a. John-ga Mary-ni soodansita.

John-NOM Mary-DAT consult-PAST

‘John consulted Mary.’

b. Mary-ga John-ni soodans-(r)rare-ta.

Mary-NOM John-DAT consult-PASS-PAST

‘Mary was consulted by John.’

For Woolford, the reason why the dative in the monotransitive sentence in (29a) switches to

nominative in (29b) is simply the result of the fact that Japanese prohibits dative cases in

intransitive constructions. Once one controls for this factor, Woolford argues, as in the

diatransitive sentence in (30a), one can see that dative cases are non-structural cases given that

they preserve their case in passive constructions:

(30) a. John-ga Mary-ni sono hon-o okutta.

23
John-NOM Mary-DAT that book-ACC send-PAST

‘John sent Mary that book.’

b. Mary-ni sono hon-ga okur-are-ta.

Mary-DAT that book-ACC send-PASS-PAST

‘Mary was sent that book.’

To conclude, Woolford (2006) splits nonstructural cases into two types, the purely

idiosyncratic lexical case, which is assigned by V to its internal arguments inside VP, and the

inherent case, which is assigned by v to its external arguments outside of VP and inside vP.

2.6.2 On default case (Schütze 2001)

Schütze (2001) argues that Universal Grammar (UG) must have in its inventory a form of

case, which is spelled out when all other ways of assigning case in the syntax proper fail. This

type of case is what Schütze (2001) calls default case. Schütze (2001: 206) defines this case as in

(31):

(31) “The default case forms of a language are those that are used to spell out nominal

expressions (e.g. DPs) that are not associated with any case feature assigned or otherwise

determined by syntactic mechanisms.”

24
This situation, according to Schütze (2001), describes cases where DPs enter the synactic

derivation without any case features. Therefore, they survive to the PF and LF interface levels

without causing the derivation to crash. In the morphophonological component of the grammar,

which Schütze (2001) assumes, the caseless DP would receive a morphological case realization,

which varies from one language to another. Crucailly for Schütze (2001), default case is not

assigned by anything to anything. This means that an argument NP cannot receive default case.

Therefore, it cannot save a derivation from crashing; if it did, this would render the Case Filter

(i.e. the requirement that NPs in certain surface positions get case) vacuous. This is shown by the

fact that the accusative case, which is argued by Schütze (2001) to be the default case of English,

cannot save the derivation in the following ungrammatical sentences:

(32) a. *It seems him/he to be tired.

b. *It/*There was believed he/him.

c. *Him/*He to leave would be rude.

d. *It is important he/him to be on time.

e. *My desire he/him to succeed led me stray.

f. *Me/*I, Tuesday is fine. (Schütze 2001: 208-209)

25
Schütze (2001: 209-210) further defines default case by what it is not. He points out four other

possible ways of assigning case, all take place in the syntax although he believes that nothing

would change if some or all of these four ways of case assignment turn out to be taking place in

the morphology before default case is supplied. The first way of assigning case is that which is

assigned by a syntactic head (e.g. a verb or Infl). Schütze (2001) points out that this is not the

same as syntactic licensing of arguments (i.e. abstract case, which is concerned with the

restrictions on the surface positions of certain arguments). He gives as an example Infl in

Icelandic, which may structurally license a dative subject, while at the same time assigning

nominative case to the object or other cases, where the verb (or AgrO) would structurally license

a nominative object and fail to assign its accusative case at all. The second type of case

assignment is that of a semantically related constituent such as a left-dislocated NP, which might

get a case, which matches that of its syntactically related argumental NP. The third type of case

is that of the case, which is assigned to the D head of a DP, and is then percolated to all other

material inside the DP in what is known as concordial case. The fourth type of case is what is

called semantic or adverbial case. This is the case of bare NPs that serve an adverbial function,

and their particular meaning is dependent on the particular case they receive such as the dative of

duration and the ablative of instrumental. Default case, argues Schütze (2001), is the case

received by an NP when none of the other mechanisms of case assignment is applicable.

2.7 Summary

This chapter traces the developments of case in Chomsky’s thinking from the early days

of the Government and Binding theory (Chomsky 1980) to the recent depevelopments (Chomsky

26
2008). It is shown that in the early days of the Government and Binding theory, case is perceived

by Chomsky as a PF requirement on all lexical (phonetically realized) NPs. In the later days of

the Government and Binding (Chomsky 1986), case is now perceived as an LF requirement on

arguments so that they can be readable by the LF component of the grammar. In early

Minimalism, goals (i.e. NPs) enter the derivation fully specified for their case features, but they

require checking against matching case features of the probes (i.e. T or v). In later Minimalism,

goals enter the derivation without a case value, and they only acquire valuation through an Agree

relation with a probe that must be ɸ-complete. Thus, ɸ-defective probes cannot assign case in

this model. In derivations by phase (Chomsky 2001), goals of a lower strong phase are still

accessible even if their features are deleted; they only freeze (cease to be accessible) when a

higher strong phase is introduced. In three factors in language design (Chomsky 2005), only the

head of the phase is accessible for further syntactic computations. In uproaching UG from below

(Chomsky 2007), C and v* have uninterpretable functional features, which they transfer to their

labels (i.e. immediately selected heads). Thus, the lable of C is T, and the label of v* is V. In this

model and in on phases (Chomsky 2008), CP, v*P and DP are phases.

Two other refinements to case theory are also introducd. Woolford (2006) splits non-

structural cases into two. The first is the purely idiosyncratic lexical case which is assigned by

certain verbs and prepositions to their complements. This case is assigned strictly inside VP. The

second non-structural case is the inherent case which is linked to thematic roles and is assigned

outside VP but inside vP exclusively to external arguments. Schütze (2001) further refines case

theory by claiming that there is another case which is assigned whenever all other case

assignment mechanisms fail. He calls this case the default case which is assigned at PF.

27
Chapter Three

Introduction to SA and previous accounts of case

3.1 Introduction

This chapter introduces some aspects of SA that are relavant for this thesis. The chapter

also gives an overview of how medieval Arab grammarians accounted for the case alternations

on nouns. Then, the previous generative accounts of case in SA are introduced and critically

discussed.

3.2 The syntax of SA: A brief overview

Below I introduce some of the relavant structures of SA. These particular syntactic

structures are the ones that will be discussed when the case assignment facts of SA are addressed

in chapter five. Introducing these particular structures will thus set the stage for the analysis of

case assignment to be developed in chapter five.

(a) There are two major orders in SA. These are VSO and SVO. All of the other possible

orders are also instantiated in the language with various pragmatic effects (e.g. Focus and Topic).

The examples in (1a-b) show the VSO and SVO orders. The example in (1c) shows the order

when the object is focalized, and the example in (1e) shows when the preverbial DP is a base-

generated topic. The example in (1f) is an example where the object is shifted to a position

preceding the postverbal subject for reasons of topcalization or focalization. The example in (1g)

is an example, where the object ʔal-masʔalat ‘the problem’ is not one of the arguments of the

sentence, but is adjoined to the sentence.

28
(1) a. VSO ḥall-a l-ṭaalib-u l-masʔalat-a.

solved-3MSG the-student-NOM the-problem-ACC

‘The student solved the problem.’

b. SVO l-ṭaalib-u ḥall-a l-masʔalat-a.

the-student-NOM solved-3MSG the-problem-ACC

The student solved the problem’

c. OVS L-MASʔALAT-A hall-a-(haa) l-ṭaalib-u.

the-problem-ACC solved-3MSG-it the student-NOM

‘It was the problem that the student solved.’

e. OVS l-masʔalat-u, hall-a-*(haa) l-ṭaalib-u.

the-problem-NOM sloved-3MSG-it the-student-NOM

‘As for the problem, the student solved it.’

f. VOS hall-a L-MASʔALAT-A l-ṭaalib-u.

sloved-3MSG the-problem-ACC the-student-NOM

‘It was the problem that the student solved.’

g. OSV l-masʔalat-u, l-ṭaalib-u hall-a-*(haa)

the-problem-NOM the-student-NOM solved-3MSG-it

29
‘The problem, the student, he solved it.’

(b) There is an agreement asymmetry in SA so that the number feature is realized on the verb

only in the SVO order but not in the VSO order. The verb in (2a) is singular even though the

subject is plural. This contrasts with the example in (2b), where the plural feature on the verb

matches that of the subject. This contrasts with the the gender feature, which is manifest on the

verb in both orders.

(2) a. hall-a/*-uu l-ṭullaab-u l-masʔalat-a.

solved-3MSG/*3PLM the-students-NOM the-problem-ACC

‘The students solved the problem.’ (VSO)

b. l-ṭullaab-u hall-uu/*-a l-masʔalat-a.

the-students-NOM solved-3MPL/*3MSG the problem-ACC

‘The students solved the problem.’ (SVO)

(c) SA has two major types of sentences. The first is the verbal sentence (3a), and the second

is the ‘verbless sentence’ (i.e. clauses lacking any verbal predicates) (3b).4

(3) a. hall-a l-ṭullaab-u l-masʔalat-a.

solved-3MSG the-students-NOM the-problem-ACC

4
The term ‘verbless sentences’ is used to describe snetences that lack a verbal predicate (cf. Benmamoun 2008 and
Hazout 2010).
30
‘The students solved the problem.’ (Verbal sentence)

b. l-ṭullaab-u mujtahid-uun.

the students-NOM hardworking-PLM.NOM

‘The students are hardworking.’ (Verbless sentence)

(d) In addition to verbal and verbless sentences, SA has complex tense structures (i.e.,

structures with auxiliary verbs as well as lexical verbs), as in (4).

(4) kaan-at l-ṭaalibaat-u y-alʕab-na.

was-3FSG the-students.F-NOM 3-play-FPL

‘The students were playing.’

(e) SA has structures with encliticized backgrounding topics. This is shown in (5), where the

encliticized backgrounding topic–hu ‘it’ is used. This particular structure is interesting because

the status of the clitic –hu is not agreed upon. It is a backgrounding topic for Fassi Fehri (2012),

but a preverbal expletive clitic for Mohammad (2000: 108) and Aoun et. al. (2010: 17). This

makes this structure interesting from a case-theoretic perspective. Mohammad (2000) and Aoun

et.al. (2010) argue that that there are two structural cases here, the accusative case is assigned to

the preverbal expletive clitic -hu ‘it’ by the complementizer and the nominative case is assigned

to the postverbal subject ʔal-musaafir-uun ‘the passengers’ by T.

31
(5) qul-tu ʔinna-*(hu) waṣal-a l-musaafir-uun.5

said-1SG that.EMPH-it.ACC arrived-3MSG the-passenger-MPL.NOM

‘I said it that indeed the passengers have arrived.’

(f) SA has structures with there-type expletives, as in (6). The important question that these

structures raise is the following: what is the status of hunaaka ‘there’, and does it receive case?

(6) hunaaka rajul-un fii l-daar-i.6

there man-NOM in the-house-GEN

‘There is a man in the house.’

(g) SA has exceptional case marking (ECM) structures. In (7), the embedded DP subject,

ʔal-musaafir-iina ‘the passengers’ does not receive the nominative case; rather, it is marked for

morphological accusative case. The SA embedded clause of ECM constructions is finite both

morphologically (having ɸ-features) and semantically (having a distinct tense specification from

that of the matrix clause). The question here is what sort of case the subject of the embedded

clause receives, and where it receives this case from.

5
The counterpart of (5) without a backgrounding topic is as shown in (i)

(i) qul-tu ʔinna l-l-musaafir-iina waṣal-uu

Said-1Sg that.EMPH the-passenger-MPL.ACC arrived-3PLM
‘I said that passengers indeed have arrived.’
6
Strictly speaking, the –an part of the NP rajulun ‘a man’ is two morphemes. The –a marker is an accusative case
morpheme, and the -n marker is either an indefinite marker (Kouloughli 2007) or the head of a possessive phrase,
which marks the absence of the possessor (Fassi Fehri 2012, fn 2: 294). Throughout this chapter, I gloss the –n
morpheme as part of the case marker unless the separation is essential for the point under discussion, as will be
shown when case assignment in PPs is discussed later in the thesis.

32
(7) ḏanan-tu l-musaafir-iina waṣal-uu.

believed-1SG the-passenger-MPL.ACC arrived-3MPL

‘I believed the passengers to have arrived.’

(h) SA instantiates double-object constructions. The DPs ʔibnat-a-haa ‘her daughter’ and

laban ‘yogurt’ are both objects of the complex verb ʔa-ṭʕam ‘cause to eat/feed’ in (8), and both

objects receive the accusative case. The question is why both objects receive the accusative case

and what happens in the context of passive sentences.

(8) ʔa-ṭʕam-at l-ʔumm-u ʔibnat-a-haa laban-an.

CAUS-eat-3FSG the-mother-NOM daughter-ACC-her yogurt-ACC

‘The mother fed her daughter yogurt.’

(i) SA instantiates two aspectual forms of the verb, the imperfective and the perfective. The

imperfective form of the verb is characterized by encoding mood markers, as in (9a-c). Three

major types of mood are encoded, and these are the indicative, as in (9a), the subjunctive, as in

(9b), and the jussive, as in (9c).7 The perfective form of the verb in SA bears no morphological

marking of mood, as in (9d). In this form of the verb, all the phi-features are realized as a suffix.

The imperfective verbs in (9a-c) are marked for mood as a suffix, and the phi-features appear as

7
According to Wright (1967, as cited in Al-Balushi 2011: 61), there are other moods in SA, and these are the
imperative and the energetic (or emphatic).
33
both prefixes and suffixes. In contrast, the perfective verb in (9d) is not marked morphologically

for mood, and the phi-features are all realized as a suffix:

(9) a. y-arsum-Ø-u l-ṭaalib-u lawḥat-an.

3-draw-SG-INDIC the-student.MSG-NOM picture-ACC

‘The student usually draws pictures/The student is drawing a picture.’

b. l-ṭullaab-u lan y-arsum-uu-Ø lawḥat-an.

the-students-NOM NEG.FUT 3-draw-PL-SUBJ picture-ACC

‘The students will not draw a picture.’

c. l-ṭullaab-u lam y-arsum-uu-Ø lawḥat-an.

The-students-NOM NEG.PAST 3-draw-MPL-JUSS picture-ACC

‘The students did not draw a picture.’

d. rasam-at l-ṭaalib-at-u lawḥat-an.

drew-3FSG the-student-FSG-NOM picture-ACC

‘The student drew a picture.’

(i) SA instantiates pronominal copulas. These are pronouns that can be used as pronominal

linking verbs by linking the subject and the predicate in a clause. Their use in these structures is

34
one of disambiguation, as it is through them that clauses are distinguished from noun phrases.8 In

the examples in (10a-e), the third person pronoun is used as a linking verb, and it always agrees

with the preverbal subject in number and gender, but not person.

(10) a. huwa huwa l-muʕallim-u.

he be. MSG the-teacher-NOM

‘He is the teacher.’

b. ʔanaa huwa l-muʕallim-u.

I be. MSG the-teacher-NOM

‘I am the teacher.’

8
For example, names can only be modified inside the NP by a postnominal definite NP modifier. Thus, the
expressions zayd-un l-muhandis-u ‘zayd, the engineer’ and zayd-un l-muḥaasib-u ‘zayd, the accountant’ are both
definite NPs with postnominal modifiers which agree with the head noun in definiteness. They both lack the
pronominal copula in (i); therefore, they can only be understood to be NPs with postnominal definite NP modifiers.
The use of the copula in (ii), however, makes the only reading available that of the clause. In (iii), the only available
reading of the example is that of a clause, given that the postnominal predicate is indefinite, hence, does not agree
with the head noun in definiteness, and no NPs need to be disambiguated from clauses. Therefore, no pronominal
copula is needed in (iii):
(i) zayd-un l-muhandis-u waṣal-a, ʔamma zayd-un l-muḥaasib-u
Zayd-NOM the-engineer-NOM arrived-3MSG as.for Zayd-NOM the-accountant-NOM
fa-lammaa y-aṣil-Ø-Ø baʕdu.
FA-NEG 3-arrive-MSG-JUSS yet
‘Zayd, the engineer arrived. As for Zayd, the accountant, he has not arrived yet.’
(ii) zayd-un huwa l-muhandis-u.
Zayd-NOM be.3MSG the-engineer-NOM
‘Zayd is the engineer.’
(iii) zayd-un muhandis-u-n.
Zayd-NOM engineer-NOM-INDEF
Zayd is an enginner.’
On the use of the pronominal copular verbs as a means of disambiguating NPs from clauses in SA, the reader is
referred to Al-Naḥas (1995: 13-34) and Eid (1983).
35
c. ʔanta huwa l-muʕallim-u.

you be.MSG the-teacher-NOM

‘You are the teacher.’

d. ʔanti hiya/*huwa ʔanaa

you.F be. FSG/*be.MSG I.NOM

‘You are me.’ (i.e., you are playing me in a play, etc.)

e. ʔanta huwa/*hiya hiya.

you.MSG be. MSG/*be.F.SG she.F.NOM

‘You are her.’ (i.e., you are playing her in a play, etc.)

(j) There are two major complementizers in SA. The first is the indicative complementizer.

This has two forms. The first of these forms is the emphatic (or assertive) form, ʔinna. This form

is characterized by being able to introduce root clauses (11a), matrix or embedded verbless

clauses (11b-c), and tensed embedded clauses (11d), and it typically follows verbs of saying in

embedded contexts. This complementizer assigns the accusative case to its adjacent NP.

(11) a. ʔinna l-ʔamṭaar-a ʔa-ġraq-at l-madiinat-a

that.EMPH the-rains-ACC CAUS-flooded-3FSG the-city-ACC

‘Indeed, the rain flooded the city.’

36
b. ʔinna ʔal-ṭaqs-a ḥaar-un l-yawm-a.

that.EMPH the-weather-ACC hot-NOM the-day-ACC

‘Indeed, the weather is hot today.’

c. qul-tu ʔinna l-ṭaqs-a ḥaar-un l-yawm-a.

said-1SG that.EMPH the-weather-ACC hot-NOM the-day-ACC

‘I said that the weather is indeed hot today.’

d. qult-tu ʔinna l-ʔamṭaar-a ʔa-ġraq-at

said-1SG that.EMPH the-rains-ACC CAUS-flooded-F

l-madiinat-a

the-city-ACC

‘I said that the rain has indeed flooded the city.’

The second form of the indicative complementizer is the neutral form, ʔanna, which can

introduce only embedded tensed clauses (12a) and embedded verbless sentences (12b).9This

variant also assigns the accusative case to its adjacent NP.

9
Traditional Arab grammarians treat the complementizer ʔanna as an emphatic (assertive) particle on a par with
ʔinna. However, Muṣṭafa (1957: 5) and Al-Maxzuumy (1986) argue that ʔanna does not bear any emphatic or
assertive force. Al-Maxzuumy (1986: 316-317), in particular, shows that ʔanna cannot possibly emphasize that the
eventuality expressed in its clause actually took place because this particle can introduce embedded sentences that
express the meaning of doubt, as in (i):
(i) ʔ-ašukk-u ʔanna-ka muṣiib-un.
1SG-doubt-INDIC that-you.MSG right-NOM
‘I doubt that you are right (Al-Maxzuumy 1986: 317)
I follow Muṣṭafa (1957) and Al-Maxzuumy (1986) on this issue. Note further that ʔanna, but not ʔinna can be found
in conditional clauses, which encode hypothetical situations, as in (ii):
(ii) law ʔanna /*ʔinna zayd-an saʔal-a-n-ii, la-ʔajab-tu-hu.
37
(12) a. ʕalim-tu ʔanna l-ʔamṭaar-a ʔa-ġraq-at

learned-1.SG that.NEUT the-rains-ACC CAUS-flooded-3FSG

l-madiinat-a

the-city-ACC

‘I learned that the rain had flooded the city.’

b. ʕalim-tu ʔanna l-ʔamṭaar-a ġaziirat-un.

learned-1SG that.NEUT the-rains-ACC heavy-NOM

‘I learned that there is heavy rain.’

The second major type of complementizers in SA is the subjunctive complementizer, ʔan,

which obligatorily introduces embedded clauses in the subjunctive form, as in (13) below:10

(13) ʔaraad-at l-ṭaalibat-u ʔan t-arsum-Ø-a/*u/*Ø

wanted-3FSG the-student-NOM that.SUBJ F-draw-SG-SUBJ/*INDIC/*JUSS

lawḥat-an.

picture-ACC

if that.NEUT/that.EMPH Zayd-ACC asked-3MSG-n-me CONJ-answered-1SG-him

‘Had Zayd asked me, I would have answered him.’
The sentence in (ii) confirms the proposal that ʔanna, but not ʔinna has a neutral sense, for one can only confirm
things that took place, not things that might have taken place, but did not.
10
I will argue in Chapter 5 that both the indicative complementizers ʔinna/ʔanna and the subjunctive
complementizer ʔan have phasal status.
38
‘The student wanted to draw a picture.’

(l) SA has structures equivalent to obligatory control structures in other languages, where

the subject of the null category in the embedded clause is controlled by the referent in the matrix

clause, as in (14):

(14) a. ḥaawal -a zayd-uni [CP ʔan ei/*j y-uġaadir-Ø-a.]

tried-3MSG Zayd-NOM that.SUBJ 3-leave-MSG-SUBJ

‘Zayd tried to leave.’

b. ʔistaṭaaʕ-a zayd-uni [CP ʔan ei/*j y- uġaadir-Ø-a.]

managed-3MSG Zayd-NOM that.SUBJ 3-leave-MSG.SUBJ

‘Zayd managed to leave.’

This type of sentences is important, as the status of the subject of the embedded clauses (pro,

PRO or NP-trace) has consequences from a case-theoretic perspective.

3.3 Case in Medieval Arabic grammar

Medieval Arab grammarians of the 8th centrury and their followers had their account of

case alterations on nouns in SA. The major contribution of the Medieval Arab grammarians of

the 8th century is the discovery that certain lexical items can influence the case of the nouns.

Thus, case endings on nouns (rafʕ ‘nominative case’, naṣb ‘accusative case’, and jarr ‘genitive

case’) can be determined by the presence or lack thereof of certain lexical items that precede

39
them, which the Medieval Arab grammarians called operants (see Baalabaki 2008: 32-33). The

following is just a quick overview of the basic ideas expressed in this rich tradition:

(a) The subject is obligatorily postverbal (this is the view of the predominant Baṣrah

grammarians of the 8th century, specifically Sibawayhi and his followers). The subject in this

view is assigned the nominative case by the operant, the verb, as can be seen in the following

example (cf. Ibn ʕaqiil 13th c./1980, Vol.2: 76):

(15) ʔataa Zayd-un.

came.3MSG Zayd-NOM

‘Zayd came/Zayd has come’ (Ibn ʕaqiil 13th century/1980: 76)

(b) When the NP appears before the verb, the NP is not a subject but a preverbal topic phrase,

and the subject is a covert pronoun (this is also the predominant view of the Baṣrah grammarians

of the 8th century, specifically Sibawayhi and his followers) (cf. Ibn ʕaqiil 13th c./1980, Vol.2:

77). This is illustrated by the example in (16):

(16) Zayd-un qaam-a (*huwa)

Zayd-NOM stood.up-3MSG (*he)

‘Zayd, he stood up.’

In (16), the Baṣrah grammarians of the 8th century propose that the preverbal NP Zayd ‘Zayd’ is

a topic phrase (called a mubtadaʔ in their terminology), and the true subject of the sentence is the

obligatorily covert (phonetically unpronounced pronominal subject huwa ‘he’). In this view, the

preverbal NP Zayd ‘Zayd’ receives its nominative case as a reflex of the fact that it introduces

40
the sentence and nothing else precedes it. Since no overt operant seems to influence the

nominative case of the topic phrase, Medieval Arab grammarians propose that there is a covert

operant which determines the nominative case on the topic phrase, and they term this covert

operant ʔal-ʔibtidaaʔ ‘being the first lexical item that introduces the sentence and no overt

operants precede it’. As for why nominative case should mark topic phrases, they propose that

nominative case is the first state of the nouns (i.e. the origin), and any other cases including

nominative cases on NPs other than preverbal topics are to be explained by the influence of

either an overt or a covert operant (cf. Baalabaki 2008: 76).

(c) The subject of verbless (or nominal sentences in Medieval Arabic grammar) is also a topic

phrase (termed ʔal-mubtadaʔ) rather than a subject. The NP predicate is termed ʔal-xabar ‘the

comment’. This is illustrated in (17):

(17) Zayd-un ʕaalim-un.

Zayd-NOM scientist-NOM

‘Zayd is a scientist.’

In this tradition, the NP Zayd ‘Zayd’ is a topic phrase and the NP predicate ʕaalim ‘scientist’ is a

comment. The nominative case on the topic phrase is assigned by the covert operant ‘ʔal-

ʔibtidaaʔ (i.e. being the first lexical item that introduces the sentence and is not preceded by any

overt operants), and the nominative case on the NP predicate is assigned by the topic phrase

itself (cf. Ibn ʕaqiil 13th c. 1980, Vol. 2: 200).

(d) The subject of passive sentences is assigned the nominative case by the verb (cf. Ibn ʕaqiil

13th c. 1980, Vol. 2: 111). This is illustrated in (18):

41
(18) qutil-a Zayd-un

killed-3MSG Zayd-NOM

‘Zayd was killed.’

(e) The active transitive verb can assign the accusative case to one object, two objects or three

objects (cf. Hasan 1962: 7). This is illustrated in (19):

(19) a. ʔištary-tu qamḥ-an

bought-1SG wheat-ACC

‘I bought wheat.’

b. ḏann-a l-ṭayyaar-u l-biyuut-a ʔakwaax-an

believed-3MSG the-pilot-NOM the-houses-ACC cottages-ACC

‘The pilot believed the houses to be cottages.’ (Hasan 1962: 7)

c. ʔ-aray-tu l-xubaraaʔ-a l-ʔaaṯaar-a kunuuz-an

CAUS-showed-1SG the-experts-ACC the-ruins-ACC treasures-ACC

‘I showed the experts that the ruins were treasures.’ (Hasan 1962: 58)

In (19a), traditional Arab grammarians hold that the object NP qamḥ ‘wheat’ is assigned the

accusative case by the verb. In (19b), they hold that the NPs l-biyuut ‘the houses’ and ʔakwaax

‘the cottages’ are both objects of the matrix verb ḏann ‘to believe’, and that the verb assigns

accusative case to both objects. In (19c), traditional Arab grammarains hold the view that the

NPs l-xubaraaʔ ‘the experts’, l-ʔaaṯaar ‘the ruins’, kunuuz ‘treasures’ are all objects of the
42
matrix verb raʔara ‘to show/to cause to see’, and that the matrix verb assigns accusative cases to

all three objects (cf. Hasan 1962: 150).

(f) The auxiliary verb ‘k-w-n’ can be added to verbless (i.e. nominal sentences in Medieval Arbic

grammar), and it assigns the nominative case to the topic phrase and the accusative case to the

NP predicate (or adjectival predicate). Other verbs which belong to the same class are the verbs

ḏalla, baata, ʔaḍḥaa, ʔaṣbaḥa ‘to change from one state to another’,zaala, baraḥa, fatiʔa,

ʔinfakka, and all have the meaning of ‘to continue’. To illustrate, consider the following example

with the auxiliary verb ‘k-w-n’:

(20) kaan-a Zayd-un ʕaalim-an

was-3MSG Zayd-NOM scientist-ACC

‘Zayd was a scientist’

Tradional Arab grammarians hold the view that the auxiliary verb assigns nominative case to the

first NP Zayd ‘Zayd’, and accusative case to the predicate nominal ʕaalim ‘scinetist’ (cf. Ibn

ʕaqiil 13th c./1980, Vol. 1: 261).

(g) Grammarians of the Baṣrah school of grammar hold the view that the particles ʔinna ‘that of

emphasis’, ʔanna ‘that’, laakinna ‘but’, and the verbs layta ‘wish’, laʕalla ‘beseach’ can all be

added to verbless sentences, and they have the opposite effect to that of the auxiliary ‘k-w-n’ and

the other verbs in its class in that they assign the accusative case to the topic phrase and the

nominative case to the predicate nominal (cf. Ibn ʕaqiil 13th c./1980, Vol. 1: 345). This is shown

in the following example:

43
(21) ʔinna Zayd-an ʕaalim-un

that.EMPH Zayd-ACC scientist-NOM

‘Indeed, Zayd is a scientist.’

In contrast to the view of the Baṣrah grammarians, grammarians of the Kuufa school of grammar

hold the view that the nominative case of nominal predicate in examples such as (21) is not

assigned; instead, this case for them is the result of the absence of any overt operants that assign

case to the predicate nominal (cf. Ibn ʕaqiil 13th c./1980, Vol. 1: 348).

(h) To account for those cases where the topic phrase receives the accusative case instead of the

predicted nominative case assigned by the covert operant ‘ʔal-ʔibtidaaʔ (i.e. being the first

lexical item that introduces the sentence and is not preceded by any overt operants), traditional

Arab grammarians claim that the topic phrase in such constructions is assigned accusative case

by a covert verb, which is identical to the overt verb that preceded the overt postverbal subject.

Thus, they propose that the sentence in (22a) is covertly understood as (22b):

(22) a. Zayd-an ḍarab-tu-hu

Zayd-ACC hit-1SG-him

‘Zayd, I hit him’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 156)

b. (*ḍarab-tu) Zayd-an ḍarab-tu-hu

hit-1SG Zayd-ACC hit-1SG-him

‘I hit Zayd, I hit him.’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 156)

44
(22a) is the actual sentence where the preverbal NP Zayd ‘Zayd’ receives the accusative case

instead of the nominative case that topic phrases should receive according to the theory put

forward by traditional Arab grammarians. To account for this deviation, traditional Arab

grammarians (cf. Ibn ʕaqiil 13th c./1980, Vol. 2: 150) claim that verb ḍarab ‘to hit’ assigns the

accusative case to the bound pronominal object -hu ‘him’. As for the preverbal topic phrase, it is

assigned the accusative case by an elided token of the overt verb ḍarab ‘to hit’. Thus, the setnece

in (22b) is ungrammatical when the initial token of ḍarab ‘to hit’ is overt.

(i) The accusative case borne by the cognate object (called absolute object in traditional Arabic

grammar) is assigned to it by either the nominlized verb (23a), the verb (23b) or the participle

(23c) (cf. Ibn ʕaqiil 13th c./1980, Vol. 2: 170):

(23) a. ʕajib-tu min ḍarb-i-ka Zayd-an ḍarb-an

shocked-1SG by hitting-GEN-your.MSG Zayd-ACC hitting-ACC

šadiid-an

brutal-ACC

‘I was shocked by your brutal hitting of Zayd.’ (Ibn ʕaqiil 13th c./1980, Vol. 2:

150)

b. ḍarab-tu Zayd-an ḍarb-an

hit-1SG Zayd-ACC hitting-ACC

‘I brutally hit Zayd.’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 150)

45
c. ʔanaa ḍaarib-un Zayd-an ḍarb-an

I hit.PTPL-NOM Zayd-ACC hitting-ACC

‘I am hitting Zayd brutally/I will be brutally hitting Zayd.’(Ibn ʕaqiil 13th c./1980,

Vol. 2: 150)

In (23a), traditional Arab grammarians hold that the absolute object (=cognate object) ḍarb

‘hitting’ is assigned the accusative case by the nominalized verb ḍarb-i-ka ‘your hitting’. In

(23b), they hold the view that the absolute object is assigned the accusative case by the verb, and

in (23c), it is assigned the accusative case by the present participle ḍaarib ‘hitting’.

(j) The object of reason or purpose is assigned the accusative case by the preceding verb (24a):

(24) jud-Ø-Ø šukr-an

be.generous-2MSG-JUSS gratefulness-ACC

‘Be generous as a gesture of gratefulness.’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 192)

(k) The object of accompaniment (i.e. comitative object) is assigned the accusative case either by

the preceding verb (25a) or the preceding participle (25b):

(25) a. siir-ii wa l-ṭariiq-a

walk-2FSG and the-road-ACC

‘Walk along the road.’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 202)

b. Zayd-un saaʔir-un wa l-ṭariiq-a

Zayd-NOM walk.PTPL-NOM and the-road-ACC

46
‘Zayd is walking along the road/Zayd will be walking along the road.’ (Ibn ʕaqiil

13th c./1980, Vol. 2: 202)

(l) Adverbial NPs are assigned the accusative case either by the verb (26a) or by the nominalized

verb (26b) (cf. Ibn ʕaqiil 13th c./1980, Vol. 2: 192):

(26) a. ḍarab-tu Zayd-an yawm-a l-jumuʕat-i

hit-1SG Zayd-ACC day-ACC the-Friday-GEN

‘I hit Zayd on Friday.’ (Ibn ʕaqiil 13th c./1980, Vol. 2: 192)

b. ʕajib-tu min ḍarb-i-ka Zayd-an

shocked-1SG by hitting-GEN-your.MSG Zayd-ACC

yawm-a l-jumuʕat-i

day-ACC the-Friday-GEN

‘I was shocked by your hitting of Zayd on Friday.’ (Ibn ʕaqiil 13th c./1980, Vol. 2:

192)

3.4 Previous generative accounts of case in SA

Various treatments of case in SA have been offered in Agree based accounts of case, as

in Raḥḥali (2003), Ouhalla (2005), Soltan (2007), Al-Balushi (2011, 2012), and Leung (2011).

Another account is that of Fassi Fehri (1993), which is the only work in the literature on SA

where a hierarchical case account is proposed. The following subsections address these accounts

and discuss the problems that they encounter.

47
3.4.1 Agree-based account of case in SA

Raḥḥali (2003: 147) proposes that in SA, structural nominative case is licensed by T, and

he proposes that case on an NP is checked in one of two ways.11 Assuming Chomsky (1998,

1999), Raḥḥali (2003: 147) proposes the nominative parameter in (27):

(27) The nominative parameter:

The nominative feature is checked:

(a) Either via Agree, or

(b) via incorporation

Raḥḥali (2003: 147) proposes that in the VSO order, the subject does not raise to Spec,

TP. Instead, it is in Spec, vP, and nominative case is checked via an Agree relation between the

functional head T and the subject in Spec, vP. This is illustared by (28):

(28) [TP naam-a T [vP l-ʔawlaad-u v VP]]

slept-3MSG the-boys-NOM

‘The boys slept.’ (Raḥḥali 2003: ex. 45a: 147)

11
There is controversy in the literature on whether the SA verbal inflection shows a tense distinction (traditional
Arab grammarians) or an aspectual distinction (Caspari 1859, as cited in Fassi Fehri 1993: 141). Following Fassi
Fehri (1993, chapter 4: 141-212), I assume that the verbal inflection in SA shows tense, aspect and mood features.
48
The author claims that the subject DP l-ʔawlaad ‘the boys’ checks its [uCase: NOM] feature

against the [uCase: NOM] feature of the functional head T via the relation Agree.

For Raḥḥali (2003), another way of checking the structural nominative feature of the

subject DP is via incorporating the subject into the functional head T, and this will yield the null

subject cases (i.e. cases with no overt lexical subject postverbally). This is illustrated in (29):

(29) [TP naam-uu [vP <-uu> v VP]].

slept-3MPL

‘They slept.’ (Raḥḥali 2003, ex. 45b: 147)

Following Fassi Fehri (1990, 1993), Raḥḥali (2003) assumes that pronominal subjects

incorporate into the functional head T, which hosts the lexical verb, the latter raises to T from v-

V. The pronominal subject in (29) is in Spec, vP, and it raises to incorporate to T in order to

check its [uCase: NOM] feature against the matching [uCase: NOM] feature of the functional

head T. As for the SVO order, Raḥḥali (2003) treats the preverbal lexical DP as a topic in a left-

peripheral position.

The major problem with Tense as the locus of case licensing is that it is not tenable cross-

linguistically. A number of linguists argue that DPs bear nominative case even when the

functional head T is non-finite (where finiteness is defined morphologically and semantically as

49
the availability of tense and agreement). For example, McFadden and Sunderasen (2011) provide

the example in (30) from Tamil:

(30) a. [vasu poori porikk-a] raman maavu

vasu.NOM poori.ACC fry-INF Raman.NOM flour.ACC

vaangi-n-aan

buy-PST-M.3SG

‘Raman bought flour for Vasu to fry pooris’ (McFadden and Sunderasen (2011,

ex. 7a: 5)

b. [naan poori porikk-a] raman maavu

[I.NOM poori.ACC fry-INF Raman flour.ACC

vaangi-n-aan

buy-PST-M.3SG

‘Raman bought flour for me to fry pooris’ (McFadden and Sunderasen (2011, ex.

7b: 5)

The authors point out that in (30a), the DP Vasu gets nominative case even though it cannot be

said to have agreed with the embedded verb, since the latter is clearly an infinitive, heading a

tenseless clause. The example in (30b) shows that even agreement with the matrix verb is ruled
50
out, since the pronoun naan ‘I’ and the agreement markers on the matrix verb are not the same.

In addition, the embedded clause in the examples above is an adjunct clause, therefore, any

argument that the DP in the embedded clause gets its nominative case via Agree with the matrix

verb does not hold.

McFadden and Sunderasen (2011: 6) provide the examples in (31) from Icelandic, which

point to the same conclusion:

(31) a. Jóni ?*virðist/virðast vera *talið/taldir

John.DAT seem.SG/PL be believed.NT.SG/M.PL

líka hestarnir.

like horses.NOM.M.PL

‘John seems to be believed to like horses.’ (McFadden and Sunderasen (2011, ex.

10a: 6)

b. Mér virðist/?*virðast Jóni líka hestarnir.

Me.DAT seem.SG/PL John.DAT like horses.NOM.M.PL

‘It seems to me that John likes horses.’ (McFadden and Sunderasen (2011, ex.

10b: 6)

51
In (31a), the object DP hestarnir ‘horses’ is unambiguously in a nonfinite clause. The subject DP

Jóni starts in the embedded clause and gets a dative case before raising to the matrix clause,

since the matrix predicate is clearly a raising verb that does not assign a theta role. According to

the authors, one can clearly claim that the subject DP fails to trigger agreement, being dative. In

this case, the object DP establishes an Agree relation with the matrix clause, triggers agreement

and gets nominative case as a result. However, (31b) shows that the object DP is still nominative

even when there is a dative subject intervening between the object DP and the matrix verb. In

other words, the object DP still gets the nominative case even when agreement is blocked by the

intervening dative subject Jóni ‘John’. That agreement is blocked is obvious from the fact that

the matrix verb gets default agreement. Crucial to this example is the fact that the object DP does

not lose its morphological nominative case when blocking takes place and no agreement is

established with the matrix verb.

Another argument against the view that Tense is the locus of structural nominative case

comes from sentences with participials in SA, as can be shown in (32):

(32) zayd-un qaariʔ-un l-kitaab-a

Zayd-NOM reading.PTPL-NOM the-book-ACC

‘Zayd is reading the book.’

‘Zayd will be reading the book sometime in the future.’

*’Zayd was readeing the book.’

The problem can be stated as follows. Fassi Fehri (1993: 181-184) and Al-Balushi (2011: 262-

264) offer convincing evidence that sentences with participials lack a tense specification in SA.

52
Al-Balushi derives this conclusion from the fact that participials do not encode the [± Past]

distinction, as can be seen in (33)

(33) a. l-mudarris-aa-t-u ʕaarif-aa-t-un l-xabar-a

the-teacher-p-f-Nom knowing-p-f-Nom the-news-Acc

‘the female teachers know the news today.’

*’the female teachers knew the news yesterday’

?? ‘the female teachers will know the news tomorrow’ (Al-Balushi 2011, ex. 140:

263)

b. l-malik-u ʕaazil-un ʔibn-a-hu

the-king-Nom firing-Nom son-Acc-his

? ‘The king is firing his son today’

* ‘the king was firing his son yesterday’

‘the king will be firing his son tomorrow’ (Al-Balushi 2011, ex. 141: 263)

Given examples such as (33), Al-Balushi establishes that participials lack the

interpretable feature [Precedence], which, according to Cowper (2005), T must have. Assuming

this analysis to be on the right track, we can proceed by raising the following question: If

sentences with participials such as those in (33) lack a tense specification (or a tense projection
53
in the case of Al-Balushi 2011), how is the nominative case of the subject of those sentences

licensed? In other words, if sentences with participials lack a tense specification, how can this be

reconciled with the claim that Tense is the locus of structural case? Clearly, the subject of the

sentences in the examples above receives nominative case despite the fact that the sentences do

not encode a tense specification. If this the case, then examples such as these run counter to the

predictions of the proposal that tense is the locus of case licensing in SA. Note that T in SA may

encode the features of T(tense and ɸ), Mood, and Aspect (cf. Fassi Fehri 1993: 151-152). In

principle, each of these features can license the nominative case of the subject. I take this as

evidence that the nominative case in (33a-b) is not default case, since default case is the case

mechanism that is bled by Agree-based case rather than vice versa (more on this in chapter 4).

The same arguments laid out against Raḥḥali (2003) also carry over to Benmamoun

(1999), where structural nominative case is also the result of an agreement relation between T

and the subject DP.

The second Agree-based account of case is that of Ouhalla (2005). In a novel attempt to

derive categorial features from independently needed features, and specifically from agreement

features, Ouhalla (2005) makes two major claims: (a) he uses data essentially from Berber to

advance the proposal that feature matching and deletion is categorization by computation. In

other words, following Chomsky’s (2001) claim that categorial features such as [N] and [V] have

no theoretical status, Ouhalla (2005) argues that categories such as [N] and [V] are not

primitives, which are transferred from the lexicon to the synatactic component; rather, these

labels are determined in the syntactic component of the grammar during the syntactic processes

of matching and deletion. This way, categories such as [N] and [V] are the result of checking

abstract features in the syntax. Subject-verb agreement, according to this view, reduces to the

54
process of matching and deleting the abstract agreement features of the verb and its related

functional heads against those features of the subject. Ouhalla provides arguments to claim that

the nominal category can be reduced to the feature [CLASS] and the verbal category, to the

feature [PERSON], and (b) the agreement properties normally associated with T in some

languages (e.g. English) is associated with an intermediate functional head, which Ouhalla calls

Pred, as in Berber. In this account, subject-verb agreement is essentially the result of an

agreement relation between a subject with the features [CLASS, PERSON, NUMBER] and a

functional head Pred/T with the features [PERSON, CLASS]. In the process of agreement, the

[PERSON] feature of the subject matches with the [PERSON] feature of the Pred/T. The result

of matching is that the [PERSON] feature must survive in the feature content of the Pred/T;

otherwise, Pred/T would fail to be categorized as a verbal category. By the same token, the

feature [PERSON] must be deleted from the feature content of the subject; otherwise, the subject

would fail to be categorized as a nominal category. Similarly, the feature [CLASS], which can

have different values in different languages (e.g. GENDER), must survive in the feature content

of the subject; otherwise, the subject would fail to be categorized as a nominal category, and the

feature [CLASS] must be deleted from the feature content of Pred/T; otherwise they would fail

to be categorized as verbal categories. In other words, both Pred/T and the subject have

conflicting categorial features, [CLASS], which is a nominal feature, and [PERSON], which is a

verbal feature. The only way to resolve the categorial conflict and be categorized as a verbal or a

nominal category is for the subject to delete its verbal [PERSON] feature and retain its nominal

[CLASS] feature, and for Pred/T to eliminate their nominal [CLASS] features and retain their

verbal [PERSON] features. Ouhalla (2005: 672) argues that the difference in subject-verb

agreement relative to T and the verb can be captured, as in (34):

55
(34) a. Pred [Person, Class] (LP)

b. T[Person, Class] (UP)

According to Ouhalla (2005: 672), the distribution of agreement features in (34a) describes

languages such as Berber and SA, where the subject does not raise to Spec, TP, and instead

remains in the lower phase (LP). By contrast, the distribution in (34b) describes languages such

as English, French and German, where the subject raises to Spec, TP, and agreement takes place

between T and the subject; therefore, agreement takes place in the upper phase (UP).

Ouhalla (2005: 681) argues that the feature [CASE] is likely to match with the verbal

feature [±Past] of T and the aspectual feature [±perfective] of Aspect. This means, Ouhalla

argues, that [CASE] is a verbal feature on nouns, and must therefore be deleted so that nouns can

be categorized as nominal categories.

Ouhalla (2005: 682) further claims that the feature [CASE] is available for nouns, but not

for pronouns, which only need [PERSON]. He states that “[t]here is a straightforward sense in

which [PERSON] plays a role in the interpretation/reference of personal pronouns, and there is

an equally straightforward sense in which the interpretation/reference of nouns does not depend

on [PERSON].” This, according to Ouhalla, is what justifies the existence of the verbal feature

[PERSON] alongside the verbal feature [CASE]. Ouhalla (2005: 682), moreover, claims that the

conclusion that pronouns are unspecified for [CASE] does not necessarily mean that the

distinction in the forms of the grammatical functions between the subject and the object cannot

be maintained. This distinction, Ouhalla argues, can still be maintained in terms of the feature

with which [PERSON] is matched. He writes that “[t]he subject form [of pronouns] is the result
56
of matching [PERSON] with the tense feature of T and the object form the result of matching

[PERSON] with the aspectual feature of the verb.”

This account also faces problems. Consider first the following problem with Ouhalla’s

(2005) account of case. Assume that pronouns are unspecified for [CASE], and that they have a

[PERSON] feature instead. Assume further that the subject and object forms of the grammatical

functions of pronouns can still be maintained in terms of the feature with which [PERSON] is

matched. Let us also assume with Ouhalla (2005: 682) that that “[t]he subject form [of pronouns]

is the result of matching [PERSON] with the tense feature of T and the object form the result of

matching [PERSON] with the aspectual feature of the verb.”

Given the above set of assumptions, the null subject of imperative sentences which lack a

tense specification would end up not being categorized according to the following reasoning: the

grammatical function of subject in this account can only determined by checking the [PERSON]

feature of the null subject pronoun against the tense feature of T. However, since T in imperative

sentences lack a tense feature, it follows that the null subject pronoun of imperatives would not

be determined. It is worth noting here that imperative sentences also cause a problem for the

claim that it is the tense feature of T that checks the nominative case on the subject (see Raḥḥali

2003 above), since these sentences lack a tense specification, and have instead a mood

specification.

Let us consider a second problem for Ouhalla’s (2005) account. Ouhalla (2005: 681)

claims that there is “[e]vidence that noun phrases seek out the category with a verbal feature

nearest to them […]” He uses the data in (35) from SA to illustrate his point:

57
(35) a. Kaanat Zaynab-u mudarrisat-an.

be.3F.S Zaynab-NOM teacher-F.ACC

‘Zaynab was a teacher.’ (Ouhalla 2005, ex. 38a: 681)

b. Zaynab-u mudarrisat-un.

Zaynab-NOM teacher-F.NOM

‘Zaynab is a teacher.’ (Ouhalla 2005, ex. 38b: 681)

Ouhalla (2005: 681) claims that the nominal predicate in (35a) appears with the accusative case

because the [CASE] feature of the nominal predicate is matched with the aspectual feature of the

verb ‘be’, given that this verbal feature is the structurally closest one to the predicate nominal.

He also claims that in (35b), the predicate nominal appears with the nominative case because

there is no verb in the sentence; therefore, the [CASE] features of the predicate nominal and the

subject are both matched with the tense feature of T. He claims that the verbal tense feature of T

can be targeted twice in (35b) given that it is a verbal feature on T, which needs not be deleted,

as T is a verbal category.

Keeping Ouhalla’s claims in mind, let us now consider sentences with psychological

predicates, where both the subject and the object are assumed to be base-generated in the lexical

domain VP (Chomsky 1995 and subsequent work). This is illustrated in (36):

(36) y-axšaa zayd-un l-ḏalaam-a

58
3-fear.3MSG Zayd-NOM the-darkness-ACC

‘Zayd fears darkness.’

On standard accounts (see Chomsky 1986, Baker 1997), the predicate in (36) is unaccusative, the

subject is base-generated in Spec, VP and the object is base-generated in the complement of VP

position. Ouhalla’s (2005) account would predict that the [CASE] feature of the subject of (36)

would appear as accusative, given that the aspectual feature of Asp would be structurally closer

to both the subject and the object than the tense feature of T. However, this is not borne out by

the facts, as the subject appears with nominative case rather than with accusative case. Notice

that Ouhalla (2005) might object by claiming that the unaccusative-unergative distinction does

not hold in SA. While this might turn out to be true, abandening the idea that psychological verbs

are internal to VP would also mean abandoning the The Universal Theta Role Assignment

Hypothesis UTAH (Baker 1997), an unwelcome result.

The third Agree-based account of case is that of Soltan (2007), who makes the following

claims:

(a) Phi-feature/CLASS agreement is the locus of case licensing. In Soltan (2007: 16-17), case is

not a probing feature on case checking/valuation heads.

(b) There is no evidence of any type of A-movement in SA. Particularly, SA does not show any

evidence for A-movement in typical A-movement structures such as passives and raising

constructions of the seem-type.

59
(c) An Agree-based syntax properly tackles the facts of SA. Agree can establish a relation

between a probe and a goal, the only constraint being minimality considerations such as closet c-

command. This way, agreement and case in SA can be established without either of these being

driven by A-movement.

(d) T in SA has the following inventory of uninterpretable features: (i) ɸ features for person and

number, which may also be realized as default; (ii) the peripheral P-feature (i.e. the EPP feature),

(iii) CLASS feature, which surfaces as a gender feature in many languages. Following Ouhalla

(2003, 2005, as cited in Soltan 2007), Soltan (2007: 69) points out that gender may be able to

probe separately if it is not part of the ɸ-complex.

(e) The well-established asymmetry in agreement between the SV and VS orders are the result of

two distinct syntactic derivations. The example in (37) (based on Soltan 2007:70-71) illustrates

that T in the SV order agrees with a null subject, pro, hence the full agreement on T, and the

preverbal DP is a topic phrase, situated in Spec, TP. The example in (38), on the other hand,

illustrates that T agrees with a phonetically realized DP, and agreement in person and number on

T is therefore default, but Agree does take place between T and the postverbal lexical subject in

CLASS, which is realized in SA as GENDER:

(37) SV order: [CP C [TP DP T EPP/ɸ/CLASS [v*P pro v* [VP V…]]]]

Agree

(38) VS order: [CP C [TP T DEFAULT/CLASS [v*P DP v* [VP V…]]]]

Agree

60
Unfortunately, Soltan (2007) does not address structures with first and second person subjects.

However, given his system, these would fall into the SV order in (37), as agreement in person

and number is realized in these structures. The fact that the EPP feature is available in the SVO

order but not the VSO order seems to indicate that T in Soltan’s system can be lexically specified

differently in the two orders of SA.

(d) Following Uriagereka (2006, as cited in Soltan 2007), Soltan (2007:76) maintains that the left

periphery is a zone rather than a position. Based on this conjecture, Soltan argues that the left

periphery in SA is any position higher than v*P. He therefore maintains that the Spec positions

of the functional projections, TP, NegP, Mod(ality)P, AspP are all left peripheral positions. This

assumption will become relavant when we address some of the problems with this account of

case.

Soltan’s account (2007, 2011) also faces some problems. This account considers that

the preverbal DP is a topic situated in Spec, TP (or Spec, TopicP in Soltan 2011), which, for

Soltan (2007), is an A’-position. Consider first the following problem: Soltan (2007) does not

address what happens when the subject of the sentence happens to be a first or second person.

However, given his system, this type of structures will presumably have a postverbal null subject

pro, which is identified by agreement in ɸ, which is person and number as well as CLASS, i.e.

GENDER which probes separately from ɸ in this system. To conform with the representation

proposed for the SV order, the structure will also have a preverbal null pro, which is a topic in

Spec, TP. This is illustrated by the following examples of structures with first and second person

subjects:

61
(39) a. proi katab-naa/*katab-a/*katab-at proi l-risaalat-a.

wrote-1PLM/*wrote-3MSG/*wrote-3FSG the-letter-ACC

‘We wrote the letter.’

b. proi katab-tunna/*katab-ti proi l-risaalat-a

wrote-2FPL/wrote-3FSG the-letter-ACC

‘You (F.PL) wrote the letter.’

Note that sentences such as (39) cannot have a postverbal lexical subject. When this happens (as

in 40), the sentences are ill-formed; they are only well-formed on a reading where the postverbal

lexical pronominal subjects are interpreted as topics or focalized DPs:

(40) a. *katab-naa naḥnu l-risaalat-a.

wrote-1PLM we.NOM the-letter-ACC

‘We wrote the letter.’ (on a neutral reading of naḥnu ‘we’)

b. *katab-tunna ʔantuna l-risaalat-a12

wrote-3FPL/wrote-3FSG you.FPL.NOM the-letter-ACC

‘You (F.PL) wrote the letter.’ (on a neutral reading of ʔantuna ‘you’)

This means that these structures will have the representation given in (37) and repeated below as

(41):

12
These examples are perfectly grammatical on a focalized reading of the pronouns.
62
(41) SV order: [CP C [TP proi T EPP/ɸ/CLASS [v*P proi v* [VP V…]]]]

Agree

Notice, however, that this creates the following problem for Soltan (2007, 2011). What we have

now are structures with a null preverbal topical pro coindexed with the postverbal null subject.

This raises the following question: what exactly is the discourse function of a null topic? Why

should a topic whose discourse function is presumably to establish an entity or an individual as

the frame of reference for the discourse or to make an entity or an individual salient in the

discourse be phonetically null?

Second, topics are, by definition, optional elements. This makes the following prediction:

if Soltan (2007, 2011) is right in claiming that preverbal DPs are topical elements, then in the

context of ECM strcutures and structures with both lexical and auxiliary verbs, the sentences in

both (a) and (b) should be grammatical, contrary to the fact, as is illustrated in (42) and (43),

where only the examples in (a) are grammatical:

(42) a. ḥasib-a zayd-un [TP l-ʔawlaad-ai y-alʕab-uu-na proi].

believed-3.M.SG Zayd-NOM the-boys-ACC 3-play-MPL-INDIC

‘Zayd believed the boys to be playing.’

b. *ḥasib-a zayd-un [TP proi y-alʕab-uu-na proi].

believed-3.M.SG Zayd-NOM 3-play-MPL-INDIC

‘Zayd believed them to be playing.’

63
(43) a. kaan-at l-banaat-ui y-alʕab-na proi

was-3FSG the-girls-NOM 3-play-FPL

‘The girls were playing.’

b. *kaan-at proi y-alʕab-na proi

was-3FSG 3-play-FPL

‘They (F.PL) were playing.’

Sentences such as (42) and (43) raise the following question: if the DP l-ʔawlaad ‘the boys’ in

(42a) and l-banaat ‘the girls’ in (43a) are topical elements, then why are they obligatory?

Third, one of Soltan’s (2007) major arguments is that structural case is licensed via a

long-distance Agree between a probe and a goal in its c-command domain. Soltan (2007: 70)

claims that “[b]y assumption, then, T can appear with ɸ, CLASS [which, for Soltan 2007, can

probe separately from ɸ, and is realized in many languages as Gender], EPP, or any

combinations of these three, subject to lexical parametrization.” Soltan (2007: 70-71) then claims

that “agreement with a pro subject is only compatible with a full T, necessarily required so pro

can be identified and the derivation converges at the interface.” Keeping these theoretical

assumptions in mind, we can now consider how Soltan (2007) treats the case facts of verbless

sentences. To illustrate this, we can consider the example in (44) (adapted from Soltan 2007, ex.

24a: 55):

(44) Zayd-un fii-l-dar-i

Zayd-NOM in-the-house-GEN

‘Zayd is in the house.’

64
Soltan (2007: 54) treats the DP Zayd ‘Zayd’ as a topic, which receives default nominative case.

Given Soltan’s theoretical assumptions, the subject of (44) must then be a null pro in the Spec,

position of some functional head. This assumption is necessary; without it, the sentence would

end up without a subject. The question that we should ask now is the following: how is the

structural case of the null subject pro licensed? Clearly, the structural case of the null subject is

licensed because the derivation does not crash, and the sentence is fully grammatical. Let us note

that this question is crucial given Soltan’s assumption that agreement with a pro subject is only

compatible with a full T. This means, on Soltan’s assumptions, that T in (44) must have EPP, ɸ

and CLASS features. For Soltan, the topic phrase Zayd ‘Zayd’ in Spec, TP is licensed by the EPP

feature on T. This leaves us with uninterpretable ɸ and CLASS features that have to be checked,

valued and deleted in order for the derivation to converge at LF. Now, the pressing question is

the following: how is pro in (44) licensed given that there is no morphological evidence that T

has checked, valued and deleted its ɸ and CLASS features? To put it differently, what evidence

do we have that T in (44) checked, valued and deleted its uniterpretable ɸ and CLASS features?

The answer is clearly none given that T in (44) does not host any lexical element that can bear ɸ

and CLASS features.

It is crucial to note that the claim that structural case is inherently linked with agreement

(i.e. ɸ-feature checking and valuation) has already been challenged in the literature (see Carstons

2001, Alboiu 2006, 2009).

The fourth Agree-based account of case is that of Al-Balushi (2011), who argues against

Soltan’s (2007) account that ɸ/CLASS-feature agreement is the locus of case licensing in SA and

65
against the view that considers Tense to be the locus of case assignment in SA (as has been

argued to be the case in Benmamoun 1999, Raḥḥali 2003).

Al-Balushi (2011, 2012) makes the following claims:

(a) Structural case licensing in SA is the result of an unvalued verbal case [VC] feature on both

heads, I and v*, which probe for a Goal, and they both receive valuation from a matching but

valued [VC] feature on the head of finiteness, Fin, located in the C domain. As a result of this

checking and valuation, a subject DP in Spec, v*P with a [uCase] feature searches upwards for

the probe I, which has now a valued [VC] feature and checks and values its [uCase] feature as

structural NOM. Similarly, the object DP with a [uCase] feature searches upwards for the probe,

v*, which has now a valued [VC] feature. As a result, the object DP in the complement of V

checks and values its [uCase] feature as structural ACC. In such a system, either element (Probe

or Goal) may c-command the other.

(b) In disagreement with Chomsky (2001), where only a probe with the full set of ɸ-features can

license the [Case] feature on the Goal, Al-Balushi (2011: 36-37) argues that the [Case] features

of subjects and objects in SA are licensed even in the presence of a defective Probe (i.e. a Probe

which does not have the full set of ɸ-features) in the VS order, which is the unmarked order in

the language. The author takes this as evidence against the hypothesis that structural case

licensing is a reflex of ɸ-agreement, contra Schütze (1997), Chomsky (2001 and subsequent

work) and Soltan (2007).

The crucial argument of Al-Balushi (2011, 2012) is that argumental lexical DPs can only

be case-licensed as a result of verbal licensing. Thus, for arguments to be licensed for case, both

T and v must have a [V] feature, which needs to be valued before the latter can license case on

66
argumental DPs. For that to happen, the head, Fin0 must c-select “an XP that has both a

categorial [V] feature and at least one I-finiteness ([T], or [Mood], or [ɸ]” (Al-Balushi 2012: 10).

Given this account, the categorial [V] features of T and v* can only be valued against a valued

[V] feature of Fin0. It is only after such valuation takes place that the functional heads, T and v*

can case-license their respective arguments, namely the subject DP and the object DP

respectively.

There are a number of problems that can be pointed out here. First, Al-Balushi (2011,

2012) uses the idea that argumental DPs can only be case-licensed when verbs are case-licensed

to argue that argumental subject pros in verbless sentences in SA need not be case-licensed. In

other words, given that such sentences lack verbs, it follows, Al-Balushi argues, that the

argumental subject pro in this type of sentences does not need to be case-licensed. Thus, Al-

Balushi argues that the null subject, pro in the example in (45) is licensed and made visible at LF

not by case, but rather by coindexation with the topic phrase, ʔar-rajul ‘the man’.

(45) l-rajul-ui [proi] mariiḍ -un.

the-man-NOM sick-NOM

‘The man is sick.’ (adapted from Al-Balushi 2012: ex.31: 8)

The problem with this argument can be stated as follows: Al-Balushi (2012: 1) states that

he “assumes the Visibility Condition, under which structural Case is necessary to make

arguments visible at LF for θ-role assignment[…][emphasis added]” Yet, in his discussion of

verbless sentences, he argues that the argumental pro does not need to receive case. In other

67
words, if arguments need case to be visible at LF, and if pro is an argument, then surely pro must

also get case to be visible at LF.

Second, in his discussion of verbless sentences, which are c-selected by the emphatic

comlplementizer, ʔinna, Al-Balushi (2012) offers the examples in (46) and states that “[t]hese

sentences show that ʔinna assigns ACC to a following noun, which always has the status of

topic[emphasis added]” (Al-Balushi 2012: 19).

(46) a. ʔinna l-mudarris-ii-n mujtahd-uun.

COMP the-teacher-PL.ACC hard.working-PL.NOM

‘Certainly the teachers are hardworking.’ (Al-Balushi 2012: ex.60: 19)

b. ʔinna l-mudarris-ii-n ya-jtahid-uu-n.

COMP the-teacher-PL.ACC IMP-work.hard-3MPL-INDIC

‘Certainly the teachers work hard.’ (Al-Balushi 2012: ex.61: 19)

As a matter of fact, there are contexts such as the example with the expletive in (47) in

which the noun bearing ACC and following ʔinna can only be interpreted as an argumental

subject DP rather than a topic, contrary to Al-Balushi’s claim.

(47) ʔinna hunaaka rajul-an fii l-daar-i.

that.EMPH there man-ACC in the-house-GEN

‘Indeed, there is a man in the house.’

68
Crucially, the subject in Spec, PP (or Spec, PredP on Al-Balushi’s account), namely, rajul ‘man’

still gets ACC from the complementizer, the latter assigning lexical accusative case to the DP on

Al-Balushi’s account. Notice, though, that such sentences pose a problem for Al-Balushi for the

following reasons. First, the DP, rajul ‘a man’ cannot be treated as a topic, because it has one

reading in which it is non-specific indefinite. Second, the existential expletive, hunaaka ‘there’

cannot be a topic given that it lacks meaning. Given the above reasoning, the only option left on

Al-Balushi’s account is to argue that the DP, rajul ‘a man’ is a subject in Spec, PP, and it

receives lexical ACC from the complementizer. As for the existential locative pro-form, the only

option left on Al-Balushi’s account is to argue that it occupies the Spec, TP position. If such is

the case, the question immediately arises as to how the nonmodified indefinite subject, namely,

rajul ‘a man’ in the Spec, PP position is licensed. Notice that this question is crucial, given that

there are no verbs in the structure to license the subject in Spec, PP; yet, the sentence is

grammatical. On Al-Balushi’s account, the mechanism available for licensing the subject in

Spec, PP has to be coindexation. Notice, however, that this is not possible here given the absence

of a topic phrase that can license the subject via coindexation. On standard accounts (Chomsky

1991: 442-443), the existential expletive in such constructions is an LF affix, and the associate

NP raises at LF to adjoin to the affix, and both form one NP. The only reasonable analysis of

such sentences on Al-Balushi’s account is to assume that the pleonastic expression is a topic

phrase licensed at LF via the abstract feature [Topic] and coindexation is therefore possible

between the pleonastic expression and the lexical DP in Spec, PP. However, pleonastic

expressions cannot be topics because they are not referential.

Third, consider sentences such as (48):

69
(48) fii l-daar-i rajul-un.

in the-house-GEN man-NOM

‘In the house is a man.’ (adapted from Fassi Fehri 1986, ex. 38: 113)

Fassi Fehri (1986: 113) treats the DP rajul ‘man’ in (48) as the subject of the sentence, and the

PP fii l-daar ‘in the house’ as a topic phrase or a focus phrase. Assuming this analysis to be on

the right track, this presents the following problem for Al-Balushi. The DP rajul ‘man’ is

licensed for case without there being a coindexing topic that licenses it at LF. Also, the sentence

lacks a verb with a [VC] feature that can license the subject. In other words, here we have a

sentence in which the subject’s case is liecensed neither by a [VC] feature nor by coindexation,

contrary to Al-Balushi’s (2011) predictions. Notice especially that Al-Balushi (2011, 2012)

might claim that both the PP fii l-daar ‘in the house’ and the DP rajul ‘man’ are topics in his

account, and the subject is thus a null pro coindexed with the topic phrase rajul ‘man’ in (48).

However, such a claim would create the following problem: the PP would have to be base-

generated within the TP domain; otherwise, we end up with a sentence with no predicate. Givn

this scenario, the PP must have moved to its surface position in Spec, TopP from its base-

position within TP. However, such a movement entails that the PP moved across another topic

phrase, namely the DP rajul ‘man’, a movement which is barred on standard accounts.

The fifth Agree-based account of case is that of Leung (2011), who puts forward the

hypothesis that Mood in the C domain is the locus of structural case assignment in SA. The

author notices that there is a correlation in SA between the grammatical mood of the embedded

clause and the choice of the embedded complementizer such that embedded clauses with the

70
indicative mood are associated with the embedded indicative complementizer ʔinna and its

variant ʔanna. He also notices that embedded clauses with the subjunctive mood are associated

with the subjunctive complementizer ʔan. The correlation can be shown by the pair of examples

in (49) and (50):

(49) samiʕ-tu ʔanna/*ʔan Zayd-an

heard-1SG that.INDIC/*that.SUBJ Zayd-ACC

y-uriid-Ø-u/*y-uriid-Ø-a l-safar-a

3-want-MSG-INDIC/*3-want-MSG-SUBJ the-traveling-ACC

‘I heard that Zayd wanted to travel.’

(50) ʔaraad-a Zayd-un ʔan/*ʔanna

wanted-3MSG Zayd-NOM that.SUBJ/*that.INDIC

y-aktub-Ø-a/*y-aktub-Ø-u risaalat-an

3-write-MSG-SUBJ/*3-write-MSG-INDIC letter-ACC

‘Zayd wanted to write a letter.’

Leung (2011: 137-138) interprets this correlation as evidence that there is C-T agreement

between the embedded T and the embedded complementizer in the feature [Mood], which, he

argues to be uninterpretable on the embedded T but interpretable on the embedded C. Agreement

and valuation in Mood, argues Leung (2011), is, therefore, the locus of structural case in SA.

Adopting Chomsky (2000), Leung (2011) argues that C has a set of formal features, which are

transferred from C to T. Mood is one of those features, and structural case is a reflex of C-T

agreement in Mood. This model can be schematized as in (51) (from Leung 2011: 137):

71
(51) Stage one:

[TP DP T1 [VP V [CP C [TP DP T2 [vP [VP V]

[α mood] [ω mood] [0 mood]

[β phi] [β phi] [0 phi] [χ phi] [0 phi]

[γ tns] [0 tns] [ψ tns]

Stage two:

[TP DP T1 [VP V [CP C [TP DP T2 [vP [VP V]

[α mood] [ω mood] [0 mood]

[β phi] [β phi] [χ phi] [χ phi] [0 phi]

[γ tns] [ψ tns] [ψ tns]

Stage three

[TP DP T1 [VP V [CP C [TP DP T2 [vP [VP V]

[α mood] [ω mood] [ω mood]13

13
In Leung (2011), the value of the mood feature on T 2 in Stage 3 is 0. I assume that this is a typo. This assumption
is confirmed by the fact that on the next page (Leung 2011: 138), the author claims that the complete set of valued
formal features on C is transferred to T in Stage 3.
72
[β phi] [β phi] [χ phi] [χ phi] [χ phi]

[γ tns] [ψ tns] [ψ tns]

In (51), Leung (2011: 137-138) claims that the valued mood and ɸ-features of the matrix T are

inaccessible for further computation, as indicated by the strikethrough notation. He claims that

the embdded C has an interpretable mood feature, but uninterpretable ɸ and tense features in

Stage one. He also claims that in Stage two, the uninterpretable ɸ feature on the embedded C is

valued via agreement with the interpretable ɸ feature of the embedded DP, the uninterpretable

tense feature of the embedded C is valued via agreement with the interpretable tense feature of

T2. In Stage 3, Leung claims that the complete valued features of mood and phi are transferred to

T through the machansim of feature inheritance (as in Chomsky 2000).

To show the close relationship between mood on C and accusative vs. nominative

subjects, Leung (2011: 141) provides the schematic structures in (52) and (53):

(52) [TP DP T1 [VP V [CP ʔinna [TP DP T2 [vP [VP V]…

[+ indic] [+ indic] [+ acc] [+ indic]

[β phi] [β phi] [χ phi] [χ phi] [χ phi]

[γ tns] [ψ tns] [ψ tns]

(53) [TP DP T1 [VP V [CP ʔan [TP DP T2 [vP [VP V]…

[+ subj] [+ subj] [+ nom] [+ subj]

[β phi] [β phi] [χ phi] [χ phi] [χ phi]

[γ tns] [ψ tns] [ψ tns]

73
The example in (52) represents embedded sentences that are introduced by the indicative

complementizer ʔinna. According to Leung, the indicative complementizer ʔinna has an

interpretable mood feature with the value [+ indic]. This feature c-commands the embedded

subject in Spec, TP and assigns to it structural accusative case. The valued mood feature of C is

also transferred to the embedded T via feature inheritance. As for (53), Leung claims that the

embedded subjunctive complementizer has an interpretable mood feature with the value [+ subj],

that it assigns the nominative case feature to the embedded subject under c-command. The

valued mood feature is also transferred to the embedded T via feature inheritance.

Having discussed Leung’s (2011) conjecture, there are a number of problems that face

this account.

First, to provide evidence for his account, Leung (2011: 143) offers (74)

(54) samiʕ-tu ʔanna-hu sa-t-usaafir-Ø-u l-bint-u

heard-SG that-it.ACC will-3F-travel-SG-INDIC the-girl -NOM

ġad-an.

tomorrow-ACC

‘I heard that the girl will travel tomorrow.’ (adapted from Leung 2011, ex. 40c: 143)

Leung (2011: 143) claims that (54) provides evidence for his account that C and T have the same

set of formal features. He claims that the indicative complementizer ʔanna in (54) has a strong

and uninterpretable [+D] feature, which attracts the pronominal clitic -hu to it for valuation. In

74
his characterization of the pronominal clitic-hu, Leung (2011, fn. 18: 143) cites Ryding (2005) as

claiming that “[t]he pronoun clitic -hu is argued to function as a generic buffer pronoun that is

independent of the subject of the embedded clause[…]. This is analogous to English ‘that’ which

incorporates a [+D] feature (i.e. it refers to the embedded clause).” Leung does not clarify how

case assignment works in (54). Assuming Leung’s account, we can claim that the indicative

complementizer ʔanna has an interpretable mood feature with the value [+ indic]. This means

that the complementizer assigns the structural accusative to “the generic buffer pronoun that is

independent of the subject of the embedded clause.” This account faces the following problem:

on the assumption that the indicative complementizer has a [+ indic] feature, and the assumption

that T inherits the valued mood feature from C, how do we explain the nominative case on the

the subject l-bint ‘the girl’ in (54)? In other words, given that C in (54) has a [+ indic] mood

feature, this feature is transferred to T. T now has a valued [+ indic] mood feature. T agrees with

the subject l-bint (or with C); yet, instead of the predicted structural accusative case, we have the

unpredicted structural nominative case.

Second, Leung’s (2011) account implies that the subject of root sentences should receive

the structrual accusative case, contrary to fact. To illustrate, let us consider (55):

(55) y-aktub-Ø-u zayd-un/*zayd-an risaalat-an l-ʔaan-a

3-write-MSG-INDIC Zayd-NOM/*Zayd-ACC letter-ACC the-now-ACC

‘Zayd is writing a letter now.’

In (55), T, which hosts the verb, bears the indicative mood marker -u. Adopting Leung’s (2011)

analysis, this should mean that T inherited its mood feature value [+ indic] from the null

75
counterpart of the overt indicative complementizer ʔinna, which bears the interpretable

counterpart of the mood feature. This in turn would predict, on Leung’s (2011) account, that the

subject zayd ‘Zayd’ should receive the structural accusative case, contrary to fact.

Having discussed the major problems with the Agree-based accounts of case, there are

some other problems that can be laid out against more than one of these accounts. First, the view

that the tense feature of T is responsible for nominative case on the subject (Raḥḥali 2003) and

the view that case is a verbal feature on nouns (Ouhalla 2005) both face problems with case in

imperative sentences. Al-Balushi (2011: 46-50) provides convincing evidence that imperative

sentences lack a TP specification in SA. Among the arguments that he uses is the fact that

imperative verb forms do not realize the [±Past] distinction, and this distinction is not observed

semantically given that the only temporal interpretation of imperatives is that of future

orientation. Assuming that this claim is on the right track, the view that Tense is the locus of

structural nominative case is difficult to maintain if imperatives whose null subjects are licensed

for case lack a Tense specification. Note that the null subject of imperative sentences has to be

marked for nominative case rather than accusative case because in SA the subject, in the absence

of an accusative case assigner such as the indicative compelmentizer ʔinna and its variant ʔanna

or the matrix verbal predicate of an ECM construction, always surfaces with the nominative case.

Similarly, the null subject pronoun of imperative sentences would fail to be categorized as a

subject in Ouhalla’s (2005) account if imperative sentences lack a tense specification, the feature

responsible for checking the [PERSON] feature of pronouns in this account.

Second, accounting for case assignment in ECM constructions would be difficult for the

view that tense feature is the locus of nominative case in SA (Raḥḥali 2003), the view that case is

76
a verbal feature (Ouhalla 2005) and the view that Agree in GENDER or ɸ-features (Soltan 2007,

2011) is the locus of case checking in SA. This is because the embedded clause in these

sentences is finite both morphologically (having ɸ-fetures) and semantically (the ability to have

distinct tense specification) (more on this in chapter 5). This means that the embedded T should

be able to assign it nominative case rather than the obligatory accusative case it surfaces with,

contrary to the predictions of Raḥḥali (2003), Ouhalla (2005) and Soltan (2007, 2011). Note that

this problem can be overcome if multiple case checking is allowed. However, none of these

authors, as far as I can tell, allows multiple case checking in their accounts.

Third, structures with event-denoting nominals provide evidence against all Agree-based

account of case. In these structures, the lexical root starts the derivation as a verb before being

nominalized later in the derivation. These structures project a vP, witnesses the fact that they can

bind an anphor, as in (56b), as well as the fact that the structures can be modified by agentive

adverbials, as in (56a-b). However, instead of the predicted structural accusative case, the theme

object inside the DP receives an obligatory genitive case.

(56) a. y-uriid-Ø-u [DP ntiqaad-a PRO l-rajul-i/*l-rajul-a]

3-want-MSG-INDIC criticizing-ACC the-man-GEN/*the-man-ACC

bi-qaswat-in/*l-qaasiy-a

with-bitterness-GEN/the-bitter-ACC

‘He wants to criticize the man with violence.’ (adapted from Fassi Fehri 1993, ex.

65a: 242)

b. y-uriid-Ø-u [DP ntiqaad-a PRO nafs-i-hi/*nafs-a-hu]

3-want-MSG-INDIC criticizing-ACC self-his-GEN/*self-his-ACC

77
bi-qaswat-in/*l-qaasiy-a

with-bitterness-GEN/the-bitter-ACC

‘He wants to criticize himself with violence.’ (adapted from Fassi Fehri 1993, ex.

65a: 242)

If structural accusative case is the result of an Agree relation between the probe v* and the goal,

the object DP l-rajul ‘the man’ in (56a), then an Agree-based account of case cannot explain why

the theme object surfaces with an obligatory genitive case instead of the predicted structural

accusative case. Note especially that the accusative case is realized on the theme object inside the

DP when the subject in the structure is a lexical DP l-rajul ‘the man’ rather than PRO. This is

illustrated in (57):

(57) ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i l-mašruuʔ-a]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

‘The man criticizing the project annoyed me.’ (Fassi Fehri 1993, ex. 46: 234)

Having discussed the Agree-based accounts of case in SA, I conlude that these accounts

face problems and that an alternative account is thus needed.

78
3.4.2 A mixed approach to case (Fassi Fehri 1993)

Fassi Fehri’s (1993) account of case is the only account of case in SA which incorporates

a case hierarchy where accusative case is dependent on another higher case (specifically

nominative case in the CP domain and genitive case in the DP domain). Fassi Fehri’s (1993)

account is in some respects similar to the account that will be adopted in this dissertation. Below,

I outline the major tenets of this approach to case.

(a) To account for the agreement asymmetry between the VSO order and the SVO order in SA,

Fassi Fehri (1993) advances the functional ambiguity hypothesis, according to which all subject

agreement affixes in SA are functionally ambiguous in that they can be either be pronouns

incorporated into the verb or they can be agreement affixes. To illustrate, let us consider the

examples in (58):

(58) a. jaaʔ-a l-ʔawlaad-u

came-3MSG the-boys-NOM

‘The boys came.’

b. l-ʔawlaad-u jaaʔ-uu

the-boys-NOM came-3MPL

‘The boys came./The boys, they came.’ (adapted from Fassi Fehri 1993, ex. 29:

27)

According to Fassi Fehri (1993), the verbal affix in (58a) can only be treated as an agreement

affix; it cannot be treated as a subject incorporated pronoun; otherwise, the sentence ends up

79
having two subjects, which is impossible. As for (58b), Fassi Fehri (1993: 28) holds that the

sentence has two interpretations, as is indicated by the English translations. In one reading, the

DP ʔal-ʔawlaad ‘the boys’ is interpreted as a clitic left-dislocated element (i.e. a base-generated

topic phrase), and the verbal affix is thus a subject pronoun incorporated into the verb. In another

reading, the preverbal DP is interpreted as the real subject of the sentence. In this case, the verbal

affix can only be treated as an agreement affix. When the subject is postverbal, as in (58a), Fassi

Fehri argues that the nominative case of the subject is assigned by a governing T. In the second

reading of (58b), where the preverbal DP is a subject, Fassi Fehri (1993: 33, 45) argues that the

case of the preverbal DP is default. For Fassi Fehri (1993), the nominative case of the preverbal

DP is default regardless of whether the DP is a subject or a clitic left-dislocated DP.14

(b) According to Fassi Fehri (1993), the accusative case on the object DP is assigned by V under

government. However, there are cases where the accusative case fails to be assigned to the DP

object, as can be seen in structures with process (= event-denoting) nominals when they are c-

selected by control verbs. This is illustrated by the contrast in (59):

(59) a. ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i l-mašruuʔ-a]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

‘The man’s criticizing the project annoyed me.’ (adapted from Fassi Fehri 1993,

ex. 60a: 239)

b. y-uriid-Ø-u [DP ntiqaad-a PRO l-rajul-i/*l-rajul-a]

3-want-MSG-INDIC criticizing-ACC the-man-GEN/*the-man-ACC

14
Arsalan Kahnemuyipour asks the following question: what accounts for the difference in agreement between the
second reading of (58b) and the VS order in (58a)? According to Fassi Fehri (1993: 34), this is accounted for by his
AGR Criterion, which is stated as follows: “Rich AGR is licensed by an argument NP in its Spec, and an argumental
NP in Spec AGR is licensed by rich AGR”
80
bi-qaswat-in

with-bitterness-GEN

‘He wants to criticize the man with violence.’ (adapted from Fassi Fehri 1993, ex.

65a: 242)

Fassi Fehri (1993) shows that process nominals in SA are internally verbal but externally

nominal, and that the categorial conversion takes place at some point in the syntactic derivation.

Some of the arguments that Fassi Fehri uses to show that process nominals are internally verbal

include the fact that they have the same argument structure of verbs, and their thematic objects

are assigned the accusative case, as can be seen in (59a). Another argument for the internally

categorial status of process nominals is that they can be modified by adverbial phrases, as is

shown in (59b). To account for why the thematic object of the control clause fails to be assigned

the accusative case in (77b), Fassi Fehri (1993: 242) proposes that PRO is caseless, and that

objects can only be assigned the accusative case when their subjects are assigned case. This is

regulated by his propsosed condition on case discharge, which is given in (60):

(60) “Object Case is discharged only if subject Case is discharged” (Fassi Fehri 1993: 243).

(c) The nominative case assigned to subjects of verbless sentences and subjects of sentences of

the SVO order in SA is a default case, according to Fassi Fehri (1993). In his account, the default

case of the subject of these types of sentences surfaces unless there are external case assigners

such as the the complementizer ʔinna/ʔanna ‘that’, the matrix verb in ECM structures, or D

inside DPs. This is illustrated by the contrasting examples in (61) and (62):

81
(61) a. zayd-un mariiḍ-un

Zayd-NOM sick-NOM

‘Zayd is sick.’

b. ʔinna zayd-an/*un mariiḍ-un

that.EMPH Zayd-ACC/*NOM sick-NOM

‘Zayd is indeed sick.’

c. ḥasib-tu zayd-an/*un mariiḍ-an

believed-1SG Zayd-ACC/*NOM sick-ACC

‘I believed Zayd to be sick.’

d. ʔaqlaqa-nii [DP kawn-u zayd-in/*un mariiḍ-an]

annoyed-me being-NOM Zayd-GEN/*NOM sick-ACC

‘It annoyed me to hear of Zayd being sick.’

(62) a. l-ʔawlaad-u jaaʔ-uu

the-boys-NOM came-3MPL

‘The boys came.’

b. ʔinna l-ʔawlaad-a/*-u jaaʔ-uu

that.EMPH the-boys-ACC/*NOM came-3MPL

‘Indeed, the boys came.’

c. ḥasib-tu l-ʔawlaad-a/*-u jaaʔ-uu

believed-1SG the-boys-ACC/*NOM came-3MPL

‘I believed that the boys came.’

82
This contrasts with sentences of the VSO order, where only the nominative case is allowed, as

can be seen in (63):

(63) jaaʔ-a l-ʔawlaad-u/*-a

came-3MSG the-boys-NOM/*ACC

‘The boys came.’

To account for the obligatory nominative case in the VSO order and the availability of non-

nominative case in the nominal sentences and sentences of the SVO order, Fassi Fehri (1993: 50)

proposes the nominality parameter in (64):

(64) I is ± nominal

Fassi Ferhi (1993: 50) claims that the nominality parameter can be translated as follows: I

includes T, which is verbal and AGR, which can or cannot be nominalized, and either one or the

other will be favored as the dominant category, which imposes its nature on I. According to this

parameter, AGR in the VSO order is strong in that it is verbal and cannot be nominalized.

Therefore, the nominative case of the postverbal subject is assigned under government by I,

which hosts AGR and T. In other words, I in the VSO order is verbal; therefore, it is a case

assigner, which assigns its case to the postverbal subject under government. In contrast to this,

AGR in the SVO order is weak; therefore, it can be nominalized by an external case assigner

(such as the indicative complementizer ʔinna or the ECM predicate or D), which assigns its case

83
to AGR. In other words, I in the SVO order is nominal; therefore, it can receive case from

external case assigners. According to Fassi Fehri (1993: 50). AGR can discharge the case of the

external case assigner because it is nominalized in these types of sentences.

Although the account proposed in Fassi Fehri (1993) captures all of the case facts in SA,

there are some theoretical problems that can be pointed out. First, this account rests on

parameterizing I in one and the same language, namely SA, and while parameterizing I might be

justifiable cross-linguistically, it is hard to see how I can be parameterized intralinguistically. If

one were to choose between this account, and one that does away with this complication, then

the second account would be preferred for simplicity reasons, if for nothing else. To put it

differently, in the spirit of the Minimalist Program (Chomsky 1995), any account that requires

fewer assumptions should, by hypothesis, be favored over another that requires more

assumptions.

Second, Fassi Fehri (1993) claims that the accusative case of the object is only

discharged when the case of the subject is discharged. This is regulated by his proposed

condition on case discharge, which is repeated in (65) for convenience:

(65) “Object Case is discharged only if subject Case is discharged” (Fassi Fehri 1993: 243).

The question now arises as to what the status of the condition in (65) is? Specifically, can

the condition in (65) be derived from any theories of case? Fassi Fehri (1993, fn. 31: 278) claims

that the condition in (65) is analogous to the case tier approach of Yip, Maling and Jackendoff

(1987). However, in Yip et al.’s (1987) theory, (i) cases in a case tier (NOM, ACC) are mapped

84
onto grammatical functions (GFs=subject, object) in a separate tier by principles of association,

(ii) the notion of domain of case assignment is central, (iii) PRO receives case, and (iv) sentences

(S), but not v*Ps, supply a case tier. On the other hand, in Fassi Fehri’s (1993) account, (i)

condition (65) neither maps cases onto GFs nor refers to domains of case assignment, (ii) PRO

doesn’t receive case, and (iii) the account implies that the case hierarchy applies at VP (v*P in

current theories) as well as S.

Third, in Fassi Fehri’s (1993) account, the AGR feature in T is responsible for assigning

the structural nominative case to the postverbal subject in the VSO order, and V (v in current

theory) assigns accusative to the object. However, in the context of process nominals in control

structures, the assignment of the accusative case to the object is dependent on the assignment of

case to the subject, as is regulated by Fassi Fehri’s (1993) condition on case discharge. It is

worth pointing out that the account does not offer any explanation as to why the accusative case,

which is the result of the syntactic relationship of government between V (= v) and the object

should in any way be dependent on a hierarchy of cases between the subject and the object, a

hierarchy that has nothing to do with the syntactic relation of government. To state the problem

differently, the account derives the state of inactivity that V (= v) experiences in these structures

from a hierarchy of cases rather than from a reassessment of the nature of V(= v) itself.

To summarize, Fassi Fehri (1993), makes the following claims:

(66) a. The postverbal subject receives a structural obligatory nominative case via agreement

with AGR in T.

b. The preverbal subject and the subject of verbless sentences receive an unmarked

default nominative case unless there are external case assigners such as the indicative

complementizer, the ECM predicate or D.

85
c. The object receives a structural accusative case from V (v in current theory) as long as

the subject receives a case.

d. The first NP complement in DP receives a structural genitive case from D.

While the above claims capture the empirical facts properly, they are not explanatory

adequate. I take these arguments as evidence that Fassi Fehri’s (1993) account needs to be

reframed within an alternative theory of case.

3.5 The proposed account

To address the above problems with the previous analyses of case facts in SA, I adopt a

new analysis, which is based on the updated version of the dependent case theory (Baker 2015).

This analysis has the merit of incorporating both Agree-based accounts of case and the

dependent case theory in one theoretical model. The major claim of this analysis is the

hypothesis that Agree-based case is bled by dependent case assignment, when the latter is

applicable in a language. For the analysis to work, there have to be at least two NPs in the spell

out domain of a phase (Baker 2015). Thus, syntactic structural accusative case in this theory is a

dependent case, which is assigned when the NP is c-commanded by another NP in either the

same phasal domain or in a higher phasal domain depending on whether the lower phase is a soft

phase or a hard phase. Broadly speaking, there are two major phases in the updated version of

the dependent case theory (Baker 2015). The first such phase is vP. When the head of this phase

is merged into the structure, its spell out domain, namely VP is sent to Spell-Out. The second

major phase is CP. When the head of this phase C is merged into structure, its spell out domain

TP is sent to Spell-Out. According to this theory, if the lower phase is a soft phase, then the NPs

86
inside this phase are still accessible to the higher phase. If, on the other hand, the lower phase is a

hard phase, then the NPs inside this phase are inaccessible to the higher phase. Suppose, for

example that vP is a phase, as is standardly taken to be the case in agreement-based theories of

case (Chomsky 2000 and subsequent work). Suppose further that there are two NPs inside the

VP, which is the spell out domain of this phase. If vP in such a language is a hard phase, then the

case of these NPs have to be both calculated inside this domain. In such a language, the higher

NP inside the VP domain c-commands the lower NP inside the same domain; therefore, the

lower NP receives the dependent accusative case, and the higher NP receives its case through the

case-assigning Agree relation with the higher functional head. The other case, which competes

with the dependent case would thus be nominative case in a nominative-accusative language.

This would be the result of the middle NP Agreeing with T. In an ergative-absolutive case

system, the dependent case would be the higher ergative case, and the other case would be the

absolutive case, which would be the result of Agreeing with the functional head v.15 Suppose

now that in a nominative-accusative language, vP is a soft phase. Suppose further that there are

two NPs inside the phasal domain of v, which is VP. In such a language, there is one of two

options. In one option, the language exhibits a “strict cycle” effect such that the dependent case

cannot be assigned in the VP domain. This is the case in languages such as Japanese and Korean,

as will be shown later in the thesis. In such a language, the lower NP cannot be assigned the

dependent case even though it is c-commanded by another NP in the same Spell out domain. In

these languages, the assignment of dependent case can only happen at the Spell Out of TP, which

is the Spell out domain of the phasal head C. This means that when TP is sent to Spell-Out in

15
This is the type of languages, where the subject of intransitves and the object of transitives would pattern together.
This contrasts with the nominative-accusative case system, where the subject of transitives and the subject of
intransitives would pattern together.
87
these languages, only the higher NP in the matrix clause will get the Agree-based case, but all

other NPs in the structure will get the dependent accusative case. The other option that languages

of the nominative-accusative type have when vP is a soft phase is when the language does not

exhibit a “strict cycle” effect such that the dependent case is assigned to the lower of two NPs in

any phasal domain be it VP or TP. This is the case in Amharic, as is argued in Baker (2015), and

this is also the case with SA, as will be argued throughout this thesis. A full exposition of the

proposed account will be introduced in Chapter 4.

3.6 Summary

This chapter reviews the previous generative accounts of case. The chapter discusses

Agree-based accounts of case in SA, and Fassi Fehri’s mixed account (1993), where a mixed

approach is developed. The chapter argues that all of the above accounts of case are problematic

either theoretically or empirically. The chapter sketches out an alternative account, namely the

dependent case theory of Baker (2015).

88
Chapter Four

Alternate accounts to case with a focus on Baker’s dependent case and relavant theoretical

assumptions

4. 1 Introduction

In this chapter, alternative generative of accounts of case to the Chomskian accounts

reviewed previously are introduced. The case tier theory of Yip, Maling and Jackendoff (1987)

and the original version of the dependent case (Marantz 1991) are discussed together with the

problems that they face. This is then followed by the updated version of the dependent case

theory, as developed in Baker (2015).

4.2 Alternative accounts of case: The Case Tier theory

Yip, Maling and Jackendoff (1987) propose a theory of case assignment in syntax which

is analogous to the autosegmenatal theory in phonology and morphology. They further propose

that NPs form an NP tier and cases form a distinct case tier, and each case is associated with an

NP via association lines. They propose that the unmarked association is from left (L) to right (R).

This is the case of nominative-accusative languages. They argue that some languages are marked

in that the association takes place from R to L. These are the ergative-absolutive languages. In

this theory, crossing association lines is not allowed. To illustrate how this theory works, let us

consider the following simple transitive sentence from English:

(1) John hit Mary (the NP tier)

NOM ACC (the case tier) (adapted from Yip et. al. 1987, ex. 1: 219)

89
In (1), the NPs form an NP tier and the cases form a case tier, and the cases are mapped onto the

NPs in a L-to-R association. An example of a simple intransitive sentence is given in (2):

(2) John paused.

NOM ACC (adapted from Yip et. al. 1987, ex. 2: 219)

The NP John in (2) is associated with the nominative case, and the accusative case remains

unassociated, as there is no NP to associate with. As a result, the accusative case is not realized.

Yip et. al. (1987: 221) argue that in double object constructions, some languages have a

rule of spreading case across two sister complements. This is the case, they argue, in Swedish, as

is illustrated by (3):

(3) Kunggen gav honom(*hon) henne(*hon) (till maka).

the.king gave him(*he) her(*she) (as wife)

NP NP NP

NOM ACC

‘The king gave her to him as a wife.’ (adapted from Yip et. al. 1987, ex.3: 221)

The example in (3) shows, the authors argue, that Swedish has a spreading rule, which spreads

the accusative case on the goal object to its sister complement theme object. The authors point
90
out that the case on the second object cannot be a default case, as nominative but not accusative

is the default case in Swedish.

In structures with lexical case, the authors argue that this case is a special case lexically

assigned by some verbs to one of their arguments. This is illustrated with the following example

from Icelandic:

DAT

(4) Siggi leyndi konuna sannlikanum.

Siggi concealed the.woman the.truth

NP NP NP

NOM ACC

‘Siggi concealed the truth from the woman.’ (adapted from Yip et. al. 1987, ex. 8a: 223)

In (4), the theme object receives a lexical dative case from the verb, which is annotated as a case

above the NP rather than below it. The syntactic accusative caseis associated with konuna ‘the

woman’, and the nominative case is associated with Siggi. Let us consider the following

example, where the goal object receives a lexical dative case from the verb:

91
DAT

(5) Siggi sagði barninu söguna.

Siggi told the.child the.story

NP NP NP

NOM ACC

‘Siggi told the child the story.’ (adapted from Yip et. al. 1987, ex. 8b: 223)

In (5), the syntactic nominative case is associated with the subject, and the syntactic accusative

case is associated with the theme object. As for the goal object, it receives a lexical case from the

verb. Similar to objects, the authors argue that subjects in Icelandic may also receive lexical case,

as is illustrated by (6):

DAT

(6) Barninu batnaði veikin.

the.child recovered.from the.dosease

NP NP

NOM ACC

In (6), the subject is assigned a lexical dative case by the verb, and the syntactic nominative case

is assigned to the object. As for the syntactic accusative case, it remains unassociated, and is

therefore, not realized.

92
Yip et. al. (1987) also propose that not only sentences provide a case tier but NPs do too.

In their discussion of Icelandic, they propose that it has one case to be associated with a bare NP

modifier, namely the syntactic genitive case. This case is assigned to the first bare NP

complement; all other bare NP complements inside the NP domain must therefore get their case

from prepositions; otherwise, the structure is ill-formed. The authors argue that the gentive case

inside the NP is assigned regardless of what thematic roles the gentive NP gets. The following

are some illustrative examples adapted from Yip et. al. (1987, ex. 15: 234):

(7) a. Active sentence:

Konan þýddi bókina.

the.woman translated the.book

NP NP

NOM ACC

Case tier in S(entence)

93
b. NP with prenominal genitive:

Jóns þýðing á bókinni

J. translation of the.book

NP PP

GEN

Case tier in NP

c. NP with postnominal genitive ‘subject’:

þýðing Jóns á bókinni

translation J. of the.book

NP PP

GEN

Case tier in NP:

94
d. NP with postnominal genitive ‘object’:

þýðing bókarinnar

translation the.book

NP

GEN

Case tier in NP

e. Ill-formed sequences of two bare NPs:

* Jóns(GEN) þýðing bókarinnar(GEN)

* þýðing Jóns(GEN) bókarinnar(GEN)

* þýðing Jóns(GEN) bókina(ACC)

Discussing the data in (7), the authors show the following: in (7a), there is an active sentence

with two NPs in the NP tier and two syntactic cases in the case tier, and each case is associated

with one NP in a one-to-one L-to-R association. This is an example of case assignment in the S

domain. All other examples in (7) are examples of case assignment in the NP domain. In (7b),

the head of the NP is the deverbal noun of the transitive verb for translate, namely þýðing

‘translation’. Here, the genitive case assigned in the NP domain goes to the first NP complement

of the head, namely the prenominal possessor Jóns ‘John’. In (7c), the genitive case is assigned

to the first NP complement, anmely the subject Jóns ‘John’. The other NP complement receives

95
its case from the preposition. In (7d), the genitive case is assigned to the first NP complement,

namely the object bókarinnar ‘the book’. In (7e), all the examples are ill-formed because they

involve assigning case to the second NP complement without a preposition.

To handle sentences with PRO subjects, Yip et. al. (1987: 239) propose that case

assignment applies strictly to items in a domain, where the domain is either an S or an NP. They

offer the following definition of domain:

(8) a. “A node Y is IN THE CASE DOMAIN of a node X iff Y is dominated by X, and

X supplies a case tier C.

b. A node Y is IN THE CASE DOMAIN of a node X iff Y is in the case domain of

X, and there is no node Z such that X dominate Z and Y is in the case domain of

Z.”

Given the definition of domain in (8), a syntactic case in a matrix clause (S2) cannot be

associated with an NP in the embedded clause (S1), as case assignment is strictly domain-

specific. Similarly, a syntactic case in S2 cannot be associated with any NP arguments inside an

NP, as NPs are themselves case domains. Having identified case domains, the authors offer their

proposal for how case assignment works in sentences with PRO subjects. This is illustrated in (9)

adapted from Yip et. al. (1987, ex. 21: 239):

96
(9) a. [S2 Bill tried [S1 PRO to help me]].

NP NP NP

S1 tier NOM ACC

S2 tier NOM ACC

According to the authors, the sentence in (9) can be analyzed as follows: In the S1 tier, there are

two NPs, PRO and the object NP, and two syntactic cases. Each syntactic case is assigned to

each NP in the NP tier in a one-to-one L-to-R association. In the S2 tier, there is one NP in the

NP tier but two syntactic cases in the case tier. The first syntactic case NOM is associated with

the only NP in the NP tier. As for the second syntactic ACC case, it remains unassociated, as

there is no other NP in the S tier to associate with. Crucially for the authors, the syntactic

accusative case in the S2 tier does not associate with any NP in the S1 tier, as case assignment is

strictly domain-specific. The same reasoning, the authors argue, applies to sentences where PRO

is controlled by the object of the matrix clause, as in (9b) adapated from Yip et. al. (1987, ex. 21:

239):

(9) b. Bill forced him [PRO to help me]].

NP NP NP NP

NOM ACC

NOM ACC

97
Note that it is crucial for the authors to claim that PRO receives case. On the assumption that

PRO does not get case, their theory would make the wrong predictions, as is illustrated by the

example in (10) adapted from Yip et. al. (1987, ex. 20: 238):

(10) [S2 Bill tried [S1 PRO to help I/me]].

NP NP NP

S1 tier NOM ACC

S2 tier NOM ACC

‘*Bill tried to help I.’

In (10), PRO of S1 does not receive a syntactic case; therefore, the syntactic case NOM is

associated with the object, and the syntactic case ACC of S1 remains unassociated; the result is

ungrammatical. Therefore, PRO must receive a syntactic case for the theory to make the right

predictions.

To account for case assignment in rasing-to-object constructions or what is known in the

GB literature as exceptional case marking (ECM) structures, Yip et. el. consider two

possibilities. According to the first possibility, the embedded subject of believe-type predicates

raises to the matrix object position. In this scenario, the authors offer the following analysis:

98
(11) a. Bill believes [she/her to like me]

NOM ACC NOM ACC

b. Bill believes heri [ti to like me]

NOM ACC

NOM ACC (Yip et.al. 1987, ex. 22: 241)

‘John believes her to like me.’

In (11), Yip et.al. (1987: 241) argue that the subject of the embedded clause is associated with

the nominative case in the complement cycle (i.e. the embedded clause). Upon raising to the

matrix object position, the subject of the embedded clause is now in the case domain of the

matrix cycle (i.e. the matrix clause) where the previously assigned nominative case is now

overlaid with the accusative case which is associated with the object in the matrix cycle.

The second possibility that the authors consider is one where the subject of the embedded

clause does not raise to the matrix clause. They offer the following example from Icelandic as in

illustration:

(12) Jón telur [hana hata mig]

John believes her to.hate me

NOM ACC

NOM ACC

‘John believes her to hate me.’

99
In (12), Yip et.al. (1987: 241-242) assume that the subject of the embedded clause does not raise

to the matrix object position. In this scenario, the embedded subject is associated with the

nominative case in the complement cycle. Due to the mechanism of S’-deletion, the subject of

the complement cycle is now in the case domain of the matrix cycle. The nominative case of the

embedded cycle is now overlaid with the accusative case of the matrix cycle.

To summarize, the Case Tier theory posits the hypothesis that the syntactic cases, NOM

and ACC, which are often called structural cases, are assigned neither to certain syntactic

configurations, nor to certain grammatical functions. Instead, they are assigned along a hierarchy

of grammatical functions, where NOM is higher on the hierarchy than ACC (Maling 2009: 87).

The theory makes the following predictions:

(13) “a. NOM is assigned before ACC.

b. Only one XP can get assigned NOM; any remaining NPs get ACC via case-spreading.

c. Which XP gets NOM reflects the hierarchy of G[rammatical]F[unction]s, where

SUBJ[ect] > OBJ[ect]…” (Maling 2009: 82).

4.2.1 Problems with the Case Tier theory

As discussed in the previous section, there are two case domains in the Case Tier theory,

S and NP. Given this theoretical assumption, there are structures in SA that pose a problem for

the Case Tier theory. One such structure includes sentences with process nominals, as can be

seen by the example in (14):

100
(14) ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i l-mašruuʔ-a bi-šiddat-in]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC by-bitterness-GEN

‘The man’s criticizing the project bitterly annoyed me.’ (adapted from Fassi Fehri, ex.

60a: 239)

In (14), the lexical root starts the derivation as a verb and is then nominalized later on in the

derivation. According to the Case Tier theory, this would mean that there are two NPs in the NP

tier and two cases in the S case tier. This in turn means that the underlying structure of the

sentence in (14) can be represented as in (15):

(15) [DP ntiqaad-u l-rajul-i l-mašruuʔ-a bi-šiddat-in] ʔaqlaqa-nii

criticizing-NOM the-man-GEN the-project-ACC by-bitterness-GEN annoyed-me

NP NP

NOM ACC

In (15), there are two NPs in the NP tier and two syntactic cases in the case tier, and each case is

associated with each NP one-to-one in a L-to-R fashion. Assuming that the object pronominal

DP (n)ii ‘me’ is cliticized onto the verb, and that the verb then raises from v to T, as is standardly

assumed for SA, would account for the surface structure shown in (14) above. So far, the Case

101
Tier theory can account for the case facts of SA. The problem starts when we want to explain the

accusative case marking on the thematic object l-mašruuʔ ‘the project’ inside the first NP. From

the perspective of the Case Tier theory, the NP is also a case domain. The only case available

inside the NP domain in this theory is the genitive case. Suppose that the genitive case is indeed

available and that it is associated with the first bare NP complement of the nominalized V, as is

shown in (16):

(16) [DP ntiqaad-u l-rajul-i l-mašruuʔ-a bi-šiddat-in]

criticizing-NOM the-man-GEN the-project-ACC by-bitterness-GEN

NP NP

GEN

If we now assume that the language-specific case spreading rule, proposed for some languages in

Yip et. al. (1987) is operative in SA, we end up making the wrong predictions, as is shown in

(17):

(17) *[DP ntiqaad-u l-rajul-i l-mašruuʔ-i bi-šiddat-in]

criticizing-NOM the-man-GEN the-project-GEN by-bitterness-GEN

NP NP

GEN

‘the man’s criticizing the project bitterly’

102
The DP in (17) is ungrammatical because the NP l-mašruuʔ ‘the project’ obligatorily surfaces

with the accusative case rather than the genitive case. One might want to argue that the syntactic

genitive case is associated in (17) with the first bare NP complement l-rajul ‘the man’, and that

the second NP l-mašruuʔ ‘the project’ receives its accusative case as a default/unmarked case.

However, this argument is untenable; given that the default/unmarked case in SA is standardly

taken to be the nominative case. Also, if accusative case were the default case inside the NP

domain, one would expect the phrase in (18) to be grammatical, contrary to the facts:

(18) *[DP ṣuurat-u zayd-in qays-an]

picture-NOM Zayd-GEN Qays-ACC

Intended meaning, ‘Zayd’s picture of Qays.’

The only way for (18) to be grammatical is when the second NP inside the DP domain is

assigned the genitive case via the prepositional strategy, as in (19):

(19) *[DP ṣuurat-u zayd-in li-qays-in]

picture-NOM Zayd-GEN of-Qays-GEN

‘Zayd’s picture of Qays.’

Notice further that the Case Tier theory cannot explain why the theme object inside the

nominalized DP of (14) may surfaces with the accusative case simply because there is only a vP

inside the DP domain, and vPs do not supply a case tier in Yip et. al.’s (1987) theory. Thus, the

Case Tier theory can not account for the accusative case of the theme object inside the DP, when

the head of this DP is a nominalized verb.

103
Let us now consider a similar problem for the Case Tier theory. This has to do with

control predicates that take as their complemnts a nominalized verb, as in (20):

(20) y-uriid-Ø-u proi [DP ntiqaad-a PROi l-rajul-i/*l-rajul-a]

3-want-MSG-INDIC criticizing-ACC the-man-GEN/the-man-ACC

bi-šiddat-in

by-bitterness-GEN

‘He wants to bitterly criticize the man.’ (adapted from Fassi Fehri 1993, ex. 65a: 242)

The example in (20) is problematic for the Case Tier theory in the following manner: In this

theory, PRO must receive a syntactic case. Suppose that PRO in (20) does receive the only

syntactic case available inside the NP domain, namely the genitive case, as proposed in the Case

Tier theory. If this is the case, then case assignment inside the DP in (20) would be as shown in

(21):

(21) [DP ntiqaad-a PROi l-rajul-i/*l-rajul-a]

criticizing-ACC the-man-GEN/the-man-ACC

NP NP

GEN

In (21), we have two bare NP complements inside the DP domain, but one syntactic case. The

problem that (21) poses for the Case Tier theory can be stated as follows: if the only syntactic

104
case inside the DP domain is associated with the first NP complement, namely PRO, how does

the second NP l-rajul ‘the man’ inside the DP domain receive its genitive case, especially given

our previous conclusion that case spreading is not operative in SA?

Given the above problems with the Case Tier theory, I conclude that this theory does not

make the right predictions for SA.

4. 3 The proposed account of case for SA

In this thesis, I develop an account of case in SA based on the dependent case approach

as developed in Baker (2015). Baker’s (2015) theory of case assignment is a development of

Marantz’ (1991) original theory of dependent case. I begin by introducing the original version of

dependent case theory as first proposed in Marantz (1991). I point out some of the problems that

face this version of the theory. I follow that by introducing Baker’s (2015) developed version of

the dependent case theory, the version that forms the theoretical framework of the proposed

account of case in SA. Following that, I lay out the technical details and theoretical assumptions

that form the basis of the analysis to be developed in chapter 5. I conclude this chapter with a

summary.

4.3.1 The original version of dependent case (Marantz 1991)

In response to abstract case theory as developed in Chomsky (1980 and subsequent

work), Marantz (1991) claims that NP licensing is distinct from the morphological assignment of

case on NPs. In this work, Marantz uses the examples in (22) from Icelandic to show that NPs
105
can get morphological case without being licensed by abstract case. Recall that this distinction is

the same distinction discussed earlier between syntactic structural cases and syntactic but non-

structural cases.

(22) a. María óskaði (ólafi) alls goðs.

Mary-NOM wished Olaf-DAT everything-GEN good-GEN

b. þess vas óskað.

this-GEN was wished

c. Henni vas óskað þess.

her-DAT was wished this-GEN ( Marantz 1991, ex. 15: 18)

Marantz (1991) reasons as follows: in (22a) both objects receive a quirky morphological case

(i.e. a case lexically specified by certain verbs), which is not the expected structural accusative

case for objects; yet, the NPs are not licensed as objects because they get these cases. This is

because these cases are not structural cases (i.e. cases determined by the structural relationship

holding between a V or v and the object in the complement position of the verb). When the verb

is passivized, as in (22b), the object has to raise to the subject position, but its case does not

change. On the face of it, this means that the object is not licensed for structural case in the

object position. Moreover, (22c) makes it clear that the movement of the object in (22b) is driven

by the need to check the Extended Projection Principle (EPP) feature rather than to get licensed

by case. (22c) shows that when the subject position is occupied by a dative NP, the genitive

object is now licensed as an object even though it is not licensed by structural accusative case. If

one wants to claim that the genitive object in (22b) has to move because it lacks structural case

106
as the object of a passive verb, then one has to explain how the object in (22c) is licensed in the

base position of a passive sentence even though it does not receive the structural accusative case.

Marantz (1991) also gives the Icelandic example in (23) to show that NPs can be licensed

as objects without having structural case:

(23) Ég tel henna hafa alltaf þótt ólafur leiðinlegur.

I believe her-DAT to-have always thought Olaf-NOM boring-NOM

‘I believe her to have always thought of Olaf as boring.’

(Marantz 1991, ex. 16: 19)

For the example in (23), Marantz (1991) provides the following account: the embedded sentence

in (23) has a non-structural dative subject and a structural nominative object in the complement

clause of believe. T cannot possibly license the object with structural nominative case, as the

complement clause is non-finite lacking the feature [Tense]. The object is however licensed in

the object position even though it does not have a structural accusative case.

As an alternative to abstract structural case theory, Marantz (1991: 20-21) offers the

principle in (24):

(24) “Nominal arguments are licensed by (extended) projection, not by Case or by

morphological properties.”

Marantz (1991) acknowledges the fact that the principle in (24) does not cover the

distribution of PRO. He also notes that the principle in (24) does not apply to the distribution of

pro, given that the latter is licensed by the morphological properties of agreement systems. To

account for the distribution of PRO, Marantz (1991) offers the residue of case theory in (25):
107
(25) “RES(Case Theory): an NP is PRO iff not governed at S-structure by a lexical item or

[+tense] INFL[ection]”.

Marantz (1991: 19) assumes a model of grammar without case theory. The model is

represented in (26):

(26) Projection

DS

SS

MS LF

PF

MS = “Morphological Structure”

Given this model of the grammar, Marantz (1991: 23-25) explains how case is realized in the

following manner: case realization takes place in the MS component, which maps SS (i.e.

Surface Structure) to PF (i.e. Phonological/Phonetic Form). At MS, the case affix in languages

that have morphological cases, is attached to the noun affix. Which case feature the case affix

requires (e.g. nominative, accusative, genitive, dative, etc.) depends on the syntactic properties of

the element, which governs the maximal projection of the noun to which the case affix

108
attaches.16 The relevant objects at MS are not NPs per se but rather chains, in particular A-chains

(argument chains), which include the traces of NP-movement. Marantz (1991: 23) states that

“the case features on the case affix may depend on what governs any link in the chain of the NP

headed by the N+Case.” This is covered by the principle in (27):

(27) “CASE features are assigned/realized based on what governs the chain of the NP headed by

N+Case” (Marantz 1991: 23).

To illustrate how case is assigned, Marantz (1991: 23) offers the representation in (28),

where the subject and object NPs are links in the same chain:

(28) IP
3
NPi I’
5 3
N+CASE I VP
3 3
Vj I V NP

tj ti

Marantz then explains how the case of the case affix is determined in the following manner:

there are three case candidates for the NP in the structure of (28), and these are lexical (e.g.

dative), nominative or accusative. If the verb requires lexical dative on the object NP, then that

NP will preserve this case no matter where it ends up at SS, MS or PF. The case preservation is

explained in this theory by the fact that the object NP and the subject NP are both members of

the same chain in (28). The case affix may receive accusative feature in the object position. This

is explained in this theory on the grounds that the NP is governed by the trace of V in (28). The

16
Crucially, note that the case feature that the case affix acquires depends, in the theory of Marantz (1991), on the
syntactic properties of the element that governs the maximal projection of the noun to which the affix is attached.
This casts doubt on the assumption that case assignment is a PF phenomenon, as argued in Marantz (1991). I thank
Gabriela Alboiu for pointing out this to me.
109
NP in the subject position may receive nominative case. This is explained by the fact that this

position is governed by the complex head V+I in (28). Case realization obeys a disjunctive

hierarchy typical of morphological Spell-Out. In this hierarchy, the more specific case feature

wins out over the less specific case feature. The hierarchy is disjunctive in that going down the

list, as soon as the case affix finds a case feature it qualifies for, it picks up that feature and

leaves the hierarchy. Marantz (1991: 24) offers the disjunctive hierarchy in (29):

(29) case realization disjunctive hierarchy:

- lexically governed case

- “dependent” case (accusative and ergative)

- unmarked case (environment-sensitive)

- default case

Marantz (1991: 24) explains the disjunctive hierarchy as follows:

“Lexically determined case takes precedence over everything else, explaining the

preservation of quirky case when an NP moves from a position governed by a quirky case

verb to a position of NOM or ECM ACC case realization; […] Unmarked case may be

sensitive to the syntactic environment; for example, in a language GEN may be the

unmarked case for NPs inside NPs (or DPs) while NOM may be the unmarked case

inside IPs. Finally, there is a general default case in the language when no other case

realization principle is applicable.”

According to this hierarchy, if a verb requires a lexical dative case on the object NP, then this is

the case the NP will have at any position, as lexical case ranks higher than any other case feature

in the hierarchy. If lexical case is unavailable, the object NP will receive a dependent case if

110
certain syntactic conditions are met, as will be explained below. If dependent case is

inapplicable, an NP will receive an unmarked case. For example, a language may have the

genitive case as the unmarked case in the NP domain, but nominative case as the unmarked case

in the TP domain. If no other case realization principles are applicable, a default case is supplied.

Notice that the difference between unmarked case and default case in Marantz’ (1991) theory

seems to be that the former is domain-sensitive whereas the latter is domain-insensitive.

Marantz (1991: 25) spells out what dependent case is in the following manner: accusative

is a dependent case assigned downward to an NP when the subject NP, which is governed by

V+I has certain syntactic properties. Ergative case is a dependent case assigned upward to a

subject NP when V+I governs downward another NP with certain syntactic properties. These

certain syntactic properties are listed in (30):

(30) Dependent case is assigned by V+I to a position governed by V+I when a distinct

position governed by V+I is:

c. not “marked” (not part of a chain governed by a lexical case determiner)

d. distinct from the chain being assigned dependent case

Dependent case assigned up to subject: ergative

Dependent case assigned down to object: accusative

The syntactic properties listed in (30) explain why the subject NP in the representation in (28)

above can never be assigned the accusative case. The idea is that the subject and the object NPs

in the representation are both members of the same chain governed by V+I. In other words, there

are no distinct positions governed by V+I. To illustrate how the subject and the object receive

111
their cases in Marantz’ (1991) theory, we can consider the following example of a transitive

sentence:

(31) John likes Mary.

Updating Marantz’ (1991) theory, the sentence in (31) presumably has the following syntactic

representation:

(32) IP
3
DP I’
John 3
I vP
3
NP v’
<John>
3
v VP
3 3
V v V NP
likes <likes> Mary

In (32), the NP John in Spec, TP is governed by the the functional head I. The subject is not

marked for lexical case, as there is no lexical case assigner. The subject and the object are not

members of the same chain. Therefore, the object qualifies for the dependent accusative case. As

for the subject, it receives unmarked nominative case in the IP domain. Let us now discuss how

an intransitive sentence can be accounted for in Marantz’ (1991) theory by considering the

following example:

(33) Mary arrived.

Presumably, the sentence in (33) has the following representation:

112
(34) IP
3
DP I’
Maryi 3
I vP

3
v VP
3 3
V v V NP
arrived <arrived> < Maryi>

In (34), the NP Mary is not lexically marked, as there is no lexical case assigner. The subject

does not qualify for the dependent accusative case because the subject is a member of the A-

chain <Mary, Mary> whose head is in the subject position in Spec, TP and whose tail is in the

object position in VP. Therefore, the subject receives the unmarked nominative case in the TP

domain.

As is pointed out by a number of authors (e.g. Legate 2008, Marantz’ 1991) PF version of

dependent case assignment faces problems in light of the fact that some syntactic phenomena

(e.g. A-bar movement) occur after dependent case assignment. For example, if an object

receives dependent case in its base-position, and then moves to the Spec, TopP position or Spec,

FocP position for topicalization or focalization purposes, then Marantz’ (1991) theory faces the

following problem: On the assumption that case assignment takes place at PF after syntax, and

the fact that A-bar movement takes place in syntax before PF, the object will be higher than the

subject at PF. This would then change the case assignment pattern such that the object receives

the unmarked nominative case and the subject receives the dependent accusative case, contrary

the observational fact that objects retain their accusative case after movement to the A-bar

113
position. Baker’s (2015) updated version tries to salvage the dependent case assignment by

updating it and moving it back to syntax. The next section discusses this updated approach.

4.3.1.1 Problems with the original version of the dependent case theory

As has been discussed in the previous section, the gist of Marantz’ (1991) account is that

abstract licensing of arguments in certain syntactic configurations is distinct from case

realization. For Marantz, the licensing of arguments in certain syntactic configurations (e.g.

Spec, TP for the subject or the compelement of V for the object) takes place in syntax proper. As

for case realization, Marantz argues that it is strictly postsyntactic; it takes place in the

Morphological Structure of the grammar, the point in the derivation where the surface structure

maps phonological/phonetic structure.

The view that case realization is postsyntactic is also defendend in Bobaljik (2008). As is

argued in Baker (2015), case assignment cannot be a purely PF phenomenon. One piece of

evidence against this view comes from the fact that the dependent case assigned to NPs in their

base-position is retained after movement of these NPs to the left periphery. On the standard

assumption that movement to the left-periphery takes place in the syntax proper, this runs

counter to the Marantzian (1991) claim that case assignment is a PF phenomenon. To illustrate

this point, let us consider the example in (35):

(35) katab-a zayd-un risaalat-an.

wrote-3MSG Zayd-NOM letter-ACC

‘Zayd wrote a letter.’

114
Here the NP risaalat ‘a letter’ receives the dependent accusative case at the Spell Out of TP.

According to the Marantzian (1991) theory, the prediction would be that the calculation of case

assignment would change when the object NP risaalat ‘a letter’ raises to the left periphery for

the purpose of focalization, for instance.17 More specifically, the prediction would be that the

subject NP zayd ‘Zayd’ would receive the dependent accusative case, and the object NP risaalat

‘a letter’ would receive the unmarked nominative case (or through the case-assigning Agree

given the theoretical framework we will adopt in this thesis), when the object NP risaalat ‘a

letter’ is raised to the left periphery. This prediction is based on the standard assumption that

movement to the left periphery is a syntactic rather than a PF phenomenon. However, such a

prediction is not borne out, as shown by the contrasting pair in (36):

(36) a. *RISAALAT-UNi katab-a zayd-an ei.

letter-NOM wrote-3MSG Zayd-ACC

Intended meaning ‘It was a letter that Zayd wrote.’

b. RISAALAT-ANi katab-a zayd-un ei.

letter-ACC wrote-3MSG Zayd-NOM

‘It was a letter that Zayd wrote.’

On the assumption that case assignment is a purely PF phenomenon, the object NP risaalat ‘a

letter’ c-commands the subject NP zayd ‘Zayd’ at PF, as shown in the ungrammatical example

(36a). Therefore, it follows, according to Marantz (1991), that the subject NP zayd Zayd’ in

Note that this prediction does not run counter to Marantz’ (1991) claim that case is assigned to chains because
17

Marantz’ (1991) claim only includes A-chains and does not extend to A’-chains.
115
(36a) should be assigned the dependent accusative case, and the object NP risaalat ‘a letter’

should receive the unmarked nominative case, contrary to fact. The grammatical example in

(36b) shows that the dependent case assigned to the object in its base-position is retained when

the object moves to the left periphery. This casts doubts on Marantz (1991) proposal that case

assignment is a purely PF phenomenon.

I therefore conclude that the original version of the dependent case theory of Marantz

(1991) cannot be on the right track.

4.3.2 The updated version of dependent case (Baker 2015)

Baker (2015: 48) updates Marantz’s (1991) dependent case theory by proposing the rule

of dependent case assignment in (37):

(37) “If there are two distinct NPs in the same spell-out domain such that NP1 c-commands NP2,

then value the case feature of NP2 as accusative unless NP1 has already been marked for case”.

Given that c-command relationship is essential in the updated version of dependent case,

Baker (2015: 49) offers the definition of c-command in (38):

(38) “X c-commands Y if X does not contain Y and the first node that properly contains X

contains Y.”.

To flesh out the mechanics of dependent case, Baker (2014a: 355, 2015: 233) makes the

assumptions in (39):

(39) a. C and v are phase heads.

b. Their complements (IP, VP) are Spell-Out domains.

116
c. Spell-Out involves mapping relevant c-command relations onto linear order

statements, case assignments, and so on.

d. CP is always a “hard phase”: its complement is invisible for later operations.

e. vP may be a “hard phase” or a “soft phase.” If it is soft, the contents of its

complement do remain visible in the next stage of derivation, but only new c-

command relationships are considered at later Spell-Outs.

Baker (2015: 234) explains the assumptions in (39) as follows: (39a, b, d) are standard

assumptions in phase theory (Chomsky 2000 and subsequent work). The difference between

(39d) and (39e) captures the fact that some languages have differential case marking while others

do not. Baker (2015: 146) proposes to consider vP as either a soft phase or a hard phase. Baker

(2015, ex. 59: 149) explains the distinction between the two phases, as shown in (40):

(40) “a. If v is a hard phase head, then the contents of its VP complement are invisible for the

subsequent syntactic derivation after spelling out.

b. If v is a soft phase head, then the contents of its VP complement undergo spell out (e.g.

they may get their case features fixed) but they remain active in the derivation.”

To illustrate the concept of vP being a hard phase, we can consider the examples in (41)

from Sakha:

117
(41) a. Masha salamaat-y sie-te.

Masha porridge-ACC eat-PAST-3sS

‘Masha ate the porridge.’

b. Masha salamaat sie-te.

Masha porridge eat-PAST-3sS

‘Masha ate porridge.’ (Baker 2015, ex. 22: 125).

Baker (2015: 125) explains the difference between the examples in (41) in the following

manner: the object in (41a) receives a specific or definite interpretation, and is assigned the

dependent accusative case. In contrast, the object in (41b) receives an indefinite or non-specific

interpretation, and is assigned an unmarked nominative case. If we assume that v is a hard phase

head, we can account for the differential object marking in the language. We can assume that

when the object is specific or definite, it shifts out of VP into vP. When the complement of v,

namely VP is sent to Spell-Out, there are no NPs available; therefore no dependent case is

applicable. Given that the object is now in the Spec of vP, it is now part of the domain of the

higher phasal head, namely C. Thus, when the complement of C is sent to Spell-Out, the object is

subject to the dependent case, and it receives the dependent accusative case. In contrast, the

object in (41b) does not receive a specific or definite interpretation. Therefore, we can assume

that the object in this structure remains in situ next to the verb. When the complement of the

phasal head v is sent to Spell-Out, the object is not subject to dependent case because it is the

only NP in this domain. Therefore, the object receives an unmarked bare nominative case. To

further strengthen the argument that this analysis is correct, Baker shows that the neutral word

118
order for specific or definite objects is before VP adverbs, whereas the indefinite or non-specific

objects must be after the VP adverbs. This is shown in (42):

(42) a. Masha salamaat-y türgennik sie-te.

Masha porridge-ACC quickly eat-PAST.3sS

‘Masha ate the porridge quickly.’

b. Masha türgennik salamaat sie-te.

Masha quickly porridge eat-PAST.3sS

‘Masha ate porridge quickly.’ (Baker 2015, ex. 23: 126).

Baker shows that the same facts are found in ergative languages. He illustrates with data

from Eastern Ostyak, a Finno-Ugric language spoken in Siberia. When the object in this

language is indefinite, it does not shift out of VP, and thus the subject is not marked ergative, as

in (43a). In contrast, when the object is definite, it shifts over a PP, and the subject is marked

ergative, as in (43b):

(43) a. Mä t’ ǝkäjǝɣlämnä ula mǝnɣälǝm.

we.DU.NOM younger.sister.COM berry pick.PAST.1pS

‘I went to pick berry with my younger sister.’

b. Mǝ-ŋǝn lǝɣǝ ǝllǝ juɣ kanŋa amǝɣaloɣ.

we-ERG them large tree beside put.PAST.3pO/1pS

We put them (pots of berries) beside a big tree.’ (Baker 2015, ex. 26: 128).

119
In addition to languages where the unmarked case for an NP at the spell-out of VP and

TP is the same, as is the case in Sakha, Baker shows that the unmarked case needs not be the

same in these two spell-out domains. Baker illustrates this with the examples in (44) from

Finnish, where the unmarked case at the spell-out of VP is partitive case, whereas the unmarked

case at the spell-out of TP is nominative:

(44) a. Tuo-n karhu-n/karhu-t.

bring-1sS bear-ACC/bear-PL.ACC

‘I’ll bring the (a) bear/the bears.’

b. Tuo-n karhu-j-a.

bring-1sS bear-PL-PART

‘I’ll bring (some) bears.’ (Baker 2015, ex. 45: 141).

According to Baker, the examples in (44) can be accounted for as follows: in (44a), the object

refers to a fixed quantity; therefore, the object shifts out of VP. At the Spell-Out of VP, the

object is not inside VP; therefore, the object does not receive the unmarked case. Instead, the

object, being out of VP is inside the Spell-Out of C, namely TP; therefore, the object is subject to

dependent case, and receives the structural dependent accusative case. In (40b), the object does

not refer to a fixed quantity; therefore, it remains in situ at the Spell-Out of VP. This way, the

object receives the unmarked case in the VP domain, namely the partitive case.

Baker (2015: 141) shows that accusative is clearly a dependent case in Finnish in that it

only shows up when a certain kind of subject is available. Thus, Baker (2015: 141) notes that

when the subject is the featureless subject of an imperative, as in (45), the definite object shows

120
up with nominative rather than accusative case. Baker accounts for this on the grounds that the

object being of a fixed quantity shifts out of VP; therefore, at the spell-out of VP, the object is

not available. At the Spell Out of TP, the definite object is inside the domain of TP, but given

that subject is featureless, the object is not subject to the dependent case, and instead, it receives

the unmarked case. However, given that the object is inside the TP domain, it does not receive

the unmarked case for the VP domain, namely the partitive case. Instead, the object receives the

unmarked case in the TP domain, namely the nominative case.

(45) Tuo karhu! (*karhu-n)

bring.IMPER bear.NOM bear-ACC

‘Bring the (a) bear!’ (Kiparsky 2001, as cited in Baker 2015, ex. 46: 141).

Baker shows that partitive is not a dependent case. Thus, when a bare-plural object with an

indefinite interpretation is found in imperative sentences, it remains partitive. This is shown in

(46):

(46) Tuo karhu-j-a!

bring.IMPER bear-PL-PART

‘Bring bears!’ (Kiparsky 2001, as cited in Baker 2015, ex. 47: 141).

Baker accounts for the example in (46) as follows: the object is indefinite; therefore, it does not

shift out of VP, and remains inside VP. Given that VP is a hard phase in Finnish, the object is the

only available NP at the Spell-Out of VP, and it receives the unmarked case in the VP domain,

namely the partitive case.

121
In contrast to languages where the VP is clearly a hard phasal domain, and v a hard phase

head such as Sakha, Eastern Ostyak, and Finnish (also known as differential object/subject

marking languages), Baker (2015) shows that there are other accusative and ergative languages,

where the VP is clearly a soft phase, and v a soft phase head. As an example of an accusative

language, where VP is a soft phasal domain, Baker illustrates with Cuzco Quechua. In this

language, Baker (2015) shows that the indefinite objects, which do not shift out of VP, still

receive the dependent accusative case. This is shown in (47):

(47) Juan wawakuna-man miski-*(ta) qunpuni.

Juan children-DAT candy-ACC give.HAB.3sS

‘Juan gives candy to the children.’ (Baker 2015, ex. 54: 146).

According to Baker (2015: 149), examples such as (47) receive an explanation if we assume that

the VP domain is a soft phasal domain, and the contents of VP remain visible to the CP spell out

domain. In other words, the indefinite object in (47) remains inside VP. At the Spell Out of VP,

the case of the object is considered, but is not valued yet. Instead, the object receives its case

value at the spell-out domain of CP. In this domain, both the object and the subject are

considered for case purposes; therefore, the object is subject to the dependent case, and it

receives the structural dependent accusative case. Thus, Cuzco Quechua contrasts with Finnish

or Sahka, where the object receives the dependent accusative case if and only if the object is

definite, and where the definite object shifts out of VP.

Baker shows that the same pattern is observed in ergative languages. He illustrates with

the example in (48) from Lezgian:

122
(48) Gada.di wiči-n žibin.da-j c’akul aqud-na.

boy.ERG self-GEN pocket-INEL feather take.out-AOR

‘The boy took a feather out of his pocket.’ (Haspelmath 1993, as cited in Baker 2015, ex.

55: 146).

Baker shows that the subject in Lezgian, as in (48), uniformly receives ergative case regardless

of the position or the specificity of the object. Thus, in (48), the object is indefinite; therefore, it

does not shift out of VP; yet, the subject still receives the dependent ergative case. Thus, Lezgian

contrasts with Eastern Ostyak, where the subject receives a dependent ergative case if and only if

the object is definite.

In addition to languages having special structural cases inside the TP cycle (special

structural dependent accusative in the accusative languages and special structural dependent

ergative in ergative languages), which is the most common type of languages, Baker shows that

there is analogue to that in that there are languages, where the higher of the two NPs inside the

VP cycle receives a special structural dependent case, namely dative. In other words, dative case

is the VP-internal analogue of ergative case inside the TP cycle. Baker (2015: 131-134) shows

that this is the case in Sakha. Baker illustrates with an example of ditransitive verbs in Sakha

(49), where the goal argument, which is generally higher than the theme argument inside the VP

domain, receives dative case.

(49) Min [VP Masha-qa kniige-ni bier-di-m] (*with Masha-ny)

I Masha-DAT book-ACC give-PAST-1sS Masha-ACC

‘I gave Masha the book.’ (Baker 2015, ex. 30: 132)

123
In order to confirm the claim that dative in Sakha is structural rather than lexical inherent case,

Baker provides the example in (50b), where the dative is clearly assigned to an NP if the root

verb of a morphological causative is transitive. Baker (2015) shows that the higher NP in the VP

domain has the agent thematic role rather than the goal role, which is normally associated with

lexical inherent dative case cross-linguistically.

(50) a. Sardanna [VP Aisen-y yta(a)-t-ta (*Aise-ŋa)].

Sardanna Aisen-ACC cry-CAUS-PAST.3sS Aisen-DAT

‘Sardaana made Aisen cry.’

b. Misha [VP Masha-qa miin-i sie-t-te].

Misha Masha-DAT soup-ACC eat-CAUS-PAST.3sS

‘Misha made Masha eat the soup.’ (Baker, ex. 31b: 132)

Baker further shows that the dative case on the higher NP of the VP cycle cannot be

thematically related, as is shown by the fact that the same NP receives accusative case rather

than dative case if the root verb of the morphological causative is intransitive, as shown in (50a).

Baker (2015: 131), therefore, claims that Sakha has a rule, which can be stated as in (51):

(51) “If XP c-commands ZP in VP, then assign Case U (dative) to XP” (Baker 2015, ex. 29a:

131).

Just as dative case is the VP-internal analogue of ergative case inside the TP, Baker

(2015: 137-138) shows that there are languages, where the lower NP inside the VP cycle receives

a special structural dependent case, namely structural oblique. In other words, this is the VP-

124
internal analogue of accusative case inside the TP cycle. Baker argues that Chamorro is a

language, which shows structural oblique case on the lower of two NPs inside the VP cycle. To

substantiate his claim, Baker (2015: 137-138) provides three arguments: First, oblique is used to

mark the theme but not the goal argument of dyadic verbs, as shown by the example in (52):

(52) Ha-na’i si nana-ṅa ni buteya-n ketchup

3s-give PN.UNM mother-3.POSS OBL bottle-LK soy sauce

‘He gave his mother the bottle of soy sauce.’ (Baker 2015, ex. 39: 137)

Second, oblique is used in morphological causatives when the base verb is transitive, as

shown in (53b), but not when it is intransitive, as shown in (53a). In this case, oblique marks the

theme argument, not the causee argument.

(53) a. Hu na’-poddung i bola

1s CAUS-fall the ball

‘I dropped the ball.’

b. Ha na’-taitai häm i ma’estru ni esti na lebblu

3s CAUS-read us the teacher OBL this LK book

‘The teacher made us read this book.’ (Baker 2015, ex. 40: 138)

Third, oblique is used to mark the theme argument of dyadic verbs that have experiencer

subjects. This is shown by the example in (54):

125
(54) Maleffa yu’ ni lebblok-ku

forget 1s OBL book-1sPOSS

‘I forgot my books.’ (Baker 2015, ex. 41: 138)

Baker, therefore, claims that Chamorro has a rule, which can be stated as in (55):

(55) “If XP is c-commanded by ZP in VP, then assign Case V (oblique) to XP” (Baker 2015,

ex. 29: 131).

Another pattern of case assignment that Baker describes is one, where the language has a

single rule, which assigns a particular dependent case to the higher of two NPs in the same spell-

out domain regardless of whether the spell-out domain is TP or VP. Baker claims that this is the

case in Ika, where the suffix seʔ is used to mark the agent of a transitive verb (56a) and the goal

of a ditransitive verb (56b):

(56) a. Tigri-seʔ tšina kΛ-ga-na

Jaguar-ERG pig PERI-eat-DIST

‘A jaguar ate his pig.’

b. Abran-di Juan-seʔ kafé aʔbe u-ž-in

Arban-TOP Juan-LOC(=ERG) coffee deliver AUX-MED-WIT

‘Arban delivered coffee to Juan.’

c. Juan-di Abram-seʔ kafé k-i-san u-ž-in

Juan-TOP Abram-LOC(=ERG) coffee PERI-?-buy AUX-MED-WIT

‘Juan bought coffee from Abram.’ (Baker, ex. 36: 136)

126
Baker (2015: 136) claims that the example in (56c) shows clearly that the higher argument

receives a structural dative(ergative) rather than lexical inherent dative given that the higher

argument in (56c) has the thematic role SOURCE. According to Baker (2015: 136); therefore,

Ika has a single rule of case assignment, which can be stated as in (57):

(57) “If NP1 c-commands NP2, and both are in the same spell-out domain (VP or TP), assign

NP1 case X.” (Baker 2015, ex. 35: 136)

According to Baker (2015: 136) X is covered in Ika by ergative and structural dative case.

Similar to the syncretism between ergative and dative case on the higher of two NPs in the same

spell-out domain (VP or TP), Baker (2015: 139) claims that there are also languages, where there

is a syncretism between structural accusative and structural oblique case on the lower of two NPs

in the same spell-out domain (VP or TP). Baker claims that this is the case in Amharic, where the

dependent accusative case is marked on the two internal arguments of ditransitive verbs as well

as on the causee and the lower object of a morphological causative. This is shown by the

examples in (58):

(58) a. Ləmma Almaz-in Tarik-u-n nəggər-at

Lemma Almaz.F-ACC story-DEF-ACC tell-(3mS)-3fO

‘Lemma told Almaz the story.’

b. Ləmma Aster-in gənzəb-u-n sərrək’-at

Lemma Aster.F-ACC money-DEF-ACC rob-(3mS)-3fO

‘Lemma robbed Aster of the money.’

127
c. Ləmma Aster-in səa-u-n as-k’orrət-at

Lemma Aster.F-ACC meat-DEF-ACC CAUS-cut-(3mS)-3fO

‘Lemma made Aster cut the meat.’ (Baker 2015, ex. 43: 139)

Baker (2015: 139) claims that the same case should also be found on the theme/stimulus

argument of an experiencer or possessive verb, but he acknowledges that this would be

indistinguishable from an ordinary transitive constructions, as the pattern would be NOM-ACC-

verb. Baker (2015: 139), therefore, claims that languages such as Amharic have a single rule,

which can be stated as in (59):

(59) “If NP1 is c-commanded by NP2, and both are in the same domain (TP or VP), assign

NP1 case Y” (Baker 2015, ex. 42: 139)

For languages such as Amharic, Y would cover both structural dependent cases, the accusative

and oblique case, which are syncretized as accusative.

Returning to the assumptions made in Baker (2015) in (39) above, (39c) includes the

novelty that when c-command pairs are considered for the purposes of determining linear order,

case assignment is also determined. Thus, both linear order and case assignment happen at Spell-

Out, the point in the derivation when syntax interfaces with PF. (39e) also includes another

novelty, namely the idea that NPs in the VP domain are still accessible for case assignment in the

CP cycle, but only if they are part of new c-command pairs at the larger spell-out domain TP.

Having discussed Baker’s (2015) updated version of the Marantzian dependent case

theory of (1991), it is important to note that the major distinction between this theory of case and

the Agree-based theory of case (Chomsky 2000, 2001) lies in the fact that in Baker (2015), case
128
is a relationship between two NPs in a local domain (VP of the phasal head v or TP of the phasal

head C). In other words, the structural dependent accusative/ergative case is only assigned in the

presence of another NP in the same local domain (VP or TP). Otherwise, dependent case is not

assigned, and the only NP in the structure receives an unmarked case (e.g. nominative if the

spell-out domain is TP, genitive inside the NP domain, or partitive in the VP domain). This

approach to case contrasts with an Agree-based approach to case (Chomsky 2000, 2001), where

case is the result of an Agree relation between an NP and a functional head (e.g. structural NOM

is the result of an Agree relation between a finite T and an NP subject in the c-command domain

of T, and structural ACC is the result of an Agree relation between the strong phasal head v* and

an object NP in the c-command domain of v*).

To summarize, in the updated version of the dependent case theory, Baker (2015)

proposes the following rules of dependent case assignment:

(60) a. If XP is c-commanded by ZP in TP, the spell-out domain of CP, assign XP the

dependent accusative case (vP is a hard phase = differential object/subject marking

languages)

(the unmarked case in the VP and TP domains is the nominative case = Sakha)

(the unmarked case in the VP domain is partitive; the unmarked case in the TP domain is

nominative = Finish)

129
(61) a. If XP is c-commanded by ZP in TP, the spell-out domain of CP, assign ZP the

dependent ergative case (vP is a soft phase, the subject receives the unmarked nominative

case in TP, Eastern Ostyak)

b. If XP is c-commanded by ZP in TP, the spell-out domain of CP, assign XP the

dependent accusative case (vP is a soft phase, and the language exhibits a “strict cycle”

effect such that the dependent accusative case is assigned only at the TP cycle, Cuzco

Quechua, Korean, Japanese)

c. If XP is c-commanded by ZP in TP, the spell out domain of CP, assign ZP the

dependent ergative case (vP is a soft phase, and the language exhibits a “strict cycle”

effect such that the dependent ergative case is assigned only at the TP cycle, Lezgian)

(62) a. If XP c-commands ZP in VP, the spell out domain of vP, assign the dependent (dative)

case to XP (Sakha)

b. If XP is c-commanded by ZP in VP, the spell out domain of vP, assign the dependent

(oblique) case to XP (Chamorro)

(63) a. If XP c-commands ZP in the same spell out domain (VP or TP), assign XP the

dependent (ergative) case to XP (Ika)

b. If XP is c-commanded by ZP in the same spell out domain (VP or TP), assign XP the

dependent (accusative/oblique) case (Amharic).

130
In this thesis, the updated version of the dependent case, as developed in Baker (2015)

will be adopted as the theoretical framework.18

4.3.3 Theoretical assumptions

In what follows, I lay out the theoretical assumptions that form the basis for the analysis

to be developed of the case assignment facts of SA in chapter 5.

(a) Following Baker (2014a: 355, 2015: 233), I adopt the theoretical assumptions in (64):

(64) a. C and v are phase heads.

b. Their complements (IP, VP) are Spell-Out domains.

c. Spell-Out involves mapping relevant c-command relations onto linear order

statements, case assignments, and so on.

d. CP is always a “hard phase”: its complement is invisible for later operations.

e. vP may be a “hard phase” or a “soft phase.” If it is soft, the contents of its

complement do remain visible in the next stage of derivation, but only new c-

command relationships are considered at later Spell-Outs.

f. Following Baker (2015: 294-295) I will assume the following hierarchy of case

assignement in SA:

18
The dependent case of Baker (2015) has recently been criticized by Omer Preminger and Ted Levin (2015) on the
grounds that two modalities of case assignment, the dependent case and the Agree-based case are undermotivated.
They argue that the Agree-based case assignment can be dispensed with when accounting for case assignment in
languages such as Sakha. At this point, I am unable to evaluate this critique. I thank Arsalan Kahnemuyipour for
directing my attention to this work.
131
(65) lexical case < dependent case < Agree-based case < unmarked/default case

Lexical case = I will argue that this case is assigned by two lexical items in SA.

The first lexical case assigner in SA is the indicative complementizer ʔinna and its

variant ʔanna. I claim that the second lexical case assigner in SA is the copular

verb k-w-n ‘to be’.

Dependent case = This is the case that is assigned to a lower NP when it is c-

commanded in a new c-command relationship by a higher NP in the same spell

out domain, VP or TP.

Agree-based case = This case is assigned as the result of an Agree relation (in

tense, mood, aspect) between a functional head T and the subject in SA. This case

is phase-bound in the sense that it must be assigned inside the CP phase, and

cannot be determined by higher phases. I claim that v does not engage in an Agree

relation with the object in SA. Accusative case in this thesis is always the result of

the dependent case.

Unmarked/default case = This is the case that is assigned when neither of he

higher-ranked case mechanisms applies. I claim that it is nominative in the CP

domain, bur accusative in the PP domain.

Semantic/adverbial case = This case is not included in the hierarchy above.

However, I claim in this thesis that this case is needed to mark adverbial NPs as

having the adverbial function.

132
Given the hierarchy in (61), it will be shown that the unmarked/default case in the

CP domain is nominative case, but the unmarked case in the PP domain is the

accusative case. The claim that the accusative case is the unmarked case inside the

PP domain is also taken from Baker (2015: 296), who claims that “[s]ome

languages might have a special unmarked case triggered by P heads (oblique in

the Hindi sense), which may or may not be the same as nominative-absolutive

and/or genitive.” I therefore claim that the unmarked case in the PP domain is not

the same as the unmarked case in the CP domain. Thus, the unmarked case in the

PP domain of SA is accusative case, whereas the unmarked case of the CP

domain is the nominative case.

g. Following Soltan (2007) and Al-Balushi (2011, 2012), I will assume that SA

includes lexical case in addition to structural case. I will assume that the

indicative C head ʔinna/ʔanna assigns a purely idiosyncratic lexical accusative

case to its linearly adjacent NP. Following Fassi Fehri (1993: 88) and Al-Balushi

(2011), I will also assume that the copular verb in SA assigns a purely

idiosyncratic lexical accusative case to its NP/AP complement.

h. Following Rizzi (1997), I will assume the split CP analysis of the left-periphery,

where the CP domain splits into a number of projections, which are shown by the

hierarchy given in (66):

(66) ForceP > TopP > FocP > TopP > FinP

133
The hierarchy in (66) shows that a focus phrase (FocP) can be sandwiched by two

topic phrases (TopP). In the case of SA, I follow (Shlonsky 2000) and Aoun et. al.

(2010), where the claim is made that clitic left-dislocated elements are always

base-generated, and they occupy the higher TopP position (or they are adjoined to

projections in the left periphery of the clause, as I will argue in this thesis). I will

also assume with Shlonsky (2000) that the reason why SA does not exhibit a

lower TopP projection is because the FocP has an adjacency requirement for the

verb given that focus phrases in SA induce subject-verb inversion.

(b) Based on Baker (2015), I will argue that the case assignment facts of SA will be shown to be

determined by one of the rules proposed in Baker, a rule offered in Baker to account for case

assignment in languages such as Amharic. The rule is stated here as (67):

(67) If XP is c-commanded by ZP in the same spell out domain (VP or TP), assign XP the

dependent accusative case

(c) I will claim that vP in SA is a soft phase in the sense of Baker (2015).

(d) One of the major claims of this thesis is that even in those structures where the dependent

case fails to apply, v does not engage in an Agree relation with the object, and does not assign

the structural accusative case as a result. This, I argue, follows naturally from the hypothesis that

v is a soft phase in SA. Thus, Fassi Fehri’s (1993) condition on case discharge, where the

accusative case is assigned only when the subject receives case is now interpreted as a reflex of v

being inactive rather than as a reflex of the subject not receiving case in certain structures.

134
(e) Following Fassi Fehri (1993), I claim that T can host features of tense, mood or aspect.

Therefore, I claim that any of these features in T can establish an Agree relation with the subject

and assign the subject the structural nominative case as a result.

(f) I will claim that DP and PP in SA are hard phases in the sense of Baker (2015).

(g) Following Baker (2015), multiple case assignment is allowed across the Spell Out of the

phasal head v*, as this is the only soft phasal head in SA.19 I claim that multiple case assignment

does not apply across the phasal heads C, D, or P, as these are claimed in this thesis to be hard

phases in SA where case decisions are made exclusively from within.

(h) I will claim that the null subject pro does compete for case, wheras tha null subject PRO does

not.

In addition to the above theoretical assumptions, I will take two further theoretical

positions, one regading the status of the preverbal DPs in the SVO order, and another regarding

the nature of the default/unmarked case in SA. These two theoretical positions will be discussed

later in this chapter. Before that, the rules of dependent case assignment have to be spelled out.

This is addressed in the next subsection.

4.3.4 The rules of dependent case assignment

Following Baker (2015: 174-175), I make the following two theoretical assumptions

about the rules of dependent case assignment:

(a) If X has a referential index and X is c-commanded by Y such that Y has a distinct index in

the same spell out domain, assign X accusative.

19
Earlier in this thesis, I argued against some of the Agree-based accounts of case (e.g. Raḥḥali 2003, Soltan 2007,
2011) on the grounds that their accounts face problems with ECM constructions. As far as I am concerned, there is
evidence that multiple case checking should be allowed (cf. Bejar and Massam 1999, and more recently Alboiu and
Hill 2016).
135
What (a) means is that for the dependent case to apply, the following two conditions must be

met: both the case competitor (i.e. the higher NP in the spell out domain) and the case undergoer

(i.e. the lower NP in the same spell out domain) have to have (a) referential indices, and (b) the

referential index of the case competitor has to be distinct from the referential index of the case

undergoer (cf. Marantz’s (1991) claim that the NPs engaged in dependent case assignment must

be distinct). In the theory of lexical categories laid out in Baker (2003: 104), only nouns and their

projections have a referential index, which is conceived of as having an ordered pair of intigers

in the syntactic representation of nouns such that nouns have syntactic representations of the

form X[j,k].20 The first integer introduces a new referent to the syntactic representation, and the

second intiger must be shared with something else in the structure, and the two integers are equal

by virtue of the fact that they are in the same index. The syntactic representation X[j,k] is read as j

is the same X as k. Having two intigers in the syntactic representation of nouns is, according to

Baker (2003: 101-109), dictated by the fact that nouns, but not verbs or adjectives, have a

criterion of identity. In other words, only nouns can fill the blank in the frame “X is the same ___

as Y”, as can be verified by the contrasts between (68) and (69):

(68) a. That is the same man as you saw yesterday.

b. Those are the same women as we saw last night.

c. That is the same water as was in the cup this morning.

d. The French want to have the same liberty as the Americans have. (Baker 2003, ex. 14:

101)

(69) a. #That is the same long as this.

20
It is true that in Baker’s (2003) theory of lexical categories only nouns and their projections have a referential
index. I take this to mean that null categories such as pro and PRO also have a referential index, since these are NPs
as well.
136
b. #She is the same intelligent as he is.

c. #I saw Julia the same sing as Mary did.

d. #I watched Nicholas the same perform a stunt as Kate performed. (Baker 2003, ex. 15:
101)

Baker (2003: 101-102) points out that the reason why the examples in (69) are ill-formed cannot

be attributed solely to syntactic reasons on the grounds that only nouns can follow determiners

such as the and adjectives such as same. Rather, the examples are ill-formed because they are

both ungrammatical syntactically and uninterpretable semantically. Thus, they are different from

ungrammatical examples such as Chomsky’s (1957, as cited in Baker 2003: 102) example given

in (70):

(70) The child seems sleeping.

As pointed out in Baker (2003: 102), the example in (70) is ill-formed for purely syntactic

reasons even though it is semantically interpretable.

One of the axioms that Baker (2003: 104) adopts for common nouns is the mathematical

property of transitivity, which is given in (71):

(71) N{i, k} and N{k, n} → N {i, n} (transitivity: if i is the same N as k and k is the same N as n,

then i is the same N as n) (Baker 2003, ex. 18d: 104).

Given the property of transitivity offered in (71), Baker offers the following example to illustrate

his system:

(72) I bought a pot[i,k] and a basket[l,m]. The pot[n,i] is heavy (Baker 2003, ex. 19c: 105)

137
In (72), the second token of pot has the index [n, i]. There is a new integer in the index, namely

n, and n is equal to the second integer i because they are in the same index. Since n is equal to i

in the index of the second token of pot and since i of the index of the first token of pot is equal to

k of the first token of pot, it follows by transitivity that k of the first token of pot is equal to n of

the second token of pot.

Thus, to engage in dependent case calculation, the two NPs involved in a spell out

domain must have distinct referential indices.

4.3.5 The preverbal position is a subject position

The first thing to notice about SA is the fact that the VSO order is the unmarked order of

the language. However, the SVO order is also used. This can be illustrated with the examples in

(73):

(73) a. katab-a t l-banaat-u risaalat-an

wrote-3FSG the-girls-NOM letter-ACC

‘The girls wrote a letter.’ (VSO)

b. l-banaat-u katab-na risaalat-an

the-girls-NOM wrote-3FPL letter-ACC

‘The girls wrote a letter/(The girls, they wrote a letter.)’ (SVO)

As can be seen in (73a), the postverbal DP is unambiguously a subject. The status of the

preverbal DP has been a subject of ongoing debate between those who consider them to be

subjects (see Benmamoun 1999, 2000), and others who consider them to be topics (see for

138
example Soltan 2007, Al-Balushi 2011). Fassi Fehri (1993) argues that the status of the preverbal

DP is ambiguous, as it can have a subject reading or a topical reading, as is shown in the English

translation in (73b). Following Benmamoun (1999, 2000, Mohammed 1990, 2000), I will take

the position that the preverbal DP is indeed a subject. There are a number of arguments in

support of this position. First, if we assume that the preverbal DP is a topic rather than a true

subject, then we have to explain why the the verb bears rich agreement (i.e. agreement in number

and gender) in this position only but not in the postverbal position. Fassi Fehri (1993) accounts

for the agreement asymmetry in the two orders by arguing that rich agreement can be an

incorporated pronoun linked anaphorically to the preverbal DP, when the latter is interpreted as a

topic. However, as noted in Aoun et. al. (2010: 79), the incorporation account is problematic

given that the ɸ-features are realized as both prefixes as well as suffixes in the

imperfective/present form of the verb. A proponent of the incorporation analysis would have to

assume that the pronoun has somehow split such that part of it appears on the prefix and part of it

appears as a suffix. Clearly, any analysis along these lines is problematic, given that pronouns

are always realized as one phonological unit. Soltan (2011) also argues that the preverbal DP is a

topic rather than a true subject. To account for the agreement asymmetry, Soltan (2011) argues

that the verb in the SV order agrees with a postverbal null subject, pro and rich agreement on the

verb is therefore necessary in order to identify the null subject. This position is also problematic

for a number of reasons. First, Fassi Fehri (1993) shows that agreement with pronouns is not

different from agreement with lexical DPs, as the agreement asymmetry obtains regardless of

whether the verb agrees with a pronoun or a lexical DP. This is illustrated by the contrast in (74)

and (75) (the examples are taken from Fassi Fehri 1993):

139
(74) a. l-nisaaʔ-u nabiil-aat-un/* nabiil-at-un.

the-women-NOM noble-FPL-nom/*noble-FSG-NOM

‘The women are noble.’

b. ʔa-nabiil-at-un l-nisaaʔ-u?

Q-noble-FSG-NOM the-women-NOM

‘Are the women noble?’

(75) a. ʔantunna nabiil-aat-un.

you.FPL.NOM noble-FPL-NOM

‘You (fpl) are noble.’

b. ʔa-nabiil-at-un ʔantunna?

Q-noble-FSG-NOM you.FPL.NOM

‘Are you (fpl)noble?’

The examples in (74) show that there is a plural agreement in the DP + adjective order, but plural

agreement is lacking in the adjective + DP order. This is the case when the DP is a lexical DP.

The examples in (75) show that the same agreement asymmetry is observed when the DP is a

pronominal DP. Based on examples like these, Fassi Fehri (1990) concludes that there is no

distinction between agreeing with lexical DPs and agreeing with pronominal DPs since in both

cases, the same asymmetry is observed. Other examples to show the same asymmetry are (76)

and (77) (the examples in (76b) and 77b) are from Al-Astrabaaði13th c./1996: 127):

140
(76) a. humaa qaaʔim-aan/*qaaʔim-un.

they.MDU.NOM standing.up-MDU.NOM/*standing.up-MSG.NOM

‘The two of them are standing up.’

b. ʔa-qaaʔim-un humaa

Q-standing.up-MSG.NOM they.MDU

‘Are they standing up?’

(77) a. ʔantumaa qaaʔim-aan/*qaaʔim-un.

you.MDU.NOM standing.up-MDU.NOM/standing.up-MSG.NOM

‘You two are standing up.’

b. maa qaaʔim-un ʔantumaa.

NEG standing.up.MSG-NOM you.MDU

‘You two are not standing up.’

In addition to the fact the agreement asymmetry is the same regardless of whether the

subject is a pronoun or a lexical DP, the complex tense constructions provide another context,

where the preverbal DP can only be interpreted as a subject. This is illustrated in (78a):

(78) a. kaan-at l-banaat-u y-aktub-na-Ø risaalat-an.

was-3FSG the-girls-NOM 3-write-PL.F-INDIC letter-ACC

‘The girls were writing a letter.’

141
In (78), the DP l-banaat ‘the girls’ is preverbal, as it triggers rich agreement on the lexical verb.

Here, no argument can be made that the preverbal DP is a topic, since the sentence does not have

the illicit interpretation ‘as for the the girls were, they were writing a letter’. The only reading the

preverbal DP has is that of the subject. Furthermore, on the assumption that the DP l-banaat ‘the

girls’ is a topic, the subject must be a null pro following the lexical verb, as shown in (78b):

(78) b. kaan-at l-banaat-u y-aktub-na-Ø pro risaalat-an.

was-3FSG the-girls-NOM 3-write-PL.F-INDIC letter-ACC

‘The girls were writing a letter.’

Following Soltan (2007, 2011), one would have to assume that the auxiliary verb k-w-n ‘be’ and

the lexical verb both establish an Agree relation with the null subject pro. The DP l-banaat ‘the

girls’ should not act as an intervener, as it is in a left-peripheral A-bar position. The auxiliary

verb would have to Agree in GENDER with the null subject, and the lexical verb would have to

establish an Agree relation with the postverbal subject in GENDER, EPP and ɸ. The problem

that this analysis creates is the following: the auxiliary verb shows agreement of the VS order on

Soltan’s (2007, 2011) account despite the fact that it is in an Agree relation with a null subject

pro, an Agree relation which should exhibit full egreement, contrary to Soltan’s (2007, 2011)

predictions. Notice that (78b) contrasts with (78c) which is in line with Soltan’s (2007, 2011)

predictions, as both the auxiliary verb and the lexical verb show full agreement when they Agree

with the null subject pro.

(78) c. l-banaat-u kunna y-aktub-na-Ø pro risaalat-an.

the-girls-NOM were.3FPL 3-write-PL.F-INDIC letter-ACC

‘The girls were writing a letter.’

142
To account for the agreement asymmetry, I follow Wurmbrand and Haddad (2016: 23-

24), who claim that v is specified for a gender feature only, wheras T is fully specified for the

features of person, number and gender.21 Wurmbrand and Haddad (2016) argue that there are

two ways of satisfying the EPP feature of T. This can be achieved directly by raising the subject

from Spec, vP to Spec, TP, or indirectly by raising v to T. They argue that v has a gender feature

only, which is valued against that of the subject, when the latter is merged in Spec, vP. In the

SVO order, T (with the features of person, number and gender) agrees with the subject, and the

subject is raised to Spec, TP. The result of this movement is that T values its person, number and

gender features. They further argue that another way of satisfying the EPP feature of T is by

rasing v to T. The functional head v is specified for a gender feature only, which is valued

against that of the subject in Spec, vP. When v raises to T to satisfy the latter’s EPP feature, v

agrees with T and values the latter’s gender feature. As for the person and number features of T,

they receive a default value [3 SG]. Wurmbrand and Haddad’s (2016) model for agreement in

the SVO order and the VSO order can be schematized, as in (79) and (80) (from Wurmbrand and

Haddad 2016: 23-24):

(79) Agreement in the SVO order when the preverbal DP is specified as [3.PL.F]22

21
Note that the claim that T is fully specified in the syntax for the features of person, number and gender does not
mean that all these features should be realized morphophonologically. As I argued in Ahmed (2015), the
morphophonological realization of T in the SVO order includes only number and gender, but not person. In other
words, only a subset of the full ɸ-set of T is realized morphophonologically in the SVO order of SA.
22
Although the structure in (78) does not show it, the assumption is is that v-V always moves to T in SA.
143
TP
3
DP T’
iɸ:3.F.PL 3
T vP
uɸ:3.F.PL 3
DP v’
iɸ:3.F.PL 3
V+v …
uɸ: F

(80) Agreement in the VSO order when the postverbal DP is specified as [3.PL.F]

TP
3
T vP
3 3
V+v T DP v’
uɸ: F uG:F iɸ:3.F.PL 3
u#: V+v …
uP:

That the lexical verb in SA raises to T is confirmed by the fact that the verb may precede

vP adverbs, as shown in (81) and (82):

(81) a. ʔakal-a Zayd-un kaṯiir-an l-tuffaḥ-a

ate-3MSG Zayd-NOM abundantly-ACC the-apples-ACC

‘Zayd ate apples abundantly.’ (Raḥḥali and Souâli 1997, ex. 8a: 321)

b. t-aṭbux-Ø-u Hind-un daaʔim-an l-samak-a

F-cook-SG-INDIC hind-NOM always-ACC the-fish-ACC

‘Hind always cooks fish.’ (Raḥḥali and Souâli 1997, ex. 8b: 321)

144
(82) a. Zayd-un ʔakal-a kaṯiir-an l-tuffaḥ-a

Zayd-NOM ate-3MSG abundantly-ACC the-apples-ACC

‘Zayd ate apples abundantly.’

b. *kaṯiir-an Zayd-un ʔakal-a l-tuffaḥ-a

abundantly-ACC Zayd-NOM ate-3MSG the-apples-ACC

‘Zayd ate the apples abundantly.’23

To summarize, the position taken in this thesis is that the subject may appear either

preverbally to satisfy the EPP feature of T directly or postverbally, in which case the EPP feature

of T is satisfied by raising v to T.

4.3.6 Nominative is the unmarked/default case in SA

In this thesis, I will follow the modern generative literature on SA, where the claim is that

nominative case is the default case in SA. This is the claim made in Fassi Fehri (1993), Ouhalla

(1994), Mohammed (1990, 2000), Raḥḥali (2003), Soltan (2007) and Al-Balushi (2011). One

piece of evidence that this is true is that nominative case can surface on the clitic left-dislocated

NP, as in (83):

(83) l-ḍuyuuf -ui, ʔistaqbal-tu-humi

the-guests-NOM received-1SG-them

‘The guests, I received them.’ (Moutaoukil 1987, ex. 109a: 88)

23
According to Raḥḥali and Souâli (1997, fn. 12: 321), the adverb kaṯiir-an ‘abundantly’ is a restrictiver modifier
of the verb; therefore it can only be conceived as a vP adverb and not as a TP adverb.
145
4.4 Summary

In this chapter, alternative generative accounts of case to the Chomskian accounts are

discussed. One such account is that of Yip et al. (1987). Another account is that of the original

version of the dependent case theory, as developed in Marantz (1991). The chapter discusses the

problems that both of these accounts face. The chapter introduces the theoretical framework of

this thesis, namely the updated version of dependent case theory, as developed in Baker (2015).

Crucially, vP in SA is claimed to be incapable of assigning the accusative case to the object NP,

as it is a soft phase in the sense of Baker (2015). In addition, the preverbal NP is considered to be

a subject, and the nominative case is assumed to be the unmarked/default case in SA.

Another crucial claim of this chapter is that dependent case applies only when the NPs

competing for case have distinct referential indices.

146
Chapter Five

Case assignment of core arguments in Standard Arabic

5.1 Introduction

In this chapter, I develop a dependent case analysis (Baker 2015) of the case assignment

facts in Standard Arabic (SA). This chapter will focus on the case assignment facts of the core

arguments (i.e. the subject and the object).

5.2 vP is a soft phase in SA

One of the major theoretical arguments of this thesis is that the lower phase vP is a soft

phase in SA in the sense of Baker (2015). It is worth recalling that a soft phase in Baker’s (2015)

theory is a phase whose NPs are still accessible to a higher phase even though case assignment

may be determined in this phase. This is in contrast to a hard phase, whose complement NPs are

inaccessible to a higher phase. There are four pieces of evidence to support the claim that vP is a

soft phase in SA. The first piece of evidence is theory-internal. In double object constructions,

the middle argument, (be it goal or source) still receives the dependent accusative case rather

than the structural nominative case assigned via Agree with T, as is shown in (1):

(1) wahab-at Hind-un Zayd-an ṯawb-an

gave-3FSG Hind-NOM Zayd-ACC dress-ACC

‘Hind gave Zayd a dress.’

This indicates that the middle argument, which is in Spec, VP is still accessible to the higher

phase, CP at the point in the derivation when case assignment is calculated. The subject is

147
assumed to be in Spec, vP and the theme object is assumed to be in the complement of VP (more

on this later in this chapter).

The second piece of evidence comes from passive sentences of double object

constructions. Here, the theme argument still receives an accusative case despite the standard

assumption in Agree-based accounts of case that passive verbs lose their ability to assign

structural accusative case to the object. This is shown in (2):

(2) wuhib-a Zayd-un ṯawb-an

given.PASS-3MSG Zayd-NOM dress-ACC

‘Zayd was given a dress.’

From the perspective of dependent case theory, this is also an indication that the vP in SA is a

soft phase rather than a hard phase. If we assume that vP is a soft phase, the facts follow

straightforwardly; the middle argument c-commands the theme argument in vP, and the theme

argument receives the dependent accusative case at the Spell Out of vP. However, if we assume

that vP is a hard phase, the facts do not follow; as the c-command relationship between the

middle argument and the theme argument is old information at the TP cycle; therefore, the

prediction is that no dependent case on the theme argument is possible, contrary to fact.

The third clue that vP is a soft phase in SA comes from the feature specification of the phasal

heads v and C. As argued in Wurmbrand and Haddad (2016), v is impoverished in SA in that it is

specified only for a gender feature. This is in contrast to T, which, according to Wurmbrand and

Haddad, is fully specified for the person, number and gender features. Assuming that v is

148
deficient in its ɸ-specification would then explain why vP is a soft phase in SA. This is

illustrated by the agreement asymmetry shown in (3):

(3) a. katab-at l-banaat-u risaalat-an

wrote-3FSG the-girls.3FPL-NOM letter-ACC

‘The girls wrote a letter.’ (VS order)

b. l-banaat-u katab-na risaalat-an

the-girls.3FPL-NOM wrote-3FPL letter-ACC

‘The girls wrote a letter.’ (SV order)

Assuming that this analysis is on the right track, this explains why vP but not CP is a soft phase

in SA. In other words, vP is a soft phase because the lower phasal head v is impoverished; the

higher phase CP is a hard phase because the functional head T, which inherits its features from C

(see Chomsky 2000, 2001), is not impoverished, as it is specified for the features of person,

number and gender.

The fourth piece of evidence, which indicates that v is a soft phase in SA comes from a

reassessment of the nature of v itself. In the context of process nominals in control structures, v

in SA is inactive (probably due to its impoverished ɸ-feature specification à la Wumbrand and

Haddad 2016) in that even in those cases where the dependent case fails to apply, v does not

establish an Agree relation with the object, and therefore, does not assign the structural

accusative case to it. This is illustrated in (4):

149
(4) y-uriid-Ø-u [DP ʔintiqaad-a [vP PRO

3-want-MSG-INDIC criticizing-ACC

l-mašruuʕ-i/*l-mašruuʕ-a bi- taʕammud-in]

the-project-GEN/*the-project-ACC with-intention-GEN

‘He wants to intentionally criticize the project.’

In chapter 7, I show that the structural accusative case on the object is the result of the object

receiving the dependent accusative case rather than the result of the object agreeing with v.

5.3 Case assignment in simple transitive sentences

The example in (5) is a simple transitive sentence:

(5) [CP C [TP katab-a [vP zayd-un[NP1] [VP <kataba> risaalat-an[NP2]]]]]

wrote-3MSG Zayd-NOM letter-ACC

‘Zayd wrote a letter.’

The derivation of the simple transitive sentence (5) proceeds along the lines of the theoretical

framework adopted in this thesis. This is shown in (6):

(6) a. [VP V NP2]

b. [vP NP1 v [VP V NP2]], v a phase head

Spell out v’s complement: NP2

→ Case NP2 = NA (= not applicable)

150
c. [CP C [TP T [vP NP1 v [VP V NP2]]]], C a phase head

Spell-out C’s complement: NP1 c-c’s NP2 → (NP1 < NP2)

→ Case NP2 = ACC

→ Case NP1 = structural NOM via Agree with T

(c-c = c-command; → = if x then y; x < y = x linearly precedes y)

In (6), NP2 is merged into the structure as the complement of the lexical head V, and VP is

projected. The light verb v is then merged into the structure with the external argument NP1

merged in the specifier position of vP. Given that vP is a phase, its domain VP is spelled out. The

only argument in this domain is NP2; therefore, the dependent case does not apply. The

functional head T is now merged into the structure and the verb raises to it. The phasal head C is

now merged, and its TP domain is spelled out at this point. In this domain, NP1 now

c-commands NP2; therefore, NP1 linearly precedes NP2 in accordance with Kayne’s (1994)

Linear Correspondence Axiom. NP2 receives the dependent accusative case given that NP2 is

part of a soft phase whose contents are visible to C. As for NP1, it receives the structural

nominative case via Agree with T.

5.4 Case assignment in simple intransitive sentences

The example in (7) is a simple intransitive sentence, where the verb is unergative with

one NP, which is the external argument of the verb, and there are no internal arguments.

(7) [CP C [TP sakat-at [vP zaynab-u[NP1] [VP…<V> ]]]].

stopped.talking-3FSG Zaynab-NOM

151
‘Zaynab stopped talking.’

The derivation of (7) unfolds as in (8):

(8) a. [VP V]

b. [vP NP1 v [VP V]], v a phase head

Spell out v’s complement: NA

c. [CP C [TP T [vP NP1 v [VP V]], C a phase head

Spell out C’s complement: NP1 = structural NOM via Agree with T

In (8), the external argument is located in Spec, vP on standard accounts, and it is the only NP in

the sentence; therefore, the subject NP is not assigned the dependent case, and it is instead

assigned the structural NOM case via Agree with T. Crucially, note that even though dependent

case is ranked higher in the adopted theoretical model than Agree-based case, the former can

only apply when there are at least two NPs competing for case in the same phasal domain;

dependent case does not apply when there is only one NP at the Spell-Out of a phase.

Another example of a simple intransitive sentence is the type of sentence known as the

unaccusative sentence. An example of unaccusative sentences is given in (9):

(9) [CP C [TP waṣal-a [vP v [VP <V> l-musaafir-uun[NP1]]]]].

arrived-3MSG

152
‘The passengers arrived.’24

The derivation of (9) is as shown in (10):

(10) a. [VP V NP1]

b. [vP v [VP V NP1]], v a phase head

Spell out v’s complement: NP1

Case NP1 → not applicable (= NA)

c. [CP C [TP T [vP v [VP V NP1]]]], C a phase head

Spell out C’s complement: NP1 = structural NOM via Agree with T

In (10), the internal argument is base-generated in the complement of VP and it does not raise to

Spec, vP, as the latter is not projected in these structures on standard accounts. There is only one

NP at the Spell-Out of TP; therefore, the dependent accusative case cannot be assigned and the

NP is assigned the structural NOM case via Agree with T. Note that the theme object does not

agree with v, for the latter is, I claim, incapable of establishing an Agree relation with the object

and assigning structural accusative to it. This is in addition to the fact that unaccusative verbs

are, on standard accounts, incapable of assigning case to their objects.

24
I assume that unaccusatives project a vP but they doe not project an external argument.
153
5.5 Case assignment in sentences with double object constructions

The example in (11) is a double object construction (or ditransitive), where the verb has

two internal arguments and one external argument.

(11) [CP C [TP ʔaʕṭ-at [vP zaynab-u[NP1] [VP zayd-an[NP2] <V> risaalat-an[NP3]]]]].

gave-3FSG Zaynab.F-NOM Zayd-ACC letter-ACC

‘Zaynab gave Zayd a letter.’

It is worthwhile to note at this point that the higher object (i.e. the goal argument) and the

lower object (i.e. the theme argument) in (11) receive the structural dependent case. That this is

the case is confirmed by the fact that both objects alternate their accusative case with nominative

case in passives, as is shown in (12):

(12) a. ʔuʕṭiy-at risaalat-un[NP1] li-zayd-in[NP2].25

gave.PASS-3FSG letter.F-NOM to-Zayd-GEN

‘A letter was given to Zayd.’

b. ʔuʕṭiy-a zayd-un[NP1] risaalat-an[NP2].

gave.PASS-3MSG Zayd-NOM letter-ACC

‘Zayd was given a letter.’

25
The reason why the agent is demoted to an adjunct position has to do with the fact in SA, there is a hierarchy of
thematic roles such that the theme cannot be projected higher than the goal in the presence of the latter (cf. Fassi
Fehri 1999 on this issue).
154
That the middle argument in double object constructions receives a structual accusative case is

also confirmed by the fact that this argument receives this case regardless of its thematic role in

the sentence, as can be shown by the following examples:

(13) a. manaḥ-at Zaynab-u Zayd-an jaaʔizat-an

gave-3FSG Zaynab.F-NOM Zayd.M-ACC prize-ACC

‘Zaynb gave Zayd a prize.’

b. salab-at Zaynab-u Zayd-an maal-a-hu

deprived-3FSG Zaynab.F-NOM Zayd-ACC money-ACC-his

‘Zaynab deprived Zayd of his money.’

As can be observed, the middle argument in (13) receives the same case even though it has

distinct thematic roles in the sentence. Thus, it is a goal in (13a), and a source in (13b). This

strongly indicates that the middle argument receives a structural accusative case rather than an

inherent case in the sense of Woolford (2006).

Following the standard assumption in the generative literature (see Baker 2015, Woolford

2006), the goal argument zayd ‘Zayd’ in (11) is projected either in the Spec, VP position or in the

Spec, vGP between VP and the higher vP that projects an external argument. For the purposes of

this dissertation, I assume that the goal argument is projected in the Spec, VP position.

The derivation of the active double object construction in (11) proceeds as in (14):

(14) a. [VP NP2 V NP3]

b. [vP NP1 v [VP NP2 V NP3]], v a phase head

155
Spell out v’s complement: NP2 c-c’s NP3 → (NP2 < NP3)

→ Case NP2 = 0

→ Case NP3 = ACC

c. [CP C [TP T [vP NP1 v [VP NP2 V NP3]]]], C a phase head

Spell-out C’s complement: NP1 c-c’s NP2 → (NP1 < NP2)

→ Case NP2 = ACC

NP1 c-c’s NP3→ (NP1 < NP3)

→ Case NP3 = ACC (redundantly)

→ Case NP1 = structural NOM via Agree with T

In (14), there are two NPs in the spell-out domain of the lower phase vP. In the VP spell-out

domain, NP2 c-commands NP3; therefore, NP2 linearly precedes NP3. As a result of the c-

command relationship, NP3 is assigned the dependent accusative case inside the VP domain. As

for NP2, it does not get a case value yet. Given that the lower phase is a soft phase and it is part

of another phase, NP2 does not get a case value yet. T is merged into the structure, the verb

raises to it. C is merged into the structure, and the domain TP is spelled-out. In this domain, NP1

c-commands NP2, which does not have a case value yet. The result is that NP1 linearly precedes

NP2. As a result of this relationship, NP2 is assigned the dependent accusative case. In this

domain, NP1 also c-commands NP3. The result is that NP1 linearly precedes NP3, and NP3 is

redundantly assigned the dependent accusative case. Note that NP3 receives the dependent case

twice, the first time at the Spell-Out of vP, and the second time at the Spell-Out of CP. Assuming

that NP3 receives the dependent case at the Spell-Out of CP is in accordance with Baker’s (2015)

theory of case given that the c-command relationship between NP1 and NP3 is a new one rather

156
than an old one; therefore, case assignment is available. As for the linear relationship between

NP2 and NP3, it has already been considered in the lower phase vP such that NP2 linearly

precedes NP3. The subject NP1 ends up with the structural nominative case via Agree with T. It

is important to recall here that the case assignment in active double object constructions provides

a strong argument for the claim made in this thesis that vP is a soft phase in SA. The goal

argument NP2 and the theme argument NP3 are still accessible in the TP cycle given that vP is a

soft phase. Therefore, the c-command relationship in the TP cycle between NP1 and NP2 and

that between NP1 and NP3 are both new c-command relationships; as a result, the dependent

case applies for the first time to the goal argument and applies for the second time to the theme

argument.

When the theme argument of (11) is passivized, we get (15):

(15) a. ʔuʕṭiy-at risaalat-un[NP1] li-zayd-in[NP2].

gave.PASS-3FSG letter.F-NOM to-Zayd-GEN

‘A letter was given to Zayd.’

b.*? ʔuʕṭiy-at risaalat-un zayd-an.

gave.PASS-3FSG letter-NOM Zayd-ACC

Intended meaning ‘A letter was given [to] Zayd.’

The example in (15b) is accepted by some traditional Arab grammarians, but rejected as

ungrammatical by others (see Hasan 1962, fn. 1: 112). Among modern linguists, Fassi Fehri

(1999: 51) treats it as ungrammatical and Moutaoukil (1987: 23) considers it to be highly

marginal. For Moutaoukil (1987: 23), sentences such as (15b) become more acceptable only if

the theme argument risaalat ‘a letter’ is referential (i.e. definite or specific indefinite). The
157
reason why (15b) might be judged as grammatical can be explained as follows: on the

assumption that the goal argument (PP or NP) is merged in Spec, VP and the theme argument in

the complement of VP, both are in the same Minimal domain of the head V; therefore, both are

equidistant from the target position, Spec, vP (or any other position for that matter). According to

Hornstein et al. (2005: 19) whose definition is based on Chomsky (1993), Minimal Domain is

defined as in (16):

(16) “The Domain of α, or MinD(α), is the set of categories immediately contained or

immediately dominated by projections of the head α, excluding projections of α.”

They define Containment and Domination as in (17):

(17) a. “Containment: A category α contains β iff some segment of α dominates β.”

b. “Domination: A category α dominates β iff every segment of α dominates β.”

As for equidistance, Hornstein et al (2005: 163) define it as in (18):

(18) “Equidistance (final version): If two positions α and β are in the same MinD, they are

equidistant from any other position.”

In (11), both internal arguments are equidistant from the position Spec, vP. In its original

position, the MinD(V) comprises the set [goal, theme]. The goal is in MinD(V) because a

projection of the head V, namely VP immediately contains and immediately dominates the goal

argument in Spec, VP. The theme argument is in MinD(V) because the projection V’

immediately contains and immediately dominates the theme argument in the complement of V.

Given the standard assumption that the lexical verb in SA moves to adjoin to v and then to T,

158
MinD(V) now includes Spec, TP as well as vP. It follows, therefore, that the Spec, vP position is

equidistant from both the goal argument and the theme argument and either one of the internal

arguments can therefore move to Spec, vP without violating Minimality.

Having established that both internal arguments are equidistant from the Spec, vP

position, let us consider for the purpose of illustration, the derivation of (15a).

I assume that (15a) has the structural representation in (19):

(19) [CP C [TP ʔuʕṭiy-at [vP [VP risaalat-un[NP1] <V>] [PP li-zayd-in[NP2] ]]]].

gave.PASS-3FSG letter.F-NOM to-Zayd-GEN

‘A letter was given to Zayd.’

The derivation of (15a) proceeds as in (20):

(20) a. [vP v [VP NP1 V ][PP P + NP2]]], v a phase head

Spell out v’s complement: NP1

→ Case NP1 = 0

b. [CP C [TP T [vP v [VP NP1 V ]]]]], C a phase head

Spell-out C’s complement: No new c-command relationships are considered for

case assignment

→ Case NP1 = structural NOM via Agree with T

When the vP is projected in (15), its domain VP is spelled-out. In this domain, only NP1 is

available, as NP2, the subject of VP, is demoted to an adjunct PP position, where case is
159
assigned by the preposition. The reason why NP2 is demoted to an adjunct position has to do

with the fact that there is a hierarchy of thematic roles such that the theme cannot be projected

higher than the goal in the presence of the latter. In other words, when the goal is higher than the

the theme in the structure, the goal is an argument, and therefore it is not preceded by a

preposition. Conversely, when the theme is projected higher in the structure, the goal has to be

demoted to an adjunct position, which is embedded in a PP (cf. Fassi Fehri 1999 on the hirerchy

of thematic roles in SA).26 NP1 is not considered for case in this domain, as it is not engaged in a

c-command relation with any other NP for the purpose of case assignment. On standard

accounts, NP1 does not raise to Spec, vP in passive sentences, as Spec, vP is not projected in

these strcutures, on standard accounts. The functional head T is merged into the structure and the

verb raises to it. C is merged into the structure, and its domain TP is spelled-out. NP1 is assigned

the the structural nominative case via Agree with T. Note that v does not Agree with the internal

argument, as the verb is passive, which is incapable of assigning an accusative case to its object

on standard accounts. In passive sentences, I assume that the lexical head V is always c-selected

by the functional head v although v in this type of sentences does not have an external argument.

When the goal argument in (11) is passivized, we get the sentence in (21):

(21) [CP C [TP ʔuʕṭiy-a [vP [VP zayd-un[NP1] <V> risaalat-an[NP2]]]]].

gave.PASS-3MSG Zayd-NOM letter-ACC

‘Zayd was given a letter.’

26
Later in this chapter, I claim that PP is a strong phase in SA, and that the genitive case assigned by the preposition
is structural rather than inherent or lexical.
160
The derivation of (21) proceeds as in (22):

(22) a. [VP V NP2]

b. [vP v [VP NP1 V NP2]], v a phase head

Spell out v’s complement: NP1 c-c’ NP2→ (NP1 < NP2)

→ Case NP1 = 0

→ Case NP2 = ACC

c. [CP C [TP T [vP v [VP NP1 V NP2]]]], C a phase head

Spell-out C’s complement: → No new c-command relations are considered for

case assignment

→ Case NP1 = structural NOM via Agree with T

At the merger of vP in (22), its domain VP is spelled-out. On standard accounts, the goal

argument is base-generated in Spec, VP, and the theme argument is base-generated in the

complement of VP. In the VP cycle, NP1 c-commands NP2. As a result, NP1 linearly precedes

NP2, and NP2 is assigned the structural dependent accusative case inside the VP cycle. At the

merger of CP, its domain TP is spelled-out. In this domain, there are no new c-command

relationships. NP1 receives the structural nominative case via Agree with T. The derivation in

(22) shows clearly that SA is similar to Amharic in that the dependent accusative case is assigned

on the VP cycle. This contrasts with languages such as Korean, where the dependent case is

assigned only on the TP cycle and cannot be assigned based on c-command relationships that

hold only in the VP domain (see Baker 2015: 236-237). In other words, SA and Amharic do not

show a “strict cycle” effect, where the VP cycle inhibits the assignment of dependent case in this

cycle. Had SA been a language with a “strict cycle” effect, the prediction would be that the

161
dependent case is assigned only at the TP cycle. This means that in the context of double object

structures in the passive form, the prediction would be that both internal arguments would

receive the structural nominative case via Agree with T. This is because the c-command

relationship between the goal argument and the theme argument would be old c-command

relationships at the TP cycle, and old information is not considered in case assignment, according

to the adopted model. This would prevent the application of the dependent case. In the absence

of the dependent case, Agree would take over. However, v is not a case assigner, as I claim in

this thesis, being a defective probe in the sense of Chomsky (2000, 2001). In this scenario, the

prediction would be for T to agree with both internal arguments and assign the structural

nominative case to both via multiple Agree. However, this prediction is not borne out by the

facts, which suggests that SA is unlike Korean in that it does not show a “strict cycle” effect.

5.6 Case assignment in sentences introduced by the indicative (emphatic or neutral)

Complementizer

The example in (23) is a complex sentence, where the embedded clause is introduced by

the indicative emphatic complementizer ʔinna ‘that’.

(23) qaal-at zaynab-u ʔinna zayd-an[NP1]

said-3FSG Zaynab-NOM that.EMPH Zayd.M-ACC

katab-a risaalat-an[NP2].

wrote-3MSG letter-ACC

‘Zaynab said that indeed Zayd wrote a letter.’

162
Before attempting to show the derivation for (23), we need to address the status of the accusative

case assigned by the complementizer ʔinna ‘that’. Specifically, is this a structural case or

lexical/inherent case?

Fassi Fehri (1993: 32-33, 2012) suggests that the complementizer assigns a structural

accusative to the NP in its TP complement. This claim is supported by the fact that the

complementizer seems to be able to assign this case both when the NP is linearly adjacent to it

(24a) and when it is not (24b):

(24) a. ʔinna rajul-an fii l-daar-i

that.EMPH man-ACC in the-house-GEN

‘Indeed, a man is in the house.’ (adapted from Fassi Fehri 1993, ex. 95a: 46)

b. ʔinna fii l-daar-i rajul-an

that.EMPH in the-house-GEN man-ACC

‘There is indeed a man in the house.’ (adapted from Fassi Fehri 1993, ex. 95b: 46)

The subject in (24a) is base-generated in Spec, PP before it raises to Spec, TP. To account for

how rajul ‘ a man’ receives its case in (24b), Fassi Fehri claims that the NP is extraposed from

the specifier position of the PP fii l-daar ‘in the house’ to the Spec,TP position, where it can

receive its accusative case from the complementizer.

Similar to Fassi Fehri (1993), Aoun et. al. (2010, fn. 8: 17) claim that the accusative case

assigned by the complementizer is structural rather than inherent. For them, the complementizer

163
cannot be assigning an inherent case to the NP in its TP complement because the NP does not

receive any thematic role from the complementizer.

In contrast to Fassi Fehri (1993, 2012) and Aoun et. al. (2010), Shlonsky (2000: 332-

333), Soltan (2007) and Al-Balushi (2011, 2012) claim that the accusative case assigned by the

complementizer is not syntactic structural. Shlonsky (2000: 232-233), for example, takes the

post-complementizer DP “to be endowed with a morphological feature, [+F], a nominal feature,

which happens to have the same phonetic realization as accusative Case.” He continues to add

that “[t]his feature is not a structural Case feature [emphasis in the original]”. In fact, the

accusative case assigned by the indicative complementizer (ʔinna or ʔanna) cannot be structural.

If it were, the derivation of sentences such as (25), where the accusative case is not assigned after

the complementizer, should crash, contrary to the facts:

(25) a. ʔinna fii l-ʕiraaq-i ḥaṣal-a l-ʔittifaaq-u

that.EMPH in the-Iraq-GEN happened-3MSG the-agreement-NOM

‘It is in Iraq that the agreement took place.’

To account for sentences such as (25), Shlonsky (2000, fn. 9: 333) assumes that “under certain

circumstances and perhaps only marginally, locative PPs can qualify as [+F] bearers ([+F]

remaining phonetically unexpressed) and can thus satisfy ʔanna.” To account for why the

postverbal subject l-ʔittifaaq ‘the agreement’ receives nominative case rather than the purely

idiosyncratic lexical accusative case assigned by the complementizer ʔinna, I claim that this is

due to the fact that the postverbal subject is an internal argument of the verb, which is located in

the complement of VP. In order for the purely idiosyncratic lexical accusative case assigned by

164
the complementizer to be assigned to the subject l-ʔittifaaq ‘the agreement’, the latter needs to be

linearly adjacent to the former (i.e. in the CP domain or in Spec, TP). However, this is not the

case; therefore the lexical case assigned by the complementizer ends up not being assigned, and

the Agree-based case mechanism applies such that the postverbal subject receives its nominative

case via Agree with T. Note especially that the locative PP in (25) is not a subject that qualifies

for structural case; rather, it is a focalized PP. The facts are the same regardless of word order, as

can be seen in (25b) below:27

(25) b. ʔinna l-ʔitifaaq-a ḥaṣal-a fii l-ʕiraaq-i

that.EMPH the-agreement-ACC happened-3MSG in the-Iraq-GEN

‘Indeed the agreement took place in Iraq.’

In (25b), the the subject satisfies the (+F) feature of the the complementizer ʔinna. Note,

however, that Shlonksy (2000) claims that the case assigned by the complementizer is not even a

case feature. I suggest that this claim is unwarranted, as it makes the following wrong prediction:

if the (+F) feature of the complementizer can be satisfied by the PP, then the prediction is that in

those sentences where the subject DP follows the PP, the subject should surface with the

nominative case rather than the accusative case, contrary to the facts, as is illustrated by the

following example:

(26) ʔinna fii l-ʕiraaq-i rajul-an/*rajul-un

that.EMPH in the-Iraq-GEN man-ACC/*man-NOM

y-afʕal-Ø-u haaðaa

3-do-MSG-INDIC this

27
On the assumption that the locative PP is focalized in (25a), it is not possible to test what happens when the
postverbal subject is placed before the verb following the focalized locative PP, since focalization in SA induces
subject-verb inversion (cf. Shlonsky 2000).
165
‘It is in Iraq that someone does this.’

The fact that the (+F) feature in (26) cannot be satisfied by the PP but only by the subject

indicates that the (+F) feature of the complementizer is a case feature, contrary to Shlonsky

(2000). In (26), I assume that the purely idiosyncratic lexical accusative case is assigned by the

complentizer to the preverbal subject because the latter is linearly adjacent to the

complementizer being either in the Spec, TP position or even in the C domain.

I suggest that the accusative case assigned by the complementizer is a purely

idiosyncratic lexical case feature rather than a structural case feature. One piece of evidence to

support this claim is that the same case obligatorily surfaces even on clitic left-dislocated NPs

(see also Shlonsky 2000: 233 for a similar claim), as is shown in (27):

(27) ʔinna l-ʔawlaad-a qaabal-a-hum Zayd-un.

that.EMPH the-boys-ACC met-3MSG-them.PL.F.ACC Zayd-NOM

‘ As for the the boys, they were indeed met by Zayd.’

Fassi Fehri (1986) and Aoun et.al. (2010) show clearly that resumptive pronouns such as

hum‘them.F.PL.ACC’ in (27) void islands; therefore, clitic left-dislocated NPs such as l-banaat

‘the boys’ must be base-generated in a left peripheral position. If this is the case, then the

complementizer must be assigning a lexical accusative case to the clitic dislocated left-peripheral

NP.

Having established that the complementizer ʔinna (and its variant ʔanna) assigns a purely

idiosyncratic lexical accusative case to the NP in its complement, we can proceed to show how

the sentence in (23) is derived. I assume that (23) has the structural representation in (28):

166
(28) qaal-at zaynab-u [CP ʔinna [TP zayd-an[NP1]

said-3FSG Zaynab-NOM that.EMPH Zayd.M-ACC

katab-a [vP <NP1> v [VP <V> risaalat-an[NP2]]]]].

wrote-3MSG letter-ACC

‘Zaynab said that indeed Zayd wrote a letter.’

The derivation of (28) is as shown in (29):

(29) a. [VP V NP2]

b. [vP NP1 v [VP <V> NP2]], v a phase head

Spell out v’s complement: NP2

→ Case NP2 = NA

c. [CP C [TP NP1 T [vP <NP1> v [VP V NP2]]]], C a phase head

Spell-out C’s complement: NP1 c-c’s NP2 → (NP1 < NP2)

→ Case NP2 = ACC

→ Case NP1 = purely idiosyncratic ACC assigned by C

At the merger of v of the embedded clause in (29), its domain VP is spelled-out. Given that there

is only one NP, namely NP2 in this spell-out domain, no c-command relationship is considered,

and hence no case assignment is considered. At the merger of the embedded C, its domain TP is

spelled-out. In this domain, NP1 in Spec, TP is in a new c-command relationship with NP2. As a

result of this relationship, NP2 is assigned the dependent accusative case. As for the embedded

agent argument zayd ‘Zayd’, it is assigned the purely idiosyncratic lexical case assigned by C.

167
5.7 Case assignment in sentences with backgrounding encliticized topics

The example in (30) is a complex sentence, where the embedded clause is introduced by

the indicative complementizer and an embedded pronoun is encliticized onto the

complementizer.

(30) qul-tu proi [ForceP ʔinna-hu [TopP <hu [NP1] >

said-1SG that.EMPH-it.ACC it.3MSG

[TP waṣal-a [vP v [VP <V> l-musaafir-uun[NP2] ]]]]].

arrived-3MSG the-passenger-MPL.NOM

‘I said that indeed the passengers arrived.’

I follow Fassi Fehri (2012: 279) who claims that the clitic pronoun –hu ‘it’ in (30) is a

backgrounding topic. If this analysis is correct, then clearly the complementizer ʔinna ‘that’ to

which the clitic pronoun is encliticized must be base-generated higher than a topic phrase,

presumably in the head of the highest functional projection ForceP of the cartographic structure

of Rizzi (1997). This claim is supported by the fact that nothing can precede the emphatic

complementizer, and the fact that the complentizer also clause-types the sentence as an assertion.

The derivation for (30) is as shown in (31):

(31) a. [VP V NP2]

b. [vP v [VP V NP2]], v a phase head

Spell out v’s complement: NP2

168
→ Case NP2 = NA

c. [CP C NP1 [TP T [vP v [VP V NP2 ]]]], C a phase head

Spell-out C’s complement: NP2

→ Case NP2 = structural NOM via Agree with T

→ Case NP1= purely idiosyncratic ACC assigned by C

In (31), the NP of the embedded clause, NP2 cannot receive the dependent accusative case. On

the standard assumption that NP2 starts the derivation in the object position of the embedded

clause, being the argument of an unaccusative predicate, the embedded NP cannot be assigned a

dependent case given that it is the only NP in the VP and TP domains. The internal argument

does not raise to Spec, vP on standard accounts. Therefore, the embedded subject receives the

structural nominative case via Agree with T. Notice that the embedded subject cannot agree with

the embedded v, given that v in SA is incapable of establishing an Agree relation with its object.

The backgrounding topic, -hu ‘it’ is base-generated in the Spec, TopP position in the embedded

clause, and it is encliticized onto the indicative complementizer, which assigns a purely

idiosyncratic lexical accuasative case to it. In the matrix clause, the dependent case is

unavailable, as there is one NP available, namely the null subject pro, which gets the structural

nominative case via Agree with T.

5.8 Case assignment in sentences with existential constructions

The example in (32) is an existential sentence:

169
(32) kaan-a hunaaka[NP1] rajul-un[NP2] fii l-daar-i[NP3]

was-3MSG there man-NOM in the-house-GEN

‘There was a man in the house.’

Fassi Fehri (1993: 156-159) claims that structures with auxiliary verbs in SA are naturally

analyzed as bi-inflectional in the sense that thay have two Ts rather than one. Among the pieces

of evidence that he brings to support this hypothesis is the fact that in these structures, both the

auxiliary verb and the lexical verb are inflected for their ɸ-features, as well as the fact that

negative markers can negate either the auxiliary verb or the lexical verb, as shown in (33):

(33) a. qad t-akuun-Ø-u l-banaat-u ʔakal-na

may F-be-3SG-INDIC the-girls-NOM eaten-3FPL

‘The girls may have eaten.’ (adapted from Fassi Fehri 1993, ex. 37: 157)

b. kaan-at l-bint-u lam t-aʔkul-Ø-Ø

was-3FSG the-girl-NOM NEG.PAST F-eat-3SG-JUSS

‘The girl had not eaten.’ (adapted from Fassi Fehri 1993, ex. 38: 157)

c. kaan-a l-junuud-u laa y-aʔkul-uu-na

was-3MSG the-soldiers-NOM NEG 3-eat-MPL-INDIC

‘The soldiers were not eating.’ (adapted from Fassi Fehri 1993, ex. 39: 157)

d. lam y-akun-Ø-Ø l-rajul-u (qad) ḥaḍar-a

NEG.PAST 3-be-MSG-JUSS the-man-NOM alrady came-3MSG

‘The man had not already come.’ (adapted from Fassi Fehri 1993, ex. 40: 157)

170
While it is true that the structure in (33) shows neither of the properties associated with complex

tense structures (i.e. both the auxiliary verb and the lexical verb are inflected for their ɸ-features,

as well as the fact that negative markers can negate either the auxiliary verb or the lexical verb), I

still claim that it has two instances of T rather than one. There are three reasons for this claim.

First, the structure projects an existential NP, namely hunaaka ‘there’. On standard accounts, the

expletive there must be base-generated in Spec, TP given that it lacks a thematic role, being

meaningless. It follows, therefore, that the existential NP must be in Spec, TP in (32). Second,

the existential NP is obligatory; and is therefore needed for structural reasons. The sentence in

(32) has one reading in which the NP rajul ‘a man’ is non-specific indefinite, and the sentence is

ungrammatical without the existential, as is shown in (34):

(34) *kaan-a rajul-un[NP2] fii l-daar-i[NP3]

was-3MSG man-NOM in the-house-GEN

Intended to mean,‘There was a man in the house.’ (on the reading where rajul ‘a man’ is

non-specific indefinite)

The fact that (34) is ungrammatical suggests that the existential NP must be base-generated in a

functional head higher than vP. I take that head to be T following standard accounts of the

expletive in other languages. Third, no plausible argument can be made that the copular verb in

(32) is in a position higher than T, as the sentence lacks neither a topic nor a focus reading, and

the copular verb is not stressed. It is worth noting that it is not possible to claim that the

171
expletive in (32) is base-generated in Spec, vP, where it does not receive a theta role.28 If it did,

then the DP rajul ‘a man’ must be generated in the complement of VP, a position, in which it

would qualify for the purely idiosyncratic lexical accusative case assigned by the copular verb.

The prediction would then be that the DP rajul ‘a man’ would surface with the purely

idiosyncratic lexical accusative assigned by the copular verb, contrary to fact.

Fassi Fehri (2012, fn. 24: 280-281) claims that the locative hunaaka ‘there’ is not an

expletive but a true locative adjunct. However, there are some good reasons to believe that there

are contexts in which hunaaka ‘there’can be used as a true expletive and not as a pure locative

adjunct, contra Fassi Fehri (2012). First, unlike true locative adjuncts, the locative expletive

cannot be questioned, as the ungrammaticality of (35) shows.

(35) a.: kaan-a hunaaka l-kaṯiir-u min l-rijaal-i.

was-3MSG there the-many-NOM of the-men

‘There were many men.’

b. A: ʔayna kaan-a l-kaṯiir-u min l-rijaal-i

where was-3MSG the-many-NOM of the-men-GEN

‘Where were many of the men?’29

B: *hunaaka

there

28
This possibility was suggested to me by Gabriela Alboiu.
29
Notice that the locative hunaaka ‘there’ cannot co-occur with the question word ʔayna ‘where’, as shown by the
ungrammatical example in (i):
(i) *ʔayna kaan-a hunaaka l-kaṯiir-u min l-rijaal-i?
Where was-3MSG there the-many-NOM of the-men-GEN
‘*Where were there many of the men.’
The example in (i) is as ungrammatical as the English counterpart with the expletive there is, as shown by the
ungrammatical English translation.

172
‘There.’

The example in (35b) contrasts with the example in (35c), where the locative hunaaka is a true

locative adjunct, hence it can be questioned:

(35) c. A: Zayd-un y-aqif-Ø-u hunaaka

Zayd-NOM 3-stand -MSG-INDIC there

‘Zayd is stnding there.’

B: ʔayna y-aqif-Ø-u Zayd-un?

Where 3-stand -MSG-INDIC Zayd-NOM

‘Where is Zayd standing?’

A: Hunaaka

there

Second, unlike true locative adjuncts, hunaaka ‘there’ cannot be focused, as shown by the

ungrammaticality in (36a), and the expletive hunaaka can also co-occur with the locative adjunct

hunaaka, as shown by the grammatical example in (36c):

(36) a. *HUNAAKA rajul-un fii l-daar-i

there man-NOM in the-house-GEN

‘THERE is a man in the house.’

b. *kaan-a rajul-un fii l-daar-i

was-3MSG man-NOM in the-house-GEN

Intended to mean ‘There is a man in the house.’ (on the reading where where rajul

‘a man’ is non-specific indefinite)

173
c. hunaaka rajul-un hunaaka

there man-NOM there

‘There is a man (over) there.’30

Third, in declarative sentences, the locative hunaaka ‘there’, unlike true locative adjuncts, cannot

be omitted, as illustrated in (36b). The proposal that the locative pronoun hunaaka and its variant

hunaalika have an existential use in SA is also made in Ryding (2005: 321).

Given the above arguments, I take sentences such as (32) to be projecting two instances

of T, where the lower T has the existential NP in its specifier position, and the higher T hosts the

copular verb. I therefore assume that (32) has the structural representation in (37):

(37) [TP kaan-a [TP hunaaka[NP1] [vP rajul-un[NP2] [vP [PP fii l-daar-i[NP3]]]]]]

was-3MSG there man-NOM in the-house-GEN

‘There was a man in the house.’

The derivation of (37) unfolds as in (38):

(38) a. [vP NP2 v [PP P + NP3]], v a phase head

Spell out v’s complement: <P + NP3>

→ Case NP3 = GEN assigned by P

b. [CP C [TP [TP NP1 [vP NP2 v]]]], C a phasal head

Spell out C’s complement: NP1 c-c’s NP2 → (NP1 < NP2)

30
I thank Youssef Haddad for drawing my attention to examples such as (36c).
174
→ Case NP2 = structural NOM via Agree with T

→ Case NP1 = No case is assigned, as NP1 has a referential index which is not

distinct from its associate31

In (38), v is merged into the structure, and its complement VP is sent to Spell-Out. In this

domain, there is only one NP, namely NP3, which is assigned the genitive case by the

preposition. At the merger of C, its complement TP is sent to Spell-Out. In this domain, NP1 c-

commands NP2. As a result of this relationship, NP1 linearly precedes NP2. However, NP1 is an

expletive, which is not a case competitor, as it has a referential index which is not distinct from

its associate rajul ‘man’. The proposal that the expletive has a referential index which is not

distinct from that of the associate can be exaplained as follows using the adopted theoretical

model. Suppose that the expletive hunaaka ‘there’ has the referential index [k, i] and its associate

rajul ‘man’ has the referential index [n, i]. The intigers k and i of the expletive are equal because

they are in the same index. Similarly, the intigers n and i of the associate are equal because they

are in the same index. By transitivity, the intiger n of the associate is equal to the intiger k of the

expletive. It follows therefore that the referential index of the expletive and that of the associate

are not distinct, and the dependent case does not apply to the associate. Given this state of affairs,

NP2 is assigned the structural nominative case via Agree with T. NP1, being an expletive that

does not trigger the assignment of dependent case, does not receive any case. One piece of

evidence that NP1 does not receive any case comes from the fact that when NP1 is embedded in

a clause, which is introduced by the indicative complementizer, which assigns a purely

31
Chomsky (2000: 122-126) argues that the expletive there has an uninterpretable ɸ-feature, which renders the case
feature for the expletive unnecessary.
175
idiosyncratic lexical accusative case, NP1 does not receive the lexical accusative case. Instead, it

is NP2 that receives the accusative case assigned by the complementizer, as is shown in (39):

(39) ʔinna hunnaka[NP1] rajul-an[NP2]/*rajul-un fii l-daar-i.

that.EMPH there man-ACC/man-NOM in the-house-GEN

‘Indeed, there is a man in the house.’

In (39), it is the NP2 rajul ‘a man’ rather than the NP1 hunnaka ‘there’, which bears the purely

idiosyncratic lexical accusative case assigned by the complementizer.

5.9 Case assignment in sentences with believe-type predicates

The example in (40) is a sentence where the embedded clause is the complement of

believe-type predicates (i.e. exceptional case marking).

(40) ḥasib-tu zayd-ank[NP1] y-uriid-Ø-u

believed-1SG Zayd.ACC 3-want-MSG-INDIC

l-safar-a[NP2].

the-traveling-ACC

‘I believed that Zayd wanted to travel.’

In the literature on believe-type predicates, these verbs are argued to c-select a TP rather

than a CP (see Johnson, 1991, Koizumi 1993, 1995, Runner 1995, 1998, Lasnik 1999, 2001,

176
Chomsky 2000). Thus, while CPs can be pseudo-clefted, TPs cannot. This explains the contrast

found with examples such as (41) (the examples are from Adger 2003: 313)

(41) a. What Medeea arranged was [CP for her children to be poisoned].

b. What Medeea attempted was [CP to poison her children].

c. *What Medeea believed was [TP Jason to be a murderer]

The examples in (41) show that only CPs but not TPs can be pseudo-clefted. The prediction then

is that the clausal complements of believe-type predicates in SA should not be allowed to be

pseudo-clefted if they were TPs. This is borne out, as can be shown by the contrast between CPs

introduced by overt complementizers and sentences with believe-type predicates in (42) and (43)

respectively:

(42) a. nasiy-a zayd-uni [CP ʔan ei y-aqfil-a

forgot-3MSG Zayd-NOM that.SUBJ 3-lock-3MSG.SUBJ

l-ʔabwaab-a].

the-doors-ACC

‘Zayd forgot to lock the doors.’

b. llaðii nasiy-a-hu zayd-uni huwa

that forgot-3MSG-it Zayd-NOM be.3MSG

[CP ʔan ei y-aqfil-a l-ʔabwaab-a].

177
that.SUBJ 3-lock-3MSG.SUBJ the-doors-ACC

‘What Zayd forgot was to lock the doors.’

(43) a. ḥasib-tu zayd-an y-uriid-Ø-u

believed-1SG Zayd.ACC 3-want-MSG-INDIC

l-safar-a.

the-traveling-ACC

‘I believed that Zayd wanted to travel.’

b. *llaðii ḥasib-tu-hu huwa zayd-an/un32

that believed-1SG-it be.3MSG Zayd-ACC/NOM

y-uriid-Ø-u l-safar-a.

3-want-MSG-INDIC the-traveling-ACC

Intended meaning ‘What I believed was that Zayd wanted to tavel.’

The examples in (42) and (43) might suggest that the embedded clausal complements of believe-

type predicates are actually TPs, not CPs. However, Fassi Fehri (2012) shows that these

embedded clauses can still support an independent temporal modifier, as is shown in (44):

(44) ʔamsi ḥasib-tu l-rajul-a y-aktub-Ø-u l-risaal-a

yesterday believed-1SG the-man-ACC 3-write-MSG-INDIC the-letter-ACC

ġad-an.

32
Pseudo-clefted sentences in SA are discussed in Moutaoukil (1985: 27-30).
178
tomorrow-ACC

‘Yesterday, I believed that the man would write the letter tomorrow.’ (adapted from Fassi

Fehri 2012, ex. 58: 249)

Following Fassi Fehri (2012: 249), I therefore assume that the believe-type constructions

start the derivation with two CPs, one of which is later truncated.33

Having established that the embedded clause of the believe-predicate in (40) is a CP,

which is truncated at TP, we now need to consider the position of the embedded subject of the

complement clause of the believe-type predicates. Specifically, does the subject of the embedded

clause remain in the Spec, TP of the embedded clause or does it raise to the object position of the

matrix clause? Fassi Fehri (2012: 249) claims that the subject of the embedded clause raises to

the object position of the matrix clause. In contrast to this position, Soltan (2007: 155-157) and

Al-Balushi (2011: 219-222) both claim that the subject of the embedded clause is in a left-

peripheral position of the embedded clause and does not raise to the matrix clause. For Soltan

(2007), the embedded subject is located in the Spec, TP of the embedded clause, which is a left-

peripheral position for him; For Al-Balushi (2011), the embedded subject is in the Spec, TopP

position of the embedded clause. The claim that the embedded subject in these constructions is in

a left-peripheral position seems, for both of these authors, to be dictated by their other claim that

there is no A-to-A movement in SA. As far as case assignment is concerned, Soltan (2007)

claims that the embedded subject, which is in a left-peripheral position, can still receive a

structural accusative case via an Agree relation with the matrix functional head v*. For

33
Alternatively, it could be the case that the lower CP of an ECM construction is defective in some way. I do not
adopt this option in this thesis.
179
Al-Balushi (2011), on the other hand, the case assigned to the embedded NP in the Spec, TopP

position is lexical rather than structural. The position that I take in this thesis is that of Fassi

Fehri (2012), namely that the embedded subject of these constructions raises to the object

position of the matrix clause. There are three pieces of evidence for this claim. First, the

embedded subject of these constructions is clearly in the matrix clause when the subject is a

pronominal clitic, as the clitic is encliticized to the matrix predicate. This is illustrated in (45):

(45) ḥasib-tu-hu daxal-a l-qaaʕat-a

believed-1SG-him entered-3MSG the-hall-ACC

‘I believed he entered the hall.’ (adapted from Fassi Fehri 2012, ex. 57: 249)

Second, the embedded subject can clearly be passivized, as is shown by the pairs in (46) through

(48); this indicates that it is accessible to the matrix clause T.

(46) a. ḏanan-tu Zayd-an ʔax-aa-ka

beleived-1SG Zayd-ACC brother-ACC-your

‘I believed Zayd to be your brother.’ (Al-Waraaq 10th c./1999: 285)

b. ḏunn-a Zayd-un ʔax-aa-ka

believed.PASS-3MSG Zayd-NOM brother-ACC-your

180
‘Zayd was believed to be your brother.’ (Al-Waraaq 10th c./1999: 285)

(47) a. ḥasib-a Zayd-un Khaled-an šaaʕir-an

believed-3MSG Zayd-NOM Khaled-ACC poet-ACC

‘Zayd believed Khaled to be a poet.’ (Moutaoukil 1987: 103)

b. ḥusib-a Khaled-un šaaʕir-an

believed.PASS-3MSG Khaled-NOM poet-ACC

‘Khaled was believed to be a poet.’ (Moutaoukil 1987: 103)

(48) a. ḏann-a Khaled -un Hind-an musaafir-at-an

believed-3MSG Khaled-NOM Hind-ACC traveling-3FSG-ACC

‘Khaled believed Hind to be traveling.’ (Moutaoukil 1987: 103)

b. ḏunn-at Hind-un musaafir-at-an

believed.PASS-3FSG Hind-NOM traveling-3FSG-ACC

‘Hind was believed to be traveling.’ (Moutaoukil 1987: 103)

Third, Fassi Fehri (1986: 232) shows that the topic hypothesis of the embedded subject of ECM

constructions cannot be correct once we consider the example in (49):

(49) man t-aḍunn-Ø-u Zayd-an ḍarab-a _____

whom 2-believe-MSG-INDIC Zayd-ACC hit-3MSG

181
‘Who do you believe Zayd hit?’ (Fassi Fehri 1986, ex. 136: 232)

On standard assumptions, movement across a topicalized element is barred. However, the

example in (49) shows that movement of the wh-word across the embedded preverbal NP does

not induce ungrammaticality, which suggests that the embedded NP is a subject rather than a

topicalized NP.

Having established that the embedded subject of the complement of believe-type

predicates raises to the object position of the matrix clause in SA, we can now proceed to show

how the sentence in (40) is derived. I assume that (40) has the structural representation in (50):

(50) [CP [TP pro[NP1] ḥasib-tu <pro[NP1]> zayd-an[NP2]

believed-1SG Zayd.ACC

[TP <NP2> y-uriid-Ø-u [vP <NP2> [VP <V> l-safar-a[NP3] ]]]]].

3-want-MSG-INDIC the-traveling-ACC

‘I believed that Zayd wanted to travel.’

The derivation of (50) is as in (51):

(51) a. [VP V NP3]

b. [vP NP2 v [VP V NP3]], embedded v a phase head

Spell out embedded v’s complement: NP3

→ Case NP3 = NA
182
c. [VP V [TP NP2 T [vP <NP2> v [VP V NP3]]

d. [vP NP1 v [VP V NP2 [TP <NP2> T [vP <NP2> v [VP V NP3]]]]], matrix v a phase head

Spell out matrix v’ complement: NP2 c-c’s NP3 → (NP2 < NP3)

→ Case NP2 = 0

→ Case NP3 = ACC

e. [CP C [TP NP1 T [vP <NP1> v [VP V NP2 [TP <NP2> T [vP<NP2> v [VP V NP3]]]]]],

matrix C a phase head

Spell out C’s complement: NP1 c-c’s NP2 → (NP1 < NP2)

NP1 c-c’s NP3 → (NP1 < NP3)

→ Case NP2 = ACC

→ Case NP3 = ACC (redundantly)

→ Case NP1 = structural NOM via Agree with T

In (51), the embedded v is merged into the structure, and the complement domain VP is spelled-

out. In this domain, there is only one NP, namely NP3. No c-command relationships are

considered, hence no case assignment is considered either. The embedded TP is now the

complement of the matrix V, as the CP projection is deleted. At the merger of the matrix v, the

complement matrix VP, which includes in it the embedded TP is now sent to Spell-Out. In this

domain, NP2 c-commands NP3. Therefore, NP2 linearly precedes NP3; the dependent case of

NP3 is also considered, and NP3 is assigned the dependent accusative case. At the merger of the

matrix C, the complement TP is sent to Spell-Out. In this domain, NP1 c-commands NP2, and at

the same time NP1 also c-commands NP3 in this domain, given that the embedded vP is a soft

phase in SA and the fact that both of these c-command relations are new rather than old. The

result of the new c-command relationships is that NP2 is assigned the dependent accusative case,
183
and NP3 is redundantly assigned the dependent accusative case, whereas the matrix NP1 is

assigned the structural nominative case via Agree with T. Note that case assignment in these

structures in the theory of dependent case (Marantz 1991, Baker 2015) differs from how case in

these structures is assigned in theories that depend on a case hierarchy such as that of Yip,

Maling and Jackendoff (1987). Thus, unlike in Yip et al. (1987), where the assignment of

accusative case to the embedded object depends crucially on the embedded subject being

assigned nominative case in the complement cycle, no such assumptions are required in the

dependent case theory. To illustrate this point, let us consider the following example of an ECM

construction:

(52) John[NOM] believed [TP him[ACC] to be stalking her[ACC]].

Recall that in Yip et. al.’s (1987) theory of case tiers, there is a dependency in case assignment

such that the accusative case can only be assigned if nominative case is assigned first. For this

account to work, Yip et. al (1987) assume that the embedded subject Him in (52) receives a

nominative case in the embedded cycle (i.e the embedded clause). It is this case that allows the

accusative case to be associated with the embedded object her. To account for why the

embedded subject of an ECM construction surfaces with an accusative case, Yip et al. (1987)

claim that the embedded clause is defective in some way (specifically S’-deletion) such that the

embedded domain can still be accessed by the matrix cycle (i.e the matrix clause). In this cycle,

the accusative case in the matrix cycle is associated with the embedded subject, and the

nominative case of the embedded subject is therefore overlaid by the accusative case of the

matrix cycle.

In contrast to the Case Tier theory, the adopted theoretical framework does not need to

make extra theoretical claims for the account to work.

184
5.10 Case assignment in sentences with control structures

The issue of control sentences in SA is much more complicated than the way it is

presented here, and semantics seems to be involved; therefore the discussion here is not

exhaustive.34 Let us consider the example in (53), which is an instance of obligatory control in

SA; in particular, obligatory subject control.35

(53) ḥaawal-a l-luṣuuṣ-ui ʔan

tried-3MSG the-thieves-NOM that.SUBJ

y-ahrab-uu-Ø ei/*j.

3-escape-MPL-SUBJ

‘The thieves tried to escape.’36

The question that needs to be answered first is whether the null category of the embedded

clause is the null category PRO, the null category pro, or a lower copy of the matrix subject NP.

This question is important because the verb in SA control constructions is unlike its equivalent in

English control constructions, where the verb is non-finite. As is clear from the example in (53),

34
For example with verbs such as qarrara ‘to decide’ control is not obligatory, as shown by the following examples
(the examples are from Youssef Haddad):
(i) qarrar-a [ʔan t-usaafir-Ø-a bnat-a-hu]
decided-3MSG that.SUBJ F-travel-SG-SUBJ daughter-ACC-his
‘He decided that his daughter should travel.’
(ii) *staṭaaʕ-a [ʔan t-usaafir-Ø-a bnat-a-hu]
managed-3MSG that.SUBJ F-travel-SG-SUBJ daughter-ACC-his
35
Recall that in these structures, the subject NP of the embedded clause is coindexed with an argument of the matrix
clause.
36
The SV order of (53) is possible in which case the subject surfaces with nominative case as well, as showon in (i):
(i) l-luṣuuṣ-u ḥaawal-uu ʔan y-ahrab-uu-Ø
the-thieves-NOM tried-3MPL that.SUBJ 3-escape-MPL-SUBJ
‘The thieves tried to escape.’
185
the embedded verb bears ɸ-features [3MPL] as well as mood features [subjunctive]. However,

Landau (2004) shows that control is independent of finiteness. Thus, Landau (2004) argues that

the Balkan languages (Greek, Romanian, Bulgarian, Albanian) show instances of what Landau

calls ‘finite control’ in the sense that an NP of the matrix clause controls into a finite clause. To

illustrate, let us consider the examples in (54) from Landau (2004, ex. 16: 826-827):

(54) a. I Maria1 prospathise PRO1/*2 na divasi

the Mary tried.3s ec PRT read.3s.Sub

‘Mary tried to read’ (Greek)

b. L1-am indemnat ca de miine

him-I.have urged that from tomorrow

PRO1/*2 sa mearga la scoala cu bicicleta

ec PRT go.3s.Sub to school with the.bike

‘I urged him to ride his bike to school from tomorrow on’ (Romanian)

c. Ivan1 uspja PRO1/*2 da ostane pri nego

Ivan managed.3s ec PRT stay.3s.Sub with him

‘Ivan managed to stay with him’ (Bulgarian)

d. I1 kerkova PRO1/*2 te recitoje nje poezi

him asked.1s ec PRT recite.3s.Sub a poem

‘I asked him to recite a poem’ (Albanian)

186
The examples in (54) show clearly that the null subject PRO is controlled by an NP in the matrix

clause even though the verb of the embedded clause is finite bearing ɸ-features [3 SG] and mood

features [subjunctive]. In other words, finite control in (54) parallels finite control in SA, where

the embedded verb is also marked for ɸ-features and mood [subjunctive] features.

This parallelism might at first sight suggest that SA control structures are indeed

equivalent to their Balkan counterparts. However, a closer look seems to indicate that the null

category in these structures is actually the null pronoun pro rather than the null pronoun PRO.

More specifically, there are other interpretative diagnostics, which show that in SA, the

embedded subjunctive clauses of control verbs include the null category pro. For example,

Landau (2004: 823) states that only PRO supports a sloppy reading under ellipsis. This is not

borne out in SA, as is illustrated by the example of subject control in (55):

(55) a. waʕad-a zayd-uni saalim-anj

promised-3MSG Zayd.M-NOM Saalim.M-ACC

[CP ʔan y-uhaðib-Ø-a ei/*j nafs-a-hui/*j]

that.SUBJ 3-behave-MSG-SUBJ self-ACC-his

‘Zayd promised Saalim that he (Zayd) behave himself.’

wa kaðaalika faʕal-a qays-un.

and so did-3MSG Qays.M-NOM

‘…and so did Qays.’

Sloppy reading or strict reading ‘Qays also promised Saalim that he (Qays/Zayd)

behave himself’

187
The subjunctive complementizer in examples such as (55) is obligatory. This example shows

clearly that the null category can have either a sloppy reading or a strict reading under ellipsis.

This indicates that the null category in these constructions must be pro and not PRO. The same

interpretation is obtained in cases of object obligatory control, where the subjunctive

complementizer is also obligatory, as can be verified by the example in (56):

(56) a. ṭalab-a zayd-unj min qays-ini [CP ʔan

asked-3MSG Zayd-NOM from Qays-GEN that.SUBJ

y-usaaʕid-Ø-a-huj/*i ei/*j]

3-help-MSG-SUBJ-him

‘Zayd asked Qays to help him.’

wa kaðaalika faʕal-at zaynab-u

and so did-3FSG Zaynab-NOM

‘…and so did Zaynab.’

Sloppy reading or strict reading ‘Zaynab also asked Qays to help her/Zayd)’

Having said that, there are clear cases where the subject of the embedded clause is

obligatorily controlled by the subject of the matrix clause. This is illustrated in (57) through (60):

(57) ḥaawal-a l-luṣuuṣ-ui ʔan

tried-3MSG the-thieves-NOM that.SUBJ

ei/*j y-ahrab-uu-Ø <ei/*j>.

188
3-escape-MPL-SUBJ

‘The thieves tried to escape.’

(58) nasiy-a zayd-uni ʔan y-aqfil-Ø-a ei/*j

forgot-3MSG Zayd-NOM that.SUBJ 3-lock-MSG-SUBJ

l-ʔabwaab-a

the-doors-ACC

‘Zayd forgot to lock the door.’

(59) ʔistaṭaaʕ-at l-ṭaalibat-u ʔan t-ajtaaz-Ø-a

managed-3FSG the-student.F-NOM that.SUBJ F-pass-SG-SUBJ

ei/*j l-ʔixtibaar-a

managed-3FSG

‘The female student managed to pass the test.’

(60) ʔamar-a zayd-un l-ʔawlaad-ai ʔan

ordered-3MSG Zayd-NOM the-boys-ACC that.SUBJ

ei/*j y-aktub-uu-Ø <ei/*j> maqaalat-an

3-write-MPL-SUBJ essay-ACC

‘Zayd ordered the boys to write an essay.’

The subject of the embedded clauses in (57) through (60) cannot be controlled by anything other

than the subject or object of the matrix clause. However, given the interpretive diagnostics

offered in Landau (2004: 823), I claim that the embedded subject in the examples in (57) through

(60) is also pro rather than PRO. The third logical possibility that needs to be considered is

189
whether the embedded subject of control sentences is actually a lower copy of the matrix subject,

as proposed in Hornstein (1999 and subsequent work). This claim, however, necessitates that the

subjunctive complementizer of the control strcutures of SA cannot be a phasal head, since case

assignment in the adopted model is determined in the CP domain, if C is phasal. To test whether

the subjunctive complementizer ʔan is a phasal head or not, we use the following test: if the CP

headed by the subjeunctive complementizer can be clefted or moved, it is phasal; if not, then it is

non-phasal.37 Using this diagnostic, we have to conclude that the subjunctive complementizer is

a phasal head, as verified by the following examples

(61) tamanny-tu [ʔan y-usaafir-Ø-a zayd-un]

wished-1SG that.SUBJ 3-travel-MSG-SUBJ Zayd-NOM

‘I wihsed that Zayd would travel.’

(62) a. ʔinna ʔakṯar-a maa tamanny-tu huwa [ʔan

that.EMPH most-ACC what wished-1SG be.3MSG that.SUBJ

y-usaafir-Ø-a zayd-un]

3-travel-MSG-SUBJ Zayd-NOM

‘Indeed, what I wished most was for Zayd to travel,’38

b. [ʔan y-usaafir-Ø-a zayd-un] huwa

that.SUBJ 3-travel-MSG-SUBJ Zayd-NOM be.3MSG

37
I thank Gabriela Alboiu for suggesting this test to me.
38
Using the same diagnostic, one can also show that the indicative complementizer ʔanna is also phasal, as can be
verified by the fact that the whole CP introduced by it can be pseudoclefted:
(i) saʔal-tu zayd-an
asked-1SG Zayd-ACC
‘I asked Zayd.’
(ii) ʔinna aqṣaa maa faʕal-tu huwa [CP ʔanna-nii saʔal-tu zayd-an]
that.EMPH most.ACC what did-1SG be.3MSG that-me asked-1SG Zayd-ACC
‘Indeed, what I mostly did was ask Zayd.’
190
ʔakṯar-u maa tamanny-tu

most-NOM what wished-1SG

‘For Zayd to travel is something that I mostly wished.’

The example in (61) is the unmarked example. The example in (62a) shows that bracketed CP

introduced by the subjunctive complementizer can be pseudoclefted, and the example in (62b)

shows that the CP can be moved to the beginning of the sentence. I take this as evidence that the

subjunctive complementizer ʔan is a phasal head. Given that the subjunctive complementizer ʔan

is a phasal head, I conclude that the subject of the embedded clauses in the control sentences of

SA cannot be a lower copy of the matrix subject, and must therefore be the null subject pro.

Aside from the structures discussed above where the subject of the embedded clause is

pro, Fassi Fehri (1993: 242) claims that SA exhibits cases of control in structures where the

complement of control verbs includes a process (or event-denoting) nominal, as in (63) and (64):

(63) y-uriid-Ø-u ntiqaad-a l-rajul-i

3-want-MSG-INDIC criticizing-ACC the-man-GEN

‘He wants to criticize the man.’ (adapted from Fassi Fehri 1993, ex. 65a: 242)

(64) y-uriid-Ø-u ntiqaad-a nafs-i-hi

3-want-MSG-INDIC criticizing-ACC self-GEN-his

‘He wants to criticize himself.’ (adapted from Fassi Fehri 1993, ex. 65a: 242)

191
I follow Fassi Fehri (1993) on this issue. The question of why the complement of the process

nominal receives a genitive case rather than the predicted accusative case is taken up later when

case assignment in structures with event-deonting nominalizations is discussed.

Having established that the embedded subject of control structures is a lower copy of the

matrix subject , we can now consider how the derivation of both the embedded CP and the

matrix CP in (53) above proceeds. I assume that (53) has the structural representation shown in

(65):

(65) [CP [TP ḥaawal-a [vP l-luṣuuṣ-u[NP1]… [FinP ʔan

tried-3MSG the-thieves-NOM that.SUBJ

[TP <l-luṣuuṣ-u[NP1]> y-ahrab-uu < l-luṣuuṣ-u>[NP1]>]]]].

3-escape-MPL.SUBJ

‘The thieves tried to escape.’

The derivation of (65) proceeds as in (66):

(66) a. [VP2 V]

b. [vP2 NP1 v [VP2 V]], embedded v a phase head

Spell out embedded v’s complement: No NPs to spell out

c. [FinP [TP <NP1> T [vP <NP1> v [VP V]]]], embedded Fin not a phase head

d. [vP1 NP1 v[VP1 V] [FinP [TP <NP1> T [vP <NP1> v [VP V]]]], matrix v a phase head

Spell out matrix v’s complement: NP1

Case NP1 → NA

192
e. [CP1 C [TP NP1 T [vP1 <NP1> v[VP1 V]]], matrix C a phase head

Spell out matrix C’s complement: NP1

→ Case NP1 = structural NOM via Agree with matrix T

In (66), the embedded clause has only one lexical NP, namely the lower copy of the subject ; l-

luṣuuṣ ‘the thieves’, therefore, the dependent case does not apply, and the NP is not assigned the

structural dependent case. The embedded subject raises to the matrix clause; where it does not

receive the dependent case, as it is the only NP in this clause. The subject ends up receiving the

structural nominative case via Agree with the matrix T.

5.11 Case assignment in sentences with adjectival/nominal predicates

Let us consider (67), which is an example of verbless sentences with an

adjectival/nominal predicate:

(67) zayd-un[NP1] saʕiid-un/muhandis-un[NP2].

Zayd-NOM happy-NOM/engineer-NOM

‘Zayd [is] happy/Zayd [is] an engineer.’

The question arises as to whether adjectival/nominal predicates in SA should receive a

structural dependent case. Baker (2015: 221-222), based on robust cross-linguistic evidence,

argues that predicate nominals neither undergo nor trigger the structural dependent case. I follow

Baker on this. Baker (2015, fn. 29: 222) acknowledges that the only empirical exception he finds

is Classical Arabic, where the predicate nominal does receive the accusative case but only when

193
there is a copular verb in the sentence. To account for this fact, Baker claims that Classical

Arabic should be analyzed as a language with no dependent case on predicate nominals, and that

the accusative case assigned to the predicate nominal when there is a copular verb in the sentence

is explained if we consider that “the overt copula (or some functional head associated with it)

assigns accusative case to an NP in its c-command domain under Agree.”

In the generative literature on SA, there are a number of accounts for the case patterns of

adjectival/nominal predicates. Benmamoun (2000: 43) claims that the copular verb assigns the

accusative case to its adjectival/nominal predicate, but does not spell out whether this case is

structural or lexical/inherent. Fassi Fehri (1993: 88) seems to leave it open as to whether this

case is structural or lexical. Fassi Fehri (1993: 88) discusses sentences such as (68):

(68) lays-at Hind-un mariiḍat-an

NEG-3SG Hind-NOM sick-ACC

‘Hind is not sick.’ (Fassi Fehri 1993, ex. 5: 88)

The negative particle in (68) has been argued by traditional Arabic grammarians and by Ouhalla

(1994) to be a complex of the archaic copular be and the negative particle. Therefore, this

negative particle patterns with copular verbs in that it assigns the accusative case to the adjectival

predicate in (68). In his treatment of this case, Fassi Fehri (1993) seems to be ambivalent. Thus,

he writes that “laysa assigns a (morphological) accusative case to the adjective. He goes on to

say that “[t]his case is presumably assigned (under adjacency) by the negative”. On the other

hand, Ouhalla (2005: 681) holds the view that this case is structural rather than lexical. In his

treatment of sentences similar to (68) above, he argues that the adjectival/nominal predicate

receives structural case, “the result of matching [CASE] with the aspectual feature of the
194
[copular] verb”. Baker (2015: 222) claims that the copular verb in SA assigns structural case to

the predicate via Agree. Al-Balushi (2011), on the other hand, argues that the accusative case

assigned by the copular verb is lexical rather than structural. There are reasons to believe that the

accusative case assigned by the copular verb in SA is a syntactic but purely idiosyncratic lexical

case rather than structural case.39 The first piece of evidence comes from the fact that

adjectival/nominal predicates, on standard accounts, are not arguments; therefore, they do not

receive a form of structural case. The second piece of evidence that the accusative case assigned

to adjectival/nominal predicate is a purely idiosyncratic lexical case assigned in the presence of

the copular verb is the fact that in the absence of the copular verb, the adjectival/nominal

predicate receives nominative case rather than accusative case, as is illustrated in (69):

(69) zayd-un mariiḍ-un/muhandis-un/*mariiḍ-an/*muhandis-an

Zayd-NOM sick-NOM/engineer-NOM/*sick-ACC/*engineer-ACC

‘Zayd [is] sick/an engineer.’ (predicational sentence)

The third piece of evidence in support of the claim that the copular verb assigns a purely

idiosyncratic lexical case comes from the fact that in equative sentences, where the pronominal

copula rather than the copular verb is obligataroy, no accusative case is assigned to the second

argument in the sentence, as is illustrated in (70):

(70) zayd-un *(huwa) zaynab-u/*zaynab-a

Zayd.M-NOM be. 3MSG Zaynab.F-NOM/*zaynab.F-ACC

39
Recall that non-structural cases are assigned in the syntax but they are not assigned based on a structural relation
between a functional head and an NP. In woolford’s (2006) terms, these are either internal arguments that receive
purely idiosyncratic case from certin Vs or prepositions, or else they are external arguments that receive therir case
based on their particulat thematic role. It is this reasoning that justifies the term “syntactic but not structural.”
195
‘Zayd is Zaynab (in the context of play)

At this point, the following question might be raised. Since the second NP zaynab ‘Zaynab’ in

the equative sentence of (70) is an argument rather than a predicate, why can’t it receive the

dependent case? The answer to this question can be found in Baker (2015: 174) who states that

“a noun is by definition a lexical category that bears a referential index.” Baker then adds that “it

becomes easy and natural to say that two nominals that interact case-theoretically must not only

have indices, but must have distinct indices.” To account for why the second argument in an

equative sentence does not receive the dependent case, we can follow Baker (2015: 175) who

makes the following theoretical claims:

(71) Suppose that X bears a referential index [i], no other phrase Z properly contains X and also

bears [i], and there is a term Y that bears a distinct index [k] in the same spell out domain as X.

a. Then if X c-commands Y, assign X ergative.

b. Then if X is c-commanded by Y, assign X accusative.

With these theoretical assumptions in mind, we can account for (70) as follows: Baker (2015:

225) proposes that predicate nominals are embedded in a projection which he calls EP (i.e.

equative phrase). Suppose that this is correct. Suppose further that the pronminal copula in SA is

a lexicalization of T, specifically the ɸ-features of T. If this is correct, then the sentence in (70)

has the following syntactic representation:

(72) [CP [TP zayd-un[k] *(huwa) [PredPi [EPi [DPi [NPi zaynab-u]]]]]]

Zayd.M-NOM be. 3MSG Zaynab.F-NOM

‘Zayd is Zaynab (in the context of a play)

196
In (72), the predicate nominal zaynab ‘Zaynab’ has a referential index [i], and it is properly

contained within EP and PredP both of which also bear the referential index [i]; therefore, the

dependent case does not apply by (71) even though the predicate nominal is c-commanded by the

subject zayd ‘Zayd’ which has a distinct referential index [k].

Having established that the accusative case assigned by the copular verb in SA is lexical

rather than structural, we can proceed to account for the case patterns with adjectival/nominal

predicates. Given that the dependent case does not apply in these structures and the fact that they

are not arguments that should agree with T for case-assigning purposes, the prediction is that

these predicates would receive a form of unmarked case, namely nominative in the CP domain. It

turns out that this mechanism does indeed account for the case patterns of adjectival/nominal

predicates in SA. To illustrate, let us consider first, case assignment in nominal sentences

(sentences lacking a lexical verb) together with their structural representation:

(73) a. Zayd-un saʕiid-un/*saʕiid-an.

Zayd-NOM happy-NOM/*happy-ACC

‘Zayd is happy.’

b. [CP [TP Zayd T [PredP <Zayd> Pred happy]

In (73), there is only one argument in the structure, namely zayd ‘Zayd’; therefore, the predicate

nominal receives the unmarked nominative case.

197
When there is an overt copular verb in the structure, this verb assigns a purely

idiosyncratic lexical accusative to the predicate, as is shown by the example in (74a) and its

structural representation in (74b):

(74) a. kaan-a Zayd-un saʕiid-an/*saʕiid-un

was-3MSG Zayd-NOM happy-ACC/*happy-NOM

‘Zayd was happy.’

b. [CP [TP kaana T [TP Zayd-un <kaana> [vP <Zayd> <kaana> v saʕiid-an]

In (74), the representation includes two instances of T, as it has a copular verb, and this kind of

structures has been argued in this thesis to have two projections of TP. Note especially that the

sentence does not include an expletive. This has to do with the fact that unlike with indefinite

subjects with the copular verb, where an expletive is obligatory, no expletive is needed when the

subject is definite, as is the case with (74). SA is a language, where non-specific indefinite

subjects are not allowed to occupy the Spec, TP position (see Fassi Fehri 1993: 28 on this).

When the sentence includes the indicative complementizer, the subject receives a purely

idiosyncratic lexical case from C, and the predicate realizes the unmarked nominative case in the

CP domain, as is shown by the example in (75):

(75) ʔinna Zayd-an saʕiid-un/*saʕiid-an

that.EMPH Zayd-ACC happy-NOM/*happy-ACC

‘Indeed, Zayd is happy.’

198
When there is an indicative complementizer as well as a copular verb, the subject receives an

idiosyncratic lexical case from the complementizer and the adjectival predicate receives an

idiosyncratic lexical case from the copular verb, as in (76):

(76) ʔinna Zayd-an kaan-a saʕiid-an/*saʕiid-un

that.EMPH Zayd-ACC was-3MSG happy-ACC/*happy-NOM

‘Indeed, Zayd was happy.’

When the nominal sentence is embedded in an ECM construction, the adjectival/nominal

predicates obligatorily surface with an accusative case. To account for this case pattern, I claim

that the adjectival/nominal predicate is assigned a purely idiosyncratic lexical accusative case by

the covert analogue of the copular verb. This is illustrated by (77a) and its structural

representation in (77b):

(77) a. ḥasib-tu Zayd-an saʕiid-an/*saʕiid-un

believed-1SG Zayd-ACC happy-ACC/*happy-NOM

‘I believed Zayd to be happy.’

b. ḥasib-tu Zayd-an muhandis-an/*muhandis-un

believed-1SG Zayd-ACC engineer-ACC/*engineer-NOM

‘I believed Zayd to be an engineer.’

c. …[vP v Zayd-ACC] [TP <Zayd> <kaana> [TP <Zayd> <kaana> [vP <kaana>
saʕiid-an / muhandis-an-ACC]]]

199
The question that might now be raised is the following: why should the copular verb be

covert only in embedded contexts?40 To answer this question, I tentatively assume that this is so

because the overt copular verb encodes tense, and in the context of the ECM constructions, the

tense of the embedded clause can but need not be encoded by the matrix predicate; therefore, the

use of of the overt copular verb to encode tense in the embedded clause becomes redundant.41

To summarize, the adjectival and nominal predicates do not undergo the dependent case

assignment. Instead, they surface with the unmarked nominative case unless there is an overt or

covert copular verb, which assigns a purely idiosyncratic lexical accusative to the predicate.

5.12 Case assignment in passive sentences

Compare the pair in (78):

(78) a. katab-a zayd-un risaalat-an

wrote-3MSG Zayd-NOM letter-ACC

40
I thank Arsalan Kahnemuyipour for raising this question. I realize that redundancy does not entail impossibility,
and the sentence with an overt copula is highly improbable if not ungrammatical in SA. The only instance of the
copular verb preceding a belief-type predicate that I managed to find is the following verse from (Ibn ʕaqiil 13th
c./1980, Vol. 2, ex.123: 35):
(i) ʔin t-azʕum-ii-ni kun-tu ʔajhal-Ø-u…
if F-believed-FSG-me was-1SG lack.knowledge-MSG-INDIC
‘If you believed that I lacked knowledge…’
41
Ouhalla (1994) attempts to account for the accusative case on the adjectival/nominal predicate in the context of
ECM constructions by claiming that the embedded clause forms a small clause, which raises to the matrix clause,
and the whole clause receives an accusative case from the matrix predicate. This analysis begs the question of why
should small clauses, which are predicates, need case. Fassi Fehri (1993, fn. 24: 92) accounts for the accusative case
on the adjectival predicate in the context of ECM constructions by using the notion of spreading, i.e. the accusative
case assigmed to the subject of the embedded clause spreads to the adjectival predicate (cf. Yip et. al. 1987). Al-
Balushi (2016: 26) claims that the embedded adjectival/nominal predicate in an ECM construction is assigned
lexical accusative case by the matrix ECM predicate.
200
‘Zayd wrote a letter.’

b. kutib-at risaalat-un/*risaalat-an

wrote.PASS-3FSG letter-NOM/letter-ACC

‘A letter was written.’

Suppose that the structure of (78b) is as shown in (79):

(79) [CP [TP kutib-at T [vP v [VP <V> risaalat-un]]]]

The example in (78a) is in the active form, and the theme object receives a structural dependent

accusative case. When the theme object is passivized, as in (78b), it surfaces as the subject of the

sentence with nominative case. The facts of (78b) can be accounted for using the dependent case

theory of Baker (2015) as follows: at the insertion of the soft phasal head v, its complement VP

is sent to Spell-Out. In this domain, there is only one NP, namely the theme object; therefore, no

case assignment is calculated. At the insertion of the strong phasal head C, its complement TP is

sent to Spell Out. In this domain, there is only NP, namely the theme object; therefore, the

depdendent case does not apply, and the theme object surfaces with the structural nominative

case via Agree with T. Note crucially that at the Spell Out of vP, no case applies if there is only

one NP in the spell out domain. At the Spell out of CP, case has to aply even there is only one

NP in the spell out domain. This is accounted for on the grounds in the adopted theoretical model

201
on the grounds that vP is a soft phase; therefore, there is still a chance for the case to compete for

other NPs for case; at the CP level, however, case has to apply even if there is one NP, as this is

a hard phase, and all case decions have to be made in this phase.

5.13 Case assignment in imperative sentences

The example in (80) is a sentence in the imperative form in SA:

(80) ktub-Ø-Ø risaalat-an/*risaalat-un.

write.IMP-MSG-JUSS letter-ACC/letter-NOM

‘Write a letter!’

In (80), the theme object obligatorily surfaces with the accusative case. Following Benmamoun

(2000) and Al-Balushi (2011), I assume that imperative sentences in SA do not have a Tense

specification. Suppose that the structural representation of (80) is as shown in (81):

(81) [CP [TP pro ktub [vP <pro> v [VP <V> risaalat-an]]]

Following Benmamoun (2000), I assume that T in imperative sentences does not have a Tense

specification, but is rather specified for an [IMP(erative)] feature. The fact that the theme object

in (80) obligatorily surfaces with the accusative case indicates that the null subject pro of

202
imperative sentences is a case competitor in that it triggers the structural dependent accusative

case on the theme object. Recall that I have agued that a pro subject always triggers the

dependent accusative case on the object in SA. The example in (80) can be accounted for using

the dependent case theory of Baker (2015) as follows: at the insertion of the soft phasal head v,

its complement VP is sent to Spell-Out. Howerver, there is only one NP in this domain, namely

the theme object; therefore no case assignment is calculated. At the Spell-Out of the strong

phasal head C, its complement TP (which includes a mood feature but no tense feature) is sent to

Spell-Out. In this domain, there are two NPs available, the theme object and the null subject pro.

As a result, the theme object is assigned the dependent accusative case, and the null subject is

assigned the structural nominative case via Agree with T. Crucially, note that T engages with the

NP in an Agree relation and the null subject pro does receive its nominative case even though T

is not specified for Tense here. This is accounted in the adopted theoretical framework on the

grounds that any feature in T, be it tense, mood or aspect can check the nominative case on the

subject (see point (e) in section 3.4.3 on the theoretical assumptions adopted in this thesis).

5.14 Case assignment in sentences with participials

According to Hasan (1963: 238-245), present participles in SA are derived from verbs

and obligatorily encode the non-past (i.e. either deictic present or future) temporal specification.

This is illustrated in (82):

(82) zayd-un[NP1] qaariʔ-un l-kitaab-a[NP2].

Zayd-NOM reading.PTPL-3MSG.NOM the-book-ACC

203
‘Zayd is reading the book now.’

‘Zayd will be reading the book sometime in the future’

‘*Zayd was reading the book in the past.’

Fassi Fehri (1993:181-184) provides a number of arguments to show that participles in

SA do not bear a Tense specification. In terms of their aspectual properties, Fassi Fehri (1993:

178-181) shows that participles are similar to adjectives in some respects, but similar to verbs in

other respects. Thus, similar to adjectives, participles encode states. However, the states that

participles encode are not the same as those that adjectives encode. Specifically, adjectives

encode non-contingent states, whereas participles encode contingent states. Thus, adjectives such

as ġariiq ‘drowned’ is not the same as the active participle ġaariq ‘drowning’. The adjective

ġariiq ‘drowned’ describes someone who is in the state of drowning, and there is no attention

paid to how the event is unfolding in time. In contrast to this, the participle ġaariq ‘drowning’

means that someone is drowning now but may not be after a while, or that someone has entered

the state of drowning (i.e. inceptive aspect). Similarly, participles can even be derived from

purely stative lexical roots, such as faahim ‘understanding’ from fahima ‘to understand’, ʕaalim

‘knowing’ from ʕalima ‘to know’. However, even here, the participles faahim ‘understanding’

and ʕaalim ‘knowing’ encode a transitional or temporary state of understanding or knowing

rather than a permanent state. In addition to their ability to express states, participles also show

other properties that are characteristic of adjectives. For example, they bear case, as shown in

(85b) above. They also agree in gender and number with the subjects. This is illustrated by the

comparison shown between (86) and (87):

204
(86) a. l-ʔawlaad-u muhaððab-uun.

the-boys-M.PL.NOM well-mannered-MPL.NOM

‘The boys are well-mannered.’

b. l-banaat-u muhaððab-aat-un.

the-girls.FPL-NOM well-mannered-FPL-NOM

‘The girls are well-mannered.’

(87) a. l-ʔawlaad-u ʕaarif-uun l-jawaab-a.

the-boys.MPL-NOM knowing.PTPL-MPL.NOM the-answer-ACC

‘The boys are now in the stage of knowing the answer.’

b. l-banaat-u ʕaarif-aat-un l-jawaab-a.

the-girls.FPL-NOM knowing.PTPL-FPL-NOM the-answer-ACC

‘The girls are now in the stage of knowing the answer.’

Moreover, Fassi Fehri (1993: 187) shows that similar to adjectives, participles occur in positions

that are occupied by adjectives. This is shown by the comparison in (88):

(88) a. daxal-a zayd-un l-bayt-a ḥaziin-an

entered-3MSG Zayd-NOM the-house-ACC sad-ACC

‘Zayd was sad when he entered the house.’

b. daxal-a zayd-un l-bayt-a raakib-an

entered-3MSG Zayd-NOM the-house-ACC riding.PTPL-ACC

ḥiṣaan-an

horse-ACC

205
‘Zayd entered the house riding a horse.’

c. daxal-a zayd-un l-bayt-a mumtaṭiy-an

entered-3MSG Zayd-NOM the-house-ACC mounting.PTPL-ACC

ḥiṣaan-an

horse-ACC

‘Zayd entered the house mounting a horse.’ (Fassi Fehri 1993, ex. 132: 187)

The example in (88a) shows that an adjective modifies the subject of the sentence. The examples

in (88b-c) show that this position can also be occupied by the active and passive participles

respectively.

On the other hand, participles show other properties which are similar to those of verbs.

Thus, similar to fully inflected verbs, participles have an argument structure, which parallels that

of their corresponding verbs, and their complements are marked with the accusative case. This is

shown by the comparison in (89) and (90):

(89) a. ʕaraf-a zayd-un l-jawaab-a

knew-3MSG Zayd-NOM the-answer-ACC

‘Zayd knew the answer.’

b. zayd-un ʕaarif-un l-jawaab-a.

Zayd-NOM knowing.PTPL-NOM the-answer-ACC

‘Zayd is now in the state of knowing the answer.’

206
(90) a. salab-a zayd-un l-rajul-a maal-a-hu

deprived-3MSG Zayd-NOM the-man-ACC money-ACC-his

‘Zayd deprived the man of his money.’

b. zayd-un saalib-un l-rajul-a maal-a-hu

Zayd-NOM depriving.PTPL-NOM the-man-ACC money-ACC-his

‘Zayd is now in the state of deprving the man of his money.’

The example in (89b) shows that a participle may have a theme object as its complement in the

same way that a monotransitive verb can in (89a). Similarly, the theme object of the participle in

(89b) is also marked in the accusative case in the same way that the theme object of the verb in

(89a) is. The example in (90b) shows that a participle may have two objects as complements in

the same way that its corresponding verb can in (90a). Also, the objects of the participle are both

marked in the accusative case in the same way that the objects of a ditransitve verb are.

Moreover, as pointed out in Fassi Fehri (1993: 186-187), participles have the same selectional

properties that their corresponding verbs have. This is illustrated by the comparison in (91):

(91) a. zayd-un y-uʕraf-Ø-u [PP bi-nazaahat-i-hi]

Zayd-NOM 3-know.PASS-MSG-INDIC for-honesty-GEN-his

‘Zayd is known for his honesty.’

b. zayd-un maʕruuf-un [PP bi-nazaahat-i-hi]

Zayd-NOM known.PTPL-NOM for-honesty-GEN-his

‘Zayd is known for his honesty.’

207
The examples in (91) show that the participle in (91b) c-selects the same PP complement that its

corresponding verb does in (91a).

Similar to Fassi Fehri (1993), Al-Balushi (2011: 262-267), also argues that participle

sentences lack a tense specification, and the proposal is made that such sentences project an

AspP instead of TP.

Given the above hybrid categorial properties of participles, Fassi Fehri (1993: 187-190)

proposes that participles are internally verbal but externally adjectival. He further proposes that

the categorial conversion takes place in syntax. I adopt this proposal, together with the claim that

T in these structures encodes aspect but not tense. I therefore propose the structure in (92) for

participles:

(92): CP
3
C TP
3
DP T’
3
T AP
3
A vP
3
<DP> v’
3
v VP
3
<V> …

The structure in (92) shows that the participle starts the derivation as a verb, before it is

converted to an adjective later on in the derivation. The participle head-moves from V to v and

then to A. The subject DP of the participle moves from its base-generation position in Spec, vP

before moving moving to Spec, TP. Based on Fassi Fehri (1993) T may bear mood, aspect or

tense features. Thus, T in the structure in (92) has aspect features but no tense features. What the
208
structure in (92) shows is that sentences with participles are actually verbless sentences in that

the predicate is not externally a verbal one but rather an adjectival one.

Having established the temporal, aspectual and categorial properties of participles, let us

consider how sentences with particples can be accounted for in the dependent case theory of

Baker (2015). Let us consider first the example in (93):

(93) [CP C [TP zayd-un[NP1] [AP qaariʔ-un-A [vP <NP1> <v>

Zayd-NOM reading.PTPL-3MSG.NOM

[VP <V> l-kitaab-a[NP2]]]]]].

the-book-ACC

‘Zayd is reading the book.’

The sentence in (93) therefore has the derivation in (94):

(94) a. [vP <NP1> v [VP V NP2]], v a phase head

Spell out v’s complement: NP2

→ Case NP2 = NA

b. [CP [TP NP1 T [AP [vP <NP1> v [VP V NP2]]]]], C a phase head

Spell out C’s complement: NP1 c-c’s NP2

→ (NP1 < NP2)

→ Case NP2 = ACC

→ Case NP1 = structural NOM via Agree with T

209
In (94), vP is projected, and its complement VP is sent to Spell-Out. In this spelled out domain,

there is only one NP, namely NP2, which is not considered for case assignment, as it is the only

NP in this domain. At the spell out of C, its complement TP is sent to Spell-Out. In this spelled

out domain, NP1 c-commands NP2. As a result of this relationship, NP1 linearly precedes NP2,

and NP2 is assigned the structural dependent accusative case. NP1 receives the structural

nominative case via Agree with T. Given that the participle is externally an adjectival predicate,

it follows that the participle receives the unmarked nominative case in the CP domain. The fact

that the participle receives the unmarked case even though it is externally an adjectival predicate

can be attributed to a PF requirement of SA, which requires that adjectival predicates bear a

morphological form of case.

Let us consider next how the sentence with auxiliaries such as (95) can be accounted for

in this theory.

(95) [CP C [TP kaan-at [TP l-ʔamṭaar-u[NP1] <kaan-at> ʔamsi

was-3FSG the-rains-NOM yesterday

[AP ġaasil-at-an-A [vP <NP1> v [VP V l-ʔašjaar-a[NP2]]]]]]].

cleaning.PTPL-FSG-ACC the-trees-ACC

‘The rains were washing the trees yesterday.’ (Hasan 1963: 248)

Here, I assume that T has a tense specification given that the copular verb bears tense

information. I also assume that there are two instances of T given the presence of the copular

verb. The derivation of the sentence in (95) proceeds as in (96):

210
(96) a. [vP <NP1> v [VP V NP2]], v a phase head

Spell out v’s complement: NP2

→ Case NP2 = NA

b. [CP C [TP kaanat T [TP NP1 <kaanat> T [AP V-v-A [vP <NP1> v [VP V NP2]]]]]], C a

phase head

Spell out c’s complement: NP1 c-c’s NP2

→ (NP1 < NP2)

→ Case NP2 = ACC

→ Case participial = a purely idiosyncratic lexical ACC assigned by the copular

verb

→ Case NP1 = structural NOM via Agree with T.

In (96), vP is projected and its complement VP is sent to Spell-Out. In this domain, there is only

one NP, namely NP2, which is not considered for case assignment, as it is the only NP in this

domain. At the Spell Out of the phasal head C, its complement TP is sent to Spell-Out. In this

domain, NP1 c-commands NP2. As a result of this relationship, NP1 linearly precedes NP2 and

NP2 is assigned the structural dependent case. The participial is assigned the purely idiosyncratic

lexical accusative case assigned by the copular verb. As for NP1, it receives the structural

nominative case via Agree with T.

5.15 Case assignment in the PP domain

In SA, NPs embedded in the prepositional phrases (PPs) receive structural genitive case

assigned by the preposition (more on this below). This is illustrated in (97):

211
(97) a. saafar-at zaynab-u ʔilaa l-ʕiraaq-i

travelled-3FSG Zaynab-NOM to the-Iraq-GEN

‘Zaynab travelled to Iraq.’

b. qadim-at zaynab-u min l-maġrib-i

came-3FSG Zaynab-NOM from the-Morocco-GEN

‘Zaynab came from Morocco.’

c. ʕaad-at zaynab-u li-l-laʕib-i

returned-3FS Zaynab-NOM to-the-playing-GEN

‘Zaynab returned to playing.’

d. takallam-at zaynab-u ʕan l-faqr-i

spoke-3FSG Zaynab-NOM about the-poverty-GEN

‘Zaynab spoke about poverty.’

e. saqaṭ-at zaynab-u ʕalaa l-ʔarḍ-i

fell-3FSG Zaynab-NOM on the-floor-GEN

‘Zaynab fell on the floor.’

f. ʔintaḍar-at zaynab-u ʕinda l-baab-i

waited-3FSG Zaynab-NOM at the-door-GEN

‘Zaynab waited at the door.’

g. ḥallaq-at l-ṭaaʔirat-u fawqa l-saḥaab-i

flew-3FSG the-plane-NOM over the-clouds-GEN

‘The plane flew over the clouds.’

h. takallam-at zaynab-u maʕa l-muʕallim-i

spoke-3FSG Zaynab-NOM with the-teacher-GEN

212
‘Zaynab spoke with the teacher.’

i. ʔiltaq-at zaynab-u bi-zayd-in

met-3FSG Zaynab-NOM with-Zayd-GEN

‘Zaynab met with Zayd.’

j. jalas-at zaynab-u bayna l-muʕallimat-i wa

sat-3FSG Zaynab.F-NOM between the-teacher.F-GEN and

l-ṭaalibat-i

the-student.F-GEN

‘Zaynab sat between the teacher and the student.’

There is, however, a class of nouns known in the literature on SA as diptotes, which can

only surface with a morphologically accusative case following a preposition. Diptotes are a class

of nouns whose declension is characterized by the two properties shown in (98):

(98) a. They are declined with the accusative case when they are indefinite and when

they function as the complement of a preposition.

b. They lack the indefinite marker –n when they are used as indefinite.42

There are a number of factors which determine when a noun is a diptote. The following are some

of these factors. According to Ibn Al-Saraaj (10th c./1996: 80-92), a noun is considered to be a

diptote whenever two of any of the following factors are available, or whenever one factor is

repeated more than once with one noun.

42
The status of the final -n, called the tanwiin or nunation in traditional grammars is not resolved. Kouloughli
(2007) defends the thesis that it is an indefinite marker. Fassi Fehri (2012, fn 2: 294), however, argues that “it is the
head of Poss(essive) phrase, which marks the absence of the possessor constituent, or absence of individuation.” For
the purpose of this study, I treat it as an indefinite morpheme.
213
(99) a. The noun is of the template ʔafʕal, as in the proper name ʔaḥmad.

b. The noun is feminine either morphologically, as in the proper name ḥamdat or

it is made up of more than three radicals (i.e. consonants), the second of which is

not followed by a vowel and it is inherently feminine, as in the proper names

Zaynab or Suʕaad.

c. The noun ends in –an, as in the proper name ʕuṯmaan.

d. The noun is definite.

e. The noun is derived from another noun resulting in a change in meaning.

f. The noun is plural, which is derived from another plural noun, such that no other

plural can be derived from the newly derived plural.

g. The noun is a foreign proper name.

h. The noun is a compound noun formed out of two simple nouns.

To illustrate, the proper name ʔaḥmad qualifies as a diptote because it is of the templatic form

ʔafʕal and it is definite. Similarly, the proper name ḥamdat is a diptote because it is

morphologcially feminine bearing the feminine suffix -at and it is definite. The proper name

ʕuṯmaan is a diptote because it ends in -an and it is definite The noun maṯnaa ‘two two in a raw’

is a diptote because it is derived from the noun ʔiṯnayn ‘two’ thus changing its templatic form,

and the derived noun maṯnaa also changes the meaning ʔiṯnayn ‘two’ into ‘two two in a raw’.

Thus, with maṯnaa ‘two two in a row’, the same factor, change, is repeated twice. The noun

ʔakluban ‘dogs’ is the plural form of kalb ‘dog’, and the noun ʔakaalib is a diptote because it is

the plural form of ʔakluban, the latter itself is the plural form of kalb ‘dog’. Here also, the same

factor, pluralization, is repeated twice.The fact that no other plural can be formed out of ʔakaalib

makes it qualify to be a diptote. The nouns ʔibraahiim ‘Ibraham’, ʔisḥaaq ‘Isac’, yaʕquub

214
‘Jacob’ are diptotes because they are foreign names and they are definite. The proper nouns

ḥaḍramawta and baʕlabakka are diptotes because both are compounds, one is derived from

ḥaḍram ‘Hadoram’ and mawta ‘land’, and the other is derived from baʕl ‘owner’ and baq

‘Biqaaʕ’ meaning ‘plain’ respectively.43

The examples in (100) contain an example of a diptote, namely Zaynab ‘Zaynab’:

(100) a. jaaʔ-at zaynab-u-Ø

came-3FSG Zaynab-NOM-INDEF

‘Zaynab came.’

b. raʔay-tu zaynab-a-Ø

saw-1SG Zaynab-ACC-INDEF

‘I saw Zaynab.’

c. marar-tu bi-zaynab-a-Ø

passed-1SG by-Zaynab-ACC-INDEF

‘I passed by Zaynab.’

The examples in (100) show that the proper noun Zaynab ‘Zaynab’, a member of the minority

class of nouns known as diptotes in SA, declines in the following manner: the proper noun is

assigned morphological nominative case when it functions as the subject of the sentence, as in

(100a); it is assigned a morphological accusative case when it functions as the object of the

sentence (100b) as well as when it functions as the object of a preposition, as in (100c). In

43
There is no agreement on the original meaning of ḥaḍramawta. The meaning I ascribe to it here is based on just
one of the possible original meanings ascribed to it.
215
addition to the case endings, the diptote Zaynab ‘Zaynab’ lacks the indefinite marker -n, which is

normally attached to proper nouns in SA (for a discussion of diptotes in SA, see Al-Hawary

2011: 343-348).

Diptotes are to be distinguished from another declensional class of nouns, which make up

the majority of nouns in SA, namely triptotes. Triptotes have distinct case markings in the

nominative, accusative and genitive, and they bear the indefinite maker –n when used as

indefinite. The examples in (101) contrast with the diptotes shown in (100):

(101) a. jaaʔ-a zayd-u-n

came-3MSG Zayd-NOM-INDEF

‘Zayd came.’

b. raʔay-tu zayd-a-n

saw-1SG Zayd-ACC-INDEF

‘I saw Zayd.’

c. marar-tu bi-zayd-i-n

passed-1SG by-Zayd-GEN-INDEF

‘I passed by Zayd.’

The examples in (101) show that the proper noun, zayd ‘Zayd’, a member of the majority

declensional class known as triptotes in SA declines in the following manner: it is assigned a

morphological nominative case when it functions as the subject of the sentence, as in (101a); it is

assigned a morphological accusative case when it has the function of the object, as in (101b), and

216
it is assigned a morphological genitive case when it functions as the object of a preposition, as in

(101c).

Having described the declensional properties of triptotes and diptotes in SA, I now

address the question of case marking on the complements of prepositions. Specifically, is the

genitive case assigned by prepositions to their complements a structural or a lexical/inherent

case? I claim that the genitive case assigned by the prepositions in SA is a structural case. This

claim is supported by two arguments against the view that this case is lexical/inherent. The first

argument against the view that the genitive case assigned by prepositions is an inherent case is

that objects of prepositions all receive a genitive case regardless of their thematic role in the

sentence. This is shown by the examples (97) above and further in (102):

(102) a. saafar-at zaynab-u ʔilaa l-ʕiraaq-i

travelled-3FSG Zaynab-NOM to the-Iraq-GEN

‘Zaynab travelled to Iraq.’

b. qadim-at zaynab-u min l-maġrib-i

came-3FSG Zaynab-NOM from the-Morocco-GEN

‘Zaynab came from Morocco.’

c. ʔaʕṭ-at zaynab-u l-kitaab-a li-zayd-in

gave-3FSG Zaynab-NOM the-book-ACC to-Zayd-GEN

‘Zaynab gave the book to Zayd.’

d. zaynab-u fii l-maġrib-i

Zaynab-NOM in the-morocco-GEN

‘Zaynab [is] in Morocco.’

217
The examples in (102) clearly show that the genitive case assigned by the prepositions to their

complements is not thematically linked. Thus, the NP l-ʕiraaq ‘Iraq’ in (102a) receives the

genitive case from the preposition ʔilaa ‘from’, and the NP has the thematic role GOAL.

Similarly, the NP l-maġrib ‘Morocco’ in (102b) receives the genitive case from the preposition

min ‘from’, and the NP has the thematic role SOURCE. In (102c), the NP Zayd ‘Zayd’ receives

the genitive case from the preposition li ‘to’, and the NP has the thematic role BENEFICIARY.

In (102d), the NP zaynab ‘Zaynab’ receives the genitive case from the preposition fii ‘in’, and

the NP has the thematic role LOCATION.

A second argument in support of the claim that prepositions in SA assign structural

genitive case to their complements rather than a purely idiosyncratic lexical case in the sense of

Woolford (2006) is that the assignment of genitive case is a property that applies to all

prepositions in SA, not just to particular ones, as shown by the examples in (97) and (102) above,

the only apparent exception noted is the case of the minority class of declensional nouns known

as diptotes, as discussed. To account for the morphological accusative case marking on diptotes

following a preposition, I follow Embick and Noyer’s (2005) analysis of diptotes in SA. In their

Distributed Morphological (DM) analysis of diptotes in SA, Embick and Noyer (2005: 17-20)

assume that the case features for SA are those in (103):

(103) Case features for SA

NOM ACC GEN

Oblique - - +

Superior + - -

218
Given the case features in (103), Embick and Noyer assume that the Vocabulary Items in

(104) are in a competition to realize the case morphemes of SA:

(104) -u ↔ [+superior]

-i ↔ [+oblique]

-a ↔ elsewhere

For the indefinite morpheme, Embick and Noyer assume that the Vocabulary Items in

(104) compete for insertion, as in (105):

(105) -n ↔ [-definite]

-Ø ↔ elsewhere

To account for the indefinite and case declensional patterns of diptotes in SA shown in

(100) above, namely the syncretism of genitive and accusative into the accusative form -a, and

the lack of the indefinite marker –n, Embick and Noyer claim that the Impoverishment rules in

(106), which precede the process of Vocabulary Insertion in the Morphophonological Structure

of DM, are responsible for the surface realization of these morphemes on diptotes in SA:

(106) SA diptote Impoverishement

a. [+oblique] → Ø/ [diptote] + ___ + [- definite]

b. [- definite] → Ø/ [diptote] + case/number + ___

219
According to Embick and Noyer, the Impoverishement rule in (106a) prevents the insertion of

the genitive morpheme –i on diptotes in SA, and the Impoverishement rule in (106b) prevents the

insertion of the indefinite morpheme on dipototes in SA.

Given Embick and Noyer’s (2005) analysis of the case and indefinite patterns of the

declensional class of diptotes in SA, it follows that prepositions always assign a structural

genitive case to their complements in the syntax. In the case of the accusative case on diptotes

following a preposition, this is due to an Impoverishment rule in the Morphophonological

component, which maps the syntactic component to the PF interface level. I take the fact that

accusative case rather than nominative case is assigned to diptotes in the PP domain as an

indication that accusative case is the unmarked case in this domain in SA. The proposal that

accusative case is the unmarked case in the PP domain is supported by Baker’s (2015: 296)

conjecture that

“[s]ome languages might have a special unmarked case triggered by P heads

(oblique in the Hindi sense), which may or may not be the same as nominative-

absolutive and/or genitive.”

One piece of evidence that the case assigned by prepositions to diptotes is syntactically genitive

but morphophonologically accusative can be gleaned from the agreement that takes place

between NPs and their modifying adjectives, which is morphologically overt in SA. It is a fact of

SA that adjectives agree with their head nouns in number, gender and case. The example in (107)

clearly shows that when an NP with a modifying adjective and a head noun, which is a diptote, is

assigned a structural genitive case by the preposition, the modifying adjective, which is not a

220
diptote in this case, surfaces with a genitive case and an indefinite morpheme even though the

head noun bears a morphological accusative case with no indefinite morphemes.44

(107) marar-tu bi-talaamiið -a judud-i-n

passed-1SG by-students-ACC new-GEN-INDEF

‘I passed by new students.’

The example in (107) shows unambiguously that the object of the preposition bears a structural

genitive case but a morphological accusative case, given that adjectives lack inherent ɸ-features

and simply copy the ɸ-features of their head nouns.

The question that remains to be answered is why the NP inside the PP domain is immune

to the assignment of structural dependent accusative case. I claim that this is due to the fact that

PPs in SA are hard phases in the sense of Baker (2015 and references therein) (on the view that

PPs can be phases in some languages, see Baker 2015: 81). One piece of evidence that PPs in SA

are hard phases comes from extraction structures. In SA, the NP complements cannot be

extracted out of the PP domain. In cases of extraction, the extracted element has to pied-pipe the

PP as a whole in order for the sentence to be grammatical. This is shown in (108):

(108) a. saafar-at zaynab-u ʔilaa l-ʕiraaq-i

travelled-3FSG Zaynab-NOM to the-Iraq-GEN

‘Zaynab travelled to Iraq.’

There are other adjectives, which are also diptotes. An example is the plural feminine adjective ʔuxar ‘other’ (Al-
44

Hawary 2011: 347).

221
b. ʔilaa l-ʕiraaq-i saafar-at zaynab-u ___

to the-Iraq-GEN travelled-3FSG Zaynab-NOM

‘It was to Iraq that Zaynab travelled.’

c. *l-ʕiraaq-i/u/a saafar-at zaynab-u ʔilaa ___

the-Iraq-GEN/NOM/ACC travelled-3FSG Zaynab-NOM to

Intended meaning ‘It was to Iraq that Zaynab travelled.’

The examples in (108) show that only when the whole PP is fronted is the sentence grammatical,

as is shown in (108b). When the NP inside the PP is extracted out of the PP, the sentence is

ungrammatical, as is shown in (108c). This, in addition to the fact that the NPs inside the PP

domain are not subject to the structural dependent case, might be taken as evidence that PPs in

SA are hard phases in the sense of Baker (2015).

5.16 Summary

In this chapter, I have shown that the case assignment facts of the core arguments in SA

can be properly accounted for using the dependent case theory of Baker (2015). The crucial

claim has been that v in SA is a soft phase in the sense of Baker (2015). The phasal head v has

been shown to be incapable of assigning a structural accusative case to the object. This has been

attributed to v’s being impoverished having only a gender feature. The structural accusative case

on the object has been argued to be the result of the dependent case assignment. The chapter has

also shown that the subject (i.e. the highest argument in Spec, vP or Spec, TP) receives its

structural nominative case via Agree with T. This mechanism of case assignment has been

argued to be available when it is not bled by dependent case assignment. This chapter has also

shown that purely idiosyncratic lexical case does apply in SA, and that that there are two lexical
222
case assigners. The first lexical case assigner in SA is the indicative complementizer. The second

lexical case assigner is the copular verb. These lexical case assigners have been shown to assign

a purely idiosyncratic lexical accusative case to NPs in their complements. The chapter has

moreover shown that NPs inside the PP domain receive a form of structural genitive case

assigned by P. Crucially, the chapter also shows that for NPs to compete for the dependent case,

they have to have distinct referential indices.

Based on the results of the analysis conducted, I have shown that the following properties are

characteristic of SA:

(a) The CP and PP are hard phases, but the vP is a soft phase.

(b) The null category pro is a case competitor in SA in that it triggers the assignment of

structural dependent accusative case on the object DP in the sentence.

(c) NOM can be the result of an Agree relation with T, or it can be the result of the assignment

of unmarked/default case in the CP domain; GEN is a structural case assigned by D to its

complement in the DP domain, and ACC is the unmarked case in the PP domain.

The goal of this chapter has been to show that in order for the structural dependent case to

apply in a spell out domain, the NPs involved must have distinct referential index. Thus, a there-

type expletive is not capable of triggering the structural dependent case on an NP because its

referential index is not distinct from that of its associate. Similarly, the lower argument in an

equative sentence does not get the dependent case. Thus, although it has a referential index, it is

not the only maximal projection bearing that index.

223
Chapter Six

Case assignment of non-arguments in SA

6.1 Introduction

In this chapter, I will discuss the case assignment of adverbial NPs and NPs in the left

periphery of the clause. I will claim that all non-arguments receive a form of non-structural non-

syntactic case referred to as semantic case (see Maling 2009, Schütze 2001, Fassi Fehri 1986),

which is homophonous to the structural accusative case.45

6.2 The case assignment of adverbial NPs

Let us consider the examples of adverbial NPs in (1):

(1) a. katab-a zayd-un risaalat-an ṣabaaḥ-an.

wrote-3MSG Zayd-NOM letter-ACC morning-ACC

‘Zayd wrote a letter in the morning.’

45
Other cases of case syncretism in SA include those cases where the class of nouns known in traditional Arabic
grammar as sound feminine plural nouns (i.e. nouns that pluralize thorugh suffixation rather than through a change
of the root/stem) realize their syntactic structural or syntactic non-structural (i.e. purely idiosyncratic lexical)
accusative case as morphologically genitive, as can be seen by the following examples:
(i) qaabal-a zayd-un l-ṭaalibaat-i/*l- ṭaalibaat-a
met-3MSG Zayd-NOM the-students.F-GEN/*the-students-ACC
‘Zayd met the students.’ (syntactic structural ACC on the theme object but morphologically GEN)
(ii) ʔinna l-ṭaalibaat-i/*l-ṭaalibaat-a fii l-ṣaff-i
that.EMPH the-students.F-GEN/*the-students.F-ACC in the-class-GEN
‘Indeed, the students are in the class.’ (syntactic purely idiosyncratic lexical ACC on the theme object but
morphologically GEN)
224
b. waḍaʕ-a zayd-un l-ʕalaamat-a yamiin-a l-ṭariiq-i

placed-3MSG Zayd-NOM the-sign-ACC right-ACC the-road-GEN

‘Zayd placed the sign on the right side of the road.’

c. zayd-un y-uṭaaliʕ-u l-ṣuḥuf-a kaṯiir-an.

Zayd-NOM 3-read-M.SG.INDIC the-newspaper-ACC often-ACC

‘Zayd often reads the newspapers.’

In (1a-c), the adverbial NPs receive a morphological accusative case. There are two questions

that need to be answered. First, are these NPs true arguments of the verb, or are they adverbial

NPs? The second question is whether the accusative case borne by these adverbial NPs is a

structural dependent case, Agree-based case, a semantic/adverbial case or a default case.

There are arguments against the claim that adverbials in SA are actually true arguments

of the verb. For instance, unlike true objects, adverbial NPs cannot be passivized, as is shown by

the contrasting pairs in (2):

(2) a. kutib-at risaalat-un ṣabaaḥ-an.

wrote.PASS-3FSG letter.F-NOM morning-ACC

‘A letter was written in the morning.’

b. *kutib-a ṣabaaḥ-un risaalat-an.

wrote.PASS-3MSG morning.M-NOM letter-ACC

225
c. wuḍiʕ-at l-ʕalaamat-u yamiin-a l-ṭariiq-i

placed.PASS-3FSG the-sign.F-NOM right-ACC the-road-GEN

‘The sign was placed on the right side of the road.’

d. * wuḍiʕ-a yamiin-u l-ṭariiq-i l-ʕalaamat-a

placed.PASS-3MSG right.M-NOM the-road-GEN the-sign-ACC

e. t-uṭaaliʕ-u l-ṣuḥuf-u kaṯiir-an.

3FSG-read.PASS-INDIC the-newspapers.F-NOM often-ACC

‘Newspapers are often read.’

f. *y-uṭaaliʕ-u kaṯiir-un l-ṣuḥuf-a

3MSG-read.PASS-INDIC often.M-NOM the-newspapers-ACC

Having established that adverbial NPs are not true arguments of the verb, it is crucial to

point out that place and time NPs as well as amount NPs can still be passivized on the condition

that they lose their adverbial function, and assume instead the function of subjects or the function

of objects of prepositions. This can be illustrated by the contrasts in (3) through ( 6):

(3) a. ṣaam-at zaynab-u yawm-a l-jumuʕat-i

fasted-3FSG Zaynba-NOM day-ACC the-Friday-GEN

‘Zaynab fasted on Friday.’

226
b. ṣiim-a yawm-u l-jumuʕat-i.

fasted.PASS-3MSG day.M-NOM the-Friday-GEN

‘?*Friday was fasted on.’ (adapted from Al-Ašram 2003: 347)

(4) a. waqaf-at zaynab-u ʔamaam-a l-nahr-i

stood.3FSG Zaynab-NOM front-ACC the-river-GEN

‘Zaynab stood in front of the river.’

b. wuqif-a ʔamaam-u l-nahr-i.

stood.PASS-3MSG front.M-NOM the-river-GEN

‘?*The front of the river was stood at.’ (adapted from Al-Ašram 2003: 348)46

(5) a. katab-at zaynab-u l-kaṯiir-a ʕan l-faqr-i

wrote-3FSG Zaynab-NOM the-lot-ACC about the-poverty-GEN

‘Zaynab wrote a lot about poverty.’

b. kutib-a l-kaṯiir-u ʕan l-faqr-i

wrote.PASS-3MSG the-lot.M-NOM about the-poverty-GEN

‘A lot was written about poverty.’

46
This example might sound degraded to many Modern speakers of SA. However, Al-Ašram 2003: 348) takes these
examples from traditional Arab grammarians who were native speakers of SA.
227
(6) a. raʔay-tu-haa fii yawm-i l-jumuʕat-i

saw-1SG-her.ACC on day-GEN the-Friday-GEN

‘I saw her on Friday.’

b. zayd-un y-uṭaaliʕ-u l-ṣuḥuf-a fii

Zayd-NOM 3-read-M.SG.INDIC the-newspaper-ACC on

kaṯiir-in mina l-ʔaḥyaan-i.

lot-GEN of the-times-GEN

‘Zayd often reads newspapers.’

I take the fact that the NPs in (3) through (6) can be passivized and bear nominative case, and

their ability to function as objects of prepositions and bear genitive case as evidence that these

NPs are no longer adverbial NPs, but assume other grammatical functions. This is also the

position taken by traditional Arab grammarians (see, for example Al-Warraaq 10th c./1999: 281).

In other words, these NPs take whatever structural cases regular NPs take when they function as

subjects of passives or objects of prepositions.

Baker (2015: 215-2016) argues, based on cross-linguistic evidence, that adverbial NPs

can still receive the structural dependent case in some languages such as Cuzco Quechua,

Finnish and Korean. Baker (2105: 217) shows that in Finnish, certain adverbs receive structural

case because they show case alternation. For example, the duration adverb is accusative in (7a)

228
but bare-nominative in (7b), an impersonal passive where the covert agent is not a case

competitor:

(7) a. Opiskel-i-n vuode-n

study-PAST-1sS year-ACC

‘I studied (for) a year.’ (Baker 2015, ex. 63a: 217)

b. Opiskel-tiin vuosi

study-PAST.PASS year.NOM

‘People/we studied (for) a year.’ (Baker 2015, ex. 63b: 217)

To see whether this is actually the case in SA, we can employ the same diagnostics used

in Baker (2015: 216-217).

For example, adverbials bear the accusative case in active sentences in SA, as can be

illustrated in (8):

(8) daras-tu ṣabaaḥ-an/*ṣabaaḥ-un.

studied-1SG morning-ACC/*morning-NOM

‘I studied in the morning.’

229
If the accusative case borne by the adverbials is structural, then the prediction would be that in

the context of impersonal passives, the adverbial NP would show case alteration by bearing the

nominative case. This prediction is not borne out by the facts, as is shown in (9):

(9) y-uxtaṣam-u [PP ʔilaa zayd-in] ṣabaaḥ-an/

3MSG-arbiterate.PASS-INDIC to Zayd-GEN morning-ACC/

*ṣabaaḥ-un.

morning-NOM

‘Zayd is referred to for arbitration in the morning.’

In (9), the NP zayd ‘Zayd’ is part of a PP, and the the adverbial NP obligatorily surfaces with the

accusative case. The fact the advberial in (9) does not show case alteration in active and passive

sentences suggests strongly that the accusative case borne by the adverbial is not syntactic

structural. It is also crucial to point out the accusative case borne by averbials cannot be some

kind of case concord. This is because the accusative case realized on the adverbial in (9) without

there being another accusative-marked NP in the sentence.

The other question that needs to be answered is the following: why can’t we assume that

adverbial NPs receive an unmarked/default case? This possibility is also ruled out given that the

unmarked/default case in SA is the nominative case rather than the accusative case in the CP

domain.

230
Another question that might be raised is why we cannot assume that the accusative case

borne by adverbial NPs is a form of lexical/inherent case. This possibility is also ruled out when

we consider the fact that in the sense of Woolford (2006), lexical case is only assigned to internal

arguments, whereas inherent case is only assigned to external arguments. Since the adverbial

NPs in SA are neither internal nor external arguments, it follows that the accusative case borne

by these NPs is neither lexical nor inherent. Furthermore, assuming that adverbial NPs receive a

form of purely idiosyncratic lexical case entails that there be lexical case assigners, something,

which is not available in the case at hand. The argument that adverbial NPs receive a form of

inherent case is also ruled out, as these NPs are not arguments of the verb to begin with.

We are thus left with one possibility to consider, which is that the accusative case borne

by adverbial NPs in SA is a form of semantic/adverbial case. One piece of evidence that the

accusative case borne by adverbial NPs in SA are probably a form of semantic/adverbial case

comes from Schütze (2001: 209) who states that “[…] in rich case-marking languages,

specifically bare DPs can serve various adverbial functions, with their particular meaning

dependent on the choice of case (e.g. dative of duration, ablative of instrument)[…]” Given that

the case system of SA is not as rich as that of other languages, with only three morphological

cases, it should not be surprising that adverbial NPs bear one of these cases, namely the

accusative form in order to mark them as having adverbial functions. Other linguists (see Fassi

Fehri 1986: 186) also claim that adverbial NPs in SA bear a form of semantic case rather than

structural case.

231
6.3 Accusative case assignment of some other case-marked NPs

In this section, I discuss the case facts of some accusative-marked NPs in SA such as the

absolute objects (= cognate objects), the circumstances, objects of reason/purpose, and objects of

accompaniment (comitative objects). I show that these NPs function as adverbial NPs rather than

true objects. I also show that the accusative case borne by these NPs is a form of semantic case

rather than structural dependent case.

SA has a number of accusative-bearing NPs, some of which are treated as objects in

traditional Arabic grammar. These NPs are cognate objects, objects of reason, and comitative

objects.

The examples below are respectively called the absolute objects (10), the objects of

purpose (11), and the comitative objects (12) in traditional Arabic grammar:

(10) dammar-a l-zilzaal-u l-qaryat-a tadmiir-an.

destroyed-3MSG the-earthquake-NOM the-village-ACC destruction-ACC

‘The earthquake literally destroyed the village.’

(11) ʔalqaa zayd-un qaṣiidat-an takriim-an li-l-ʕaalim-i

recited.3MSG Zayd-NOM poem-ACC honor-ACC for-the-scientist-GEN

‘Zayd recited a poem to honor the scientist.’

(12) sir-tu wa l-niil-a.

walked-1SG and the-Nile-ACC

‘I walked along the Nile.’

232
However, there are reasons to believe that these NPs are not true objects. Instead, I argue that

they are adverbial NPs. Semantically, traditional Arab grammarians (see Ibn ʕaqiil (13th c./1980:

169, 186) hold the view that absolute objects encode the meaning of corroboration, and that

objects of purpose, encode the meaning of reason or purpose. This indicates that their function in

the sentence is one of modification. Syntactically, there are other reasons to believe that these

objects are adverbial NPs rather than true objects. First, unlike true objects, which can be

passivized, these objects cannot, as can be seen by the contrasting pairs in (13) through (15):

(13) a. dammar-a l-zilzaal-u l-qaryat-a tadmiir-an.

destroyed-3MSG the- earthquake -NOM the-village-ACC destruction-ACC

‘The earthquake literally destroyed the village.’

b. dummir-at l-qaryat-u tadmiir-an.

destroyed.PASS-3FSG the-village.F-NOM destruction-ACC

‘The village was literally destroyed.’

c. *dummir-a tadmiir-un l-qaryat-a.47

destroyed.PASS-3MSG destruction.M-NOM the-village-ACC

(14) a. ʔalqaa zayd-un qaṣiidat-an takriim-an li-l-ʕaalim-i

recited.3MSG Zayd-NOM poem-ACC honor-ACC for-the-scientist-GEN

‘Zayd recited a poem to honor the scientist.’

b. ʔulqiy-at qaṣiidat-un takriim-an li-l-ʕaalim-i

recited.PASS-3FSG poem.F-NOM honor-ACC for-the-scientist-GEN

‘A poem was recited in honor of the scientist.’

c. *ʔulqiy-a takriim-un li-l-ʕaalim-i

47
On the ungrammaticality of such examples of absolute objects in passives, see (Hasan 1962: 115).
233
recited.PASS-3MSG honor.M-NOM for-the-scientist-GEN

qaṣiidat-an48

poem-ACC

(15) a. sir-tu wa l-niil-a.

walked-1SG and the-Nile-ACC

‘I walked along the Nile.’

b. *siir-a ` wa l-niil-u.

walked.PASS-3MSG and the-Nile-NOM

Second, unlike true objects, which are obligatory, these objects are optional, as can be

seen by the contrasting examples in (16) through (18):

(16) a. dammar-a l-zilzaal-u l-qaryat-a tadmiir-an.

destroyed-3MSG the-volcano-NOM the-village-ACC destruction-ACC

‘The earthquake literally destroyed the village.’

b. dammar-a l-zilzaal-u l-qaryat-a.

destroyed-3MSG the-volcano-NOM the-village-ACC

‘The earthquake destroyed the village.’

c. * dammar-a l-zilzaal-u tadmiir-an.

destroyed-3MSG the-earthquake-NOM destruction-ACC

48
On the impossibility of passivizing the objects of reason/purpose and comitative objects, see Hasan (1962: 122).
234
(17) a. ʔalqaa zayd-un qaṣiidat-an takriim-an li-l-ʕaalim-i

recited.3MSG Zayd-NOM poem-ACC honor-ACC for-the-scientist-GEN

‘Zayd recited a poem to honor the scientist.’

b. ʔalqaa zayd-un qaṣiidat-an

recited.3MSG Zayd-NOM poem-ACC

‘Zayd recited a poem.’

c. *ʔalqaa zayd-un takriim-an li-l-ʕaalim-i

recited.3MSG Zayd-NOM honor-ACC for-the-scientist-GEN

(18) a. sir-tu wa l-niil-a.

walked-1SG and the-Nile-ACC

‘I walked along the Nile.’

b. sir-tu

walked-1SG

‘I walked.’

Having established that absolute objects are not true objects but are rather adverbial NPs,

it is worth pointing out that there are other cases, where absolute objects do behave like true

objects. This is exclusively the case with absolute objects that are quantified, as can be seen in

(19):

(19) saar-a zayd-un sayr-an ḥaṯiiṯ-an.

walked-3MSG Zayd-NOM walking-ACC swift-ACC

‘Zayd walked a swift walk.’

235
The quantified absolute object in (19) can be passivized, just like true objects, as can be seen in

(20):

(20) siir-a sayr-un ḥaṯiiṯ-un.

walked.PASS-3MSG walking-NOM swift-NOM

‘A swift walking was walked.’

Quantified objects such as those in (20) are therefore true objects that receive the structural

dependent accusative case that other regular objects do.

Other linguists (e.g. Pereltsvaig 1999, de Hoop and Zwarts 2009: 174, Maling 2009: 75-

76) also treat cognate objects in other languages as adverbs of manner rather than as true objects.

Having established that the absolute objects, objects of reason/purpose, and the

comitative objects are not true objects but are instead adverbial NPs, the question remains

whether the accusative case that these adverbial NPs bear is a structural dependent case or not.

There is reason to believe that the accusative case borne by these adverbial NPs is a form of

semantic/adverbial case rather than structural case. We have established earlier that PPs are not

case competitors in SA in that they do not trigger the dependent case on another NP in the same

spell-out domain. If this is correct, then there are two predictions that can be made in order to test

whether or not the accusative case borne by these NPs is structural dependent case or not. If

these objects bear the dependent case, then the prediction would be that in the context of

passives, which do not project an external argument, they should surface with the nominative

case. However, this prediction is not borne out, as is shown in (21):

236
(21) a. ʔiḥtufiy-a [PP bi-zayd-in] ʔiḥtifaaʔ-an/

welcomed.PASS-3MSG for-Zayd-EGN welcoming-ACC

*ʔiḥtifaaʔ-un

welcoming-NOM

‘Zayd was warmly welcomed.’

b. ʔiḥtufiy-a [PP bi-zayd-in] takriim-an/

welcomed.PASS-3MSG for-Zayd-GEN rewarding-ACC/

*takriim-un li-juhuud-i-hi

rewarding-NOM for-efforts-GEN-his

‘Zayd was celebrated for his efforts.’

c. siir-a [PP bi-zayd-in] wa l-niil-a/*l-niil-u

walked.PASS-3MSG with-Zayd-GEN and the-Nile-ACC/*the-Nile-NOM

‘Zayd was walked with along the Nile.’

The examples in (21) show that absolute objects, objects of purpose and comitative objects all

obligatorily surface with the accusative case. This shows that the accusative case borne by these

NPs is not the structural dependent case. Given that the accusative case borne by these NPs is not

237
structural, together with the claim that these NPs have an adverbial function, it follows that these

NPs bear a form of semantic/adverbial case rather than structural case. Here also, the possibility

that the accusative case borne by these NPs is default/unmarked is ruled out, as it is the

nominative case rather than the accusative case that is the default case in the CP domain of SA.

To summarize, I have shown that SA has a number of accusative-bearing NPs, some of

which are treated as objects in traditional Arabic grammar. Using a number of diagnostics, I have

demonstrated that these NPs are not actually objects, but are better treated as adverbial NPs. I

have also shown that the accusative case borne by these NPs is a form of semantic/adverbial case

rather than structural dependent accusative case or default case.

6.4 Case assignment of NPs in the left periphery of the clause

In this section, I discuss the case facts of NPs in the left periphery of the clause, i.e. the

CP domain. The example in (22) is an illustrative example.

(22) l-kitaab-u, qaraʔ-tu-hu

the-book-NOM read-1SG-it

‘The book, I read it.’

First, we need to address the status of resumptive pronouns in SA. I consider resumptive

pronouns in SA to be full pronominal arguments which are encliticized to lexical or functional

heads in the structure. This claim is supported by Fassi Fehri’s (1986) observation and Aoun et.

238
al.’s (2010) observation that resumptive pronouns void islands in SA, as can be shown by the

following examples:

(23) l-mariiḍ-u, qaabal-tu [NP l-rajul-a ll-aðii zaar-a-*(hu)]

the-sick.person-NOM met-1SG the-man-ACC who visited-3MSG-him

‘The sick person, I met the man who visited him.’

(24) l-mariiḍ-u, xaraj-tu [CP qabl-a ʔan ʔ-azauur-Ø-a-*(hu)]

the-sick.person-NOM left-1SG before that.SUBJ 1SG-visit-3MSG-SUBJ-him

‘The sick person, I left before I visited him.’

(25) l-mariiḍ-u, samiʕ-tu ʔanna zayd-an y-aʕlam-Ø-u

the-sick.person-NOM heard-1SG that Zayd-NOM 3-know-MSG-INDIC

[CP man zar-a-*(hu)]

who visited-3MSG-him

‘The sick person, I heard that Zayd knows who visited him.’

The examples in (23-25) show clearly that the complex NP island (23), the adjunct CP island

(24), and the wh-island (25) can all be voided using the resumptive pronoun. I take this as

evidence that the resumptive pronouns are true arguments of the verb and that the NP ʔal-mariiḍ

‘the sick.person’ is base-generated in the left-periphery rather than moved from within the

clause.

Let us consider next the contrasting pair in (26a-b) below, which show that the sentence

is rendered ill-formed if the focalized NP Hind ‘Hind’ is followed by the SV order . However,

the sentence is well-formed when the focalized NP is followed by VS order.

239
(26) a. *HIND-AN Zayd-un raʔaa-haa

Hind-ACC Zayd-NOM saw-her

‘It was Hind that Zayd saw.’

b. HIND-AN raʔaa-haa Zayd-un

Hind-ACC saw-her Zayd-NOM

‘It was Hind that Zayd saw.’

To account for these sentences, I follow Shlonsky’s (2000: 330) claim that focalization in SA

induces subject-verb inversion. Thus, the sentence in (26a) is ill-formed because the focalized

NP Hind ‘Hind’ is followed by the subject Zayd ‘Zayd’ rather than by the verb. This is in

contrast with the grammatical example in (26b), where the focalized NP Hind ‘Hind’ is followed

by the verb.

In the cartographic theory of Rizzi (1997), it is argued that question words universally

occupy the Spec, FocP position and that question particles universally occupy the head position

of FocP. In addition, Chomsky (1986), Marantz (1991) and Baker (2015) argue that case is

assigned to chains. Assuming that these two claims are correct, we can now address the

following questions: what makes the left peripheral DP in (27-31) ungrammatical when the left

peripheral DP is marked with the accusative or nominative cases and the sentence has a gap? (b)

why are the sentences in (27-31) grammatical when the left-peripheral DP surfaces with the

nominative case and the sentence has the resumptive strategy?

240
(27) a. hal ḍarab-ta Zayd-an

Q hit-2MSG Zayd-ACC

‘Did you hit Zayd?’

b. *Zayd-an hal ḍarab-ta _____

Zayd-ACC Q hit-2MSG

(Fassi Fehri 1986, ex. 73: 121)

c. *Zayd-un hal ḍarab-ta _____

Zayd-NOM Q hit-2MSG

d. Zayd-un hal ḍarab-ta-hu

Zayd-NOM Q hit-2MSG-him

‘Zayd, did you hit him?’

(28) a. ʔa-ḍarab-ta Zayd-an

Q-hit-2MSG Zayd-ACC

‘Did you hit Zayd?’

b. *Zayd-an ʔa- ḍarab-ta _____

Zayd-ACC Q-hit-2MSG

241
c. *Zayd-un ʔa- ḍarab-ta _____

Zayd-NOM Q-hit-2MSG

d. Zayd-un ʔa- ḍarab-ta-hu

Zayd-NOM Q-hit-MSG-him

‘Zayd, did you hit him?’

(29) a. man daʕaa llaah-a

who prayed.3MSG God

‘Who prayed God?’

b. *llaah-a man daʕaa _____

God-ACC who prayed.3MSG

(Fassi Fehri 1986, ex. 72: 121)

c. *llaah-u man daʕaa _____

God-NOM who prayed.3MSG

d. llaah-u man daʕaa-hu

God-NOM who prayed-him

‘God, who prayed him?’

242
(30) a. kam ḍarab-ta Zayd-an

how.much hit-2MSG Zayd-ACC

‘How much did you hit Zayd?’

b. *Zayd-an kam ḍarab-ta _____

Zayd-ACC how.much hit-2MSG

(Fassi Fehri 1986, ex. 54: 116)

c. *Zayd-un kam ḍarab-ta _____

Zayd-ACC how.much hit-2MSG

d. Zayd-un kam ḍarab-ta-hu

Zayd-NOM how.much hit-2MSG-him

‘Zayd, how much did you hit him?’

(31) a. hallaa ʔakram-ta Zayd-an

PRT.why.not were.generous.with-2MSG Zayd-ACC

‘Why were you not generous with Zayd?’

b. *Zayd-an hallaa ʔakram-ta _____

Zayd-ACC PRT.why.not were.generous.with-2MSG

(Fassi Fehri 1986, ex. 52:116)

243
c. *Zayd-un hallaa ʔakram-ta _____

Zayd-NOM PRT.why.not were.generous.with-2MSG

d. Zayd-un hallaa ʔakram-ta-hu

Zayd-NOM PRT.why.not were.generous.with-2MSG-him

‘Zayd, why were you not generous with him?’

Let us assume (following Chomsky 1977, Fassi Fehri 1986, Aoun et. al. 2010) that the presence

of gaps indicates movement, and the presence of resumptive pronouns indicates the absence of

movement, given that resumptive pronouns void islands. The examples in (27-31) all show the

following pattern: when the left-peripheral accusative-marked NP Zayd ‘Zayd’ is focalized or

topicalized leaving a gap as the result of movement, as in (27b-31b), the sentence is

ungrammatical; when the left-peripheral nominative-marked NP Zayd ‘Zayd’ is focalized or

topicalized as the result of movement, as in (27c-31c), the result is ungrammatical; when the left-

peripheral nominative-marked NP Zayd ‘Zayd’ is base-generated in the left-periphery as in (27d-

31d) and the resumptive strategy is used, the result is grammatical. These examples can all be

accounted for if we assume the cartographic approach of Rizzi (1997). To account for the

ungrammatical examples in (27b-31b), we can proceed along the following manner: suppose that

the question words and particles ʔa-, hal, man, kam, hallaa all project a FocP. Suppose further

that a gap indicates movement, and the resumptive pronoun indicates lack of movement, given

that resumptive pronouns void islands in SA. If this is the case, then the examples in (27b-31b)

are all barred because the NP Zayd ‘Zayd’ is raised from its base position in the complement of

244
VP, where it receives a dependent accusative case to the Spec, FocP position or the Spec, TopP

position in the sentence. This movement induces ungrammaticality for one of two reasons: (a)

the DP Zayd ‘Zayd’ is raised to the Spec, FocP position, and the ungrammaticality is due to the

fact that it is impossible to have two projections of FocP in the architecture of the sentence, à la

Rizzi (1997). In other words, the underlying representation of the sentences in (27b-31b) can be

represented as is shown in (32):

(32) *[FocP Zayd-an [FocP ʔa, hal, man, hallaa, kam…<Zayd-an>]

(b) the DP Zayd ‘Zayd’ is raised to the Spec, TopP position, and the ungrammaticality is due to

the fact that the DP Zayd ‘Zayd’ moved across the question words, which occupy the FocP

projection, and movement across elements in the C domain is banned on standard accounts. In

other words, the underlying derivation of the sentences in (27b-31b) can be represented as

follows:

(33) *[TopP Zayd-an [FocP ʔa, hal, man, hallaa, kam…<Zayd-an>]

To account for why the examples in (27c-31c) are ungrammatical, we can proceed along the

following manner: these examples are all ungrammatical due to one of two reasons (a) the

pronounced copy of the NP Zayd ‘Zayd’is raised to the Spec, FocP position; therefore, this

structure incurs two violations, namely the structure projects two instances of FocP projections,

thus violating the universal ban on having two FocP projections in the structure, and the fact that

the two links in the chain <Zayd, Zayd> end up receiving two distinct cases, thus violating the

245
condition that stipulates that cases are assigned to chains (cf. Chomsky 1986, Marantz 1991,

Baker 2015) (b) the pronounced copy of the NP Zayd ‘Zayd’is raised to the Spec, TopP position;

therefore, the structure incurs two violations. The first violation has to do with the fact that the

two links in the chain <Zayd, Zayd> end up receiving two distinct cases, accusative case in the

base position and nominative case in the landing position, thus violating the condition stipulated

in Chomsky (1986), Marantz (1991) and Baker (2015) that case is assigned to chains. The

second violation has to do with the fact the DP Zayd ‘Zayd’ moved across an element in the C

domain, a movement, which is banned on standard accounts. The underlying representation of

(27c-31c) can therefore be represented as in (34):

(34) *[TopP Zayd-un[NOM] [FocP ʔa, hal, man, hallaa, kam…<Zayd-an[ACC]>]

To account for the grammatical examples in (27d-31d), we can assume that the lack of a gap in

the structure and the presence of a resumptive pronoun indicate lack of movement. Thus, the NP

Zayd ‘Zayd’ in these sentences must be one of two things: (a) a higher topic phrase base-

generated in its surface position, and it therefore surfaces with the unmarked/default case in the

CP domain, namely the nominative case. Therefore, nothing is violated, and the sentences are

grammatical. The underlying representation of (27d-31d) can therefore be shown as in (35):

(35) [TopP Zayd-un[NOM] [FocP ʔa, hal, man, hallaa, kam…]

246
Alternatively, the NP Zayd ‘Zayd’ is adjoined to FocP, and it receives the unmarked/default

nominative case in SA. The underlying structure of the sentences in (27d-31d) must, according to

this option be represented as follows:

(36) [FocP Zayd-un[NOM] [FocP ʔa, hal, man, hallaa, kam…]

If the above analysis is correct and if question particles occupy the head Foc of the FocP

projections, the question now arises as to how to account for the following sentence:

(37) ʔa-Zayd-an ḍarab-ta _____

Q-Zayd-ACC hit-2MSG

‘Was it Zayd that you hit?’

In (37), there is a gap, which indicates that the NP Zayd ‘Zayd’ is moved from its base position

as the object of the verb to the CP domain for focalization purposes. If focalization can be

encoded by question particles and if the NP Zayd ‘Zayd’ is also focalized (or topicalized), as

seems to be the case, the question is how to account for the surface order of the question particle

and the focalized NP. This sentence is accounted for once we consider the fact that it is a yes/no

question, and in the cartographic system of Rizzi (1997), question particles of the yes/no type

occupy the highest position in the clause, namely ForceP. Therefore, I suggest that the bound

247
question particle is located in the highest position of the clause, and the DP Zayd ‘Zayd’ is the

only element that moves to Spec, FocP (or Spec, TopP). 49

Now we can consider the sentences in (38-41) and ask the following questions: what

makes the left peripheral DP in (38-41) ungrammatical when the left peripheral DP is marked

with the accusative or nominative cases and the sentence has a gap? (b) why are the sentences in

(38-41) grammatical when the left-peripheral DP surfaces with the nominative case and the

sentence has the resumptive strategy?

(38) ʔin laqii-ta Zayd-an fa-ʔkrim-Ø-hu

if met-2MSG Zayd-ACC CONJ-be.generous.with-MSG-him

‘If you meet Zayd, be generous with him.’

b. *Zayd-an ʔin laqii-ta _____

Zayd-ACC if met-2MSG

fa-ʔakrim-hu

CONJ-be.generous.with-him

(Fassi Fehri 1986, ex. 51: 116)

c. *Zayd-un ʔin laqii-ta _____ fa-ʔakrim-Ø-hu

Zayd-NOM if met-2MSG CONJ-be.generous.with-MSG-him

49
I thank Gabriela Alboiu for suggesting this analysis to me.
248
d. Zayd-un ʔin laqii-ta-hu fa-ʔakrim-Ø-hu

Zayd-NOM if met-2MSG-him CONJ-be.generous.with-MSG-him

‘Zayd, if you meet him, be generous with him.’

(39) a. law qaabal-at Hind-un Zayd-an la-ʔakram-at-hu

if met-3FSG Hind-NOM Zayd-ACC CONJ-was.generous-3FSG-him

‘Had Hind met Zayd, she would have been generous with him.’

b. *Zayd-an law qaabal-at Hind-un _____,

Zayd-ACC if 2-meet-MSG-JUSS

la-ʔakramat-hu

CONJ-was.generous.with-him

c. *Zayd-un law qaabal-at Hind-un _____,

Zayd-NOM if met-3FSG Hind-NOM

la-ʔakramat-hu

CONJ-was.generous.with-him

d. Zayd-un law qaabal-at-hu Hind-un,

Zayd-NOM if met-3FSG-him Hind-NOM

la-ʔakram-at-hu

249
CONJ-was..generous.with-him

‘Zayd, had Hind met him, she would have been generous with him.’

(40) a. la-ʔanaa ʔ-uḥibb-u Zayd-an

PRT.verily-I 1SG-love-INDIC Zayd-ACC

‘Verily, I love Zayd.’

b. *Zayd-an la-ʔanaa ʔ-uḥibb-u _____

Zayd-ACC PRT-I 1SG-love-INDIC

(Fassi Fehri 1986, ex. 53: 116)

c. *Zayd-un la-ʔanaa ʔ-uḥibb-u _____

Zayd-NOM PRT-I 1SG-love-INDIC

d. Zayd-un la-ʔanaa ʔ-uḥibb-u-hu

Zayd-NOM PRT-I 1SG-love-INDIC-him

‘Zayd, I verily love him.’

(41) a. ʔinn-ii ḍarab-tu Zayd-an

that.EMPH hit-1SG Zayd-ACC

‘Indeed, I hit Zayd.’

250
b. *Zayd-an ʔinn-ii ḍarab-tu _____

Zayd-ACC that.EMPH-me hit-1SG

(Fassi Fehri 1986, ex. 74: 121)

c. *Zayd-un ʔinn-ii ḍarab-tu _____

Zayd-NOM that.EMPH-me hit-1SG

d. Zayd-un ʔinn-ii ḍarab-tu-hu

Zayd-NOM that.EMPH-me hit-1SG-him

‘Zayd, I hit him.’

To account for the ungrammatical examples in (38b-41b), I argue that the particles law and ʔin of

conditions, la of initiation and swearing/promising, and ʔinna of assertion and corroboration are

all base-generated in the highest position of the architecture of the CP domain, namely the head

Force. This claim is supported by the fact that nothing in SA (aside from adjuncts) can ever

precede these particles. This is in addition to the fact these particles clause-type the sentence.

Thus, law and ʔin clause-type the sentence as a condition, la- clause-types the sentence as a

promise, and ʔinna clause-types the sentence as an assertion. If this claim is on the right track,

then the facts follow straightforwardly. The examples in (38b-41b) are all ungrammatical

because the NP Zayd ‘Zayd’ is raised to a position higher than ForceP, thus violating the

universal architecture of the clause. The underlying representation of (38b-41b) can therefore be

shown as in (42):

251
(42) *[FocP/TopP Zayd-an [ForceP law, ʔin, la, ʔinna…<Zayd-an>]

Notice crucially that the NP Zayd ‘Zayd’ cannot be argued to be adjoined to ForceP, as there is a

gap in these sentences, which indicates movement from within the TP domain to the CP domain.

A similar explanation can account for the ungrammatical sentences in (38c-41c) in addition to

the fact that the sentences incur another violation, namely the two links in the chain receive

distinct cases. As for the examples in (38d-41d), they are grammatical because no violations are

incurred. The NP Zayd ‘Zayd’ is base-generated as an adjunct, which adjoins to ForceP, and it

receives the unmarked/default nominative case in the CP domain. The underlying representation

of (38d-41d) can be shown as in (43):

(43) [ForceP Zayd-un [ForceP law, ʔin, la, ʔinna…]

The claim that law, ʔin, la, and ʔinna occupy the highest projection in the structure of the clause,

namely ForceP makes the following prediction: the accusative-marked NP Zayd-an in (43)

should be fine below these heads, on the assumption that Zaydan can be occupying the Spec,

TopP position or Spec, FocP position.50 In principle, the prediction should be borne out.

However, there are other interfering factors that make it difficult to test this prediction at least

with some of these heads. For example, law and ʔin of conditions seem to force the VSO order

50
I thank Arsalan Kahnemuyipour for raising this issue.
252
suggesting that they induce subject-verb inversion, as can be seen by the following contrasts in

grammaticality:

(44) a. law jaaʔ-at Hind-un la-ʔakram-tu-haa

if came-3FSG Hind-NOM CONJ-was.generous.with-1SG-her

‘Had Hind come, I would have been generous with her.’

b. *law Hind-un jaaʔ-at la-ʔakram-tu-haa

if Hind-NOM came-3FSG CONJ-was.generous.with-1SG-her

(45) a. ʔin jaaʔ-at Hind-un fa-ʔakrim-Ø-haa

if came-3FSG Hind-NOM CONJ-be.generous.with-MSG-her

b. *ʔin Hind-un jaaʔ-at fa-ʔakrim-Ø-haa

if Hind-NOM came-3FSG CONJ-be.generous.with-MSG-her

However, with the heads la and ʔinna, placing the NP Zayd ‘Zayd’ below them is fine, as can be

seen by the examples below:

(46) a.? la-Zayd-an ʔanaa ʔ-uḥibb-u51

PRT.verily-Zayd-ACC I 1SG-love-INDIC

51
This example is grammatical on the reading whereby the DP ʔannaa ‘I’ is interpreted as a preverbal subject rather
than as a topic. It is ungrammatical on the reading whereby the DP is interpreted as a topic. This explains its
marginal acceptability.
253
‘Verily, Zayd, I love.’

b. ʔinna Zayd-an qaabal-a-hu l-ʔawlaad-u

that.EMPH Zayd-ACC met-3MSG-him the-boys-NOM

‘Indeed, Zayd, the children met him.’

Assuming this analysis to be correct, the obligatory case on the left-peripheral NPs follows

straightforwardly. Thus, the left-peripheral NP receives its dependent case before it raises to the

left-periphery for focalization or topicalization purposes.

To summarize, I have argued in this section that the different patterns of case in the left-

periphery can be accounted for using the cartographic approach/mapping of Rizzi (1997)

together with the assumption made in Chomsky (1977), Fassi Fehri (1986) and Aoun et. al.

2010) that resumptive pronouns indicate lack of movement, wheras gaps indicate the presence of

movement.

6.5 Summary

In this chapter, I have shown that adverbial NPs in SA receive a form of dverbial/semantic

case rather than structural dependent case or default case. By considering the case assignment

facts of the left-peripheral NPs, I have also shown how case is assigned to chains in that two

links in the same chain necessarily bear the same case.

254
Chapter Seven

Case assignment in the DP domain

7.1 Introduction

This chapter looks at case assignment in the DP domain of SA. In SA, there are two types

of DPs, simple DPs and derived DPs. Simple DPs are those DPs, where the lexical root starts the

derivation as a noun and ends the derivation as a noun. By contrast, derived DPs are those, where

the lexical root is a V, which is then categorially converted into a noun at a later point in the

derivation. Following Baker’s (2015) conjecture that DPs are universally hard phases, I assume

that the DP is a hard phase in SA. This decision is based on the fact that cases assigned outside

the DP do not affect the cases of the complements of D in SA. I begin by discussing simple DPs,

and then address derived DPs.

7.2 Case assignment in simple DPs

Let us consider the DP in (1) and its structural representation in (2). The internal structure

of the DP in SA is adapted from Fassi Fehri (1993).

(1) xams-u ṣuwar-i-n jamiil-at-i-n

five-NOM pictures.F-GEN-INDEF beautiful-F-GEN-INDEF

‘five beautiful picures’

(2) [DP xams-u [QP <xams-u> [nP ṣuwar-i-n [NP <ṣuwar-i-n> jamiil-at-i-n]]]]

five-F-NOM pictures.F-GEN-INDEF beautiful-F-GEN-INDEF

‘five beautiful pictures’ (in a nominative environment, where the DP occupies the

Spec, vP or Spec, TP position)

255
The structure of (1) can be represented in a tree-diagram, as in (3):

(3)
DP
3
D QP
xams [NOM] 3
five <Q> nP
3
n NP
ṣuwar[GEN]3
Spell-Out of pictures AdjP NP
D’s complement 5
jamiilat[GEN] <N>
beautiful

In (3), the lexical head noun ṣuwar ‘pictures’ is merged into the structure, and the AP is left-

adjoined to NP.52 NP is c-selected by the functional categorizing head n, and N is raised to n.

The assumption that NP is always c-selected by a categorizing N head is based on proposals

made in Distrubuted Morphology (see among others, Embick and Noyer 2005). The nP is c-

selected by the functional head Q(= quantifier), and QP is now projected. The functional

projection QP is c-selected by the functional head D and the DP is now projected. At the merger

of D, its QP complement is sent to Spell-Out. When the complement domain of D is spelled-out,

the head noun, namely ṣuwar ‘pictures’ is assigned genitive case by D, following Fassi Fehri’s

(1993) proposal that D assigns a structural genitive case to the first of its NP complements. The

52
Following the category theory of Baker (2003), I assume that nouns cannot license a specifier given that they have
a referential index. Therefore, the AP is an adjunct rather than a specifier of N in (1). Assuming that nouns can have
a referential index and license a specifier entails a violation of Bakers’ (2003) Reference Predication Constraint.

256
AP, which is adjoined to NP then copies the case features of the head noun and thus is marked

for genitive as a result of case concord. As for the quantifier head xams ‘five’, which is moved to

D, it receives the structural nominative via Agree with T, assuming that the DP is the subject of

the sentence in which it is used. Note crucially that only D, which hosts the quantifier, ends up

realizing the structural nominative case, whereas all the complements of D do not. This is

expected if DP, as is claimed in this thesis, is a hard phase. In other words, the complements of D

do not realize the nominative case, simply because these complements are not visible to the

domains outside the DP. As for D itself, it is still visible to the higher domains because it is the

head of the phase, and it is standardly assumed in Phase Theory (Chomsky 2000, 2001 and

onwords) that heads of phases and their specifiers are visible to the immediately higher phasal

domains.53

In (1), I have claimed that D in SA assigns the structural genitive case to its complement.

But is there any evidence that this is actually the case? Why can’t we, for example, claim that

genitive is the unmarked case inside NPs, a possibility, which is suggested in Marantz (1991: 24)

as being available in some languages? There isreason to believe that this cannot be true of SA.

Instead, the evidence weighs in favor of the proposal first made in Fassi Fehri (1993) that D in

SA assigns structural genitive to its complement. Baker (2015: 164) mentions Japanese and

Tamil as some of the languages, where the genitive case seems to be the unmarked case inside

53
Note that when the DP lacks the quantifier xams ‘five’, it is the head noun, which raises to D that will end up
receiving the nominative case assigned via Agree with T, when the former functions as the subject tof the sentence,
as in (i)
(i) ṣuwar-un jamiilat-un ʔiltuqiṭ-at yawm-a ʔams-i
pictures.F-NOM beautiful-NOM took.PASS-3FSG day-ACC yesterday-GEN
‘Beautiful picutres were taken yesterday.’
257
the NP, as more than one nominal inside the larger nominal can be assigned the genitive case, as

can be illustrated by the Tamil example in (4):

(4) John-ooʈa Mary-ooʈa padam

John-GEN Mary-GEN picture

‘John’s picture of Mary.’ (Baker 2015, ex. 80c: 164)

In (4), both John and Mary receive the genitive case, thus indicating that the genitive case is the

unmarked case in Tamil. If genitive case were not the unmarked case in Tamil, the prediction in

the theoretical framework adopted is to see the theme object Mary ‘Mary’ realizing the

dependent accusative case, as there are two NPs in the DP domain, and the agent NP John

‘John’c-commands the theme object Mary ‘Mary’. However, this prediction is not borne out by

the facts. In SA, the equivalent of (4) is ruled out, as shown in (5), thus suggesting that genitive

case is not the unmarked case in the DP domain of SA:

(5) *ṣuurat-u zayd-in hind-in

picture-NOM Zayd-GEN Hind-GEN

Intended meaning ‘Zayd’s picture of Hind.’

258
SA behaves like English in that (5) can only be ruled in when the prepositional strategy is used,

as shown in (6):54

(6) ṣuurat-u zayd-in li-hind-in

picture-NOM Zayd-GEN of-Hind-GEN

‘Zayd’s picture of Hind.’

The examples in (5) and (6) indicate that the genitive case is not the unmarked case inside the

NP. Had it been the unmarked case, (5) should have been grammatical, as nothing, in principle,

should prevent the unmarked case from being assigned more than once in the same spell-out

domain. As an example of a language, where the genitive case is assigned as a result of

agreement with D, Baker mentions Turkish, as shown in (7):

(7) Ali-nin kalem-i

Ali-GEN pencil-3sP

‘Ali’s pencil’ (Baker 2015, ex. 81: 165)

54
Note that in NPs such as (i), the two NPs um ‘mother’ and Zayd ‘Zayd’ do get the genitive case:
(i) sayyaarat-u umm-i Zayd-in
car-NOM mother-GEN Zayd-GEN
‘Zayd’s mother’s car.’
However, this does not provide evidence that GEN is the unmarked case in SA. The NP ummi Zayd-in ‘Zayd’s
mother’ is a construct state in SA. Therefore, the underlying representation of (i) is as shown in (ii):
(ii) [DP sayyaarat-u [PossP umm-i Zayd-in]]
In (ii), D, to which N raises in a head-to-head movement, assigns the genitive case to the construct state (i.e. the
PossP as a whole), which then percolates down to its constituent NPs.
259
In (7), the possessum agrees with the possessor, as shown by the morphology on the possessum,

and the possessor receives the genitive case as a result. Is there any evidence in SA that the

possessum agrees with the possessor? As a matter of fact, there is evidence in SA that D agrees

with its complement although the evidence is not quite as obvious as it is in Turkish. For

example, Fassi Fehri (1993) shows that D in SA inherits definiteness from the possessor, as is

shown when the head noun is modified by an adjective or a relative clause. This is illustrated in

(8) and (9) respectively:

(8) a. kitaab-u zayd-in l-ṯamiin-u/* ṯamiin-u-n

book-NOM Zayd-GEN the-precious-NOM/*precious-NOM-INDEF

‘Zayd’s precious book’

b. kitaab-u-n ṯamiin-u-n/* l-ṯamiin-u

book-NOM-INDEF precious-NOM-INDEF/*the-precious-NOM

‘a precious book’

(9) a. kitaab-u zayd-in *(llaðii) qaraʔ-tu-hu

book-NOM Zayd-GEN that.DEF read-1SG-it

‘Zayd’s book, which I read’

b. kitaab-u-n (*llaðii) qaraʔ-tu-hu

book-NOM-INDEF that.DEF read-1SG-it

260
‘a book, which I read.’

The example in (8a) shows that a definite DP is grammatical only when the modifying AP is

definite. By contrast, the example in (8b) shows that an indefinite DP is grammatical only when

the modifying AP is indefinite. Likewise, the example in (9a) shows that the definite DP is ruled

in only when the modifying relative clause is headed by the definite complementizer. This

contrasts with (9b), where the posessum is indefinite; therefore, no definite complementizer is

allowed in the modifying relative clause. These examples show clearly that D in SA, where the

possessum moves to, agrees with the possessor in definiteness. Thus, the possessum in SA agrees

with the possessor although in a more subtle manner than that shown in Turkish. If this is the

case, then SA belongs to those languages, where D assigns the genitive case to its complement,

as was first proposed in Fassi Fehri (1993).

Having established that genitive case is a structural case assigned by D in the DP domain,

let us consider the DP in (10) and its structural representation in (11):

(10) kutub-u zayd-i-n l-ṯamiin-at-u

books.F-NOM Zayd-GEN-INDEF the-precious-F-NOM

‘Zayd’s precious books’

(11) [DP kutub-u [PossP zayd-i-n [nP <kutub-u> [NP <kutub-u>

books.F-NOM Zayd-GEN-INDEF

261
l-ṯamiin-at-u]]]]

the-precious-F-NOM

‘Zayd’s precious books.’

The DP in (10) can be represented as the tree-diagram in (12):

(12)

DP
3
D PossP
kutub[NOM] 3
books zayd[GEN] Poss’
3
<Poss> nP
3
<n> NP
3
→Spell-Out AP <N>
D’s 5
Complement l-ṯamiin-at[NOM]
the-precious

In (12), the lexical head noun kutub ‘books’ is merged into the structure and the AP l-ṯamiin-at

‘the precious’ is left-adjoined to NP. The NP is c-selected by the functional head n, and N is

raised to n for categorization. The nP is c-selected by the functional head Poss(essive). The

possessor Zayd ‘Zayd’ is merged into the structure in the Spec, PossP position. The functional

head D is merged into the structure, and the complex head N+n+Poss is raised to it in order to

check the latter’s [uN] feature. D is a phasal head. This means that the complement PossP is sent

to Spell-Out at the merger of D into the structure. In the spell-out domain of D, the possessor

Zayd ‘Zayd’ in the Spec, PossP position is assigned the structural genitive case by D in the DP
262
domain. The AP copies the [NOM] feature of the head N given that N is a member of the chain

whose head is now in D. The head D, which hosts the head N, is now visible to the higher phasal

domains of which DP is the subject. Therefore, D receives the structural nominative case via

Agree with T.

Let us consider next the DP in (13) and its structural representation in (14):

(13) kitaab-u ʔibn-i ʕamm-i zayd-i-n l-ṯamiin-u

book.M-NOM son-GEN uncle-GEN Zayd-GEN-INDEF the-precious.M-NOM

‘the precious book of the son of the uncle of Zayd/The precious book of Zayd’s cousin.’

(14) [DP kitaab-u [PossP ʔibn-i [PossP ʕamm-i [PossP zayd-i-n

book-NOM son-GEN uncle-GEN Zayd-GEN-INDEF

[nP <kitaab-u> [NP <kitaab-u> l-ṯamiin-u ]]]]

the-precious-M.NOM

‘the precious book of Zayd’s cousin’

The structural representation (13) can be represented as a tree diagram in (15):

263
(15)

DP
3
D PossP
kitaab 3
[NOM] PossP Poss’
3 3
DP Poss’ <N+n+ Poss > nP
ʔibn 3 3
[GEN] Poss PossP <n> NP
3 3
DP Poss’ AP <N>
ʕam 3 l-ṯamiin
[GEN] Poss PossP the-precious
3 [NOM]
DP Poss’
Zayd 3
[GEN] Poss

In (15), the lexical head noun kitaab ‘book’ is merged into the structure, and the AP is left-

adjoined to NP. The NP is c-selected by the functional categorizing head n, and N is raised to n

for categorization. The functional head Poss is merged into the structure, and the possessor ʔibn

ʕam Zayd ‘the son of the uncle of Zayd’ is merged in the Spec, PossP position. The complex

head N+n+Poss is raised to D to check the [uN] feature of D. At the merger of D, its complement

PossP is sent to Spell-Out. In this spelled-out domain, the DP in PossP is assigned the structural

genitive case assigned by D in the DP domain, which is then percolated down to all the elements

inside the PossP.55 The head D, which hosts the head N, is now visible to the higher phasal

55
That the case feature of the whole nominal is percolated down to all the members of the nominal is confirmed in
Baker (2015: 175) who states that “[t]here are, of course, also languages in which every element inside the nominal
264
domains of which DP is the subject. Therefore, D receives the structural nominative case via

Agree with T. The AP later copies the [NOM] case feature of the head noun in D given that N

and D are members of the same chain.

7.3 Case assignment in derived nominals

Process (or event) nominals are particularly interesting from the perspective of case

theory in that the lexical root starts the derivation as a verb but ends up becoming a noun later in

the derivation. This hybrid categorial nature of process nominals has effects on the way NPs

inside the larger DP are assigned case. To illustrate, let us consider (16) and (17) from Fassi

Fehri (1993):

(16) ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i l-mašruuʔ-a]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

‘The man criticizing the project annoyed me.’ (Fassi Fehri 1993, ex. 46: 234)

(17) ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i li-l- mašruuʔ-i]

annoyed-me criticizing-NOM the-man-GEN of-the-project-GEN

‘The man’s criticizing of the project annoyed me.’ (Fassi Fehri 1993, ex. 47: 234)

(more or less) is marked for case, as in IE languages with case concord; these would be the result of the case feature
distributing from the nominal as a whole to all the words inside the nominal, perhaps at PF.”
265
Fassi Fehri (1993: 234) shows that the process nominals in (16) and (17) are both internally

verbal but externally nominal. Thus, process nominals in (16) and (17) take whatever arguments

that their inflected verbs take, and they both can be modified by manner adverbs, as in (18):

(18) a. ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i l-mašruuʔ-a

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

bi-šiddat-in]56

with-violence-GEN

‘The man criticizing the project with violence annoyed me.’

b. ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i li-l- mašruuʔ-i

annoyed-me criticizing-NOM the-man-GEN of-the-project-GEN

bi-šiddat-in]

with-violence-GEN

‘The man’s criticizing of the project with violence annoyed me.’

56
The NP šiddat ‘violence’ in the adverbial PP receives its genitive case from the preposition bi- ‘with.’
266
To derive the different cases assigned to the theme arguments in (16) and (17), Fassi

Fehri (1993) argues that the categorial conversion takes place at different stages in the

derivation. To derive the case facts of (16), Fassi Fehri (1993: 240) provides the structure in (19):

(19) DP
3
D NP
3
N VP
Affix 3
DP V’
l-rajul 3
the-man V DP
ntqd l-mašruuʔ
the-project

In (19), the verb is nominalized after VP is projected. This allows the verb to assign the

accusative case to the theme object. By contrast, the case facts of (17), can, according to Fassi

Fehri (1993: 235), be derived if we assume that the structure is as in (20):

(20) DP
3
D NP
3
DP N’
l-rajul 3
the-man N KP
3 3
V Affix li-l-mašruuʔ
Ntqd of-the-project

267
In (20), the abstract nominal affix is adjoined to V early on in the derivation; therefore, the theme

argument cannot be assigned the accusative case by the verb, and instead the genitive case is

assigned to the theme object via the prepositional strategy.

Given that the theory adopted in this dissertation is different from that adopted by Fassi

Fehri (1993), how can the case facts of the theme object in (16) and (17) be derived using a

dependent case approach? Let us consider (16) repeated here as (21) for convenience:

(21) ʔaqlaqa-nii [DP ntiqaad-u r-rajul-i l-mašruuʔ-a]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

‘The man criticizing the project annoyed me.’ (Fassi Fehri 1993, ex. 46: 234)

Following Fassi Fehri (1993), I assume that (21) has the structure shown in (22):

268
(22) DP
3
D NP
ntiqaad-u 3
criticizing-NOM <N> vP
3 3
V Affix DP v’
l-rajul 3
the man <v> VP
3
<V> DP
Spell-Out of Spell-Out ntqd l-mašruuʔ
DP of vP criticize the-project

In (22), the lexical root starts the derivation as a verb, which moves to v, and is then nominalized

by adjoining to an abstract nominal affix. N is then moved to D. At the Spell-Out of the soft

phasal head v, its complement is sent to Spell-Out. However, no case assignment takes place, as

there is only one NP in the domain, namely l-mašruuʔ ‘the project’. At the Spell-Out of the hard

phasal head D, its complement NP is sent to Spell-Out. In this domain, there are two NPs. The

agent argument l-rajul ‘the man’ c-commands the theme argument l-mašruuʔ ‘the project’. As a

result, the theme argument is assigned the dependent accusative case. As for the agent argument

(the possessor), it is assigned the structural genitive case assigned by D. The theme argument is

accessible at the Spell-Out of D, as the former is part of a soft phase vP. The DP as a whole will

receive the structural nominative case via th case-assigning Agree with T. However, the

complements of D are at this point in the derivation inaccessible to T, given that DP is a hard

phase. D, on the other hand, is accessible to T, given that D is the head of the phase DP;

therefore, it is accessible to T, and thus realizes the nominative case assigned to the DP as a

whole.

269
Let us now consider (17) repeated here as (23) for convenience:

(23) ʔaqlaqa-nii [DP ntiqaad-u l-rajul-i li-l- mašruuʔ-i]

annoyed-me criticizing-NOM the-man-GEN of-the-project-GEN

‘The man’s criticizing of the project annoyed me.’ (Fassi Fehri 1993, ex. 47: 234)

To account for the genitive case of the theme argument, I follow Fassi Fehri (1993) by assuming

that the theme argument is assigned the gentive case via the prepositional strategy because V (or

v in current minimalist theory) adjoins to the nominal affix early on in the derivation. The same

analysis adopted for (16) applies here with the exception that the theme argument is assigned the

gentive case by the preposition rather than the dependent accusative case.

More complicated and interesting are cases where the complement of an event-denoting

nominal receives a genitive case rather than an accusative case, as in (24):

(24) a. y-uriid-Ø-u ntiqaad-a l-rajul-i/*l-rajul-a

3-want-MSG-INDIC criticizing-ACC the-man-GEN/the-man-ACC

bi- šiddat-in

with-violence-GEN

‘He wants to criticize the man with violence.’ (adapted from Fassi Fefri 1993, ex.

65a: 242)

270
b. y-uriid-Ø-u ntiqaad-a nafs-i-hi/*nafs-a-hu

3-want-MSG-INDIC criticizing-ACC self-GEN-his/self-ACC-his

bi- šiddat-in

‘He wants to criticize himself with violence.’ (adapted from Fassi Fehri 1993, ex.

65b: 242)

To account for the puzzling genitive case on the complement of the event-denoting nominals in

(24), Fassi Fehri (1993: 242-243) proposes the following solution: the event-denoting nominal in

(24b) has the internal structure in (25):

(25)

DP
3
D NP
ntiqaad-a 3
criticizing-ACC N VP
3
DP V’
PRO 3
V DP
ntqd nafsi-hi
criticize himself [GEN]

In this account, V raises to N before raising further to D. To account for the genitive case on the

complement of the event-denoting nominal, Fassi Fehri (1993: 242-243) claims that the genitive

271
case is checked only at LF, and that the null subject PRO inside the structure of the event-

denoting nominal is caseless. Fassi Fehri claims that the thematic object (i.e. the reflexive) in

(24b) must raise to the Spec, DP position at LF to discharge the genitive case assigned to it by D.

To account for why the reflexive in (24b) cannot receive the accusative case, Fassi Fehri (1993:

243) attributes this to the condition on case discharge in (26):

(26) Object case is discharged only if subject case is discharged.57

Fassi Fehri claims that the transitive verb ntqd ‘to criticize’ in (24b) cannot assign the accusative

case to the reflexive pronoun because the subject PRO in its Spec does not discharge its case, as

the latter is caseless, being PRO. This, together with the fact that PRO cannot absorb the genitive

case assigned by D, explains, according to Fassi Fehri, why the reflexive pronoun cannot receive

the accusative case and surfaces instead with the genitive case assigned to it by D.

Let us point out that Fassi Fehri’s (1993) solution to the puzzling genitive case in (26) is

not available to us for the following reasons. First, in current case theory, it is no longer possible

to claim that the assignment of case takes place at LF. Resort to case assignment at LF has been

abandoned ever since the operation Agree was introduced into the Minimalist program, as a

syntactic operation. In other words, in current generative accounts of case, case checked at LF

has been reformulated as case checked in the syntax via a long-distance Agree. Second, given the

theoretical framework adopted in this work, structural case is assigned at Spell-Out, the point

where syntax interfaces with PF. Therefore, any resort to case assignment at LF is not possible.

This means that a different account is needed to explain the puzzle. I assume along with Fassi

57
Fassi Fehri (1993) acknowledges that the condition in (26) is analogous to the case tier approach (Yip et al. 1987),
where there is a hierarchy of case assignment such that the accusative case can only be assigned when the
nominative case is assigned. Fassi Fehri (1993, fn. 31: 278) claims, analogous to Yip et al. (1987), that case
assignment follows a hierarchy in which nominative case is the highest.
272
Fehri (1993) that the internal structure of the event-denoting nominal of (24b) is that in (25).

Specifically, I assume that the internal structure of the event-denoting nominal in (24b) includes

the null subject PRO in the Spec, vP position. However, I take the morphological fact that the

theme object can only receive the genitive case but not the accusative case as an indication that

the null subject category PRO is not a case competitor in SA.58’59 Once this assumption is made,

the facts follow straightforwardly. Thus, I argue that the structural representation of (24b) is as in

(27):

(27) [CP [TP y-uriid-Ø-u proi [DP ntiqaad-a-N-D [NP <v-N>

3-want-MSG-INDIC criticizing-ACC

[vP PROi <v> nafs-i-hi bi-šiddat-in]]]]]]

self-GEN-his with-violence-GEN

‘He wants to criticize himself with violence.’ (adapted from Fassi Fehri 1993, ex. 65b:

242)

58
The claim that PRO is not a case competitor in SA might be a language-specific option, as it has been argued that
PRO receives case in many languages.
59
Youssef Haddad raises the following question: could it be that PRO is actually a case competitor but that GEN is
the dependent case while ACC is the unmarked case inside DP? I believe that this suggestion faces the following
problem: It makes the wrong prediction that the dependent case is always GEN inside the DP, as the accusative case
obligatorily surfaces on the theme object when the subject (i.e. the possessor/agent) inside process-nominals is a
lexical NP rather than PRO. Even if we make the suggestion that GEN is a higher dependent case in the DP of SA,
we face the following two problems: first, SA would be a language where the dependent case is the lower case
inside the CP domain, but it is the higher case inside the DP domain. Second, when the agent/possessor is PRO, both
the agent/possessor and the theme would surface with GEN and there will be no case competition, contrary to the
observational fact that GEN surfaces on the higher of the two NPs and ACC surfaces on the lower of the two NPs in
process nominals where the higher NP is a lexical NP rather than PRO..
273
The derivation of (24b) proceeds as in (28):

(28) a. [vP PRO v [VP V NP[reflexive]]], v a phase head

Spell out v’s complement: NP[reflexive]

→ Case NP[reflexive] → NA

b. [DP N-D [NP <v-N> [vP PRO <v> [VP V NP[reflexive]]]]], D a phase head

Spell out D’s compement: PRO c-c’s NP[reflexive]

→ (PRO < NP[reflexive])

Case NP[reflexive] → genitive inside the DP domain assigned by D,

as PRO is not a case competitor in SA

The derivation in (28) shows that the theme object, namely the reflexive NP, is not assigned the

dependent case at the Spell-Out of the phasal head v, as there is only one NP in this domain.

Therefore, no case calculation is considered. At the merger of the hard phasal head D, its

complement is sent to Spell-Out. In this domain, the reflexive NP inside the vP domain is

accessible to case calculation, as v is a soft phase. Crucially, the NP reflexive does not receive

the dependent accusative case even though it is c-commanded by another NP, namely PRO at the

Spell-Out of D. This is because PRO is not a case competitor in SA. Therefore, the reflexive

object is assigned genitive case by D. The reflexive is accessible to D because it is part of a soft

phase, namely vP. Thus, unlike pro, which has been argued in this dissertation to be a case

competitor, PRO is not. To explain why certain null categories can in some languages be case

competitors but not others, Baker (2015: 201) proposes the hierarchy of case competitors in (29):

274
(29) Overt NPs and clitics > pro > controlled PRO > arbitrary PRO > implicit agent of

passives/agent of nominal nonspecific object > PP, VP, etc.

Baker (2015: 201) states that “categories to the right on the scale have a subset of nominal

features that categories to the left on the scale have. Languages then vary as to which of these

features are minimally necessary in order to participate in dependent case assignment.” Baker

then continues to state that “some [languages] rule out any covert NPs from being case

competitors, some allow pro to be a case competitor but not PRO, some allow PRO but not the

implicit agent in a passive, and so on[emphasis added].”60 I take the facts of genitive case

assignment in structures with event-denoting nominals as a confirmation of Baker’s (2015)

proposed hierarchy. Specifically, the facts from SA show that pro is a case competitor but PRO

is not. This means that PRO in SA must lack some nominal feature, which is minimally required

in SA, in order for a category to be a case competitor.

7.4 Testing Baker’s dependent case theory against the Case Tier theory and the Agree-

based theory of case: Evidence from structures with event-denoting nominals

Let us consider again the structures in (24) repeated here as (30):

(30) a. y-uriid-Ø-u ntiqaad-a l-rajul-i/*l-rajul-a

3-want-MSG-INDIC criticizing-ACC the-man-GEN/the-man-ACC

60
The fact that PRO is caseless is surprising in the adopted theoretical feramework. In the present thesis, to
participate in case assignment, the two NPs involved must have distinct referential indices. PRO is an NP; therefore,
it must have a referential index, following Baker’s (2003) theory of lexical categories which we adopt. In structures
with event-denoting nominals; the prediction would then be that PRO should be a case competitor given that it has a
referential index which is distinct from that of the theme object in these structures; yet, it does not. At this point in
time I have no explanation as to why this is the case, and the proposal that PRO in SA does not participate in
dependent case assignment remains a stipulation dictated by Baker’s (2015) hierarchy of dependent case triggers
which is itself a stipulation.
275
bi- šiddat-in

with-violence-GEN

‘He wants to criticize the man with violence.’ (adapted from Fassi Fefri 1993, ex.

65a: 242)

b. y-uriid-Ø-u ntiqaad-a nafs-i-hi/*nafs-a-hu

3-want-MSG-INDIC criticizing-ACC self-GEN-his/self-ACC-his

bi- šiddat-in

‘He wants to criticize himself with violence.’ (adapted from Fassi Fehri 1993, ex.

65b: 242)

It is crucial at this juncture to point out that the obligatory genitive case of the thematic

complement of process nominals offers a testing ground for a comparison between the updated

version of dependent case (Baker 2015) on the one hand, and the Case Tier approach of Yip et al.

(1987) and the agreement-based approaches to case (Chomsky 2000, 2001) on the other.

Specifically, this is a case, where only the updated version of dependent case (Baker 2015)

makes the right predictions, and where the Case Tier approach (Yip et. al.) and the agreement-

based approach (Chomsky 2000, 2001) make the wrong predictions. Thus, in the approach

adopted in this dissertation, the genitive case on the thematic object of the process nominal is

predicted on the grounds that PRO in SA is not a case competitor in the sense of Baker (2015)

together with the assumption that vP is a soft phase in SA, which makes it accessible to the

276
genitive-assigning head, D via Agree. However, in the Case Tier approach, the genitive case of

the thematic object complement of the process nominal can only be explained on the grounds

that both PRO and the thematic object complement are in the same case domain, namely DP

(termed NP in Yip et al. 1987), and the DP in this approach supplies a case tier with one case

namely genitive case (see specifically Yip et al. 1987: 233), which is spread from PRO to the

theme object. Note, however, that this solution faces the problem of undergeneralization, since

this approach would predict that the theme argument can never be realized with the accusative

case, contrary to the facts, as shown in (21) above, which is repeated as (31) below:

(31) ʔaqlaqa-nii [DP ntiqaad-u r-rajul-i l-mašruuʔ-a]

annoyed-me criticizing-NOM the-man-GEN the-project-ACC

‘The man criticizing the project annoyed me.’ (Fassi Fehri 1993, ex. 46: 234)

Similarly, it is hard to see how an agreement-based account of case can explain the

obligatory genitive case on the thematic object complement of the process nominal. This is

because the verb in these structures starts the derivation as a transitive verb. This means that an

agreement-based account would predict that the thematic object of the process nominal would

always receive the structural accusative case via an Agree relation with the functional head v*,

contrary to the facts. If, however, v in SA is not capable of establishing an Agree relation with

the object and assigning the accusative case to it, as I claim in this thesis, then the facts follow. In

the absence of the dependent case mechanism, the Agree-based case assigning mechanism takes

over. In the structure of (30), there are two potential case assigners, D, and v. However, v does

not assign case in SA; therefore, D is the only functional head capable of establishing an Agree

relation with the theme object and assigning the structural genitive case to it.

277
To summarize, structures with event-denoting nominals provide evidence for the

dependent case approach and against the Agree-based approach and the Case Tier approach.

7.5 Case assignment in structures with pronominal possessives

Let us consider the following pair of examples with process nominals and pronominal

possessives:

(32) a. ʔaqlaqa-nii [DP ntiqaad-u-hu zayd-an]

annoyed-me criticizing-NOM-his.GEN Zayd-ACC

‘His ciriticism of Zayd annoyed me.’ (Fassi Fehri 1993, ex. 73: 245)

b. *ʔaqlaqa-nii [DP ntiqaad-u-hu zayd-in]61

annoyed-me criticizing-NOM-his.GEN Zayd-GEN

Intended to mean ‘Zayd’s criticism of him annoyed me.’ (Fassi Fehri 1993, exs.

74a-b: 245)

c. ʔaqlaqa-nii [DP ntiqaad-u zayd-in la-hu/ʔiyyaa-hu]

annoyed-me criticizing-NOM Zayd-GEN of-him.GEN/ʔiyyaa-him.GEN

‘Zayd’s criticism of him annoyed me.’

61
Non-nominative pronouns in SA are syncretic in accusative and genitive environments. In the examples under
discussion, the non-nominative pronominal clitic hu- can be an accusative form translating into English as ‘him’ and
it can also be a genitive form translating into English as ‘his’.
278
The question of interest is how to account for the grammaticality judgments in (32a-b) using the

dependent case theory of Baker (2015). The example in (32a) can be easily accounted for as

follows: D is a hard phasal head in SA. At the insertion of D, its complement is sent to Spell-Out.

In this domain, two NPs are available, namely the agent pronominal clitic hu- ‘his’ and the theme

object Zayd ‘Zayd’; therefore, the theme object zayd ‘Zayd’ receives the dependent accusative

case, and the agent pronominal clitic is assigned the genitive case by D. To account for the

judgment in (32b), we have to ask the following question: What makes the following analysis

wrong: the theme pronominal clitic is assigned the dependent accusative case and the agent NP

zayd ‘Zayd’ is assigned the genitive case by D. We can rule out (32b) by claiming that ntiqaadu

zaydin ‘Zayd’s criticism’ is a construct state, which forms one phonological unit such that

nothing can linearly intervene between the nominalized verb in D and the possosser/agent Zayd

‘Zayd’.62 The only way for (32b) to be grammatical is when the clitic is hosted by a preposition,

as in (32c).

Let us consider another more interesting example of process nominals that are derived

from ditransitive verbs, as in (33):

(33) [DP salb-u [NP [vP Zayd-in [VP l-rajul-a maal-a-hu ]]]]

depriving-NOM Zayd-GEN the-man-ACC money-ACC-his

‘Zayd’s depriving the man of his money’ (Fassi Fehri 1993, ex. 78a: 247)

62
I thank Youssef Haddad for suggesting this analysis to me.
279
The example in (33) can be easily accounted for using the dependent case theory as follows: At

the insertion of the soft phasal head v, its complement VP is sent to Spell-Out. In this domain,

there are two NPs available, the source argument and the theme object. As a result, the theme

object is assigned the dependent accusative case, and the source argument remains without a case

value. At the insertion of the hard phasal head D, its complement NP is sent to Spell-Out. In this

domain, three NPs are available, the agent NP, the source NP and the theme NP. The agent NP c-

commands both the source NP and the theme NP. As a result, the source NP is assigned the

dependent accusative case, and the theme NP is redundantly assigned the dependent accusative

case. The theme NP is available at the Spell-Out of D because it is part of a soft phase vP. As for

the agent NP, it is assigned the genitive case by D. If this DP is in the subject position of the

clause, as is hypothesized here, it receives the structural nominative case via Agree with T.

However, aside from D, the complements of D are no longer accessible at the point in the

derivation when T agrees with the DP, given that DP is a hard phase. D, on the other hand, is still

accessible to T, as it is the head of the phase DP. There, the D head of DP realizes the structural

nominative case assigned via Agree with T.

7.6 Summary

Following Baker (2015), I have assumed that the DP is a hard phase in SA. I have

demonstrated that the complement of D is assigned the structural genitive case by D. I have also

shown that structures with event-denoting nominals can be properly accounted for using the

updated version of dependent case (Baker 2015) as a theoretical framework. I have further

shown that structures with event-denoting nominals provide a testing ground for the dependent

case theory against the Case Tier approach and the Agree-based approach. Using these
280
strucutures, I have provided evidence that only the dependent case theory makes the right

predictions for case assignment/values in SA.

281
Chapter Eight

Conclusions

8.1 Conclusions

In this thesis, I have argued that agreement-based theories of case assignment in SA (e.g.

Raḥḥali 2003, Soltan 2007, 2011, Al-Balushi 2011, 2012) face problems. I have also argued that

the updated dependent case theory (Baker 2015) can better account for the case facts of SA and

that such a theory overcomes the problems that other theories face. The major argument of this

thesis is that case-assigning Agree does apply but only in those cases when it is not bled by the

dependent case. Adopting the updated dependent case theory of Baker (2015), I have shown that

SA exhibits the properties listed in (1):

(1) a. CPs, PPs and DPs are hard phases, but vPs are soft phases. T assigns the structural

nominative case to the subject in the CP domain; P assigns the structural gentive

case inside the phasal PP domain, whereas D assigns the structural genitive case

inside the phasal DP domain.

b. SA is governed by one rule of case assignment according to which an NP1, which

is c-commanded by NP2 in any phasal domain (VP or TP) is assigned the

dependent case Z in its respective domain. Thus, if an NP1 is c-commanded by an

NP2 in VP, which is the spell out domain of the phasal head v, NP1 is assigned

the dependent accusative case in VP. Similarly, if an NP1 is c-commanded by

NP2 in TP, which is the spell out domain of the phasal head C, NP1 is assigned

the dependent accusative case in TP. This assumption has been crucial in

282
analyzing cases of double object constructions as well cases of ECM

constructions in SA.

c. v is a soft head in SA, which is not even capable of establishing an Agree relation

with the object in the absence of the dependent case mechanism. This is due to its

impoverished feature specification, which includes nothing more than a gender

feature. This contrasts with T, which has the full complement of ɸ-features, and is

therefore capable of assigning the structural nominative case to the subject.

d. In order to take part in dependent case, the NPs contained in a spell out domain

should have distinct referential indices.

e. The genitive case assigned to the NPs inside PPs is structural rather than purely

lexical or inherent.

f. NPs in the left periphery of the clause may appear with either a nominative case,

or with an accusative case. If the NP in the left periphery appears with the

nominative case, this is the unmarked case. If, on the other hand, the NP in the left

periphery appears with the accusative case, this case is the dependent case

assigned to a lower link in a chain whose head is the NP in the left periphery.

g. Adverbial NPs in SA receive the semantic/adverbial case.

h. Adjectival and nominal predicates in SA receive an unmarked nominative case in

the CP domain, unless there is a copular (overt or covert) in the structure, in

which case these predicates receive a purely idiosyncratic lexical case from the

copular verb.

283
i. The indicative complementizer and the copular verb in SA assign a purely

idiosyncratic lexical accusative case rather than structural accusative case.

j. The null category pro is a case competitor which triggers the dependent

accusative case on the lower NP in its spell out domain. This contrasts with the

null category PRO which is not a case competitor in that it does not trigger the

dependent case on the lower NP inside its spell out domain.

At this point, one might raise the following question: If v is incapable of establishing

Agree and assigning accusative case to the object because v’s ɸ-specification is impoverished,

being specified only for a gender feature, what enables v in Indo-European languages to establish

Agree and assign the accusative case to the object, despite the fact that v in these languages is

also impoverished?63 To answer that, I tentatively suggest that it could be the case that lacking

person and number features together, as is the case of v in SA might be the reason why v in SA is

weaker than other Indo-Eurpoean languages (where v might not be specified for a person feature

only or a gender feature only).

I conclude therefore that the updated version of the dependent case theory (Baker 2015)

makes the right predictions for SA.

Let us now consider the categories that participate in dependent case assignment in SA.

Table 1 shows the categories observed in this study and their participation or lack thereof in case

assignment:

63
I thank Gabriela Alboiu for raising this question.
284
Table1. The categories of case triggers in SA.

Categories Participation in triggering the dependent case

on another NP in the same spell-out domain

Overt referential NPs YES

Null referential pro YES

PRO NO

Overt existential expletives NO

The important question to answer with reference to Table 1 is what makes some of these

categories trigger the dependent case on another NP in the same spell-out domain, and what

makes others incapable of acting as triggers for dependent case. In this thesis, I suggest,

following Baker (2003, 2015) that it is the distinct referential indices of the NPs contained in a

spell out domain that makes some NPs receive the dependent accusative case and others do not.

In Baker’s (2015: 174) words, “a noun is by definition a lexical category that bears a referential

index.” Baker then adds that “it becomes easy and natural to say that two nominals that interact

case-theoretically must not only have indices, but must have distinct indices.” One way of trying

to explain the facts of SA summarized in Table 1 might be as follows: overt referential NPs

trigger the assignment of dependnt case on other NPs in the same spell-out domain because they

(a) have referential indices and (b) they bear distinct referential indices from those of the NPs,

which bear the dependent case. That is why members of the same chain do not receive distinct

cases.

How can we explain the fact that although both pro and PRO bear referential indices and

both bear distinct referential indices from the other NPs in their spell-out domains, only pros

285
trigger the dependent case on another NP in the same spell-out domain? At this point in time, this

claim is a stipulation dictated by Baker’s (2015) hierarchy of the triggers of dependent case.

The next question is this: why can’t existential expletives act as triggers of case

assignment in SA? I believe that this can be attributed to the idea that the referential indices

borne by expletives are not distinct from those of their associates.

Lastly, one might raise the following question: Why can’t the dependent case mechanism

apply to adjectival and nominal predicates? In this thesis, I suggest, following Baker (2015), that

the referential index borne by predicate nominals is also borne by other projections which

contain the predicate nominal, specifically EP and PredP. Therefore, the predicate nominal

cannot be assigned the dependent case since it is not a maximal projection with a referential

index. As for why adjectival predicates do not receive the dependent case, this is quite predicted

in the theoretical framework adopted, as adjectives and verbs do not, according to the theory of

lexical categories of Baker (2003) which we adopt, have a referential index.

One final question to be addressed before this thesis is concluded is the following: I have

argued, following Baker (2015), that there are two mechanisms of case assignment that can

account for the case alteration of what is called structural or abstract cases in the Minmalist

literature. These two mechanisms are the dependent case of Baker (2015) and the Agree-based

case of Chomsky (2000, 2001). Following Baker (2015), I have also posited a hierarchy of case

mechanisms such that dependent case, when it applies, bleeds Agree-based case. The question

that might now be asked is the following: are we not losing the explanatory adequacy that Agree-

based accounts of case afford us when opting for a dependent case account which is rule-based

rather than parameter-based?64 In other words, are we not risking the loss of explanatory

64
I thank Youssef Haddad, Arsalan Kahnemuyipour, Ruth King and Gabriela Alboiu for rasing this question.
286
adequacy when we opt for the descriptive adequacy that can be achieved by taking into account a

dependent case approach to case? I believe that there are two reasons why a dependent case

account should be seriously entertained. The first reason is that the Agree-based account of case

is now controversial and has been criticized by a number of linguists (see for example, Marantz

1991, McFadden 2004, Diercks 2012). The second reason is that there does not seem to be

conclusive evidence that a parameter theory of language variation is in any way superior to a

rule-based theory of language variation. As pointed out in Newmeyer (2004: 189), “[…] in all

cases, a rule-based account is either more adequate than a parameter-based one or that, when the

facts are taken into account, they are empirically indistinguishable.”

To conclude, this thesis advances the following hypothesis: some linguists (see Al-

Balushi: 2011, 2012) argue against Chomsky’s (2000, 2001) hypothesis that only a probe with

the full complement of ɸ-features can assign case on the grounds that SA is a language, where T

is a defective probe (lacking the feature number in the VSO order); yet, it can still assign the

structural nominative case to the subject. In contrast to this hypothesis, I argue in this thesis that

(a) T is a probe with the full set of ɸ-features; therefore, it can assign the structural nominative

case to the subject;

(b) v is a defective probe (following Wumbrand and Haddad 2016); therefore, it cannot assign

the structural accusative to the object;

(c) the structural accusative case assigned to the object in SA is not the result of v agreeing with

the object, as v is incapable of assigning case to the object; rather, this case is the result of the

dependent case mechanism, which bleeds the Agree-assigning case mechanism.

287
To conclude, the case facts of SA can be best be accounted for using the dependent case

theory of Baker (2015). This theory overcomes the problems that other accounts of case face

without any loss of descriptive adequacy.

8.2 Future research

One of the issues that I have not addressed in this thesis is the issue of case alternation in

vocative constructions in SA. In these constructions, the DP might surface with either the

nominative case or the accusative case, as shown below:

(2) yaa rajul-an, t-akallam-Ø-Ø

O man-ACC 2-speak-MSG-JUSS

‘O man, speak!’ (Moutaoukil 1986, ex. 21a: 168)

(3) ʔayyuhaa l-naaʔim-u, ʔistayqiḏ-Ø-Ø

O the-sleeper-NOM wake.up-MSG-JUSS

‘O the sleeper, wake up!.’ (Moutaoukil 1986, ex. 43c: 175)

. The question that immediately arises is why. I leave this issue for future research.

Another very important question which is not answered here is why pro but PRO

competes for case. I leave this for future research.

288
References

Adger, David. 2003. Core syntax: A minimalist approach. Oxford: Oxford University Press.

Ahmed, Amer. 2015. On agreement affixes, incorporated pronouns and clitics in Standard

Arabic. SKY Journal of Linguistics, 28, 67-102.

Al-Ašram, Al-Mitwalli Ali Al-Mitwalli. 2003. ʔal-ḏarfu: xaṣaaʔiṣuhu wa tawḏiifuhu l-

naḥawiyyu [Adverbs: Their properties and their syntatictic functions]. Al-Manṣuura,

Egypt: Jaziratu l-Wardi Publishing Company.

Al-Balushi, Rashid. 2016. The ‘believe’- construction in Standard Arabic. Journal of Linguistics,

52, 1-36.

Al-Balushi, Rashid. 2012. Why verbless sentences in Standard Arabic are verbless. Canadian

Journal of Linguistics, 57, 1, 1-30.

Al-Balushi, Rashid. 2011. Case in Standard Arabic: The untraveled paths. Doctoral dissertation,

University of Toronto.

Alboiu, Gabriela. 2009. Null explatives and Case. The view from Romance. In Romance

languages: Structures, interfaces and microparametric variation. Amsterdam: John

Benjamins.

Alboiu, Gabriela. 2006. Are we in agreement? In Agreement systems, ed. Cedric Boekcx, 13-139,

Amsterdam: John Benjamins.

Alboiu, Gabriela and Virginia Hill. 2016. Root gerunds in Old Romanian. In Romance

Linguistics 2013: Papers selected from the 43rd Symposium on Romance Linguistics

(LSRL) New York, 1-21, ed. by Christina Tortora et al. Romance Languages and

Linguistic Theories (9). Amsterdam/Philadelphia: John Benjamins.

289
Al-Hawary, Mohammad T. 2011. Modern Standard Arabic: A learner’s guide. Malden, MA:

Wiley Blackwell.

Al-Maxzuumy, Mahdi 1985. Fii l-naḥwi l-ʕarabiyi: naqdun wa tawjiihun. 2nd ed [On Arabic

grammar: A critique and a direction for the future]. Beirut: Al-Raaʔidu l-ʕarabiyu

Publishing House.

Al-Naadiry, Mohammad Asʕad. 1995. Naḥu l-luġatil-ʕarabiyati: kitaabun fii qawaaʕid-i l-naḥui

w l-ṣarfi [The grammar of Arabic: A book on the rules of syntax and morphology].

Beirut: Al-Matabatu l-ʕaṣriyatu for Publishing and Distribution.

Al-Naḥas, Mustafa. 1995. Min qaḍaaya l-luġa [Some Linguistic issues]. Kuweit: Kuweit

University Press.

Al-Waraaq, Abu l-Hasan. 10th c./1999. ʕilalu l-naḥwi [The arguments from syntax]. Al-Riyaḍ,

Saudi Arabia: Al-Rušd Publishing Company.

Aoun, Joseph, Elabbas Benmamoun and Lina Choueiri. 2010. The syntax of Arabic. Cambridge:

Cambridge University Press.

Aoun, Joseph, Benmamoun, Elabbas, Sportich, Dominique. 1994. Agreement and conjunction in

some varieties of Arabic. Linguistic Inquiry, 25, 195-220.

Baalabki, Ramzi. 2008. The legacy of the Kitāb. Leidin: E.J. Brill.

Babby, L. 1986. The locus of case assignment and the direction of percolation: Case theory in

Russian. In Case in Slavic, ed. by R. Brecht and J. Levine, 170-219. Columbus, OH:

Slavica.

Babby, L. Freidin, Robert. 1984. On the interaction of lexical and syntactic properties: Case

structure in Russian. Cornell Working Papers in Linguistics, 6: 71-105.

290
Baker, Mark. 2015. Case: Its principles and parameters. Cambridge: Cambridge University

Press.

Baker, Mark. 2014. On dependent ergative case (in Shipibo) and its derivation by phase.

Linguistic Inquiry 45,3, 341-379.

Baker, Mark. 2003. Lexical categories: Verbs, nouns and adjectives. Cambridge: Cambridge

University Press.

Baker, Mark. 1997. Thematic roles and grammatical categories. In Elements of grammar:

Hanndbook of generative syntax, ed. by L. Haegeman. Doredrecht: Kluwer, 73-137.

Bejar, Susana and Daine Massam. 1999. Multiple case checking. Syntax 2, 2, 65-79.

Benmamoun, Elabbas. 2008. Clause structure and the syntax of verbless sentences. In

Foundational issues in linguistic theory, ed. by Robert Freidin, Carlos P. Otero, and Maria Luisa

Zubizarreta, 105–131. Cambridge, MA: MIT Press.

Benmamoun, Elabbas. 2000. The feature structure of functional categories: A comparative study

of Arabic dialects. USA: Oxford University Press.

Benmamoun, Elabbas. 1999. Spec-head agreement and overt Case in Arabic. In Specifiers:

Minimalist approaches, ed. by David Adger, Susan Pintzuk, Bernadette Plunkett and

George Tsoulas, 110-125, Oxford: Oxford University Press.

Bin Muqbil, Musaed S. 2006. Phonetic and phonological aspects of Arabic emphatics and

guturals. Doctoral dissertation, University of Wisconsin-Madison

Burzio, Luigi. 1986. Italian syntax. Reidel: Doredrecht.

291
Bobaljik, Jonathan. 2008. Where’s phi? Agreement as a postsytactic operation. In Phi theory:

Phi features across interfaces and modules, ed. by Alana Johns, Diane Massam, and

Juvenal Ndayiragije, 47-77, Doredercht: Springer.

Carstons, Vicky. 2001. Multiple agreement and case deletion: Against ɸ-(in)completeness.

Syntax, 4, 3, 147-163.

Chomsky, Noam. 2008. On phases. In Foundational Issues in Linguistic Theory: Essays in

Honor of Jean-Roget Vergnaud, ed. by Robert Freidin, Carlos P. Otero, and Maria Luisa

Zubizarreta, 133-166, Cambridge: The MIT Press.

Chomsky, Noam. 2007. Approaching UG from below. In Interfaces + recursion=language? :

Chomsky’s Minimalism and the view from syntax/ semantics, ed. by Uli Sauerland and

Hans-Martin Gartner, 1-31. New York: Mouton de Gruyter.

Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36, 1, 1-22.

Chomsky, Noam. 2001. Derivation by phase. In A life in language, ed. by Michael Kenstiwics,

1-52. MIT: MIT Press.

Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays on

Minimalist syntax in honor of Howard Lasnik, ed. by R. Martin, D. Miachaels, and Juan

Uriagereka, 89-155. Cambridge, MA: MIT Press.

Chomsky, Noam. 1995. The Minimalist Program. MIT Press.

Chomsky, Noam. 1991. Some notes on economy of derivation and representation. In Principles

and Parameters in comparative grammar, ed. by Robert Freidin, 417-454. Cambridge,

Massachusetts: The MIT Press.

Chomsky, Noam. 1986. Knowledge of language: Its nature, origin, and use. Prager: New York.

Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Fortis.

292
Chomsky, Noam. 1980. On binding. Linguistic Inquiry, 11, 1, 1-46.

Chomsky, Noam. 1977. On wh-movement. In Formal Syntax, ed. by P.W. Culicover, T. Wasow,

and A. Akmajian, 71-132, New York, NY: Academic Press.

Chomsky, Noam & Howard Lasnik. 1993. The theory of principles and parameters. In. Syntax:

An International handbook of contemporary research, ed. by J. Jacobs, von Stechow, W.

Sternefield, and T. Vennemann, 506-569. Berlin: Walter de Gruyter.

Cowper, Elizabeth. 2005. The Geometry of Interpretable Features: Infl in English and Spanish.

Language, 81, 1:10-46.

De Hoop, Helen and Zwarts, Joost. 2009. Case in formal semantics. In The handbook of case, ed.

by Andrej Malchucov and Andrew Spenser, 170-181. Oxford: Oxford University Press.

Diercks, Michael. 2012. Parameterizing case: Evidence from Bantu. Syntax 15, 3, 253-286.

Eid, Mushira. 1983. The copula function of pronouns. Lingua 59: 197-207.

Embick, David, and Rolf Noyer. 2005. Distributed morphology and the syntax/morphology

interface. In The Oxford handbook of linguistic interfaces, ed. by G. Ramchand and C.

Reiss, 1-28. Oxford: Oxford University Press.

Fassi Fehri, Abdelkader. 2012. Key features and parameters in Arabic grammar. Amsterdam:

John Benjamins.

Fassi Fehri, Abdelkader. 1999. ʔal-muʕjamu l-ʕarabiyy: namaaðiju taḥliiliyatun jadiida [The

Arbic lexicon: New analytical models], Casablanca: Tubqal Publishing Company.

Fassi Fehri, Abdelkader. 1993. Issues in the structure of Arabic clauses and words. Kluwer

Academic publishers: Dordrecht.

293
Fassi Fehri, Abdelkader. 1990. ʔal-binaaʔ l-muwaazi: Naḏariyya fii binaaʔ l-kalima wa l-jumla

[The parallel structure: A theory of word snd clause structure]. Casablanca: Tubqal

Publishing Company.

Fassi Fehri, Abdelkader. 1986. ʔal-lisaaniyaatu wa l-luġatu l-ʕarabiyyatu [Linguistics and the

Arabic language]. ʕuwaidaat Publishing House: Beirut, Lebanon.

Freidin, Robert and Rex A. Sprouse. 1991. Lexical case phenomena. In Principles and

parameters in comparative grammar, ed. by Robert Freidin, 392-416. The MIT Press:

Massachusetts

Haegeman, Liliane. 1991. Introduction to Government and Binding theory.2nd ed. Blackwell:

Oxford.

Hasan, Abbas. 1962. ʔal-naḥu l-waafii [The comprehensive grammar]. 2nd Vol. 3rd ed. Cairo: Al-

maʕaarif Publishing House.

Hasan, Abbas. 1963. ʔal-naḥu l-waafii [The comprehensive grammar]. 3rd Vol. 3rd ed. Cairo: Al-

maʕaarif Publishing House.

Hazout, Ilan. 2010. Verbless sentences and clause structure. Linguistic Inquiry, 41, 3, 471-485.

Hornstein, Norbert. 1999. The movement theory of control. Linguistic Inquiry, 30, 1, 69-96.

Hornstein, Norbert, Jairo Nunes, and Kleenthes K. Grohmann. 2005. Understanding

Minimalism. Cambridge: Cambridge University Press.

Ibn Al-Anbari, Abdulraḥmaan Ibn Muḥammad 11th c./1961. Al-Insaafu fii masaaʔili l-khilaafi

bayna l-naḥawiina l-Baṣriyiina wa l-Kuufiyiina, Al-ṯabʕatu l-raabiʕatu [The proper

judgment on the issues of contention between the Basran and the Kufan grammarians 4th

ed.]. Cairo: Al-Saʕaada Publishing Company.

294
Ibn Al-Saraaj, Abu-Bakr. 10th c./1996. ʔal-ʔuṣuul-u fii l-naḥw [The principles of grammar]. 3rd

ed, Vol.2. Beirut: Al-Resalah Publishing Company.

Ibn ʕaqiil, Bahaaʔuldiin. 13th c./1980. šarḥu Ibn ʕaqiil ʕalaa ʔalfiyati Ibn Maalik. Al-mujallad

Al-ṯaani [Ibn ʕaqiil’s interpretation of Ibn Maalik’s thousand verses of of Arabic

grammatical theory. Vol. 2]. Cairo: Al-Turaaṯ Publishing House.

Ibn ʕaqiil, Bahaaʔuldiin. 13th c./1980. šarḥu Ibn ʕaqiil ʕalaa ʔalfiyati Ibn Maalik. Al-mujallad

Al-ṯaani [Ibn ʕaqiil’s interpretation of Ibn Maalik’s thousand verses of of Arabic

grammatical theory. Vol. 1]. Cairo: Al-Turaaṯ Publishing House.

Johnson, Kyle. 1991. Object positions. Natural Language and Linguistic Theory 9: 577-636.

Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA. : MIT Press Working

Papers in Linguistics 18, 99-148. .

Koizumi, M. 1995. Object agreement phrases and the slpit VP hypothesis. In Papers on case and

agreement I: MIT Working Papers in Linguistics 18, ed. by Philips, C. & Jonathan

Bobalijk, 99-148.

Koizumi, M. 1993. Phrase Structure in Minimalist Syntax. PhD dissertation, MIT.

Kouloughli, Djamel. 2007. Sur la valeur du tanwiin: Nouvelle contribution à l’étude du système

determinative de l’arabe. Arabica 54: 94-131.

Landau, Idan. 2004. The scale of finiteness and the calculus of control. Natural language and

linguistic theory, 22,4, 811-877.

295
Lasnik, H. 2008. On the development of Case theory: Triumphs and challenges. In Foundational

issues in linguistic theory: Essays in Honor of Jean-Roger Vergnaud, ed. by R. Freidin,

C. Otero, and M.-L. Zubizarreta. MIT Press, pp. 17-41.

Lasnik, Howeard. 2001. Subjects, objects, and the EPP. In Objects and other subjects:

Grammatical functions, functional categories, and configurationality, ed. by Davies, W.

D. and Stanley Dubinsky, Doredrecht: Kluwer Academic Publishers: 103-121.

Lasnik, Howard. 1999. Minimalist analysis. Blackwell.

Legate, Julie Anne. 2008. Morphological and Abstract Case. Linguistic Inquiry, 39:1, 55-101.

Leung, Tommy. 2011. Mood feature as case licensor in Modern Standard Arabic. In Perspectives on

Arabic Linguistics XXII-XXIII, ed. by Ellen Broselow and Hamid Ouali, 127-148. Amsterdam:

John Benjamins.

Maling, Juan. 2009. The case tier: A hierarchical approach to morphological case. In the Oxoford

handbook of case, ed. by A. Malchukov & A. Spenser, 72-87. Oxford: Oxford University

Press.

Marantz, Alec. 1991. Case and licensing. In Proceedings of ESCOL 91, ed. by Westphal, G., Ao,

B., & Chae, H.-R., 11-29, Cornell University, Cornell Linguistics Club, Ithaca, NY.

McFadden, Thomas. 2004. The position of morphological case in the derivation. Doctoral

dissertation, University of Pennsylvania.

McFadden, Thomas and Sandhya Sunderasen. 2010. Nominative case is independent of

finiteness and agreement. BCGL5, Case at the Interfaces, Brussels.

Mohammad, Mohammad.2000. Word Order, Agreement, and Pronominalization in Standard

and Palestinian Arabic. Amsterdam and Philadelphia: John Benjamins.

296
Mohammad, Mohammad. 1990. The problem of subject-verb agreement in Arabic: Towards a

solution. In Perspectives on Arabic linguistics I, ed. by Mushira Eid, 95-125. Amsterdam:

John Benjamins.

Moutaoukil, Ahmed. 1987. Min l-binyati l-ḥamiliyati ʔilaa l-binyati l-mukawwiniyati: ʔal-

waḏiifatu l-mafʕuul-u fii l-luġat-i l-ʕarabiyati [From the predicational structure to the

constitument structure: The object’s function in Arabic]. Casablanca: Al-ṯaqaafatu for

publishing and distribution.

Moutaoukil, Ahmed. 1985. ʔal-waḏaaʔifu l-tadaawiliyatu fii l-luġat-i l-ʕarabiyati [The

pragmatic functions in Arabic]. Casablanca: Al-ṯaqaafatu for publishing and distribution.

Musṭafa, Ibraahiim. 1957. ʔiḥyaaʔu n-naḥui [Reviving grammar]. Cairo: The Committee for

authorship, translation and distribution.

Newmeyer, Frederic J. 2004. Against parameter-setting to typological variation. In Linguistic

Variation Yearbook 4, ed. by Piera Pica, 181-234. Amsterdam: John Benjamins.

Ouhalla, Jamal. 2005. Agreement and antiagreement. Natural Language and Linguistic Theory,

23, 655-686.

Ouhalla, Jamal. 1994. Verb movement and word order in Arabic. In Verb movement, ed. by

David Lightfoot and Norbert Hornstein, 41-72, Cambridge: Cambridge University Press.

Pereltsvaig, Asya. 1999. Cognate objects in Russian. Is the notion “cognate” relevant to syntax?

Canadian Journal of Linguistics 44: 267-291.

Preminger, Omer and Ted Levin. 2015. Case in Sakha: are two modalities really necessary?

Natural Language and Linguistic Theory, 33, 1, 231-250.

297
Raḥḥali, Mohammed. 2003. Tarkiibu j-jumlati al-ʕarabiyyati: Muqaarabatun naḏariyyatun

jadiidatun [The syntax of Arabic sentences: A new theoretical approach]. Casablanca:

Tubqal Publishing Company.

Raḥḥali, Mohammed and El-Hassan Souâli. 1997. A minimalist approach to verb movement in

Standard Arabic. Studia Linguistica 51, 3, 317-338.

Ryding, Karin C. 2005. A reference grammar of Modern Standard Arabic. Cambridge:

Cambridge University Press.

Richardson, Kylie. 2007. Case and aspect in Slavic. Oxford: Oxford University Press.

Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar, ed. Liliane

Haegeman, 281-337. The Netherlands: Kluwer Academic Publishers.

Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Fortis.

Runner, J.T. 1998. Noun phrase licensing. Outstanding dissertations in Linguistics. New York:

Garland.

Runner, J. T. 1995. Noun phrase licensing and interpretation. PhD dissertation, University of

Massachusettes, Amherst.

Schütze, Carston. 2001. On the nature of default case. In Syntax, 4, 3, 205-238.

Shlonsky, Ur. 2000. Remarks on the complementizer layer of Standard Arabic. In Research in Afroasiatic

grammar, ed. by J. Lecarme, J. Lowenstamm, and U. Shlonsky, 325-344. Amsterdam: John

Benjamins.

298
Soltan, Usama. 2011. On issues of Arabic syntax: An essay in syntactic argumentation. Brill’s

Annual of Afroasiatic Languages and Linguistics, 3, 236-280.

Soltan, U. 2007. On formal feature licensing in Minimalism: Aspects of Standard Arabic

morphosyntax. Doctoral dissertation, University of Maryland, College Park.

Vergnaud, Jean-Roger. 2008. Letter to Noam Chomsky and Howard Lasnik on “Filters and

Control,” April 17, 1977, ed. by Freidin R., Carlos Otero & Maria Luisa Zubizarreta, 3-

15.

Woolford, Ellen. 2006. Lexical case, inherent case, and argument structure. Linguistic Inquiry,

37,1, 111-130.

Wurmbrand, Susi and Youssef A. Haddad. 2016. Cyclic spell-out derived agreement in Arabic

raising constructions. In Perspectives on Arabic Linguistics XXVIII: Papers from the

Annual Symposium on Arabic Linguistics, Grainesville, Flordia 2014, ed. by Youseff A.

Haddad and Eric Potsdam, pp. 193-226.

Yip, Moira, Joan Maling & Ray Jackendoff. 1987. Case in tiers. Language, 63,2, 217-250.

299