From Root to Nunation: the Morphology of Arabic Nouns

From Root to Nunation: The Morphology of Arabic Nouns

Abdullah S. Alghamdi

A thesis in fulfillment of the requirements for the degree of Doctor of Philosophy

School of Humanities and Languages Faculty of Arts and Social Sciences

March 2015

PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet Surname or Family name: Alghamdi First name: Abdullah Other name/s: Abbreviation for degree as given in the University calendar: PhD School: Humanities and Languages Faculty: Arts and Social Sciences Title: From root to nunation: The morphology of Arabic nouns.

Abstract 350 words maximum: (PLEASE TYPE) This thesis explores aspects of the morphology of Arabic nouns within the theoretical framework of Distributed Morphology (as developed by Halle and Marantz, 1993; 1994, and many others). The theory distributes the morphosyntactic, phonological and semantic properties of words among several components of grammar. This study examines the roots and the grammatical features of gender, number, case and definiteness that constitute the structure of Arabic nouns. It shows how these constituents are represented across different types of nouns. This study supports the view that roots are category-less, and merge with the category-assigning feature [n], forming nominal stems. It also shows that compositional semantic features, e.g., ‘humanness’, are not a property of the roots, but are rather inherent to [n]. This study supports the hypothesis that roots are individuated by indices and the proposal that these indices are conceptual in nature. It is shown that indices may activate special language-specific rules by which certain types of Arabic nouns are formed. Furthermore, this study argues that the masculine feature [-F] is prohibited from remaining part of the structure of Arabic nonhuman plurals. Thus, crucially a feature-changing operation that changes it into [+F] is introduced and defended. In the light of several modifications that the structure of Arabic nouns is claimed to undergo, this study argues that the component where these modifications take place is divided into phases and each phase is restricted to a certain type of modification. The study also reviews and counters some previous accounts for the morphology of Arabic nouns. For example, it shows that the number feature [Pl] of certain plural nouns, descriptively known as broken plurals, is not pre-syntactically configured. It also shows that the behaviour of certain structures, known as diptotes, is phonologically rather than morphosyntactically motivated. Finally, contrary to studies that do not regard the suffix -n, known as nunation, as an indefinite marker, this study provides evidence which shows that nunation is indeed an indefinite marker in nouns, but a marker of specificity in proper names.

Declaration relating to disposition of project thesis/dissertation I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350-word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only). Signature:...... Witness:...... Date:...... The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

ORIGINALITY STATEMENT

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed ......

Date ......

iii

Publications and Conference Presentations Arising from the Writing of this

Thesis:

Alghamdi, A. (2012). The Structure of Arabic Dual and Sound Masculine Plural Nouns: Evidence for Impoverishment and Fusion. Paper presented at the Annual Conference of the School of International Studies, University of New South Wales, Australia.

Alghamdi, A. (2013). The Importance of Root Indices: Evidence from Arabic Nouns. Paper presented at the International Linguistics Beyond and Within (LingBaW) Conference, John Paul II Catholic University of Lublin, Poland.

Alghamdi, A. (2015). The Importance of Root Indices: Evidence from Arabic Nouns. In A. Bondaruk and A. Prażmowska (Eds.), Within Language, Beyond Theories: Studies in Theoretical Linguistics (Vol. 1). Newcastle: Cambridge Scholars Publishing.

iv Dedication

To all those who pray and wish the best for me

v Acknowledgments

Acknowledgments

First and foremost, all praises and thanks to God for giving me everything to accomplish this thesis successfully: patience, health, wisdom, and blessings. I would like to express the most sincere gratitude to my joint supervisors, Dr. Hugues Peters and Dr. Mengistu Amberber, for guiding me and giving me their precious time. They have been a great source of help, support and inspiration, and were always by my side during the ups and downs I encountered during my candidature. I cannot thank them enough. I am immensely grateful to my friend and fellow linguist Dr. Abdelhak El Hankari for reading the early drafts of this thesis and providing me with his valuable comments. I will not forget that day when we spent over seven hours discussing – and at times arguing about – my analysis. I would like to apologise for any inconvenience I may have caused him on that day. Lest I forget those who have contributed to my knowledge since I was a child, I would like to express my appreciation to everyone who taught me during all levels of my education. Special thanks to Dr. Abdul Ghaffar Khan, Dr. Ahmad Al-Izabi, Dr. Yousef Shaheen and Dr. Osama Mufti from Jeddah Teachers’ College, Saudi Arabia, and also equally to Dr. Rob Pensalfini, Dr. Mary Laughren, Dr. John Ingram and Dr. Michael Harrington from the University of Queensland, Australia. I would like to express my deep gratitude to my employer Albaha University, Saudi Arabia, and their representative the Saudi Cultural Mission in Canberra for funding me to pursue my dream of continuing my higher education. Without their support, this dream would not have come true. Who are we without friends? I am grateful to my friends for being there when I needed them. I would particularly like to acknowledge Tariq Alhawassi and Atheer Jasim for being family to me in Australia, and Fahad Alnotaifi, Ghazi Alqethami, Naief Albaraq, Saleh Alshehri, Saleh Alzahrani, Saud Alboqami and Abdullah Alqarni for the fruitful discussions we have had. Special thanks to Brett Todd for making the time we have spent together at the study lab fun and for editing my thesis. The greatest THANK YOU goes to my family for everything I achieved. I apologise to my parents for not being there when they needed me – I promise not to do it again. I appreciate the efforts that my brothers and sisters have exerted to fill my position and take care of our parents. I have the utmost admiration for my wife, Rana, for enduring the many hours alone and bearing extra burdens while I laboured at this endeavour. I thank you from the bottom of my heart.

vi Table of Contents

Table of Contents

1. List of Tables ...... x 2. List of Figures ...... xi 3. List of Abbreviations ...... xii 4. Arabic Phonemic Inventory ...... xiii 1. Chapter One – Introduction ...... 1 1.1. Thesis Overview ...... 1 1.1.1. Statement of the Problem ...... 1 1.1.2. Objectives and Questions of the Study ...... 1 1.1.3. Significance of the Study ...... 2 1.2. Source of Data ...... 2 1.3. Navigating Through the Chapters ...... 4 2. Chapter Two - Arabic and Its Nouns ...... 9 2.1. Introduction ...... 9 2.2. The Historical Development of Arabic ...... 10 2.3. Roots and the Formation of Stems ...... 13 2.4. The Grammatical Features of Arabic Nouns ...... 17 2.4.1. Singular Nouns ...... 19 2.4.2. Dual Nouns ...... 21 2.4.3. SndM Plural Nouns ...... 22 2.4.4. SndF Plural Nouns ...... 23 2.4.5. Broken Plural Nouns ...... 25 2.5. Is -at the only Feminine Suffix? ...... 28 2.6. Concluding Remarks ...... 32 3. Chapter Three - Theoretical Framework ...... 34 3.1. Introduction ...... 34 3.2. Overview ...... 35 3.3. The Narrow Lexicon ...... 38 3.4. The Morphology ...... 40 3.4.1. Impoverishment ...... 42 3.4.2. Fusion ...... 44 3.5. The Vocabulary ...... 47 3.6. The Encyclopaedia ...... 51 3.7. Concluding Remarks ...... 53 4. Chapter Four – Roots and Nominal Stems ...... 55 4.1. Introduction ...... 55 4.2. Roots and Compositional Semantic Features ...... 57 4.3. Root Individuations ...... 62 4.3.1. Conceptual Individuation ...... 63 4.3.2. Phonological Individuation ...... 66 4.3.3. Abstract Individuation ...... 68 4.4. In Favour of Conceptual Individuation ...... 71 4.4.1. DM and Phonological Individuation ...... 72

vii Table of Contents

4.4.2. Conceptually Empty Roots ...... 73 4.4.3. Roots with Multiple Concepts ...... 75 4.5. Cross-Dialectal and Cross-Linguistic Variations ...... 79 4.6. Root Categorisation ...... 81 4.7. Conclusion ...... 89 5. Chapter Five – Gender and Number ...... 91 5.1. Introduction ...... 91 5.2. Singular Nouns ...... 94 5.2.1. Gender in Masculine and Overt Feminine Nouns ...... 94 5.2.2. Gender in Covert Feminine Nouns ...... 98 5.2.3. Number in Singular Nouns ...... 101 5.3. Dual Nouns ...... 103 5.4. Nonhuman Plurals ...... 106 5.4.1. The Development of Deflected Agreement ...... 109 5.4.2. Gender in Nonhuman Plurals ...... 112 5.5. Sound Plurals ...... 119 5.5.1. SndM Plurals ...... 120 5.5.2. SndF Plurals ...... 122 5.6. Broken Plurals ...... 124 5.6.1. The Status of [Pl] in Broken Plurals ...... 125 5.6.2. The Formation of Broken Plurals in the Morphology ...... 129 5.7. Default Plural Formation ...... 134 5.8. Insight into the Morphology ...... 138 5.9. Conclusion ...... 143 6. Chapter Six – Case and Definiteness ...... 146 6.1. Introduction ...... 146 6.2. Case in SA Nouns ...... 149 6.2.1. Singular and Broken Plural Nouns ...... 150 6.2.2. SndF Plural Nouns ...... 152 6.2.3. Dual and SndM Plural Nouns ...... 154 6.3. (In)Definiteness in SA Nouns ...... 159 6.3.1. The Standard (In)Definiteness-Marking System in SA ...... 160 6.3.2. Construct State ...... 164 6.4. The Internal Structure of N ...... 168 6.5. SA Diptotes ...... 172 6.6. SA Defective, Indeclinable and Invariable Nouns ...... 176 6.6.1. Defective Nouns ...... 177 6.6.2. Indeclinable Nouns ...... 181 6.6.3. Invariable Nouns ...... 183 6.7. Case and Definiteness in DA Nouns ...... 185 6.8. Conclusion ...... 187 7. Chapter Seven – Nunation ...... 190 7.1. Introduction ...... 190 7.2. A Different View of Nunation ...... 193 7.2.1. Why Nunation Cannot Mark Indefiniteness ...... 193 7.2.2. The Alternative Functions of Nunation ...... 198 7.3. -n(noun) and /nV/ in Vocative Contexts ...... 201

viii Table of Contents

7.4. Nunation on Names ...... 208 7.5. Nunation in the Bedouin Dialects ...... 215 7.6. Conclusion ...... 218 8. Chapter Eight - Conclusion ...... 221 8.1. The Formation of Arabic Nouns ...... 221 8.2. Nunation ...... 226 8.3. The Theory ...... 227 8.4. Suggestions for Future Studies ...... 230 9. References ...... 232 10. Appendices ...... 247 Appendix 1: Ethics Approval Letter ...... 247 Appendix 2: Samples of the Collected Data ...... 248

ix List of Tables

1. List of Tables

Table 2.1: Stems Derived from the Tri-Consonantal Root /ktb/ ...... 14

Table 2.2: The Grammatical Features of Arabic Nouns ...... 17

Table 2.3: Singulars vs. Broken Plurals ...... 25

Table 3.1: Strong and Weak Norwegian Adjectival Suffixes ...... 43

Table 4.1: Prefix-Stem English Verbs with Common Phonological Units . . 66

Table 4.2: Prefix-Stem English Nouns with Common Phonological Units . . 67

Table 4.3: Nominal Binyanim and Root Categorisation in Arabic ...... 84

Table 4.4: Root Gemination in Nominal Stems ...... 87

Table 5.1: Paradigms of Demonstratives and Subject Pronouns ...... 112

Table 5.2: Nonhuman Masculine Singulars Pluralised as SndF ...... 113

Table 5.3: Gender and Number VIs in SA and DA Nouns ...... 145

Table 6.1. SA Diptotes ...... 172

Table 6.2: SA Defective Nouns ...... 177

Table 6.3: SA Indeclinable Nouns ...... 182

Table 6.4: SA Invariable Nouns ...... 184

Table 6.5: DA Subject, Object and Possessive Pronouns ...... 186

x List of Figures

2. List of Figures

Figure 2.1: Interactions of Gender, Number, and Case in SA Nouns ...... 27

Figure 2.2: Interactions of Gender and Number in DA Nouns ...... 27

ʕ Figure 2.3: Arabic Scripts of s aħraːʔ (Desert) and ðikraː (Remembrance)... 29

Figure 3.1: The Architecture of the Grammar of DM ...... 38

Figure 4.1: Arabic Stems Derived from /ktb/, /lʕb/, and /ʕlm/ ...... 75

Figure 4.2: The Recurrence of Jar in the Story ...... 79

Figure 5.1: Deflected and Strict Agreements across the Stages of Arabic . . . 109

Figure 5.2: Gender and Number Rules across the Morphological Phases . . . 142

xi List of Abbreviations

3. List of Abbreviations

1st First Person LF Logical Form

2nd Second Person M Masculine

3rd Third Person n Nominal Category Feature

Acc Accusative N Noun

Adj Adjective Nom Nominative

AGR Agreement Num Number

BFR Broken Formation Rule O Object

C Consonant Obl Oblique

CA Classical Arabic P Phrase

CS Construct State PF Phonological Form

CSF Compositional Semantic Feature Pl Plural

D Determiner Poss Possessive

DA Dialectal Arabic Pres Present

Def Definite S Subject

Dl Dual SA Standard Arabic

DM Distributed Morphology Sg Singular

F Feminine Snd Sound

Gen Genitive SPEC Specifier

Gend Gender UG Universal Grammar

Hum Human v Verbal Category Feature

Idef Indefinite V Verb

K Case VI Vocabulary Item

xii Arabic Phonemic Inventory

4. Arabic Phonemic Inventory

Consonants

alveolar - Velar Dental Palatal Uvular Glottal Bilabial Alveolar Pharyngeal Labiodental Post

Plosive b t d k g q ʔ

Nasal m n

Trill r

Fricative f Ө ð s z ∫ ʒ x χ ħ ʕ h

Glide w y

Lateral l

Pharyngealised Consonants: tʕ, dʕ, ðʕ and sʕ. ʒ = the IPA symbol dʒ. y = the IPA symbol j.

Vowels

Front Central Back

High i iː u uː

Low a aː

xiii Chapter One - Introduction

1. Chapter One – Introduction

1.1. Thesis Overview

1.1.1. Statement of the Problem

This thesis aims to explore Arabic nouns within the framework of the theory of

Distributed Morphology (DM), as introduced by Halle and Marantz (1993, 1994) and

pursued by many others. In DM, roots are category-less constituents that need to merge

with category-assigning elements such as the nominal morpheme [n] or the verbal

morpheme [v], as will be illustrated in Chapter Three (e.g., Marantz, 1996, 2001; Harley

and Noyer, 1998, 1999; Arad, 2003, 2005; Harley, 2005; Embick and Noyer, 2007;

Embick and Marantz, 2008; Lowenstamm, 2008). As far as Arabic nouns are

concerned, this thesis investigates this notion and its implications with regard to the

features of gender, number, case and definiteness with reference to the theoretical views

of DM.

1.1.2. Objectives and Questions of the Study

Using data representing the nouns of Standard Arabic (SA) and Dialectal Arabic (DA),

the thesis examines the basic structure of Arabic nouns and the various ways it can be

modified yielding different types of nouns. This involves examining how the

constituents of this structure are handled during the word formation process that is

proposed by DM. The investigation essentially leads to figure out solutions to the

following questions:

(1) What are the basic Arabic roots?

(2) How does the attachment of the nominal morpheme [n] create certain implications

with regard to gender, number, case and definiteness?

1 Chapter One - Introduction

(3) Are there any structural differences in formation between SA and DA nouns?

(4) What is nunation in SA nouns and names?

(5) How does the theory of DM accounts for the structure of Arabic nouns in general

and the implications found in question (2) in specific?

1.1.3. Significance of the Study

As is well known, Arabic is a morphologically rich language. Besides the distinctive

stem formation strategy it has by virtue of its membership of the Semitic family, its

nouns encapsulate a variety of grammatical features in several configurations. In this

regard, it will be shown that SA and DA nouns share almost the same formation

strategy, but they usually differ in the way their stems and grammatical features are

phonologically expressed. One of the main differences, for example, is that case and

indefiniteness are not overtly marked in DA nouns.

It is important to point out that although some aspects of Arabic nouns have been

previously addressed in some studies within the framework of DM, none of these

studies claims to provide a comprehensive account for the formation of Arabic nouns.

To the best of my knowledge, there is no study within DM that claims to provide such

an account. Moreover, in some of the previous studies where certain aspects of Arabic

nouns have been addressed, the lack of a holistic view of the different types of Arabic

nouns has led to inaccurate conclusions. Hence, in addition to delivering a

comprehensive investigation of the structure of Arabic nouns within the theory of DM,

this thesis endeavours to review those studies and tackle their deficiencies.

1.2. Source of Data

This study is based mainly on data from two varieties of Arabic: Standard Arabic (SA)

and Dialectal Arabic (DA). DA is represented by three dialects: Egyptian, Jordanian,

2 Chapter One - Introduction

and Saudi. Although some sub-dialects do indeed exist, e.g., Cairene and Sa’idi in

Egypt, Rural and Urban in Jordan, and Najdi and Hijazi in Saudi, the data collected for

this thesis is considered a sample that, to a large degree, represents the general

characteristics of the structure of DA nouns. In fact, as will be shown in this thesis, the

dialects representing DA have a mutual noun formation mechanism and the same

phonological representations for their grammatical features, but they often employ

different phonological representations for the constituents of their stems.

Besides my intuition as a native speaker of Arabic, the data used in this study was

collected from various sources. As to SA, the data was sourced from several descriptive

and analytic studies in the literature (Haywood and Nahmad, 1965; Al-Afghani, 1971;

Abboud and McCarus, 1983; Belnap and Shabaneh, 1992; Fassi Fehri, 1993, to name a

few) with the main contribution coming from Ryding (2005). Other sources of SA data

were ArabiCorpus1 as well as TV and radio shows and interviews in which SA is the

main language (e.g., shows related to religion and literature). Finally, some verses of the

Quran are also used for the purpose of comparing some aspects of SA nouns to their

counterparts in Koranic Arabic.

In terms of DA, the data was collected by conducting recorded interviews with

some native speakers of the varieties under investigation. The interviews involved

general conversations and narrating a wordless picture-based story named Frog, Where

Are You? (Mayer, 1969) (see Appendix 2 for samples). Having the same pictures

described in different dialects provided this study with a wide range of data in which the

similarities and differences between the dialects are highlighted. In order to achieve the

same clarity about the similarities and the differences between SA and DA, a university

1 ArabiCorpus is a corpus developed and maintained by Prof. Dilworth Parkinson, Brigham Young University. The corpus contains 173,600,000 words collected from different types of sources, e.g., newspapers, Islamic discourse and modern literature. Users can access and search the corpus at http://arabicorpus.byu.edu. 3 Chapter One - Introduction

academic specialising in Arabic volunteered to provide a SA narration for the story. A

supporting source of DA data was TV and radio shows from the local channels where

these dialects are spoken. These channels were accessible from Australia through the

internet.

1.3. Navigating Through the Chapters

Chapter Two provides a description of Arabic in general, and the morphological system

of its nouns in particular. It provides a brief overview of the history, importance, and

developmental stages of the language. Most of the chapter describes the morphological

system of Arabic nouns by showing the characteristics of each type of nouns and how

DA nouns differ from their SA counterparts. Psycholinguistic and neurolinguistic

studies supporting the notion that Arabic roots are consonantal are briefly reviewed. The

chapter concludes by addressing a traditional claim that the suffix -at is not the only

feminine suffix in Arabic nouns. It will be shown that this claim is based on analysing

the written form of the nouns, which resulted in mistakenly considering /ʔaː/ and /aː/ as

suffixes rather than parts of the stems.

Chapter Three is concerned with the theoretical framework of DM. It provides an

overview of how the theory has developed and how it contrasts with some other

approaches. It demonstrates the overall architecture of the grammar that DM proposes

and the mechanism by which the word formation process is carried out. The chapter

focuses on describing how the morphosyntactic, phonological and semantic properties

of linguistic structures are regarded in DM as well as how these properties are

distributed among the proposed grammar. In short, DM employs a syntactic approach

towards the formation of words in which a single generative engine in conjunction with

different components produces syntactically built, morphologically modified,

phonologically expressed, and semantically interpreted structures. The most relevant 4 Chapter One - Introduction

components for the purposes of this thesis are the Morphology and the Vocabulary,

where structures are morphologically modified and phonologically expressed

respectively. Thus, most of the chapter is devoted to providing a detailed description of

these components and their properties.

The descriptive background of Arabic and the theoretical framework have been

introduced in Chapter Two and Chapter Three respectively. Chapter Four signals the

beginning of the analysis where Arabic nouns and the theory come together constituting

the core of the original contribution of the thesis. It is hoped that such a separation

between the literature of Arabic nouns and the literature of the theory would effectively

familiarize readers with the essential elements of the thesis before being analysed.

Chapter Four seeks to explore the nature of Arabic roots, the formation of

nominal stems, and the related phonological and semantic matters within DM. It

provides evidence that, in contrast with what Marantz (1995, 1996) implies,

compositional semantic features cannot be a property of the roots. The identity of roots

is also discussed and support given to the view that roots obtain their identity from

indices. However, contrary to arguments that root indices are phonological (Borer,

2009, 2013) or abstract (Harley, 2011), Pfau’s (2000) proposal that they are of a

conceptual nature is supported. In the subsequent chapters, it is suggested that, in

addition to the Vocabulary, root indices are probably able to communicate with the

Morphology as well. The chapter also supports the view that roots are not inherently

specified for a part of speech (e.g., Marantz, 1996, 1997, 2001; Harley and Noyer, 1998,

1999; Arad, 2003, 2005). Therefore, roots of Arabic nouns merge with the category-

assigning morpheme [n], which is also a functional set (Kihm, 2005) and the location of

the compositional semantic features, forming nominal stems.

5 Chapter One - Introduction

Chapter Five investigates how Arabic nouns express their gender and number

features. It shows that although gender and number are independent features, they are

quite dynamic in that they can interact with each other and with other features, yielding

different types of nouns. The chapter also addresses a phenomenon in which Arabic

nonhuman plurals agree as singular feminine. It will be argued that the singular aspect

of the agreement is probably motivated by a gap in the agreement system, while the

feminine aspect is motivated by the plurals themselves. Specifically, it will be argued

that if the plurals are not structured as feminine, they undergo a gender-changing

process by which they become feminine. Hence, a feature-changing operation will be

proposed. In addition, the chapter discusses the sound and broken pluralisation

strategies in Arabic, and whether the number feature [Pl] is pre-syntactically inherent to

[n] in broken plurals as argued by Noyer (1992). Finally, in the light of the several

modifications that the structure of Arabic nouns undergoes in the Morphology with

respect to gender and number, Chapter Five examines the internal structure of the

Morphology. It will be argued that the Morphology is organised into phases depending

on the nature and the motivation of the rules by which the modifications are carried out.

Chapter Six focuses on the case and definiteness features of Arabic nouns. It

provides an analysis that accounts for the standard case, definiteness and indefiniteness

marking systems of Arabic nouns as well as how some nouns deviate from them. It will

be argued that some of these deviations are morphologically motivated, while some

others are simply phonological. For a certain type of nouns known as diptotes, it will be

shown that their deviation is likely to be phonologically motivated, which is a view that

contrasts with an analysis provided by Embick and Noyer (2007). The interactions

between the case and number features in some nouns are also analysed, and the

outcomes will not only explain how these nouns are formed, but also support the

6 Chapter One - Introduction

argument made in Chapter Five that the Morphology is internally organised into phases.

The case and definiteness features of nouns in genitive constructions are touched upon

in the light of previous studies by Ritter (1988, 1991) that analyse similar constructions

in Hebrew. The analysis of these constructions lends support to the view that nouns are

structured as DPs that involve an N-to-D raising (Abney, 1987). In this regard, the

underlying structure of Arabic nouns will be considered as consisting of two levels: N

as the locus of roots, [n], gender, number and case – whose configuration inside N will

be demonstrated – and D as the landing site for N where the different values of the

definiteness feature can be assigned.

Part of the analysis provided in Chapter Six is that indefiniteness in SA noun is

phonologically expressed by the suffix -n, known as nunation. Chapter Seven is mainly

a response to a claim made by Fassi Fehri (1993) and Lyons (1999) that nunation

cannot be an indefinite marker. The chapter is basically a critique of the bases of this

claim and the alternative functions that the authors suggest for nunation. The authors’

arguments are reviewed and a detailed refutation is provided. The chapter addresses the

link that the authors establish between the nunation of nouns and the /ni/ and /na/ parts

of certain suffixes on the one hand, as well as the link that Fassi Fehri (1993)

establishes between the nunation of nouns and the nunation of proper names on the

other hand. It will be concluded that the /ni/ and /na/ parts cannot be nunation, while the

nunation of names cannot be functionally the same as the nunation of nouns. The

nunation of names will be considered a marker of specificity, while the nunation of

nouns will continue to be considered a marker of indefiniteness. Finally, examples from

nouns of Bedouin Arabic where nunation can appear under certain circumstances are

provided. The purpose is to show that the nunation that can appear in this variety is

7 Chapter One - Introduction

more likely to be a vestige of the nunation of names rather than a vestige of the

indefinite marker as argued by Holes (1990).

Chapter Eight concludes the thesis and provides a broad picture of the

contributions this thesis attempts to make to the understanding of both the morphology

of Arabic nouns and the theory of DM. Some suggestions for future studies are also

provided.

8 Chapter Two – Arabic and Its Nouns

2. Chapter Two - Arabic and Its Nouns

2.1. Introduction

This chapter aims to provide a brief overview of the historical development of Arabic,

as well as the relevant aspects of the nouns of the varieties under investigation. These

aspects include the nature of Arabic roots, the formation of Arabic stems, and how the

grammatical features of gender, number, case and definiteness are expressed. It is meant

to be a general description of the major characteristics of Arabic nouns, while

descriptions of some relevant minor issues will be provided in later chapters. The

description offered in this chapter is expected to provide an insight into the

morphological system of the nouns that will be useful when the underlying structure of

these nouns is explored within the framework of DM in the subsequent chapters.

This chapter is organised as follows. Section 2.2 presents the distribution,

importance, historical background, and current status of Arabic. It is a brief outline of

the major steps leading up to the emergence of the contemporary varieties of Arabic,

and how they are currently regarded and used.

Section 2.3 provides an overview of the nature of Arabic roots and the formation

of its stems, which appear to be in line with the distinctive nature of the morphology of

Semitic languages in general. As is well known, Arabic roots are made up of

consonantal elements that cannot emerge independently before combining with patterns

that mainly consist of vowels. This is the mechanism by which stems are formed and

equipped with the basic semantic and functional properties. External evidence

supporting this view of the nature of roots will be provided by briefly going through

some psycholinguistic and neurolinguistic experiments that have focused on this issue.

9 Chapter Two – Arabic and Its Nouns

Relying on the descriptive literature of Arabic and data collected for this study,

Section 2.4 will show how Arabic nouns express their grammatical features of gender,

number, case and definiteness. The vast majority of the section will be a description of

how these features are expressed in each type of noun in SA and DA. The main

morphological differences between the nouns of SA and DA will be highlighted, along

with the several ways the grammatical features interact with each other in each variety.

These issues will constitute most of the areas that will be investigated and analysed

within the framework of DM in the subsequent chapters.

One of the features that will be described in Section 2.4 is gender, which is

usually overtly marked by the suffix -at only when feminine. Section 2.5 addresses a

traditional claim that -at is not the only feminine suffix. This claim will be reviewed and

a refutation will be provided. It will be shown that it is probably the traditional practice

of analysing the written forms of the language that has led – or rather misled – to

considering -at a member of a set of three feminine suffixes in Arabic. The section will

conclude with dismissing this claim and considering feminine nouns to be either overtly

marked with -at, or simply covert feminine nouns. This will pave the way for a further

investigation that will be presented in Chapter Five.

Section 2.6 will conclude the chapter and remark on the points that will be

discussed throughout the remainder of the thesis.

2.2. The Historical Development of Arabic

Arabic is the official language in 22 countries constituting what is known as the Arab

World, or the League of Arab States.2 In addition, due to the fact that it is the language

of the Quran, which is the key text and the Holy Book of Islam, Arabic has become the

2 A statistical report issued by the League of Arab States indicates that the population of Arab countries was over 350 million people in 2012. 10 Chapter Two – Arabic and Its Nouns

liturgical language of religious practice and study for all the world’s Muslims (Kaye,

1987; Versteegh, 1997; Ryding, 2005).

In terms of its origin, Arabic is a member of the Semitic family of languages. The

Semitic family, in turn, belong to a wider group of languages called Afroasiatic

(formerly Hamito-Semitic), which also includes the Berber, Cushitic, Chadic, Egyptian,

and Omotic3 families (Hetzron, 1987; Hayward, 2000; Childs, 2003; Crass, 2006). The

extant languages of the Semitic family include: (1) Ethiopian Semitic languages such as

Amharic, Tigrinya, Harari, and Gurage; (2) Hebrew; (3) South Arabic languages like

Mehri and Soqotri; (4) Aramaic, which is spoken by groups of Christians in Iran, Iraq

and Syria; and (5) Arabic (Hetzron and Kaye, 1987; Bateson, 2003; Ryding, 2005).

By tracing its historical development, it can be found that Arabic went through

different stages before the emergence of what is currently used and known as Standard

Arabic (SA),4 and the different spoken regional varieties of the language (Dialectal

Arabic = DA). Ryding (2005) lists the stages of Old and Early Arabic that roughly

lasted from the seventh century BC to the sixth century AD. Versteegh (1997) argues

that the inscriptions of what is known as Old Arabic are likely to belong to the South

Arabic languages since they are largely related to the present forms of Soqotri and

Mehri. In terms of the inscriptions of Early Arabic, Versteegh (1997) states that they are

closely related to its following stage, which is known as Classical Arabic (CA).

At the beginning of the CA stage, specifically in the pre-Islamic period, Arabic

was looked upon as the language of literacy and the prosperous era of Arabic literature,

3 Greenberg (1966) argues that Omotic languages form a sub-division of Cushitic rather than a direct branch of Afroasiatic. Diakonoff (1998), on the other hand, suggests that considering the grammatical isoglosses between Semitic and Omotic as well as the classification of the neighbouring languages, Omotic languages may not belong to the Afroasiatic group at all. 4 Which is also termed Modern Standard Arabic, as found in Kaye (1970); Ferguson (1989); Ryding (2005); Alhawary (2011), among many others, probably because of its relation to the modern styles of writing and literature. 11 Chapter Two – Arabic and Its Nouns

mainly the poetic forms. Then, the revelation of the Quran5 in the seventh century made

CA the language of Islam and the Islamic empire, which expanded from the Arabian

Peninsula to the Iberian Peninsula in the west, and to Central and South Asia in the east.

After the empire had weakened, CA remained the main language used in the Arabic

Peninsula, the Levant and North Africa. In the thirteenth century, Arabic entered a

transitional stage called Middle Arabic during which CA preserved its position as the

language of Islam and literature, and the spoken regional varieties started to emerge

(Versteegh, 1997; Ryding, 2005).

Scholars became aware of SA, possibly as the substitute of CA, approximately at

the end of the eighteenth century. Its emergence coincides with the start of a modern era

of Arabic literature that is different from the literature of CA due to the exposure to

modern knowledge and writing styles. For example, Western innovations contributed to

the addition of new vocabulary, while the modern journalistic and literary styles of

writing had a significant impact on the creation of new literary forms like novels,

dramas, and autobiographies (Versteegh, 1997; Bateson, 2003; Holes, 2004; Ryding,

2005).

Nowadays, despite the fact that Arabs speak different regional varieties in their

everyday communication, SA is still the language used in formal contexts such as

official meetings, media, religious ceremonies, written correspondence, and even in

reciting a type of ode known as fusʕħa (literally, eloquent) on casual and formal

occasions. This phenomenon of having two varieties of Arabic used side by side within

communities has been taken as a clear example of what is referred to as diglossia,6 as

first introduced into the field of sociolinguistics by Ferguson (1959). In this view, SA is

the High variety of the language. It is learned through formal education and used, in

5 The language used in the Quran is usually called Koranic Arabic. 6 For more on diglossia in Arabic, see, e.g., Ryding (1991); Kaye (2001); Saiegh-Haddad (2003). 12 Chapter Two – Arabic and Its Nouns

spoken and written forms, as the formal and prestigious variety of the language. DA, on

the other hand, represents the Low varieties of Arabic. They are spoken as the colloquial

forms of the language, and acquired as the first language without any kind of formal

learning. The varieties of DA are usually referred to by terms that indicate a lower

position such as ʕaːmmiyah (the common language) or suwqiyah (the language of the

market) (Kaye, 1987; Versteegh, 1997; Ryding, 2005).

2.3. Roots and the Formation of Stems

What might be a characteristic property of the morphology of Arabic, and Semitic

languages in general, is the way stems are formed. Arabic stems are usually formed by

combining consonantal roots – mostly tri-consonantal7 – and patterns of (short and

long) vowels that may sometimes include consonants (Ryding, 2005:47-48). These

patterns are also known as binyanim (singular = binyan), which is the term adopted by

this study.8 Roots and binyanim are combined by interspersing the consonants of the

roots with the constituents of the binyanim. To illustrate, the following table contains

examples of Arabic stems that are derived from the root /ktb/, which can roughly

convey the meaning of writing:

7 As is also in Hebrew (Harris, 1941:160; Aronoff, 2007:820). According to Moussa (1996, as cited in Boudelaa and Gaskell, 2000:50), Arabic has 11,978 roots of which 7,597 are trilateral, 4,081 are quadrilaterals and 300 are quinquiliterals. 8 Binyanim and binyan are Hebrew terms that have been widely used in the literature of Arabic and Hebrew (for example, McCarthy, 1981; Yip, 1988; Noyer, 1992; Aronoff, 1994; Ravid and Schiff, 2006; Harley, 2011; Fassi Fehri, 2012). Lists of the most common Arabic binyanim, especially in the domain of verbs, can be found in Ziadeh and Winder (1955), Wright (1967), Yaquub (1993), and Ryding (2005). In Israeli Hebrew, it has been found that the majority of the words are made up of about 1800 roots combining with about 50 binyanim (Choueka, 1996; as cited in Levin, Ravid, and Rapaport, 2001:744). 13 Chapter Two – Arabic and Its Nouns

Table 2.1: Stems Derived from the Tri-Consonantal Root /ktb/ -kitaːb- -kutub- ‘Book’ (N) ‘Books’ (N) -kaːtib- -kuttaːb-9 ‘Writer’ (N) ‘Writers’ (N) -maktab- -makaːtib- ‘Office/Desk’ (N) ‘Offices/Desks’ (N) -aktub- katab- ‘Write’ (V) ‘Wrote’ (V) -uktab- kutib- ‘Is written’ (V) ‘Was written’ (V)

For a stem such as kitaːb (book) above, the root consists of the consonants /ktb/

10 while the binyan is /C1iC2aːC3/. This binyan, like most binyanim, can interlock with

several roots to produce different stems. For example, from the binyan /C1iC2aːC3/,

different stems such as ħisaːb (calculation, root /ħsb/), qitʕaːʕ (sector, root /qtʕʕ/) and

ʕiqaːb (punishment, root /ʕqb/) can be formed.11 Although the consonantal roots are

usually able to convey a primitive meaning, they cannot be used independently. As can

be seen in the above examples, roots have to be mapped onto binyanim in order for

stems with more specific meanings and denotations to be created. As might have been

noticed, roots carry the primitive semantic meaning of the stems, whereas in general

binyanim make that meaning more specific while carrying functional properties by

which stems obtain some of their grammatical features, especially their categories (i.e.,

part of speech).

It is a well-established and largely accepted conception that roots in Arabic – and

Semitic languages in general – are consonantal. For Arabic in particular, this notion can

be traced back to Ibn Jinniy and Ibn Faris (traditional grammarians of Arabic in the 9th

century, as cited in Prunet, 2006:42). For Semitic languages in general, the notion has

been discussed in the works of de Alcala (1505); de Volney (1787); de Sacy (1810);

9 This stem involves doubling the middle consonant of the root in what is known as gemination, as will be explained in Chapter Four. 10 Cx represents the slots where root consonants can reside. 11 This type of morphology is labelled with different terms by different authors, e.g., non-concatenative (McCarthy, 1981), transfixing or root-and-pattern (Bauer, 2003), and intro-flectional (Watson, 2006). 14 Chapter Two – Arabic and Its Nouns

Ewald (1827) (to name a few of the works cited in Ussishkin, 2006:37). The notion of

consonantal roots has become overwhelmingly adopted in the more recent literature of

Semitic languages (see, e.g., Harris, 1941; Yushmanov, 1961; Haywood and Nahmad,

1965; McCarthy, 1979, 1981; Versteegh, 1997; Watson, 2002, 2006; Bauer, 2003;

Holes, 2004; Ryding, 2005; Alhawary, 2011).12

Evidence supporting this notion from the field of psycholinguistics has been

obtained via segment-shifting tasks (Feldman, Frost, and Pnini, 1995), a variety of

priming experiments (Frost, Forster, and Deutsch, 1997; Boudelaa and Marslen-Wilson,

2004a, 2004b; Velan, Frost, Deutsch, and Plaut, 2005), language games (Bagemihl,

1988, 1989), and investigating speech errors (Abd-El-Jawad and Abu-Salim, 1987;

Berg and Abd-El-Jawad, 1996). Probably the most striking and interesting evidence was

found in a case study of a bilingual Arabic-French aphasic patient conducted by Prunet,

Béland, and Idrissi (2000). In the study, the patient was exposed to 1455 stimuli in

Arabic and 1498 stimuli in French while performing different tasks of reading aloud,

repetition, writing to dictation, and picture naming. The patient produced consonantal

metathesis errors in both languages. However, the number of those errors in Arabic was

25 times more than that in French. More interestingly, all the metatheses that occurred

in Arabic targeted the consonants of the roots, while the consonants of the affixes and

the order of the vowels constituting the binyanim were never affected. Prunet et al.

(2000:619) conclude that, on the one hand, metathesising only the consonants of roots

(and sparing the consonants of the affixes and the vowels of the binyanim) in Arabic

indicates the solidarity of those root consonants. They are apparently distinct and the

effect of the aphasic condition appears only when they are accessed. On the other hand,

the significantly higher frequency of this type of error in Arabic than in French shows

12 For somewhat different views, see, Heath (1987, 1997); Bat-El (1994, 2003); Ratcliffe (1997); Benmamoun (2003). 15 Chapter Two – Arabic and Its Nouns

how the morphological systems of these languages are different.

From the field of neurolinguistics, evidence supporting the notion of Arabic

consonantal roots was obtained via an experiment conducted by Boudelaa,

Pulvermüller, Hauk, Shtyrov, and Marslen-Wilson (2010). The experiment involved

creating two types of what the authors called deviation points, then measuring and

comparing the time and the topographical nature of the brain electrical activities

(responses) associated with each type. The deviation points of the first type were

created by providing the participants with auditory pairs of words that are different only

in their consonantal roots, e.g., /ʕariːs/ (bridegroom) vs. /ʕariːf/ (corpora). The deviation

points of the second type were created by providing the participants with pairs that only

differ in their binyanim, e.g., /ʕariːs/ (bridegroom) vs. /ʕaruːs/ (bride). The first type

elicited fronto-central responses starting from 160 milliseconds, while the second type

elicited left hemispheric responses starting from 250 milliseconds. Given the significant

topographical and time differences between the responses elicited by each type,

Boudelaa et al. (2010:1007) conclude that consonantal roots and binyanim are two

different morphemic entities with different functional properties and different

electrophysiological profiles.

Given their nature, Arabic roots are seemingly different from their counterparts in

concatenative languages, e.g., English. For instance, at the surface level, the English

root cat has a grammatical category (noun), but, beyond any doubt, the Arabic root /ktb/

does not. It was shown that /ktb/ can be part of nouns and verbs depending on what

binyan it interlocks with. Semantically, cat is meaningful and has a denotation, but the

meaning of /ktb/ is, to a large extent, vague. Binyanim are what make the vague

meaning of the root more specific. Morphologically, cat, as it stands, is also a stem that

can inflect for plurality by having a suffix, but /ktb/ is not. From a DM perspective, the

16 Chapter Two – Arabic and Its Nouns

theoretical implications of these differences are important as will be shown in this

thesis, especially in Chapter Four.

2.4. The Grammatical Features of Arabic Nouns

The outcomes of the Arabic non-concatenative morphology are stems that fall under

different parts of speech. As far as nouns are concerned, these stems need to obtain

certain grammatical features in order to be used in a clause. These features are gender,

number, case and definiteness. Depending on the variety of the language (SA or DA),

the methods by which these features are usually expressed can generally be sketched out

as shown in the table below (Haywood and Nahmad, 1965; Holes, 1990; Ingham, 1994;

Brustad, 2000; Watson, 2002; Bateson, 2003; Ryding, 2005, to name a few):

Table 2.2: The Grammatical Features of Arabic Nouns Method of Expression Feature SA DA Masculine Ø Ø Gender Feminine Suffix Suffix Singular Ø Ø Number Dual Suffix Suffix Plural Suffix/Stem Changes13 Suffix/Stem Changes Nominative Case Accusative Suffix Ø Genitive Definite Prefix Prefix Definiteness Indefinite Suffix Ø14

There are some general issues about these features that may need to be explained.

As shown in the above table, Arabic nouns fall under two gender categories: masculine

and feminine. As in many other languages, e.g., French, (grammatical) gender in Arabic

13 In both varieties, nouns that express their plurality by stem changes cannot take gender suffixes. The new stems appear to suffice to express the number and the gender features of the nouns. This will be illustrated with examples later in this chapter. 14 Under certain circumstances, indefinite nouns of some Bedouin dialects can take a suffix that is similar to the suffix of SA indefinite nouns. This suffix does not seem to be entirely an indefinite marker, as will be illustrated in Chapter Seven. 17 Chapter Two – Arabic and Its Nouns

is a feature that is not exclusive to human/animate nouns. Nonhuman/inanimate nouns

can have a gender as well. Generally, for nouns referring to humans or living creatures,

their grammatical gender is determined by the inherent biological gender of the

referents. Otherwise, the grammatical gender of the nouns is arbitrarily determined

(Ryding, 2005:119).

As also shown in the above table, SA nouns employ suffixation to express their

case features. In most of the descriptive literature, Arabic is considered to have three

different case features: nominative, accusative and genitive, but these features are

overtly marked only in SA. Generally, SA nouns are considered nominative when they

refer to the subjects (of verbs or in verbless sentences)15 and when used in isolation

(citation forms). The nouns are accusative when they refer to the objects of verbs, and

genitive when they are complement of possessive structures (known as Construct

States) or when they are modified by prepositions16 (for an extensive list, see Ryding,

2005:168-182). Some SA nouns have three different markers for these features, while

some other nouns are overtly marked for being nominative or non-nominative, as will

be shown in this section and analysed in Chapter Six.

With respect to definiteness, Arabic nouns are usually considered to be definite

when they take the definite prefix al-, which can appear on SA and DA nouns in several

forms at times as will be shown soon in this section. The standard method by which SA

nouns express their indefiniteness is by taking the suffix -n (known as nunation),17

while it is the absence of the definite prefix by which DA nouns are made indefinite.

Interestingly, there are instances in SA where nunation does not appear when expected.

15 Arabic sentences can occur without a present tense copula, so they are considered verbless, as will be exemplified in Chapter Six. 16 In Chapter Six, nouns that are modified by prepositions will be considered oblique rather than genitive. 17 An argument that nunation cannot be an indefinite marker (Fassi Fehri, 1993; Lyons, 1999) will be addressed in Chapter Seven. 18 Chapter Two – Arabic and Its Nouns

These instances include a group of SA nouns known as diptotes18 and another group of

nouns known as invariables, as will be discussed in Chapter Six.

As for number, Arabic nouns can have three different number features: singular,

dual, and plural. The way the nouns express features such as case and indefiniteness,

does not only depend on the variety to which the nouns belong, but also on other

factors, namely: (a) what number feature the nouns have; (b) if plural, how the plurality

is expressed (suffixation or stem changes), and (c) what gender feature is involved when

the plurality is expressed by suffixation. According to these factors, the types of Arabic

nouns are: singular, dual, sound plural (plurals by suffixation), and broken plural

(plurals by internal changes to the stems). The sound plural nouns are further divided

into sound masculine (SndM) and sound feminine (SndF) depending on their gender.

The following subsections provide an overview of how these types of nouns express

their grammatical features in SA and DA.

2.4.1. Singular Nouns

Arabic nouns are not overtly marked for singularity. The stems of singular nouns are

simply the basic stems that are formed by combining roots and binyanim. Singular

nouns take several affixes to express the other grammatical features, as shown below:

(1) SA Singular Nouns: a. ʕaqada al-mudiːr-u ʔiʒtimaːʕ-a-n held.he Def-director.M-Nom meeting.M-Acc-Idef ‘The director held a meeting.’

b. ʔintaqala ʔilaː dawl-at-i-n ʔuxraː moved.he to country-F-Gen-Idef another ‘He moved to another country.’

c. musaːʕid-u al-waziːr-i assistant.M-Nom Def-minister.M-Gen ‘The minister’s assistant.’

18 In addition to dropping their nunation, diptotes take the accusative marker instead of the genitive one when indefinite. These issues will be discussed in Chapter Six. 19 Chapter Two – Arabic and Its Nouns

(2) DA Singular Nouns: al-walad bi-yutʕul fiː ħufr-ah Def-boy.M is-looking in hole-F ‘The boy is looking into a hole.’

In terms of gender, SA and DA singular nouns can be overtly marked for gender

only when feminine. They typically take the feminine suffix -at, which is sometimes

pronounced as /ah/ or /a/ when not followed by another sound (pause form, as termed

by Ryding, 2005:121). The typical form of the suffix is more common in SA because it

is usually followed by a case suffix. However, because DA nouns are not overtly

marked for case, the feminine suffix is usually in a word-final position, so the pause

form of the suffix is more common. There are some nouns that, despite being feminine,

do not take the feminine suffix.19 These nouns will be examined in Chapter Five.

Moreover, there is a claim, mainly in the traditional grammar of Arabic, that -at is not

the only feminine suffix in Arabic. This claim will be addressed in Section 2.5 where it

will be shown that it is an offshoot of analysing the language based on its written form,

which is a normal practice in the traditional grammar, rather than its spoken form.

With respect to the case feature, the above examples show that case is overtly

marked in SA, while the case-marking system is dropped in DA (Ingham, 1994:49;

Brustad, 2000:27). SA singular nouns usually take the suffixes -u when nominative, -a

when accusative, and -i when genitive, while DA nouns have no case suffixation as

shown in (2) above.

In terms of (in)definiteness, the above examples show that singular nouns (SA and

DA) take the prefix al- when definite, but it is only the SA nouns that are overtly

marked for indefiniteness. The definite prefix al- has different phonologically

19 For example, the feminine noun ʔardʕ (land), which agrees as feminine as shown below: al-ʔardʕ-u al-muqaddas-at-u Def-land.F-Nom Def-holy-F-Nom ‘The Holy Land.’ 20 Chapter Two – Arabic and Its Nouns

conditioned forms. The prefix appears as ʔal- in utterance-initial positions, and/or its /l/

part assimilates into the first consonant of the stem if the consonant is coronal (for more

on the phonology of the definite prefix, see, e.g., Abu-Salim, 1988; Alfozan, 1989;

Watson, 2002; Zuraiq and Zhang, 2006; Abu-Chacra, 2007; Youssef, 2013). For the

purposes of this study, the definite prefix will be presented as al-. When the nouns are

indefinite, on the other hand, they usually take a nunation in SA, while they are taken to

be indefinite when they do not take the definite prefix in DA.

2.4.2. Dual Nouns

As mentioned earlier, dual nouns express their number feature by suffixation. In SA, the

dual suffixes vary according to the case feature of the noun, so they mark number and

case. In DA, on the other hand, the dual suffix does not exhibit variation since there is

no case marking. Consider the following SA and DA examples of dual nouns:

(3) SA Dual Nouns: a. hunaːka mawdʕuːʕ-aːni muhim-aːni there subject.M-Dl.Nom important.M-Dl.Nom ‘There are two important subjects.’

b. ʕayanat dawl-at-u al-ʔimaːraːt-i safiːr-at-ayni appointed.it state-F-Nom Def-emirates-Gen ambassador-F-Dl.Acc ‘(The state of) Emirates appointed two female ambassadors.’

c. bayna safiːr-ayni between ambassador.M-Dl.Gen ‘Between (two) ambassadors.’

d. siʕr-u al-kitaːb-ayni price.M-Nom Def-book.M-Dl.Gen ‘The price of the two books.’

(4) DA Dual Nouns: a. baʕd ∫ahr-ain wu nusʕ after month.M-Dl and half ‘After two and a half months.’

b. san-at-ain taʔriːban year-F-Dl approximately ‘Approximately two years.’ 21 Chapter Two – Arabic and Its Nouns

c. ∫uft raʒaːl-ain yetkalamuːn saw.I man.M-Dl talking.they ‘I saw two men talking.’

In terms of gender, dual nouns (in SA and DA) behave similarly to their singular

counterparts. There is no overt marking for masculine, but the suffix -at is used to mark

the femininity of the nouns. When the suffix is used, it attaches directly to the stem, and

can be followed by the other suffixes.

As to the number and case features, SA dual nouns mark their duality by taking

the suffix -aːni when nominative, or the suffix -ayni otherwise. DA dual nouns,

however, always take the suffix -ain that marks the nouns for their duality regardless of

what their case feature would be, which is not overtly marked. In both varieties of

Arabic, the dual nouns take the prefix al- when definite, but no nunation is used to mark

their indefiniteness. The combined number and case suffixation as well as the absence

of the indefinite marker in SA duals will be accounted for in Chapter Six.

2.4.3. SndM Plural Nouns

SndM plurals are a group of nouns that are used with human referents, though not all

human plural nouns are SndM. These nouns represent one type of Arabic plurals that

express their number feature by suffixation. This suffixation in SA SndM plurals is

associated with the case feature of the nouns as well, as shown in (5). In DA, however,

the nouns take one suffix that is related only to their number feature, as shown in (6)

below:

(5) SA SndM Plurals: a. muraːqib-uːna rasmiy-uːna observer.M-Pl.Nom official.M-Pl.Nom ‘Official observers.’

b. ʔiʒtimaːʕ-a-n maʕa al-muwaðʕaf-iːna meeting.M-Acc-Idef with Def-employee.M-Pl.Gen ‘A meeting with the employees.’

22 Chapter Two – Arabic and Its Nouns

(6) DA SndM Plurals: a. al-baːħiӨ-iːn yeħtaːʒuːn maʕaːmil Def-researcher.M-Pl need.they lab.F.Pl ‘The researchers need labs.’

b. ∫uft al-muwaːtʕin-iːn bi al-saːħ-ah saw.I Def-citizen.M-Pl in Def-square-F ‘I saw the citizens in the square.’

As the name suggests, the gender of SndM plurals is always masculine, which is always

not overtly marked in Arabic. In terms of number and case, SA SndM plurals express

their plurality either by the suffix -uːna or -iːna depending on their case feature, while

DA SndM plurals take the suffix -iːn, which obviously is a pure number suffix. As in

dual nouns, SndM plurals take the prefix al- when definite, but they cannot take

nunation when indefinite. They are made indefinite simply by the absence of the

definite prefix. The joint expression of number and case in SA SndM plurals, as well as

their inability to be overtly marked for indefiniteness, will be discussed in Chapter Six.

2.4.4. SndF Plural Nouns

Just as SA duals and SndM plurals represent the nouns where number and case interact,

SndF plurals (in SA and DA) represent the instance where gender and number interact.

SndF plurals are unique by taking the suffix -aːt to express their plurality and

femininity, as exemplified below:

(7) SA SndF Plurals: a. tabaːdul-aːt-u-n tiʒaːriy-at-u-n exchange-F.Pl-Nom-Idef commercial-F-Nom-Idef ‘Commercial exchanges.’

b. takallamtu maʕa al-tʕaːlib-aːt-i spoke.I with Def-student-F.Pl-Gen ‘I spoke with the (female) students.’

c. yaftaħu maʒaːl-aːt-i-n waːsiʕ-at-a-n open.it field-F.Pl-Acc-Idef wide-F-Acc-Idef ‘It opens wide fields.’

23 Chapter Two – Arabic and Its Nouns

(8) DA SndF Plurals: a. marrayt bi al-tʕaːlib-aːt passed.I by Def-student-F.Pl ‘I passed by the (female) students.’

b. tʕaːwil-aːt wu karasiː table-F.Pl and chair.F.Pl ‘Tables and chairs.’

As shown above, the SndF plural suffix -aːt appears in SA and DA nouns to mark them

for their feminine gender and plural number.

With respect to the case feature, as with all DA nouns, DA SndF plurals do not

take case suffixes. However, SA SndF plurals are overtly marked for case, but they do

not distinguish between the accusative and genitive cases. They take the usual

nominative suffix -u when nominative, but they take the typical genitive suffix -i in the

non-nominative cases. A clear example can be seen in (7.c) above where both the SndF

plural noun maʒaːl-aːt-i-n (fields) and its modifying adjective waːsiʕ-at-a-n (wide) are

accusative with two different accusative markers. The noun is marked accusative by the

non-nominative (genitive) marker -i, while the adjective agrees with it by taking the

regular accusative suffix -a as the adjective is singular.20 Obviously, SA SndF plurals

have an unusual behaviour towards case, which will be discussed in Chapter Six.

In terms of (in)definiteness, SndF plural nouns behave consistently with most of

Arabic nouns. They can take the prefix al- when definite in SA and DA, but only the

SA ones can take nunation when indefinite. DA SndF plurals, like all DA nouns, do not

take nunation when indefinite. In most cases, DA SndF plurals are considered to be

indefinite when they do not take the definite prefix al-.

20 Adjectives agree with nonhuman plural nouns as singulars. This is part of a phenomenon known as deflected agreement, which will be discussed at greater length in Chapter Five. 24 Chapter Two – Arabic and Its Nouns

2.4.5. Broken Plural Nouns

Arabic (SA and DA) broken plurals are nouns that express some of their grammatical

features, especially plurality, by changing their binyanim. The below table shows how

some singular nouns change into broken plurals:

Table 2.3: Singulars vs. Broken Plurals Singulars Broken Plurals -kitaːb- -kutub- ‘Book’ ‘Books’ -dars- -duruːs- ‘Lesson’ ‘Lessons’ -madiːn-at- -mudun- ‘City’ ‘Cities’ -tʕabiːb- -ʔatʕibbaːʔ- ‘Doctor’ ‘Doctors’ -ʕaːlim- -ʕulamaːʔ- ‘Scholar/Scientist’ ‘Scholars/Scientists’

As can be seen above, some singular nouns change into plural by having new binyanim

(broken plurals) rather than by suffixation (sound plurals).21 It appears that the new

binyanim are the elements that carry the plural feature of the broken plurals.

In DA, broken plurals are overtly marked only for definiteness by having the

definite prefix al-. Broken plurals that are overtly marked for the features of

definiteness, indefiniteness and case can be found only in SA, as shown below:

(9) a. ʔiʃtara al-tʕulaːb-u al-kutub-a bought.they Def-student.Pl-Nom Def-book.Pl-Acc ‘The students bought the books.’

b. al-tadʕaːmun-u maʕa al-ʕummaːl-i Def-solidarity-Nom with Def-worker.Pl-Gen ‘The solidarity with the workers.’

c. baːq-aːt-u al-zuhuːr-i bouquet-F.Pl-Nom Def-flowers-Gen ‘Bouquets of flowers.’

21 Sound plurals and nouns with a similar formation strategy are sometimes called affixal (Carstairs, 1987; Joseph, 1997; Heath, 2005; Dammel and Nübling, 2006; Bickel, 2011), suffixing (Noyer, 1992) or suffixal (Newman, 1990, Kihm, 2011; Neme and Laporte, 2013). Broken plurals and similar nouns are sometimes called templatic (Faust, 2007; Al-Aghbari, 2012; Borer, 2013), stem-adjusting (Noyer, 1992), or stem- internal (Ratcliffe, 1998; Acquaviva, 2008). This study maintains the sound/broken distinction as it appears to be the most prominent in the literature on Arabic. 25 Chapter Two – Arabic and Its Nouns

d. al-ʕulamaːʔ-u ħaːʔir-uːna Def-scientist.Pl-Nom puzzled-Pl.Nom ‘The scientists are puzzled.’

e. kutub-u-n kabiːr-at-u-n book.Pl-Nom-Idef big-F-Nom-Idef ‘Big books.’

In terms of case, the above examples show that SA broken plurals behave identically to

their singular counterparts in expressing their case features. They take the suffix -u

when nominative, -a when accusative, and -i elsewhere. In terms of (in)definiteness, the

examples show that SA broken plurals, like SA singulars and SndF plurals, take the

prefix al- when definite and the suffix -n (nunation) when indefinite.

The major difference between broken plurals and the other types of Arabic nouns

is that the binyanim of broken plurals are the elements by which the nouns express their

number feature, as shown above, and their gender feature. The gender of broken plurals

can be determined by observing how the nouns agree. For example, the broken plural

ʕulamaːʔ (scientists) in (9.d) agrees as masculine, while the broken plural kutub (books)

in (9.e) agrees as feminine.22 As will be discussed from a DM perspective in Chapter

Five, broken plurals are not marked by suffixation for their gender and number features

because these features are embedded in the new binyanim that the nouns employ.

Given how the grammatical features of gender, number, case and (in)definiteness

are distributed among the different types of Arabic nouns, it can be said that

(in)definiteness is the feature that does not show any interaction with the other features

in SA and DA nouns. Definiteness is expressed by an independent prefix in SA and DA,

while indefiniteness is either expressed by an independent suffix (most SA nouns) or

simply not overtly marked (DA and some SA nouns). Apart from (in)definiteness, the

22 kutub (books) is the plural form of the singular kitaːb (book). The noun is masculine when singular, but feminine when plural. This is considered part of the previously mentioned phenomenon of deflected agreement in which nonhuman plurals (broken and sound) agree as feminine singular. In Chapter Five, it will be shown that nonhuman masculine singulars change into feminine when pluralised. 26 Chapter Two – Arabic and Its Nouns

rest of the features may interact with each other depending on the variety of the

language to which they belong and the type of their nouns.

In SA nouns, gender is an independent feature in singulars, duals and SndM

nouns, while it interacts with number in SndF and broken plurals. Number is an

independent feature only in singulars, while it interacts with case in duals and SndM

plurals. Finally, case is an independent feature only in singulars, SndF and broken

plurals. These interactions in SA nouns can be represented as shown below:

Figure 2.1: Interactions of Gender, Number, and Case in SA Nouns

In DA nouns, on the other hand, since there is no case-marking system, no

obvious interaction between case and the other features can be detected. As in SA,

gender in DA nouns is either an independent feature (singulars, duals and SndM plurals)

or a feature that interacts with number (SndF and broken plurals). Number is

independent in DA singulars. As a result of the absence of a case-marking system,

number is also independent in the duals and SndM plurals.

Figure 2.2: Interactions of Gender and Number in DA Nouns

27 Chapter Two – Arabic and Its Nouns

Most of this thesis will be devoted to analysing these several interactions within the

theoretical framework of DM. This analysis will provide us with a better understanding

of the structure of Arabic nouns and the theory of DM.

2.5. Is -at the only Feminine Suffix?

Recall that -at is a feminine suffix, but it is not taken as the only feminine suffix in the

traditional grammar of Arabic. Traditional grammarians, and some recent linguists such

Bateson (2003:12), Ryding (2005:123) and Alhawary (2009:5) who might be influenced

by the traditional view, take /aːʔ/ and /aː/ that appear at the end of the stems of some

feminine nouns to be feminine suffixes. Consider the following examples:

(10) a. al-ħikaːy-at- al-qadiːm-at- / *al-qadiːm- Def-story-F Def-old-F Def.old.M ‘The old story.’

b. al-sʕaħraːʔ- al-ʕarabiy-at- / *al-ʕarabiy Def-desert.F Def-Arabian-F Def-Arabian.M ‘The Arabian Desert.’

c. al-ðikraː al-sanawiy-at- / *al-sanawiy Def-remembrance.F Def-annual-F Def-annual.M ‘The anniversary.’

This study takes -at as a (feminine) suffix, but argues that /aːʔ/ and /aː/ are not. Rather,

/aːʔ/ is considered part of the binyan /C1aC2C3aːʔ/ that, in the above examples,

interlocks with the roots /sʕħr/ forming the noun sʕaħraːʔ (desert), and, similarly, /aː/ is

considered part of the binyan /C1iC2C3aː/ that forms the noun ðikraː (remembrance).

Although these two nouns are feminine and agree as feminine, this study takes them as

covert feminine nouns, like the covert feminine noun ʔardʕ (land). The structure of

covert feminine nouns will be discussed in Chapter Five.

The reason why grammarians take /aːʔ/ and /aː/ as suffixes may have to do with

the approach they follow when analysing the language. Arabic traditional grammar is

mainly based on analysing the written form of words in which consonants and long 28 Chapter Two – Arabic and Its Nouns

vowels are represented with actual letters, while short vowels are represented with

diacritic marks added above or under some consonants, as shown below:23

Figure 2.3: Arabic Scripts of sʕaħraːʔ (Desert) and ðikraː (Remembrance)

The above figure shows how the feminine nouns sʕaħraːʔ (desert) and ðikraː

(remembrance) are written and how each noun is divided into three different types of

elements. According to the written forms, in the circles we have the roots, which are

represented as /sʕħr/ and /ðkr/ in this study. In the triangles we have the diacritic marks

that correspond to the short vowels /a/ and /i/. This leaves us with what are claimed to

be the feminine suffixes /aːʔ/ and /aː/ in the rectangles.

Diacritic marks are named in Arabic as ħarakaːt (movements). As the name

suggests, they describe the articulation movements readers should make immediately

after pronouncing the marked consonants. The most common diacritic marks are called

fatħah (opening), dʕammah (bringing together/joining), and kasrah (breaking).24 Taking

when followed by another (ﻛـ) as an example the consonant /k/, which is written as

causes it to be pronounced as /ka/, adding ( َﻛـ) letter, adding fatħah above the consonant

turns ( ِﻛ ـ) makes its pronunciation /ku/, and adding kasrah under it ( ُﻛ ـ) dʕammah above it

it into /ki/. Although the short vowels that these marks represent are always pronounced,

23 The Arabic writing system is right-to-left, and the circles, triangles, and rectangles are added for illustration purposes. 24 The rest of the marks include sukuːn (quietness), which indicates the absence of any vowel after the consonant it marks, and ∫addah (tightening/emphasising), which indicates that the consonant is geminated (for more on the diacritic marks, see Chapter Two of Ryding, 2005). 29 Chapter Two – Arabic and Its Nouns

using these marks in writing is optional.25 In most cases, written words can be read and

pronounced correctly without any diacritic marks. However, in rare cases, mostly when

writers think the script of the intended word is likely to be confused with a similar

script, they add the appropriate diacritic marks to avoid any misreading or

mispronunciation. What is important here is that grammarians consider the diacritic

marks and the short vowels they stand for distinct from /aːʔ/ and /aː/. Removing /aːʔ/

and /aː/ from the written forms leaves us with the roots, so /aːʔ/ and /aː/ are taken to be

suffixes.

In order to examine whether /aːʔ/ and /aː/ are feminine suffixes, their behaviour is

compared to the behaviour of the feminine suffix -at in gender shifting. If /aːʔ/ and /aː/

are feminine suffixes, they are expected to behave in the same way the feminine suffix

-at does. That is, if /aːʔ/ and /aː/ are feminine suffixes, they will disappear

independently from the other elements of the words to which they are attached. An

examination of this sort is more effective with adjectives for two reasons. Firstly,

adjectives are not inherent or restricted to a specific gender. They can be masculine or

feminine depending on the gender of the nouns they modify, while the gender of many

feminine nouns, e.g., sʕaħraːʔ (desert) and ðikraː (remembrance), cannot be changed

into masculine as they have no masculine counterparts. Secondly, grammarians who

claim that /aːʔ/ and /aː/ are feminine suffixes do not discriminate between nouns and

adjectives. According to Ingham (1994:47) and Bateson (2003:9), traditional

grammarians divide words into three classes: verbs, nouns, and particles. In this

classification, adjectives and nouns are not considered distinct, so refuting the claim that

25 According to Versteegh (1997), Arabic letters used be written without diacritic marks, which were introduced later into the writing system (for more on the development of Arabic scripts, see Chapter Five of Versteegh, 1997). Moreover, Ryding (2005:25) states that diacritic marks are invisible in the modern practice of Arabic writing. 30 Chapter Two – Arabic and Its Nouns

/aːʔ/ and /aː/ are suffixes in adjectives entails refuting it in nouns, and vice versa.

Supposing that /aːʔ/ and /aː/ are suffixes, consider the following examples:

(11) a. al-ħikaːy-at- al-qadiːm-at- à al-matʕaːr- al-qadiːm- Def-story-F Def-old-F Def-airport.M Def-old.M ‘The old story.’ ‘The old airport.’

b. al-badl-at- al-zarq-aːʔ à al-ħuːt- *al-zarq- Def-suit-F Def-blue-F Def-whale.M Def-blue ‘The blue suit.’

c. al-ħarb- al-kubr-aː à al-iskandar *al-kubr- Def-war.F Def-greatest-F Def-Alexander.M Def-greatest ‘The Great War.’

The example (11.a) shows that replacing the modified feminine noun with a masculine

one entails changing the gender of the modifying adjective from feminine into

masculine. This is achieved simply by removing -at from the adjective, without

affecting anything else of its constituents. However, applying the same strategy to the

feminine adjectives zarq-aːʔ in (11.b) and kubr-aː in (11.c) results in ungrammaticality

and does not serve the purpose.

The only way to change the gender of these adjectives is to replace their

binyanim, including the /aːʔ/ and /aː/ endings, as shown below:

(12) a. al-badl-at- al-zarqaːʔ à al-ħuːt- al-ʔazraq- Def-suit-F Def-blue.F Def-whale.M Def-blue.M ‘The blue suit.’ ‘The blue whale.’

b. al-ħarb- al-kubraː à al-iskandar al-ʔakbar- Def-war.F Def-greatest.F Def-Alexander.M Def-greatest.M ‘The Great War.’ ‘Alexander the Great.’

As shown above, changing the gender of the feminine adjectives zarqaːʔ (blue) requires

replacing its binyan /C1aC2C3aːʔ/ with /ʔaC1C2aC3/, which produces the masculine

adjective ʔazraq when interlocked with the root consonants /zrq/. Similarly, in order for

the gender of the feminine adjective kubraː (greatest) to be changed into masculine, its

31 Chapter Two – Arabic and Its Nouns

binyan /C1uC2C3aː/ needs to be replaced with /ʔaC1C2aC3/, which produces the

masculine adjective ʔakbar when interlocked with the root /kbr/.26

It is now obvious that, unlike the suffix -at, /aːʔ/ and /aː/ are not suffixes that can

be added or removed in order for gender shifting to be achieved. They appear/disappear

as parts of their binyanim, whereas -at appears/disappears independently from any stem

constituent. Therefore, /aːʔ/ and /aː/ cannot be suffixes as traditionally thought. Thus,

this study divides feminine nouns into two types: overt feminine nouns, which take the

feminine suffix -at, and covert feminine nouns, which do not take the suffix. The latter

type includes feminine nouns that end with /aːʔ/ or /aː/. The gender feature of both types

will be analysed from a DM perspective in Chapter Five.

2.6. Concluding Remarks

It was shown in this chapter that the morphology of Arabic nouns is rich and complex.

It was demonstrated that roots are of a consonantal nature and they are functionally poor

and semantically vague. The non-concatenative nature of the language regarding the

way in which stems are usually formed means careful scrutiny is required when

identifying roots. The consonants of the roots spread between the constituents of the

binyanim, which may contain some other consonants as well. In Chapter Four, an

analysis of roots and stem formation in Arabic from a DM perspective will be provided.

It will be shown that the nature of Arabic roots is challenging for some theoretical

assumptions that seem to be based on the nature of the roots of concatenative languages.

It was also shown that contextualised Arabic nouns express their grammatical

features of gender, number, case and definiteness in several ways depending on the

variety of the language and the type of nouns to which they belong. Although these

26 This is possibly an indication that some adjectival binyanim can bear gender. This requires an in depth investigation of Arabic adjectives, which is a temptation that should be resisted as such an investigation is beyond the scope of this thesis. 32 Chapter Two – Arabic and Its Nouns

features can be independently expressed, they can sometimes interact with each other.

In SA nouns, some interactions between gender, number and case can be observed.

Such interactions in DA, however, can be observed only between gender and number

with no involvement of the case feature due to the absence of an overt case-marking

system. The regular ways by which Arabic nouns express these grammatical features,

and how these features interact with each other in some nouns, will be accounted for

within the word formation mechanism proposed by DM in Chapters Five (gender and

number) and Six (case and definiteness). Some exceptions such as covert feminine,

diptote and invariable nouns will be discussed in due course within these chapters.

The last point discussed in this chapter was related to the way the femininity of

some Arabic nouns is overtly expressed. It was shown that, in the traditional literature

of Arabic grammar as well as in a few recent works, the feminine suffix -at is taken to

be one of three feminine suffixes in Arabic. The other two suffixes are claimed to be the

endings /aːʔ/ and /aː/ that appear on some feminine nouns. It was argued that this claim

results from the traditional approach of analysing the written form of the language

instead of its spoken form. It was illustrated that due to the lack of written letters

representing the short vowels of the binyanim, these vowels are considered distinct from

the /aːʔ/ and /aː/ endings. Since removing the letters that represent /aːʔ/ and /aː/ results

in leaving the roots alone, /aːʔ/ and /aː/ are traditionally considered suffixes. By

comparing the behavior of /aːʔ/ and /aː/ to the behavior of the feminine suffix -at in

gender shifting, it was revealed that each of /aːʔ/ and /aː/ is a binyan ending, whose

appearance/disappearance is conditioned by the appearance/disappearance of its whole

binyan, whereas -at behaves like an independent suffix. Thus, -at was considered the

only feminine suffix of Arabic. As a result, feminine nouns are either overt or covert

feminine nouns, as will be discussed further in Chapter Five.

33 Chapter Three – Theoretical Framework

3. Chapter Three - Theoretical Framework

3.1. Introduction

This chapter describes the main principles and key properties of the theory of DM as

first proposed by Halle and Marantz (1993, 1994) and pursued by many others such as

Marantz (1995, 1996, 1997); Halle (1997); Harley and Noyer (1999); Embick and

Noyer (2007) and Embick and Marantz (2008). Although DM is over 20 years old, it is

still evolving and some of its claims are still controversial. Inevitably, the constant

development of the theory has resulted in some variations –including in terminology –

between the foundation and recent works, as well as divergences among the different

works involved in the internal debates of the theory. In addition to providing a general

description of DM, this chapter also sheds some light on the variations that are relevant

to this thesis.

Generally, DM is a theory that redefines the traditional concept of Lexicon (as in

the Lexicalist Hypothesis) by distributing its functions among different components,

and employs a syntactic approach to the morphology of words, hence its name

Distributed Morphology. Section 3.2 provides a brief overview of DM showing how it

is different from some of the prominent theories of the time, and the architecture of the

grammar it proposes. The components that replace the Lexicon have different contents,

and accordingly have different functions that will be described in some detail. These

components are known as the Narrow Lexicon (Section 3.3), the Morphology (Section

3.4), the Vocabulary (Section 3.5) and the Encyclopaedia (Section 3.6). Section 3.7

concludes the chapter by showing how DM is distinctive, and what contributions to the

theory this thesis is expected to make.

34 Chapter Three – Theoretical Framework

3.2. Overview

It is a longstanding convention in theories of Generative Grammar that the

computational system is responsible for generating sentences in hierarchical structures

using rules made available by Universal Grammar (UG), while the Lexicon is where

words are generated. The input for the computational system is syntactically,

phonologically and semantically specified words (lexical items) drawn from the

Lexicon. This view is generally known as the Lexicalist Hypothesis or Lexicalism

(Chomsky, 1970).

A debate within the Lexicalist Hypothesis revolves around how much of the

morphology of words is generated in the Lexicon. In one version of this model, known

as the Strong Lexicalist Hypothesis, there is a sharp division between sentence structure

(syntax) and word structure (morphology, including derivational and inflectional

morphology). The entire morphology of words is considered a property of the Lexicon,

while the computational system is only concerned with building sentences. However,

motivated by phenomena that show some interactions between morphology and syntax,

another version of the model, known as the Weak Lexicalist Hypothesis, distinguishes

between the derivational and the inflectional aspects of the morphology. It is argued that

inflectional morphology is performed in the computational system along with sentence

structure, while derivational morphology is maintained to be performed in the Lexicon

(for more on this debate, see, e.g., Siegel, 1974; Aronoff, 1976; Wasow, 1977; Lieber,

1980; Williams, 1981; Anderson, 1982; Kiparsky, 1982; Selkirk, 1982; Levin and

Rappaport, 1986; Di Sciullo and Williams, 1987; Booij, 1996; Dubinsky and Simango,

1996).

Apart from the status of inflectional morphology, it does not seem to be disputed

within the Lexicalist Hypothesis that the Lexicon is generative, and the phonological

35 Chapter Three – Theoretical Framework

and the semantic properties of its outputs are manipulated by the computational system.

It appears that the attention that both versions of the Lexicalist Hypothesis have paid to

identifying the boundaries between morphology and syntax, and between derivational

and inflectional morphology, have distracted them from questioning the necessity of

burdening the computational system with phonological and semantic information that is

irrelevant to its functions. Without being accessed or showing any interactions with its

principles, carrying these phonological and semantic properties into the computational

system is like carrying “excess baggage” (Marantz, 1995:399).

There have been a number of efforts to reduce the amount of irrelevant

information involved in the computational system, as in Anderson’s (1992) A-

Morphous Morphology. On the one hand, A-Morphous Morphology agrees with the

Weak Lexicalist Hypothesis that the computational system is involved in the

inflectional morphology of words (Anderson, 1992:101):

We therefore took this characteristic as potentially definitional: inflectional morphology is the area in which principles of syntactic structure and of word formation interact with one another, and a theory of inflection is in some sense a theory of the ways in which the maximally strong form of the Lexicalist Hypotheses must be relaxed.

On the other hand, A-Morphous Morphology distinguishes between the contents of

roots – as parts of the derivational morphology that is performed in the Lexicon – and

inflectional morphology. Roots are phonologically and semantically specified, while

inflectional morphology is achieved by means of purely syntactic rules where no

phonological elements are involved (Anderson, 1992:72):

We thus take the formal constituents of complex words to be not listed morphemes, but rather operations on the form of words. On this view, dogs is represented not by concatenating /dɔg/ and /-z/, but rather by having /dɔg/ undergo a rule whose change consists of /X/ -> /Xz/, whenever it represents a Noun with the property of being [+Plural].

With sentences being structured in the computational system by manipulating fully

specified roots, A-Morphous Morphology, like the Lexicalist Hypothesis, fails to lift the

burden from the computational system. Most of the excess baggage is still there as the

36 Chapter Three – Theoretical Framework

phonological and semantic features of dog that distinguish it from cat do not drive any

syntactic principle, rule or constraint (Marantz, 1996:3).

In order to accomplish a full eradication of the computationally irrelevant

features, DM argues that words are neither structured in the Lexicon, nor

phonologically or semantically pre-specified when drawn into the computational

system. Although DM acknowledges the computational system (the Syntax) as the

engine where sentences are structured, it also takes the Syntax as the engine that

structures words as well. The Syntax carries out its functions through syntactic

operations such as Merge and Move (Chomsky, 1995a) that, in DM, manipulate atomic

syntactic elements whose phonological and semantic properties are provided later in the

process. In other words, DM is a single-engine piece-based model with a separationist

character. It employs a syntactic approach towards word formation, and makes a sharp

distinction between the syntactic, phonological and semantic properties of words.

The DM view of word formation is incorporated in a model of grammar (see

Figure 3.1) in which the underlying structure of words is built in the Syntax by merging

purely syntactic items (known as morphemes). These morphemes are drawn from the

Lexicon (now, the Narrow Lexicon) where they originally reside. The semantic

representations of the syntactic structures are provided in the Encyclopaedia on the

Logical Form (LF) branch, while the phonological representations are provided in a

component known as the Vocabulary on the Phonological Form (PF) branch of the

model. The Vocabulary is mainly the storehouse of the phonological elements available

in the language, known as the Vocabulary Items (VIs). Morphemes are phonologically

represented in the Vocabulary by the insertion of the relevant VIs into them in a process

known as Vocabulary Insertion. In certain circumstances determined by particular

morphological requirements of the language, structures may need to be post-

37 Chapter Three – Theoretical Framework

syntactically modified, mainly by adding or altering constituents. This takes place in a

component called the Morphology prior to the Vocabulary on the PF branch of the

model. The basic architecture of the grammar proposed by DM can be presented as

shown below:

Figure 3.1: The Architecture of the Grammar of DM

Narrow Lexicon

Syntax PF LF

Morphology Encyclopaedia

Vocabulary

Hence, DM basically rejects the Lexicon, in its traditional sense, and distributes its

functions across several components of the new grammar it proposes (Halle and

Marantz, 1993, 1994; Marantz, 1995, 1996, 1997; Halle, 1997; Harley and Noyer, 1999;

Embick and Noyer, 2007; Embick and Marantz, 2008; Harley, 2011, to name a few). In

the following sections, the contents and functions of the components replacing the

Lexicon are illustrated.

3.3. The Narrow Lexicon

The Narrow Lexicon, which is also termed the Pure Lexicon (Marantz, 1996), is what

most directly substitutes the traditional Lexicon. It is the storehouse of the primitive

elements constituting structures, which are known as morphemes. Morphemes are the

input for the Syntax where structures are built. These morphemes are devoid of any

38 Chapter Three – Theoretical Framework

computationally irrelevant phonological and semantic information, and they are of two

types: root morphemes, and functional morphemes.27

In addition to their lack of phonological and semantic information, root

morphemes have no syntactic information by which they might be categorised as nouns,

verbs, and so forth. They are basically category-less, category-neutral, or simply

acategorial. Thus, the structure of a word is composed of at least a root morpheme

merged with a category-assigning morpheme such as [n] or [v], as will be discussed in

Chapter Four (e.g., Marantz, 1996, 2001; Harley and Noyer, 1998, 1999; Arad, 2003,

2005; Harley, 2005; Embick and Noyer, 2007; Embick and Marantz, 2008;

Lowenstamm, 2008). Given that they have no phonological, semantic or syntactic

information, root morphemes are simply indistinct, and so are often represented with

[√]28 or [√ROOT]. This has sparked an ongoing debate regarding how indistinct root

morphemes are given distinct phonological and semantic representations later in the

process. These issues will be touched upon in Sections 3.5 and 3.6, and discussed in

more detail in Chapter Four.

Unlike root morphemes, functional morphemes are distinct by the different

features and values they bear (e.g., [Pl] and [+Past]), so the determination of their

representations is much easier than for that of the roots. According to Noyer (1992), the

active features in any language are a subset of a wider range of features made available

for selection by UG. Thus, the Narrow Lexicon is the place in which the selected subset

of features are stored, as well as the place from which a subset of the stored subset is

drawn into the Syntax for manipulation.

27 Also known as l-morphemes/l-nodes (lexical) and f-morphemes/f-nodes (functional), respectively (Harley and Noyer, 1998, 1999). 28 The √ symbol notating roots is borrowed from Pesetsky (1995) and widely adopted in the literature. 39 Chapter Three – Theoretical Framework

3.4. The Morphology

The Morphology is a language-specific component that contains the morphological

rules of the language. It is an intermediate level on the PF branch that is responsible for

operating on the syntactic structures in order to make them compliant with these rules

before they are phonologically realised. The morphological modifications can be carried

out by operations that may delete the features of certain morphemes (Impoverishment),

combine the features of multiple morphemes into a single one (Fusion), or split the

features of a given morpheme into multiple morphemes (Fission).29 Among these

operations, Impoverishment and Fusion are the most relevant to this study, so they will

be explained in more detail in the coming subsections.

In DM, adding morphemes to structures in the Morphology is possible.30 The

added morphemes are known as dissociated morphemes, as defined by Embick (1997:8)

below:

Dissociation: A morpheme will be called dissociated when the morphosyntactic position/features it instantiates are not features figuring in the syntactic computation, but are instead added in the Morphological component under particular structural conditions.

The concept of morpheme addition in the Morphology appears to be employed to

account for cross-linguistic variations related to the features of case (Marantz, 2000;

McFadden, 2004),31 agreement (Halle and Marantz, 1993) and voice (Embick, 1997),

for more see, e.g., Halle and Marantz (1994); Marantz (1996); Harley and Noyer (1999)

and Embick and Noyer (2001, 2007). Given that DM is a Y-model in which the Syntax

feeds both PF and LF, and dissociated morphemes are inserted on the PF branch, it

follows naturally that the features of dissociated morphemes are not relevant to the

29 Another operation that is found in the foundation works of DM is called Morphological Merger. It will be discussed in relation to Fusion in Section 3.4.2. 30 The source of the added morphemes is not clear in the literature of DM. Logically, they are either stored in the Morphology as language-specific features, or imported from the Narrow Lexicon. The latter entails that the Morphology has access to the Narrow Lexicon in the model. 31 Marantz (2000) is a paper presented at the Eastern States Conference on Linguistics, University of Maryland, in 1991. 40 Chapter Three – Theoretical Framework

Encyclopaedia. That is, if a feature is semantically relevant, it must figure in the Syntax

so it can be part of the syntactic structure handed over to the Encyclopaedia (Embick

and Noyer, 2001:558; 2007:305; Picallo, 2008:48).

As far as Arabic nouns are concerned, this classification entails that case is a

dissociated morpheme that is added in the Morphology, while gender, number and

definiteness are syntactic morphemes that figure in the Syntax.32 This should not be

confused with the gender, number, case and definiteness features that may appear (all or

some of them) on adjectives, verbs and pronouns as agreement features. These features

are parts of the dissociated AGR(eement) morphemes of these structures. They are

copied from the nouns onto the AGR morphemes at the agreement level in the

Morphology (for more on how agreement is achieved, see, e.g., Halle and Marantz,

1993:115; Embick and Noyer, 2007:305; Arregi and Nevins, 2012:8).

There is no clear indication in the literature of DM that features can be changed

in the Morphology. Marantz (1996:14), however, makes a brief reference to the issue

without ruling it out conclusively:

We have here assumed that no feature changing operations exist in the morphophonology. Apparent feature changing might result from feature deletion (impoverishment) followed by a fill-in of the unmarked value of a feature. Note, however, that simply disallowing feature-changing doesn’t make the data that apparently motivate feature changing go away. There is class of phenomena that Zwicky handles through “rules of referral” and that might motivate feature-changing in the morphology. Careful and exhaustive evaluation of these phenomena would be necessary for anyone that wished to claim that feature-changing is impossible.

Obviously, although Marantz (1996) is sceptical about the ability of the Morphology to

perform a direct feature-changing operation, he leaves the issue open for further

investigation as such an operation might be needed in some languages. In Chapter Five,

it will be shown that a feature-changing operation is probably needed in the

32 It is not really significant to this study whether case is a dissociated morpheme or not, as this study is more concerned with how it is handled after becoming part of the structures, rather than with how it becomes part of them. What might be of interest to this study, though, is the status of gender. Given that gender can be semantically relevant (Picallo, 2008), it is considered a syntactic feature. This is consistent with several views that take gender to be (in various ways and positions) a syntactic feature (e.g., Picallo, 1991; Ritter, 1993; Pfau, 2002, 2007; Müller, 2004; Lowenstamm, 2008; Acquaviva, 2009). 41 Chapter Three – Theoretical Framework

Morphology of Arabic. An operation of this sort, which will be termed Revaluation,

will be proposed to account for a phenomenon in which nonhuman plural structures are

not allowed to surface as masculine. Given that gender is a syntactic morpheme,

attributing this restriction to the Syntax would undermine its universality. Therefore,

while assuming that gender in Arabic is fully specified, i.e., there exist [-F] and [+F] in

the Narrow Lexicon, the proposed operation will be claimed to target the value of the

gender morpheme of these structures if specified as [-F] and change it into [+F]. It will

be shown that this operation is motivated by a morphological filter (Noyer, 1992) that

prohibits the co-occurrence of the features [-F], [Pl] and [-Hum] in nominal structures.

Now, we move to the morphological operations of Impoverishment and Fusion.

3.4.1. Impoverishment

As first introduced by Bonet (1991) and adopted by others (e.g., Noyer, 1992; Halle and

Marantz, 1993, 1994; Halle, 1997), Impoverishment is a morphological operation by

which features of the targeted morphemes are deleted before they are phonologically

represented. This blocks the insertion of the VI that corresponds to the deleted feature,

which may consequently allow the elsewhere VI to be inserted into the impoverished

morpheme instead. It is simply an operation by which the relevant language-specific

rules that disallow certain features in certain contexts are enforced.

To illustrate the effect of Impoverishment, an example from Norwegian adjectival

suffixes (Sauerland, 1995, as cited in Harley and Noyer, 1999) is presented. There is a

distinction in Norwegian adjectives between strong and weak syntactic positions.33 In

the strong positions, adjectives can take -t, -e or simply Ø depending on the value of

their gender ([±Neuter]) and number ([±Pl]) features, while in the weak positions the

suffix is always -e, as shown in the table below:

33 According to Julien (2005), a Norwegian adjective is in a strong position when attributive indefinite or predicative, while it is in a weak position when attributive definite. 42 Chapter Three – Theoretical Framework

Table 3.1: Strong and Weak Norwegian Adjectival Suffixes Strong [-Neuter] [+Neuter] [-Pl] Ø -t [+Pl] -e -e

Weak [-Neuter] [+Neuter] [-Pl] -e -e [+Pl] -e -e

It would seem that -e is the least specified (elsewhere) VI as it appears in all weak

positions and all plural strong positions. Therefore, Sauerland (1995) proposes the

following set of environments in which the Norwegian adjectival VIs can be inserted:

(1) /t/ ↔ [-Pl, +Neuter] / Adj + _____

/Ø/ ↔ [-Pl, -Neuter] / Adj + _____

/e/ ↔ elsewhere

Although the above environments perfectly capture the distribution of the VIs in the

strong positions, they do not capture the fact that /e/ appears in all weak positions.

According to the above representation, /t/ and /Ø/ are predicted to appear in the weak

singular positions as well, which is not actually the case. Thus, given that each of /t/ and

/Ø/ is sensitive to a certain value of gender in the presence of [-Pl], the only way to

block their insertion in the weak positions is to assume that the morphemes are

completely genderless. That is, the gender feature of the morphemes, whether it is

[+Neuter] or [-Neuter], is Impoverished, as shown below:

(2) [±Neuter] → Ø

This will sufficiently block the insertion of /t/ and /Ø/, and allow the elsewhere /e/ to be

inserted instead.

This mechanism represents the standard formulation of Impoverishment as a

feature-deletion operation. Noyer (1998), however, argues that Impoverishment is a

feature-changing operation that deletes a feature in order to allow a less marked value of

43 Chapter Three – Theoretical Framework

the feature to take its place, as also suggested by Marantz (1996:14) above (for a

response to Noyer's argument, see Vinka, 2001). According to this view, the VI that will

be inserted into the impoverished morpheme is not necessarily the elsewhere VI. It is

simply the VI that corresponds to the new feature, which may or may not be the

elsewhere VI of the morpheme.

Applying this view of Impoverishment to the Norwegian adjectival suffixes in

weak positions will result in replacing the morpheme [-Pl, -Neuter] with [-Pl, +Neuter],

or vice versa depending on which of [-Neuter] and [+Neuter] is the less marked feature.

Either way, however, /e/ will not be inserted. In order to account for the insertion of /e/

in non-plural weak positions, a feature-deletion that render the morphemes [-Pl] must

take place before the insertion. This emphasises the necessity of maintaining

Impoverishment as a feature-deletion operation. Nevertheless, if feature-changing is

detected in a language, a new operation must be introduced separately from

Impoverishment. In this study, we will encounter situations where feature-deletion and

feature-changing are required. For feature-deletion, the original concept of

Impoverishment will be maintained and supported, while feature-changing will be

accounted for by proposing a new operation, as mentioned earlier.

3.4.2. Fusion

Fusion is a morphological operation by which features of multiple morphemes are

combined into a single morpheme. Its major effect is phonological, as the fused features

are now expressed by a single VI. In order for the operation to be performed, the

targeted morphemes must be sisters under a single category node. However, if the

morphemes are not syntactically sisters, they can be made so in the Morphology by an

operation known as Morphological Merger (M-Merger) (early formulation in Marantz,

1984, 1988) only if they are adjacent (Halle and Marantz, 1993:116):

44 Chapter Three – Theoretical Framework

We distinguish here between "merger" and "fusion." Merger, like head-to-head movement, joins terminal nodes under a category node of a head (a "zero-level category node") but maintains two independent terminal nodes under this category node. Thus, Vocabulary insertion places two separate Vocabulary items under the derived head, one for each of the merged terminal nodes…On the other hand, fusion takes two terminal nodes that are sisters under a single category node and fuses them into a single terminal node. Only one Vocabulary item may now be inserted…Unlike merger, fusion reduces the number of independent morphemes in a tree. Since both head-to-head movement and merger form structures in which two terminal nodes are sisters under a single category node, both may feed fusion.

As exemplified by the authors, head-to-head movement may include moving

English auxiliary verbs to Tense, which is then moved to Complementiser, while M-

Merger combines Tense with English main verbs. Fusion, on the other hand, is

responsible for the existence of a single affix for case and number in many Indo-

European languages, as compared to Turkish in which case and number are expressed

by separate affixes. Similarly, as seen in Bobaljik (2005), while M-Merger may not be

involved in the formation of more intelligent, smart-er is produced when M-Merger is

applied. In both cases, however, the adjectival root and the comparative node are two

terminals for two different VIs. A single overt VI for the two terminals can be seen in

worse, which may indicate the involvement of Fusion.

A more detailed example of Fusion can be seen in Acquaviva (2009:8-9). The

author observes that the plural form of ox is either oxen or oxes depending on the

humanness feature of the noun. That is, the canonical use of the noun leads to the usual

irregular plural form oxen, while using the noun to metaphorically describe a human

(i.e., large uncouth man) often leads to the regularised plural form oxes.34 Assuming that

roots are distinct before the insertion level, the author accounts for the former plural

form by proposing the following Fusion rule by which the morpheme [n, Pl] is created:

(3) [n] + [Num: Pl] → [n, Num: Pl] / [OX, CHILD, BRETHR, GOOSE, MOUSE…]

The above rule can be interpreted as: whenever the category-assigning morpheme [n]

34 Similarly, human parasites are often louses rather than lice (Acquaviva, 2009:8).

45 Chapter Three – Theoretical Framework

co-occurs with the number morpheme [Pl] in the environment of any of the roots OX,

CHILD, etc., [n] and [Pl] are fused creating the morpheme [n, Pl]. This eventually

allows the insertion of the suffix -en, as the following:

(4) [n, Num: Pl] ↔ -en / [OX, CHILD, BRETHR]

For the regularised plural form oxes, the author assumes that the human reading of

the noun involves an additional [n] morpheme hosting the feature [+Hum], as shown

below:

(5) NumP

nP Num: Pl

nP n[+Hum]

OX n

The intervening n[+Hum] disrupts the adjacency of [Num: Pl] and the lower [n] and

prevents them from fusing. This results in [Num: Pl] being phonologically represented

with the regular plural suffix -s, producing the plural form oxes. This entails that the

following is the syntactic structure of oxen in which the fusion of [n] and [Num: Pl] is

permitted:

(6) NumP

nP Num: Pl

OX n

Observe that the targeted morphemes are not sisters, but they are of course considered

fusible since M-Merger is capable of making them sisters prior to the activation of the

Fusion rule in (3) above.

Siddiqi (2010:533) describes the relation between Fusion and M-Merger as

follows:

46 Chapter Three – Theoretical Framework

The process of fusion is in many ways the extension of morphological merger. In many cases, one lexeme (VI) primarily realizes more features than are present in just one syntactic node. This is often the case with agreement marking of phi-features on the verb (-s realizes not only present tense but third person and singular as well). After morphological merger has created a complex head, the application of fusion reduces the complex head to a simplex one, containing all the features previously contained in the complex head.

M-Merger is an intermediate operation that paves the way for the application of Fusion.

Hence, as also shown in Acquaviva’s (2009) analysis, when an M-Merger that feeds

Fusion is detected in Arabic nouns, this thesis will deal with the Fusion part of the

process assuming that M-Merger has already been performed.

3.5. The Vocabulary

The Vocabulary is the inventory of the phonological expressions available in the

language. These expressions are sorted as items (VIs) that correspond to the morphemes

of structures. After structures have been built in the Syntax and, if needed, modified in

the Morphology, they become ready to be phonologically represented in the

Vocabulary. This is achieved through the insertion of VIs into their constituting

morphemes. However, given that morphemes are of two types, root and functional, each

type has its own insertion mechanism.

In early DM, compositional semantic information such as animate/inanimate and

count/mass is considered to be present in the Syntax as a property of the roots (Marantz,

1995, 1996). As far as nouns are concerned, this type of information also includes

common/proper and human/nonhuman, as found in Chomsky (1965), who takes them as

compositional semantic features (henceforth, CSFs) represented as [±Animate],

[±Count], [±Common], and [±Human]. According to Marantz (1995, 1996), CSFs and

the overall syntactic environment surrounding root morphemes are assumed to be

involved in the choice of the inserted root VIs. However, since numerous root VIs can

be eligible for insertion in any given syntactic environment and combination of CSFs

47 Chapter Three – Theoretical Framework

(consider e.g., /kæt/ and /dɔg/), it is not actually clear how the final choice is made, so

alternative proposals have emerged.

Recent proposals argue that each root morpheme has an index that gives it a

unique identity and, consequently, guides the insertion process. Indices make roots

precisely identifiable, so the insertion of root VIs is a simple and straightforward

matching process. However, the nature of these indices is still a debatable issue where it

is claimed to be conceptual (Pfau, 2000), phonological (Borer, 2009, 2013), or simply

abstract (Harley, 2011).

These different views on how root morphemes are phonologically represented and

their implications on issues like root variations (e.g., raise/rise and go/went) will be

discussed in Chapter Four. Moreover, based on analysing Arabic data, it will be shown

that CSFs do not belong to roots, so they cannot play any role in determining their

phonological or semantic representations. As a result, the notion of root indices, whose

nature will be argued to be conceptual (Pfau, 2000), will be supported.

The insertion of VIs into functional morphemes, on the other hand, is determined

by the compatibility of the VIs with the features of the morphemes. However, given that

a functional morpheme can contain multiple features (e.g., AGR or fused morphemes),

there could be more than one VI matching all or some of these features. In such cases,

the relevant VIs start competing for insertion, but what will eventually be inserted is a

single VI. As to how the choice is made, Halle (1997:427) proposes the Subset

Principle35 as what governs the competition:

The phonological exponent of a Vocabulary item is inserted into a morpheme in the terminal string if the item matches all or a subset of the grammatical features specified in the terminal morpheme. Insertion does not take place if the Vocabulary item contains features not present in the morpheme. Where several Vocabulary items meet the conditions for insertion, the item matching the greatest number of features specified in the terminal morpheme must be chosen.

35 Which, according to Harley (2011), is a version of Kiparsky’s (1973) Elsewhere Condition. 48 Chapter Three – Theoretical Framework

According to the Subset Principle, for a VI to be eligible for insertion, it must match any

feature in the morpheme, but not correspond to any features that do not exist in the

morpheme. Multiple eligible VIs will trigger a competition, and the winning VI is the

one that corresponds to the greatest number of features, as stated above.

To illustrate how the Subset Principle governs competitions, Halle (1997) presents

an example from Russian. Subject-verb agreement in present tense in Russian involves

copying the person and number features of the subject onto the AGR morpheme of the

verbs. The different Russian VIs that can be inserted into this morpheme can be

presented as the follows:

(7) /mO/ ↔ [1st, Pl] nd /te/ ↔ [2 , Pl] /u/ ↔ [1st] / [+Pres] + ____ /šO/ ↔ [2nd] /tO/ ↔ elsewhere

When an AGR morpheme occurs in the environment of [+Pres], it will ultimately be

phonologically represented by one of the above five VIs depending on its features. If,

for example, the morpheme is specified as [2nd, Pl], the VIs /te/ and /šO/ will be in a

competition as each one of them matches at least a feature of the morpheme, which is

[2nd]. The VIs /mO/ and /u/ will be out of the competition as they match the feature [1st],

which is not present in the morpheme. As mentioned earlier, the elsewhere VI /tO/ will

be inserted as the last resort when none of the other four VIs can be inserted. According

to the Subset Principle, among the competing VIs, /te/ will be chosen as it matches

more features than its competitor /šO/. With insertion of the VIs corresponding to the

root and the [+Pres] morphemes, a verb such as the following can be produced:

(8) nes- é- te carry-Pres-2.Pl ‘You (all) carry.’

49 Chapter Three – Theoretical Framework

This type of phonological entry in (7) represents situations in which the relation

between VIs and morphemes can be described as a one-to-one relation, i.e., one VI for

each morpheme.36 In some other situations, however, there could be a one-to-many or a

many-to-one relation between VIs and morphemes. A one-to-many relation can be seen

in the English suffix -s that not only corresponds to the morpheme [Pl], but also the

morpheme [3rd, Sg, Pres] (Halle, 1997). We will encounter a similar situation in Arabic

in Chapter Four where such a relation will be considered an instance of homophony.

That is, a language can have two (or more) functionally different VIs that happen to be

phonologically identical.

As found in Marantz (1995), an example of a many-to-one relation between VIs

and morphemes can be seen in the distribution of the English suffixes -ity and -ness,

which presumably have the same function as the exponents of [n]. These suffixes can be

parts of the nouns rational-ity and purple-ness, but they are not interchangeable, e.g.,

*rational-ness and *purple-ity. Marantz (1995:405) states that -ity and -ness may have

different contextual features limiting their insertion in the environment of other VIs.

That is, in the environment of /ræʃənəl/ (rational), the insertion of the VI /əti/ (-ity) will

be permitted while the insertion of the VI /nəs/ (-ness) will be blocked. In the

environment where /pɜːrpəl/ (purple) is inserted, the insertion of /əti/ will be blocked

while the insertion of /nəs/ will be permitted.37

Insertion of VIs is not necessarily the last step before words can surface. DM

acknowledges what is known as post-insertion readjustment rules that, in certain

36 This should not be confused with the relation between VIs and the features of morphemes. In a one-to- one relation between a VI and a morpheme, the VI may correspond to the multiple features present in the morpheme, which makes the relation between the VI and the features a one-to-many relation. 37 A similar observation in Halle and Marantz (1993) involves the English VIs of the tense morpheme [Past], which can be /Ø/ (e.g., put), /t/ (e.g., dwelt) and /d/ (e.g., played). Halle and Marantz (1993:123) consider the choice of the inserted VI to be determined not only by the compatibility of the VI with the features of the morpheme (context-free competition, as in (7)), but also depending on its contextual features (context-dependent competition, or conditioned allomorphy).

50 Chapter Three – Theoretical Framework

contexts, change some phonological properties of the words. For example, Halle and

Marantz (1993) argue that among the three English VIs /Ø/, /t/ and /d/ that correspond

to the tense morpheme [Past], /t/ is inserted in the environment of /send/. However, such

affixation triggers a post-insertion readjustment rule that leads to a stem-internal change

by which the stem of the combination *send-t loses its final consonant, so the word can

surface as sent. Similarly, the affixation of /d/ to /sel/ (sell) and /tel/ (tell) results in a

stem-internal vowel change yielding the past verbs /sol-d/ (sold) and /tol-d/ (told)

instead of *sel-d and *tel-d (for more examples, see Halle and Marantz, 1993:128-129).

Observe that, in these instances, post-insertion readjustment rules are employed to

account for root alternations that cannot be accounted for by proposing a competition

between root VIs. The discussion in Chapter Four will shed more light on the different

approaches to this issue. In addition, it will be shown that post-insertion readjustment

rules in Arabic can apply to root VIs as well as VIs of functional morphemes.

3.6. The Encyclopaedia

In the view of DM, the Encyclopaedia is a component in which syntactic structures are

semantically interpreted. Being fed immediately from the Syntax, it comes naturally that

the post-syntactic modifications that structures may undergo in the Morphology are

invisible to the Encyclopaedia. That is, it is not relevant to the Encyclopaedia whether

[n] and [Pl] in oxen have been fused (Acquaviva, 2009) or not. Moreover, as mentioned

in Harley (2012:2152), the Encyclopaedia does not see whether [Pl] is phonologically

realised as -en or -s, which may indicate a lack of communication between the

Vocabulary and the Encyclopaedia.

However, in early DM, a link between the Vocabulary and the Encyclopaedia was

assumed to exist. Such a link appears to be established mainly to facilitate the process

of semantically interpreting roots. Given that roots, which probably are the most 51 Chapter Three – Theoretical Framework

semantically relevant constituents, have no distinct identities in the Encyclopaedia or at

any point prior to the insertion level, interpreting structures would need to be delayed

until roots become identifiable. This is possible only after the corresponding root VIs

have been inserted into root morphemes, which entails that the outputs of the

Vocabulary have a role to play in the Encyclopaedia (Marantz, 1995, 1996; Harley and

Noyer, 1999; Embick and Noyer, 2007). That is, retrieving the encyclopaedic entry that

corresponds to the canonical interpretation of cat or dog is made based on the choice

made in the Vocabulary between /kæt/ and /dɔg/. Prior to this point, what the

Encyclopaedia has is a standard indistinct root morpheme whose semantic interpretation

cannot be provided.

If one accepts the notion of root indices, however, such a link between PF and LF

can be abandoned since roots can now be identified in the Encyclopaedia by their

indices in the same way they are identified in the Vocabulary (Harley, 2011). Such a

view would reduce the complexity of the grammar as a whole, accord with the

separationist character of DM by exhibiting a full separation between the phonology

and the semantics of structures, and entail a near equal treatment of root and functional

morphemes in the Vocabulary and the Encyclopaedia. That is, each morpheme is now

distinguishable either by the features it contains (functional morpheme) or by its index

(root morpheme), and is provided with its canonical phonological and semantic

representations depending on its own properties. Nevertheless, as the canonical

phonological representations may change for contextual reasons in the Vocabulary,

special contextual meanings that differ from the canonical ones can be assigned to

structures in the Encyclopaedia, e.g., kick the bucket (for more on how phrasal idioms

are interpreted in the Encyclopaedia, see Marantz, 1996).

52 Chapter Three – Theoretical Framework

3.7. Concluding Remarks

Given the architecture and the principles of DM, Halle and Marantz (1994) list three

properties that, together, distinguish DM from other theories: Syntactic Hierarchical

Structure All the Way Down, Late Insertion and Underspecification. As opposed to

theories where word formation is pre-syntactic, all structures in DM, including the

structure of words, are syntactic. Although these structures can be modified in the

Morphology, these modifications are constrained, limited and do not result in major

deviations from the syntactically motivated configurations.

Late Insertion reflects the DM view that the phonological representations of

morphemes are provided late in the process. This is a view that obviously contrasts with

other theories where constituents can be handled in the Syntax while phonologically

pre-specified. In a related matter, Underspecification indicates that, at the insertion

level, VIs do not need to be fully specified for the morphemes into which they will be

inserted. As mentioned earlier, VIs are subject to insertion as long as they match any of

the features of the targeted morphemes and do not correspond to any feature that is not

present in the morphemes. This may allow more than one VI to be eligible for insertion,

which will consequently trigger a competition. Such a mechanism cannot exist in early

insertion theories in which the phonological representations of constituents are fully

specified for their features.

Having shown how DM addresses the process of word formation, most of the

discussions that follow will be concerned with how DM can account for the formation

of Arabic nouns. This is expected to contribute to the better understanding of the

morphosyntax of Arabic nouns, and to the refinement of DM. However, as previously

illustrated, the morphology of Arabic nouns is complex, so there are a number of

challenges in the application of DM. For the theory-related issues, subsequent chapters

53 Chapter Three – Theoretical Framework

will reveal that the CSFs of Arabic nouns do not belong to the roots, which bolsters the

notion of root indices. Furthermore, it will be shown that root indices may have a role to

play in the Morphology by triggering special rules leading to morphological

modifications. In addition to supporting the existing morphological operations of DM, a

new operation by which feature-changing is achieved will be introduced and defended.

54 Chapter Four – Roots and Nominal Stems

4. Chapter Four – Roots and Nominal Stems

4.1. Introduction

This chapter seeks to examine the nature of Arabic roots and nominal stems in the light

of the word formation mechanism provided by DM. Although examining these aspects

of Arabic appears to be a language-oriented goal, it cannot be achieved without

embarking on a number of theoretical issues from a cross-linguistic perspective.

Starting with roots, as mentioned in the previous chapter, DM takes root morphemes to

be devoid of syntactic, phonological or semantic features. They acquire the appropriate

phonological and semantic representations only in the Vocabulary and the

Encyclopaedia, respectively. At any point prior to this, root morphemes are simply

formally indistinct. Such a radical underspecification of root morphemes has raised a

fundamental question: on what basis do indistinct root morphemes receive distinct

representations in these components?

This chapter discusses two current approaches towards this issue. In the first

approach, it is argued that, in conjunction with the surrounding syntactic environment,

the CSFs of the roots are involved in determining their phonological representations,

while the semantic representations of the roots are determined by their phonological

ones (Marantz, 1995, 1996). The second approach takes roots to be individuated by

indices – albeit their nature is a heavily debatable matter – that make them accurately

recognisable in these components, especially in the Vocabulary (Pfau, 2000; Borer,

2009; Harley, 2011). Based on conceptual and empirical grounds, this chapter shows

that CSFs cannot be involved in the choice of root representations. Alternatively, it

supports the view that roots are identified by indices, which are likely to be conceptual

in nature as proposed by Pfau (2000). It also addresses the implications of adopting the

55 Chapter Four – Roots and Nominal Stems

conceptual individuation approach on the theoretical and empirical (cross-

linguistic/dialectal) levels.

In addition, based on data from Arabic nouns, this chapter supports the DM view

that root morphemes are category-less (acategorial) (see, e.g., Marantz, 1996, 1997,

2001; Harley and Noyer, 1998, 1999; Arad, 2003, 2005). This requires a discussion of

how Arabic roots acquire their categories and how nominal stems are formed.

Following Arad’s (2003, 2005) analysis of Hebrew, it will be proposed that Arabic

binyanim are the phonological representations of the category-assigning [n] morpheme.

The [n] morpheme, however, will be considered a functional set of different members

(Kihm, 2005) based on the different values of features each member has. An account for

how every member of [n] can be expressed by various binyanim and how a single

binyan can sometimes match different members of [n] is provided. Finally, this chapter

touches on the post-insertion process by which the phonological representations of root

and [n] morphemes interlock with each other yielding non-concatenative structures that

make Arabic, and Semitic languages in general, morphologically distinctive.

This chapter is organised as follows. Section 4.2 sketches out and discusses

Marantz’s (1995, 1996) view of how root representations are determined and how CSFs

are involved. Section 4.3 outlines the notion of root individuation and the different

proposals in which the nature of this individuation is considered conceptual (Pfau,

2000), phonological (Borer, 2009, 2013), and abstract (Harley, 2011). By elaborating on

some implications of the proposed phonological and abstract natures of individuation,

and by providing evidence from Arabic roots, Section 4.4 argues in favour of

conceptual individuation. Section 4.5 shows how conceptual individuation can account

for cross-dialectal/linguistic variations and where these variations exist. Section 4.6

investigates the categorisation of Arabic roots, the formation of Arabic nominal stems

56 Chapter Four – Roots and Nominal Stems

and the related phonological modifications they may undergo after the insertion. Section

4.7 concludes the chapter.

4.2. Roots and Compositional Semantic Features

Marantz (1996), which is mainly a discussion of Jackendoff’s (1994, 1995) view of

idiomatic meanings, makes reference to two types of semantic information, namely:

compositional semantic features (CSFs) and non-compositional semantic meanings. As

mentioned earlier, CSFs include [±Animate], [±Count], [±Common], and [±Human]

(Chomsky, 1965). As will be shown soon, Marantz (1995, 1996) takes these features to

be present in the Syntax and involved in the choice of the appropriate root VIs in the

Vocabulary. Non-compositional semantic meanings, on the other hand, are considered a

property of the Encyclopaedia. They are simply the encyclopaedic knowledge that is

irrelevant to the Syntax, e.g., cat ↔ furry domestic feline (Marantz, 1996:5).

Departing from the latter type of semantic information, according to Marantz

(1996), the choice of the appropriate meaning of a structure in the Encyclopaedia is

made in accordance with (a) the VI inserted into its root morpheme in the Vocabulary,

and (b) the surrounding syntactic environment. Considering the first factor, this view

entails that the inserted root VIs serve as input to the Encyclopaedia (Marantz, 1996:3).

Thus the use of the Encyclopedia to interpret LF semantically must involve knowledge of Vocabulary insertion at PF, where a choice between “dog” and “cat” may have been made.

Hence, for example, after the root VI /kæt/ has been chosen for insertion into a root

morpheme, the Encyclopaedia provides the canonical non-compositional semantic

meaning associated with it. The surrounding syntactic environment (the second factor)

usually comes into play when the structure is a phrasal idiom, e.g., kick the bucket,

where a special contextual non-compositional meaning is provided instead (for more on

how idioms are interpreted in the Encyclopaedia see Marantz, 1996:9-10). This

57 Chapter Four – Roots and Nominal Stems

mechanism per se may well explain how roots obtain their semantic interpretations, but

it does not explain how the choice of root VIs is made in the first place.

As for the CSFs, Marantz (1995, 1996) discusses their location and function.

According to Marantz (1995:403), CSFs are present in the Syntax and, in conjunction

with the syntactic features of the structure, they can condition the insertion of VIs:

The syntactic and compositional semantic features that condition the insertion of Vocabulary items at PF are taken to be involved in the grammar qua computational system while the idiosyncratic properties of “words” are not.

In such a broad statement, neither the morphemes that carry these features into the

Syntax nor the VIs whose insertion is co-conditioned by them are precisely specified.

Specifying the carrier of the CSFs will determine what VI insertion they co-condition in

the Vocabulary. Marantz (1996), however, establishes a link between the VIs of roots

and the CSFs, as shown below:

The choice of “cat” as opposed to “dog” (and as opposed to any of the numerous other Vocabulary items that might be inserted into something like an “animate count noun” node from the syntax) must be registered as part of the derivation subject to semantic interpretation with the help of the Encyclopedia. (Marantz, 1996:3)

As far as the licensing structure of grammar is concerned, all we need to know about “cat” is that a particular bunch of phonology can be associated with a syntactic node with the features, singular count noun (perhaps also, animate). (Marantz, 1996:8)

The syntactic nodes that Marantz refers to as containing CSFs are the ones into which

/kæt/ and /dɔg/ are inserted, which are merely the root morphemes (see Harley, 2011:4

for a similar understanding of Marantz's view). With respect to the function, as shown

above, Marantz (1995:403) posits that CSFs along with the syntactic features of the

structures can condition the insertion of VIs at PF. Given the location of CSFs, the

insertion referred to here is clearly the insertion of root VIs into root morphemes.

Marantz’s view of the location of CSFs and how roots are semantically interpreted can

be presented as the following:

58 Chapter Four – Roots and Nominal Stems

(1) Narrow Lexicon [√[+Count,+Animate]]

Syntax PF LF

Vocabulary Encyclopaedia

/kæt/ “furry domestic feline”

The insertion of root VIs may not be the last step in giving roots their

phonological representations. Root VIs may undergo some phonological readjustments

after the insertion. For example, the English rise and raise are not taken to be two

different root VIs. Rather, Marantz (1997:13) states that raise is a special pronunciation

of rise in a certain context, i.e., a transitive context created by a special version of the

functional head v (termed as v-1) that projects an agent. That is, rise is inserted into the

root morpheme and, then, a post-insertion readjustment rule changes it into raise (also

see Harley and Noyer, 1999:5). This clearly obviates any competition between root VIs

in the Vocabulary.

The rejection of root VI competition in this view is mainly based on the absence

of any root identity. To illustrate the basis of this stance, Marantz (1996:17) states that

if a competition between root VIs does exist and if, hypothetically, /hawnd/ is the plural

form of /dɔg/ (dog), /hawnd/ would not only block the insertion of /dɔg/ in plural

contexts, it will also block the insertion of /kæt/ and any similar root VI. That is,

/hawnd/ will be inserted into any [√[+Animate, +Count]] in the presence of [+Pl], even if the

root morpheme is supposed to be occupied by /kæt/. Thus, according to this view,

/hawnd/, as an allomorph of /dɔg/, is created by readjusting /dɔg/ in plural

environments. This becomes possible only when the root morpheme obtains a unique

59 Chapter Four – Roots and Nominal Stems

identity by the insertion of /dɔg/ into it. Prior to this point, the identity of the root is

simply indiscernible.

The link that Marantz (1995, 1996) establishes between CSFs and the insertion of

root VIs has been criticised for not providing a solid explanation as to why a certain

root VI is inserted into a root morpheme whose CSFs and surrounding syntactic features

can be perfectly matched by many other VIs (see, e.g., Pfau, 2000; Harley, 2011). If

CSFs are a property of root morphemes, they would at best minimise the number of root

VIs eligible for insertion, but they would never play any role in making the choice. That

is, if a root morpheme is [√[+Animate, +Count]], root VIs such as /kæt/ and /dɔg/, among

many others, will be equally eligible for insertion; any other root VIs, e.g., /desk/, will

be excluded. When it comes to the insertion, this theory stops short from providing an

explanation of how the final choice between the eligible /kæt/ and /dɔg/ is made. Thus,

CSFs would neither be sufficient to individuate root morphemes nor able to condition

the insertion of root VIs into them.

The proposed location of CSFs might have been based on analysing English

nouns such as cat and dog, as they have been repeatedly mentioned in Marantz’s works.

One can immediately understand that these nouns refer to entities that are animate,

nonhuman, and count. Given that /kæt/ and /dɔg/ are root VIs, it comes naturally that

CSFs are located in the root morphemes into which these VIs are inserted. However,

this cannot be the case with Arabic roots. Consider the following examples:

(2) Nouns Derived from the Root /ktb/:

a. /ktb/ + /C1iC2aːC3/ = kitaːb- (book)

b. /ktb/ + /C1aːC2iC3/ = kaːtib- (writer)

60 Chapter Four – Roots and Nominal Stems

(3) Nouns Derived from the Root /bħӨ/:

a. /bħӨ/ + /C1aC2C3/ = baħӨ- (research)

b. /bħӨ/ + /C1aːC2iC3/ = baːħiӨ- (researcher)

(4) Nouns Derived from the Root /lʕb/:

a. /lʕb/ + /C1aC2iC3/ = laʕib- (playing)

b. /lʕb/ + /C1aːC2iC3/ = laːʕib- (player)

It can be seen in (2.a), (3.a), and (4.a) that the roots /ktb/, /bħӨ/, and /lʕb/ respectively

can be parts of nominal stems with nonhuman38 references. However, as seen in (2.b),

(3.b), and (4.b) above, the same roots can also be parts of stems with human references.

What determines whether the above nouns have human or nonhuman references is their

binyanim rather than the roots. 39

If /ktb/, /bħӨ/, and /lʕb/ are the VIs that encode root morphemes, as argued in this

study, it is obvious that these morphemes cannot be the location of CSFs. Thus, in

contrast with Marantz, CSFs cannot constrain or condition the insertion of root VIs in

Arabic. They cannot be involved in the insertion of root VIs not because there are too

many root VIs matching every combination of CSFs, as argued by Pfau (2000) and

Harley (2011). Rather, it is because CSFs do not belong to root morphemes in the first

place. If CSFs are a property of root morphemes, every root will be predicted to

participate in either human or nonhuman structures, but not in both. Arabic nouns

cannot be identified as count/mass, in-/animate, non-/human, or common/proper by the

consonants of their roots. Although Arabic roots alone may convey some primitive

meanings, as will be shown in Section 4.4, they cannot convey comprehensible

denotations or referents that can be associated with certain values of CSFs. What is

38 Humanness is an essential CSF in Arabic nouns. The way nouns, especially when plural, agree with verbs, adjective, and pronouns depends on the value of this CSF. This issue will be discussed at greater length in the next chapter. 39 Besides being the carriers of CSFs, binyanim have different functions that will be discussed soon in this chapter. 61 Chapter Four – Roots and Nominal Stems

important here is that CSFs cannot be a property of root morphemes, so they cannot be

involved in the insertion of their VIs. As to their location, it will be argued in Section

4.6 that CSFs are a property of the category-assigning [n] morphemes.

4.3. Root Individuations

Given that CSFs and syntactic features cannot decisively determine what phonological

forms and, consequently, what semantic interpretations root morphemes should take, it

becomes imperative to explain how the choices are made, especially in the Vocabulary.

The most prominent approach to this issue is root individuation, which was firstly

introduced by Pfau (2000) and pursued by many authors afterwards.40 Basically, root

individuation proposes that every root is made unique in the Narrow Lexicon by an

index that demonstrates its identity. Consequently, this individuation allows the

Vocabulary to precisely identify the root in order to provide it with the appropriate

phonological form and, as argued by Harley (2011), it also allows the Encyclopaedia to

supply the appropriate semantic interpretations.

However, there is a lively debate about the nature of this individuation. There are

essentially three positions. Pfau (2000) argues that the individuation of roots is

conceptual, Borer (2009, 2013) argues that it is phonological, while Harley (2011)

argues that the individuation can neither be conceptual nor phonological. Alternatively,

she proposes that root indices are abstract. The following subsections discuss the

arguments for these positions and show their motivations. The compatibility of these

positions to the present study, especially with the nature of Arabic roots, will be

addressed in Section 4.4.

40 In order to mark the difference that Pfau’s proposal made within DM, Siddiqi (2006, 2009) terms the traditional view of the theory as the pre-Pfau (model of) DM. 62 Chapter Four – Roots and Nominal Stems

4.3.1. Conceptual Individuation

Pfau (2000) assumes that the semantic difference between, for example, cat and dog is

present at an early stage of the grammar. He claims that each root in the Narrow

Lexicon is equipped with an index that individuates it throughout the process by

showing the conceptual features of the root. His claim is based on empirical evidence

obtained from investigating spontaneous speech errors in English and German. In his

investigation, Pfau (2000) identifies two types of speech errors, namely: meaning-based

substitutions and anticipations/perseverations of semantic features, which are both

semantically motivated. Consider the following examples:

(5) Meaning-Based Substitutions (Pfau, 2000:115):

a. hast du einen Radiergummi da ← einen Spitzer have you an eraser there ← a pencil.sharpener

b. in welcher Höhe, äh, Tiefe haben sie gegraben in what height, er, depth have they dug

c. damit kommst du auf keinen grünen Baum ← grünen Zweig with.that come you on no green tree ← green branch ‘With that you’ll never get anywhere.’

d. ich habe keine Tafel mehr ← keine Kreide I have no blackboard more ← no chalk

It can be seen above that the desired nouns are wrongly substituted with semantically

related ones. For example, the desired nouns Spitzer (pencil sharpener) and Kreide

(chalk), in (5.a) and (5.d), are replaced with their co-hyponyms Radiergummi (eraser)

and Tafel (blackboard) respectively. Tiefe (depth), in (5.b), is replaced with its antonym

Höhe (height). Finally, Zweig (branch), in (5.c), is replaced with Baum (tree) where

both have a part-whole relationship.

The other type of error is slightly different. It involves replacing the desired word

with a word that has no semantic relation to it, but is related to another word that is

present in the context. Consider the following examples:

63 Chapter Four – Roots and Nominal Stems

(6) Semantic Anticipations/Perseverations (Pfau, 2000:119):

a. ich wollte den Hund anbellen ← anbinden I wanted the dog bark.at ← tie.up

b. ich hätte meine Bohne mit Karotten und Erbsen mitbringen können ← meine Dose I have.COND my bean with carrots and peas bring can ← my can ‘I could have brought my can with carrots and peas.’

c. they even fly on the wing ← sleep on the wing (Harley, 1984:201)

d. a branch falling on the tree ← on the roof (Fromkin, 1973:262)

In (6.a) above, it is shown that anbinden (tie up) is replaced with anbellen (bark at).

Unlike the examples in (5), the words have no semantic relations to which the

substitution can be attributed. The undesired presence of the latter word is rather

motivated by its semantic relation to Hund (dog). The same can justify the error in (6.b)

where Dose (can) is replaced with Bohne (bean) due to the semantic relation that the

latter has with Karotten (carrots) and Erbsen (peas). In (6.c), sleep is replaced with fly

due to its relation with wing. In (6.d), replacing roof with tree is due to the relation that

the latter has with branch.

Based on this analysis, Pfau (2000) claims that roots are individuated by indices

that encode their conceptual features. Conceptually related roots are linked to each other

in the Narrow Lexicon in a way that expresses the nature of the relations, e.g., “X is Y”,

“X is the opposite of Y”, “X is part of Y”, and so on. Hence, when a speech error

occurs, the activation of the desired root in the Narrow Lexicon, during a stage termed

the Conceptual Preparation, spreads to a conceptually related root resulting in drawing

it into the Syntax. The intruding root may replace the desired one, as shown in (5), or

replace a different root in the context to which it has no relation, as shown in (6). In

both cases, the undesired root enters the Syntax and becomes part of the structure. In the

Syntax, indices cannot make any contributions. All root morphemes are the same to the

Syntax; thus, no syntactic rules/principles are driven by the conceptual identity of the

64 Chapter Four – Roots and Nominal Stems

root. Post-syntactically, the Vocabulary identifies the root by its index, and the

corresponding root VI is inserted despite it being the wrong root. That is, speech errors

of these sorts occur at the first step of the process, which is drawing root morphemes

from the Narrow Lexicon into the Syntax. If this is the case, these speech errors are not

the result of inserting the wrong root VI into the right root morpheme. Rather, they

result from inserting the right root VI into the wrong root morpheme.

Since each root is now uniquely distinct and individuated by its index, the

insertion of root VI becomes a straightforward process. The Vocabulary contains root

VIs, and each VI matches a certain root. At the moment of the insertion, the Vocabulary

precisely identifies the root, with the help of its index, and the insertion becomes a

matter of choosing the VI that appropriately corresponds to it.41 As to the implication of

this proposal in the Encyclopaedia, it can be hypothesised that, in contrast with Marantz

(1995, 1996), the Encyclopaedia does not need to wait for the insertion to happen in

order to provide the roots with their non-compositional semantic meanings. Rather, the

Encyclopaedia identifies the root by its index (as the Vocabulary does) and supplies the

meaning that is relevant to its index and the syntactic environment. If this is the case,

root VIs do not need to serve as the input to the Encyclopaedia as traditionally thought

in DM. The modification that Pfau (2000) proposes for DM can basically be seen as

assigning a unique identity, of a conceptual nature, to each root embedded in its index.

Determining the phonological representation in the Vocabulary (and perhaps the

semantic meanings in the Encyclopaedia) is guided by the identity of the root. Thus,

according to Pfau (2000), what is retrieved from the Narrow Lexicon can be represented

41 Pfau (2000) does not make any claim about competition amongst root VIs. The competition that is repeatedly mentioned in his analysis is claimed to take place in the Narrow Lexicon due to the activation of two conceptually related roots when speech errors occur. However, based on Pfau’s proposal, a claim that root VIs can compete was introduced to account for root alternations (Harley, 2011), as will be discussed in this chapter. 65 Chapter Four – Roots and Nominal Stems

as [√(CAT)] rather than just [√ROOT], where the notation attached to the √ symbol is a

label referring to the index that is carried by the root.

4.3.2. Phonological Individuation

Borer (2005a, 2005b) claims that roots (designated as listemes) are pairings of sounds

and meanings. That is, a listeme is a sealed package of conceptual properties that is

stamped with a unique identifying phonological index. Listemes, as such, constitute

what is termed the Conceptual Array from which they are drawn to be embedded into

structures. Listemes receive the appropriate phonological representations based on the

phonological indices they bear, and their interpretations based on their conceptual

contents (for more, see Borer, 2005a:11-15). This proposal contrasts with the traditional

view of DM in that roots are individuated, and differs from the proposal of Pfau (2000)

in that the phonology of the root is involved in its individuation.

Nonetheless, later Borer (2009, 2013) abandons the term listeme and the notion

that roots have conceptual properties, but maintains that roots are phonologically

indexed. This conversion is motivated by the fact that different English words with

different meanings share one phonological unit that, itself, has no meaning in isolation,

as shown below (Borer, 2013:383):

Table 4.1: Prefix-Stem English Verbs with Common Phonological Units MIT DUCE SUME CEIVE SCRIBE42 e-mit re-duce re-sume re-ceive sub-scribe sub-mit in-duce sub-sume con-ceive de-scribe per-mit de-duce pre-sume de-ceive pre-scribe trans-mit pro-duce con-sume per-ceive trans-scribe

Given that the above common units are phonologically consistent, but conceptually

empty, Borer (2013) claims that they are the realisations of certain phonological

42 The author notes that although scribe does have a meaning alone with transcribe being closely (and describe being potentially) related to this meaning, it is hardly the case in subscribe and ascribe. 66 Chapter Four – Roots and Nominal Stems

properties that have no conceptual contents. That is, from a DM perspective, mit, for

example, is inserted based on the phonological index of the root whose conceptual

properties are determined according to the environment in which it occurs.

Moreover, according to Borer (2013), the phonological properties associated with

each unit above do not necessarily guarantee a rigid phonological form in every

occurrence. Rather, since each unit can take a slightly different, but still consistent,

phonological form in different environments, as shown below, it can be assumed that

these phonological properties serve as guiding features (indices) that ensure a certain

degree of a phonological consistency for each unit across all possible occurrences.

Compare the following examples to their counterparts in Table 4.1 (Borer, 2013:383):

Table 4.2: Prefix-Stem English Nouns with Common Phonological Units MIT DUCE SUME CEIVE SCRIBE e-mis re-duct re-sumpt re-cept sub-script sub-mis in-duct sub-sumpt con-cept de-script per-mis de-duct pre-sumpt de-cept pre-script trans-mis pro-duct con-sumpt per-cept trans-script

In Borer’s view, it is clear that mit and mis, for example, are different forms of an entity

that bears nothing but a phonological index. The index guarantees the insertion of a

specific phonological exponent that is determined by the environment of the occurrence.

Any conceptual properties that can be matched with this entity must not be attributed to

this index. For a further illustration, Borer (2013:386) assumes that for a root such as

π√BANK,43 there exists only one phonological representation that is given to the root –

with the help of its index – regardless of any conceptual content that might be matched

to the root later, be it roughly SHORE or FINANCIAL INSTITUTION.

In addition, in what can be related to our discussion of Arabic roots, Borer

(2013:384-385) discusses some examples from Semitic languages, specifically Arabic

43 π designates the phonological property of the root. 67 Chapter Four – Roots and Nominal Stems

and Hebrew. She provides examples of different words derived from the Arabic root

π√KTB as well as the Hebrew roots π√XŠB and π√PQD. Although Borer (2013) does

not explicitly rule out any conceptual relation between the Arabic words katab (write),

hiktib (dictate) and miktab (letter),44 she states that it might be hard to find a relation

between the Hebrew words xiššeb (calculate) and hexšib (esteem) or, more clearly,

between mipqad (census) and mipqada (army HQ).

Finally, a major implication of phonological individuation, as stated by Borer

(2013:398-399), is that alternations such as go/went, good/better/best, and

bad/worse/worst – which are radically phonologically distinct as compared to mit/mis,

duce/duct, and sume/sumpt – are different realisations of different phonological

roots/indices:

If, e.g., /πgo/ and /πwent/ must be characterized as spellouts of the same root, then the inevitable conclusion is that roots cannot be in any explanatory sense phonological indices. The same, of course, goes for the comparative/superlative paradigms.

That is, /goʊ/ (go) and /wɛnt/ (went) are considered two different root VIs inserted into

two different root morphemes without the need for a post-insertion readjustment, as

discussed in Section 4.2 earlier, or for a competition between root VIs (Harley, 2011),

as will be discussed in Section 4.3.3 below.

4.3.3. Abstract Individuation

Harley (2011) acknowledges that roots need to be individuated in the Narrow Lexicon

in order to receive the appropriate VIs in the Vocabulary as well as the appropriate

meanings in the Encyclopaedia. However, she argues that the individuation is rather

abstract. Her argument is based on rejecting the proposed conceptual (Pfau, 2000) and

44 The Arabic examples provided by Borer (2013:384) are not accurate. katab should be the equivalent of wrote rather than write or correspondent. In addition, hiktib (dictate), hitkatteb (correspond), miktab (letter), and makteba (desk) would be more accurate if they were presented as ʔuktub (you write!), (n)- ukaːtib ((we) correspond), maktuːb (written/letter), and maktab (desk), respectively. However, these technical errors should not affect Borer’s (2013:384) statement that “[it] might just be sufficiently robust to find it tempting to say that, e.g., π√KTB has some conceptual Content to the effect that it is related to writing.” 68 Chapter Four – Roots and Nominal Stems

phonological (Borer, 2009, 2013) nature of indices.

Part of Harley’s argument – specifically, the part that rejects phonological

individuation – is based on the assumption that root alternations are neither resulting

from phonological post-insertion readjustments (Marantz, 1997), nor from the insertion

of two different root VIs into two different root morphemes (Borer, 2009, 2013).

Alternatively, Harley (2011:6-9) argues that such alternations are better accounted for

as competition-driven. That is, with roots being individuated in the Narrow Lexicon, the

hypothetical /hawnd/ and /dɔg/ alternation (Marantz, 1996) are two different root VIs

that coexist in the Vocabulary of English. Both VIs correspond to one specific root

morpheme that is well identified in the Vocabulary by its index. At the moment of the

insertion, /hawnd/ and /dɔg/ will be the only competing root VIs, whereas /kæt/ and any

other VI will be excluded as they do not match the identity of the root under question. If

this is the case, /hawnd/, as the most specified VI, wins the competition only if [+Pl] is

present in the structure, while /dɔg/ wins elsewhere. This can be equally applied to

actual alternations such as rise/raise (Marantz, 1997) and go/went (Borer, 2009, 2013)

(for more, also see Siddiqi, 2006, 2009; Chung, 2009).

Harley (2011:10) rejects the proposal that roots are phonologically individuated

for two reasons. Firstly, as a supporter of root VI competition, Harley argues that this

competition proves that root morphemes are subject to late insertion, which is a major

DM principle. Thus, assigning phonological properties to roots prior to the insertion

would simply be a violation of this principle and, subsequently, a rejection of the

competition altogether, which is exactly what Borer (2013:398-399) claims about this

type of competition. Secondly, phonological individuation, as it takes go and went as

different phonological representations of different root morphemes, focuses on the

phonological differences between the words at the expense of the semantic similarities

69 Chapter Four – Roots and Nominal Stems

they have. It creates a sharp distinction between roots based on their phonological forms

ignoring their mutual semantic distribution, especially in idioms as in, for example, go

around the bend vs. went around the bend.

As to conceptual individuation, in an argument that is similar to Borer’s (2009;

2013:383), Harley (2011:13) provides the following examples of roots that cannot be

semantically interpreted outside their contexts.45

(7) a. kit and caboodle “everything”

b. run the gamut “includes a whole range”

c. in cahoots “conspiring”

d. high jinks “mischief”

According to Harley (2011), no interpretations to the bolded words above can be given

without consulting the whole contexts in which they occur. Speakers learn the meaning

of each expression as a whole without having any knowledge about the individual

meanings of these words. As opposed to cat, which can convey a meaning by itself,

caboodle, gamut, cahoots and jinks are meaningless when taken out of the above

contexts. Therefore, since there is no concept behind these words in isolation, and one

does not know how to interpret them outside their contexts, Harley (2011) argues that

root indices cannot be of a conceptual nature.

Hence, according to this view, there is no such a root as [√(CAT)] (Pfau, 2000) – or

presumably [π√CAT] (Borer, 2013) – in the Narrow Lexicon. Alternatively, in an

attempt to emphasise the abstractness of the individuation, Harley (2011) proposes that

a unique numerical notation should be given to the index of each root, e.g., [√239].

Moreover, Harley (2011) makes it clear that based on these indices, the Vocabulary and

the Encyclopaedia provide the appropriate representations to the roots. That is, indices,

45 Harley (2011:11-14) provides examples of roots with multiple concepts, e.g., the Hebrew root √kv∫ from which the semantically distinct kvi∫ (highway) and kava∫ (pickle) can be derived. These examples were skipped as similar ones, e.g., √XŠB and √PQD (Borer, 2013), had been discussed earlier. 70 Chapter Four – Roots and Nominal Stems

as such, link each root to a certain set of instructions in each component on how the root

should be phonologically and semantically represented.

4.4. In Favour of Conceptual Individuation

The inability of CSFs (even if they belong to root morphemes) and the surrounding

syntactic features to totally condition the insertion of root VIs has given more

popularity to the notion of root individuation within DM (regardless of the nature of the

individuation). What makes individuation even more important for Arabic roots is their

lack of CSFs, as shown in Section 4.2. As long as roots are not fully phonologically and

semantically specified, and the principles/rules of the Syntax cannot be influenced by

the indices of the roots, root individuation should not be problematic for DM whether it

is conceptual (Pfau, 2000), phonological (Borer, 2009, 2013), or abstract (Harley,

2011).

Although assigning indices to roots might be an undesirable enrichment for some

linguists, it is, however, important and inevitable. Without indices, a plausible

explanation as to how the Vocabulary and the Encyclopaedia precisely identify roots

and give them the appropriate representations becomes difficult to reach. If we, in

principle, agree that roots are individuated by indices, the nature of these indices, which

is still a debatable issue, becomes a secondary but still important point of discussion.

Although the recent attempts to explore the nature of root indices are significant, they

might still be far from conclusive. In fact, extensive diachronic and cross-linguistic

studies are still required in order for a much stronger conclusion to be reached.

Meanwhile, this study contributes to the ongoing debate about the nature of root

individuation by supporting the proposal that such individuation is conceptual (Pfau,

2000). It will be shown that phonological individuation (Borer, 2009, 2013) cannot be

compatible with DM, and abstract individuation (Harley, 2011) is probably an 71 Chapter Four – Roots and Nominal Stems

overgeneralisation that stands heavily on limited and isolated cases, i.e., conceptually

empty roots, that can be accounted for within DM independently and within the

proposal of conceptual individuation. Examples showing that the conceptual identity of

Arabic roots is real, and an account for roots that seem to have multiple unrelated

concepts will be provided.

4.4.1. DM and Phonological Individuation

Recall that phonological individuation, as introduced by Borer (2009, 2013), entails that

radically phonologically distinct root alternations, e.g., go/went, are two different

morphemes. Besides the fact that this proposal disallows root VI competition and

ignores the mutual semantic distribution the roots have, as argued by Harley (2011), it

distinguishes roots not only by their phonological forms, but also by their grammatical

features. Taking π√GO and π√WENT (using Borer’s notations) as two distinct root

morphemes simply means that, at least, the latter is inherently specified for [+Past].

Such a proposal cannot be brought into the realm of DM since it is a clear violation.

DM takes roots to be devoid of any grammatical features, including grammatical

categories such as [n] and [v], as will be discussed at greater length in the next section,

and tense of course. Moreover, having a root morpheme that is specified for tense

beside a separate [+Past] morpheme in one structure is nothing but a redundancy.

In DM approaches, where /goʊ/ (go) and /wɛnt/ (went) are considered different

representations of a single morpheme, there is no room for such a violation to occur.

The grammatical (tense) distinction between the words is left to the Syntax to create

(independently from the roots) by adding the appropriate tense morpheme to the

structure, whereas the phonological distinction is simply the responsibility of the

Vocabulary to create. The phonological distinction is achieved either by post-insertion

readjustment (Marantz, 1997) or by a competition between root VIs (Harley, 2011),

72 Chapter Four – Roots and Nominal Stems

which is still a debatable issue in DM. In both cases, the final phonological form is

determined in accordance with the syntactically selected tense morpheme. However,

with two separate phonologically pre-specified roots, as argued by Borer, none of these

mechanisms would be possible.

4.4.2. Conceptually Empty Roots

It was shown earlier that the proposal of abstract individuation (Harley, 2011) is based

on rejecting both phonological and conceptual individuations. Although this study

agrees with Harley (2011) in rejecting phonological individuation, it does not agree with

rejecting the conceptual version. Recall that the arguments made by Borer (2009, 2013)

and Harley (2011) against conceptual individuation are motivated by the existence of

roots that cannot be semantically interpreted outside their contexts. These roots are

either (a) conceptually empty, e.g., caboodle, gamut, cahoots, and jinks, as will be

discussed in this subsection, or (b) able to have multiple conceptually unrelated

interpretations, e.g., mipqad (census) vs. mipqada (army HQ) from the Hebrew root

√PQD, bank vs. bank in English (Borer, 2013:384-385), and similarly kvi∫ (highway) vs.

kava∫ (pickle) from the Hebrew root √kv∫ (Harley, 2011:11), as will be discussed along

with some Arabic examples in the next subsection.

Conceptually empty roots can be a real challenge to the notion of conceptual

individuation only if they form a widely spread phenomenon. Words of the caboodle-

type may not exist in every language, especially in Semitic languages, and when they do

exist in a language, e.g., in English, they apparently represent a minority. One can come

up with a very long list of English roots that have meanings in isolation, but when it

comes to conceptually empty roots, the list will be significantly limited and shorter. It is

a radical overgeneralisation to say that all roots are conceptually empty because

caboodle, gamut, cahoots, and jinks (or even -mit, -duce, -sume, and -ceive, as shown in

73 Chapter Four – Roots and Nominal Stems

Borer’s analysis) are empty. If an explanation for these empty roots needs to be

provided, it must conform with and must not ignore the fact that the majority of roots

are conceptually real. Ignoring the concepts that the majority of roots have in isolation,

as well as how roots are conceptually related to each other, as shown in Pfau (2000), is

like ignoring the mutual semantic distribution that go and went have. That is exactly

what Harley (2011:10) herself considers a drawback of Borer’s assumption that go and

went are two different roots. In fact, it is counterintuitive to the proposed abstractness of

root indices to assume that go and went are different representations of a single root that

has certain semantic properties ensuring identical distribution of these representations.

In addition, since the proposal of conceptual individuation is based on

semantically motivated speech errors that show real conceptual relations between roots,

any counter argument must provide an alternative explanation that effectively accounts

for these relations. If indices are abstract or phonological, what would be the alternative

explanation that accounts for a speech error where, for example, fly on the wing is

produced instead of sleep on the wing, as shown in (6.c)? If fly is not mistakenly used

instead of sleep because of the conceptual relation it has with wing, what would be the

alternative motivation for this error? The conceptual relation between fly and wing is too

obvious to be dismissed, and neither phonological individuation nor abstract

individuation can provide an alternative explanation.

As mentioned earlier, any attempt to account for conceptually empty roots must

not come at the expense of the conceptually defined ones. Thus, this study assumes that

conceptually empty roots have indices, like any other root, in order to be precisely

identified by the Vocabulary and, consequently, given the appropriate phonological

form. However, the indices of these specific and special roots are assumed to be

conceptually empty, i.e., defective. That is, they serve as identity tags that work well in

74 Chapter Four – Roots and Nominal Stems

the Vocabulary, but are useless in the Encyclopaedia. When indices of this sort enter the

Encyclopaedia, the Encyclopaedia searches for a canonical interpretation to provide as

it always does with any root, such as √CAT. However, in this case, the Encyclopaedia

finds no match, so it consults the whole structure as it also does with any phrasal idiom.

The difference here is that consulting the whole structure in the case of kit and caboodle

is triggered by the defectiveness of the index of caboodle, whereas consulting the

structure of a phrasal idiom is motivated by the existence of a listed special non-

compositional meaning that matches the structure. If this is on the right track, empty

roots of the caboodle-type are better seen as the exceptions that prove the rule rather

than undermine it.

4.4.3. Roots with Multiple Concepts

Before addressing the issue of roots that seem to have multiple unrelated concepts, it is

important to establish that Arabic roots show a great deal of consistency with the notion

that root indices are conceptual. As shown earlier, Arabic roots are consonantal and

cannot be pronounced or become completely comprehensible in isolation. In order for

roots to emerge as fully comprehensible, stems need to be formed. This is achieved by

interlocking the roots with the appropriate binyanim, as shown in the figure below:

Figure 4.1: Arabic Stems Derived from /ktb/, /lʕb/, and /ʕlm/

kitaːb- kaːtib- luʕb-ah ʕilm- book knowledge writer -alʕab- toy ʕaːlim- kaːtab- play corresponded scholar ʕalim- m-aktab- laːʕib- knew ktb office lʕb player ʕlm m-aktuːb- written -uʕalim- katab- laʕib- m-alʕab- teach -aʕlam- -aktub- wrote playing stadium know write

The above examples show how different words can be derived from each of the Arabic

roots /ktb/, /lʕb/, and /ʕlm/. It is obvious that the words derived from the root /ktb/ share

75 Chapter Four – Roots and Nominal Stems

a primitive meaning that can be related to the concept of writing, the words derived

from /lʕb/ have a primitive meaning of playing, and the words of /ʕlm/ are related to the

concept of knowledge. In each group above, the source of the primitive meaning the

words share is the root, and cannot be the binyanim. Some binyanim can occur across

the groups without affecting the primitive meanings assigned by the roots. For example,

the binyan /C1aːC2iC3/ can combine with each of the above roots, yielding kaːtib

(writer), laːʕib (player), and ʕaːlim (scholar), without interfering with the primitive

meaning of each root. Another examples can be seen in maktab (office) vs. malʕab

(stadium/playground), -alʕab- (play) vs. -aʕlam- (know), and laʕib (playing/played) vs.

ʕalim (knew).

Given that the sources of the primitive meanings are the roots that, in turn, are the

VIs inserted into root morphemes equipped with indices, these indices must contain

conceptual information from which these primitive meanings are generated. The

conceptual similarities between these words are obvious and cannot be ignored.46 Any

explanation purporting to deny that these indices have conceptual contents would be

implying that the conceptual similarities between the words are simply coincidental.

As mentioned earlier, despite the fact that Borer (2009, 2013) and Harley (2011)

do not agree with each other about the nature of root individuation, they both disagree

with Pfau’s (2000) claim that the individuation is conceptual. Their argument against

conceptual individuation is based on the existence of conceptually empty roots, which

was discussed earlier, and the existence of multiple concepts of some roots, e.g. bank

(SHORE) vs. bank (FINANCIAL INSTITUTION), and the Hebrew mipqad (census)

vs. mipqada (army HQ). As a matter of fact, roots with multiple concepts do exist in

46 Although Borer (2013) argues against conceptual contents of roots, she could not deny these contents in Arabic. Recall Borer’s (2013:384) statement that: “[it] might just be sufficiently robust to find it tempting to say that, e.g., π√KTB has some conceptual Content to the effect that it is related to writing”. As just mentioned, this study takes these conceptual properties as part of the indices attached to the roots. 76 Chapter Four – Roots and Nominal Stems

Arabic too. Compare the following examples to the words shown in the Figure 4.1

above:

(8) a. katiːb-at- (battalion)

b. luʕaːb- (saliva)

c. ʕalam- (flag)

The examples in (8) above are words that are made up of the roots /ktb/, /lʕb/, and /ʕlm/

that were previously claimed to be related to the concepts of writing, playing, and

knowledge respectively. However, the above words do not seem to have any relation to

these concepts. The only common thing between the roots of the words in (8) and the

roots of the words in Figure 4.1 is the phonological form. Borer’s treatment of similar

cases entails that katiːb-at- (battalion) and kitaːb- (book), for example, are derived from

one phonologically, and only phonologically, individuated root that can be matched

later with different concepts.

However, given the incompatibility of phonological individuation with DM as

well as the strength of Pfau’s (2000) evidence of conceptual individuation, this study

assumes that katiːb-at- (battalion) and kitaːb- (book) – and, consequently, bank vs. bank

and mipqad (census) vs. mipqada (army HQ) – are derived from two different

conceptually individuated roots that happen to receive identical phonological

representations later.47 That is, the words in (8) are considered to be derived from the

48 roots [√(ktb)], [√(lʕb)], and [√(ʕlm)] respectively, but these roots are different from the

roots from which the words in Figure 4.1 are derived.

For more illustration, by using more conventional conceptual notations, the roots

47 Extensive diachronic studies may reveal that katiːb-at- (battalion) and kitaːb (book) were conceptually related at some point. If this is not the case, the only option we have is to assume that they are derived from different conceptual roots that happen to have identical phonological representations. 48 Since Latin letters are not used in the writing system of Arabic, the indices of Arabic roots are designated with subscripted phonological symbols, e.g., [√(ktb)] and [√(ʕlm)], instead of subscripted upper- case letters of the written forms of the target words, e.g., [√(CAT)]. This should not be taken as a phonological individuation. 77 Chapter Four – Roots and Nominal Stems

in Figure 4.1 can be represented as [√(WRITING)], [√(PLAYING)], and [√(KNOWLEDGE)],

whereas the roots in (8) can be represented as [√(BATTALION)], [√(SALIVA)], and [√(FLAG)].

In other languages, the phonological representations of the first set of roots would be

completely different from the phonological representations of the roots of the second

set. However, in the Vocabulary of Arabic, it happens that the phonological

representation of each root from the first set is identical to the phonological

representation of a root from the second set. Cross-linguistic and cross-dialectal

variations of conceptual roots will be discussed briefly in the next section.

Obviously, identical phonological representations for distinct morphemes are not

unusual. This phenomenon can easily be seen in Arabic (as will be shown in Section

4.6), English, and in many other languages. Thus, consider the suffixes of the following

English examples:

(9) a. teach/teacher vs. nice/nicer

b. book/books vs. I walk/she walks

As shown above, /ər/ (-er) can be a nominal suffix (teach vs. teacher) and a marker for

comparative adjectives (nice vs. nicer), and /s/ can be a plural marker, and a marker for

the third person singular in the present tense. If we sort things based on their

phonological forms, we may end up claiming that /ər/ in teacher and /ər/ in nicer

represents a single morpheme that has no specific functional properties, but matched

with two distinct functions at a later stage of the formation process. This is as similar

and as inadequate as Borer’s claim that bank (SHORE) and bank (FINANCIAL

INSTITUITION) – and similarly katiːb-at- (battalion) and kitaːb (book) – are derived

from a single root morpheme that has no specific conceptual properties, but is matched

with two different concepts later.

78 Chapter Four – Roots and Nominal Stems

4.5. Cross-Dialectal and Cross-Linguistic Variations

Given that the Vocabulary is a language-specific component and, according to Harley

(2011), it contains sets of instructions on how roots should be phonologically

represented, it follows naturally that cross-dialectal variations in the representation of

roots that have the same conceptual features, i.e. indices, are due to the different sets of

instructions the Vocabulary of each dialect has. The data collected from SA and the

Egyptian, Jordanian, and Saudi dialects using Mayer’s (1969) picture-based story shows

that, in some cases, a certain entity was referred by the participants using the same

noun. For instance, in all these varieties of Arabic, the noun kalb was used to refer to a

dog, ∫aʒar to refer to trees, and naħl to refer to bees. Assuming that these nouns are

made from the root morphemes [√(klb)], [√(∫ʒr)], and [√(nħl)] respectively, it can be said

that the Vocabulary of each variety of the language has the same sets of instructions on

how these specific roots should be phonologically represented.

In some other cases, the participants employed different nouns to refer to identical

entities. For example, a jar has appeared six times in the story, as shown below:

Figure 4.2: The Recurrence of Jar in the Story

79 Chapter Four – Roots and Nominal Stems

The jar was the centre of an incident in which the dog fell from the window with its

head stuck in the jar. In describing the incident, every participant referred to the jar with

a different noun, as shown below:

(10) a. zuʒaːʒ-ah (SA)

b. bartʕamaːn (Egyptian)

c. gatʕramiːz (Jordanian)

d. ʕilb-ah (Saudi)

The above nouns refer to one entity/concept, but with different phonological

representations. As far as roots are concerned, it can be assumed that these nouns are

derived from one root with a certain index, i.e., the same conceptual root. In the

Encyclopaedia, this root is associated with a semantic interpretation that might be

something like: cylindrical glass vessel with a wide mouth and used as a container.

However, in the Vocabulary, each variety of the language has its own instructions for

how this root is phonologically represented. As a result, a different root VI will be

inserted into this root morpheme in each variety. Given that the same semantic

interpretation might also be available in the Encyclopaedia of English, it can be said the

root morpheme that is phonologically expressed with /dʒɑːr/ (jar) in English is the same

root morpheme from which the above Arabic nouns are derived.

This assumption entails that the differences between roots across

languages/dialects lie in the different sets of phonological instructions available in the

Vocabulary of each language/dialect. This view is, to a large extent, consistent with the

traditional view of DM, which takes root to be indistinct in the Narrow Lexicon, but

made distinct later in the Vocabulary. On the other hand, it contrasts with the Lexicalist

Hypothesis in which roots are phonologically and semantically specified in the Lexicon,

making it the place where languages/dialects differ from each other. However, given

80 Chapter Four – Roots and Nominal Stems

that roots are now taken to be conceptually individuated, the generalisation that cross-

dialectal/linguistic variations come from the Vocabulary can be true only if all

languages/dialects share the same conceptual roots, which is not always the case. For

example, the concept of river that is phonologically represented with /rɪvər/ in English

and /nahr/ in Arabic has two conceptually and phonologically different forms in French,

namely: rivière (a river that flows into another river) and fleuve (a river that flows into

an ocean/sea).49 Although it might be true that the English /rɪvər/ and the Arabic /nahr/

are the ultimate phonological representations of one mutual conceptual root, it cannot be

true that the same conceptual root exists in the Narrow Lexicon of French. This is

indicative that root differences across languages/dialects do not solely lie in the

Vocabulary of each language/dialect. Languages/dialects can also have conceptual

variations in the indices of their roots in the Narrow Lexicon leading to semantic

variations in the Encyclopaedia as well as phonological variations in the Vocabulary.

4.6. Root Categorisation

In earlier linguistic frameworks, e.g., the Lexicalist Hypothesis and X-bar Theory

(Chomsky, 1970; Jackendoff, 1977), a noun such as growth is taken to be derived from

the pre-existing verb grow, in what is known as (deverbal) nominalisation. Besides

implying roots are phonologically and semantically specified, this view also suggests

that roots are pre-categorised in the Lexicon. To the contrary, Marantz (1997:9) argues

that roots are never verbs or nouns by themselves at any stage, i.e., they are acategorial.

What determines whether a root is categorised as a noun or a verb is the syntactic

environment in which it occurs. That is, [√(GROW)], as an acategorial root, becomes a

verb when merged with a [v] morpheme, and a noun when merged with a [n]

morpheme, i.e., merged with a category-assigning morpheme, as shown below (for

49 Thanks to Hugues Peters for this example. 81 Chapter Four – Roots and Nominal Stems

more, see, e.g., Marantz, 1996, 2001; Harley and Noyer, 1998, 1999; Arad, 2003, 2005;

Harley, 2005; Embick and Noyer, 2007; Embick and Marantz, 2008; Lowenstamm,

2008):

(11) grow = vP growth = nP

[√(GROW)] [v] [√(GROW)] [n]

/ɡrəʊ/ /Ø/ /ɡrəʊ/ /Ө/

When it comes to this notion of acategorial roots and category-assigning

functional morphemes, concatenative languages, especially English, can sometimes be

misleading. This notion creates a formal syntactic distinction between roots, e.g.,

[√(GROW)], and stems, e.g., [[√(GROW)] v] and [[√(GROW)] n]. However, on the surface

level, this distinction is unnoticed sometimes since grow the root is identical to grow the

verbal stem due to the lack of an overt marker for [v]. More ambiguity can be seen in

homophones such as hammer and to hammer. Although there is a distinction in their

underlying structures (i.e., [[√(HAMMER)] n] vs. [[√(HAMMER)] v] respectively), this

distinction does not show up on the surface since both category-assigning morphemes

have no overt phonological representations.

(12) hammer = nP to hammer = vP

[√(HAMMER)] [n] [√(HAMMER)] [v]

/hæmər/ /Ø/ /hæmər/ /Ø/

As compared to the roots of concatenative languages, Arabic roots and stems can

serve as more explicit examples supporting the notion that roots are acategorial (see

Marantz, 2001). As mentioned earlier, Arabic roots are consonantal and cannot stand

alone as functioning and meaningful elements, so they interlock with binyanim in order

82 Chapter Four – Roots and Nominal Stems

to form stems. It is possible for one set of consonants constituting a root to interlock

with different binyanim to form several stems with different categories and

interpretations. A root like /ktb/, for example, can take part in different nominal and

verbal structures, as shown below:

(13) a. /ktb/ + /C1iC2aːC3/ = kitaːb- (book)

b. /ktb/ + /C1aːC2iC3/ = kaːtib- (writer)

c. /ktb/ + /aC1C2uC3/ = -aktub- (write)

d. /ktb/ + /C1aC2aC3/ = katab- (wrote)

In the above structures, it can be seen that the root /ktb/ is the same, its consonants are

always in the same sequence, and, obviously, it is not exclusively associated with nouns

or verbs. Given that /ktb/ is the VI inserted into the root morpheme, this morpheme is,

consequently, not specified for any grammatical category on its own. What determines

the category of each of the above words is its binyan rather than the root: the binyanim

in (13.a-b) are nominal and the binyanim in (13.c-d) are verbal.

Following some analyses of the morphology of Hebrew, specifically those of

Arad (2003, 2005), this study assumes that Arabic binyanim are the VIs that are inserted

into the category-assigning morphemes. That is, as far as the above nouns are

concerned, the binyanim /C1iC2aːC3/ and /C1aːC2iC3/ are the VIs inserted into the

category-assigning [n] morphemes. Moreover, given that CSFs are always expressed by

the binyanim of nouns (so CSFs are not a property of roots as discussed in Section 4.2),

and given that those binyanim are the VIs inserted into [n], this study assumes that

CSFs are a property of the [n] morphemes that also serve as the category-assigning

morphemes for the acategorial roots, as shown below:

83 Chapter Four – Roots and Nominal Stems

(14) kitaːb (book) = nP kaːtib (writer) = nP

[√(ktb)] [n[-Hum]] [√(ktb)] [n[+Hum]]

/ktb/ /C1iC2aːC3/ /ktb/ /C1aːC2iC3/

The table below shows a variety of nominal binyanim and how they categorise roots.

Although these binyanim are quite productive, they are of course not the only binyanim

available in Arabic:

Table 4.3: Nominal Binyanim and Root Categorisation in Arabic Binyan Root Word /ħrs/ ħaːris- (guard)

/C1aːC2iC3/ /qtl/ qaːtil- (killer) /ʕml/ ʕaːmil- (worker) /tʕbx/ tʕabbaːx- (cook/chef) 50 /C1aC2C2aːC3/ /ħml/ ħammaːl- (carrier) /tʕbl/ tʕabbaːl- (drummer) /∫rb/ ∫araːb- (drink) ʕ ʕ /C1aC2aːC3/ /t ʕm/ t aʕaːm- (food) /sʕbħ/ sʕabaːħ- (morning) /ʒbl/ ʒabal- (mountain) ʕ ʕ /C1aC2aC3/ /χd b/ χad ab- (anger) /ʕml/ ʕamal- (work)

There are two things about nominal binyanim that can be noticed: (a) /C1aC2aC3/ can

serve as a nominal binyan, as shown in the above table, as well as a verbal binyan, as

shown in (13.d), at the same time; (b) there are different binyanim that can serve as the

VIs inserted into [n].

The first observation, which includes but is not limited to /C1aC2aC3/, can be

accounted for simply by assuming that there are two phonologically similar copies of

50 This binyan has four empty slots for three consonants. When a root interlocks with it, the middle root consonant is doubled in what is known as gemination, as will be discussed later in this section. 84 Chapter Four – Roots and Nominal Stems

/C1aC2aC3/ (homophony) in the Vocabulary of Arabic with a different function for each

copy. It is not unusual in many languages that different functional morphemes are

expressed by identical VIs. For example, /Ø/ can assign nominal and verbal categories

to English roots (hammer vs. to hammer, as discussed earlier), -er is a nominal suffix

(teach vs. teacher) and a marker for comparative adjectives (nice vs. nicer), and -s is a

plural marker (book vs. books) and a marker for the third-person singular present tense

(I walk vs. he walks), as discussed earlier. Since the Vocabulary is a language-specific

component, it should not be problematic that it contains VIs that have the same form but

serve as the exponents of different functional morphemes.

The second observation is similar to the issue of how the English suffixes -ity and

-ness are distributed, i.e., why rational-ity and purple-ness but not *rational-ness and

*purple-ity. In Chapter Three, it was shown that Marantz’s (1995:405) account for this

phenomenon involves the assumption that VIs have contextual features according to

which their insertion is limited to the environments of other VIs. That is, in the

environment of rational, -ity will be permitted and -ness will not, while -ness will be

permitted in the environment of purple and -ity will not. If these VIs are considered

different exponents of [n], this view entails there being only one [n] for which these VIs

(perhaps among others) are competing. The competition ends with the selection of the

VI whose contextual features match the environment in which the terminal [n]

morpheme occurs.

However, given that CSFs are a property of [n] and they can appear in different

values, i.e., [±Animate], [±Count], [±Common], and [±Human], it does not seem

plausible to assume that there is only one [n] for which all the nominal binyanim of

Arabic are competing. Therefore, as also in Kihm’s (2005:18) analysis of Spanish

nouns, this study assumes that [n] is a functional set that contains different members.

85 Chapter Four – Roots and Nominal Stems

These members in Arabic, however, are distinguished by the different values of the

CSFs they carry, as exemplified in (14) above. In the Vocabulary of Arabic, each

nominal binyan matches a certain member of [n] and serves as the VI that expresses (a)

the category the morpheme assigns to the root and (b) the CSFs it carries.

Marantz’s (1995:406) hypothesis about contextual features of VIs can come into

play only if there is a member of [n] that can be expressed by more than one nominal

binyan, which is actually the case in Arabic. For example, it can be seen in the above

table that the nominal binyanim /C1aːC2iC3/ and /C1aC2C2aːC3/ can give a human

reference to the roots they combine with. This makes each of these binyanim a perfect

match to a member of [n] that is specified as [n[+Hum]]. The binyan /C1aːC2iC3/ can

combine with /ħrs/ to form the human noun ħaːris- (guard), but it cannot combine with

the root /tʕbl/ as this will produce *tʕaːbil, which has no meaning as a noun. The same

ʕ thing applies to the binyan /C1aC2C2aːC3/ that can combine with /t bl/ to produce the

noun tʕabbaːl- (drummer), but cannot combine with /ħrs/ as this will produce *ħarraːs,

which is not meaningful as a noun. As compared to -ity and -ness, it can be said that

although /C1aːC2iC3/ and /C1aC2C3aːC4/ can function as nominal human binyanim, they

are not interchangeable since each binyan has its own contextual features that limit its

insertion into [n[+Hum]] in the environment of other VIs. That is, according to the above

table, /C1aːC2iC3/ wins the competition in the environment of /ħrs/, /qtl/, or /ʕml/, and

ʕ ʕ loses to /C1aC2C2aːC3/ in the environment of /t bx/, /ħml/, or /t bl/.

Up to this point, root VIs (the consonants) and binyanim are two separate

phonological entities inserted into two different morphemes. They need to combine with

each other in order to generate pronounceable and meaningful nominal stems. In

contrast with concatenative languages where the VIs inserted into the root and category-

assigning morphemes are usually ready to surface, these VIs in Arabic are not. Arabic

86 Chapter Four – Roots and Nominal Stems

root VIs need to be mapped onto the binyanim (VIs of [n] morphemes) in order to

surface, as shown below:

(15) /k t b/ /k t b/

/C1 i C2 aː C3/ /C1 aː C2 i C3/

k i t aː b- (Book) k aː t i b- (Writer)

This is an entirely phonological process that only takes place after the insertion of the

VIs. It aims to readjust/reorder the elements of the VIs in a way that is consistent with

the non-concatenative nature of the language. This can be looked at as a post-insertion

readjustment rule by which the mapping is performed.51 Since no root VI or binyan

appears to be able to escape this rule, it must be registered as permanently active rule in

the language. That is, the mapping rule does not require special environments to be

activated. It operates constantly on every VI inserted into root morphemes and on every

binyan inserted into [n] morphemes.

There are some cases in Arabic where a binyan comes with four slots among

which the middle ones are both reserved for the middle consonant of the roots. In such

cases, the middle consonant is doubled, or geminated. 52 Consider the following

examples:

Table 4.4: Root Gemination in Nominal Stems Root Non-Geminated Nouns Geminated Nouns kaːtib- kuttaːb- /ktb/ Writer (Male) Writers tʕaːlib- tʕullaːb- /tʕlb/ Student (Male) Students tʕabx- tʕabbaːx- /tʕbx/ Cooking Chef ħaml- ħammaːl- /ħml/ Carrying/Pregnancy Carrier

51 For more on mapping in Arabic, see McCarthy (1981). 52 For more on germination in Arabic, see McCarthy (1981), Ratcliffe (1997), Hassan (2002), and Ryding (2005). 87 Chapter Four – Roots and Nominal Stems

It can be seen above that each pair of the geminated/non-geminated stems has the same

primitive meaning, but different binyanim and one extra copy of C2 in the geminated

form. Changing the binyan can result in pluralisation, as seen in the first two nouns, or

in creating a different noun with a slightly different denotation, as seen in the last two

nouns. It is obvious that these changes are obtained by changing the binyanim rather

than the roots. That is, the above nouns, whether geminated or not, are all made from

/ktb/, /tʕlb/, /tʕbx/, and /ħml/, which are the only root VIs in the language that match the

ʕ ʕ indices of the root morphemes [√(ktb)], [√(t lb)], [√(t bx)], and [√(ħml)] respectively.

Regardless of what [n] morpheme is available in the structure, each root VI above is

inserted uncontested.

Obviously, geminating the middle consonant of the above root VIs is a post-

insertion process that is motivated by the binyanim and probably performed as part of

the mapping rule. When the number of slots the binyan has equals the number of the

consonants available in the root VI, gemination does not occur. As a result, mapping the

root VI onto the binyan becomes a one-to-one process after which the stem can surface,

as shown in (15). However, when gemination occurs, the slots of the binyan outnumber

the consonants of the root VI, so the middle consonant spreads in order for an

appropriate stem to be formed, as shown below:

(16) /k t b/

/C1 u C2 C2 aː C3/

k u t t aː b

Whether spreading the middle consonant is part of the mapping rule or performed by a

special gemination rule, it must be considered the result of a post-insertion process

rather than the result of inserting a different root VI with two identical middle

consonants.

88 Chapter Four – Roots and Nominal Stems

4.7. Conclusion

In this chapter, it has been shown that Arabic roots are not restricted to specific

grammatical categories or CSFs. They can freely participate in nominal and verbal

structures as well as in human and nonhuman nouns. Hence, Arabic roots were

considered acategorial (consistent with the architecture of DM) and CSF-free.

Therefore, CSFs cannot play any role in determining the representations of roots, which

is a view that contrasts with Marantz (1995, 1996). This gives importance to the notion

that roots are individuated by indices that guide the choice in the Vocabulary and the

Encyclopaedia. Furthermore, given the shortcomings of the proposed phonological

(Borer, 2009, 2013) and abstract (Harley, 2011) natures of root individuation, it was

shown that the identity of Arabic roots is conceptual. This lends support to the proposal

that root individuation is conceptually based (Pfau, 2000).

Given that roots are conceptual in nature, it was shown that it is not always

necessarily the case that root variations across languages/dialects are attributed to the

different phonological forms available in the Vocabulary of each language/dialect. In

some cases, languages can differ by having conceptual roots in their Narrow Lexicons

that do not exist in other languages.

With respect to the formation of nominal stems in Arabic, it was illustrated that

since roots are acategorial, roots gain their category and become part of nominal

structures only when a category-assigning [n] morpheme is present. The phonological

representations of the [n] morphemes were shown to be the binyanim that interlock with

the root VIs forming the stems. Binyanim can come in different forms with different

CSFs, which indicates that the [n] morphemes they encode are the location of the CSFs

of the nouns. Given that CSFs can have several values, [n] was considered a functional

set of different members (Kihm, 2005). Each member has a different combination of the

89 Chapter Four – Roots and Nominal Stems

possible values of CSFs. When several binyanim match a certain member of [n], which

is usually the case, it was shown that the insertion becomes a matter of selecting the

binyan whose contextual features (Marantz, 1995) match the VI inserted into the root

morpheme.

Interlocking the inserted binyan with the inserted root VI was considered a post-

insertion phonological process in which the root VI is mapped on the binyan. It is

simply a process that involves distributing the root consonants among the designated

slots of the binyan. In some cases, however, a binyan contains two slots that are

designated for the middle consonant of the root VI, so the consonant is geminated. The

overall outcomes of the process are nominal stems whose gender, number, case, and

definiteness features are the topics of the next chapters.

90 Chapter Five – Gender and Number

5. Chapter Five – Gender and Number

5.1. Introduction

This chapter is mainly concerned with the gender and number features of Arabic nouns.

Within the framework of DM, it investigates how these features are handled throughout

the word formation process, especially in the Morphology component. This

investigation will potentially provide us with a better understanding of the Morphology

and how it might be organised.

There are different views about the position of gender and number cross-

linguistically. For example, Ritter (1993) argues that gender in some Romance

languages is a feature of Num(ber) that may move to N, while gender in Hebrew is

generated as a feature of N. Pfau (2002, 2007) argues that gender in German is a

property of the roots, while Picallo (1991) proposes that gender and number in Catalan

nouns are independent features that constitute syntactic projections. Although the

authors propose different positions for gender and number, they all agree that these

features are syntactic, i.e., they figure in the Syntax. Picallo (2008) makes it clear that

gender, in particular, cannot be considered a feature of a post-syntactically added

(dissociated) morpheme due to its intervention in LF.

Gender and number are probably the most dynamic features in Arabic. In this

chapter, it will be shown that across the spectrum of different types of Arabic nouns,

gender and number may occur as independent features, interact with each other, or

interact with other features, including [n]. Features with such a high level of flexibility

motivate the assumption that they are generated as independent features. Hence,

similarly to the proposal of Picallo (1991), this study assumes that gender and number

are generated as two independent features under two syntactic projections: GendP and

91 Chapter Five – Gender and Number

NumP. In the Morphology, this independence may or may not be disrupted depending

on the syntactic environment in which they occur. Moreover, this study assumes that

Arabic nominal structures are fully specified for these features, especially gender. That

is, gender morphemes are specified as masculine [-F] or feminine [+F], while number

morphemes are singular [Sg], dual [Dl], or plural [Pl]. Although some of these

morphemes have no phonological representations in the language, their presence,

especially [-F], in the structures is essential to the Morphology at times, as will be

shown in relation to nonhuman plurals in Section 5.4.

Gender seems to behave consistently in SA and DA, while number is more

dynamic in SA than in DA. Specifically, number is associated with case in SA dual and

SndM plural nouns, while it is independent in these nouns in DA due to the absence of

case marking. Since case is involved, number in SA duals and SndM plurals will be

addressed in the next chapter.

This chapter is organised as follows. Section 5.2 investigates how gender and

number are morphologically treated and phonologically represented in singular nouns.

Building on the previous discussion of nominal stems, this section suggests positions

for the GendP and NumP in the internal structure of N. Evidence to support the

suggested positions will be provided throughout the discussion in the subsequent

sections. The section, also, addresses the phonological exponents of gender and number

in singular nouns, and accounts for the absence of the usual feminine marker in covert

feminine nouns. Section 5.3 focuses on gender in SA and DA dual nouns, as well as

number in the DA ones. It will be shown that, in comparison to singular nouns, gender

morphemes receive the same phonological representations in the Vocabulary, while the

number morpheme of DA dual nouns receives a different representation by virtue of the

new value it bears.

92 Chapter Five – Gender and Number

Due to the complexity and diversity that plural nouns exhibit with respect to

gender and number, they will be analysed across different sections. Section 5.4 aims to

account for why nonhuman plurals are treated as singular feminine in terms of

agreement, in a phenomenon known as deflected agreement (Ferguson, 1989, and many

others). It will be suggested that the singular treatment is a result of a deflection in the

agreement system, while the feminine treatment is a consequence of gender changing. A

new morphological operation by which gender changing can be achieved will be

proposed.

Plural nouns have two forms: sound (masculine ‘SndM’ and feminine ‘SndF’) and

broken. The formation of sound plurals is discussed in Sections 5.5. in which the gender

feature of SA and DA SndM plurals and the number feature of DA SndM plurals will be

addressed. The section will also investigate the interaction between gender and number

in SndF plurals in both varieties. Section 5.6 focuses on the formation of broken plurals

by showing that, in contrast with Noyer (1992), the number feature of these nouns is not

pre-syntactically configured. In an alternative analysis, it will be shown that the

structural distinction between sound and broken plurals is created in the Morphology

instead. In the light of the proposed word formation strategies of sound and broken

plurals, Section 5.7 concentrates on establishing which strategy is the default in the

language.

Finally, throughout the chapter, a number of morphological modifications to the

structure of Arabic nouns with respect to gender and number will be assumed. These

modifications will be collectively examined in Section 5.8 as an attempt to explore how

the Morphology is internally organised. It will be suggested that the Morphology of

Arabic is divided into several phases across which the morphological operations are

carried out. Section 5.9 concludes the chapter and highlights its findings.

93 Chapter Five – Gender and Number

5.2. Singular Nouns

Arabic singulars are simply nominal stems (root consonants and binyanim) that do not

take a number suffix, but can take a gender suffix only when feminine. Thus, at the

surface level, singular nouns are considered the base forms from which dual and plural

nouns are derived in various ways. This section focuses on providing a DM account for

the gender and number features of singular nouns, which mostly behave in the same

way in SA and DA.

It is generally assumed that the grammatical gender of Arabic nouns that refer to

humans and living creatures reflects the natural gender of the referents, whereas the

grammatical gender of inanimate nouns is arbitrarily assigned as in many other

languages, as mentioned in Chapter Two. Regardless of the basis of the assignment,

singular nouns are not overtly marked for gender when masculine, but are followed by

the suffix -at when feminine with the exception of a few nouns. Based on this, singular

nouns will be divided into three types with respect to gender: masculine, overt feminine,

and covert feminine.

Section 5.2.1 examines gender in both masculine and overt feminine nouns, while

Section 5.2.2 tackles the absence of the feminine suffix in the covert feminine ones.

Finally, Section 5.2.3 discusses the number feature of singular nouns.

5.2.1. Gender in Masculine and Overt Feminine Nouns

When a singular noun is masculine, whether it is referring to an animate or inanimate

entity, it does not take an overt gender marker. Since there are some feminine nouns that

are not overtly marked as well, a simple test to discover the gender of the nouns is to

observe the behaviour of the surrounding elements that exhibit gender agreement.

Masculine nouns will, then, be distinguished from covert feminine nouns by the

masculine form of agreement they motivate. For example, Arabic singular (proximity) 94 Chapter Five – Gender and Number

demonstratives come in two forms in terms of gender: the masculine form haːðaː and

the feminine form haːðihi (Ryding, 2005:315). Naturally, masculine nouns take the

former and cannot take the latter, as shown below:

(1) a. haːðaː al-tʕaːlib- / *haːðihi al-tʕaːlib- this.M.Sg Def-student this.F.Sg Def-student ‘This (male) student.’

b. haːðaː kalb- / *haːðihi kalb- this.M.Sg dog this.F.Sg dog ‘This is a (male) dog.’

c. haːðaː al-kitaːb- / *haːðihi al-kitaːb- this.M.Sg Def-book this.F.Sg Def-book ‘This book.’

Assuming that a gender morpheme specified as [-F] is part of the underlying structure

of masculine nouns,53 one possibility would be that the morpheme is embedded in the

underlying structure of the stems, which is [[√(index)] n[CSF]]. That is, gender is hosted

either by the root or by the category-assigning [n] morpheme, which already is the

location of CSFs as proposed in the previous chapter. Since the roots of the above nouns

can also be parts of feminine structures, as will be shown soon, roots do not appear to

have any gender restrictions, so they cannot be the host. This leaves us with the option

that if gender is part of the stems, [n] would be its host, i.e., [[√(index)] n[CSF, -F]].

However, If [n] shows no capability of hosting gender, the inevitable conclusion would

be that gender is a syntactically independent morpheme, i.e., [[[√(index)] n[CSF]] -F], and

this is what singular feminine nouns effectively demonstrate.

Most Arabic singular feminine nouns differ from the masculine ones only by

taking the suffix -at, immediately after the stem. Consequently, when a singular

demonstrative – or any other agreeing element – is present, it exhibits feminine

agreement, as shown below:

53 The discussion of nonhuman plurals in Section 5.4 will support the assumption that masculine nouns are syntactically specified for gender. 95 Chapter Five – Gender and Number

(2) a. haːðihi al-tʕaːlib-at- / *haːðaː al-tʕaːlib-at- this.F.Sg Def-student-F this.M.Sg Def-student-F ‘This (female) student.’

b. haːðihi al-kalb-at- / *haːðaː al-kalb-at- this.F.Sg Def-dog-F this.M.Sg Def-dog-F ‘This (female) dog.’

c. haːðihi al-ziyaːr-at- / *haːðaː al-ziyaːr-at- this.F.Sg Def-visit-F this.M.Sg Def-visit-F ‘This visit.’

Observe that the binyanim of the masculine animate nouns in (1.a-b) are the same

binyanim as their feminine counterparts in (2.a-b), which is a clear indication that these

binyanim are genderless and the gender of the feminine nouns is expressed by the suffix

-at. Having the feminine suffix in a position that is external to the nominal stems is an

indication that the suffix is the phonological representation of an independent gender

morpheme that is specified as [+F] in the underlying structure of the nouns, i.e.,

[[[√(index)] n[CSF]] +F].

Building on the structure of the nominal stems discussed in the previous chapter, a

general structure for Arabic singulars with respect to their stems and gender can be

represented as the following:

(3) GendP

nP [Gend]

[√(index)] [n[CSF]]

The [Gend] morpheme above can be specified either as [-F] to mark the masculine, or

as [+F] to mark the feminine. Hence, when nominal syntactic structures arrive in the

Vocabulary, this study assumes that their gender morphemes are supplied with /Ø/

when specified as [-F], but supplied with the VI /at/ when specified as [+F], as shown

below:

96 Chapter Five – Gender and Number

(4) /at/ ↔ [+F] /Ø/ ↔ elsewhere54

It is important to point out that the above entries only cover gender morphemes that are

independent. Gender morphemes remain independent in singular nouns, as well as in

dual and SndM plural nouns, as will be shown in this chapter. The other types of nouns

are instances of interactions between gender and other features that yield a competition

between different VIs.

There are two aspects of the feminine suffix -at that need to be explained. Firstly,

when the suffix appears on a feminine noun that has a masculine counterpart, i.e., some

human and animate nouns, removing the suffix usually results in changing the gender

into masculine, as can be seen by comparing (2.a-b) to (1.a-b). On the other hand,

removing the suffix from inanimate feminine nouns, e.g., the noun shown in (2.c), is not

possible as it makes the resulting form uninterpretable. Removing the feminine suffix

does not reverse the gender feature back to masculine, probably because gender is an

arbitrary property of these nouns. Thus, the suffix becomes a sort of compulsory and

immovable addition to the stems. Whether -at is detachable or not, it remains the VI that

gives [+F] its phonological representation in the Vocabulary.

Secondly, although the suffix is widely taken to be represented as -at following its

typical pronunciation, it is sometimes pronounced as /ah/ or /a/, which is termed the

pause form by Ryding (2005:121). As the name suggests, the suffix takes this form

when followed by a pause, i.e., not followed by another sound. In SA feminine nouns,

the typical form -at is more common since it is usually followed by a case suffix. In

DA, on the other hand, the pause form is more common as the suffix usually falls in a

word-final position due to the lack of case marking in the dialects, as will be discussed

in Chapter Six. This study takes /at/ as the standard and the only VI available for

54 Up to this point of the analysis, the elsewhere environment corresponds to the morpheme [-F]. 97 Chapter Five – Gender and Number

insertion in the Vocabulary. The change of the inserted /at/ into the pause form is

considered the result of a post-insertion readjustment process that is performed in

accordance with the phonological environment in which the inserted /at/ occurs.

5.2.2. Gender in Covert Feminine Nouns

As compared to overt feminine nouns, the covert ones are simply the nouns that behave

and agree as feminine without having the suffix -at, as shown below with the nouns

corresponding to English ‘desert’, ‘remembrance’, ‘war’, ‘mother’, and ‘donkey’:55

(5) a. al-sʕaħraːʔ- al-ʕarabiy-at- / *al-ʕarabiy Def-desert.F Def-Arabian-F Def-Arabian ‘The Arabian Desert.’

b. al-ðikraː al-sanawiy-at- / *al-sanawiy Def-remembrance.F Def-annual-F Def-annual ‘The anniversary.’

c. al-ħarb- al-ʕaːlamiy-at- / *al-ʕaːlamiy Def-war.F Def-global-F Def-global ‘The World War.’

d. al-ʔumm- al-mariːdʕ-at- / *al-mariːdʕ Def-mother.F Def-sick-F Def-sick ‘The sick mother.’

e. al-ʔataːn- al-barriy-at- / *al-barriy Def-donkey.F Def-wild-F Def-wild ‘The wild (female) donkey.’

If the above feminine nouns have their gender morphemes specified as [+F], these

morphemes are expressed by /Ø/, which usually expresses [-F], instead of the standard

VI /at/. This gives rise to the question: why is the elsewhere VI used to correspond to a

gender morpheme that is perfectly matched by another VI?

This study assumes that all feminine structures, whether they later surface as overt

or covert feminine nouns, are syntactically equipped with [+F]. The presence of this

55 In a few cases, nouns of female animates have different stems than their masculine counterparts. For example, the masculine counterparts of the nouns in (5.d-e) are ʔabb (father) and himaːr (male donkey) respectively. Feminine nouns of this sort will be treated as covert feminine. 98 Chapter Five – Gender and Number

morpheme allows adjectives – and any other agreeing expressions – to agree with these

nouns in gender by having [+F] in their structures as well. This agreement is probably

achieved by copying the gender feature of the nouns onto the agreement node of the

adjectives in the Morphology (Halle and Marantz, 1993:115). If the nominal structure is

to eventually surface as an overt feminine noun, [+F] remains active until the structure

enters the Vocabulary, in which case /at/ will be inserted into it as well as into the [+F]

morpheme of the agreeing adjective.

If, on the other hand, the feminine noun is meant to surface as covert feminine,

the [+F] morpheme of the nominal structure is modified in the Morphology in a way

that prevents the insertion of /at/ into it in the Vocabulary. This modification, by which

the distinction between overt and covert feminine structures is created, takes the form of

impoverishing [+F], as shown below:

(6) [+F] → Ø

Hence, when a modified feminine structure arrives in the Vocabulary, the insertion of

/at/ into its [+F] morpheme is avoided. The Impoverishment rule above does not

necessarily operate on the [+F] morphemes of the adjectives allowing the VI /at/ to be

inserted into them. This is an indication that this morphological modification takes

place after the agreement level in the Morphology.56

There is no clear reason in Arabic as to why covert feminine nouns lose their

feminine suffix. In the underlying structure of these nouns, there is no special feature by

which one can predict that a certain structure should surface as covert or overt feminine

noun. More technically, no one can associate the activation of the rule by which [+F] is

impoverished in covert feminine nouns with any specific feature available in the

underlying structure of these nouns. As shown in the above examples, covert feminine

56 Not all morphological modifications are performed after the agreement level in the Morphology. The following discussions will provide an insight into the Morphology that will be presented in Section 5.8. 99 Chapter Five – Gender and Number

nouns can be human, nonhuman animate or inanimate, so any attempt to associate these

nouns with a certain value of CSFs does not seem plausible.

Prior to entering the Morphology, the syntactic structures of overt and covert

feminine nouns must not be different from each other. In both types, there is a root

morpheme with a unique index, i.e., [√(index)], a category-assigning [n] morpheme that is

also the location of CSFs, and a feminine morpheme [+F]. Since the Morphology is a

language-specific component, it can be expected that it is capable of distinguishing

between feminine structures of the overt type and those of the covert one, with the latter

being the special case that motivates a specific morphological alteration.

Given that this modification cannot be related to the CSFs of the [n] morphemes,

and all other morphemes are not distinct except for the root morphemes, there is a

possibility that root indices have a role to play in making covert feminine structures

identifiable to the Morphology. That is, any given covert feminine noun is probably a

special sequence of morphemes such that when identified by the index of the root in the

Morphology, the [+F] morpheme is impoverished.

If this is actually the case, root indices are capable of influencing the Morphology

just as they are also capable of influencing the Vocabulary and the Encyclopaedia.

These components of DM are language-specific, as mentioned in Chapter Three, and

since root indices can influence the Vocabulary and the Encyclopaedia, it does not seem

to be theoretically problematic that root indices can influence the Morphology as well.

Root indices play a role in the choice of root VIs in the Vocabulary and the choice of

the appropriate semantic interpretations in the Encyclopaedia, as discussed in Chapter

Four. What is hypothesised here is an extension of the role of these indices. The

Morphology is the storehouse of language-specific rules, and it is not unreasonable that

a portion of the responsibility for activating some of these rules is associated with root

100 Chapter Five – Gender and Number

indices. This is similar to the concept that the choice of language-specific root VIs in

the Vocabulary is associated with these indices. As the discussion of the other types of

nouns unfolds, this possibility becomes even more likely.

5.2.3. Number in Singular Nouns

Arabic singular nouns in SA and DA, be they masculine, overt feminine, or covert

feminine, show no overt number marking. Hence, they are considered the base forms

from which the other types of nouns are derived (Ryding, 2005:119). However,

although the singular feature of the nouns is not overtly marked, it can be discovered

through the nouns’ agreement with, for instance, demonstrative pronouns. Singular

nouns take the singular demonstrative haːðaː (this) when masculine and its feminine

counterpart haːðihi when feminine, but they neither take the dual masculine

demonstrative haːðaːni, nor its feminine counterpart haːtaːni, as shown below:

(7) a. haːðaː al-tʕaːlib- / *haːðaːni al-tʕaːlib- this.M.Sg Def-student this.M.Dl Def-student ‘This (male) student.’

b. haːðaː al-kitaːb- / *haːðaːni al-kitaːb- this.M.Sg Def-book this.M.Dl Def-book ‘This book.’

c. haːðihi al-tʕaːlib-at- / *haːtaːni al-tʕaːlib-at- this.F.Sg Def-student-F this.F.Dl Def-student-F ‘This (female) student.’

d. haːðihi al-ziyaːr-at- / *haːtaːni al-ziyaːr-at- this.F.Sg Def-visit-F this.F.Dl Def-visit-F ‘This visit.’

In the above examples, the nouns have two features that determine what demonstrative

pronoun should be used, namely gender, which can be masculine or feminine, and

number, which is always singular. Apart from the requirement of gender agreement, the

above nouns cannot occur with any dual demonstrative, indicating that the nouns are

strictly singular despite the absence of any singular suffix. Hence, if the underlying

101 Chapter Five – Gender and Number

structure of the above singular nouns has a number morpheme, this morpheme can be

nothing but [Sg], which, in contrast with [Dl] and [Pl] as will be shown soon, has no

phonological representation in the language, i.e., /Ø/. That is, from a DM perspective,

when the number morpheme of a nominal structure is specified as [Sg], /Ø/ is the only

exponent that can phonologically express it in the Vocabulary.

(8) /Ø/ ↔ [Sg]

Investigating how the other variations of [Num], i.e., [Dl] and [Pl], are phonologically

expressed in the Vocabulary will reveal different number VIs and provide us with a

comprehensive view of how number is represented in SA and DA nouns.

Since singular nouns do not have overt number markers, they cannot be used for

exploring the primary location of [Num] in the underlying structure of Arabic nouns.

However, by comparing singular nouns to the other types, especially dual feminine

nouns, it becomes clear that [Num] is an independent morpheme under NumP that is

located immediately higher than the GendP in the tree diagram shown in (3) above.

Consider the following examples:

(9) a. ʕayanat dawl-at-u al-ʔimaːr-aːt-i safiːr-at-ayni appointed state-F-Nom Def-emirate-F.Pl-Gen ambassador-F-Dl.Acc ‘(The state of) Emirates appointed two female ambassadors.’

b. san-at-ain taʔriːban57 year-F-Dl approximately ‘Approximately two years.’

The SA sentence in (9.a) above contains the dual feminine noun safiːr-at-ayni (two

female ambassadors), and the DA sentence in (9.b) contains the dual feminine noun

san-at-ain (two years). The duality of the nouns is expressed by the suffixes -ayni and

-ain respectively; if these are removed, the nouns will automatically revert to singular

without affecting their feminine gender, which is expressed by the suffix -at. It is

57 This sentence was produced by an Egyptian participant. The glottal sound /ʔ/ in taʔriːban (approximately) is usually pronounced as /g/ in the other varieties, and as /q/ in SA. 102 Chapter Five – Gender and Number

assumed that these suffixes are different phonological representations of [Dl], which, in

turn, is a variation of [Num]. The absence of any singular suffix in singular nouns does

not necessarily indicate the absence of [Num] in the basic structure. Rather, it is an

indication that, although the morpheme is there, it has no overt phonological

representation in the Vocabulary of the language when specified as [Sg], as suggested in

(8) above. Hence, after the addition of [Num], the structure shown in (3), can be

updated as follows:

(10) NumP

GendP [Num]

nP [Gend]

[√(index)] [n[CSF]]

The upcoming investigation of sound plurals, especially SndF, will reveal that the

primary position of [Num] suggested in the above structure offers the best account for

the gender and number features of Arabic nouns.

5.3. Dual Nouns

Dual nouns embody the first significant difference between SA and DA nouns,

specifically in number. With respect to the gender feature, SA and DA dual nouns are

not different from each other, and they are also not different from singular nouns. With

respect to the number feature of the nouns, however, SA and DA dual nouns exhibit two

different marking systems. In SA dual nouns, the determination of the appropriate dual

suffix is always based on the case feature of the nouns. Nominative dual nouns take the

suffix -aːni, while non-nominative (accusative, oblique and genitive) nouns take the

103 Chapter Five – Gender and Number

suffix -ayni. This indicates a strong interaction between number and case in SA dual

nouns, which will be discussed from a DM perspective in the next chapter.

DA dual nouns, on the other hand, show a simple and straightforward number

marking system in which -ain is the only marker. DA dual nouns take this suffix

regardless of their function in the sentence, thus indicating a total separation between

number and case, which can be attributed to the lack of overt case marking in DA, as

will be discussed in Chapter Six. Consider the following examples:

(11) SA Dual Nouns: a. wasʕala safiːr-aːni arrived ambassador-Dl.Nom ‘Two ambassadors arrived.’

b. hunaːka mawdʕuːʕ-aːni muhimm-aːni there subject-Dl.Nom important-Dl.Nom ‘There are two important subjects.’

c. ʕayanat dawl-at-u al-ʔimaːr-aːt-i safiːr-at-ayni appointed state-F-Nom Def-emirate-F.Pl-Gen ambassador-F-Dl.Acc ‘(The state of) Emirates appointed two female ambassadors.’

d. fiː madiːn-at-ayni in city-F-Dl.Obl ‘In two cities.’

(12) DA Dual Nouns: a. baʕd ∫ahr-ain wu nusʕ after month-Dl and half ‘After two and a half months.’

b. ðʕifdaʕ-ain xalf al-∫aʒar-ah frog-Dl behind Def-tree-F ‘Two frogs behind the tree.’

c. ∫uft raʒaːl-ain yetkalamuːn saw man-Dl talking ‘I saw two men talking.’

d. al-∫aʕb magsuːm nusʕ-ain Def-nation divided half-Dl ‘The nation is divided (into two halves).’

e. san-at-ain taʔriːban year-F-Dl approximately ‘Approximately two years.’ 104 Chapter Five – Gender and Number

Starting with gender, the above examples show that dual nouns, in SA and DA, do not

take any gender suffix when masculine. The absence of a masculine suffix in these

nouns leads to placement of the dual suffixes immediately after the stems, as can be

seen in (11.a-b) and (12.a-d) above. However, when the nouns are feminine, as shown

in (11.c-d) and (12.e), the standard feminine suffix -at is inserted between the stems and

the dual suffixes.

The way Arabic dual nouns are marked for gender is consistent with the way

singular nouns are marked, which suggests that the gender VIs of singular nouns shown

in (4), and reproduced below, also capture the gender marking system in dual nouns.

(13) /at/ ↔ [+F] /Ø/ ↔ elsewhere

Moreover, the position of the feminine VI /at/ between the stems and the dual suffixes

asserts the independence of the [+F] morpheme it encodes – and, consequently, the

independence of [Gend] in general – from any other constituents of the structure, as

discussed earlier. Subsequent discussions will reveal that the independence that [Gend]

enjoys in singular and dual nouns will also be found in SndM, but is significantly

disrupted in the other nouns.

As for number, which is where the difference between singular and dual nouns

lies, dual nouns are always overtly marked for duality by a suffix. This study assumes

that the [Num] morpheme in dual structures is specified as [Dl]. However, since the

phonological representation of [Dl] in SA is associated with case, discussion of the

whole number marking system of SA dual nouns will be postponed to Chapter Six to

avoid any repetition. As for DA, -ain is purely a number suffix that expresses the

duality of its nouns. This makes /ain/ the only phonological representation of [Dl]

available in the Vocabulary of DA. With respect to what the Vocabulary also provides

105 Chapter Five – Gender and Number

for [Sg], the phonological entries for number in DA singular and dual nouns can be

represented as follows:

(14) /ain/ ↔ [Dl]

/Ø/ ↔ elsewhere

According to the above representation, which will be updated when plural nouns are

investigated, whenever [Num] is specified as [Dl], the VI /ain/ will be inserted into it.

Otherwise, the Vocabulary provides /Ø/, which is the elsewhere VI that corresponds to

[Sg]. It is important to emphasise that in order for the VI /ain/ to be inserted, [Dl] must

be independent. This is the case only in DA.

5.4. Nonhuman Plurals

The humanness feature is more essential in Arabic plurals than in singulars and duals. It

plays an important role in determining the form of agreement the plurals take.58 Human

plurals agree according to their gender and number, as shown in (15). Nonhuman

plurals, on the other hand, agree as feminine singulars, as shown in (16) below:

(15) Human Plurals Agreement:59

a. ʔaħad- al-sukkaːn- al-ʔasʕliy-iːna one Def-resident.M.Pl Def-indigenous-M.Pl ‘One of the indigenous residents.’

b. sa-y-adʕamu mura∫aħ-iːna mustaqill-iːna will-he-support candidate-M.Pl independent-M.Pl ‘He will support independent candidates.’

c. al-tʕaːlib-aːt al-naːʒiħ-aːt Def-student-F.Pl Def-successful-F.Pl ‘The successful (female) students [i.e., successful in passing the exams].’

58 Although case and definiteness are among the relevant agreement features, humanness does not affect them as it affects gender and number. When the agreement of Arabic plurals is under investigation, the focus is usually directed to the correlation between humanness, on the one hand, and gender and number on the other. Hence, the relevant words of the examples used in this section will be classified according to their gender and number only, e.g., M, F, Sg, and Pl. 59 Collective human plurals, e.g., sukkaːn (residents), are usually masculine despite the fact that they can also be used for referring to a group of males and females. However, when referring to a group of females only, a SndF plural form is usually used instead, e.g., saːkin-aːt (residents ‘F Pl’). 106 Chapter Five – Gender and Number

(16) Nonhuman Plurals Agreement:

a. furasʕ- ʒadiːd-at- opportunity.F.Pl new-F.Sg ‘New opportunities.’

b. maxluːq-aːt- mufiːd-at- creatur-F.Pl beneficial-F.Sg ‘Beneficial (helpful) creatures.’

As found in Ferguson (1989) and others, the first type of agreement is called strict

agreement and the second type is called deflected agreement. This section is concerned

with the second type, in which nonhuman plurals agree as feminine singular. Adopting

the terms used by Corbett (1983) in investigating Slavic agreement,60 nonhuman plurals

will be referred to as controllers whenever agreement is involved, while the expressions

that agree with them will be referred to as targets, which include adjectives, verbs, and

pronouns (including anaphoric, possessive, relative, and demonstrative).

Although the phenomenon of deflected agreement will be involved in the

discussion, it is important to emphasise that this section is more concerned with

exploring the structure of nonhuman plurals in terms of gender and number than

investigating the phenomenon. In fact, deflected agreement has drawn a lot of attention

and been extensively studied.61 A further investigation would be beyond the scope of

this thesis for two reasons. Firstly, it would require a comprehensive examination of the

wide variety of the targets involved. Secondly, it would require a thorough diachronic

study as the phenomenon, most notably in adjectives, has developed gradually through

different stages of the language. Any investigation that engages with such a challenging

phenomenon while focusing on a subset of its variables is likely to lead to inaccurate

conclusions. In Section 5.6, it will be shown how Noyer’s (1992) attempt to use

60 Which are also used in the literature of Arabic agreement, e.g., Belnap and Shabaneh (1992). 61 For more on deflected agreement, see Beeston (1975), Ferguson (1989), Belnap (1991, 1999), Belnap and Shabaneh (1992), Belnap and Gee (1994), Herin and Al-Wer (2013), among many others. 107 Chapter Five – Gender and Number

deflected agreement as evidence without enough consideration of its variables has lead

to the inaccurate conclusion that [Pl] is pre-syntactically configured in broken plurals.

Deflected agreement involves singular feminine forms of agreement with

nonhuman plurals, but the term deflected does not actually specify whether it is a

number or a gender deflection, and whether the deflection is associated with the

agreement system of the language or with the nouns themselves. To have a better

insight into the structure of nonhuman plurals, this study deals with deflected agreement

as a phenomenon with two instances of agreement: singular agreement and feminine

agreement. The purpose is to determine what nonhuman plurals have to do with each

kind of agreement, i.e., why nonhuman plurals agree as singulars, and why they agree as

feminine. It will be shown that singular agreement may have more to do with a possible

deflection in the agreement system than with the plurals themselves, while feminine

agreement can be mainly attributed to changes to the underlying structure of the plurals.

Relying mostly on a study by Belnap and Shabaneh (1992), which considers

many of the variables associated with deflected agreement, section 5.4.1 is a brief

overview of how the phenomenon has developed through the history of Arabic. The

overview is followed by a discussion in which we will see that the singular agreement

aspect of the phenomenon is a property of the agreement system of the language rather

than the structure of the nouns. The most relevant discussion is in Section 5.4.2 where it

will be shown that the feminine agreement aspect is probably motivated by the internal

structure of the plurals to which the agreement system simply responds. Specifically,

the subsection investigates the gender feature of nonhuman plurals and accounts for the

changes it may undergo within the framework of DM.

108 Chapter Five – Gender and Number

5.4.1. The Development of Deflected Agreement

Using a corpus of randomly selected texts representing the pre-Islamic, Koranic,

Classical, and the modern (referred to as SA in this thesis) era of Arabic, Belnap and

Shabaneh (1992) present a statistical study showing the distribution of strict and

deflected agreements with nonhuman plurals (controllers) across these stages.62 The

corpus contains approximately 90,000 words of which more than 700 controllers were

identified as being agreed with by adjectival, verbal, anaphoric and demonstrative

targets. The overall results are shown in the graph that follows:

Figure 5.1: Deflected and Strict Agreements across the Stages of Arabic 63

120

100

60 Delected Strict 40

0 Pre-Islamic Koranic Classical SA

The findings show how deflected agreement has shifted from the least common form of

agreement in the pre-Islamic stage into the dominant form in the subsequent stages.

This supports an earlier argument made by Beeston (1975:66) that deflected agreement

is a phenomenon that has gradually become the norm in Arabic. Beeston assumes that it

first appeared in pronouns then started affecting the other types of targets.

62 The vast majority of the texts represent the Classical and the SA stages of the language, as they were the main concern of the study. 63 Figure 5.1 is a modification of Table 2 of Belnap and Shabaneh (1992:254). In the original table, controllers are classified by their forms: sound and broken. The data is classified by the sources: Imru’ Al-Qays (pre-Islamic), Quran (Koranic), Maqamat and Ibn Khaldoun (Classical), Taha Hussein and modern press (SA). Finally, targets are classified as feminine singulars (deflected), as well as broken plurals and feminine plurals (strict). 109 Chapter Five – Gender and Number

Although deflected agreement became more common in Koranic Arabic, this

stage of Arabic appears to have been a transitional period for adjectives whose form of

agreement was dependent on the form of their controllers. Belnap and Shabaneh

(1992:256) show that adjectives targeted by broken controllers in the Quran are in

variation between deflected (50 per cent) and strict (50 per cent), whereas those targeted

by sound controllers are exclusively strict (100 per cent). This can be illustrated by the

following Koranic examples:

(17) Variation of Adjectival Agreement with Broken Controllers in Koranic Arabic:

a. fiː sʕuħuf-i-n mukarram-at-i-n marfuːʕ-at-i-n mutʕahar-at-i-n [Quran 80:13-14] in sheet.F.Pl- honoured-F.Sg- exalted-F.Sg- purified-F.Sg- ‘(It is recorded) in honoured, exalted, and purified sheets.’

b. wa fiː al-ʔardʕ-i qitʕaʕ-u-n mutaʒaːwir-aːt-u-n [Quran 13:4] and in Def-land.F- piece.F.Pl- adjacent-F.Pl- ‘And in the Earth are adjacent tracts.’

(18) Strict Adjectival Agreement with Sound Controllers in Koranic Arabic:

a. ʒann-aːt-u-n maʕruː∫-aːt-u-n [Quran 6:141] garden-F.Pl- trellised-F.Pl- ‘Trellised gardens.’

b. ʔaːy-aːt-u-n muħkam-aːt-u-n [Quran 3:7] verse-F.Pl- firm-F.Pl- ‘Firm (precise) verses.’

As the analysis of Belnap and Shabaneh (1992) suggests, deflected agreement

(adjectival and otherwise) appears to have developed further in Classical Arabic and

SA. As far as SA is concerned, the analysis shows that deflected agreement has become

dominant to such an extent that it is probably safe for the rare occurrences of strict

agreement to be considered circumstantial and negligible. In fact, out of the 433

controllers the authors found in SA, only five (1.15 per cent) showed strict agreement.

Four of these instances were found in one novel called Al-Ayyam (the days) by Taha

Hussain, as exemplified below (Belnap and Shabaneh, 1992:250):

110 Chapter Five – Gender and Number

(19) wa haːðihi al-ʔiʒaːz-aːt al-qisʕaːr allatiː kaːn-at and this.F.Sg Def-vacation-F.Pl Def-short.F.Pl that.F.Sg was-F.Sg

t-ataxallal diraːsat al-ʔazhariy-iːna F.Sg-interrupt studying Def-Azhari-M.Pl ‘…and these vacations that were interrupting Azharites’ studying.’

The strict agreement occurring in the above example is between the controller ʔiʒaːz-aːt

(vacations ‘F Pl’) and the adjective qisʕaːr (short ‘F Pl’).64 However, as evidence for the

prevalence of deflected agreement, the rest of the targets modifying the same controller

are all deflected, e.g., haːðihi (this ‘F Sg’), allatiː (that ‘F Sg’), kaːn-at (was ‘F Sg’),

and t-ataxallal (interrupt ‘F Sg’).

Given this brief overview of deflected agreement, it can be seen that the main

aspect of deflected agreement is that nonhuman plurals are treated as singulars. This

might have more to do with the agreement system of the language than with the plurals

themselves. That is, what is taken to be a gradual development of the phenomenon of

deflected agreement is likely to have been a gradual deflection in the agreement system

by which the system has progressively become unable to provide plural forms of

agreement to nonhuman plurals. The lack of nonhuman plural forms of agreement can

be clearly seen in the paradigms of subject, object, possessive, relative and

demonstrative pronouns, as presented in Ryding (2005). Taking the paradigms of

(proximity) demonstratives and (independent 3rd person) subject pronouns as examples,

consider the table below:

64 This reflects a tendency by some modern novelists who occasionally use strict adjectival agreement, probably to resemble the Koranic and pre-Islamic style in order to distinguish their works. Another example in Belnap and Shabaneh (1992:250) shows Taha Hussain using both strict and deflected adjectival agreement in one sentence. 111 Chapter Five – Gender and Number

Table 5.1: Paradigms of Demonstratives and Subject Pronouns Masculine Feminine

Demonstratives Singular haːðaː (this) haːðihi (this) Dual haːðaː-aːni (these two) haːt-aːni (these two) Plural (human only) haːʔulaːʔi (these)

Subject Pronouns Singular huwa (he/it) hiya (she/it)

Dual humaː (they both)

Plural (human only) hum (they) hunna (they)

The above table shows that although there are plural forms in the system, these forms

are reserved for human nouns. Therefore, when a nonhuman plural occurs, there is

nothing the system can offer but the singular forms.65 As both masculine and feminine

singular forms are prima facie eligible, all examples of this subsection have shown that

it is only the feminine ones that are supplied. This is another aspect of deflected

agreement that will be the topic of the next subsection.

5.4.2. Gender in Nonhuman Plurals

From a DM perspective, this subsection tries to explore why the provided singular

targets are always feminine. Is it another deflection in the agreement system, or is it that

the controllers actually are feminine?

It is not surprising that nonhuman feminine singulars remain feminine in their

plural forms. What might be unusual is that masculine singulars change into feminine

when plural, as shown below:

(20) Nonhuman Masculine Singulars à Feminine Plurals:

a. haːðaː waːʒib- min waːʒib-aːt- this.M.Sg duty.M.Sg from duty-F.Pl ‘This is a duty of (many) duties.’

65 Not offering the dual pronouns as an alternative may suggest that the singular forms are the default. 112 Chapter Five – Gender and Number

b. kalb- dʕaːl- / kalb-at- dʕaːʔiʕ-at- à al-kilaːb- al-dʕaːl-at- dog.M.Sg M.Sg-stray / dog-F.Sg missing-F.Sg Def-dog.F.Pl Def-strayF.Sg ‘A stray (male) dog.’ ‘A missing (female) dog.’ ‘The stray dogs.’

c. dars- muhim- à duruːs- muhim-at- lesson.M.Sg important.M.Sg lesson.F.Pl important.-F.Sg ‘An important lesson.’ ‘Important lessons.’

d. tabaːdul- tiʒaːriy- à tabaːdul-aːt- tiʒaːriy-at- exchange.M.Sg commercial.M.Sg exchange-F.Pl commercial-F.Sg ‘A commercial exchange.’ ‘Commercial exchanges.’

The above examples show that nonhuman masculine singulars, including animate

singulars that can have both genders as seen in (20.b), are all grammatically feminine in

their plural forms. Their feminine gender when plural can be determined by observing

the overt feminine suffix (i.e., the SndF plural suffix -aːt) they take as in (20.a), the

gender of their targets as in (20.b-c), or by both approaches as seen in (20.d).

Overt feminine plurals, as in (20.a) and (20.d), are problematic for any approach

in which the gender of nonhuman plurals is determined based on the gender of their

singular forms. These plurals are overtly feminine (SndF) and agree as feminine despite

being masculine when singular. The following table shows more examples of masculine

singulars that change into overt feminine when plural.

Table 5.2: Nonhuman Masculine Singulars Pluralised as SndF Masculine Singular SndF Plurals mugtamaʕ (society) mugtamaʕ-aːt (societies) tartiːb (arrangement) tartiːb-aːt (arrangements) ʕ ʕ qit aːr (train) qit aːr-aːt (trains) ʕ ʕ mat aːr (airport) mat aːr-aːt (airports) ʔiʕlaːn (announcement) ʔiʕlaːn-aːt (announcements) tawatur (tension) tawatur-aːt (tensions) ʕ ʕ muħit (ocean) muħit -aːt (oceans) ʔinʕikaːs (reflection) ʔinʕikaːs-aːt (reflections) istiӨmaːr (investment) istiӨmaːr-aːt (investments)

This should leave no doubt that nonhuman masculine singulars, in effect, change their 113 Chapter Five – Gender and Number

gender when plural. Determining the gender of nonhuman plurals by relying on the

gender they possess when singular may lead to overlooking the phenomenon of gender

alteration and complicates the phenomenon of deflected agreement. When nonhuman

plurals of masculine origin are considered to preserve their gender, deflected agreement

will always involve a number deflection (F Pl vs. F Sg), as well as a gender deflection

at times (M Pl vs. F Sg). This study separates number agreement from gender

agreement. With respect to nonhuman plurals, number agreement is always deflected

(Pl vs. Sg) as a result of a gap in the agreement system, as discussed earlier, whereas

gender agreement is not deflected as the plurals are actually feminine, so their targets

are feminine as well (F vs. F). This gives rise to the question: why are nonhuman plurals

always feminine?

Given that gender is a syntactic feature (as also considered in a number of studies,

e.g., Picallo, 1991, 2008; Ritter, 1993; Pfau, 2002, 2007; Müller, 2004; Lowenstamm,

2008), there are two possibilities in the grammar of DM to explain why nonhuman

plurals are always feminine. The first possibility is that nonhuman plurals are never

structured as masculine. That is, they always emerge from the Syntax with their gender

morphemes specified as [+F]. Such an assumption entails that the Syntax is somehow

prevented from building structures with [-F], [n[-Hum]] and [Pl] morphemes. Needless to

say, such a hypothesis will automatically lead to parameterizing the Syntax and

undermining its universality. Since it is likely that masculine plurals exist in other

languages whose nonhuman nouns can be marked for gender, 66 attributing the

prohibition of these structures in Arabic to the Syntax appears to be arbitrary and

theoretically undesirable. Therefore, this study assumes that the Syntax is capable of

building masculine nonhuman plural structures, i.e., [[[[√(index)] n[-Hum]] -F] Pl], but the

66 Examples of nonhuman masculine plurals from other languages can be los libros blancos, and les livres blancs (the white books) in Spanish and French respectively. 114 Chapter Five – Gender and Number

structures are prohibited from surfacing as such in Arabic. This leads to the other

possibility that whenever a syntactic structure of this sort emerges from the Syntax, the

Morphology, as a language-specific component, is where the structure becomes

feminine.

Noyer (1992)67 observes some relatively similar phenomena in Arabic verbs.

Among these phenomena, for example, is the absence of gender distinction in first-

person verbs, e.g., ʔ-aktub- (I write) and n-aktub- (we write), that are both used for

masculine and feminine. To account for this, Noyer (1992:43) proposes a

morphological filter68 that prohibits the co-occurrence of [1st] and [F] in structures.

(21) a. *[1st F]

According to the above filter, the Morphology restricts the co-occurrence of the features

[1st] and [F], so one must be impoverished. Noyer (1992:46) proposes that the features

related to his observations are hierarchically ordered as follows:69

(22) 1st > 2nd > Pl > Dl > F

Hence, when a choice between [1st] and [F] is to be made in the Morphology, it is the

latter that will be impoverished as it is the lower feature in the hierarchy. This study

adopts Noyer’s notion of morphological filters. However, before applying a filter to

nonhuman plurals, Noyer’s treatment of gender distinctions in Arabic needs to be

discussed, as it is directly related to our analysis.

Noyer (1992) does not consider masculine, which is represented as [-F] in this

study, an active feature in Arabic.70 As mentioned in Noyer (1992:41-42), previous

phonological experiments (Eimas, Siqueland, Jusczyk, and Vigorito, 1971; Streeter,

67 Noyer (1992) is a doctoral thesis submitted to MIT and published as a book in 1997. 68 According to Noyer (1992:41-43), the notion of morphological filters is a modified morphological application of Calabrese's (1988, 1991) phonological co-occurrence restrictions (phonological filters). 69 For more on Feature Hierarchy, see Noyer (1992:44-49). 70 This assumption appears to be based on the fact that Arabic has no overt masculine marker in its phonological system. 115 Chapter Five – Gender and Number

1976; Werker, 1989) have shown that infants perceive more phonological distinctions

than what they become able to perceive as they grow. This supports Calabrese’s (1988,

1991) proposal of phonological alphabets, which Noyer (1992:43) converts into

morphosyntactic terms as follows:

Converting these ideas into the realm of morphosyntax, I will assume that there exists a universal set of features for at least person, number, tense and aspect. Gender features (Corbett, 1991) and the classification of verbal action called Aktionsart are possibly universally defined but show such variability and are so poorly understood that I will have little to say about these here.

Based on this, Noyer (1992:43) argues that the active features of a language are a

chosen subset of the features made available by UG, whereas the rest are simply

inactive in the language although they are universally available:

Of the universal set of morphosyntactic features, some are active in a given morphosyntactic system, while others are inactive. Of the active features, not all will fully cross-classify with each other, giving the effect of ‘gaps’ in inflectional paradigms.

Noyer’s treatment of gender in Arabic implies that [F] is the only active gender feature

in the language, so it is not necessary for gender to be represented as a binary feature,

[±F], since it is either [F] in feminine structures or simply nothing otherwise.71

If masculine gender is not present as a feature in syntactic structures, and

restricting nonhuman plurals from being masculine cannot take place in the Syntax, the

only way for the Morphology to change the gender of nonhuman plurals into feminine

is to add a gender morpheme specified as [F] to their structures, so they can later surface

as nonhuman feminine plurals. That is, according to this hypothesis, whenever the co-

occurrence of [n[-Hum]] and [Pl] with the absence of [F] is detected in the Morphology, a

gender morpheme specified as [F] will be added. This clearly declares gender a

dissociated morpheme, which cannot be the case as mentioned earlier.

71 A similar assumption in Spanish can be seen in Harris (1991:60) who emphatically states that “masculine gender is not marked in Spanish in any way, lexically, morphologically, or phonologically. No binary feature [-feminine], or any formal equivalent, plays any role in the grammar of Spanish”. 116 Chapter Five – Gender and Number

The detection of a feature representing the masculine gender is a key requirement

for gender changing in nonhuman plurals to be performed, which lends support to our

earlier assumption that [Gend] is specified as [-F] in masculine nouns and as [+F] in the

feminine ones.72 With the presence of [-F], one might think of adopting Impoverishment

as a feature-changing operation (Noyer, 1998) to account for changing the [-F] feature

of nonhuman plurals into [+F] in the Morphology. Recall that, within this hypothesis,

Impoverishment is capable of deleting a feature in order to allow a less marked value of

the same feature to take its place. What we have in Arabic nonhuman plurals, however,

is the opposite. Gender changing in Arabic nonhuman plurals involves replacing the less

marked feature [-F] with the more marked feature [+F]. Hence, Impoverishment as a

feature-changing operation cannot be adopted.

What is needed here is a morphological operation that is capable of performing a

direct feature-changing regardless of whether the new feature is more or less marked

than the old one. As mentioned in Chapter Three, Marantz (1996:14) indicates that one

needs to be careful to claim that feature-changing is impossible since data motivating

such an operation may actually exist. This study argues that the behaviour that Arabic

nonhuman plurals exhibit regarding gender requires such an operation, and this is what

will be proposed in the remainder of this subsection.

Following the notion of morphological filters, this study assumes the existence of

the following filter in the Morphology of Arabic:

(23) *[n[-Hum] -F Pl]

According to the above filter, it is prohibited for any nominal, nonhuman, masculine,

and plural structure to pass through the Morphology unchanged. In the context of verbs,

as shown in Noyer (1992), feature-deleting is adopted as the operation by which

72 Since structures need to be syntactically fully specified for gender, it is possible that [Sg], which is a variant of [Num] as [Dl] and [Pl], is also present although it does not have an overt phonological representation in the language. 117 Chapter Five – Gender and Number

structures become compliant with the restrictions imposed by the filters. However, what

we are dealing with here is different. Unlike the first-person verbs that need to surrender

their gender to become compliant, nonhuman masculine plurals emerge from the

Morphology with a new gender that is then reflected on the gender of their targets. This

implies two things about the Morphology of Arabic. Firstly, it suggests that the

Morphology is in fact capable of changing the value of the [-F] morpheme into [+F].

Secondly, it implies that changing the value of the feature is performed prior to copying

it onto the agreement nodes of the targets, which is not the case with the structure of

covert feminine nouns, as discussed earlier. The second issue requires a further

investigation into the internal organisation of the Morphology, which will be discussed

in Section 5.8 after all the modifications related to gender and number have been

illustrated. Modifications that deal with case and definiteness will be discussed in

Chapter Six and they will provide further support the assumption that will be made in

this chapter about the internal organisation of the Morphology.

Given that gender changing is performed in the Morphology, there must be an

operation by which the masculine feature is changed into feminine. This study argues

that changing the masculine feature into feminine is achieved by changing its value

from (-) into (+), so the feature of the existing gender morpheme is changed from [-F]

into [+F], allowing the structure to surface as feminine. If this is on the right track, it

entails the existence of an operation that, when activated, is capable of targeting the

value of the feature – let us call it Revaluation – as shown below:

(24) [-F] → [+F] / [n[-Hum]] + ____ + [Pl]

The above rule can be interpreted as: [-F] changes into [+F] whenever it co-occurs with

[n[-Hum]] and [Pl]. This Revaluation process is motivated by the morphological filter

shown in (23).

118 Chapter Five – Gender and Number

Revaluation in this sense has a less drastic effect than the existing DM operations

that are capable of combining the features of multiple morphemes into one (Fusion),

and splitting the feature of a single morpheme into multiple morphemes (Fission).

Revaluation does not increase or decrease the number of the syntactically defined

morphemes or the number of the phonological terminals into which VIs will be inserted.

In this instance, it simply changes the value of an existing morpheme so the whole post-

syntactic structure becomes compliant with the local rules of the language. Revaluation,

however, must remain a tentative proposal until supported by further evidence from

Arabic as well as other languages.

If this is on the right track, deflected agreement is in fact a number deflection in

the agreement system of Arabic. It can be seen as a gap due to which the system lacks

nonhuman plural targets. The development of deflected agreement throughout the

history of Arabic reflects the gradual expansion of this gap. The lack of plural targets

forces the system to alternatively provide singular forms of agreement when nonhuman

plurals occur. However, since nonhuman plurals are either syntactically feminine or

changed into feminine before the agreement level in the Morphology, the agreement

system supplies the feminine singular forms of targets in order to match the gender of

the controllers.

5.5. Sound Plurals

As mentioned in Chapter Two, sound plurals are divided into two types depending on

their gender: sound masculine (SndM) and sound feminine (SndF) plurals. Both types

are composed of basic nominal stems followed by a suffix. In SA SndM plurals, the

suffix expresses the number and the case features of the noun, while it only expresses

the number feature in DA. In SndF plurals, on other hand, it is a gender and number

suffix in both SA and DA. Form a DM perspective, the following subsections strive to 119 Chapter Five – Gender and Number

account for how the gender and number features of these nouns are handled during the

word formation process.

5.5.1. SndM Plurals

SndM is a group of human masculine plurals that are marked differently in SA and DA.

SA SndM plurals take the suffix -uːna when the nouns are nominative, and -iːna

otherwise. DA SndM plurals, on the other hand, take the suffix -iːn regardless of the

noun’s function in the sentence. Consider the following examples:

(25) SA SndM Plurals:

a. muraːqib-uːna rasmiy-uːna observer-Pl.Nom official-Pl.Nom ‘Official observers.’

b. ʕaqada al-mudiːr-u ʔiʒtimaːʕ-a-n maʕa al-muwaðʕaf-iːna held Def-director-Nom meeting-Acc-Idef with Def-employee-Pl.Obl ‘The director held a meeting with the employees.’

(26) DA SndM Plurals:

a. al-baːħiӨ-iːn yeħtaːʒ-uːn maʕaːmil Def-researcher-Pl need-M.Pl lab.F.Pl ‘The researchers need labs.’

b. ∫uft al-muwaːtʕin-iːn bi al-saːħ-ah saw Def-citizen-Pl in Def-square-F ‘I saw the citizens in the square.’

In both varieties, removing the SndM suffixes turns the nouns into singulars without

affecting their gender. This is an indication that these suffixes are mainly number

markers.

As seen previously, even though Arabic has no overt masculine marker, this study

assumes that the underlying structure of masculine nouns is fully specified for gender

with a [-F] morpheme. As far as SndM plurals are concerned, this morpheme is

independent, so its phonological representation in the Vocabulary can be captured by

the gender VIs proposed for singular and dual nouns, as shown below:

120 Chapter Five – Gender and Number

(27) /at/ ↔ [+F]

/Ø/ ↔ elsewhere

That is, the gender morpheme of SndM plurals falls in the elsewhere environment in

which /Ø/ is provided. Recall that this environment also covers the independent [-F]

morphemes in singulars and duals, as well as the impoverished [+F] in covert feminine

nouns. In order for any gender morpheme to be captured by the above VIs, it should

arrive in the Vocabulary as an independent morpheme. SndM plurals are the last type of

Arabic nouns for which it is assumed that gender morphemes are independent.

As for number, it appears that the case feature plays an essential role in the choice

of the SndM suffix in SA. This can be taken as an interaction between number and case

that is similar to the interaction found in SA dual nouns. Therefore, the number feature

of SA SndM plurals will be discussed in the next chapter. In DA SndM plurals,

however, -iːn appears to be a pure number marker because there is no case marking in

DA. Since -iːn is what gives nominal stems their plural feature in DA, it can be assumed

to be the number VI inserted into [Pl]. Along with the other number VIs found in DA

nouns so far, the representation of DA number VIs can be updated as follows:

(28) /iːn/ ↔ [Pl]

/ain/ ↔ [Dl]

/Ø/ ↔ elsewhere

Up to this point, the above representation entails that, in the Vocabulary, /ain/ is inserted

whenever the number morpheme is specified as [Dl] (DA dual nouns), /iːn/ is inserted

whenever it is specified as [Pl] (DA SndM plurals), and /Ø/ is inserted elsewhere (DA

singulars). Observe that the insertion of /ain/ and /iːn/ is conditioned by the

independence of the number morphemes. This independence is interrupted in SA SndM

plurals, as well as in SA and DA SndF and broken plurals, as will be discussed.

121 Chapter Five – Gender and Number

5.5.2. SndF Plurals

Unlike SndM plurals, SndF is a group that is open to both human and nonhuman nouns.

In terms of gender and number, SndF plurals behave in the same way in SA and DA.

They are simply basic nominal stems followed by the suffix -aːt in both varieties, as

shown below:

(29) SA SndF Plurals:

a. takallam-tu maʕa kulli al-tʕaːlib-aːt-i spoke-I with all Def-student-F.Pl-Obl ‘I spoke with all the (female) students.’

b. ya-ftaħu maʒaːl-aːt-i-n waːsiʕ-at-a-n it-open field-F.Pl-Acc73-Idef wide-F-Acc-Idef ‘It opens wide fields.’

c. ʔaʒraː muħaːdaӨ-aːt-i-n held conversation-F.Pl-Acc-Idef ‘He had conversations.’

(30) DA SndF Plurals:

a. marrayt bi-tʕaːlib-aːt passed-I by-student-F.Pl ‘I passed by (female) students.’

b. tʕaːwil-aːt wa karasiː table-F.Pl and chair.F.Pl ‘Tables and chairs.’

The SndF plurals in the above examples are composed of nominal stems followed

immediately by the suffix -aːt. The SndF plurals in (29.a-b) and (30.a) can be reverted

to their basic forms simply by removing the SndF suffix -aːt. This will automatically

result in changing the gender and number of these nouns back into masculine74 and

singular. However, for the SndF plurals in (29.c) and (30.b), which are feminine in their

singular forms, removing the suffix -aːt must be followed by adding the normal singular

73 SndF plurals take the usual oblique/genitive suffix /-i/ when accusative, as will be discussed in Chapter Six. 74 Recall that maʒaːl-aːt (fields ‘F Pl’) in (29.b) is a nonhuman plural whose gender is masculine in the singular form. 122 Chapter Five – Gender and Number

feminine suffix -at, i.e., muħaːdaӨ-at (a conversation) and tʕaːwil-at (a table). The

singular forms of these nouns cannot stand without a feminine marker (whether -at or -

aːt) since they have no masculine counterparts.

In both cases, however, the suffix -aːt suggests a strong interaction between

gender and number in this type of nouns. To capture this interaction, this study takes

/aːt/ to be the VI that phonologically represents the morpheme [+F, Pl].

(31) /aːt/ ↔ [+F, Pl]

The existence of such a morpheme composed of two features is likely to be the result of

a Fusion operation by which [+F] and [Pl] are combined in the Morphology.

Based on the basic syntactic structure of Arabic nouns that we have seen so far, as

proposed in Section 5.2, it is predicted that feminine plurals are generally structured as

follows:

(32) [[[[√(index)] n[CSF]] +F] Pl]

According to the above order of the morphemes, [+F] and [Pl] are two separate

phonological destinations that will be represented with one VI when the plural is SndF.

This requires a morphological intervention by which they ultimately become a single

destination for a single VI. Thus, it is a language requirement that motivates the

Morphology to combine [+F] and [Pl], as shown in the Fusion rule below:

(33) [+F] [Pl] → [+F, Pl]

The above rule creates the modified morpheme [+F, Pl]. For the insertion into this

morpheme, there will be a competition between the VIs that match any of its features,

and do not correspond to a feature that does not exist in the morpheme. In SA, the

competition is between /at/ and /aːt/, while the competing VIs in DA are /at/, /aːt/ and

/iːn/. This is when the Subset Principle (Halle, 1997) comes into play to regulate the

competition. According to the Subset Principle, the VI /aːt/ must be chosen as it

123 Chapter Five – Gender and Number

matches the greatest number of features of the morpheme, as opposed to the other

competing VIs of which each VI matches just one of the two features of the morpheme.

Having shown the effect of the above Fusion rule in the Vocabulary, it is

important to emphasise that in order for the rule to be activated in the Morphology, [+F]

and [Pl] must be features of two independent morphemes. If, for any reason, these two

morphemes are modified in such a way that their independence is disrupted, the rule

will be automatically supressed. This restriction will prevent the rule from operating on

the [+F] and [Pl] morphemes of feminine broken plurals that must have a different

configuration, as will be shown in the next section.

5.6. Broken Plurals

This section is concerned with how broken plurals are formed. Broken plurals, in SA

and DA, exhibit a special case where the genderless and numberless binyanim that are

consistently found in singular, dual and sound plural nouns are replaced by new

binyanim that express the gender and the plurality of the nouns, as shown below:

(34) a. ʕaːlim- à ʕulamaːʔ- scholar scholar.M.Pl ‘A male scholar.’ ‘Male scholars.’

b. fursʕ-at- à furasʕ- opportunity-F opportunity.F.Pl ‘An opportunity.’ ‘Opportunities.’

c. dars- à duruːs- lesson.M lesson.F.Pl ‘A lesson.’ ‘Lessons.’

In Chapter Four, it was shown that binyanim are what give structures their nominal

category, as well as their CSFs. The broken examples above show that when these

binyanim are replaced in order for broken plurals to be formed, the new binyanim also

become what give the structures their gender and number features. There is a claim that,

unlike sound plurals, broken plurals are pre-syntactically configured by having a special 124 Chapter Five – Gender and Number

[n] with an inherent [Pl] (Noyer, 1992). As a response to this claim, Section 5.6.1 shows

that broken plurals cannot be pre-syntactically formed. Section 5.6.2, then, provides an

alternative analysis in which it will be claimed that the distinction between sound and

broken plurals is created in the Morphology.

5.6.1. The Status of [Pl] in Broken Plurals

Noyer (1992) presents an analysis of the structural differences between Arabic sound

and broken plurals. The author concludes that the [Pl] feature of broken plurals is pre-

syntactically (i.e., in the Narrow Lexicon) inherent to [n], while the [Pl] feature that is

phonologically realised as an affix in sound plurals is a syntactically added feature.

(35) a. [n, Pl] (broken)

b. [n] [Pl] (sound)

Hence, when a broken structure is being built in the Syntax, the pre-existence of [Pl] as

an inherent feature blocks any further application of it.

To support his argument, Noyer (1992) provides two observations as evidence.

Firstly, he argues that the independence of [Pl] in sound plurals gives it the freedom to

fuse with case in SndM or to be “separately realised” in SndF.75 Secondly, and most

importantly, Noyer (1992:32-33) argues that having [Pl] as a pre-syntactically inherent

feature makes broken plurals special so they behave differently from sound plurals in

terms of agreement. In order to demonstrate the special status of broken plurals, he

shows how adjectives agree with sound plurals as compared to their agreement with

broken plurals in Koranic Arabic,76 as shown below (Noyer, 1992:33):

75 This study takes [Pl] to be fused with gender in SndF, as discussed earlier, and fused with case in SndM in SA, but not in DA, as will be discussed in the next chapter. Generally, this study agrees with Noyer (1992) that the independence that [Pl] enjoys is in fact what enables it to fuse or not with other features. 76 Although (36.a) and (36.c) may resemble how broken-adjective agreement is deflected sometimes in Koranic Arabic, the examples are not Koranic in the sense that they are extracted from the Quran. 125 Chapter Five – Gender and Number

(36) a. kutub-un77 kabiːr-at-un books big-F.Sg- ‘Big books.’

b. ʒann-aːt-un maʕruː∫-aːt-un [Quran 6:141] gardens trellised-F.Pl- ‘Trellised gardens.’

c. mudun-un kabiːr-at-u-n cities big-F.Sg- ‘Big cities.’

The purpose of the above examples is to show that the agreement between adjectives

and sound plurals, as in (36.b), cannot be deflected as opposed to the agreement

between adjectives and broken plurals, as shown in (36.a) and (36.c). Thus, the

deflected agreement that broken plurals exhibit is indicative that these plurals are

formed in a special way that involves assigning their [Pl] features pre-syntactically.

This study agrees that [Pl] is independent from [n] in sound plurals, while both

features are associated in broken plurals. However, this study argues that the association

between [Pl] and [n] in broken plurals is resulting from a morphological modification,

as will be discussed in Section 5.6.2, rather than a pre-syntactic inheritance.

Noyer’s employment of the phenomenon of deflected agreement as a special

behaviour of a certain form of Arabic plurals (broken plurals) is problematic for two

reasons. First, recall that deflected agreement is associated with nonhuman plurals, as

mentioned in Section 5.4. If, according to Noyer’s argument, deflected agreement is a

reflection of a special formation of broken plurals, we should not expect strict

agreement to occur with broken plurals whether the plurals are human or nonhuman,

which is not the case. Below are Koranic examples of strict agreement with human and

nonhuman broken plurals:

77 Noyer (1992) considers the nonhuman broken plural kutub (books) masculine probably based on its former gender in the singular form. In Section 5.4.2, it was shown that nonhuman plurals of masculine origins change their gender into feminine, and surface as overt SndF plurals sometimes. 126 Chapter Five – Gender and Number

(37) a. wildaːn-u-n muxallad-uːna [Quran 56:17] youth.M.Pl- immortal-M.Pl ‘Immortal youths.’

b. qitʕaʕ-u-n mutaʒaːwir-aːt-u-n [Quran 13:4] piece.F.Pl- adjacent-F.Pl- ‘Adjacent tracts.’

Second, as mentioned in Section 5.4, adjectives are not the only targets involved in

strict and deflected agreements. Thus, if deflected agreement is a special behaviour of

broken plurals, we should not expect other targets to be in a deflected agreement with

sound plurals. Yet pronominal and demonstrative targets in Koranic Arabic exhibit

deflected agreement with sound plurals, as exemplified below:

(38) a. tuʔaduː al-ʔamaːn-aːt-i ʔila ʔahl-i-haː [Quran 4:58] render.you Def-trust-F.Pl- to owners-Obl-its.F.Sg ‘(God commands) you to render trusts to their owners.’

b. tilka ʔaːy-aːt-u Allah-i n-atluː-haː [Quran 2:252] that.F.Sg verse-F.Pl-Nom Allah-Gen we-recite-it.F.Sg ‘These are the verses of Allah that We recite.’

The above examples show how the sound plurals ʔamaːnaːt (trusts) and ʔaːyaːt (verses)

agree with their targets -ha (its/it ‘F Sg’) and tilka (that ‘F Sg’). Although the nouns are

sound plurals, the targets are in their singular forms,78 indicating a deflected agreement.

As shown in Belnap and Shabaneh (1992), Koranic Arabic was a transitional

period for the phenomenon of deflected agreement, especially the adjectival type.

Although deflected agreement became more common in that stage, adjectives were

fluctuating between deflected and strict agreement depending on the plural form of their

controllers. With sound plurals, adjectival agreement was entirely strict. However, with

broken plurals, adjectives were able to have deflected or strict agreement. The below

78 The plural forms of these targets are -hunna ~ -hinna (they/their ‘F Pl’) and -ʔulaːʔika (those ‘Pl’), respectively (Ryding, 2005:309; 316). 127 Chapter Five – Gender and Number

examples reveal the freedom that adjectival agreement had with nonhuman broken

plurals in Koranic Arabic: 79

(39) a. ʔayaːm-a-n maʕduːd-at-a-n [Quran 2:80] day.F.Pl-Acc-Idef counted-F.Sg-Acc-Idef ‘A few days.’

b. ʔayaːm-a-n maʕduːd-aːt-i-n [Quran 3:24] day.F.Pl-Acc-Idef counted-F.Pl-Acc-Idef ‘A few days.’

However, given that deflected agreement is a phenomenon that has become the norm in

Arabic (Beeston, 1975:66), it is to be expected that adjectival agreement with

nonhuman plurals (sound and broken) was overwhelmingly strict in the pre-Koranic

stages and became overwhelmingly deflected in the post-Koranic stages. The Koranic

stage was just a period during which there was, to a certain degree, a relation between

the form of the nonhuman controller (sound or broken) and the agreement type (strict or

deflected) (Belnap and Shabaneh, 1992). This may reflect how plurals were subject to

the spreading phenomenon that dominated some of the adjectival domain of the broken

forms, but did not reach to the sound forms at that stage. It by no means can be taken as

evidence that broken plurals are special in such a way that their form is pre-syntactically

determined.

As mentioned earlier, relying on a few of the many variables involved in

deflected agreement is likely to lead to inaccurate conclusions. This seems to be what

led Noyer (1992) to consider the phenomenon of deflected agreement as a special

behaviour motivated by a special pre-syntactic status of [Pl] in broken plurals. The data

that Noyer (1992) selected to demonstrate deflected agreement is limited and far from

representative. He attempted to link the formation of all broken plurals (human and

nonhuman) to a behaviour that is sometimes seen in the agreement of some nonhuman

79 The adjectives in both examples are accusative. However, the first is singular and the second is SndF, and this is why they have different suffixes for the same case, as will be shown in the next chapter. 128 Chapter Five – Gender and Number

plurals with adjectival targets at a specific stage of the language. Noyer’s analysis

misses the agreement of human broken plurals, as well as the agreement of the

nonhuman ones with non-adjectival targets in Koranic Arabic – not to mention the other

stages of the language. If these variables were taken into consideration, it would become

clear that deflected agreement is about the humanness feature of the plurals rather than

their pluralisation form.

By demonstrating how Noyer’s (1992) analysis of the relationship between

deflected agreement and broken plurals is problematic, this study does not adopt his

view that the [Pl] feature of broken plurals is pre-syntactically inherent to [n]. Instead,

this study argues that the structural distinction between sound and broken plurals, which

involves combining the [Pl] feature of broken plurals with [n], is made in the

Morphology. The next subsection will address how broken plurals are formed in that

component.

5.6.2. The Formation of Broken Plurals in the Morphology

Comparing SA and DA broken plurals to their singular, dual, and sound plural

counterparts shows that broken plurals are composed of the same roots, but different

binyanim. Binyanim that appear as parts of the singulars remain the same in duals and

sound plurals, but change in the broken forms, as shown below:

(40) a. tʕabiːb (M Sg) tʕabiːb-aːni (M Dl) ʔatʕibbaːʔ (broken M) tʕabiːb-at (F Sg) tʕabiːb-at-aːni (F Dl) tʕabiːb-aːt (SndF) Doctor Two doctors Doctors

b. ʒundiy (M Sg) ʒundiy-aːni (M Dl) ʒunuːd (broken M) ʒundiy-at (F Sg) ʒundiy-at-aːni (F Dl) ʒundiy-aːt (SndF) Soldier Two soldiers Soldiers

c. sʕabiy (M Sg) sʕabiy-aːni (M Dl) sʕibyaːn (broken M) sʕabiy-at (F Sg) sʕabiy-at-aːni (F Dl) sʕabaːyaː (broken F) Youngster Two youngsters Youngsters

129 Chapter Five – Gender and Number

d. baːħiӨ (M Sg) baːħiӨ-aːni (M Dl) baːħiӨ-uːna (SndM) baːħiӨ-at (F Sg) baːħiӨ-at-aːni (F Dl) baːħiӨ-aːt (SndF) Researcher Two researchers Researchers

e. baħӨ (M Sg) baħӨ-aːni (M Dl) ʔabħaːӨ (broken F) Research Two researches Researches

f. fursʕ-at (F Sg) fursʕ-at-aːni (F Dl) furasʕ (broken F) Opportunity Two opportunities Opportunities

g. tʕaːwil-at (F Sg) tʕaːwil-at-aːni (F Dl) tʕaːwil-aːt (SndF) Table Two tables Tables

In the previous analyses, it was shown that gender and number may remain independent

from each other (in singulars, DA duals and DA SndM plurals) or interact with each

other (in SndF plurals). The next chapter will reveal that while gender is independent in

SA duals and SA SndM plurals, number interacts with the case feature of the nouns.

Despite these variations of gender and number, the above examples of singulars, duals,

and sound plurals show that binyanim are not affected and they keep the same shapes

across these nouns. Given that binyanim are the VIs of the [n] morphemes, as discussed

in Chapter Four, it is clear that whatever the status of [Gend] and [Num] is in these

types of nouns, those features cannot influence [n] as long as they remain external to it.

In broken plurals, however, the barriers between gender, number and [n] seem to

be removed. The binyanim of broken plurals do not only express the grammatical

category and the CSFs of their nouns, they also express their gender and number

features, thereby preventing broken plurals from taking any suffixes to express their

gender or plurality.

Given that the structural distinction between sound and broken plurals cannot be

pre-syntactically made, it is assumed that broken plurals are formed as such in the

Morphology. This study adopts Noyer’s (1992) representation of [n] and [Pl] in broken

plurals, shown in (35) above. However, given that Arabic nominal structures need to be

fully specified for gender, as discussed in Section 5.4, and the gender of broken plurals

130 Chapter Five – Gender and Number

is expressed by whatever expresses their [n] and [Pl], Noyer’s representation of broken

plurals can be modified as follows:

(41) [n[CSF], Gend, Pl]

According to the above representation, the three features, which are originally

independent from each other, are fused into one morpheme (henceforth, broken

morpheme) in the Morphology. Consequently, when a broken morpheme is in the

Vocabulary, the binyanim that usually express [n[CSF]] alone are no longer inserted, even

though they match a subset of the features specified in the new morpheme. The new

binyanim that broken plurals take do not only signal that the features of their terminal

morphemes are different, they also indicate the existence of more specified binyanim

(broken binyanim) in the Vocabulary that, according to the Subset Principle (Halle,

1997), win the competition because they match a greater number of features specified in

the broken morphemes than the binyanim that match [n[CSF]] alone.

80 (42) /Broken Binyan/ ↔ [n[CSF], Gend, Pl]

It is likely that among the many broken binyanim available in the Vocabulary,

more than one of them may match a given broken morpheme and have equal eligibility

to be inserted. For example, the broken plurals ʔabħaːӨ (researches) and furasʕ

(opportunities) are feminine, plural, and likely share the same combination of CSFs,

which is an indication that both nouns have identical broken morphemes in their

underlying structures. However, they are made of two different binyanim:

81 /ʔaC1C2aːC3/ and /C1uC2aC3/ respectively. According to Marantz (1995:406), as also

mentioned in Chapter Four, VIs have different contextual features that limit their

insertion in the environment of other VIs. That is, although the broken binyanim

80 Wright (1967:199-234) lists 32 broken binyanim that are commonly used in Arabic nouns and adjectives. The list does not include the binyanim of DA broken plurals, which may vary from SA and from one dialect to another. 81 Binyanim can sometimes contain consonants such as /ʔ/, as mentioned in Chapter Two. 131 Chapter Five – Gender and Number

/ʔaC1C2aːC3/ and /C1uC2aC3/ are perfect matches for a single broken morpheme, the

contextual features of the former limit its insertion to certain environments among

which is the root VI /bħӨ/, while the insertion of the latter is limited to other

environments among which is the root VI /frsʕ/.

If the features constituting broken morphemes are in fact grouped together, this

must result from the activation of a Fusion rule that can be generally represented as

follows:

(43) Broken Formation Rule (BFR):

[n[CSF]] [Gend] [Pl] → [n[CSF], Gend, Pl]

As can be seen, the above rule is unusually broad. The only invariable element of the

rule is [Pl]. Apart from that, there is no particular combination of CSFs or a particular

gender that make the rule more specific. The examples in (40) show that broken plurals

are highly unpredictable. They can be human and nonhuman, as well as masculine and

feminine. The only limitation we know is that nonhuman broken plurals – along with

nonhuman sound plurals – cannot be masculine in gender, as discussed earlier. Based on

this exception, the general BFR above can possibly have three more specific forms, as

shown below:

(44) The Possible Forms of the BFR:

a. [n[+Hum]] [-F] [Pl] → [n[+Hum], -F, Pl] (Human Masculine Broken)

b. [n[+Hum]] [+F] [Pl] → [n[+Hum], +F, Pl] (Human Feminine Broken)

c. [n[-Hum]] [+F] [Pl] → [n[-Hum], +F, Pl] (Nonhuman Feminine Broken)

Although the above BFRs may explain how the features are fused, none of them

actually explains how the Fusion rules are activated in the first place. That is, even

though the specifications of every plural shown in the examples (40) match the

specifications of a BFR, not every plural is broken. For instance, despite the fact that the

132 Chapter Five – Gender and Number

plurals ʔatʕibbaːʔ (doctors), ʒunuːd (soldiers), sʕibyaːn (youngsters) and baːħiӨ-uːna

(researchers) are all human masculine, the former three are pluralised as broken while

the latter is sound. Similarly, the plurals tʕabiːb-aːt (doctors), ʒundiy-aːt (soldiers),

baːħiӨ-aːt (researchers) and sʕabaːyaː (youngsters) are all human feminine, but only the

latter is broken. Finally, the plurals ʔabħaːӨ (researches) and furasʕ (opportunities) are

broken, but tʕaːwil-aːt (tables) is sound although all of them are nonhuman feminine.

The inconsistency of the application of the BFRs is indicative of a possible

selectivity criterion that is external to the features of broken morphemes. The only

variations between sound and broken plurals that share the same CSFs and gender

specifications are the roots. If root morphemes are involved with the activation of the

BFRs, their indices are probably what guide the Morphology to the structures to which

the BFRs should be applied. For example, although the co-occurrence of the

morphemes [n[+Hum]], [-F] and [Pl] in a single structure is a requirement for the BFR

shown in (44.a) to be performed, the activation of the rule seems to be determined by

the appearance of certain root morphemes on the same structure. Within the examples in

ʕ ʕ (40), these root morphemes can be [√(t bb)], [√(ʒnd)] and [√(s by)] from which the broken

plurals ʔatʕibbaːʔ (doctors), ʒunuːd (soldiers) and sʕibyaːn (youngsters) are derived, but

not [√(bħӨ)]. When [√(bħӨ)] occurs in a human masculine plural structure, the BFR is not

activated, and alternatively the whole structure takes a sound form, as seen in baːħiӨ-

uːna (researchers).

If root morphemes can play a role in the activation of the BFRs, this is further

evidence that root indices are able to influence the Morphology. Recall that

impoverishment of the gender morpheme of covert feminine nouns is a morphological

modification applied to certain structures that are probably identified by the indices of

their root morphemes. The formation of broken plurals is, to a large degree, similar to

133 Chapter Five – Gender and Number

the deletion of [+F] in covert feminine nouns. The rules by which broken plurals are

formed and some feminine nouns are made covert are too powerful to be left

unrestrained. Not every structure meeting the specifications of these rules is supposed to

be modified. Certain structures are subject to the rules, and root indices appear to be the

tags that the Morphology needs to identify these structures.

5.7. Default Plural Formation

In the previous two sections, it was shown how sound and broken plurals are formed in

the Morphology. In fact, sound and broken are descriptive terms that make reference to

the status of the singular stems after pluralisation. That is, apart from any affixes, when

a plural noun surfaces with a stem that is identical to the stem of its singular

counterpart, the plural noun is called sound; otherwise, the noun is called broken.82

There could be a number of ways to determine which plural formation strategy is

the default. Given that Arabic stems are generally formed by interlocking root

consonants with binyanim rather than affixation, one could argue that the formation of

broken plurals is the default since it resembles the stem formation strategy, while the

formation of sound plurals is the special case as it involves affixation. One could

instead argue that the default word formation strategy of Arabic nouns, in general, is

mixed. That is, the strategy of stem formation is interlocking roots and binyanim, while

affixation is the default strategy for stems to express their grammatical features (gender

and number, as well as case and definiteness). As far as plural nouns are concerned, this

approach entails viewing the formation of sound plurals as the default, while the

formation of broken plurals as the special case because they do not express their gender

and number features by affixation. A third view may simply depend on the quantitative

82 According to this, dual nouns are also technically sound, but it would appear to be pointless to label them as such since they have no broken counterparts. 134 Chapter Five – Gender and Number

productivity of each plural form to determine which formation strategy is the default in

the language.

McCarthy and Prince (1990b:212-213) employ an approach that is a combination

of the last two views, as shown below:

“Although the term "sound plural" suggests normality – and indeed its form is entirely predictable from gender and other grammatical information – the sound plural is in no way the regular or usual mode of pluralization . . . In Arabic the “special case” system is fully articulated and relatively few items escape it to end up with the default “sound” suffix. For the lexicon as a whole, then, broken plural formation is by far the norm rather than the exception”.

With respect to the formation strategy of each type of plural, the authors consider sound

plurals the default and broken plurals the special case. However, considering the

quantitative productivity of each type, the authors take sound plurals as the exception,

i.e., minority default, while broken plurals are seen to be the norm. McCarthy and

Prince (1990b:210) explain that their generalisation regarding the productivity of each

type is based on scanning a database of about 3500 singular/plural pairs extracted from

the first half of Wehr’s (1971) Dictionary of Modern Written Arabic. The main

objective of their investigation was to provide evidence from broken plurals to support

the theory of Prosodic Morphology, developed in McCarthy and Prince (1986 et seq.).

As a response to the quantitative aspect of McCarthy and Prince’s (1990b)

investigation, Eisele (2002:22-33) provides an extensive critique in which he concludes

that the authors’ investigation is “seriously flawed” for different reasons. Firstly, Eisele

(2002:24) argues that “using a dictionary literally as a lexicon is a very naive mistake”

since dictionaries do not usually list all the possible formations of words.83 Secondly, he

claims that the methodology which the authors followed in defining the domain of

applicability of each type of plural is inconsistent, as the authors classify the domain of

sound plurals on morphological (e.g., deverbals and participles) and semantic (e.g.,

83 For example, most English dictionaries list the irregular plurals feet and geese as separate entries, but not balls or houses. 135 Chapter Five – Gender and Number

proper names and names of the letters) bases, while they classify the domain of broken

plurals on a phonological basis (e.g., all canonically-shaped nouns). Thirdly, Eisele

(2002) points out that the comparison that the authors make between the domain of

Arabic broken plurals they define and the significantly smaller domain of the English

irregular plurals has led to a significant magnification of the presence of broken plurals,

and, consequently, a significant marginalisation of the presence of sound plurals.84

Finally, and to shed even more doubt on the authors’ investigation, Eisele (2002:26)

refers to an investigation conducted by Boudelaa and Gaskell (2000) who counter

McCarthy and Prince’s argument by using a similar dictionary-count approach.

Boudelaa and Gaskell (2000) analysed Khouloughli’s (1992) Basic Lexicon of

Modern Standard Arabic, which appears to include listings of plural forms. Out of the

3000 words that the dictionary describes as the most frequent words in Arabic,

Boudelaa and Gaskell (2000) found that almost two thirds of the 1670 nominal entries

are pluralised in the sound forms. This led the authors to conclude that sound plurals are

more quantitatively productive than broken plurals. The authors describe the opposite

claim as resulting from the practice of Arabic lexicographers who list only broken

plurals because they are unpredictable.

Whether or not the dictionary count approach followed by McCarthy and Prince

(1990b) and Boudelaa and Gaskell (2000) is effective to determine which plural form is

more productive, and which form may be more productive should await further

research. For its purpose, this study will focus on the qualitative investigation for

several reasons. Firstly, determining which plural form is more productive is an issue

that is limited to plural nouns, while this study is concerned with a wider variety of

nouns in which singular and dual nouns are also included. Secondly, finding that broken

84 For more on the proposed domains of applicability and the comparison between Arabic broken plurals and English irregular plurals see McCarthy and Prince (1990b:212-213). 136 Chapter Five – Gender and Number

plurals or sound plurals are more productive in Arabic is a quantitative matter, while

exploring their formation strategy within a specific and strictly formal theoretical

framework, which is one of the goals of this study, is more of a qualitative nature.

Finally, besides Eisele’s (2002) reservations, using dictionaries for determining the

actual weight of certain structures in the lexicon of a given language might add to our

knowledge only if we take that lexicon as the storehouse of phonologically and

semantically specified items. To a study that adopts the theory of DM, such knowledge

is simply irrelevant since what is known as the lexicon is distributed among different

components. Among these components, the closest to the lexicon, in the traditional

sense, is the Narrow Lexicon whose items are neither associated with phonological and

semantic properties nor with any specific type of structure.

Therefore, this study assumes that word formation of Arabic nouns has two

strategies: (a) the stem formation strategy, which involves interlocking roots with

binyanim, that applies to all types of nouns; (b) the affixation strategy by which stems

express their grammatical features, which is the default mechanism that applies to all

grammatical features in all types of nouns except for gender and number in broken

plurals. This is what makes broken plurals the special case.

The way broken plurals deviate from the default formation strategy of Arabic

nouns with respect to gender and number can be represented as the following:

(45) a. [√(index)] + [n[CSF]] + [Gend] + [Num] (Default)

Consonants + Binyan Stem + Affix + Affix

b. [√(index)] + [n[CSF], Gend, Pl] (Broken Plurals)

Consonants + Binyan Stem

137 Chapter Five – Gender and Number

As shown above, broken plurals are the type of Arabic nouns whose internal structure

deviates from the default formation strategy of the language by having the [Gend] and

[Pl] morphemes fused with [n], as discussed in Section 5.6. Although the other types of

nouns, including sound plurals, may also undergo some morphological modifications,

these modifications do not affect their [n] morpheme or cause any changes that result in

a deviation from the default formation strategy of the language. For example, although

[+F] and [Pl] in SndF plurals are fused, as claimed in Section 5.5.1, the overall structure

of the nouns preserves the major characteristics of the default formation strategy by

having their grammatical features (of gender and number) expressed externally to their

stems (root and binyanim) by affixation. In the next chapter, it will be revealed that SA

SndM plurals follow the same strategy despite the morphological modifications that

affect their number, case and indefiniteness features.

Given that plurals are made sound or broken in the Morphology, considering

sound plurals the default and broken plurals the special case will have some

implications on classifying the rules by which each type is formed, as will be discussed

in the next section.

5.8. Insight into the Morphology

In the past sections, we have come across a number of morphological rules that may

explain how gender and number are handled in the Morphology. In the light of these

rules, this section aims to explore the Morphology of Arabic and how these rules might

be organised in it. Below are the rules that have been discussed so far:

138 Chapter Five – Gender and Number

(46) Gender and Number Rules in the Morphology of Arabic:

a. [-F] → [+F] / [n[-Hum]] + ____ + [Pl] ‘Revaluation in Nonhuman Masculine Plurals’

b. [+F] → Ø ‘Impoverishment in Covert Feminine Nouns’

c. [n[CSF]] [Gend] [Pl] → [n[CSF], Gend, Pl] ‘General BFR’

d. [+F] [Pl] → [+F, Pl] ‘Fusion in SndF plurals’

Starting with the rule in (46.a) above, it is a filter rule because it is motivated to

eliminate the prohibited co-occurrence of [n[-Hum]], [-F] and [Pl] in one structure. As a

result of applying the rule, the [-F] morpheme undergoes a Revaluation process that

changes its value into [+F] allowing nonhuman plurals to surface as feminine.85

If modifying post-syntactic structures in the Morphology is carried out through

phases, fulfilling the requirement of the filter is assumed to take place in the first phase.

Prohibited co-occurrences are possibly the first thing the Morphology needs to detect

before any further modifications can be applied. For example, if application of the rule

in (46.a) does not take place first, nonhuman plural structures with [-F] will remain

masculine and, consequently, they will escape the SndF rule in (46.d) since their gender

morphemes are not specified as [+F]. If this is what actually happens, we would expect

SndF nouns to be strictly limited to plurals that descend from feminine origins, which is

not the case. It was shown earlier that many nonhuman plurals of masculine origins can

take the SndF form, e.g., muħitʕ (ocean ‘M Sg’) à muħitʕ-aːt (oceans ‘SndF’). In fact,

applying the filter rule first is what paves the way for the gender and number features of

these nouns to be fused by the SndF rule. In other words, changing [-F] into [+F] in

(masculine) nonhuman plural structures is what creates the environment in which some

of these structures are formed in the SndF form.

85 Recall that restricting the Syntax from building nonhuman masculine plural structures cannot be acceptable. If these structures are prohibited from surfacing as masculine, the Morphology is the ideal location for the occurrence of gender changing. 139 Chapter Five – Gender and Number

As to the agreement level, it is assumed that this level is located immediately after

all the prohibited co-occurrences are terminated, i.e., immediately after the filter rules

and prior to the application of any further modification. Having the agreement level

immediately after the phase of filter rules accounts for the feminine gender of the

targets of nonhuman plurals in the most economical way. This location entails that

syntactically masculine nonhuman plurals change their gender into feminine before

their gender feature is copied onto the agreement node of their targets. If the agreement

level is prior to this phase, the targets of nonhuman masculine plurals will agree as

masculine, and then Revaluation will need to apply to the gender features of both the

plurals and their targets.

In addition, having the agreement level at this location explains why the targets of

covert feminine nouns can be overtly feminine although the gender feature of their

controllers is impoverished. Recall that covert feminine nouns undergo a morphological

modification by which their [+F] morphemes are impoverished, but the nouns still agree

with their targets as feminine. If impoverishment of the feminine feature of covert

feminine nouns takes place before the agreement level, the targets must not be able to

surface as overt feminine since the [+F] feature of their controllers, which will have

already been impoverished, will not be available to be copied onto the targets’

agreement nodes. However, if agreement takes place before impoverishment of the [+F]

morphemes of the covert nouns, the feature will have been already copied onto the

targets by the time it is impoverished. Thus, this is the location this study proposes for

the agreement level.

The rules in (46.b-c) represent the operations by which feminine nouns are made

covert and plural nouns are made in the broken form, respectively. These rules are more

general and, as they stand, can be applied to virtually every structure that matches their

140 Chapter Five – Gender and Number

specifications, which is not actually the case. Not every feminine noun is covert and not

every plural is broken. Therefore, it was assumed that the choice of the feminine and the

plural structures that are subject to these rules is determined by the Morphology – as a

language-specific component – with a possible signal from the root indices. Due to the

unpredictability of their activation, this study assumes that these rules are the special

rules of the language. That is, covert feminine and broken plural nouns are special

structures formed by special rules, while nouns whose structures are not subject to these

rules are consequently the default. Among the nouns discussed so far, default nouns

include singulars, DA duals and DA SndM plurals – whose structures do not appear to

be altered in the Morphology – as well as SndF plurals86 whose [+F] and [Pl] are always

fused by the rule in (46.d), which is considered one of the default rules of the language.

By comparing the special (broken) rule in (46.c) to the default (SndF) rule in

(46.d), it becomes clear that one of them must suppress the other. The BFR combines

[Gend] with [n] and [Pl], so the SndF rule cannot be activated as [+F] and [Pl] are no

longer independent as required. The SndF rule, on the other hand, combines [+F] and

[Pl], so the BFR, in the case of a broken feminine plural, cannot be activated as the

[Gend] and [Pl] morphemes are no longer independent. As mentioned earlier, McCarthy

and Prince (1990b:213) state that “in Arabic the ‘special case’ system is fully articulated

and relatively few items escape it to end up with the default ‘sound’ suffix”. Regardless

of the word few that follows from a quantitative viewpoint, as discussed in Section 5.7,

this study also takes the default sound plurals as the structures that escape the special

system of broken plurals.

Therefore, as far as the Morphology of Arabic is concerned, every plural structure

is examined at a certain phase in the Morphology by a BFR. If the structure somehow

86 Recall that sound plurals follow the default formation strategy in Arabic, as illustrated in Section 5.7. 141 Chapter Five – Gender and Number

activates the BFR, a broken plural is formed. Otherwise, the structure escapes the rule

and, consequently, defaults to the sound form. For example, if, on the one hand, the

morphemes [n[-Hum]], [+F] and [Pl] of the feminine broken plurals ʔabħaːӨ (researches)

and furasʕ (opportunities) are not fused first, the latter two morphemes will remain

independent and will, consequently, be sufficient to activate the default SndF rule

whose application will undesirably produce *baħӨ-aːt and *fursʕ-aːt. On the other hand,

the application of the special BFR to the structure of these nouns first will prevent the

production of the ill-formed nouns by disrupting the independence of their [+F] and [Pl]

morphemes, so the default SndF rule cannot be activated. Hence, this study assumes

that the phase where plural structures are examined for their capability of activating a

special BFR is immediately above the phase where the structures can default to the

relevant sound form. In general, special rules that conflict with the default rules of the

language, e.g., BFRs vs. SndF rule, take precedence in order to be able to suppress the

default rules in question.

Considering the phases proposed in this section, it can be assumed that the

Morphology of Arabic is internally organised as shown in the following figure:

Figure 5.2: Gender and Number Rules across the Morphological Phases Phase One: Filter Rules

[-F] → [+F] / [n[-Hum]] + ____ + [Pl]

Agreement Level

Phase Two: Special Rules

[+F]→ Ø [n[CSF]] [Gend] [Pl] → [n[CSF], Gend, Pl]

Phase Three: Default Rules [+F] [Pl] → [+F, Pl]

142 Chapter Five – Gender and Number

In the analysis of SA dual and SndM plural nouns in the next chapter, we will come

across more rules that need to be applied to the structure of these nouns in the third

phase. It will be shown that the activation of some of these rules is conditioned by the

application of other rules, which lends further support to the proposed internal

organisation of the Morphology.

Finally, it is important to emphasise that the assumed organisation of Arabic

Morphology is based solely on analysing Arabic nouns. The validity of this assumption

needs to be examined through analysing the remaining parts of speech of the language,

particularly the formation of verbs. Moreover, further cross-linguistic studies may or

may not support the assumed organisation. However, if this, or a similar, organisation is

found in other languages, it may indicate that although the Morphology is a language-

specific component that contains the local morphological rules of the language, it is

probably organised in a universal way. However, due to the limitations of this thesis,

these issues will not be pursued here.

5.9. Conclusion

This chapter showed that gender and number in Arabic nouns are independent syntactic

features that are subject to various modifications in the Morphology. These

modifications include impoverishing the feminine feature of covert feminine nouns,

changing the masculine feature of nonhuman plurals into feminine, fusing the feminine

and plural features of SndF plurals, and fusing the gender and plural features of broken

plurals with [n]. The modifications were assumed to be performed by the standard DM

operations, except with gender changing in nonhuman plurals.

It was assumed that gender changing in nonhuman plurals is motivated by a

morphological filter that restricts the co-occurrence of [n[-Hum]], [-F] and [Pl] in one

structure. Thus, in order for the structure to comply with the requirements imposed by 143 Chapter Five – Gender and Number

the filter, [-F] must change into [+F]. This study argued that this modification is carried

out by a new morphological operation, labelled Revaluation, that simply changes the

value of the feature. It is important to emphasise that proposing this operation is based

empirically on the behaviour of a particular type of nouns in Arabic. Further evidence

from within Arabic and other languages would be needed to establish the rule on a

firmer ground.

The modifications that nominal structures may undergo in the Morphology

indicate that this component of DM works in a certain sequence of phases. It was

assumed that filter rules by which prohibited co-occurrences are eliminated are the first

priority in the Morphology. When prohibited co-occurrences are eliminated or not

detected, structures enter the agreement level in which certain nominal features are

copied onto the agreement nodes of their targets. The agreement level is followed by a

phase in which the special rules of the language are applied. These rules may render

feminine nouns covert or plural structures broken. In the subsequent phase, the

application of the only default rule we have so far, which is the SndF rule, is

conditioned by whether or not the special broken rule has been activated in the previous

phase. If the broken rule has been activated, the SndF rule is blocked as the morphemes

it operates on are no longer independent. If, on the other hand, the broken rule has not

been activated, plural structures default to their sound forms. The appropriate sound

form is then determined by the value of the gender morphemes. That is, the SndF rule is

activated whenever the gender morpheme of the plural structure is specified as [+F].

Otherwise, another default rule that yields SndM plurals in SA will be activated, as will

be shown in the next chapter.

Having all the necessary morphological modifications performed in the

Morphology, structures proceed to the Vocabulary in which their gender and number

144 Chapter Five – Gender and Number

morphemes, be they modified or not, are phonologically represented as shown in the

following table:

Table 5.3: Gender and Number VIs in SA and DA Nouns SA Gender and Number VIs (excluding number in duals and SndM plurals) /at/ ↔ [+F] • Gender in feminine singulars and duals. /aːt/ ↔ [+F, Pl] • Gender and number in SndF plurals.

/Broken Binyan/ ↔ [n[CSF], Gend, Pl] • Gender and number in broken plurals. • Gender in masculine singulars, duals and SndM plurals. /Ø/ ↔ elsewhere • Gender in covert feminine nouns. • Number in singulars. DA Gender and Number VIs /at/ ↔ [+F] • Gender in feminine singulars and duals. /aːt/ ↔ [+F, Pl] • Gender and number in SndF plurals.

/Broken Binyan/ ↔ [n[CSF], Gend, Pl] • Gender and number in broken plurals. /ain/ ↔ [Dl] • Number in duals /iːn/ ↔ [Pl] • Number in SndM plurals • Gender in masculine singulars, duals and SndM plurals. /Ø/ ↔ elsewhere • Gender in covert feminine nouns. • Number in singulars.

If no post-insertion readjustment process is required, e.g., changing the standard

feminine VI /at/ into the pause form, the VIs inserted into the above gender and number

morphemes are ready to surface.

145 Chapter Six – Case and Definiteness

6. Chapter Six – Case and Definiteness

6.1. Introduction

The main objective of this chapter is to provide a DM analysis for the various

behaviours that Arabic nouns show with respect to their case and (in)definiteness

features. In the previous chapters, the structure of Arabic nouns in terms of the

formation of their stems and the several ways they express their gender and number

features was discussed. Case and (in)definiteness will be discussed in this chapter as the

last two pieces constituting the structure of Arabic nouns.

In contrast with gender and number, case and (in)definiteness are more overtly

marked in SA nouns than in DA. Thus, the SA will be the main focus in this chapter.

Moreover, as compared to gender and number, most of the changes that case and

(in)definiteness undergo are more phonologically than morphologically motivated. This

might have to do with the fact that the phonological exponents of those features are

usually positioned at the edges of the nouns, so they become affected by the

phonological elements of the surrounding words. In terms of their morphological

interactions, case is found to be associated with number only in SA duals and SndM

plurals, while (in)definiteness is always disassociated.

This chapter is organised as follows. Section 6.2 investigates the case-marking

system of SA singulars and broken plurals, as well as the question of why SndF plurals

deviate from the system when accusative. The section also investigates the case-

marking system of SA duals and SndM plurals, the interaction between their case and

number features, and the implications of this interaction to the Morphology and the

Vocabulary. The overall picture that emerges from this section is a DM account for the

formation of SA nouns with respect to case and (in)definiteness and a comprehensive

146 Chapter Six – Case and Definiteness

account of VI competition in the domain of case that represents the standard case-

marking system of SA.

Section 6.3 addresses the (in)definiteness features of SA nouns. It provides a DM

analysis for the presence of the definite prefix al- and the indefinite suffix -n (known as

nunation) on most SA nouns. This section also addresses the absence of nunation from

indefinite duals and SndM plurals, and shows that this absence is morphologically

motivated. Moreover, by treating nouns as DPs (Abney, 1987), both affixes will be

considered the phonological representations of the (in)definiteness features generated

under the D(eterminer) node. This view of the position of the features and the treatment

of nouns as DPs will be supported by showing how genitive constructions, known as

construct states (CSs), are structured. In Arabic and Modern Hebrew, CSs have a very

large degree of similarity, so an analysis of Hebrew CSs by Ritter (1988, 1991), which

has been widely adopted in the literature of Arabic linguistics, will be presented and

adopted. It accounts for the structure of CSs and, most importantly, for the absence of

the (in)definiteness features from their head nouns as well as for the assignment of the

genitive case to their complement nouns.

In the light of the analyses provided in the previous chapters as well as Sections

6.2 and 6.3, Section 6.4 explores the basic internal structure of Arabic Ns. With the

(in)definiteness features being generated under D and outside N, this section shows how

stem constituents and the features of gender, number and case are organised inside N to

allow the modifications by which the different types of nouns are formed to take place.

Section 6.5 focuses on a type of SA singulars, broken plurals and names known as

diptotes. These structures do not accept nunation and they take the usual accusative case

marker when oblique and genitive when their nunation does not show up. An attempt by

Embick and Noyer (2007) to employ the operation of Impoverishment to account for the

147 Chapter Six – Case and Definiteness

behaviour of diptotes will be reviewed. It will be shown that diptotes cannot be formed

as such in the Morphology and their behaviour is likely to be phonologically motivated

after the insertion level in the Vocabulary. Some evidence to counter the morphological

analysis will be presented, but it will remain unclear how exactly this behaviour is

motivated in the Vocabulary.

Section 6.6 addresses some other types of SA singulars and broken plurals known

as defective, indeclinable and invariable nouns. In several ways, these nouns deviate

from the standard case and indefiniteness marking systems of SA. By relying on an

analysis provided by Abboud and McCarus (1983), which addresses the first two types,

it will be shown that the structure of these nouns does not undergo any morphological

modifications with respect to case and indefiniteness. Rather, the nouns deviate from

the standard case and indefiniteness systems for purely phonological reasons that can be

discovered only after the insertion level. Moreover, plural defective nouns, which are

diptotes at the same time, will provide much stronger evidence against the

morphological treatment of diptotes proposed by Embick and Noyer (2007).

Section 6.7 is concerned with the case and (in)definiteness features of DA nouns.

It will be shown that although DA nouns have maintained the definite marker /al/, they

take no case suffixes or nunation. It will be shown that case and indefiniteness in DA

have become silent features as no phonological representations are available in the

Vocabulary of DA for them. The absence of case markers and nunation eliminates the

differences between regular nouns, on the one hand, and diptotes, defectives,

indeclinables and invariables, on the other. Finally, Section 6.8 concludes the chapter

and summarises its findings.

148 Chapter Six – Case and Definiteness

6.2. Case in SA Nouns

According to Blake (2001:1), “case is a system of marking dependent nouns for the type

of relationship they bear to their heads”. In SA, there are various environments in which

nouns are overtly marked (by suffixation) for case. Ryding (2005:168-182) provides an

extensive list and examples of these environments under the assumption that SA nouns

are overtly marked for three cases: nominative, accusative and genitive. The latter case

includes complement nouns in CSs as well as nouns that are complements of

prepositions, both exhibiting identical case marking.

Following some standard assumptions, e.g., Chomsky (1981), this study takes the

case feature of preposition complements as oblique. The case feature of complement

nouns in CSs will remain to be considered genitive. However, since there is no

phonological distinction between these two cases, they will simply be referred to as

oblique/genitive or [Obl/Gen] unless a distinction needs to be made. Thus, in this study,

SA nouns are considered to be overtly marked for four cases: nominative, accusative,

oblique and genitive. As a general rule, an SA noun in a sentence is nominative in the

subject position, accusative in the object position, oblique when modified by a

preposition, and genitive when a complement noun in a CS. Moreover, when a noun is

used in isolation (citation form) or in a verbless sentence, 87 its case is usually

nominative.

There are various approaches from different theoretical perspectives as to how

case is generally assigned (see, for example, Chomsky, 1981, 1995b; Bittner and Hale,

1996; Marantz, 2000) and different proposals as to how case is assigned to Arabic

87 An English sentence such as the man is a teacher has no present tense copula in Arabic, so it is verbless: al-raʒul-u muʕallim-u-n Def-man-Nom teacher-Nom-Idef

149 Chapter Six – Case and Definiteness

nouns in particular. These proposals cover the general case assignment mechanism in

Arabic (Homeidi, 1994; Soltan, 2007; Al-Balushi, 2011;) with particular reference to

case (and agreement) in VSO/SVO word order (Mohammad, 1990; Fassi Fehri, 1993,

2012; Aoun, Benmamoun, and Sportiche, 1994; Ouhalla, 1994; Benmamoun, 1999,

2000; Harbert and Bahloul, 2002; Al-Shorafat, 2012), case in verbless sentences (Eid,

1991; Fassi Fehri, 1993; Benmamoun, 2000, 2008; Ouhalla, 2005; Al-Horais, 2006;

Bahloul, 2008), as well as the genitive case (and definiteness) in CSs, which will be

addressed in Section 6.3.2.

For lack of space, this study will not go into details of this rich literature on the

syntactic properties of case assignment, but rather will focus on how case is handled in

the Morphology and how it is phonologically represented in the Vocabulary after the

assignment, regardless of how it is assigned. The following subsections include a DM

account for the standard case-marking system of the SA nouns. After discussing the

(in)definiteness feature in Section 6.3, nouns that deviate from both the case and the

indefiniteness marking systems of the language will be addressed.

6.2.1. Singular and Broken Plural Nouns

Most SA singular and broken plural nouns exhibit a three-way case-marking system, in

which a noun takes the suffix -u when nominative, -a when accusative, and -i when

oblique/genitive. Consider the case-marking system of the following singular nouns:

(1) a. χaːdara ʔammaːn-a musaːʕid-u al-waziːr-i left Amman-Acc assistant-Nom Def-minister-Gen ‘The minister’s assistant left Amman.’

b. ħasʕaltu ʕalaː al-ʒinsiy-at-i obtained.I on Def-citizenship-F-Obl ‘I obtained the citizenship.’

In the above example, the object ʔammaːn (Amman) is followed by the accusative

suffix -a, the subject musaːʕid (assistant) is followed by the nominative suffix -u, and

150 Chapter Six – Case and Definiteness

the noun waziːr (minister), which is the possessor noun of a CS, and the noun ʒinsiy-at

(citizenship), which is an object of a preposition, are followed by the oblique/genitive

suffix -i. Observe that the case suffixes are independent and not associated with any

other features of their nouns. An identical case-marking system is also found in SA

broken plurals, as shown below:

(2) a. ʔiʃtara al-tʕulaːb-u al-kutub-a bought Def-students-Nom Def-books-Acc ‘The students bought the books.’

b. al-tadʕaːmun-u maʕa al-ʕummaːl-i Def-solidarity-Nom with Def-workers-Obl ‘The solidarity with the workers.’

c. baːq-aːt-u al-zuhuːr-i bouquet-F.Pl-Nom Def-flowers-Gen ‘Bouquets of flowers.’

The behaviour of SA singulars and broken plurals towards case can be an

indication that, during the word formation process, the morpheme that carries the case

feature, i.e. [K]88, of these nouns remains independent. It is not affected by any

modification the structures may undergo in the Morphology. Depending on the function

of the nouns in the sentence, the [K] morpheme can be specified for different types of

case features that are phonologically expressed in the Vocabulary as the following:

(3) /u/ ↔ [Nom]

/a/ ↔ [Acc]

/i/ ↔ [Obl/Gen]

The above VIs of case represent the suffixes that singular and broken plural nouns take

to express their case features. It can be seen that each VI above corresponds to a single

case feature, except /i/ that corresponds to two case features. This may indicate that /i/ is

the elsewhere VI of case in SA. This assumption, which contrasts with Embick and

88 K is an abbreviation of case as found in Bittner and Hale (1996).

151 Chapter Six – Case and Definiteness

Noyer (2007) as will be discussed in Section 6.5, will be supported by evidence from

the case-marking system of SndF plurals in the next subsection.

6.2.2. SndF Plural Nouns

SA SndF plurals show a slightly different case-marking system than the system of

singulars and broken plurals. SndF plural nouns take the suffixes -u when nominative

and -i when oblique/genitive, as shown in (4), but they do not take the usual a- suffix

when accusative. Alternatively, when SndF plurals occur in an accusative position, they

take the suffix -i instead of -a, as shown in (5) below:

(4) a. ʔaʕlanat al-wilaːy-aːt-u al-mutaħid-at-u taxfiːf-a announced Def-state-F.Pl-Nom Def-united-F-Nom easing-Acc ʔuquːb-aːt-i-haː sanction-F.Pl-Gen-its ‘The United States announced the easing of its sanctions.’

b. fiː buħayr-aːt-i ʔifriːqiyaː89 in lake-F.Pl-Obl Africa ‘In the lakes of Africa.’

(5) a. taʕlama kalim-aːt-i-n maʕduːd-at-a-n90 learned.he word-F.Pl-Acc-Idef counted-F-Acc-Idef ‘He learned a few words.’

b. yuʒriː ʔitisʕaːl-aːt-i-n run.he contact-F.Pl-Acc-Idef ‘He is making contacts/phone calls.’

As shown above, the nominative and oblique/genitive SndF nouns in (4) follow the

same case-marking system found in singulars and broken plurals, while the accusative

SndF plurals do not. The SndF nouns kalim-aːt (words) and ʔitisʕaːl-aːt in (5) are in the

position of object nouns, but instead of taking the usual accusative suffix -a, they take

the suffix -i, which is the oblique/genitive case marker in singulars and broken plurals.

89 ʔifriːqiyaː (Africa) is not overtly marked for case because it belongs to the class of invariable nouns, as will be discussed in Section 6.6. 90 Observe that the SndF noun and its adjective are both accusative with two different markers. The adjective follows the standard case-marking system since it is singular (deflected agreement), while the SndF noun deviates from it only when accusative. 152 Chapter Six – Case and Definiteness

SndF nouns appear to have lost the distinction between accusative and

oblique/genitive cases. Alternatively, they show a two-way case-marking system that

distinguishes only between nominative and non-nominative cases. That is, /u/ becomes

the most specified VI whose insertion is conditioned by the presence of [Nom], while /i/

is simply the elsewhere VI that is inserted otherwise. Having /i/ as the elsewhere VI

captures its appearance in two environments (oblique and genitive) in singular and

broken plural nouns, which supports the earlier assumption that /i/ is also the elsewhere

VI in these nouns. Based on this, the phonological entries shown in (3) above can be

updated as follows:

(6) /u/ ↔ [Nom]

/a/ ↔ [Acc]

/i/ ↔ elsewhere

Although the above representation captures the case-marking system of singulars and

broken plurals, it is still far from capturing the behaviour of SndF plurals towards case,

and specifically the accusative case. In fact, it gives rise to the question: why is /a/ not

inserted when the case morpheme of SndF plurals is specified as [Acc]?

This study argues that in order for the elsewhere /i/ to be inserted into a given case

morpheme, the morpheme must not be specified as [Nom] or [Acc] unless it has been

impoverished by the time it enters the Vocabulary. The presence of /i/ in accusative

SndF plurals is indicative that their [Acc] feature is impoverished, and /i/ is inserted by

virtue of being the elsewhere VI of case. As to how Impoverishment takes place and

why it targets the [Acc] feature in SndF plurals but not the same feature in singulars and

broken plurals, this analysis assumes that impoverishing the [Acc] feature in the

Morphology is conditioned by the existence of the morpheme [+F, Pl], which is the

153 Chapter Six – Case and Definiteness

distinctive characteristic of the structure of SndF plurals, as discussed in the previous

chapter. An Impoverishment rule illustrating this process can be as shown below:

(7) [Acc] → Ø / [+F, Pl] + ______

The above rule entails that [K] is impoverished if specified as [Acc] in the environment

of [+F, Pl]. As a result, the insertion of /a/ will be blocked in the Vocabulary, and the

insertion of the elsewhere /i/ will take place instead.

Should this turn out to explain why SndF plurals do not receive the standard

accusative case marker, it may provide a further insight into how the Morphology might

be organised. Recall that the morpheme [+F, Pl], which is the environment that

conditions the activation of the above rule, is itself created by a default (Fusion) rule in

the last morphological phase proposed in the previous chapter. This can be an indication

that the Morphology does not stop modifying structures once an operation has been

performed. Rather, it is active until all necessary modifications are accomplished. More

importantly, it suggests the existence of a certain order inside that phase since the

activation of one operation is conditioned by the accomplishment of another operation.

This order does not seem to be related to the nature of the operations themselves,

e.g., Fusion is hypothetically stronger than Impoverishment or vice versa. Rather, it

appears to follow from the order according to which the targeted morphemes are

merged in the Syntax. This issue will become clearer when the final structure of Arabic

nouns is determined in Section 6.4.

6.2.3. Dual and SndM Plural Nouns

The case feature of SA duals and SndM plurals shows a strong association with the

number feature of these nouns. These nouns are composed of basic stems followed by a

suffix that can take four different forms depending on the number feature of the nouns

and whether the nouns are in a nominative or non-nominative position. Although the

154 Chapter Six – Case and Definiteness

nouns have a similar behavior to SndF plurals in the sense that they make no distinction

between the accusative and oblique/genitive cases, duals and SndM plurals show two

different case-marking systems that are also different from that of SndF plurals due to

the involvement of their number features.

In SA dual nouns, case does not seem to enjoy the independence it has in singular,

broken plural, and SndF plural nouns. These nouns express their duality and case by one

suffix, which is either -aːni when the nouns are nominative, as shown in (8), or -ayni

otherwise, as shown in (9) below:

(8) hunaːka mawdʕuːʕ-aːni muhim-aːni there subject-Dl.Nom important-Dl.Nom ‘There are two important subjects.’

(9) a. ʕayanat dawl-at-u al-ʔimaːr-aːt-i safiːr-at-ayni appointed state-F-Nom Def-emirate-F.Pl-Gen ambassador-F-Dl.Acc ‘(The state of) Emirates appointed two female ambassadors.’

b. bayna safiːr-ayni between ambassador-Dl.Obl ‘Between (two) ambassadors.’

c. siʕr-u al-kitaːb-ayni price-Nom Def-book-Dl.Gen ‘The price of the two books.’

SA SndM plurals show the same behaviour towards number and case, but with a

different set of suffixes. The nouns take the suffix -uːna in the nominative case, and the

suffix -iːna in all non-nominative cases. Consider the following examples:

(10) muraːqib-uːna rasmiy-uːna observer-Pl.Nom official-Pl.Nom ‘Official observers.’

(11) a. faransaː tuħaðiru al-naː∫itʕ-iːna al-ʔislaːmiy-iːna France warn Def-activist-Pl.Acc Def-islamic-Pl.Acc ‘France warns the Islamic activists.’

b. ʕadad-u-n mina al-baːħiӨ-iːna al-misʕriy-iːna number-Nom-Idef of Def-researcher-Pl.Obl Def-egyptian-Pl.Obl ‘A number of Egyptian researchers.’

155 Chapter Six – Case and Definiteness

c. ziyaːd-at-u-n fiː ħuquːq-i al-musaːhim-iːna increase-F-Nom-Idef in rights-Obl Def-shareholder-Pl.Gen ‘An increase in the shareholders’ dividends.’

Removing the suffixes -aːni and -ayni from the dual nouns or -uːna and -iːna from the

SndM plural nouns above changes the nouns into singulars that would, consequently,

take the case suffixes -u, -a, and -i. The appearance of the dual and SndM suffixes

prevents the nouns from being marked for case by these suffixes.

To account for the joint number/case-marking system of dual and SndM nouns, it

is assumed that the above suffixes are the phonological representations of morphemes

that are specified for both features, i.e., [Num, K] morphemes. Of these, [Num] is

specified as [Dl] in the dual and as [Pl] in the SndM plural nouns, while [K] can be

specified with any of the available case features in the language, i.e., [Nom], [Acc],

[Obl], and [Gen]. However, the Vocabulary does not appear to contain distinct VIs for

the non-nominative features when combined with [Num], so, for the purpose of this

study, the non-nominative features of these nouns will simply be labeled as [-Nom]

when referred to within the domain of the Vocabulary, as shown below:91

(12) /aːni/ ↔ [Dl, Nom]

/ayni/ ↔ [Dl, -Nom]

/uːna/ ↔ [Pl, Nom]

/iːna/ ↔ [Pl, -Nom]

Assuming that [Num] and [K] are originally independent, their presence in a

single morpheme indicates a modification to the structure of dual and SndM nouns in

the Morphology. For SA dual nouns, such a modification takes place in a

straightforward fashion. Since gender is not involved and dual nouns come in one form,

i.e., there are no broken and sound duals, it can be considered a default rule that every

91 The /ni/ and /na/ parts of the dual and SndM suffixes are dropped when they occur as part of the head nouns in CSs. This behavior will be discussed in the next chapter. 156 Chapter Six – Case and Definiteness

[Dl] morpheme is fused in the Morphology with [K] regardless of the specification of

the latter, as shown below:

(13) [Dl] [K] → [Dl, K]

Observe that in order for the above rule to be activated, [Dl] and [K] must be originally

independent when they arrive in the phase of default rules in the Morphology.

Whenever these morphemes are present as such on a structure, the Fusion rule is

activated and the features are combined into one morpheme.

Recall that dual nouns in DA are usually expressed by the suffix -ain that has no

variants. This was taken as an indication that [Dl] is an independent morpheme that is

always represented with /ain/ in the Vocabulary of DA. That is, the above Fusion rule is

a property of the Morphology of SA, but probably not the Morphology of DA.

The formation of SA SndM plurals, in terms of number and case, is more complex

than the formation of the duals. It was shown in the previous chapter that plural

structures are formed either as broken, i.e., by a special BFR that fuses their [n], [Gend]

and [Pl], or as sound. It was also argued that sound plurals, including SndM, represent

the default formation strategy of the language. As far as sound plurals are concerned,

gender appears to play a major role in determining how the constituents of their

structures must be handled in the Morphology. When the gender morpheme is specified

as [+F], a SndF plural is formed by fusing its [+F] and [Pl], as discussed in the previous

chapter. However, when the gender morpheme is specified as [-F], the number and case

features will be fused,92 as shown below:

(14) [Pl] [K] → [Pl, K] / [-F] + _____93

92 This fusion was briefly indicated by Noyer (1992) in his analysis of the nature of [Pl] in broken and sound plurals, as mentioned in the previous chapter. 93 This is a further support to the assumption made in the previous chapter that Arabic nouns are fully specified for gender, including the masculine gender [-F]. Without [-F], every [Pl] that is not part of a broken plural structure would be fused with [K]. Assuming that fusing [Pl] and [K] is conditioned by the presence of [-F] excludes the [Pl] feature of SndF plurals and limits the application of the rule to the SndM ones. 157 Chapter Six – Case and Definiteness

The above rule entails that whenever an SA structure enters the phase of the default

rules carrying independent number and gender morphemes that are specified as [Pl] and

[-F] respectively, the former is fused with [K], forming what is known as SndM plural

and allowing the appropriate SndM number/case suffix to be inserted.

SndM plurals in DA differ from their SA counterparts by taking the suffix -iːn,

which does not show any interaction with case, as illustrated in the previous chapter. As

with DA duals, this can be an indication that [Pl] in DA SndM plurals remains an

independent morpheme, which makes the above SndM rule a property of the

Morphology of SA, but not the Morphology of DA.

If the analyses of case provided so far with respect to SA nouns are correct, a

comprehensive list of case VIs, representing the standard case-marking system of SA,

can be presented as follows:

(15) /aːni/ ↔ [Dl, Nom]

/ayni/ ↔ [Dl, -Nom]

/uːna/ ↔ [Pl, Nom]

/iːna/ ↔ [Pl, -Nom]

/u/ ↔ [Nom]

/a/ ↔ [Acc]

/i/ ↔ elsewhere

According to the Subset Principle (Halle, 1997), a VI is eligible for insertion into a

morpheme only when (a) it matches all or a subset of the features specified in the

morpheme, and (b) it does not correspond to a feature that is not present in the

Morpheme. That is, when a morpheme such as [Dl, Nom] is in the Vocabulary, /aːni/

and /u/ will be eligible as each VI is specified for [Nom]. All the other VIs will not be

eligible as they match no features, or correspond to features that are not present in the

158 Chapter Six – Case and Definiteness

morpheme. The eligible VIs /aːni/ and /u/ will enter a competition and, according to the

Subset Principle, the former will be inserted as it matches the greatest number of

features specified in the morpheme in question. In a representation where a case

morpheme is impoverished, e.g., [Acc] in SndF plurals, the elsewhere /i/ will be

inserted since all other VIs will become ineligible because they correspond to features

that are not present in the morpheme. As discussed earlier, this explains why SndF

plurals take the suffix -i instead of -a when accusative.

6.3. (In)Definiteness in SA Nouns

As might have been noticed in the SA examples provided throughout this thesis so far,

SA nouns generally express their definiteness by taking the prefix al- and their

indefiniteness by taking the suffix -n, the latter known as nunation.94 The definite prefix

al- is allowed to attach to all types of nouns unless the noun is the head of a CS. The

absence of nunation, on the other hand, follows a more complex pattern. Nunation is not

allowed to appear on all duals and SndM plurals, some singulars and broken plurals

known as diptote and invariable nouns, in addition to any noun that is a head of a CS.

This section aims to examine the standard (in)definiteness-marking system of SA

nouns, the position where the (in)definiteness features are originally generated, why

nunation does not appear on indefinite duals and SndM plurals, and why al- and

nunation do not appear on head nouns of CSs. This section relies on treating nouns as

DPs (Abney, 1987) in determining the position of the (in)definiteness features, and on

an analysis of CSs in Modern Hebrew (Ritter, 1988, 1991) in accounting for the

elimination of the features in head nouns as well as for how CSs are generally

94 Nunation is the English equivalent to tanwiin in Arabic. It is likely to have been named as such after the /n/ sound it adds to the nouns. 159 Chapter Six – Case and Definiteness

structured. The absence of nunation and the unusual case-marking system in SA

diptotes and invariables will be discussed in the subsequent sections.

6.3.1. The Standard (In)Definiteness-Marking System in SA

As mentioned in Chapter Two, all types of SA nouns can take the prefix al- when

definite, as shown below:

(16) a. fiː al-ðʕalaːm-i (Singular) in Def-darkness-Obl ‘In the darkness.’

b. siʕr-u al-kitaːb-ayni (Dual) price-Nom Def-book-Dl.Gen ‘The price of the two books.’

c. ʔiʃtara al-tʕulaːb-u al-kutub-a (Broken Plural) bought Def-students-Nom Def-books-Acc ‘The students bought the books.’

d. ʔiʒtimaːʕ-a-n maʕa al-muwaðʕaf-iːna (SndM Plural) meeting-Acc-Idef with Def-employee-Pl.Obl ‘A meeting with the employees.’

e. al-tʕaːlib-aːt al-naːʒiħ-aːt (SndF Plural) Def-student-F.Pl Def-successful-F.Pl ‘The successful (female) students [i.e., successful at passing the exams].’

There does not seem to be any restrictions on the distribution of al- in SA as the prefix

is allowed to attach to any type of noun. From a DM perspective, if /al/ is a VI that is

inserted into a morpheme of the structure of these nouns, this morpheme must be

specified as [+Def].

(17) /al/ ↔ [+Def]

This study takes /al/ to be the standard definite VI in the language. However, the

insertion of the VI is not necessarily the last step in expressing the definiteness of the

nouns. In fact /al/ is subject to a number of well understood phonologically motivated

readjustments that include adding the glottal stop /ʔ/ before the prefix when it falls in an

utterance-initial position, and assimilating the lateral /l/ part of the prefix to the first

160 Chapter Six – Case and Definiteness

consonant of the stem if the consonant is coronal (for more on the phonology of the

definite prefix in SA and the different varieties of Arabic, see, e.g., Abu-Chacra, 2007;

Abu-Salim, 1988; Alfozan, 1989; Watson, 2002; Youssef, 2013; Zuraiq and Zhang,

2006). These are highly systematic phonological changes, so they are assumed to take

place following the insertion of the VI /al/, i.e., they are post-insertion readjustments.95

Indefiniteness, on the other hand, is expressed by the suffix -n (nunation) in some

SA nouns, and simply by the absence of the definite al- in some others. Namely, nouns

that allow nunation are singulars, broken plurals, and SndF plurals, while nunation is

not allowed to appear on duals and SndM plurals. The following are examples of nouns

that take nunation to mark indefiniteness:

(18) a. ʔintaqala ʔilaː dawlat-i-n ʔuxraː (Singular) moved.he to country-Obl-Idef another ‘He moved to another country.’

b. ʔa∫hur-u-n qaliːl-at-u-n (Broken Plural) months-Nom-Idef few-F-Nom-Idef ‘A few months.’

c. yuʒriː ʔitisʕaːl-aːt-i-n (SndF Plural) run.he contact-F.Pl-Acc-Idef ‘He is making contacts/phone calls.’

As would be expected given their definiteness value, nunation never co-occurs with the

definite prefix. With respect to al- and nunation, each noun above is either a definite

noun with the prefix al-, an indefinite noun with nunation, or a bare noun without al- or

nunation (for example, when it is a head noun of a CS).96 As nouns never take both

affixes simultaneously, e.g., *al-dawlat-i-n, the strict contrastive distribution of al- and

nunation confirms that the former is a definite marker, while the latter is an indefinite

95 For the convenience of this study, the definite prefix is presented as al- despite the changes it may undergo after the insertion. 96 e.g., dawl-at-u al-ʔimaːr-aːt-i state-F-Nom Def-emirate-F.Pl-Gen ‘The state of Emirates.’ 161 Chapter Six – Case and Definiteness

marker. That is, from a DM viewpoint, /al/ is the VI inserted into [+Def], as shown

earlier, and /n/ is the VI inserted into [-Def].

(19) /n/ ↔ [-Def]

Following the DP-Hypothesis (Abney, 1987), this study assumes that Arabic

nouns are DPs headed by D and dominating NPs.97 In this configuration, D is the place

where [±Def] is generated as well as the landing site for N when it rises, as shown

below:

(20) DP

D NP [±Def]

Two things can be accomplished by the movement of N to D, namely: N recieves its

[±Def] feature, and also receives the case feature assigned to it by whatever governs the

DP. In addition to the conceptual motivation of the notion, taking Arabic nouns as DPs

is crucial to analysing CSs, as will be shown in the next subsection.

This view entails that when N rises to D, either a [+Def], which is eventually

realised as /al/, or [-Def], eventually realised as /n/, attaches to it. It might have been

noticed that /al/ surfaces as a prefix, whereas most of the feature VIs we have

encountered in this thesis, including nunation, surface as suffixes. This may raise the

question: how is the final distribution of VIs determined? According to Marantz

(1996:15), “Vocabulary items may specify whether they are prefixes, infixes, or

suffixes and may have contextual restrictions that put constraints on their distribution”.

That is, no matter where [±Def] is attached, /al/ is specified for its position as a prefix,

97 Analysing Arabic nouns as DPs is a fairly common practice in the recent literature of Arabic linguistics (see, e.g., Mohammad, 1988; Fassi Fehri, 1999; Benmamoun, 2000; Bardeas, 2008; Al-Khatib, 2009, among many others). 162 Chapter Six – Case and Definiteness

while nunation is specified for its positions as a suffix.98

Nouns that are not overtly marked for indefiniteness by disallowing nunation are

duals and SndM plurals, as shown below:

(21) a. bayna safiːr-ayni /*bayna safiːr-ayni-n between ambassador-Dl.Obl between ambassador-Dl.Obl-Idef ‘Between (two) ambassadors.’

b. maʕa musaːʕid-iːna /*maʕa musaːʕid-iːna-n with assistant-Pl.Obl with assistant-Pl.Obl-Idef ‘With assistants.’

These nouns are considered indefinite simply because they have no definite marker. In

order to account for the absence of an overt indefinite marker, this analysis assumes that

the [-Def] feature of these nouns is impoverished. As shown in the previous section, the

formation of duals and SndM plurals involves fusing their [Num] and [K] morphemes.

These are exactly the same structures whose [-Def] is assumed to be impoverished. In

other words, in order for the Impoverishment rule that targets [-Def] to be activated, the

environment [Num, K] must have been created, as shown below:

(22) [-Def] → Ø / [Num, K] + ______

If this is in fact the case, the above rule captures impoverishing [-Def] in duals, whose

structure contains the fused morpheme [Dl, K], and SndM plurals, whose structure

contains [Pl, K]. In addition, similarly to impoverishing [Acc] in SndF plurals, it

indicates a possible order according to which [Num] and [K] are fused then [-Def] is

impoverished in duals and SndM plurals. This order will be discussed in Section 6.4.

Having shown the function of nunation in SA nouns, it is worth pointing out that,

surprisingly, nunation can appear on names as well. However, for several reasons, this

study does not take the presence of nunation on names to follow from the insertion of

/n/ into [-Def]. That is, nunation on names is not an indefinite marker as it is on

98 To emphasise this point, Halle and Marantz (1993) annotate VIs in a way that shows their positions, e.g., the VIs in question would be presented as /al-/ and /-n/. 163 Chapter Six – Case and Definiteness

singular, broken plural and SndF plural nouns. One of the reasons why nunation on

names cannot be an indefinite suffix is its failure to affect the words that modify the

names to which it is attached, as shown below:

(23) muħammad-u-n al-xaːmis-u Muhammad-Nom-nunation Def-fifth-Nom ‘Muhammad V (a former King of Morocco).’

As shown above, the name has nunation while the adjective that modifies it does not,

but rather is marked as definite by al-. Such co-occurrence (of nunation and definiteness

marking) is not acceptable outside of proper names, for example with indefinite nouns,

as shown below:

(24) a. *raʒul-u-n al-xaːmis-u man-Nom-Idef Def-fifth-Nom

b. *mutasaːbiq-u-n al-xaːmis-u racer-Nom-Idef Def-fifth-Nom

This clearly indicates that names are not morphosyntactically indefinite, so the nunation

that shows up on them is not an indefinite marker. As will be discussed at greater length

with more examples in the next chapter, nunation that appears on names is more likely

to be a semantic marker of specificity.

6.3.2. Construct State

The construct state (CS) is an extensively studied phenomenon both in Arabic and

Hebrew (see, e.g., Mohammad, 1988; Ritter, 1988, 1991; Fassi Fehri, 1989, 1993, 1999;

Siloni, 1991, 2001; Borer, 1999; Benmamoun, 2000; Shlonsky, 2004; Bardeas, 2008;

Al-Khatib, 2009). This section will focus only on those properties of CS that are

relevant to the word formation process of nouns.

164 Chapter Six – Case and Definiteness

A CS is a structure that consists of two nouns that usually have some sort of a

possession relationship.99 The first noun (N1) is the head/possessed noun, while the

second noun (N2) is the complement/possessor noun, as shown below:

(25) a. daːr-u al-raʒul-i house-Nom Def-man-Gen ‘The man’s house.’

b. daːr-u raʒul-i-n house-Nom man-Gen-Idef ‘A man’s house.’

In terms of case, N1 is assigned the case feature that corresponds to its function in the

sentence (nominative, accusative, etc.), but N2 is always genitive. In terms of

(in)definiteness, N1 is always bare (no al- or nunation), but N2 is overtly marked.100

Forcing a definite or an indefinite marker on N1 will immediately result in

ungrammaticality, as shown below:

(26) a. *al-daːr-u al-raʒul-i Def-house-Nom Def-man-Gen

b. *daːr-u-n raʒul-i-n house-Nom-Idef man-Gen-Idef

An important aspect of Arabic CS is that when an attributive adjective modifies N1, the

adjective must be positioned after N2 (outside the CS), agree with N1 in case (in

addition to gender and number), and agree with N2 in (in)definiteness, as shown below:

(27) a. daːr-u al-raʒul-i al-waːsiʕ-at-u101 house-Nom Def-man-Gen Def-large-F-Nom ‘The man’s large house.’

99 There are several types of relationships between nouns that can constitute a CS in Arabic. With a different word order at times, these relationships are similar to those between the nouns of the English phrases, e.g., John’s book (possessor-possessed), the crowing of the rooster (action-agent), the raising of the flag (action-object), a decision maker (object-agent), the gold chest (content-container), the Minister of Justice (identity), and a rescue plane (purpose). For more on these relationships, see Ryding (2005:205-211). 100 Unless N2 does not accept nunation, or - as shown below - N2 is also the N1 of an embedded CS, e.g.: [taʕyiːn-u [waziːr-i al-daːxiliy-at-i]] appointment.Nom minister-Gen Def-interior-F-Gen ‘The appointment of the Minister of Interior.’ 101 The adjective is feminine as it agrees with the covert feminine noun daːr (house). 165 Chapter Six – Case and Definiteness

b. daːr-u raʒul-i-n waːsiʕ-at-u-n house-Nom man-Gen-Idef large-F-Nom-Idef ‘A man’s large house.’

The above examples indicate that N1 and N2 constitute one phrase whose

(in)definiteness feature is carried by N2 and reflected on the adjective.

With some variations, CSs are also found in Modern Hebrew. As in Arabic, N1 of

Hebrew CSs is always bare, while N2 is the element to which the Hebrew definite

marker ha- can be attached,102 as exemplified below (Ritter, 1988:915):

(28) beyt ha-mora / *ha-beyt ha-mora house the-teacher ‘The teacher’s house.’

When an adjective is involved in Hebrew CSs, it must agree with N2 in definiteness, as

shown below (Ritter, 1988:916):

(29) a. beyt ha-mora ha-yafa (F N2 + F adjective) house.M the-teacher.F the-pretty.F ‘The pretty teacher’s house.’

b. beyt ha-mora ha-yafe (M N1 + M adjective) house.M the-teacher.F the-pretty.M ‘The teacher’s pretty house.’

In a similar behaviour to Arabic adjectives in CSs, adjectives in Hebrew are positioned

outside the CS, and N2 carries the definiteness feature of the whole CS that is reflected

on the adjectives even when they modify the bare N1, as shown in (29.b).

Ritter (1991), which is a modified version of Ritter (1988), argues that Hebrew

CSs are DPs whose internal structure parallels the structure of sentences. That is, like

IPs (inflectional phrases) in sentences are headed by INFL that assigns nominative case

to its subjects, CSs are DPs that are headed by a (functional) D that assigns genitive

case to its subjects as well, i.e., Dgen. According to Ritter, in addition to the function of

Dgen as a case assigner, it is a phonetically (but morphosyntactically in this study) null D

102 According to Ritter (1991), Hebrew has no indefinite marker. In addition, from the examples provided in Ritter (1988, 1991), N2 in Hebrew CSs does not seem to be overtly marked for being genitive.

166 Chapter Six – Case and Definiteness

in the sense that it has no association with the (in)definiteness feature of the

construction, i.e., [±Def] is not generated under Dgen. According to this view, the

structure of the Hebrew CS beyt ha-mora can simply be represented as the following:

(30) DP

Dgen NP

DP N

ha-mora beyt

In the above configuration, N1 (beyt) moves to Dgen, in the same way N normally moves

to D in non-CSs, but receives no [±Def] so it is always bare. In its new position, N1

receives the case feature assigned by whatever governs the highest DP. The embedded

DP, however, is headed by a normal D to which N2 (mora) moves to receive its [+Def]

(ha-) and the genitive case assigned to it by Dgen. As to how the [+Def] feature of the

embedded DP becomes the feature of the whole CS, Ritter (1991:41) argues that a

mechanism of SPEC-head agreement between N2 and N1 allows the latter, before it

moves to Dgen, to acquire the feature. As a result of the movement, the feature is

transferred to Dgen; then it percolates to the highest DP allowing the constituents of the

CS to agree with their adjectives accordingly.

As shown in the above examples, CSs in Arabic and Hebrew are, to a large

degree, similar to such an extent that may well make the general direction of Ritter’s

(1988, 1991) analysis of Hebrew CSs applicable to their Arabic counterparts. For more

analyses of Arabic (and generally Semitic) CSs – sometimes with different theoretical

assumptions – see, Mohammad (1988); Fassi Fehri (1989, 1993, 1999); Siloni (1991,

2001); Borer (1999); Benmamoun (2000); Shlonsky (2004); Bardeas (2008); Al-Khatib

(2009), among others. 167 Chapter Six – Case and Definiteness

This study adopts Ritter’s analysis of CSs not only because it plausibly accounts

for their structure, but also because it treats nouns as DPs. This treatment entails that the

(in)definiteness features of the nouns are external to N and not positioned anywhere

between its constituent features. The next section will show that the (in)definiteness

features would be problematic for the formation of SA duals and SndM plurals if not

considered to be generated under the highest projection of the structure of SA nouns.

6.4. The Internal Structure of N

The main objective of the section is to explore the internal structure of N in the light of

its nature as a DP constituent as well as the analyses of its features that have been

provided in this study. The configuration in which nouns and CSs are considered DPs

entails that nouns are composed of NPs headed by N and dominated by DPs, which, in

turn, entails that N does not contain [±Def]. This gives rise to the question: what

features does N contain, and how are these features organised?

In Chapter Four, it was assumed that Arabic nominal stems are structured by

merging [√] and [n], with their embedded indices and CSFs respectively, under nP. In

Chapter Five, it was argued that gender and number features are added to the stems by

merging nP and [Gend] under GendP; then merging GendP and [Num] under NumP, as

shown below:

(31) NumP

GendP [Num]

nP [Gend]

[√(index)] [n[CSF]]

168 Chapter Six – Case and Definiteness

The above structure was assumed to represent the basic configuration of stems with

respect to their gender and number features. SA and DA singulars, and DA duals and

SndM plurals were claimed to maintain this configuration, while SA and DA broken

and SndF plurals were assumed to be formed by the application of some morphological

modifications to the above structure. It was shown that [n], [Gend], and [Pl] are fused in

broken plurals, while it is only [Gend], and only when [+F], that is fused with [Pl] in

SndF plurals.

Given the behaviour of [K] as an independent morpheme in some SA nouns, [K]

is assumed to be generated as an independent morpheme under KP, as shown below:

(32) KP

NumP [K]

GendP [Num]

nP [Gend]

[√(index)] [n[CSF]]

The independence of [K] allows it to remain so (for singulars, broken and SndF plurals),

or to be fused with [Num] (duals and SndM plurals), as discussed earlier. This thesis

considers the above tree representative of the basic internal structure of N that heads its

own NP under DP. That is, the changes that Arabic nouns may undergo with respect to

their [n], [Gend], [Num] and [K] take place inside and within the domain of N.

At first glance, assuming that KP is the highest projection may seem consistent

with Bittner and Hale (1996:3-4), who state the following:

Our theory relies crucially on the notion that Case is a functional head; to be precise, it is the nominal counterpart of C[omplementiser]…Case, we maintain, represents the maximal extension of nominal projection, whereas C represents the maximal extension of the verbal projection. Thus, a Case-marked nominal is KP, just as a fully extended verbal projection is CP.

169 Chapter Six – Case and Definiteness

However, the authors consider KP the highest projection in a structure that includes N,

not the highest projection of the internal structure of N. According to the authors, KP is

the highest projection in a nominal configuration that parallels the verbal configurations

in which CP is the highest projection. This parallelism can be represented as follows:103

(33) CP KP

C IP K DP

I VP D NP

V N

Although Bittner and Hale’s (1996) proposal might successfully account for the case

system in many languages, their assumption about the location of DP as an intermediate

projection between KP and NP, as shown above, is problematic for the formation of SA

duals and SndM plurals. Given that D is where [±Def] is generated, and [Num] and [K]

are always fused in SA duals and SndM plurals, as discussed earlier, the position of

[±Def] between [Num] and [K] would not allow this fusion to take place.104

This leaves us with only two options as to where DP can be positioned without

hindering the formation of SA duals and SndM plurals. That is, DP is either the highest

projection – higher than KP – in the internal structure of N, or the projection that

dominates NP under which N, with KP as its highest projection, is generated. This

analysis adopts the second option not only because it makes the fusion of [Num] and

[K] possible, but also because it effectively accounts for the case and (in)definiteness

features of CS nouns, as shown earlier in this chapter.

103 This is a simplified reproduction of the trees presented in Bittner and Hale (1996:4). 104 In addition, Bittner and Hale (1996) argue that nominative nouns are caseless, i.e., have no KP, which makes their analysis even more problematic to the fusion of [Num] and [K] in SA nominative duals and SndM plurals. 170 Chapter Six – Case and Definiteness

Considering the basic structure of N postulated in (32), N is a structure that

contains all the grammatical features that Arabic nouns take (gender, number, case),

except (in)definiteness. N rises to D where the structure is sealed by receiving its [±Def]

feature. The whole DP, which is now a fully formed phrase/clause constituent, enters

the Morphology where the necessary morphological modifications are carried out. In

SA, these modifications alter the internal structure of N, so the different types of nouns

are formed. Considering the modifications that have been proposed in this study, the

final internal configuration of N in each type can be represented as the following:

(34) a. [[[[[√(index)] n[CSF]] Gend] Sg] K] (Singular Nouns)

b. [[[[√(index)] n[CSF]] Gend] Dl, K] (Dual Nouns)

c. [[[√(index)] n[CSF], Gend, Pl] K] (Broken Plural Nouns)

d. [[[[√(index)] n[CSF]] -F] Pl, K] (SndM Plural Nouns)

e. [[[[√(index)] n[CSF]] +F, Pl] K] (SndF Plural Nouns)

As discussed in the previous chapter, some of the operations by which the above

structures are produced are motivated by special rules of the language (e.g., broken

plurals), while some other operations are carried out by default in accordance with the

value of certain features (e.g., duals and sound plurals). Among the latter type, there are

some operations whose activation is conditioned by the accomplishment of other

operations. Specifically, impoverishing [Acc] in SndF plurals is conditioned by fusing

[+F] and [Pl] (Section 6.2), and impoverishing [-Def] in duals and SndM plurals is

conditioned by fusing their [Num] and [K] morphemes (Section 6.3).

This operational order is likely to be reflecting the order by which the targeted

morphemes are added to the structure in the Syntax. For example, according to the

proposed structure of N, adding [+F] and [Pl] to structures that will eventually surface

as SndF plurals takes place prior to adding [K]. As a default rule, [+F] and [Pl] will be

171 Chapter Six – Case and Definiteness

fused regardless of the specification of [K]; then [K] will be impoverished if specified

as [Acc]. Likewise, since adding [Num] and [K] to Ns that will eventually surface as

dual or SndM plural occurs prior to adding the [-Def] feature (as a result of N-to-D

rising), fusing [Num] and [K] takes precedence over impoverishing [-Def]. In each of

these modifications, two operations are involved. The operation that is performed first is

the one that targets the morphemes that are presumably added first to the structure. At

the same time, it is the operation that paves the way for the activation of the other

operation.

6.5. SA Diptotes

Diptotes are a group of certain SA singulars and broken plurals (overwhelmingly the

latter) as well as names that do not take nunation. When diptotes resist nunation, they

also exhibit a special behaviour towards case by taking the suffix -a to mark their

oblique/genitive cases instead of -i, as shown below:

Table 6.1. SA Diptotes Definite Nouns Indefinite Nouns Nom Acc Obl/Gen Nom Non-Nom al-sʕaħraːʔ-u al-sʕaħraːʔ-a al-sʕaħraːʔ-i sʕaħraːʔ-u sʕaħraːʔ-a The desert A desert al-ruʔasaːʔ-u al-ruʔasaːʔ-a al-ruʔasaːʔ-i ruʔasaːʔ-u ruʔasaːʔ-a The presidents Presidents al-xanaːʒir-u al-xanaːʒir-a al-xanaːʒir-i xanaːʒir-u xanaːʒir-a The daggers Daggers al-mawaːdʕiːʕ-u al-mawaːdʕiːʕ-a al-mawaːdʕiːʕ-i mawaːdʕiːʕ-u mawaːdʕiːʕ-a The subjects Subjects Names Nom Non-Nom ʔaħmad-u ʔaħmad-a Ahmad ʕafaːf-u ʕafaːf-a Afaf yaʕquːb-u yaʕquːb-a Jacob

172 Chapter Six – Case and Definiteness

What is interesting about diptotes is that when they lose nunation, they deviate from the

usual case-marking system despite the nunation that names lose being functionally

different from the nunation lost by nouns, as indicated in Section 6.3.1. This can be an

indication that the nunation of both types of diptotes is targeted for its phonological

nature rather than its function.

This section reviews an analysis done by Embick and Noyer (2007) in which the

authors take the behaviour of diptotes to be the result of morphological modifications

that involve impoverishing some of their features. It will be shown that this behaviour

of diptotes is more likely to be motivated in the Vocabulary rather than the Morphology.

Nevertheless, no detailed explanation will be provided as to how exactly diptotes are

formed as such in the Vocabulary. A comprehensive investigation that covers all the

aspects of diptotes, including the diachronic aspect, is indeed needed. Thus, the analysis

provided in this section must remain tentative.

Traditionally, grammarians have classified diptotes on the basis of a number of

parameters: by gender (most feminine and certain masculine names), by etymological

origin (names derived from verbs, e.g., yaziːd ‘Yazid, literally, he increases’ and foreign

names, e.g., yaʕquːb ‘Jacob’), and by morphological pattern (nouns that have certain

105 binyanim, e.g., /C1uC2aC3aːʔ/ and /C1aC2aːC3iC4/). Although this classification

captures the majority of diptotes, there are still some exceptions. This is why Embick

and Noyer (2007, note 28) propose that diptotes form a special class that is marked with

a diacritic feature, as quoted below:

Although certain generalizations, some exceptionless, exist regarding whether a given stem will be diptote or triptote, in many cases the choice is unpredictable. For example, the proper name hind- can inflect diptote or triptote (Haywood and Nahmad, 1965:384-388). Regardless of how predictable the diptote property is, however, it remains clear that the diptotes as a class must be marked with a diacritic class feature of some kind. The feature [diptote] is used here for this purpose.

105 For more on this classification, see Ryding (2005:192-197) 173 Chapter Six – Case and Definiteness

Observe that the authors use the term triptotes when referring to the opposite class

whose case-marking system contains the three different markers /u/, /a/ and /i/. Based

on this classification, Embick and Noyer (2007:311) suggest the following VIs of case:

(35) /u/ ↔ [+superior]

/i/ ↔ [+oblique]

/a/ ↔ elsewhere

To account for the insertion of /a/ into [+oblique] (i.e., [Obl/Gen] in this study) and the

absence of nunation in diptotes, Embick and Noyer (2007:312) propose the following

Impoverishment rules:

(36) a. [+oblique] → Ø / [diptote] + _____ + [-definite]

b. [-definite] → Ø / [diptote] + case/number + ______

The above rules entail that when a structure is diptote and indefinite, its [Obl/Gen] case

morpheme is impoverished allowing the elsewhere /a/ to be inserted. The [-Def]

morpheme is then improvised, which explains the absence of nunation.

As might have been already noticed, the list that Embick and Noyer (2007)

propose for the VIs of case is different, and less comprehensive, than what was

proposed in Section 6.2. This deficiency in their proposal arises from overlooking the

fact that /i/ is the case marker used with accusative SndF plurals as well as the oblique

and genitive marker for singulars, broken and SndF plurals. As /i/ is by far the least

specified VI of case in the language, it must remain the elsewhere VI. With this point

established, impoverishing the [Obl/Gen] morpheme of diptotes, as proposed by the

authors, would not result in the insertion of /a/ instead of /i/. The latter will remain the

VI that should be inserted whether the [Obl/Gen] morpheme is impoverished or not. If

/a/ is the elsewhere VI, the insertion of /i/ instead of /a/ in accusative SndF plurals

174 Chapter Six – Case and Definiteness

would not be accounted for. According to the competition the authors propose, SndF

plurals should be phonologically marked for the accusative case by /a/, contrary to fact.

Moreover, the explanation that the authors provide about impoverishing [-Def] in

diptotes is also problematic. Although it may account for the absence of nunation from

diptote nouns, it does not capture the same absence in diptote names. It was shown

earlier that names could not be structurally marked as indefinite. Attributing the absence

of nunation in diptote names to impoverishing [-Def] entails that the appearance of

nunation on non-diptote names follows from the insertion of /n/ into a [-Def]

morpheme. Names do not appear to have such a morpheme, so taking the absence of

their nunation as a result of impoverishing their [-Def] feature does not seem plausible.

Given that the behaviour of diptotes cannot be accounted for in the Morphology, a

more reasonable speculation would be that the absence of nunation and the employment

of /a/ instead of /i/ in diptotes are outcomes that are determined in the Vocabulary. If

this is where the behaviour of diptotes is initiated, there would be two options as to how

this task is accomplished. Under the first option, nunation (be it an indefinite marker in

nouns or filling another function in names) and /i/ are not inserted; consequently /a/ is

inserted instead. The second option is that nunation and /i/ are inserted as normal, then

the former is deleted and, as a result, the latter is replaced with /a/. The first option

entails that the behaviour of diptotes is initiated at the insertion level, while the second

option makes diptotes the result of post-insertion readjustments.

In fact, both options are ad hoc and not quiet convincing. However, evidence from

defective plural nouns that are also diptote, as will be shown in the next section,

suggests that nunation and /i/ are indeed inserted, and then phonologically readjusted,

which makes the second option more likely. If post-insertion readjustments are what

create diptotes, it remains unclear how exactly those readjustments are motivated or

175 Chapter Six – Case and Definiteness

carried out. They might be the result of certain phonological restrictions on diptotes

triggered by the diacritic feature [diptote] proposed by Embick and Noyer (2007). This

issue will be left open.

What is clear though is that Embick and Noyer’s (2007) attempt to employ

Impoverishment in order to account for the behaviour of diptotes fails to capture the

entire case-marking system of the language and the absence of nunation from diptote

names, so their analysis will not be adopted. As mentioned earlier, evidence supporting

the assumption that the behaviour of diptotes is initiated after the insertion level will be

provided by analysing defective plural nouns in the next section.

6.6. SA Defective, Indeclinable and Invariable Nouns

There are nouns known as defective, indeclinable and invariable nouns that deviate

from the regular case and indefinite marking system in various ways. Relying on a

description by Abboud and McCarus (1983), especially for the first two types, this

section shows that the deviations result from certain phonological restrictions that

motivate certain post-insertion readjustment rules. These restrictions include the

occurrence of the semi-vowels /y/ or /w/ as the final root and simultaneously the final

stem element, and the occurrence of the long vowel /aa/106 as the final segment in the

binyan and simultaneously the final stem element. The discussion regarding plural

defective nouns in particular will provide us with a better insight into diptotes and why

this study assumes that they are not morphologically modified as proposed by Embick

and Noyer (2007).

106 The double vowels /aa/ and /ii/ will be used instead of /aː/ and /iː/ for illustration purposes. 176 Chapter Six – Case and Definiteness

6.6.1. Defective Nouns

Defectives are a group of singular and broken plural nouns that are overtly marked for

case only when accusative. In terms of nunation, defective singulars take nunation

whenever they are indefinite, but defective broken plurals take nunation – or what

appears to be so – only when indefinite and non-accusative. Consider the following

table:

Table 6.2: SA Defective Nouns Definite Indefinite

Non-Accusative Accusative Non-Accusative Accusative

Judge al-qaːdʕii al-qaːdʕiy-a qaːdʕi-n qaːdʕiy-a-n Challenge al-taħadii al-taħadiy-a taħadi-n taħadiy-a-n Suburbs al-dʕawaːħii al-dʕawaːħiy-a dʕawaːħi-n dʕawaːħiy-a Chairs al-karaːsii al-karaːsiy-a karaːsi-n karaːsiy-a

Notice that each defective noun has different forms depending on its case and

(in)definiteness. The accusative stems, however, are consistent and appear to be

representative of the unchanged stems. They take the standard accusative suffix -a (in

both singular and plural nouns) and nunation as an indefinite suffix (only in singular

nouns). In fact, the absence of nunation from the accusative defective plurals is

expected since all defective plurals are diptote (Abboud and McCarus, 1983:89; Ryding,

2005:198). They are diptote because they contain the binyan /C1aC2aːC3iC4/, as

mentioned in the previous section, so the absence of nunation in these nouns is in line

with their nature as diptotes. This gives rise to the question: why do the indefinite non-

accusative defective plurals dʕawaːħi-n (suburbs) and karaːsi-n (chairs) have nunation

while they are diptotes? Abboud and McCarus (1983) raised this question, but did not

provide an answer. This question will be answered after showing how defectives are

generally formed.

177 Chapter Six – Case and Definiteness

Given that accusative defectives contain the unchanged stems, Abboud and

McCarus (1983) assume that they are the stems from which all defective nouns (definite

and indefinite) are initially formed. The stems of the above nouns are the following:

(37) a. qaːdʕiy (Judge)

b. taħadiy (Challenge)

c. dʕawaːħiy (Suburbs)

d. karaːsiy (Chairs)

As shown above, all the stems end with /-iy/. The /i/ part is the final element of the

binyan inserted into [n], and the semi-vowel /y/ is the final element of the root VIs

inserted into the root morphemes. As to how the above stems sometimes change, the

authors claim that the /-iy/ ending is subject to various phonological restrictions

depending on what sequences are produced after the case suffixes and nunation are

added. The full range of these sequences can be represented as the following:

(38) Definite Defective Nouns:

a. -iyu (Nominative)

b. -iyi (Oblique/Genitive)

c. -iya (Accusative)

(39) Indefinite Defective Nouns:

a. -iyun (Nominative)

b. -iyin (Oblique/Genitive)

c. -iyan (Accusative)

According to the authors, the sequences in (38.a-b) and (39.a-b) are prohibited, so they

need to be modified. The authors propose the following phonological rules by which the

final forms of the prohibited sequences are achieved:

178 Chapter Six – Case and Definiteness

(40) a. -iyu -ii -iyi

b. -iyun -in -iyin

The accusative endings -iya (38.c) and -iyan (39.c), however, are acceptable sequences,

so they do not need to be modified.

In order to convert Abboud and McCarus’s (1983) analysis into the realm of DM,

the phonological rules proposed in (40) must be taken as post-insertion readjustment

rules. They come into effect when the inserted VIs form a prohibited sequence. The VIs

involved are root VIs that end with /y/, binyanim that end with /i/, a case VI that is

either /u/ or /i/, and the indefinite VI /n/ (nunation) when the noun is indefinite. When a

prohibited sequence is produced, the post-insertion readjustment rules replace it with

/-ii/ or /-in/ depending on the elements of the prohibited sequence. For illustration, this

subsection demonstrates this mechanism on the defective singular stem qaːdʕiy (judge),

then the defective (diptote) plural stem karaːsiy (chairs). The latter will explain why

diptotes are thought to be formed at the post-insertion level in the Vocabulary.

The defective singular stem qaːdʕiy is formed by inserting, then interlocking, the

ʕ root VI /qd y/ with the binyan /C1aːC2iC3/. A stem that ends with /-iy/ is thereby

produced. In definite structures, the stem will take the definite VI /al/ as a prefix, and

either /u/, /i/, or /a/ for its case feature as a suffix. The first two case suffixes will

produce the prohibited sequences /-iyu/ and /-iyi/, so they change into /-ii/. The

accusative suffix /a/ and the stem will form the acceptable sequence /-iya/, so nothing

will be changed. If the structure is indefinite, a nunation will be added as a suffix,

producing the sequences /-iyun/, /-iyin/ and /-iyan/. As shown in the rules of Abboud

and McCarus (1983), the first two sequences will change into /-in/, while the last

sequence remains intact. As a result, the nouns surface as shown in Table 6.2 above.

179 Chapter Six – Case and Definiteness

For the defective plural karaːsiy (chairs), the root VI /krsy/ interlocks with the

binyan /C1aC2aːC3iC4/, which is a diptote binyan. When the noun is definite, /al/ will be

prefixed, and the case VIs /u/, /i/ or /a/ will be suffixed. With /-iy/ as the ending of the

stem, the case VIs /u/ and /i/ will produce prohibited sequences, while /a/ will not. The

prohibited sequences will be replaced with /-ii/ following the same mechanism that is

also applied to definite defective singulars.

When the noun is indefinite, however, they become subject to the restrictions of

diptote nouns. That is, they are not allowed to take nunation or the oblique/genitive case

suffix -i. If, at the insertion level, /n/ and /i/ are not inserted because the features [-Def]

and [Obl/Gen] are impoverished (as argued by Embick and Noyer, 2007), the produced

sequences will be either /-iyu/ (indefinite nominative) or /-iya/ (indefinite non-

nominative). On the one hand, according to the rules in (40), /-iyu/ will be replaced with

/-ii/. This will produce karaːsii, which is not how the indefinite noun actually looks.

This form is found only when the noun is definite non-accusative. On the other hand, /-

iya/ will be deemed acceptable and the stem will surface as karaːsiy-a, which is

consistent with the indefinite accusative form we have in the table. Obviously, this

mechanism according to which /n/ and /i/ are not inserted is problematic for the

formation of indefinite non-accusative defective plurals.

In the previous section, it was suggested that /n/ and /i/ are inserted in diptotes,

then readjusted as they are not allowed to surface. Thus, at some point after the

insertion, /n/ needs to be deleted and /i/ needs to be replaced with /a/. Let us examine

this suggestion using the defective plural stem karaːsiy. If the stem can be fully

represented for case and indefiniteness at the insertion level, there will be three possible

endings: /-iyun/, /-iyin/ and /-iyan/. On the one hand, the first two sequences are

prohibited, so they will be replaced with /-in/, which is consistent with the actual form

180 Chapter Six – Case and Definiteness

of the indefinite non-accusative defective plural karaːsin. Observe that the final element

of the new ending, which is /n/, is identical to the final element of the replaced one.

Before the replacement, the /n/ sound was the VI inserted into [-Def], but after the

replacement, the /n/ sound is part of the new sequence that is employed to rectify the

prohibited occurrences. Hence, as answer to the earlier question, the /n/ that appears on

indefinite non-accusative defective plurals that are simultaneously diptote is not the

indefinite marker, which is not allowed to appear on diptotes. The /n/ is there simply by

virtue of the rule and it does not encode the [-Def] feature. The inserted /n/ that

corresponds to that feature will have been already taken away as part of the prohibited

sequences.

On the other hand, the last sequence /-iyan/ is acceptable, as far as the rules in

(40) are concerned, so it will pass. However, since the final /n/ of this sequence is a

genuine nunation that is not allowed to remain attached to diptotes, it must be removed

at some point so the noun surfaces as karaːsiy-a instead of *karaːsiy-a-n.

This mechanism shows that whatever the motivation is, targeting the /n/ and /i/

sounds of diptotes must take place after the insertion level. Defective plurals cannot be

formed without the insertion of /n/ and /i/. This supports the earlier assumption that

prohibiting indefinite diptotes from taking nunation and the oblique/genitive suffix -i is

a phonological matter that is dealt with in the Vocabulary, specifically, after the

insertion. This contrasts with Embick and Noyer’s (2007) analysis in which diptotes are

taken to be formed in the Morphology.

6.6.2. Indeclinable Nouns

Indeclinable nouns are a set of certain singulars and broken plurals that are not overtly

marked for case, but take nunation when indefinite. Consider the following examples:

181 Chapter Six – Case and Definiteness

Table 6.3: SA Indeclinable Nouns Indefinite Definite (All cases) (All cases) Stick/Cane al-ʕasʕaa ʕasʕa-n Café al-maqhaa maqha-n Villages al-quraa qura-n

Like defective nouns, indeclinable nouns are taken to be stems that form prohibited

sequences when phonologically marked for case and indefiniteness, so they undergo

some phonological changes. In order to determine the initial forms, especially the roots,

of the stems, Abboud and McCarus (1983:14-19) compare the above nouns to nouns

that are made from the same roots, as shown below:

(41) a. ʕasʕaa vs. ʕasʕaw-aːni stick stick-Dl.Nom ‘A stick.’ ‘Two sticks.’

b. maqhaa vs. qahw-at- café coffee-F ‘A café.’ ‘Coffee.’

c. quraa vs. qary-at- villages village-F ‘Villages.’ ‘A village.’

The above examples show that the roots of the indeclinable nouns included one of the

semi-vowels /w/ and /y/ as their final element. This means the indeclinable nouns that

we have in Table 6.3 are initially formed as the following:

(42) Initial Forms of Definite Indeclinable Nouns:

*a. al-ʕasʕaw-u al-ʕasʕaw-a al-ʕasʕaw-i Def-stick-Nom Def-stick-Acc Def-stick-Obl/Gen

*b. al-maqhaw-u al-maqhaw-a al-maqhaw-i Def-café-Nom Def-café-Acc Def-café-Obl/Gen

*c. al-quray-u al-quray-a al-quray-i Def-villages-Nom Def -villages-Acc Def - villages-Obl/Gen

182 Chapter Six – Case and Definiteness

(43) Initial Forms of Indefinite Indeclinable Nouns:

*a. ʕasʕaw-u-n ʕasʕaw-a-n ʕasʕaw-i-n stick-Nom-Idef stick-Acc-Idef stick-Obl/Gen-Idef

*b. maqhaw-u-n maqhaw-a-n maqhaw-i-n café-Nom-Idef café-Acc-Idef café-Obl/Gen-Idef

*c. quray-u-n quray-a-n quray-i-n villages-Nom-Idef villages -Acc-Idef villages -Obl/Gen -Idef

Based on this assumption, the authors explain how the above nouns change into their

final and acceptable forms, as seen in the table, by proposing phonological rules that are

sensitive to the occurrence of the case suffixes and nunation after the endings /-aw/ and

/-ay/, as shown below:

(44) a. -awu b. -awun -awa -awan -awi -awin -ayu - aa -ayun - an -aya -ayan -ayi -ayin

As far as this study is concerned, the above rules are applied to the nouns after the

insertion level. That is, the initial forms in (42) and (43) above show how the nouns

look immediately after the insertion of the VIs that correspond to their morphemes. The

changes that the inserted VIs of indeclinable nouns undergo are applied as post-

insertion readjustment rules that are motivated by purely phonological factors.

6.6.3. Invariable Nouns

Invariable nouns are also a group of singulars and broken plurals. These nouns do not

take nunation or any case suffix. Apart from the definite prefix al-, an invariable noun

always has the same form in all environments, as shown below:

183 Chapter Six – Case and Definiteness

Table 6.4: SA Invariable Nouns Indefinite Definite (All cases) (All cases) Complaint al-∫akwaa ∫akwaa Memorial al-ðikraa ðikraa Gifts al-hadaːyaa hadaːyaa Deserts al-sʕaħaːraa sʕaħaːraa

As can be seen above, the final root element can be a semi-vowel or a consonant. In

addition, it can also be preceded by a binyan vowel or a root consonant. What invariable

nouns have in common is their ending, which is /-aa/. 107 Unlike defective and

indeclinable nouns, this ending is part of the binyanim and has nothing to do with the

roots of the nouns. The final root elements are followed by the final elements of the

binyanim that happen to be the final elements of the stems as well. For example, in the

invariable noun ðikraa above, the root is /ðkr/ and the binyan is /C1iC2C3aa/. When the

root interlocks with the binyan, the final root consonant /r/ is not the final element of the

stem. It is followed by /aa/ that is the final segment of the binyan as well as the final

element of the whole stem.

Given how the stems of invariable nouns are formed, when the case VIs (/u/, /a/

or /i/) and nunation are inserted into the [K] and [-Def] morphemes of these nouns, the

sequences /-aau(n)/, /-aaa(n)/ and /aai(n)/ are produced. As can be noticed, a sequence

of vowels (hiatus) occurs when the /-aa/ vowel ending is followed by the case VIs,

which are vowels as well. As a way of hiatus resolution, and as a post-insertion

readjustment process, the sequences are truncated by dropping the elements that follow

the ending /-aa/. No morphological operations seem to be involved as it is simply a

phonological matter that cannot take place at any point prior to the insertion level.

107 When the ending /-aa/ appears on names, e.g. ʔifriːqiyaa (Africa), kanadaa (Canada), ʔustraːliaa (Australia), and ʔamriːkaa (America), the names show the same behaviour that invariable nouns display with respect to case and nunation. 184 Chapter Six – Case and Definiteness

6.7. Case and Definiteness in DA Nouns

While the definite prefix al- is maintained in DA, there is no overt case or indefinite

marking in DA for its nouns. Neither the usual SA case suffixes nor nunation appear on

DA nouns, as shown below:

(45) a. al-kalb raːħ yitʕaːrid al-naħl, Def-dog went chase Def-bees ‘The dog went chasing the bees.’

b. al-tʕifl tisalag al-ʃaʒarah wa kaːn yibħaӨ ʕan al-dʕifdaʕ Def-kid climbed Def-tree and was search for Def-frog ‘The kid climbed the tree searching for the frog.’

c. ʔana ʃaːyif tʕifl sʕaχiːr fiː χurfat-uh wa maʕaːh kalb wa dʕifdaʕ I see kid little in room-his and with-him dog and frog ‘I see a little kid in his room with a dog and a frog.’

d. al-walad bi-yutʕul fiː ħufrah Def-boy is-looking in hole ‘The boy is looking into a hole.’

In terms of case, the above examples show that wherever a DA noun falls in the

sentence, it does not take any overt case marker. This is clearly consistent with the fact

that Arabic dialects have dropped their case-marking system (Ingham, 1994:49;

Brustad, 2000:27).

However, DA still makes a distinction between subject and object/possessive in

pronouns. To a degree, DA resembles the case-marking system of English where

subject, object and possessive pronouns are distinct, e.g. I, me and my, although nouns

are not overtly marked for case. The difference between DA and English in this regard

is that these pronouns in DA are not always independent. DA has a wider variety of

pronouns, most of which are affixes. The forms of these pronouns may vary from their

SA counterparts, as well as from one dialect to another. A general paradigm for these

pronouns in DA can be outlined as shown in the table below (Qafisheh, 1977; Holes,

1990, 2010; Ingham, 1994; Vicente, 2007; Vollers, 2011):

185 Chapter Six – Case and Definiteness

Table 6.5: DA Subject, Object and Possessive Pronouns Subject Verbal Nominal Verbal Verbal Independent Object Possessive Affixes Affixes Pronouns Affixes Affixes (Imperfective) (Perfective) 1st Sg ana a-V V-t -V- -niː N-iː Person Pl iħnaː n-V V-naː -V- -naː N-naː M Sg inta t-V V-t -V- -ak N-ak 2nd F Sg inti t-V-iː(n) V-tiː -V- -ik N-ik Person M Pl intu(m) t-V-uː(n) V-tu(m) -V- -ku(m) N-ku(m) F Pl108 intin t-V-in/an V-tin/tan -V- -kin/kan N-kin/kan M Sg hu(wa) y-V V -V- -uh N-uh 3rd F Sg hiy(ya) t-V V-at -V- -haː N-haː Person M Pl hum(ma) y-V-uː(n) V-uː -V- -hum N-hum F Pl hin(na) y-V-in/an V-in/an -V- -hin/han N-hin/han

The distinction between subjects and objects in DA can be clearly seen when verbs

inflect for both by affixation, as shown below:

(46) a. y-iħibb-niː (Holes, 1990:114) he-like-me ‘He likes me.’

b. ∫if-naː-hum (Holes, 2010:97) saw-we-them ‘We saw them.’

c. zaːr-uː-haː (Holes, 2010:97) visted-they-her ‘They visited her.’

d. sʕaːr y-aːkul-uh (Brustad, 2000:32) began.he he-eat-it ‘He began to eat it.’

For making such a distinction between subjects and objects in pronouns, English

is typologically classified as a case-poor language, rather than a caseless language such

as Chinese (Nichols, 1992, as cited in Sigurdsson, 2001:103). This also makes DA a

case-poor language as well. Thus, one difference between SA and DA in terms of case

108 Feminine plural forms in 2nd and 3rd persons are not usually used in DA. They are common in certain, usually Bedouin, Gulf and Jordanian, sub-dialects. In Egyptian Arabic, for example, the masculine plural forms are used instead (Vollers, 2011). 186 Chapter Six – Case and Definiteness

is that the Vocabulary of DA does not provide any case VI to [K] in nouns. Case

morphemes in the structure of DA nouns are simply phonologically unrealised.

Furthermore, [K] does not appear to fuse with [Dl] or [Pl] in dual and SndM

plural nouns in DA. These nouns are marked for number by taking the suffixes -ain and

-iːn respectively, as mentioned in the previous chapter, and the case features of the

nouns do not affect the suffixes. These suffixes were taken as the VIs inserted into [Dl]

and [Pl] in DA. The fact that these morphemes are independent at the insertion level

indicates that the rules by which the morphemes are fused with [K] in SA are probably

no longer active in the Morphology of DA.

With respect to indefiniteness in DA nouns, the examples in (45) above show that

DA nouns do not take nunation when indefinite.109 Nouns such as tʕifl (kid), kalb (dog),

dʕifdaʕ (frog) and ħufrah (hole) occurred as indefinite in (45.c-d) without taking

nunation. In fact, nunation did not occur on any noun in the collected data, which may

suggest that DA has also dropped the indefinite-marking system that is active in SA.

That is, [-Def] in DA became a silent morpheme into which no VI is inserted.

The absence of case marking and nunation in DA nouns entails that, unlike in SA,

there are no diptote, defective, indeclinable and invariable nouns. In fact, all DA nouns

are now invariable in the sense that they have the same form in all case environments

and are considered indefinite by the absence of the definite prefix al-.

6.8. Conclusion

This chapter provided a DM analysis for case and (in)definiteness in Arabic nouns.

With respect to case in SA nouns, it was shown that case morphemes in singulars,

broken plurals, and SndF plurals are independent, and subject to Impoverishment when

109 A form of nunation that is different from the nunation of SA is sometimes used in Bedouin dialects in certain environments. The way this form of nunation is employed suggests a different function, as will be discussed in relation to nunation on SA names in the next chapter. 187 Chapter Six – Case and Definiteness

specified as [Acc] in the latter type. Impoverishing the [Acc] morpheme in SndF

plurals, however, is conditioned by fusing their [+F] and [Pl] morphemes.

It was suggested that fusing these morphemes takes precedence over

impoverishing the [Acc] morpheme because the fused morphemes are presumably

merged first in the Syntax. This was considered an indication that when two operations

are to be performed on a single structure in the phase of the application of default rules,

the order in which these operations are performed reflects the order in which the

targeted morphemes are added to the structure.

The chapter also showed that case in SA duals and SndM plurals is associated

with the number feature of these nouns. This association was taken to result from fusing

the features in the Morphology. A comprehensive representation of case VIs that

captures the case-marking system of SA was also introduced.

In terms of (in)definiteness in SA nouns, the chapter assumed that al- and

nunation are the phonological representations of the [±Def] features that are generated

under the D node heading the DPs. The definite al- was shown to be allowed to attach

to all types of SA nouns, while nunation was shown to be disallowed from appearing on

SA duals and SndM plurals, as the feature [-Def] it encodes is always impoverished.

This Impoverishment was shown to be conditioned by creating the fused morpheme

[Num, K] first. This was taken as supporting evidence that morphemes are modified in

the phase of default rules in the Morphology following the order according to which

they are added to their structures.

By treating nouns as DPs, two characteristics of the structure of Arabic nouns

were accounted for, namely: the absence of the [±Def] features from head nouns of CSs,

as analysed in Hebrew by Ritter (1988, 1991), and fusing [Num] and [K] in SA duals

and SndM plurals that becomes possible by taking the [±Def] features to be generated in

188 Chapter Six – Case and Definiteness

a position that is external to N. This configuration of the structure of Arabic nouns led

to assuming that stems and their features of gender, number and case are parts of the

internal structure of N that must rise to D in order for it to receive its [±Def] features.

The chapter then argued that the unusual behaviour of diptotes regarding nunation

and case is initiated in the Vocabulary. This contrasts with Embick and Noyer (2007)

who assume that this behaviour is morphologically motivated. It was shown how the

authors’ analysis fails to capture the case-marking system of SA nouns and the absence

of nunation from diptote names. By showing how defective nouns – along with

indeclinable and invariable nouns – are formed in the Vocabulary, further evidence

supporting the phonological treatment of diptotes was obtained. It was shown that,

according to the phonological readjustment rules proposed by Abboud and McCarus

(1983), defective plural nouns that are also diptotes cannot be formed properly if

nunation and the elsewhere case VI /i/ are not inserted before the nouns are readjusted.

In terms of case and indefiniteness in DA nouns, it was shown that these features

have no phonological representation in the Vocabulary. The lack of case and

indefiniteness marking in DA removes the characteristics by which some nouns are

made diptote, defective, indeclinable and invariable in SA, so such distinctions between

nouns are useless in DA. With respect to the Morphology of DA, it was shown that the

morphological rules by which number and case are fused in SA duals and SndM plurals

are probably not active in DA. As a result, the VIs that express [Num] are not affected

by the case features of these nouns in DA.

189 Chapter Seven – Nunation

7. Chapter Seven – Nunation

7.1. Introduction

Nunation, the equivalent in modern grammatical terminology of the Arabic term

Tanwiin, usually refers to the -n suffix that appears on many nouns and names in SA

and the earlier stages of Arabic. In the previous chapter, the nunation that appears on

nouns was treated as an indefinite marker. Although this function of nunation is widely

acknowledged, there exists a different view.

Fassi Fehri (1993) and Lyons (1999) claim that nunation cannot be an indefinite

marker, but each author provides a different alternative function. This view of nunation

is based on two arguments. The first argument involves treating the /ni/ and /na/ parts of

the dual and SndM plural suffixes -aːni/ayni and -uːna/iːna on a par with the -n suffix of

singulars, SndF and broken plurals. The authors claim that /ni/ and /na/ are variants of

nunation. This treatment is motivated by the fact that /ni/, /na/ and the suffix -n are

dropped from nouns that occur as the heads of CSs. However, since /ni/ and /na/ can co-

occur with the definite prefix al- in non-CS environments, the authors argue that

nunation cannot be an indefinite marker. The second argument is that nunation cannot

be an indefinite marker simply because it can also appear on names (Fassi Fehri, 1993).

According to the authors’ views, /ni/, /na/, the suffix -n of nouns and the suffix -n

of names belong to a broader definition of nunation, and whatever function this

nunation has, it cannot be a marker of indefiniteness. As an alternative function, Fassi

Fehri (1993) hypothesises that nunation is a Poss(essive) marker, while Lyons (1999)

proposes that nunation is a marker of nominality. This chapter is a detailed refutation of

these views.

190 Chapter Seven – Nunation

If /ni/, /na/, the suffix -n of nouns and the suffix -n of names are in fact different

forms of one type of nunation that has one function, there would be no doubt that

nunation is not an indefinite marker. However, this does not seem to be the case, and

this is exactly what this chapter seeks to demonstrate. The main objective of this chapter

is to show that the association that the authors make between these forms is inaccurate.

Although different issues – some of them quite puzzling – will arise as we move along,

refuting the link claimed to exist between these forms and defending the view of this

study that the nunation of nouns is what encodes the feature [-Def] should remain the

main goal of the discussion.

For ease of exposition, the -n suffix that appears on singulars, broken and SndF

plurals will be referred to as -n(noun), the -n suffix that appears on names will be referred

to as -n(name), and the /ni/ and /na/ parts of the dual and SndM plural suffixes will simply

be /nV/ (i.e., /n/ followed by a vowel) throughout most of the chapter. It will be shown

that /nV/ is not a variant of nunation, while -n(noun) and -n(name), although both called

nunation, are functionally different: -n(noun) will continue to be considered an indefinite

marker, while -n(name) will be considered a marker of specificity.

In Section 7.2, the view that nunation is not an indefinite marker will be reviewed

and some of its technicalities will be critiqued. The justifications for this view and the

alternative functions that nunation is claimed to have will be discussed. It will be shown

that these functions are implausible not only because they cannot account for the actual

behaviour of -n(noun) and its full contrastive distribution with the definite prefix al-, but

also because they create more mysteries than they purport to solve. Most of the section

will address the link that Fassi Fehri (1993) and Lyons (1999) attempt to establish

between -n(noun) and -nV, while the link that Fassi Fehri (1993) tries to establish between

-n(noun) and -n(name) will be examined in Section 7.4.

191 Chapter Seven – Nunation

In order to bolster the argument that /nV/ cannot be a nunation in the way -n(noun)

is, Section 7.3 compares the behaviour of both forms in vocative contexts. The

comparison involves observing the behaviour of singulars (where -n(noun) can occur) and

the behaviour of both duals and SndM plurals (where /nV/ occurs) in vocative contexts.

The comparison will reveal that -n(noun) is merely an indefinite marker, while /nV/ is

nothing but part of the number/case suffixes of its nouns. In addition, the comparison

will reveal that the absence/presence of -n(noun) cannot be accounted for by the new

functions that Fassi Fehri (1993) and Lyons (1999) propose for nunation.

Section 7.4 is concerned with the function of -n(name). This section argues in favour

of the view that -n(name) is not an indefinite marker. By comparing the behaviour of

names that take -n(name) to the behaviour of definite and indefinite nouns in adjectival

agreement, it will be shown that names behave as definite. However, unlike Fassi Fehri

(1993), this study distinguishes between -n(noun) and -n(name) functionally even though

both are called nunation. It will be shown that while -n(noun) is essentially an indefinite

marker, -n(name) will be assumed to be the element by which the high semantic/pragmatic

specificity of names is morphologically marked. It will also be shown that marking the

specificity of names is needed when the names are identical to the words from which

they are derived, but can be abandoned in the presence of morphological elements that

can serve that purpose.

In order to support the argument that -n(name) is a specificity marker, Section 7.5

touches on a different form of nunation (-in) that appears under certain circumstances

on the nouns of the Bedouin dialects of Central Arabia. By briefly reviewing some

analyses of the Bedouin nunation, it will be shown that the main function of this

nunation is marking specificity. Therefore, the Bedouin nunation will be considered a

192 Chapter Seven – Nunation

vestige of the specificity marker -n(name), rather than a vestige of the indefinite marker -

n(noun) as assumed by Holes (1990).

Finally, Section 7.6 concludes the chapter and summarises its discussions.

7.2. A Different View of Nunation

In the previous chapter, nunation was treated as the VI that encodes the [-Def] feature,

i.e., an indefinite marker, of nouns. Although this function is consistent with its

behaviour and how it is predominantly perceived, there is a different view that is found

in Fassi Fehri (1993:216-219; 273) and Lyons (1999:91-94). Both authors argue that

nunation cannot be an indefinite marker, but each author assigns a different function as

an alternative. Fassi Fehri (1993) suggests that nunation is a Poss(essive) marker, while

Lyons (1999) claims that it is a semantically empty marker of nominality. In the

following subsections, these different views of nunation, their motivations and the

alternative functions will be reviewed.

7.2.1. Why Nunation Cannot Mark Indefiniteness

There are two reasons as to why it is argued that nunation cannot be an indefinite

marker. The first reason is simply that nunation can appear on names, as shown below:

(1) a. hind-u-n (Fassi Fehri, 1993:216) Hind-Nom-n ‘Hind.’

b. muħammad-u-n (Fassi Fehri, 1993:217) Muħammad-Nom-n ‘Muhammad.’

Taking the appearance of nunation on names as evidence that it cannot be a marker of

indefiniteness implies that (a) names are (inherently) definite, and (b) the nunation that

appears on names is (functionally) the same nunation that appears on nouns. This study

agrees that names are definite, so the nunation they take cannot be an indefinite marker.

193 Chapter Seven – Nunation

However, it differentiates between the nunation that appears on names (= -n(name)) and

the nunation that appears on nouns (= -n(noun)). The latter is considered an indefinite

marker, as argued in the previous chapter, while the former will be considered a marker

of specificity. In Section 7.4, it will be shown why a distinction between -n(name) and

-n(noun) should be made, as well as why -n(name) is a specificity marker.

The second reason why nunation is claimed not to be an indefinite marker is much

more complex than the first one. It involves redefining the concept of nunation by

making it a set of a few variant forms some of which can co-occur with the definite

prefix al- (Fassi Fehri, 1993; Lyons, 1999). There is a restriction disallowing the co-

occurrence of the definite prefix al- and -n(noun), as mentioned in the previous chapter.

This restriction is expected to be resulting from the fact that the former is a definite

marker while the latter is an indefinite marker. In order to undermine any function that

is assigned to nunation based on this distribution, one must come up with instances

where nunation can co-occur with al-. Such instances are impossible to find unless

nunation is redefined with a new set of forms. This is what Fassi Fehri (1993) and

Lyons (1999) strive to establish.

The authors argue that the /ni/ and /na/ endings (/nV/) of the number/case suffixes

-aːni/ayni of dual nouns and -uːna/iːna of SndM plural nouns are variants of nunation.

As a consequence, nunation – which now includes /nV/ and -n(noun) – cannot be a marker

of indefiniteness since /nV/ can co-occur with the definite prefix. The argument that

/nV/ is a nunation is motivated by the fact that when duals and SndM plurals occur as

the head nouns of CSs, the /nV/ part of their suffixes is dropped. This is the same

behaviour that -n(noun) shows in the same environment. In the following examples of CSs

provided by Fassi Fehri (1993), compare the singular head noun daːr (house) to the

SndM plural head noun muslim-uː (Muslims):

194 Chapter Seven – Nunation

(2) a. daxal-tu daːr-a 110r-raʒul-i (Fassi Fehri, 1993:217) entered.I house-Acc Def-man-Gen ‘I entered the man’s house.’

b. muslim-uː l-madiːnat-i (Fassi Fehri, 1993:217) muslim-Pl.Nom Def-city-Gen ‘The Muslims of the city.’

Observe that the singular noun daːr (house) in (2.a) is bare (no nunation or al-) as it is in

the position of a head noun of a CS, as explained in the previous chapter. What is

significant for Fassi Fehri (1993) is that when nunation is absent from the singular head

noun, the /na/ part that usually appears after the -uː suffix of the SndM noun muslim in

(2.b) is absent as well. The same observation is made by Lyons (1999), who provides

the following CSs examples with dual and SndM nouns as their heads:

(3) a. ʕayn-a l-bint-i (Lyons, 1999:93) eye-Dl.Nom Def-girl-Gen ‘The (two) eyes of the girl.’

b. xaːdim-u l-malik-i (Lyons, 1999:93) servant-Pl.Nom Def-king-Gen ‘The servants of the king.’

In addition to the /nV/ dropping, which this study acknowledges, observe that Lyons

(1999) presents the long vowel /aː/ of the dual nominative suffix as /a/ (3.a), and the

long vowel of the nominative SndM plural suffix /uː/ as /u/ (3.b).

The author claims that shortening the long vowels is part of some modifications

that the number/case suffixes of duals and SndM plurals undergo only in CSs. Lyons

(1999:92-93) states the following:

The construct state differs, however, in that the final -ni is dropped, leaving the dual ending as nominative -aː (undergoing shortening to -a before the article of the possessor noun phrase) and oblique -ay (modified to -ayi before the article) … A similar situation obtains with the external plural [SndM plurals], which is formed by addition to the stem of -uːna (nominative) and -iːna (oblique), for both definites and indefinites; but the final -na is dropped (and resulting final -uː and -iː shortened before a following article) for the construct state.

110 Fassi Fehri (1993) presents the definite prefix al- in its final forms. In addition, he does not provide an example of duals in CSs perhaps to avoid repetition as duals will show the same behavior of dropping their /nV/ that is seen in the SndM example. 195 Chapter Seven – Nunation

In fact, this is not accurate. Shortening the vowels of the suffixes automatically results

in ambiguity that does not exist in SA. For example, shortening the vowel /aː/ of the

dual nominative suffix makes it identical to the singular accusative suffix /a/. This

results in having the noun ʕayn-a instead of ʕayn-aː, which changes the meaning of the

example in (3.a) into the eye of the girl instead of the intended meaning the (two) eyes of

the girl. It is obvious that after dropping the /ni/ part, the long /aː/ in CSs is what

distinguishes the dual nominative reading from the singular accusative reading of the

noun. Likewise, shortening the vowel /uː/ of the SndM nominative suffix into /u/ creates

a similar ambiguity. The misrepresentation of the suffix in (3.b), xaːdim-u instead of

xaːdim-uː, changes the intended meaning of the example from the servants of the king

into the servant of the king. The length of the vowel in CSs is what makes the

distinction between the plural and the singular meanings of the nouns.

The claimed shortening gives an impression that, in non-CS environments, the

long vowels -aː and -uː that precede the /nV/ part of the dual and SndM suffixes are also

variants of the case suffixes u-, a- and i- of the other types of nouns. Therefore, the long

vowels return to their original (short) forms in CSs. This view allows separating the

long vowels from /nV/, which entails that they are two functionally different elements

as suggested by Lyons (1999:93) below:

Now, two things are striking here. First, the plural endings -uːna, -iːna look biomorphemic; the first syllable, -uː-, -iː-, closely resembles the nominative or genitive inflection of the singular, differing only in length. Second, the second syllable of both the plural and the dual endings contain [n], thus closely resembling phonologically the “indefinite article” -n (nunation).

Observe that the author compares the first syllables -uː- and -iː- of the SndM suffixes to

the nominative and genitive case suffixes -u and -i of the singulars. While it is true that

these syllables have forms and functions that are largely similar to the singular case

suffixes, the author does not mention how the first syllables -aː- and -ay- of the dual

suffixes are different from the singular case suffixes. As can be clearly seen, -aː- in the

196 Chapter Seven – Nunation

dual suffix is nominative, while the closest singular case suffix to it is -a, which is

accusative. Likewise, -ay- that is the first syllable of the non-nominative dual suffix

does not resemble any of its singular counterparts.

Given that /nV/ is dropped in the same environment where -n(noun) is dropped,

Fassi Fehri (1993) and Lyons (1999) argue that both forms are nunation. However,

since /nV/, as a variant of nunation, can co-occur with the definite prefix al- in non-CS

environments, as shown below, nunation and al- are no longer prohibited from co-

occuring, which leads to the conclusion that nunation – including -n(noun) – is not an

indefinite marker.

(4) a. r-raʒul-aː-n (Fassi Fehri, 1993:217) Def-man-Dl.Nom-n ‘The two men.’

b. l-muslim-uː-n (Fassi Fehri, 1993:217) Def-muslim-Pl.Nom-n ‘The Muslims.’

c. al-ʕayn-aːni (Lyons, 1999:93) Def-eye-Nom.Dl ‘The two eyes.’

d. al-xaːdim-uːna (Lyons, 1999:93) Def-servant-Nom.Pl ‘The servants.’

The above are examples of definite duals and SndM plurals that are meant to show how

nunation, as redefined by the authors, can co-occur with the definite prefix. Upon this,

the authors argue that nunation is not an indefinite marker, as shown below:

The nature of nunation has been (and is still) a real puzzle for Arabic grammarians. Most of them treat this form (…) as an indefinite article, but this view is hardly tenable. (Fassi Fehri, 1993:216).

But if this is correct, the -n/-ni/-na ending cannot encode indefiniteness, since it sometimes co-occurs with al-…This analysis of nunation and the dual and plural endings differs from the accounts usually given in the manuals. But I believe it is plausible, and if it is correct, it puts Arabic among the languages which have a definite, but no real indefinite article. (Lyons, 1999:93-94).

197 Chapter Seven – Nunation

It is important to point out that in the examples (4.a-b) provided by Fassi Fehri

(1993), what is supposed to be /nV/ is presented as /n/, which makes it identical to the

nunation found in singulars, broken and SndF plurals. This is another misrepresentation

of the data that may give an impression that duals and SndM plurals can take the same

form of nunation that singulars and the other nouns take. Apart from this, Fassi Fehri

(1993) separates the /n/ part, which is supposed to be /nV/, and gives it no function in

the gloss. In the examples of Lyons (1999) in (4.c-d), however, the /nV/ parts are

represented in the right form, but they are not separated despite the author’s suggestion

that /nV/ is a variant of nunation. Putting the technicality aside, both authors use the

above examples as instances of the co-occurrence of nunation and the definite prefix,

which they claim is the evidence supporting their argument that nunation is not an

indefinite marker.

7.2.2. The Alternative Functions of Nunation

After arguing that nunation does not function as an indefinite marker, the authors

speculate on the new function it may have. Given that CSs are usually composed of

possessed (head) and possessor (complement) nouns and the former is the noun that

loses its nunation, be it -n(noun) or /nV/, Fassi Fehri (1993:273) hypothesises that

“nunation realizes Poss just in case the Possessor is not present”. In other words,

nunation is claimed to be what compensates for the absence of a possessor noun.

According to the new function, the absence/presence of nunation is determined by the

absence/presence of a possessor noun in the sentence. This leaves us with only two

options as to when nunation should or should not appear. Namely, nunation should

appear when the possessor noun is absent, e.g., in non-CS sentences, and it should not

appear when the possessor noun is present, e.g., in CS sentences.

198 Chapter Seven – Nunation

For /nV/, this new function seems plausible as the /nV/ ending strongly correlates

with the absence/presence of possessor nouns. That is, /nV/ appears on dual and SndM

plural nouns in non-CS sentences (where there are no possessor nouns) even when the

nouns are definite. It disappears when the nouns are heads of CSs, which is the

environment where the possessor nouns are present. However, the new function does

not capture the disappearance of -n(noun) from definite nouns in non-CS sentences.

Although possessor nouns are usually absent in non-CS sentences, nunation still

disappears; obviously because the definite prefix al- is present. If -n(noun) is not an

indefinite marker, but rather a Poss marker, as argued by Fassi Fehri (1993), it should

not be affected by the definiteness status of its noun in non-CS environments. It should

appear even when the definite al- is present, which is not the case as shown below:

(5) daxal-tu al-daːr-a / *al-daːr-a-n entered.I Def-house-Acc ‘I entered the house.’

According to the new function assigned to nunation by Fassi Fehri (1993), the absence

of a noun referring to the possessor of the house (daːr) should allow nunation to appear.

However, to the contrary, nunation is still not allowed to appear. This brings us back to

the contrastive distribution of -n(noun) and the definite prefix al- that simply cannot be

ignored. Lyons (1999) tries to overcome this distribution in a different way.

Unlike Fassi Fehri (1993), Lyons (1999) does not associate the new function of

nunation with possessiveness. Without elaborating much, he alternatively suggests that

nunation is a marker of nominality, as stated below (Lyons, 1999:93-94):

It is probably a semantically empty marker of nominality, which (like quasi-indefinite articles) indirectly indicates indefiniteness because of its partial complementary distribution with definite determiners.

With the new function of nunation as a marker of nominality, Lyons (1999:93)

attributes the contrastive distribution of -n(noun) and al- to phonological reasons, as stated

below:

199 Chapter Seven – Nunation

Where the noun is made definite by the attachment of the article al-, the “weaker” -n variant is dropped, but the phonologically fuller -ni and -na are not.

This makes nunation a semantically empty (morphological) marker of nominality whose

distribution is phonologically determined. If nunation is a marker of nominality and

dropping its “weaker” version -n is conditioned by the presence of al-, as argued by

Lyons (1999), why is it drooped from head nouns of CSs in the absence of al-?

It is obvious that labelling -n(noun) and /nV/ as nunation in order to account for the

mutual behaviour they have in CSs has come at the expense of the (non-discountable)

strong contrastive distribution of -n(noun) and the definite prefix al-. It is obvious that

because -n(noun) is clearly an indefinite marker, it is dropped from the head nouns of CSs

and from definite nouns in non-CSs. As argued in the previous chapter, -n(noun) is the VI

inserted into the feature [-Def]. In CSs, this feature is not generated under (the null) Dgen

where the head nouns rise (Ritter, 1991), so -n(noun) cannot be inserted. In non-CSs, on

the other hand, definite nouns are equipped with the feature [+Def] into which /al/ is

inserted, so -n(noun) as the VI of [-Def] cannot appear. For /nV/ to have a mutual

behaviour with -n(noun) in a single environment does not necessarily mean that it belongs

111 the same category to which -n(noun) belongs. They need more than one environment in

which they have something in common to be considered nunation, and CSs is the only

environment they have.

If we look closely at the relation that each of -n(noun) and /nV/ bears to the definite

prefix al-, we find that -n(noun) and al- are in a strong relationship, while the relationship

between /nV/ and al- is much weaker. On the one hand, -n(noun) disappears when al-

appears, and al- disappears when -n(noun) appears in non-CS sentences. In CS sentences,

both disappear leaving the head noun bare. It is clear that they canno co-occur. On the

111 There is a good chance that /nV/ alone is a Poss marker, and this is why it is dropped in CSs. However, this raises another question about why duals and SndM plurals are the only nouns marked for this function. For this study, /nV/ remains part of the number/case suffixes of duals and SndM plurals until proven otherwise. 200 Chapter Seven – Nunation

other hand, /nV/ and al- can appear together (definite duals and SndM plurals in non-

CSs) or disappear together (head nouns of CSs), and /nV/ can appear alone (indefinite

duals and SndM plurals in non-CSs). In the vast majority of instances, /nV/ is not

affected by the appearance/disappearance of al-, whereas in all the instances where /nV/

is involved, its appearance/disappearance is conditioned by the presence of the

case/number suffixes of duals and SndM plurals. There is no single environment in SA

where /nV/ can appear without the number/case suffixes. Even when /nV/ disappears,

the suffixes – or what is left of them – must be present. It is obvious that, unlike -n(noun),

/nV/ is a totally dependent element, and part of the number/case suffixes of its nouns.

All these differences indicate that -n(noun) and /nV/ do not belong to the same category.

They are two different elements that simply happen to have a mutual behaviour only in

one environment.

In order to emphasise this point, another behavioural difference between -n(noun)

and /nV/ will be demonstrated in the next section.

7.3. -n(noun) and /nV/ in Vocative Contexts

The aim of this section is to establish another difference between -n(noun) and /nV/,

which, in addition to the differences they already have in non-CS environments, will

suffice to demonstrate that they do not belong to the same category. Consequently, this

will show that the claim that /nV/ is a variant of nunation cannot be maintained, thus

arguing against the unified treatment of -n(noun) and /nV/. This support comes from the

different behaviour that each of these elements exhibits in vocative contexts where the

vocative particle yaː is involved.

The vocative particle yaː can be translated into English as ‘O’, mostly in religious

and literary contexts, or hey in informal settings. It serves several semantic/pragmatic

201 Chapter Seven – Nunation

purposes such as addressing,112 attracting the attention of someone, supplicating, or

expressing exclamation (Ryding, 2005). What this section is concerned with is the

vocative contexts in which the particle precedes the nouns that refer to the addressees or

the persons whose attention is to be attracted. In such context, the nouns can have three

forms depending on the specificity of their referents. These forms are described in

details in the traditional literature of Arabic grammar (Al-Afghani, 1971; Hasan, 1978;

Ghalayini, 1993; Fayadh, 1995; Al-Fozan, 1999; Haroun, 2001; Al-Dahdah, 2007;

Ashaikh, 2009, to name a few). The skeletons of these three forms can be sketched out

as the following:

(6) a. yaː + N-Acc-nunation (Indefinite non-specific)

b. yaː + N-Nom (Specific)

c. yaː + ʔayuhaː + al-N-Nom (Definite Specific)

The first form is an accusative noun with a nunation.113 This form is used when the

speech is general and there is no specific addressee. The second form is simply a bare

nominative noun that is used when the addressee is specific. However, in order for the

definite prefix al- to be used with the second form, another particle, which is ʔayuhaː

(roughly = you with a vocative sense), must precede the noun. This third form is also

used with specific addressees.

One way to discover whether /nV/ is a variant of nunation, as argued by Fassi

Fehri (1993) and Lyons (1999), is to have its behaviour in vocative environments

compared to the behaviour of -n(noun), which is claimed to be the other variant of

nunation. This section focuses on SA nouns: singular (where -n occurs), dual (where -ni

occurs), and SndM plural (where -na occurs). This comparison will introduce us to

112 It can also be used to address things, usually, in the sense of giving them orders, e.g.: yaː ʔardʕ-u ʔiblaʕ-iː maːʔ-a-ki [Quran 11:44] O earth-Nom swallow-you water-Acc-yours ‘O earth, swallow your water!’ 113 Unless the noun is not allowed to take nunation, e.g., diptote or invariable. 202 Chapter Seven – Nunation

another difference between -n(noun) and /nV/ which rules out the possibility that the latter

is a variant of nunation.

As sketched out above, a singular noun such as raʒul (man) can have three forms

in vocative contexts, as shown below:

(7) a. yaː raʒul-a-n xuð bi-yad-iː hey man-Acc-n114 take by-hand-my ‘Hey man, take my hand.’

b. yaː raʒul-u xuð bi-yad-iː hey man-Nom take by-hand-my ‘Hey man, take my hand.’

c. yaː ʔayuhaː al-raʒul-u xuð bi-yad-iː hey ʔayuhaː Def-man-Nom take by-hand-my ‘Hey man, take my hand.’

The above three examples would match a situation where the speaker is seeking help.

The sentence in (7.a) embodies a situation where no specific man is addressed, e.g., the

speaker is seeking help from any man around.115 In this situation, the noun is accusative

with a nunation. Unlike the example in (7.a), the other examples represent a situation

where the speaker is seeking help from or addressing a specific man, so the noun is

nominative without a nunation, but capable of taking the definite prefix al- only when

preceded by ʔayuhaː.

It is unclear why the presence of the definite prefix is conditioned by the presence

of ʔayuhaː. Nevertheless, for nouns to have a certain case when non-specific and a

different case when specific is of course not unusual and is known in the literature on

case as differential case marking, which, cross-linguistically, can also be related to

114 As in Fassi Fehri (1993), no function is provided for the suffix -n in the gloss since it, along with /nV/, is under investigation. 115 Seeking help from anyone can also be expressed in several ways using different words such as ʔaħad (someone) or the imperative plural verb saːʕiduːniː (help me). 203 Chapter Seven – Nunation

animacy and definiteness, among other factors.116 What is significant here is the relation

between the semantic/pragmatic (non-)specificity of the nouns, on the one hand, and the

morphological case and (in)definiteness marking, on the other. The

accusative/nominative alternation along with the absence/presence of al- or nunation are

what make the distinction between specificity and non-specificity. Changing the

elements constituting the above sentences will result in ungrammaticality, as shown

below:

(8) a. *yaː raʒul-u-n xuð bi-yad-iː (no nominative and nunation)

b. *yaː raʒul-a xuð bi-yad-iː (no accusative without nunation)

c. *yaː al-raʒul-u xuð bi-yad-iː (no al- without ʔayuhaː)

d. *yaː ʔayuhaː al-raʒul-a xuð bi-yad-iː (no accusative and al-)

The examples in (7) and (8) above show that, in vocative contexts, the specificity of the

addressee is mainly expressed by the case of the noun. The noun is accusative when its

referent is non-specific, and nominative when the referent is specific.117 Nunation and

al-, then, behave accordingly. That is, when the noun is accusative (non-specific),

nunation is allowed, but al- is not. When the noun is nominative (specific), al- can be

allowed, but nunation is not.

If /nV/ is a variant of nunation, it would be expected to follow the same pattern,

but this is not the case. Starting with dual nouns, when a dual noun is preceded by the

vocative particle yaː, it takes the non-nominative118 form when its referents are non-

specific, and the nominative form when the referents are specific. As in singular nouns,

116 For more on the phenomenon of differential case marking and its relation to animacy and definiteness cross-linguistically, see de Hoop and Narasimhan (2005); de Hoop and Malchukov (2007); Amberber (2008); Malchukov (2008); Malchukov and de Swart (2008), to name a few. 117 Perhaps among several other views, Belletti (1988) argues that specific nouns are assigned structural case, while the case of non-specific nouns is inherent. Whether this can account for SA nouns in vocative contexts needs a special investigation that should consider how case is assigned in SA. As pointed out in the previous chapter, this study is not concerned with the issue of case assignment as it requires embarking on some aspects of phrase/clause structure that may fall beyond the scope of this thesis. 118 Recall that duals and SndM plurals distinguish only between nominative and non-nominative cases. They do not have special forms for the accusative case. 204 Chapter Seven – Nunation

the definite prefix al- is allowed to appear when the addressee noun is nominative and

preceded by ʔayuhaː, as shown below:

(9) a. yaː xasʕm-ayni ʔitafiqaː O opponent-Dl.Acc agree ‘O (two) opponents, come to an agreement.’

b. yaː xasʕm-aːni ʔitafiqaː O opponent-Dl.Nom agree ‘O (two) opponents, come to an agreement.’

c. yaː ʔayuhaː al-xasʕm-aːni ʔitafiqaː O ʔayuhaː Def-opponent-Dl.Nom agree ‘O (two) opponents, come to an agreement.’

The above examples represent a situation where the speaker is addressing two

opponents and asking them to end their dispute and come to an agreement. The situation

in (9.a) is more likely to be part of a general preaching or sermon in which the speaker

is encouraging any/every two opponents/parties to end their disputes. The speeches in

(9.b-c), however, are more likely to be delivered by a person mediating between two

specific opponents and encouraging them to end their dispute. As can be seen above, the

addressee noun in the non-specific vocative context is non-nominative, while the same

noun is nominative when its referents are specific. What is striking here is the presence

of /ni/ in the specific vocative context in (9.b-c). The same behaviour can be seen in

SndM plurals as well, as shown below:

(10) a. yaː muslim-iːna tamasakuː bi-al-quraːn-i O muslim-Pl.Acc grasp by-Def-Quran-Obl ‘O Muslims, grasp (a handhold on) the Quran.’

b. yaː muslim-uːna tamasakuː bi-al-quraːn-i O muslim-Pl.Nom grasp by-Def-Quran-Obl ‘O Muslims, grasp (a handhold on) the Quran.’

c. yaː ʔayuhaː al-muslim-uːna tamasakuː bi-al-quraːn-i O ʔayuhaː Def-muslim-Pl.Nom grasp by-Def-Quran-Obl ‘O Muslims, grasp (a handhold on) the Quran.’

205 Chapter Seven – Nunation

The semantic/pragmatic implication of having a non-nominative SndM plural noun in

the vocative context in (10.a) and a nominative noun in (10.b-c) is that the former noun

refers to all Muslims, while the referents of the latter nouns are a specific group of

Muslims. Again, /na/ is allowed to appear on the nominative SndM plural nouns in

(10.b-c) even though they are specific.

In (9.b-c) and (10.b-c) above, we have seen /nV/ allowed to appear in specific

environments, while -n(noun) is not permitted in the specific environments in (7.b-c).

Besides the appearance of /nV/ as compared to the disappearance of -n(noun) in definite

non-CS environments, the different behaviour that /nV/ and -n(noun) show in vocative

contexts is another indication that they are not variants of a single category called

nunation. The absence of -n(noun) in nominative (specific) nouns in vocative contexts

was shown to be possible, or rather compulsory, while removing /nV/ from the

nominative (specific) nouns in (9.b-c) and (10.b-c) above is prohibited, as shown below:

(11) a. *yaː xasʕm-aː ʔitafiqaː

d. *yaː ʔayuhaː al-xasʕm-aː ʔitafiqaː

c. *yaː muslim-uː tamasakuː bi-al-quraːn-i

e. *yaː ʔayuhaː al-muslim-uː tamasakuː bi-al-quraːn-i

Moreover, the behaviour that /nV/ and -n(noun) show in vocative contexts is

problematic to the analyses of Fassi Fehri (1993) and Lyons (1999). Recall that Fassi

Fehri (1993) claims that nunation is a Poss marker that is employed to compensate for

the absence of a possessor noun. If this is what -n(noun) does on singular nouns, why is its

appearance/disappearance conditioned by the specificity of its noun? There are no

possessor nouns in all the singular examples in (7). This should, consequently, allow the

suffix -n to appear. However, it appears only when the noun is non-specific and

disappears when the noun is specific whether, in the latter case, the noun has al- or not.

206 Chapter Seven – Nunation

It is obvious that, in vocative contexts, -n(noun) is an indefinite marker that correlates

with the non-specificity119 of its noun rather than with the absence/presence of a

possessor noun.

In addition, recall that Lyons (1999) argues that when a noun is made definite by

the prefix al-, the -n variant of nunation (-n(noun)) is dropped because it is phonologically

weaker, while /nV/ is not dropped because it is phonologically fuller. This does not

explain why the nunation of the specific singular noun in (7.b) is dropped despite the

noun not having al-. Obviously, no phonological factors are involved here.

To sum up, it was shown in this section that when a noun occurs in a vocative

context, there is a dependency relationship between specificity and case. The noun is

accusative when its referent is non-specific, and nominative otherwise. On the one hand,

the behaviour of -n(noun) in these environments indicates its function as an indefinite

marker that appears when the noun is non-specific, and disappears when the noun is

specific. On the other hand, /nV/ has a different pattern in vocative contexts. It appears

on vocative nouns whether the nouns are specific or not, and whether the definite prefix

al- is present or absent. Unlike -n(noun), /nV/ does not show any relation to the

(in)definiteness or (non-)specificity of the vocative nouns on which it appears. It was

also shown that the analysis of Fassi Fehri (1993) cannot account for the absence of

-n(noun) in nominative vocative nouns, and nor can the analysis of Lyons (1999) capture

the disappearance of -n(noun) from nominative vocative nouns that do not have the

definite prefix.

119 Morphological (In)definiteness and semantic (non-)specificity do not necessarily correlate. As will be shown at the end of this chapter, a different form of nunation that is found in Bedouin dialects is likely to be an indefinite-specific marker. For example, in a sentence like I bought a car, the noun car is morphologically indefinite although it refers to a specific car. For more on the relation between (In)definiteness and (non-)specificity, see Heusinger (2002). 207 Chapter Seven – Nunation

7.4. Nunation on Names

An interesting fact about SA names is that they sometimes take the suffix -n (-n(name)).

Recall Fassi Fehri’s (1993) argument that since nunation can appear on names, it cannot

be an indefinite marker. This implies that -n(name) and -n(noun) are (functionally) the same,

perhaps because of their shared phonological form. This section aims to show that

although -n(name) and -n(noun) have the same phonological form and are both called

nunation, they are functionally different.

In Chapter Six, it was shown that -n(noun) is an indefinite marker, which, as shown

in the previous section, correlates with the non-specificity of the nouns in vocative

contexts. The analysis of this section will reveal that -n(name) cannot be a marker of

indefiniteness. Rather, it marks the inherent specificity of names that are derived from

existing nouns and adjectives. The analysis will also reveal that the appearance of

-n(name) becomes unnecessary when other elements that may (morphologically) express

the specificity of the names are present. If this study succeeds in demonstrating that

-n(name) is different from -n(noun), Fassi Fehri’s (1993) argument will be dismissed, and

-n(noun) will maintain its function as an indefinite marker.

In order to argue that -n(name) cannot be an indefinite marker as -n(noun) is, one

needs to show that when names take -n(name), they behave differently from nouns that

take -n(noun). For this purpose, a comparison between the behaviour of nouns and names

in adjectival environments will be made. The comparison will reveal that names that

take nunation behave similarly to definite nouns that cannot take nunation. That is,

names are definite despite having nunation, which indicates that the nunation they have

is different from the nunation that indefinite nouns take.

As in many other languages, Arabic adjectives can be attributive or predicative. In

SA, attributive adjectives simply agree with their nouns in (in)definiteness, as shown in

208 Chapter Seven – Nunation

(12.a-b). Predicative adjectives, however, are indefinite while the nouns they modify are

definite, as shown in (12.c) below (Ryding, 2005:239-244):

(12) a. al-hilaːl-u al-xasʕiːb-u (Def. N + Def. Adj) Def-cresent-Nom Def-fertile-Nom ‘The Fertile Crescent.’

b. nasʕiːħ-at-u-n maʒʒaːniy-at-u-n (Idef. N + Idef. Adj) advice-F-Nom-Idef free-F-Nom-Idef ‘Free advice.’

c. al-ħasʕaːd-u wafiːr-u-n120 (Def. N + Idef. Adj) Def-harvest-Nom abundant-Nom-Idef ‘The harvest is abundant.’

When a name that has nunation is modified by an adjective, the adjective is either

definite (attributive) or indefinite (predicative), a shown below:

(13) a. muħammad-u-n al-xaːmis-u Muhammad-Nom-nunation Def-fifth-Nom ‘Muhammad V (a former King of Morocco).’

121 b. muħammad-u-n karːim-u-n Muhammad-Nom-nunation generous-Nom-Idef ‘Muhammad is generous.’

Such a combination of an element that has a nunation being modified by a definite

adjective, as seen in (13.a), is acceptable only when the modified element is a name.

When the modified element is a noun, the noun and its modifying adjective can be

Def+Def, Idef+Idef, or Def+Idef, but never Idef+Def. As far as nunation is concerned, a

name with -n(name) is allowed to be modified by a definite adjective, as shown in (13.a)

120 This is an example of what is known as a verbless sentence, as mentioned in the previous chapter. The absence of a present tense copula in Arabic is probably what makes it considered verbless. The sentence could be a short form of the following sentence: al-ħasʕaːd-u ħasʕaːd-u-n wafiːr-u-n Def-harvest-Nom harvest-Nom-Idef abundant-Nom-Idef ‘The harvest is an abundant harvest.’ 121 Nunation appears on the adjective as an indefinite marker because the adjective is predicative, not because the name is indefinite. Again, the sentence could be a short form of the following sentence: muħammad-u-n raʒul-u-n karːim-u-n Muhammad-Nom-nunation man-Nom-Idef generous-Nom-Idef ‘Muhammad is a generous man.’

209 Chapter Seven – Nunation

above, but definite adjectives are not allowed to modify nouns with -n(noun), as shown

below:

(14) a. *hilaːl-u-n al-xasʕiːb-u cresent-Nom-Idef Def-fertile-Nom ‘A crescent the fertile.’

b. *nasʕiːħ-at-u-n al-maʒʒaːniy-at-u advice-F-Nom-Idef Def-free-F-Nom ‘An advice the free.’

c. *ħasʕaːd-u-n al-wafiːr-u harvest-Nom-Idef Def-abundant-Nom ‘A harvest the abundant.’

What is significant here is that -n(name) is permitted while its bearer is modified by a

definite adjective, but -n(noun) is not. It is clear that, unlike -n(noun), -n(name) does not

render its name indefinite. On the contrary, when -n(name) appears on a name, the name

behaves as definite, while the noun that has -n(noun) has no option but to behave as

indefinite. This is a strong indication that -n(name) is not functionally the same as -n(noun),

even though they have the same phonological form. It is obvious that names are

inherently definite, and this explains why they share the behaviour of the overtly

definite nouns. This gives rise to the question: what is the function of -n(name)?

In order to discover what -n(name) does when it occurs on names, it would be useful

to explore how names are generally derived in Arabic. According to Al-Suhaili (1114 –

1185),122 Arabic names are either transferred or non-transferred. Transferred names are

those which pre-exist in Arabic as words of different parts of speech, i.e., nouns,

adjectives and verbs that are employed as names. For example, /ʒaʕfar/ (literally = river

‘N’), /muhammad/ (literally = praised ‘Adj’) and /yaziːd/ (literally = he increases ‘V’)

are also names. Non-transferred names, on the other hand, include foreign names, e.g.,

/ʔibraːhiːm/ (Abraham) and /ʔismaːʕiːl/ (Ishmael), as well as names made of corrupted

122 The manuscripts (dictations) of Al-Suhaili were collected in a book by Albanna (1969). 210 Chapter Seven – Nunation

forms of existing words such as Omar (pronounced as /ʕumar/, which is a corrupted

form of the adjective /ʕaːmir/ that literally means ample or flourished). Al-Suhaili stated

that among these categories, names that can take nunation are those which could take

nunation before becoming names. That is, names derived from nouns and adjectives that

take nunation can also take nunation, while names derived from verbs as well as

123 foreign and corrupted names cannot. If Al-Suhaili’s generalisation is correct, -n(name)

is a marker by which a name is distinguished from its origin. It is a marker that conveys

that the names holding it are no longer normal nouns or adjectives.

Although this might be convincing, distinguishing between names and their

origins does not seem to be the only function that -n(name) has. This study assumes that

-n(name) also marks names for their semantic specificity. Unlike the nouns and the

adjectives from which they are derived, names have more specific referents and

denotations. They now work as labels by which certain entities can be identified, so

they become less likely to denote non-specific referents than their former nouns and

adjectives. If this is actually the case, -n(name) does not only distinguish names from their

origins, it also morphologically indicates that the name it marks is now more

semantically specific.

For the sake of argument, let us assume that the following names have occurred

with a nunation in the following Koranic examples124 because their referents are

semantically specific:

ʕ (15) a. falammaː ʒaːʔa ʔamrunaː naʒʒaynaː s aːliħ-a-n [Quran 11:66] when came decree.our saved.we Salih-Acc-n ‘When our decree came to pass, we saved Salih.’

123 An exception to this generalisation would be a few foreign names such as /nuːħ/ (Noah) and /luːtʕ/ (Lot) that can take nunation. 124 Names can take nunation in SA as well. However, it is easier to find names that are mentioned repeatedly across different environments in the Quran than in SA texts. The purpose will become clearer when the occurrence of the names mentioned in these examples is compared to their occurrence in some other Koranic examples that will be shown very soon. 211 Chapter Seven – Nunation

ʕ b. qaːla lahum ʔaxuhum s aːliħ-u-n ʔalaː tataquːn [Quran 26:142] said.he to.them brother.their Salih-Nom-n don’t you.fear ‘Their brother Salih asked them: Do not you fear (God)?’

c. laʔini ʔitabaʕtum ∫uʕayb-a-n ʔinakum ʔiðan la-xaːsiruːna [Quran 7:90] certainly.if follow.you Shuayb-Acc-n you.are thus certainly.losers ‘If you follow Shuayb, you will certainly lose.’

d. ʔallaðiːna kaððabuː ∫uʕayb-a-n kaːnuː humu al-xaːsiriːna [Quran 7:92] those.who distrusted Shuayb-Acc-n were them Def-losers ‘Those who called Shuayb a liar were the ones to lose.’

In the above examples, there are two names referring to two specific men: Salih (a pre-

Islamic messenger whose name means good ‘Adj’) and Shuayb (a pre-Islamic

messenger whose name, according to Duraidi (2000), means a little underground water

stream ‘N’). According to the assumption made in this study, these names have

nunation to mark the specificity of their referents as compared to the lesser specificity

of the noun and the adjective from which they are derived. The appearance of -n(name)

per se, however, cannot be strong evidence that it is a specificity marker. Therefore, the

environments where the -n(name) of these names is dropped needs to be investigated.

An environment where -n(name) is dropped is vocation. Recall that when a noun

occurs in a vocative context, it is either accusative with a nunation (non-specific), or

nominative without a nunation (specific). The behaviour of names in vocative contexts

is identical to the behaviour of specific nouns. That is, when a name is preceded by the

vocative particle yaː, the name is nominative125 without its nunation, as shown below:

ʕ (16) a. qaːluː yaː s aːliħ-u ʔiʔtinaː bimaː taʕidunaː [Quran 7:77] said.they O Salih-Nom bring.us by.what you.promise.us ‘They said: O Salih, bring us what you promise.’

ʕ b. qaːluː yaː s aːliħ-u qad kunnta fiːnaː marʒuwa [Quran 11:62] said.they O Salih-Nom aux were.you in.us hoped ‘They said: O Salih, we had hope in you.’

125 This is another piece of evidence that names behave as definite even when they have nunation. 212 Chapter Seven – Nunation

c. la-nuxriʒanaka yaː ∫uʕayb-u [Quran 7:88] certainly-we.expel.you O Shuayb-Nom ‘We will certainly expel you, O Shuayb.’

ʕ d. qaːluː yaː ∫uʕayb-u, ʔa-s allaːtuka taʔmuruka …. [Quran 11:87] said.they O Shuayb-Nom, does-prayer.your order.you…. ‘They said: O Shuayb, does your prayer order you ….’

At a glance, this may appear confusing or even counterintuitive to the assumption that

-n(name) is a specific marker. If names are specific, they would be expected to be

nominative in vocative contexts, which is exactly how they appear in the above

examples. However, if -n(name) is what marks their specificity, as assumed here, we

would expect it to remain attached to the names, which is not actually the case. Thus, a

legitimate question is: if -n(name) is a specificity marker, why is it dropped from the

semantically specific names in the above vocative contexts?

Here, we argue that -n(name) marks its names for their specificity only in the

absence of any other element or context that may explicitly serve the same purpose.

That is, it simply becomes unnecessary (or even redundant) to morphologically mark

the specificity of a name with -n(name) when the referent of the name is directly

addressed.126 Direct addressing narrows the range of reference of the name, and yaː the

vocative particle is the morphological element by which this is achieved, so -n(name) is no

longer needed. That is, the semantic/pragmatic specificity of names needs to be

morphologically expressed, and -n(name) seems to be the responsible element. However,

when another element, which is yaː in this instance, is present and can fill the same role,

-n(name) is dropped.

A similar environment where the range of reference of names is narrowed by a

morphological element, so that -n(name) as a specificity marker is dropped, can be

observed when ibn (literally = son of) is involved. The main use of ibn is to link a man’s

126 Recall that when the referents of nouns in vocative contexts are non-specific, e.g., when the speech is general, there is no direct addressing. 213 Chapter Seven – Nunation

name to the name of his father, e.g., John the son of Peter. In such a context in SA, ibn

modifies the name it follows by narrowing its range of reference to a specific person

who can now be precisely identified by the name of his father (the second name). When

a string of this sort occurs, the first name (the modified name) does not take -n(name) even

though it normally does. The second name, however, takes -n(name) as normal, as

exemplified below:

(17) ʒaʕfar-u ibn-u muħammad-i-n Jafar-Nom son-Nom Muhammad-Gen-nunation ‘Jafar (the son of) Muhammad.’127

It is important to point out that CS is involved in the above sentence. The head of the

CS is ibn and the second name is its complement. This is why the second name is

genitive. The first name, however, is not involved in the CS. If the first name is part of

the CS, i.e., its head, ibn would be expected to be genitive instead of nominative. What

is striking in the above example is that the -n(name) of the first name ʒafar is dropped.

This study argues that -n(name) is dropped because its name is modified by ibn, while the

-n(name) of the second name is maintained because of the lack of any element that could

compensate for its function as a specificity marker. Although both names are

semantically specific, the specificity of each name is morphologically expressed in a

different way. The specificity of the second name is morphologically expressed by

-n(name) (nunation), while the specificity of the first name is expressed by the presence of

ibn, which makes the appearance of -n(name) on the first name unnecessary.

To sum up, it was shown that although -n(noun) and -n(name) have the same

phonological form, they cannot have the same function. While -n(noun) is an indefinite

marker, the way names agree with their adjectives and behave in vocative contexts

indicates that names are definite, so -n(name) cannot be an indefinite marker. Relying on

127 An Islamic jurist also known as Jafar Alsadiq (702-765). 214 Chapter Seven – Nunation

Al-Suhaili’s classification of names, this study takes -n(name) to be the element by which

names are tagged differently from their words of origin, as well as the morphological

element by which names are marked for the high semantic specificity they inherently

have as opposed to their origins. In the presence of another morphological element that

can express the specificity of the names, e.g., yaː or ibn, the appearance of -n(name) on the

modified name becomes superfluous, so it is dropped.

In the next section, further evidence for the present analysis will be presented on

the basis of data from Bedouin dialects of Arabic. If the analysis of -n(name) provided in

the current section is on the right track, the nunation that appears on Bedouin nouns is

more likely to be a vestige of -n(name) rather than the indefinite -n(noun) as previously

thought.

7.5. Nunation in the Bedouin Dialects

Recall that in DA, nouns are not overtly marked for indefiniteness. A DA noun is

usually considered indefinite when it does not have the definite prefix al-. In this

respect, the nouns of the Bedouin dialects, especially the dialects of Central Arabia

known as Najd, do not behave differently. They are generally known to be indefinite

when they lack the definite prefix. In certain environments, however, a form of nunation

that is suffixed to the nouns in the form of -in128 is used (Holes, 1990; Ingham, 1994;

Brustad, 2000), as shown below:

(18) bint-in zeena, bint ʕamm-iː (Holes, 1990:116) girl-Idef good girl uncle-my ‘She is a good [looking] girl, my cousin.’

128 As in all varieties of DA, the nouns of the Bedouin dialects are not overtly marked for case (Ingham, 1994:49; Brustad, 2000:27). The /i/ part of the Bedouin nunation seems to be employed to bridge the gap made by the absence of the case suffixes. It does not change according to the position of the noun in the sentence, so it is unlikely to be related to case.

215 Chapter Seven – Nunation

While treating it as an indefinite marker, Holes (1990:115) states that -in attaches to

indefinite Bedouin nouns, i.e., nouns that do not have the definite prefix, only when

they are modified by attributive adjectives, as shown above. Holes’ (1990) treatment

entails that the Bedouin -in and the SA -n(noun) are both indefinite markers. However, the

appearance of the former is conditioned by the existence of an attributive adjective

modifying the indefinite noun. Therefore, Holes (1990:115) considers the Bedouin -in a

vestige of the SA indefinite suffix -n (-n(noun)).

Ingham (1994:47-50) also considers the Bedouin -in an indefinite marker.

However, he states that the occurrence of the suffix is not conditioned only by the

existence of an attributive adjective. Rather, it occurs in the environments where the

indefinite noun is modified by an adjective, a prepositional phrase or a modifying

clause, as shown below (Ingham, 1994:49):

(19) a. beet-in kibiːr (Modifying Adjective) house-Idef large ‘A large house.’

b. ʒizʔ-in min-h (Modifying Prepositional Phrase) part-Idef from-it ‘A part of it.’

c. kalmit-in gaːl-uː-ha l-iː (Modifying Clause) word-Idef said-they-it to-me ‘A word which they said to me.’

In addition to its function as an indefinite marker, Ingham (1994) describes the Bedouin

-in as a suffix that also functions as a juncture since it is usually used in non-pausal, and

rarely in utterance-final, positions. As far as (in)definiteness marking is concerned,

Ingham (1994:47) classifies Bedouin nouns into the following:

(20) Indefinite beet ‘house’ or ‘a house’

Indefinite (marked) beet-in ‘a (particular) house’

Definite al-beet ‘the house’

216 Chapter Seven – Nunation

Observe that the range of the overtly marked noun beet-in is narrower than the range of

the unmarked beet. The latter may refer to any house, while the reference of the former

is limited to large houses, as it occurred in (19.a) above. The range of reference of the

beet-in does not include small houses, so it certainly is more specific than beet.

Such an observation was made by Brustad (2000:28), who states that the Bedouin

-in “functions as an indefinite-specific marker” rather than just an indefinite marker, as

suggested by Holes (1990). Given that it appears on indefinite nouns (which do not

have al-) only when they are made specific by a modifier, the nouns on which the -in

suffix appears are not entirely indefinite. These nouns have a degree of specificity that

distinguishes them from the entirely indefinite nouns that do not take the suffix. The

author provides supporting evidence for this function of the Bedouin -in from old

Spanish Arabic texts, found in Corriente (1977), where this type of nunation occurred as

-an,129 as shown below:

(21) a. kalb-an abyadʕ dog-an white ‘A white dog.’

b. ʕala qalb-an kaːfir on heart-an unfaithful. ‘With an unfaithful heart.’

c. bi-xaːtʕir-an yattaqad miӨl al-naːr with-mind-an flaming like Def-fire ‘With a mind as bright as fire.’

As shown above, the nouns take the suffix -an when they are made more specific by a

modifier. The range of reference of the nouns above are limited to white dogs, unfaithful

hearts and bright minds respectively. Many entities that can be generally referred to by

the nouns dog, heart and mind are excluded.

129 Again, the /a/ part of the suffix does not seem to be related to case. The examples (21.b-c) show the nouns in oblique positions in which they would be marked for case with /i/ in SA. It appears that, unlike SA and the older stages, Spanish Arabic lacked an overt case-marking system, and the /a/ part of the suffix is employed to bridge the gap made by the absence of a case suffix. 217 Chapter Seven – Nunation

This analysis of the Bedouin -in and the Spanish Arabic -an implies that each of

these suffixes has two aspects: the morphological aspect that makes it an indefinite

marker, and the semantic/pragmatic aspect that makes it a marker of specificity. This

overlaps with the analysis of the SA -n(name) provided in the previous section in that

semantic/pragmatic specificity can be morphologically marked. The difference between

-n(name), on the one hand, and the Bedouin -in and the Spanish Arabic -an, on the other,

is that -n(name) morphologically marks (inherently definite) names for their semantic

specificity, while -in and -an mark their (indefinite) nouns for the specificity they obtain

by external modifiers. Therefore, the appearance of -in and -an is conditioned by the

appearance of a specifying element, while the disappearance of -n(name) is conditioned by

the appearance of an element that is capable of playing the morpho-semantic role of

-n(name).

If this turns to actually be the case, the Bedouin -in is more likely to be a vestige

of the specificity marker -n(name), which contrasts with the view of Holes (1990) who

suggests that it is a vestige of the indefinite marker -n(noun). To a large degree, this

should lend support to the analysis of -n(name) provided in the previous section.

Moreover, it should also lend support to the distinction this study makes between -n(noun)

and -n(name), which consequently rules out Fassi Fehri’s (1993) argument that nunation,

including -n(noun), cannot be an indefinite marker because it can appear on names.

7.6. Conclusion

The main objective of this chapter was to provide a counterargument to the claim made

by Fassi Fehri (1993) and Lyons (1999) that nunation cannot be an indefinite marker.

The backbone of this claim was shown to be the authors’ treatment of /nV/ as a variant

of nunation, as well as Fassi Fehri’s (1993) assumption that the nunation of nouns is

functionally the same nunation that appears on names. The claim was countered by 218 Chapter Seven – Nunation

demonstrating that /nV/ cannot be a variant of nunation, and what is known as nunation

is a set of two phonologically identical, but functionally different variants: -n(noun) and

-n(name).

It was shown that although /nV/ and -n(noun) have a mutual behaviour in CSs, they

still have more differences that effectively undermine any attempt to group them

together in a single category. Their many differences clearly outweigh their single

similarity. Obviously, -n(noun) contrasts with the definite prefix al-, while /nV/ does not.

The former never co-occurs with the definite prefix, while /nV/ and the definite prefix

can co-occur in definite duals and SndM plurals. To highlight another difference, the

behaviour of -n(noun) and /nV/ in vocative contexts was examined. It was shown that the

appearance/disappearance of -n(noun) in vocative contexts is strictly conditioned by the

specificity of the referent of its noun, while the whole specificity issue appears to be

totally irrelevant to /nV/. Moreover, it was shown that none of the different alternative

functions that the authors assign to the redefined nunation is fully capable of accounting

for the full range of contexts within which -n(noun) appears and disappears. Neither

labelling it as a Poss marker (Fassi Fehri, 1993) nor as a semantically empty marker of

nominality that is dropped for phonological reasons (Lyons, 1999) could effectively

account for the actual behaviour of -n(noun) in vocative and non-CS environments.

As for -n(name), it was argued that due to their behaviour as definite, names cannot

be indefinite, so the -n(name) they take cannot be an indefinite marker. Alternatively, it

was proposed that -n(name) functions as a specificity marker by which the narrower range

of reference that names have is distinguished from the wider range of their origins. The

higher semantic/pragmatic specificity of names was assumed to obligate a

morphological expression, which is usually achieved by the presence of -n(name).

However, in the environments where another element such as yaː (vocative particle) or

219 Chapter Seven – Nunation

ibn (son of) accomplishes the mission, -n(name) becomes a redundant element and,

consequently, is removed.

A different form of nunation that had been observed in certain environments in

Bedouin dialects was addressed. It was shown that in several ways in the literature, this

type of nunation has been considered the suffix by which nouns express their specificity

when they lack the definite prefix al-. Thus, given that -n(name) is a marker of specificity,

as argued in this study, the Bedouin nunation would better be regarded as its vestige

rather than the vestige of the indefinite marker -n(noun). This supports the distinction

made in this study between -n(name) and -n(noun) and the function that the former was

assumed to have.

With respect to the function of -n(noun), this study maintains that it is an indefinite

marker, as discussed in the previous chapter. This view of -n(noun) is not only motivated

by its contrast with the definite prefix al- and by how it is largely perceived, but also

motivated by its behaviour in SA that complies with the criterion that Lyons (1999:89)

sets for expressions in order to be considered indefinite markers, as stated below:

But if we do find an expression which either obligatorily marks any indefinite noun phrase, or normally does so but is absent under certain specifiable conditions, we can take it to be an indefinite article.

Given that -n(noun) normally marks indefinite nouns, but is absent under certain

morphosyntactic (duals, SndM plurals and heads of CSs), phonological (diptotes and

invariables) and semantic (non-specific vocative nouns) conditions, it certainly is an

indefinite marker.

220 Chapter Eight – Conclusion

8. Chapter Eight - Conclusion

This thesis has employed Distributed Morphology (DM) as a tool to investigate the

structure and the word formation process of Arabic nouns. DM showed a great

capability of accounting for the formation of the nouns and related phenomena. The

thesis has examined the constituting features of the nouns and the various interactions

they exhibit. Within the grammatical architecture of DM, this study has assumed several

modifications to the basic structure of Arabic nouns by which the different types of

these nouns are formed. In doing so, the study has engaged with a range of theoretical

issues in DM, including the identity of roots, the location of compositional semantic

features (CSFs), and the internal structure of the Morphology component with particular

reference to its ability to perform a feature-changing operation. It is hoped that this

theoretical engagement will provide further grounding for the theory.

In this chapter, the main findings of the thesis will be highlighted. Section 8.1

demonstrates how Arabic nouns were assumed to be formed from a DM perspective.

Section 8.2 summarises the analysis of nunation provided in this thesis. Section 8.3

highlights the contributions that this thesis is hoped to have made to the development of

DM. Finally, Section 8.4 suggests general directions for future studies.

8.1. The Formation of Arabic Nouns

The main objective of this thesis was to analyse Arabic nouns within the word

formation mechanism provided by DM. The basic syntactic structure of Arabic nouns

was assumed to contain an acategorial root morpheme that merges with the category-

assigning [n] morpheme, forming an nP. Each of the grammatical features of gender,

number and case was considered to head its own projection. The nP and the higher

221 Chapter Eight – Conclusion

projections of the grammatical features were considered to constitute the internal

structure of N that moves to D where the appropriate definiteness feature is assigned.

Since Arabic roots are not restricted to any type of CSFs, as shown in Chapter

Four, it became clear that root morphemes are CSF-free, which contrasts with what is

implied by Marantz (1995, 1996). Alternatively, it was shown that CSFs belong to the

category-assigning [n] morphemes that, along with the root morphemes under nP,

constitute the basic structure of Arabic nominal stems. However, since CSFs can come

in different values and various combinations, it was assumed that [n] in Arabic is a

functional set of members.

Given that roots have no CSFs, it became more important to investigate how roots

obtain their identity. This study supported the view that roots are equipped with indices

that identify them throughout the word formation process, especially in the Vocabulary.

Moreover, considering the nature of Arabic roots, it was suggested that root indices are

conceptual, lending further support to Pfau’s (2000) proposal.

It was assumed that across all types of Arabic nouns, root and [n] morphemes are

expressed in the Vocabulary by the insertion of the corresponding root VIs (root

consonants) and binyanim respectively. The choice of the appropriate root VI was

shown to be based on the root index, while the choice of the appropriate binyan is made

in accordance with its contextual features (Halle and Marantz, 1993; Marantz, 1995). It

was shown that stems are realised with their appropriate phonological representation

after mapping the inserted root VIs onto the inserted binyanim. These stems can be part

of different types of Arabic nouns that are determined depending on how the

grammatical features of the structure are post-syntactically modified in the Morphology.

Singular nouns were considered the most representative of the basic structure of

Arabic nouns. In most cases, structures that will eventually surface as singular pass

222 Chapter Eight – Conclusion

through the Morphology and to the Vocabulary with the configuration of their features

unchanged. The only modification that may apply to these features is impoverishment

of the [+F], blocking the insertion of the feminine VI /at/ and creating a covert feminine

noun. It was shown that while such a modification is not predictable on morphological

grounds, the root index of the targeted structure might be involved, as discussed in

Section 5.2.2.

The configuration of the morphemes constituting the structure of dual nouns may

differ from that of the basic structure of Arabic nouns depending on the variety of the

language. In DA duals, the morphemes do not appear to be subject to any

morphological modification, as shown in Section 5.3. SA duals, on the other hand,

deviate from the basic structure by having their [Dl] and [K] morphemes fused, as

shown in Section 6.2.3, and their [-Def] feature impoverished, as shown in Section

6.3.1. In the Vocabulary, these modifications lead to inserting different number and case

VIs as well as blocking the insertion of the indefinite VI /n/ (nunation).

It was assumed that Arabic plural nouns can have three structurally different

forms: broken, SndM and SndF plurals. The first post-syntactic step in the formation

process of Arabic plurals was argued to be detecting and eliminating the prohibited co-

occurrence of the features [n[-Hum]], [-F] and [Pl] when it arises before the appropriate

plural form is determined. Following the notion of morphological filters (Noyer, 1992),

this study argued that a filter, *[n[-Hum] -F Pl], that activates the morphological

modification responsible for eliminating this co-occurrence does exist in the

Morphology of Arabic, as discussed in Section 5.4.2. This modification is carried out by

changing the value of the gender feature into [+F], explaining why nonhuman plurals

are feminine in form and/or agreement. An operation by which such a modification is

performed was introduced and termed Revaluation.

223 Chapter Eight – Conclusion

The analysis showed that the second step is the activation of a Broken Formation

Rule (BFR) by which the [n], [Gend] and [Pl] morphemes are fused in the Morphology,

forming a broken plural. This view contrasts with Noyer (1992) who argues that the [Pl]

feature of broken plurals is pre-syntactically inherent to [n], as explained in Section

5.6.1. It was shown that the different forms of the BFR per se are not specific enough

since the targeted morphemes can be found in broken and non-broken plurals. Thus, it

was suggested that a selectivity criterion that is external to the targeted morphemes

might be involved, and root indices were considered the best candidates. That is, broken

plurals are probably the structures whose [n], [Gend] and [Pl] morphemes match a BFR,

and their roots (as identified by indices) are among the roots required to activate the

rule. When a BFR is activated and a broken morpheme is created, the Vocabulary

provides the morpheme with the appropriate binyan that does not only express [n], but

also expresses the gender and the plurality of the noun. This explains why broken

plurals take different binyanim and why they do not take gender or number suffixes.

In the cases where the plural structure does not activate a BFR, its final form is

determined by the value of its gender morpheme. When the gender morpheme is

specified as [+F], a Fusion rule by which the morpheme is combined with [Pl] is

activated, yielding a SndF plural structure, as shown in Section 5.5.2. However, in SA,

the existence of the fused morpheme [+F, Pl] activates an Impoverishment rule by

which the [Acc] feature is deleted, allowing the elsewhere case VI /i/ to be inserted

instead of /a/. This explains why SA SndF plurals take what is usually the

oblique/genitive suffix -i when accusative instead of the accusative suffix -a, as

discussed in Section 6.2.2.

When the gender morpheme is specified as [-F], on the other hand, the structure

may pass through the Morphology of DA unchanged, as shown in Section 5.5.1. In SA,

224 Chapter Eight – Conclusion

however, the structure undergoes a Fusion operation by which [Pl] is fused with [K],

leading to inserting different number and case VIs, as illustrated in Section 6.2.3. As

also found in SA duals, the existence of the fused morpheme [Num, K] in SA SndM

plurals activates an Impoverishment rule by which the [-Def] feature of the structure is

deleted, blocking the insertion of nunation, as discussed in Section 6.3.1. Finally,

because of the filter that does not allow masculine plural structures to surface with

nonhuman features, SndM plurals are limited to human nouns. If this filter did not exist,

we would expect SndM plurals to be open to both human and nonhuman nouns.

The morphological modifications that the structure of Arabic nouns may undergo

in the Morphology have implications to the way the affected morphemes are

phonologically represented in the Vocabulary. For example, when a SA dual noun is

nominative, the standard nominative VI /u/ loses the competition to /aːni/ because the

case morpheme of the noun has been fused with [Dl], making /aːni/ more specified for

the morpheme than /u/, as shown in Section 6.2.3. Such a different marking system is

simply a reflection of the modification that was carried out in the Morphology.

Not all deviations from the standard marking system of the language are

morphosyntactically motivated. It was shown that in some cases, the insertion process

takes place as usual, but the nouns deviate from the system as a result of post-insertion

readjustments that are phonologically motivated. Relying mostly on the analysis of

Abboud and McCarus (1983), it was shown that the behaviour of SA defective,

indeclinable and invariable nouns is phonologically motivated. It was illustrated that the

underlying structures of these nouns are not different from their regular singular and

broken plural counterparts, so they receive the same VIs at the insertion level. However,

due to the prohibited phonological sequences produced by the inserted VIs of roots, [n],

225 Chapter Eight – Conclusion

case and indefiniteness, post-insertion readjustment rules are activated for the purpose

of rectification, as discussed in Section 6.6.

In what could be the most controversial argument, this study claimed that the

unusual behaviour of SA diptotes with respect to nunation and case is phonologically

motivated as well. It was shown that the pre-insertion morphological treatment of

diptote nouns provided by Embick and Noyer (2007) neither captures the absence of

nunation in diptote names nor the behaviour of accusative SndF plurals. By showing

that the nunation of nouns is indeed an indefinite marker while the nunation of names is

functionally different, as discussed in Chapter Seven, it became clear that both types of

nunation disappear from diptote nouns and names because of their phonological form

rather than their functions. Furthermore, evidence from the phonological readjustments

that defective nouns undergo (Abboud and McCarus, 1983) points to the fact that

nunation and the standard oblique/genitive VI /i/ are inserted into diptotes. Therefore,

the modifications that these VIs undergo must be registered as part of a post-insertion

readjustment process.

8.2. Nunation

In Chapter Seven, the nature and functions of nunation were addressed as a response to

an argument made by Fassi Fehri (1993) and Lyons (1999) that nunation cannot be a

marker of indefiniteness. The bases of this argument were discussed and a refutation

was provided. It was shown that the /ni/ and /na/ parts (/nV/) of the SA dual and SndM

suffixes cannot be nunation as argued by the authors. Differences between the

behaviours of /nV/ and the nunation that appears on some SA nouns (-n(noun)) in

vocative contexts and environments that are not construct state (CS) were

demonstrated. These differences were indicative that the latter is a nunation that plays

the role of an indefinite marker, while the former is not a nunation. It was also shown 226 Chapter Eight – Conclusion

that the alternative functions that the authors assign to nunation are designed to account

for the mutual behaviour that -n(noun) and /nV/ have in CS, but fail to plausibly account

for their behaviour in some other environments as well as for the strict contrastive

distribution between -n(noun) and the definite prefix al- in non-CS environments.

This study agreed with Fassi Fehri (1993) that the nunation of names (-n(name))

cannot be a marker of indefiniteness since the behaviour of names is consistent with the

behaviour of definite nouns. However, unlike Fassi Fehri (1993), this study argued that

-n(name) is functionally different from -n(noun) and this difference should not be taken as

evidence that nunation, in all its forms, cannot express indefiniteness. Thus, it was

argued that -n(noun) must continue to be considered an indefinite marker, while -n(name)

must have a different function. Based on evidence from its behaviour in contexts where

yaː (a vocative particle) or ibn (son of) is involved, -n(name) was considered a specificity

marker. It distinguishes between the specific range of reference of names and the wider

range of reference of their words of origin. Given this function of -n(name), the slightly

different nunation that appears on Bedouin nouns in certain environments was

considered a vestige of -n(name). This contrasts with the view of Holes (1990), who takes

the Bedouin nunation as a vestige of -n(noun) that functions as an indefinite marker.

8.3. The Theory

As part of its analysis of Arabic nouns, this study addressed a number of problematic

issues that arise in the application of DM. The first issue was the association that

Marantz (1995, 1996) makes between CSFs and root VIs. As shown in Chapter Four,

Marantz’s treatment of root VIs implies that root morphemes are specified for the CSFs

of the nouns. It was shown that such a view may have resulted from focusing on

English root morphemes and their VIs. Given that the English root VIs /kæt/ and /dɔg/

are able to convey compositional semantic information, e.g., animate, nonhuman and 227 Chapter Eight – Conclusion

count, it might be a reasonable conclusion that this type of information is a property of

the root morphemes into which these VIs are inserted. By investigating the nature of

Arabic roots, however, this thesis argued that root morphemes cannot be the location of

the CSFs. Unlike the English VIs /kæt/ and /dɔg/, Arabic root VIs, e.g., /ktb/, cannot

convey this type of information before they are combined with the appropriate

binyanim. In many cases, a given root can be part of several structures with different

CSFs. Therefore, it was argued that the category-assigning [n] morphemes of Arabic

nouns are the carrier of the CSFs.

The second issue addressed was the status of root individuation. In Chapter Four,

the broad view that roots are individuated by indices (Pfau, 2000; Borer, 2009, 2013;

Harley, 2011) was supported. However, this support was not because CSFs are not

sufficient to individuate roots – even though it is a valid argument – but mainly because

roots are CSF-free. Given that Arabic roots are the source of the primitive meanings of

the words they form, as discussed in Section 4.4, the proposal that root indices are of a

conceptual nature (Pfau, 2000) was shown to provide a better explanation. Due to the

ability of root indices to communicate with the Vocabulary and the Encyclopaedia

(Harley, 2011), the view that root VIs are input to the Encyclopaedia (Marantz, 1996)

should no longer hold. Moreover, given how the morphological modifications that

convert Arabic nouns into covert feminine or broken plural are performed, it was

suggested that root indices are probably able to communicate with the Morphology as

well.

A third theoretical issue internal to DM was whether a feature-changing operation

could be performed in the Morphology. In Section 3.4, it was shown that a feature-

changing operation was not ruled out in the literature of DM and Impoverishment as an

operation with such a capability was in fact proposed by Noyer (1998). However, given

228 Chapter Eight – Conclusion

that this form of Impoverishment was proposed as an operation that deletes a feature

and inserts a less marked feature in its place, crucially it could not account for the

gender-changing phenomenon in Arabic nonhuman plurals that involves replacing the

less marked feature [-F] with [+F]. Therefore, a feature-changing operation that is

different from Noyer’s Impoverishment was proposed in this thesis. It was argued that

this operation is capable of changing the value of the targeted feature, so it was termed

Revaluation. In the case of Arabic nonhuman plurals, it was shown that Revaluation

changes the value of the gender morpheme of these nouns from [-F] into [+F],

explaining why Arabic nonhuman plurals are feminine as illustrated in Section 5.4.2. It

was also shown that the implication of Revaluation on the Vocabulary is less drastic

than the implications of the existing DM operations of Fusion and Fission. Revaluation,

like Impoverishment, does not increase or decrease the number of morphemes of the

targeted structure. Thus, whether a morpheme is revaluated or not, the Vocabulary still

provides it with a single VI. The extent to which Revaluation is an independent and new

operation should await further research.

Finally, the various morphological modifications that the structure of Arabic

nouns may undergo in the Morphology, as well as the fact that certain structures can

undergo multiple modifications in this component, motivated this thesis to attempt to

explore the internal structure of the Morphology and how it might be organised. This

thesis concluded that the Morphology is divided into three phases: (a) the phase of filter

rules followed by the agreement level, (b) the phase of special rules, and (c) the phase

of default rules, as discussed in Section 5.8.

The first phase was shown to be the place where filter-motivated modifications

take place. In Arabic nouns, revaluating the gender morpheme of nonhuman plurals is

the only modification that can be assumed to take place in this phase. It was shown why

229 Chapter Eight – Conclusion

it is important for such a modification to be performed before the agreement level. If the

agreement level precedes this phase, Revaluation would have to operate on the gender

feature of the nouns as well as the agreement node of the agreeing expressions, allowing

nonhuman plurals to agree as feminine.

The second phase was assumed to be the place where the modifications that are

motivated by the special rules of the language are performed. In Arabic nouns, it was

assumed to be the phase where certain structures, probably identified by their root

indices, are made covert feminine or broken plural.

The last phase was assumed to be where the default rules of the language are

performed when the requirements of their activation are met. This was assumed to be

the phase where the constituent morphemes of Arabic nouns can be modified in several

ways, producing the structures of duals, SndF and SndM plurals. For structures that are

subject to multiple default rules in this phase, it was assumed that rules targeting the

morphemes that were added earlier to the structure take precedence over the rules which

target the morphemes that were added later.

8.4. Suggestions for Future Studies

Due to time and space limitation, this thesis refrained from steering away from nouns,

although it would be important to investigate the structure of other parts of speech in

Arabic such as verbs and adjectives. Investigating the other aspects of Arabic in the

light of DM is highly desirable for future studies to test the claims made in this thesis.

This thesis argued that /ni/ and /na/ in SA duals and SndM plurals cannot be

nunation. Nevertheless, the behaviour that these forms share with nunation only in CSs

is worth investigating. Future studies may need to explore why /ni/ and /na/ are dropped

from head nouns in CSs. Are they Poss markers, as suggested by Fassi Fehri (1993)? If

230 Chapter Eight – Conclusion

the answer is yes, given that these forms are not nunation (as argued in this thesis), why

are the other nouns not overtly marked for Poss?

Within SA nouns, further research would be required to explore how the

behaviour of diptotes with respect to case and nunation is phonologically motivated.

This thesis provided evidence that this behaviour is likely to be a result of special post-

insertion readjustments, but it fell short of providing the exact motivation and

mechanism of these readjustments. An in-depth study focusing on this issue would

certainly add to our knowledge of the nature of this type of structure.

It is indeed important for future studies of other languages within DM to shed

more light on Revaluation as a feature-changing operation, and whether the internal

structure of the Morphology assumed in this thesis is cross-linguistically common or a

special structure of the Morphology of Arabic. It is also important for those studies to

focus on structures that may undergo multiple morphological modifications in order to

verify the assumption that the sequence of these modifications is consistent with the

sequence according to which the targeted morphemes are added to their structures.

231 References

9. References

Abboud, P., and McCarus, E. (Eds.). (1983). Elementary Modern Standard Arabic (Vol. 2). Cambridge: Cambridge University Press.

Abd-El-Jawad, H., and Abu-Salim, I. (1987). Slips of the Tongue in Arabic and Their Theoretical Implications. Language Sciences, 9(2), 145-171.

Abney, S. (1987). The English Noun Phrase in Its Sentential Aspect. Doctoral Dissertation, MIT.

Abu-Chacra, F. (2007). Arabic: An Essential Grammar. New York: Routledge.

Abu-Salim, I. (1988). Consonant Assimilation in Arabic: An Autosegmental Perspective. Lingua, 74(1), 45-66.

Acquaviva, P. (2008). Lexical Plurals: A Morphosemantic Approach. Oxford: Oxford University Press.

Acquaviva, P. (2009). Root and Lexicality in Distributed Morphology. York Papers in Linguistics, 10, 1-21.

Al-Afghani, S. (1971). Al-Muwjaz fi Qawaid Al-Lughati Al-Arabiyati [The Concise Grammar of Arabic Language]. Damascus: Al-Fikr Publishing House.

Al-Aghbari, K. (2012). Noun Plurality in Jebbali. Doctoral Dissertation, University of Florida.

Al-Balushi, R. (2011). Case in Standard Arabic: The Untraveled Paths. Doctoral Dissertation, University of Toronto.

Al-Dahdah, A. (2007). Sharhu Alfiyati Ibnu Malik [An Explanation of Ibn Malik's Alfyat]. Riyadh: Obekan Library.

Al-Fozan, A. (1999). Daleelu Assaliki ila Alfiyati Ibnu Malik [A Guide for the Pursuer of Ibn Malik's Alfyat] (Vol. 2). Riyadh: Al-Muslim Distribution and Publishing House.

Al-Horais, N. (2006). Arabic Verbless Sentences: Is There a Null VP? Pragmalinguistica, 14, 101-116.

Al-Khatib, S. (2009). On the Structure of Possessives in Palestinian Arabic. Working Papers of the Linguistics Circle, 19(1), 1-12.

Al-Shorafat, M. (2012). Subject-Verb Agreement in Standard Arabic: A Minimalist Approach. SKASE Journal of Theoretical Linguistics, 9(1), 33-49.

Albanna, M. (1969). Amaali Al-Suhaili [Dictations of Al-Suhaili]. Cairo: Al-Sa’adah Press.

232 References

Alfozan, A. (1989). Assimilation in Classical Arabic. Doctoral Dissertation, University of Glasgow.

Alhawary, M. (2009). Arabic Second Language Acquisition of Morphosyntax. New Haven, CT: Yale University Press.

Alhawary, M. (2011). Modern Standard Arabic Grammar: A Learner's Guide. Oxford: Wiley-Blackwell.

Amberber, M. (2008). Differential Case Marking of Arguments in Amharic. In A. Malchukov and A. Spencer (Eds.), The Oxford Handbook of Case (pp. 742-755). Oxford: Oxford University Press.

Anderson, S. (1982). Where’s Morphology? Linguistic Inquiry, 13(4), 571-612.

Anderson, S. (1992). A-Morphous Morphology. New York: Cambridge University Press.

Aoun, J., Benmamoun, E., and Sportiche, D. (1994). Agreement, Word Order, and Conjunction in Some Varieties of Arabic. Linguistic Inquiry, 25(2), 195-220.

Arad, M. (2003). Locality Constraints on the Interpretation of Roots: The Case of Hebrew Denominal Verbs. Natural Language and Linguistic Theory, 21(4), 737-778.

Arad, M. (2005). Roots and Patterns: Hebrew Morpho-Syntax. Dordrecht: Springer.

Aronoff, M. (1976). Word Formation in Generative Grammar. Cambridge, MA: MIT Press.

Aronoff, M. (1994). Morphology by Itself: Stems and Inflectional Classes. Cambridge, MA: MIT Press.

Aronoff, M. (2007). In the Beginning was the Word. Language, 83(4), 803-830.

Arregi, K., and Nevins, A. (2012). Morphotactics: Basque Auxiliaries and the Structure of Spellout. Dordrecht: Springer.

Ashaikh, H. (2009). Al-I'rab Al-Mahaliy li Al-Mufradat Al-Nahawiyah [The Local Analysis of the Grammatical Vocabulary]. Beirut: The Arab Institute for Research and Publishing.

Bagemihl, B. (1988). Alternate Phonologies and Morphologies. Doctoral Dissertation, University of British Columbia.

Bagemihl, B. (1989). The Crossing Constraint and ‘Backwards’ Languages. Natural Language and Linguistic Theory, 7(4), 481-549.

Bahloul, M. (2008). Structure and Function of the Arabic Verb. New York: Routledge.

Bardeas, S. (2008). A Minimalist Approach to the Semitic Construct State. York Papers in Linguistics, Series 2(9), 1-22.

233 References

Bat-El, O. (1994). Stem Modification and Cluster Transfer in Modern Hebrew. Natural Language and Linguistic Theory, 12(4), 571-596.

Bat-El, O. (2003). Semitic Verb Structure within a Universal Perspective. In J. Shimron (Ed.), Language Processing and Acquisition in Languages of Semitic, Root- Based, Morphology (pp. 29-59). Amsterdam: John Benjamins Publishing Company.

Bateson, M. (2003). Arabic Language Handbook. Washington, D.C.: Georgetown University Press.

Bauer, L. (2003). Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press.

Beeston, A. (1975). Some Features of Modern Standard Arabic. Journal of Semitic Studies, 20(1), 62-68.

Belletti, A. (1988). The Case of Unaccusatives. Linguistic Inquiry, 19(1), 1-34.

Belnap, R. (1991). Grammatical Agreement Variation in Cairene Arabic. Doctoral Dissertation, University of Pennsylvania.

Belnap, R. (1999). A New Perspective on the History of Arabic Variation in Marking Agreement with Plural Heads. Folia Linguistica, 33(2), 169-185.

Belnap, R., and Gee, J. (1994). Classical Arabic in Contact: The Transition to Near Categorical Agreement Patterns. In M. Eid, V. Cantarino and K. Walters (Eds.), Perspectives on Arabic Linguistics VI: Papers from the Annual Symposium on Arabic Linguistics (pp. 121-149). Amsterdam: John Benjamins Publishing Company.

Belnap, R., and Shabaneh, O. (1992). Variable Agreement and Nonhuman Plurals in Classical and Modern Standard Arabic. In E. Broselow, M. Eid and J. McCarthy (Eds.), Perspectives on Arabic Linguistics IV: Papers from the Annual Symposium on Arabic Linguistics (pp. 245-262). Amsterdam: John Benjamins Publishing Company.

Benmamoun, E. (1999). Spec-Head Agreement and Overt Case in Arabic. In D. Adger (Ed.), Specifiers: Minimalist Approaches (pp. 110-125). Oxford: Oxford University Press.

Benmamoun, E. (2000). The Feature Structure of Functional Categories: A Comparative Study of Arabic Dialects. Oxford: Oxford University Press.

Benmamoun, E. (2003). The Role of the Imperfective Template in Arabic Morphology. In J. Shimron (Ed.), Language Processing and Acquisition in Languages of Semitic, Root-Based, Morphology (pp. 99-114). Amsterdam: John Benjamins Publishing Company.

Benmamoun, E. (2008). Clause Structure and the Syntax of Verbless Sentences. In R. Freidin, C. Otero and M. Zubizarreta (Eds.), Foundational Issues in Linguistic

234 References

Theory: Essays in Honor of Jean-Roger Vergnaud (pp. 105-132). Cambridge, MA: MIT Press.

Berg, T., and Abd-El-Jawad, H. (1996). The Unfolding of Suprasegmental Representations: a Cross-Linguistic Perspective. Journal of Linguistics, 32(2), 291-324.

Bickel, B. (2011). Absolute and Statistical Universals. In P. Hogan (Ed.), The Cambridge Encyclopedia of the Language Sciences (pp. 77-79). Cambridge: Cambridge University Press.

Bittner, M., and Hale, K. (1996). The Structural Determination of Case and Agreement. Linguistic Inquiry, 27(1), 1-68.

Blake, B. (2001). Case. Cambridge: Cambridge University Press.

Bobaljik, J. (2005). Distributed Morphology. Ms., University of Connecticut.

Bonet, E. (1991). Morphology After Syntax: Pronominal Clitics in Romance. Doctoral Dissertation, MIT.

Booij, G. (1996). Inherent versus Contextual Inflection and the Split Morphology Hypothesis. In G. Booij and J. Marle (Eds.), Yearbook of Morphology 1995 (pp. 1-16). Dordrecht: Kluwer Academic Publishers.

Borer, H. (1999). Deconstructing the Construct. In K. Johnson and I. Roberts (Eds.), Beyond Principles and Parameters (pp. 43-89). Dordrech: Kluwer Academic Publishers.

Borer, H. (2005a). In Name Only: Structuring Sense (Vol. 1). Oxford: Oxford University Press.

Borer, H. (2005b). The Normal Course of Events: Structuring Sense (Vol. 2). Oxford: Oxford University Press.

Borer, H. (2009). Roots and Categories. Paper presented at the 19th Colloquium on Generative Grammar, University of the Basque Country, Vitoria-Gasteiz.

Borer, H. (2013). Taking Form: Structuring Sense (Vol. 3). Oxford: Oxford University Press.

Boudelaa, S., and Gaskell, M. (2000). In Search of the Minority Default: The Case of Arabic Plurals. In L. Gleitman and A. Joshi (Eds.), Proceedings of the Twenty- Second Annual Conference of the Cognitive Science Society (pp. 48-53). Mahwah, N.J.: Lawrence Erlbaum Associates.

Boudelaa, S., and Marslen-Wilson, W. (2004a). Abstract Morphemes and Lexical Representation: The CV-Skeleton in Arabic. Cognition, 92(3), 271-303.

Boudelaa, S., and Marslen-Wilson, W. (2004b). Allomorphic Variation in Arabic: Implications for Lexical Processing and Representations. Brain and Language, 90(1-3), 106-116.

235 References

Boudelaa, S., Pulvermüller, F., Hauk, O., Shtyrov, Y., and Marslen-Wilson, W. (2010). Arabic Morphology in the Neural Language System. Journal of Cognitive Neuroscience, 22(5), 998-1010.

Brustad, K. (2000). The Syntax of Spoken Arabic: A Comparative Study of Moroccan, Egyptian, Syrian, and Kuwaiti Dialects. Washington, D.C.: Georgetown University Press.

Calabrese, A. (1988). Towards a Theory of Phonological Alphabets. Doctoral Dissertation, MIT.

Calabrese, A. (1991). The Notion of Phonological Complexity in Phonological Theory: A New Approach to Markedness Theory. Ms., Harvard University.

Carstairs, A. (1987). Diachronic Evidence and the Affix-Clitic Distinction. In A. Ramat, O. Carruba and G. Bernini (Eds.), Papers from the 7th International Conference on Historical Linguistics (ICHL '84) (pp. 151-162). Amsterdam: John Benjamins Publishing Company.

Childs, G. (2003). An Introduction to African Languages. Amsterdam: John Benjamins Publishing Company.

Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT press.

Chomsky, N. (1970). Remarks on Nominalization. In R. Jacobs and P. Rosenbaum (Eds.), Readings in English Transformational Grammar (pp. 184-221). Washington, D.C.: Georgetown University Press.

Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris.

Chomsky, N. (1995a). Bare Phrase Structure. In H. Campos and P. Kempchinsky (Eds.), Evolution and Revolution in Linguistic Theory: Studies in Honor of Carlos P. Otero (pp. 51-109). Washington, D.C.: Georgetown University Press.

Chomsky, N. (1995b). The Minimalist Program. Cambridge, MA: MIT Press.

Choueka, Y. (1996). Rav milim: A Dictionary of Contemporary Hebrew. Tel Aviv: Center for Educational Technology.

Chung, I. (2009). Suppletive Verbal Morphology in Korean and the Mechanism of Vocabulary Insertion. Journal of Linguistics, 45(3), 533-567.

Corbett, G. (1983). Hierarchies,Targets and Controllers: Agreement Patterns in Slavic. London: Croom Helm.

Corbett, G. (1991). Gender. Cambridge: Cambridge University Press.

Corriente, F. (1977). A Grammatical Sketch of the Spanish Arabic Dialect Bundle. Madrid: Instituto híspano-árabe de cultura.

236 References

Crass, J. (2006). Afroasiatic Languages. In E. Brown, R. Asher and J. Simpson (Eds.), Encyclopedia of Language and Linguistics (Vol. 1, pp. 106-109). Oxford: Elsevier.

Dammel, A., and Nübling, D. (2006). The Superstable Marker as an Indicator of Categorial Weakness? Folia Linguistica, 40(1-2), 97-113.

de Alcala, P. (1505). Arte parta ligera mente saber la lengua araviga, emendada y añada y segunda mente imprimida. Granatae: Juan Verala de Salamanca. [cited in Ussishkin, 2006:37].

de Hoop, H., and Malchukov, A. (2007). On Fluid Differential Case Marking: A Bidirectional OT Approach. Lingua, 117(9), 1636-1656.

de Hoop, H., and Narasimhan, B. (2005). Differential Case-Marking in Hindi. In M. Amberber and H. de Hoop (Eds.), Competition and Variation in Natural Languages: The Case for Case (pp. 321-345). Amsterdam: Elsevier.

de Sacy, S. (1810). Grammaire Arabic à l’usage des élèves de l’Ecole spéciale des languages orientales vivantes. Paris: Imprimerie Royale. [cited in Ussishkin, 2006:37].

de Volney, C. (1787). Voyages en Syrie et en Egypte pendant les années 1783, 1784, 1785. Paris: Volland, Desenne. [cited in Ussishkin, 2006:37].

Di Sciullo, A.-M., and Williams, E. (1987). On the Definition of Word. Cambridge, MA: MIT Press.

Diakonoff, I. (1998). The Earliest Semitic Society: Linguistic Data. Journal of Semitic Studies, 43(2), 209-219.

Dubinsky, S., and Simango, S. (1996). Passive and Stative in Chichewa: Evidence for Modular Distinctions in Grammar. Language, 72(4), 749-781.

Duraidi, A. (2000). Al-A'laamu fi Al-Qurani Al-kareemi [Proper Names in the Holy Quran]. Master Dissertation, Al-Najah National University, Nablus.

Eid, M. (1991). Verbless Sentences in Arabic and Hebrew. In B. Comrie and M. Eid (Eds.), Perspective on Arabic Linguistics III: Papers from the Annual Symposium on Arabic Linguistics (pp. 31-61). Amsterdam: John Benjamins Publishing Company.

Eimas, P., Siqueland, E., Jusczyk, P., and Vigorito, J. (1971). Speech Perception in Infants. Science, 171(3968), 303-306.

Eisele, J. (2002). The Linguistic Representation of Arabic Morphology. Al-Arabiyya: Journal of the American Association of Teachers of Arabic, 35, 1-59.

Embick, D. (1997). Voice and the Interfaces of Syntax. Doctoral Dissertation, University of Pennsylvania.

237 References

Embick, D., and Marantz, A. (2008). Architecture and Blocking. Linguistic Inquiry, 39(1), 1-53.

Embick, D., and Noyer, R. (2001). Movement Operations after Syntax. Linguistic Inquiry, 32(4), 555-595.

Embick, D., and Noyer, R. (2007). Distributed Morphology and the Syntax/Morphology Interface. In G. Ramchand and C. Reiss (Eds.), The Oxford Handbook of Linguistic Interfaces (pp. 289-324). Oxford: Oxford University Press.

Ewald, G. (1827). Kritische Grammatik derhebraïsche Sprache. Leipzig: Hahn. [cited in Ussishkin, 2006:37].

Fassi Fehri, A. (1989). Generalized IP Structure, Case and VS Word Order. MIT Working Papers in Linguistics, 10, 75-113.

Fassi Fehri, A. (1993). Issues in the Structure of Arabic Clauses and Words. Dordrecht: Kluwer Academic Publishers.

Fassi Fehri, A. (1999). Arabic Modifying Adjectives and DP Structures. Studia Linguistica, 53(2), 105-154.

Fassi Fehri, A. (2012). Key Features and Parameters in Arabic Grammar. Amsterdam: John Benjamins Publishing Company.

Faust, N. (2007). In a Position to Spell Out: The Syntactic Structure of Vocalic Patterns in Hebrew. Ms., University of Paris VII.

Fayadh, S. (1995). Al-Nahu Al-Asriy: Daleelun Mubasatun li Qawaaidi Al-Lughati Al- Arabiyati [The Modern Grammar: A Simplified Guide to the Grammar of Arabic]. Cairo: Al-Ahram Center of Translation and Publishing.

Feldman, L., Frost, R., and Pnini, T. (1995). Decomposing Words into Their Constituent Morphemes: Evidence from English and Hebrew. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(4), 947-960.

Ferguson, C. (1959). Diglossia. Word, 15(2), 325-340.

Ferguson, C. (1989). Grammatical Agreement in Classical Arabic and the Modern Dialects: A Response to Versteegh's Pidginization Hypothesis. Al-Arabiyya: Journal of the American Association of Teachers of Arabic, 22, 5-17.

Fromkin, V. (Ed.). (1973). Speech Errors as Linguistic Evidence. The Hague: Mouton.

Frost, R., Forster, K., and Deutsch, A. (1997). What can We Learn from the Morphology of Hebrew? A Masked-Priming Investigation of Morphological Representation. Journal of Experimental Psychology, 23(4), 829-856.

Ghalayini, M. (1993). Jamiu Addurusi Alarabiyati [A Collection of Arabic Lessons]. Beirut: The Modern Library.

Greenberg, J. (1966). The Languages of Africa. Bloomington: Indiana University.

238 References

Halle, M. (1997). Distributed Morphology: Impoverishment and Fission. MIT Working Papers in Linguistics, 30, 425-449.

Halle, M., and Marantz, A. (1993). Distributed Morphology and the Pieces of Inflection. In K. Hale and S. Keyser (Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger (pp. 111-176). Cambridge, MA: MIT Press.

Halle, M., and Marantz, A. (1994). Some Key Features of Distributed Morphology. MIT Working Papers in Linguistics, 21, 275-288.

Harbert, W., and Bahloul, M. (2002). Postverbal Subjects in Arabic and the Theory of Agreement. In J. Ouhalla and U. Shlonsky (Eds.), Themes in Arabic and Hebrew Syntax (pp. 45-70). Dordrecht: Kluwer Academic Publishers.

Harley, H. (2005). One-Replacement, Unaccusativity, Acategorical Roots and Bare Phrase Structure. Harvard Working Papers in Linguistics, 11, 59-78.

Harley, H. (2011). On the Identity of Roots. Ms., University of Arizona.

Harley, H. (2012). Semantics in Distributed Morphology. In C. Maienborn, K. von Heusinger and P. Portner (Eds.), Semantics: An International Handbook of Natural Language Meaning (Vol. 3, pp. 2151-2171). Berlin: Walter de Gruyter.

Harley, H., and Noyer, R. (1998). Licensing in the Non-Lexicalist Lexicon: Nominalizations, Vocabulary Items and the Encyclopedia. MIT Working Papers in Linguistics, 32, 119-137.

Harley, H., and Noyer, R. (1999). Distributed Morphology. Glot International, 4(4), 3- 9.

Harley, T. (1984). A Critique of Top-Down Independent Levels Models of Speech Production: Evidence from Non-Plan-Internal Speech Errors. Cognitive Science, 8(3), 191-219.

Haroun, A. (2001). Alasaleeb Alinshaiyah fi Alnahw Alarabiy [The Compositional Styles in Arabic Grammar]. Cairo: Al-Khanji Library.

Harris, J. (1991). The Exponence of Gender in Spanich. Linguistic Inquiry, 22(1), 27- 62.

Harris, Z. (1941). Linguistic Structure of Hebrew. Journal of the American Oriental Society, 61(3), 143-167.

Hasan, I. (1978). Asraru Al-Nidai fi Lughati Al-Qurani Al-Kareemi [The Secrets of Vocation in the Language of the Quran]. Cairo: Al-Fijalah Print

Hassan, Z. (2002). Gemination in Swedish and Arabic with a Particular Reference to the Preceding Vowel Duration: An Instrumental and Comparative Approach. Fonetik 2002, 44(1), 81-85.

239 References

Hayward, R. (2000). Afroasiatic. In B. Heine and D. Nurse (Eds.), African Languages: An Introduction (pp. 74-98). Cambridge: Cambridge University Press.

Haywood, J., and Nahmad, M. (1965). A New Arabic Grammar of the Written Language. Cambridge, MA: Harvard University Press.

Heath, J. (1987). Ablaut and Ambiguity: Phonology of a Moroccan Arabic Dialect. Albany, N.Y.: State University of New York Press.

Heath, J. (1997). Moroccan Arabic Phonology. In A. Kaye (Ed.), Phonologies of Asia and Africa (pp. 205-217). Winona Lake, IN: Eisenbrauns.

Heath, J. (2005). A Grammar of Tamashek (Tuareg of Mali). Berlin: Walter de Gruyter.

Herin, B., and Al-Wer, E. (2013). From Phonological Variation to Grammatical Change: Depalatalisation of /č/ in Salti. In C. Holes and R. Jong (Eds.), Ingham of Arabia: A Collection of Articles Presented as a Tribute to the Career of Bruce Ingham (pp. 55-74). Leiden: Koninklijk Brill NV.

Hetzron, R. (1987). Afroasiatic Languages. In B. Comrie (Ed.), The World’s Major Languages (pp. 645-653). London: Croom Helm.

Hetzron, R., and Kaye, A. (1987). Semitic Languages. In B. Comrie (Ed.), The World’s Major Languages (pp. 654-663). London: Croom Helm.

Heusinger, K. (2002). Specificity and Definiteness in Sentence and Discourse Structure. Journal of Semantics, 19(3), 245-274.

Holes, C. (1990). Gulf Arabic. London: Routledge.

Holes, C. (2004). Modern Arabic: Structures, Functions, and Varieties. Washington, D.C.: Georgetown University Press.

Holes, C. (2010). Colloquial Arabic of the Gulf: The Complete Course for Beginners. Abingdon: Routledge.

Homeidi, M. (1994). Government and Binding and Case Assignment in Modern Standard Arabic. Papers and Studies in Contrastive Linguistics, 29, 123-140.

Ingham, B. (1994). Najdi Arabic: Central Arabian. Amsterdam: John Benjamins Publishing Company.

Jackendoff, R. (1977). X-Bar Syntax: A Study of Phrase Structure. Cambridge, MA: MIT Press.

Jackendoff, R. (1994). Lexical Insertion in a Post-Minimalist Theory of Grammar. Ms., Brandeis University.

Jackendoff, R. (1995). The Boundaries of the Lexicon. In M. Everaert, E.-J. van der Linden, A. Schenk and R. Schreuder (Eds.), Idioms: Structural and Psychological Perspectives (pp. 133-165). Mahwah, N.J.: Lawrence Erlbaum Associates.

240 References

Joseph, B. (1997). How General are our Generalizations? What Speakers Actually Know and What They Actually Do. In A. Green and V. Motopanyane (Eds.), Proceedings of the Thirteenth Eastern States Conference on Linguistics (ESCOL '96) (pp. 148-160). Ithaca: Cascadilla Press.

Julien, M. (2005). Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins Publishing Company.

Kaye, A. (1970). Modern Standard Arabic and the Colloquials. Lingua, 24, 374-391.

Kaye, A. (1987). Arabic. In B. Comrie (Ed.), The World’s Major Languages (pp. 664- 685). London: Croom Helm.

Kaye, A. (2001). Diglossia: The State of the Art. International Journal of the Sociology of Language, 152, 117-129.

Khouloughli, D. E. (1992). Basic Lexicon of Modern Standard Arabic. Paris: L' Harmattan.

Kihm, A. (2005). Noun Classes, Gender, and the Lexicon-Syntax-Morphology Interfaces: A Comparative Study of Niger-Congo and Romance Languages. In G. Cinque and R. Kayne (Eds.), The Oxford Handbook of Comparative Syntax (pp. 459-512). Oxford: Oxford University press.

Kihm, A. (2011). Plural Formation in Nubi and Arabic: A Comparative Study and a Word-Based Approach. Brill’s Annual of Afroasiatic Languages and Linguistics, 3, 1-21.

Kiparsky, P. (1973). "Elsewhere" in Phonology. In S. Anderson and P. Kiparsky (Eds.), A Festschrift for Morris Halle (pp. 93-106). New York: Holt, Rinehart and Winston.

Kiparsky, P. (1982). Lexical Phonology and Morphology. In I.-S. Yang (Ed.), Linguistics in the Morning Calm: Selected Papers from SICOL-1981 (pp. 3-91). Seoul: Hanshin Publishing Company.

Levin, B., and Rappaport, M. (1986). The Formation of Adjectival Passives. Linguistic Inquiry, 17(4), 623-661.

Levin, I., Ravid, D., and Rapaport, S. (2001). Morphology and Spelling Among Hebrew-Speaking Children: From Kindergarten to First Grade. Journal of Child Language, 28(3), 741-772.

Lieber, R. (1980). On the Organization of the Lexicon. Doctoral Dissertation, MIT.

Lowenstamm, J. (2008). On Little n, √, and Types of Nouns. In J. Hartmann, V. Hegedüs and H. van Rimesdijk (Eds.), Sounds of Silence: Empty Elements in Syntax and Phonology (pp. 105-144). Amsterdam: Elsevier.

Lyons, C. (1999). Definiteness. Cambridge: Cambridge University Press.

241 References

Malchukov, A. (2008). Animacy and Asymmetries in Differential Case Marking. Lingua, 118(2), 203-221.

Malchukov, A., and de Swart, P. (2008). Differential Case Marking and Actancy Variations. In A. Malchukov and A. Spencer (Eds.), The Oxford Handbook of Case (pp. 339-356). Oxford: Oxford University Press.

Marantz, A. (1984). On the Nature of Grammatical Relations. Cambridge, MA: MIT Press.

Marantz, A. (1988). Clitics, Morphological Merger, and the Mapping to Phonological Structure. In M. Hammond and M. Noonan (Eds.), Theoretical Morphology: Approaches in Modern Linguistics (pp. 253-270). San Diego, CA: Academic Press.

Marantz, A. (1995). A Late Note on Late Insertion. In Y.-S. Kim, B.-C. Lee, K.-J. Lee, H.-K. Yang and J.-Y. Yoon (Eds.), Explorations in Generative Grammar: A Festschrift for Dong-Whee Yang (pp. 396-413). Seoul: Hanakuk Publishing Company.

Marantz, A. (1996). “Cat” as a Phrasal Idiom: Consequences of Late Insertion in Distributed Morphology. Ms., MIT.

Marantz, A. (1997). No Escape from Syntax: Don’t Try Morphological Analysis in the Privacy of Your Own Lexicon. Penn Working Papers in Linguistics, 4(2), 201- 225.

Marantz, A. (2000). Case and Licensing. In E. Reuland (Ed.), Arguments and Case: Explaining Burzio's Generalization (pp. 11-30). Amsterdam: John Benjamins Publishing Company.

Marantz, A. (2001). Words. Ms., MIT.

Mayer, M. (1969). Frog, Where are You? New York: Dial Press.

McCarthy, J. (1979). Formal Problems in Semitic Phonology and Morphology. Doctoral Dissertation, MIT.

McCarthy, J. (1981). A Prosodic Theory of Nonconcatenative Morphology. Linguistic Inquiry, 12(3), 373-418.

McCarthy, J., and Prince, A. (1986). Prosodic Morphology. Ms., University of Massachusetts.

McCarthy, J., and Prince, A. (1990). Foot and Word in Prosodic Morphology: The Arabic Broken Plural. Natural Language and Linguistic Theory, 8(2), 209-283.

McFadden, T. (2004). The Position of Morphological Case in the Derivation: A Study on the Syntax-Morphology Interface. Doctoral Dissertation, University of Pennsylvania.

242 References

Mohammad, M. (1988). On the Parallelism Between IP and DP. In H. Borer (Ed.), Proceedings of the Seventh West Coast Conference on Formal Linguistics (pp. 241-254). Stanford, CA: CSLI Publications.

Mohammad, M. (1990). The Problem of Subject-Verb Agreement in Arabic: Towards a Solution. In M. Eid (Ed.), Perspectives in Arabic Linguistics I: Papers from the Annual Symposium on Arabic Linguistics (pp. 95-126). Amsterdam: John Benjamins Publishing Company.

Moussa, A. (1996). Database for Major Arabic Dictionaries. Proceedings of the 5th International Conference and Exhibition on Multilingual Computing, Cambridge.

Müller, G. (2004). Syncretism and Iconicity in Icelandic Noun Declensions: A Distributed Morphology Approach. In G. Booij and J. Marle (Eds.), Yearbook of Morphology (pp. 229-271). Dordrecht: Springer.

Neme, A., and Laporte, E. (2013). Pattern-and-Root Inflectional Morphology: The Arabic Broken Plural. Language Sciences, 40, 221-250.

Newman, P. (1990). Nominal and Verbal Plurality in Chadic. Dordrecht: Foris Publications.

Nichols, J. (1992). Linguistic Diversity in Space and Time. Chicago, IL: University of Chicago Press.

Noyer, R. (1992). Features, Positions and Affixes in Autonomous Morphological Structure. Doctoral Dissertation, MIT.

Noyer, R. (1998). Impoverishment Theory and Morphosyntactic Markedness. In S. Lapointe, D. Brentari and P. Farrell (Eds.), Morphology and Its Relation to Phonology and Syntax (pp. 264-285). Stanford, CA: CSLI Publications.

Ouhalla, J. (1994). Verb Movement and Word Order in Arabic. In D. Lightfoot and N. Hornstein (Eds.), Verb Movement, (pp. 41-72). Cambridge: Cambridge University Press.

Ouhalla, J. (2005). Agreement Features, Agreement and Antiagreement. Natural Language and Linguistic Theory, 23(3), 655-686.

Pesetsky, D. (1995). Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press.

Pfau, R. (2000). Features and Categories in Language Production. Doctoral Dissertation, Johann Wolfgang Goethe University.

Pfau, R. (2002). Local Licensing and Feature Copy in Language Production. In M. van Koppen, E. Thrift, E. J. van der Torre and M. Zimmermann (Eds.), Proceedings of ConSOLE IX (pp. 173-187). Leiden: Student Organisation of Linguistics in Europe.

243 References

Pfau, R. (2007). Cheap Repairs: A Distributed Morphology Toolkit for Sentence Construction. MIT Working Papers in Linguistics, 53, 9-33.

Picallo, M. (1991). Nominals and Nominalizations in Catalan. Probus, 3(3), 279-316.

Picallo, M. (2008). Gender and Number in Romance. Lingue e Linguaggio, 7(1), 47-66.

Prunet, J.-F. (2006). External Evidence and the Semitic Root. Morphology, 16(1), 41- 67.

Prunet, J.-F., Béland, R., and Idrissi, A. (2000). The Mental Representation of Semitic Words. Linguistic Inquiry, 31(4), 609-648.

Qafisheh, H. (1977). A Short Reference Grammar of Gulf Arabic. Tucson, AZ: University of Arizona Press.

Ratcliffe, R. (1997). Prosodic Templates in a Word-Based Morphological Analysis of Arabic. In M. Eid and R. Ratcliffe (Eds.), Perspectives on Arabic Linguistics X: Papers from the Annual Symposium on Arabic Linguistics (pp. 147-172). Amsterdam: John Benjamins Publishing Company.

Ratcliffe, R. (1998). Defining Morphological Isoglosses: The 'Broken' Plural and Semitic Subclassification. Journal of Near Eastern Studies, 57(2), 81-123.

Ravid, D., and Schiff, R. (2006). Roots and Patterns in Hebrew Language Development: Evidence from Written Morphological Analogies. Reading and Writing, 19(8), 789-818.

Ritter, E. (1988). A Head-Movement Approach to Construct-State Noun Phrases. Linguistics, 26(6), 909-929.

Ritter, E. (1991). Two Functional Categories in Noun Phrases: Evidence from Modern Hebrew. Syntax and Semantics, 25, 37-62.

Ritter, E. (1993). Where's Gender? Linguistic Inquiry, 24(4), 795-803.

Ryding, K. (1991). Proficiency Despite Diglossia: A New Approach for Arabic. The Modern Language Journal, 75(2), 212-218.

Ryding, K. (2005). A Reference Grammar of Modern Standard Arabic. Cambridge: Cambridge University Press.

Saiegh-Haddad, E. (2003). Linguistic Distance and Initial Reading Acquisition: The Case of Arabic Diglossia. Applied Psycholinguistics, 24(3), 431-451.

Sauerland, U. (1995). The Late Insertion of Germanic Inflection. Ms., MIT.

Selkirk, E. (1982). The Syntax of Words. Cambridge, MA: MIT Press.

Shlonsky, U. (2004). The Form of Semitic Noun Phrases. Lingua, 114(12), 1465-1526.

244 References

Siddiqi, D. (2006). Minimize Exponence: Economy Effects on a Model of the Morphosyntactic Component of the Grammar. Doctoral Dissertation, University of Arizona.

Siddiqi, D. (2009). Syntax within the Word: Economy, Allomorphy, and Argument Selection in Distributed Morphology. Amsterdam: John Benjamins Publishing Company.

Siddiqi, D. (2010). Distributed Morphology. Language and Linguistics Compass, 4(7), 524-542.

Siegel, D. (1974). Topics in English Morphology. Doctoral Dissertation, MIT.

Sigurdsson, H. (2001). Case: Abstract vs. Morphological. Working Papers in Scandinavian Syntax, 67, 103-151.

Siloni, T. (1991). Noun Raising and the Structure of Noun Phrases. MIT Working Papers in Linguistics, 14, 255-270.

Siloni, T. (2001). Construct States at the PF Interface. Linguistic Variation Yearbook, 1, 229-266.

Soltan, U. (2007). On Formal Feature Licensing in Minimalism: Aspects of Standard Arabic Morphosyntax. Doctoral Dissertation, University of Maryland

Streeter, L. (1976). Language Perception of Two-Month Old Infants Shows Effects of Both Innate Mechanisms and Experience. Nature, 259(5538), 39-41.

Ussishkin, A. (2006). Semitic Morphology: Root-Based or Word-Based? Morphology, 16(1), 37-40.

Velan, H., Frost, R., Deutsch, A., and Plaut, D. (2005). The Processing of Root Morphemes in Hebrew: Contrasting Localist and Distributed Accounts. Language and Cognitive Processes, 20(1-2), 169-206.

Versteegh, K. (1997). The Arabic Language. Edinburgh: Edinburgh University Press.

Vicente, A. (2007). Personal Pronouns (Arabic Dialects). In K. Versteegh and M. Eid (Eds.), Encyclopedia of Arabic Language and Linguistics (Vol. 3, pp. 584-588). Leiden: Koninklijk Brill NV.

Vinka, M. (2001). Impoverishment as Feature Deletion: Dual and Plural Agreement in Sámi. Lund Working Papers in Linguistics, 48, 183-191.

Vollers, K. (2011). The Modern Egyptian Dialect of Arabic: A Grammar with Exercises, Reading Lessons and Glossaries. New York: Cambridge University Press.

Wasow, T. (1977). Transformations and the Lexicon. In P. Culicover, T. Wasow and A. Akmajian (Eds.), Formal Syntax (pp. 327-360). New York: Academic Press.

245 References

Watson, J. (2002). The Phonology and Morphology of Arabic. Oxford: Oxford University Press.

Watson, J. (2006). Arabic as an Introflecting Language. In K. Brown (Ed.), Encyclopedia of Language and Linguistics (pp. 431-434). Oxford: Elsevier.

Wehr, H. (1971). Dictionary of Modern Written Arabic. J. M. Cowan (Ed.). Ithaca, N.Y.: Spoken Language Services.

Werker, J. (1989). Becoming a Native Listener. American Scientist, 77, 54-59.

Williams, E. (1981). On the Notions "Lexically Related" and "Head of a Word". Linguistic Inquiry, 12(2), 245-274.

Wright, W. (1967). A Grammar of the Arabic Language. Cambridge: Cambridge University Press.

Yaquub, E. (1993). Moujam Al-Awzaan Al-Sarfiyah [A Dictionary of the Morpholgical Patterns]. Beirut: Aalam Al-Kutub.

Yip, M. (1988). Template Morphology and the Direction of Association. Natural Language and Linguistic Theory, 6(4), 551-577.

Youssef, I. (2013). Place Assimilation in Arabic: Contrasts, Features, and Constraints. Doctoral Dissertation, University of Tromsø.

Yushmanov, N. (1961). The Structure of the Arabic Language (M. Perlman, Trans.). Washington, D.C.: Center for Applied Linguistics of the Modern Language Association of America. (Original work published 1938).

Ziadeh, F., and Winder, R. (1955). An Introduction to Modern Arabic. Princeton, N.J.: Princeton University Press.

Zuraiq, W., and Zhang, J. (2006). Phonological Assimilation in Urban Jordanian Arabic. Kansas Working Papers in Linguistics, 28, 33-64.

246 Appendices

10. Appendices

Appendix 1: Ethics Approval Letter

Human Research Ethics Advisory Panel B Arts, Humanities & Law

Date: 07.06.2011

Investigators: Mr Abdullah Alghamdi

Supervisors: Dr Mengistu Amberber and Dr Hugues Peters

School: School of Languages and Linguistics

Re: The Structure of Arabic Words

Reference Number: 11 055

The Human Research Ethics Advisory Panel B for the Arts, Humanities & Law is satisfied that this project is of minimal ethical impact and meets the requirements as set out in the National Statement on Ethical Conduct in Human Research*. Having taken into account the advice of the Panel, the Deputy Vice-Chancellor (Research) has approved the project to proceed.

Your Head of School/Unit/Centre will be informed of this decision.

This approval is valid for 12 months from the date stated above.

Yours sincerely

Associate Professor Annie Cossins Panel Convenor HREA Panel B for the Arts, Humanities & Law

Cc: Associate Professor Ludmila Stern Head of School School of Languages and Linguistics

* http://www.nhmrc.gov.au/

247 Appendices

Appendix 2: Samples of the Collected Data

------Standard Arabic: yuħaːwilu al-kalb-u al-wusʕuːl-a ʔila xaliy-at-i an-naħl-i he.try Def-dog-Nom Def-reaching-Acc to cell-F-Obl Def-bees.Gen

al-mutadaly-at-i mina aʃ-ʃaʒar-at-i, Def-dangling-F-Gen from Def-tree-F-Obl ‘The dog is trying to reach the beehive dangling off the tree.’

wa atʕ-tʕifl-u yabħaӨu ʕani adʕ-dʕifdaʕ-i fiː ʔiħda al-ʒuħuːr-i and Def-kid-Nom he.search about Def-frog-Obl in one Def-holes-Gen

ʔallati waʒada-ha fiː al-χaːb-at-i that found.he-it in Def-forest-F-Obl ‘And the kid is searching for the frog in one of the holes he found in the forest.’

yaxruʒu mina al-ʒuħr-i faʔr-u-n yufziʕu atʕ-tʕifl-a it.exit from Def-hole-Obl rat-Nom-Idef it.scare Def-kid-Acc ‘A rat is coming out of the hole scaring the kid.’ ------Egyptian: al-kalb raːħ yitʕaːrid an-naħl, wa an-naħl ma ʔarasʕuː-ʃ. tʕab kwayis. Def-dog went it.chase Def-bees, and Def-bees didn’t sting-Neg. OK good ‘The dog went chasing the bees, and the bees didn’t sting him. That’s good.’ al-walad ʕammaːl yidawar. liʔi xurm, basʕ fiː-h. Def-boy be he.search. found.he hole, looked.he in-it. ‘The boy is searching. He found a hole, and looked into it.’

248 Appendices

tʕiliʕ-l-uh faːr wa xaːf minn-uh, wa kida came.out-for-him rat and scared.he from-it, and like.this ‘A rat came out and he (the boy) became scared of it, and something like that.’ ------Jordanian: al-walad bi-yutʕul fiː ħufr-ah zay ħufr-at al-ʔaraːnib ʔaw ħufr-at al-xulund Def-boy is-he.look in hole-F like hole-F Def-rabbits or hole-F Def-mole ‘The boy is looking into a hole, like a rabbits’ hole or a hole of a mole.’

wa al-kalb qaːʕid yelʕab maʕa an-naħl-aːt and Def-dog be it.play with Def-bees-F.Pl ‘And the dog is playing with the bees.’

tʕiliʕ min ħufrat al-xulund ħayawaːn Өaːniː came.out from hole Def-mole animal another ‘Another animal came out of the mole hole.’

ʃikluh χaðʕbaːn ʕala al-walad. al-walad ʔitfaːʒa wa ʃikluh mitʔasif seem.it angry on Def-boy. Def-boy surprised and seem.he sorry ‘It seems to be angry at the boy. The boy was surprised and he seems to be sorry.’ ------

Saudi: kaːn atʕ-tʕifl yibħaӨ ʕan aðʕ-ðʕifdaʕ fiː ʒuħr was Def-kid he.search about Def-frog in hole ‘The kid was searching for the frog in a hole.’

al-kalb kaːn yilʕab bi-xaliy-at an-naħl, wa an-naħl kaːn yitʕiːr ħawaːlay-h Def-dog was it.play by-cell-F Def-bees, and Def-bees were it.fly around-him ‘The dog was playing with the beehive, and the bees were flying around him.’

xaraʒ li-atʕ-tʕifl faːr, wa xaːf atʕ-tʕifl min haːða al-faːr exited.it for-Def-kid rat, and scared.he Def-kid from this rat ‘A rat came out for the kid, and the kid became scared of this rat.’ ------

249