Université Mohammed V

Faculté des Lettres et des Sciences Humaines

Department of English
Rabat, Morocco
*****************

Constraint Interaction in the Phonology and Morphology

of Casablanca Moroccan Arabic

A dissertation submitted in fulfillment of the requirement for the degree of

“Doctorat d’Etat”

By: Abdelaziz BOUDLAL

Jointly Supervised by: Professor Abderrafi BENHALLAM

Université Mohammed V, Rabat

And

Professor Elisabeth SELKIRK

University of Massachusetts, Amherst

20 March 2001
 Defense Committee

Professor Abderrafi BENHALLAM

Université Mohammed V, Rabat

Professor Bouchaib IDRISSI BOUYAHYAOUI

Université Mohammed V, Rabat

Professor John J. McCARTHY

University of Massachusetts, Amherst

Professor Elisabeth O. SELKIRK

University of Massachusetts, Amherst

Professor Mohamed TAKI

Université Cadi Ayyad, Beni-Mellal

ii
 Acknowledgements

I would like to express my gratitude to my supervisors Professor Abderrafi Benhallam

and Professor Elisabeth Selkirk for their constant encouragement and guidance in the preparation
of this dissertation. Professor Benhallam ought to be praised for both his human and academic
qualities. I have learnt a great deal from him about social life and phonology over the last 15
years. I am particularly indebted to him for having encouraged me to persevere in my 1993 third
cycle thesis at a moment I was about to give it up. Professor Selkirk has given me the chance to
pursue study and research at the University of Massachusetts at Amherst by accepting to act as
co-supervisor of this dissertation. In addition to that, she has contributed substantially to improve
this work by her challenging questions, sound criticism and insightful comments and suggestions,
leading to major revisions.
Special thanks go to the staff, faculty and graduate students in the Linguistics Department
of the University of Massachusetts at Amherst for hosting me in their department for 12 months
and also for providing a wonderful research environment while I was writing parts of this
dissertation. I have benefited a great deal from the 1997 Fall Semester classes of Professor John
McCarthy and also from his comments on some earlier chapters herein contained and those of
some graduate students, namely Paul de Lacy, Steve Parker and John Alderete. For all of that, I
would like to thank them all. Thanks are also due to John Kingston for helping out with
laboratory work and to two other linguists outside the University of Massachusetts, namely
Michael Kenstowicz and Mohamed Guerssel who have read and commented on the second
chapter of this dissertation.
I am gratefully indebted to a number of other linguists for having contributed to make this
a better dissertation than it would otherwise have been. I wish to thank all the members of the
Phonology-Morphology Research Group based at the Faculty of Letters in Rabat for their
insightful discussion and challenging questions which have certainly helped sharpen my
arguments and improve the content of this dissertation. In particular, I would like to thank Samira
Rguibi, Karim Bensoukas, Abdellatif Al Ghadi, Said Imouzaz, and Abdelouahed Hammari. This

iii
work could not have been completed without precious help from Mohamed Es-Safi and Mourad
Mawhoub. The first allowed me to conduct the quantitative test about stress by hosting me at the
Faculty of Letters I in Casablanca and inviting his students to complete the questionnaire related
to that test. The second helped me pitch-track the recordings obtained from the subjects and
showed great patience during the long hours we spent in front of his computer doing this task.
I am also grateful to all the members of the Moroccan-American Commission for
Educational and Cultural Exchange for having awarded me a grant which enabled me to do
research for three periods of four months each at the University of Massachusetts at Amherst
during the Fall Semesters of the years 1997, 1998 and 1999.
I would also like to thank my friends, especially Mustapha Chebbak, Mohamed Saki,
Kamal Alaoui-Fatine, Abdelouahed Safadi, Mustapha Minoual, Driss Maskine, Brahim
Taoukoul, Abdelkrim Mabchour, Mohamed-Sghir Syad, Mohamed Rakii and those whose names
I cannot recall right now. Additional thanks, for service above and beyond duty, are due to
Brahim, Abdelkrim, Mohamed-Sghir and Mohamed who have accepted voluntarily to help with
their proofreading skills.
Thanks to all my colleagues at the Faculty of Letters in Beni-Mellal who have contributed
in a way or another in this dissertation, namely Noureddine Bourima, Cherki Karkaba, Khalid
Chaouch, Mustapha Mamaoui, Mohamed Taki, Said Bennis and Rachid Hamdi. Also, I should
not forget to thank my former professors at the Faculty of Letters I in Casablanca and the Faculty
of Letters in Rabat, for without them I would not have reached this stage.
Finally, I wish to thank all the members of my family for their encouragement, moral and
emotional support while I was finalizing this work. To my mother Fatna, to my sisters Keltoum,
Jamila and Malika, and to my brothers Mohamed and Saïd I say thank you for loving me.

iv
 Dedication

To The Memory of My Father …

v
 Abstract of the Dissertation

The objective of this dissertation is to account for some aspects of the prosodic phonology
and morphology of Casablanca Moroccan Arabic within the framework of Optimality Theory as
conceived in Prince and Smolensky (1993) and McCarthy and Prince (1993a) and developed in
the Correspondence model of McCarthy and Prince (1995, 1999) and other related works. It is
argued that prosodic aspects such as the syllable structure and the stress system and
morphological aspects such as the nisba adjective, the causative, the passive participle and the
diminutive are better understood as cases involving interaction between two types of conflicting
universal constraints: markedness constraints and faithfulness constraints.
It is shown that a division must be established between two types of syllables: a major
syllable whose nucleus is one a schwa or one of the full vowels [i, u, a], and a minor syllable
which consists solely of a moraic consonant. Granting a moraic status to this consonant is
allowed for the purpose of achieving the foot binarity requiring, especially in nonderived
trisegmental words on the pattern CCV, CCəC or CəCC, where the first consonant of the initial
cluster and the second consonant of the final cluster form minor syllables on their own. The fact
that the schwa is epenthesized before the final consonant of nonderived trisegmental verbs,
adjectives and a class of nouns follows from the requirement that the right edge of the stem be
aligned with a prominent syllable, which corresponds to a major syllable. The nominal class
showing the CəCC pattern is shown to abide by markedness constraints favoring schwa syllables
with a higher sonority coda. The dissertation also offers an adequate analysis of the problematic
cases of cyclic schwa syllabification in terms of a subset of output-output constraints, one of
which demanding phonological identity between a derived form and its morphologically-related
base form.
The theoretical framework herein conceptualized gives a straightforward answer to the
puzzling stress system of the language which shows both trochaic and iambic feet. In particular,
it is argued that in both isolation words, where the foot is trochaic and context words, where the
foot is iambic, the location of stress and consequently the foot types that emerge depend on the
hierarchical organization of prosodic words into phonological phrases. A unitary account of the
stress system is offered to the effect that both trochaic and iambic feet occur in the language.
Trochaic feet surface as optimal when the word is in isolation (i.e. when it is a phonological
phrase); iambic feet arise when the word is in context .
The dissertation also argues that morphological classes such as the causative, the passive
participle and the diminutive are governed by a prosodic constraint requiring that their output
conform to an iambic foot. The causative form, which has previously been analyzed as involving
prosodic circumscription, is now reanalyzed as a case of partial reduplication which can be
accounted for by constraints demanding correspondence between the base and its reduplicant. In
particular it is shown that the constraint calling for an iamb consisting of a sequence of two light
syllables takes priority over the constraint on the base and reduplicant identity and therefore
block total reduplication. The passive participle and the diminutive are two instances that resort
to augmentation to achieve an iambic foot type. In the case of the passive participle, it is argued
that the prefinal vowel that appears in certain classes of non derived verbs is the result of the
constraint requiring that the output conform to an iambic foot consisting of a sequence of light

vi
and heavy syllables. In case where augmentation would lead to the violation of higher ranked
constraints, the foot that emerges as optimal consists of a sequence of two light syllables. In the
case of the diminutive, if augmentation applies, it is for the sole purpose of achieving a light-light
foot. Augmentation itself takes tow different forms: either by the addition of schwa syllables to
words that are masculine, or by the suffixation of the feminine morpheme to words which are
inherently feminine.

vii
 Table of Contents

Page
Acknowledgements iii
Dedication v
Abstract vi
Table of Contents viii
Introduction 1

Chapter One : Preliminaries 6

1. Introduction 6
2. Review of the Literature on Moroccan Arabic 6
3. Casablanca Moroccan Arabic 13
3.1 Geographical Situation 13
3.2 Dialectal Situation 13
4. The Data 14
5. Basic Phonology and Morphology of Casablanca Moroccan Arabic 15
5.1 Consonant and Vowel Inventories 15
5.2 Morphology 20
5.1 Root-and-Pattern Morphology 20
5.2 Relevant Morphological Processes 21
5.2.1 The Nisba 22
5.2.2 The Causative 23
5.2.3 The Passive Participle 24
5.2.4 The Diminutive 24
6. Theoretical background 25
6.1 Prosodic Morphology before Optimality Theory 26
6.1.1 Prosodic Morphology 26
6.1.2 Circumscription within Prosodic Morphology 27
6.2 Optimality Theory 30
6.2.1 Introduction 30

viii
 6.2.2 The Construction of Grammar in Optimality Theory 32
6.2.3 Prosodic Morphology within Optimality Theory 34
6.2.4 Correspondence in Optimality Theory 35
7. Conclusion 39

Chapter Two : The Syllable Structure of Casablanca

Moroccan Arabic 41
1. Introduction 41
2. Previous account of Moroccan Arabic syllable Structure 42
3. Casablanca Moroccan Arabic Syllable Structure and Optimality Theory 51
3.1 Syllable Structure in Optimality Theory 52
3.2 Universal Constraints on Casablanca Moroccan
Arabic Syllable Structure 55
3.3 The Prosodic Word in Casablanca Moroccan Arabic 64
3.4 Syllable Structure and Geminates 75
3.4.1 Final Geminates 78
3.4.2 Initial Geminates 82
3.5 The Problem of Cyclic Syllabification 84
3.6 Sonority and Syllabification 87
4. Conclusion 96

Chapter Three : The Stress System of Casablanca

Moroccan Arabic 99
1. Introduction 99
2. Review of the Literature on Moroccan Arabic Stress 100
3. Establishing an Empirical Basis for Stress in Casablanca Moroccan
Arabic 107
3.1 The Quantitative Test 107
3.1.1 The Hypothesis 107
3.1.2 The Questionnaire 108
3.1.3 The Data 109
3.1.4 The Subjects 111
3.1.5 Results and Analysis 111
3.1.5.1 Disyllabic Words 114
3.1.5.2 Trisyllabic Words 116
3.1.5.3 Polysyllabic Words 118

ix
 3.2 The Instrumental Test 119
3.2.1 The Corpus 120
3.2.2 The Subjects 121
3.2.3 Words in Isolation 121
3.2.4 Words in Context 125
4. Words in Isolation and Words in Context: A Unified Account
of Stress in Casablanca Moroccan Arabic 129
5. Conclusion 149

Chapter Four : Output-Output Correspondence in

Casablanca Moroccan Arabic 152
1. Introduction 152
2. Transderivational Relations in Optimality Theory 155
3. Cyclic Syllabification in Casablanca Moroccan Arabic 160
3.1 Introduction 160
3.2 Cyclicity as Output-Output Correspondence 162
4. Truncation 175
5. Prosodic Circumscription as Reduplication 187
6. Conclusion 201

Chapter Five : The Prosodic Nature of the Passive

Participle in Casablanca Moroccan Arabic 203
1. Introduction 203
2. A taxonomic Survey of the Passive Participle 204
2.1 The Passive Participle of Trisegmental Verb Bases 204
2.1.1 Non-derived Verb Bases 204
2.1.2 Derived Verb Bases 207
2.2 The Passive Participle of Quadrisegmental Verb Bases 209
2.2.1 Non-Derived Verb Bases 209
2.2.2 Derived Verb Bases 210
3. Previous Account of the Passive Participle 212
4. Alternative Optimality-Theretic Analysis 216
4.1 Introduction 216
4.2 Verb Prosodization 219
4.3 An Iamb-based Analysis of the Passive Participle 223
4.3.1 Augmented Passive Participle Forms 225

x
 4.3.2 Unaugmented Passive Participle Forms 236
4.3.2.1 Derived Trisegmental Verb Bases 237
4.3.2.2 Quadrisegmental Verb Bases 241
5. Conclusion 246

Chapter Six : Augmentation in the Prosodic Morphology of

the Diminutive 248
1. Introduction 248
2. The Data 249
3. Against a Circumscrptive Analysis of the Diminutive 252
4. On the need for the Constraint INITIAL-CC 255
5. Evidence for INITIAL-CC 261
5. 1 Labialization 261
5.2 Labial Consonant Gemination 266
6. Augmentated Diminutive Forms 277
6.1 Augmentation as the Addition of the Default Syllable 277
6.2 Augmentation as the Feminine Morpheme Suffixation 288
7. Unaugmented Diminutives Forms 292
8. Conclusion 299

Conclusion 301

Appendices 306
Appendix A: The Constraints Listed in the Dissertation 306
Appendix B: The Quantitative Test: TheQuestionnaire 310
Appendix C: The Istrumental Test: The Data 313
Appendix D: Scores of Subjects in Test Items in Isolation 315
Appendix E: Pitch Tracks of Selected Test Items in Isolation 323
Appendix F: Scores of Subjects in Test Items in Context 331
Appendix G: Pitch Tracks of Selected Test Items in Context 340
Bibliography 348

xi
 Introduction

Since the establishment of morphology as an autonomous component in Generative

Grammar with Chomsky (1970) and Halle (1973), a number of theories dealing with word
structure and the organization of the lexicon have been developed. Examples of such theories
include the theory of Word Formation (Aronoff 1976), the theory of Level-Ordered Morphology
(Siegel 1974, Allen 1978), the X-bar theory of word structure (Selkirk 1982) and the theories
dealing with the interaction between morphology and phonology such as the theory of Lexical
Morphology and Phonology (Kiparsky 1982, 1985; Mohanan 1982, 1986) and the theory of
Prosodic Morphology (McCarthy 1979, 1981, 1982; McCarthy and Prince 1986 et seq.). In
phonology, parallel developments have been achieved since the standard theory of Chomsky and
Halle (1968). The traditional view that the input to phonological representations is matched to the
output by a set of rewrite rules has been challenged by theories such the theory of Constraints and
Repair Strategies (Paradis 1988) and Government Phonology (Kaye et al. 1985, 1990). The first
theory explains phonological alternations by appealing to a set of surface-unviolated constraints
complemented by repair strategies whose role is to solve any violations resulting from constraint
conflicts; the second attempts to replace rules by a set of universal principles common to all
linguistic systems along with a series of language-specific parameters.
The proposals concerning the role of constraints in grammar advanced in the
aforementioned works and others were very important to the emergence of a theory of constraints
and constraint interaction, known as Optimality Theory (Prince and Smolensky 1991, 1992,
1993, McCarthy and Prince 1993a). The central idea of Optimality Theory (henceforth OT) is
that Universal Grammar consists of a set of ranked and violable constraints on output structures
together with a general means of resolving their conflicts. The means that OT resorts to in order
to resolve conflicts is to rank constraints in a strict dominance hierarchy. Higher-ranked
constraints take precedence over lower-ranked ones. The constraints themselves fall into two
types: well-formedness constraints and faithfulness constraints. Well-formedness constraints
enforce segmental or prosodic markedness, whereas faithfulness constraints militate against any
changes between an input and its output.
The basic principles of OT have been applied to the areas where prosody and morphology
interact. The characterization of patterns of reduplication, root-and-pattern morphology and
truncation is now achieved through the interaction of more general principles of phonological
and morphological well-formedness. Within the OT model of Prosodic Morphology interaction
takes the form of constraint domination, where prosodic well-formed constraints have priority
over morphological requirements. Later developments within the OT framework have given rise
to Correspondence Theory (McCarthy and Prince 1995, 1999). In Correspondence Theory
(henceforth CT), faithfulness is conceived of as a set of constraints on correspondence relations
between the input and the output, the base and the reduplicant.
It is within the general framework of OT and CT that we propose to analyze some aspects
of the prosodic phonology and prosodic morphology of a variety of Moroccan Arabic —
Casablanca Moroccan Arabic (henceforth CMA). The aspects of CMA prosody that will be
considered are the syllable structure and the stress system which, we judge, are germane to an
adequate analysis of the prosodic morphology. As to the morphological aspects that will be
treated, they include the nisba adjective, the causative, the passive participle and the diminutive
formation. The choice of these morphological classes stems from three reasons. First, some of
these classes have received little attention from researchers. The works that have been undertaken
so far treat them as residual phenomena (for the passive participle, see Youssi 1986, Marsil 1988,
Boudlal 1993. For the nisba adjective, see Al Ghadi 1990, Boudlal 1993. For the diminutive, see
Benhallam 1980, Al Ghadi 1990). Second, the works that have dealt at length with these
morphological classes have failed to come up with an account that is explanatorily adequate and
this is either because these works have not had recourse to prosody as is the case with the
analysis of the diminutive in Lasri (1989) or because of the limitations of the theoretical
framework adopted as it is the case with the analysis of the causative in Bennis (1992). Third,
most of the analyses undertaken in these works resort to rules and/or mechanisms that are
language-specific and as such fail to recognize that individual grammars are best undertood to
derive from a large set of ranked and violable constraints pertaining to Universal Grammar.
Our objective in the present dissertation is to exploit the basic tools available in the OT
framework and see to what extent they can allow us to adequately analyze the aforementioned

2
aspects of CMA prosody and morphology. Attaining this objective means a significant move
towards the establishment of an individual grammar of CMA based on the reranking of a set of
violable universal constraints.
The dissertation is organized into six chapters. Chapter one, which contains the
preliminaries, gives details about the relevant phonology and morphology of the variety of MA to
be analyzed as well as the theoretical framework that will be adopted in this dissertation.
Chapter two offers a constraint-based account of CMA syllable structure and its
interaction with schwa epenthesis. Therein, we argue that prosodic structure assignment and
consequently schwa epenthesis are governed not by the step-by-step syllable structure building
rules proposed within the derivational frameworks, but by universal constraints such as the
constraint requiring the onset and the constraint prohibiting the coda. The recalcitrant problem of
schwa occurrences will be shown to derive from the interaction of well-formedness constraints
and faithfulness constraints and their relative ranking. In particular, it is argued that the
syllabification of words on the pattern CCəC, where the schwa appears before the final consonant
in trisegmental verbs, adjectives and a class of nouns, follows from a prosodic constraint
requiring that the stem be iambic. The nominal class showing the CəCC pattern will be shown to
abide by markedness constraints requiring that the sonority value of the consonant serving as a
coda to the schwa syllable be as close as possible to that of the nucleus. Our account of CMA
syllable structure incorporates the proposals made within the derivational frameworks undertaken
by our predecessors, subject to drastic changes and reformulations to conform to the theoretical
framework herein conceptualized.
Chapter three analyzes the stress system of the language and is divided into four major
sections. In the first section, we present a review of the previous discussions of MA stress and
state their limitations. In the second section, we set the background for an empirical study of
stress in CMA whose objective is to quantify the results obtained form a perceptive test given to
native speakers. A sample of native speakers of CMA were given a list of items and asked to
place stress relying on their intuitions. Later, the recordings of 5 subjects were subjected to an
instrumental test to see whether or not the results obtained in this test match up those obtained
from the quantitative test. It will be shown that the location of stress depends on whether or not
the items considered are in isolation or in context. When words occur in isolation, the foot types
obtained are trochaic. On the other hand, when words occur in context, the foot types obtained

3
are iambic. This puzzle about the stress system (i.e. allowing both iambic and trochaic feet in the
language) cannot be accounted for within a parametrized theory where a language is forced to
choose between iambs and trochees. Within the OT model, it will be shown that the location of
stress in both isolation and context cases leading to trochaic and iambic feet follows from the
Selkirk (1978) prosodic hierarchy, namely the organization of prosodic words into phonological
phrases.
Chapter four treats the three linguistic cases of cyclic syllabification, the nisba adjective
and the causative derivation in the light of the extended version of CT. In derivational models,
underlying sequences such as /CCC-C/, where the final consonant is a subject affix, surface as
[CCəC-C] and not as *[CəCC-əC] with the schwa being epenthesized before the final consonant
of the stem. Benhallam (1990), for example, assumes that cases such as these require a cyclic
treatment whereby schwa epenthesis applies in the first cycle to give the stem [CCəC] but is
blocked from application in the second cycle after the addition of the suffixal consonant. Such an
analysis works for quadrisegmental bases on the pattern /CCCC-C/ but cannot be generalized to
account for sequences such as /CCC-C/ where the affixal consonant is the object marker. These
sequences surface as [CəCC-əC] and not as *[ CCəC-C] as predicted by cyclic syllabification.
The analysis we will offer in this chapter derives cyclic syllabification from the interaction of
constraints requiring identity between the base form and its morphologically related output form,
much in the spirit proposed in works such as Basri et al. (1998) and Selkirk (1999). The
principles advanced in the extended version of CT will also be tested to see if they can
adequately account for the formation of the causative and the formation of the nisba adjectives
derived from compounds. It will be shown that the best way to characterize the causative in CMA
is by assuming that it involves reduplication. As to the nisba adjectives, it will be argued that a
fair account of this morphological class requires reference to principles governing morphological
structure.
Chapter five deals with the passive participle and justifies the need for an iamb-based
analysis of this morphological category. The assumption made here is that the passive participle
morpheme is the prefix [m-] and not the discontinuous morpheme [m-...-u-...]. The prefinal vowel
that shows up in some passive forms is argued to be the language default vowel which is
epenthesized for prosodic purposes. Two possible hypotheses are tested to explain the passive
participle formation. The first is based on the assumption that the verbal base to which the

4
passive participle prefix is adjoined must conform to an iambic foot without ever specifying the
nature of this iamb. The second, which is adopted in this work, posits that iambicity holds on the
passive participle form rather than on the verbal stem. It will be argued that augmentation by u-
epenthesis applies for the sole purpose of achieving an output form that conforms to the most
harmonic iambic foot consisting of a sequence of a light and a heavy syllables. The cases that do
not show this augmentation surface with an iambic foot which consists of a sequence of two light
syllables.
Finally, chapter six presents further support to the iamb-based analysis witnessed with the
passive participle in chapter five. We will show that the diminutive resorts to two different types
of augmentation to achieve an iambic foot: internal augmentation by the addition of schwa
syllables to monosyllabic bases which are masculine, and external augmentation by the
suffixation of the feminine morpheme to a class of inherently feminine monosyllabic words. In
both cases, it will be shown that the iambic foot type achieved is of the one that consists of two
light syllables. The chapter also considers the process of labialization because of its tight
relevance to the diminutive. It will be shown that the realization of the round-feature morpheme,
responsible for the labialization of the initial segment of the base form, follows from the
interaction of faithfulness constraints and markedness constraints on labialized consonants.

5
 Chapter One

Preliminaries

1. INTRODUCTION

This chapter presents the preliminaries which we judge necessary to the understanding of
the whole dissertation. It is divided into 5 major sections. Section 2 gives a brief review of the
literature on Moroccan Arabic as well as on the other varieties used in Morocco. Section 3
presents the variety of Moroccan Arabic that constitutes the subject of study in this work, i.e.
Casablanca Moroccan Arabic. In particular, we will point out some of the characteristics that
distinguish this variety from the other varieties of Moroccan Arabic. Section 4 is about the
procedure followed in the collection the data. Section 5 gives background knowledge about the
phonology and morphology of the variety to be studied. Therein, we list the consonant and
vocalic inventories as well as the relevant morphological categories that will be dealt with at
length in the core chapters. Finally section 6 establishes the theoretical background for the
analysis of some aspects of the phonology and prosodic morphology of Casablanca Moroccan
Arabic. It will especially lay down the basic tenets of Optimality Theory and Correspondence
Theory which constitute together the theoretical framework adopted in the present work.

2. REVIEW OF THE LITERATURE ON MOROCCAN ARABIC

The body of linguistic research on MA could be traced back to the early forties with
Harris's (1942) famous article which was subsequently followed by Cantineau (1950), Brunot
(1950), Harrell (1962) and Abdelmassih (1970), among others. Although the works of these
scholars were mainly didactic and descriptive in nature, they made very important contributions
to the study of MA. The main concern of these scholars was to make the preliminary

6
investigations into the MA sound system (Harris, 1942) or establish courses for foreign learners
(Brunot 1950, Harrell 1962, and Abdelmassih 1970).
The neglect of MA by Moroccan linguists was partly due to the diglossic relationship of
MA and classical Arabic (henceforth CA). Ferguson (1959) described the different functions of
CA (the high variety) and MA (the low variety). CA is the medium of written communication
because it is highly codified whereas MA is a medium of oral communication. The neglect could
also be due to the complexity of the linguistic situation in Morocco. This complexity is due to
what Youssi (1983, 1986, 1990) refers to as a triglossia situation with multilingualism. This
situation is characterized by a competition between Arabic, Berber and French, on the one hand
and between CA, MA and Middle Moroccan Arabic (MMA), on the other1.
It was not until the mid-seventies that Moroccan researchers have turned to the study of
their native language. Works such as Youssi (1977, 1986) Benhallam (1980), Benkirane (1982),
Benkaddour (1982) tried not only to set up the phonemic system of sounds but also to test the
hypotheses advanced by different schools (See Benhallam 1989 for a review of these works).
Youssi (1986) addresses different aspects of MMA. In phonology, for example, he deals
with the phonematic characteristics of MMA and the different syntagmatic relations holding
between phonemes. His presentation of these aspects is based on the relations and functions of
the double articulation. The consonantal system he established is based on three orders of
correlational oppositions: simple versus tensed, inflated versus non-inflated and rounded versus
unrounded2. The relations between phonemes as well as their arrangement are assumed to be
important to morphology. In morphology, on the other hand, he proposes three classes. He
determines the predicative function of verbs and their formal variations in relation to their classes
of compatibility (modification related to tense, aspect, person etc...).
Benhallam (1980) tackles syllable structure and the interaction of phonology and
morphology within the framework of Generative Phonology of Chomsky and Halle (1968). He
distinguishes between phonological and phono-lexical rules on the one hand, and morphological
and morpho-lexical rules, on the other. In phonology, he considers the consonantal systems of

1
According to Youssi (1989), Middle Moroccan Arabic (also called Modern Moroccan Arabic (Youssi 1992)) is
based on both Literary Arabic and Moroccan Arabic. Literary Arabic provides the essential of the lexicon while
Moroccan Arabic provides the essential of the grammar.
2
Instead of gemination and emphasis, Youssi (1998) uses the corresponding terms tensing and inflation.

7
both MA and Standard Arabic (henceforth SA) and the effect of certain phonological rules on
syllable structure. He also shows that schwa epenthesis in MA and other aspects of the
phonology of SA, such as vowel/glide alternation, can best be described in terms of a theory that
makes use of the syllable.
Benkirane (1982) tries to set up the different types of syllables in MA within a non-linear
framework. He investigates the issue of syllable boundary in MA in relation to three acoustic
parameters which are fundamental frequency, intensity and duration. He also considers some
syllable-based phonological rules such as stress and schwa epenthesis. Although his analysis is
phonetically oriented, it remains, nonetheless, one of the very few experimental studies on MA.
Benkaddour (1982) examines some aspects of the phonology and non-concatenative
morphology of the Rabati MA within the framework of non-linear phonology. He studies the
phonotactic constraints and the interaction between syllable weight and stress. He also shows the
importance of prosody in the application of certain phonological rules such as schwa deletion and
schwa epenthesis and their interaction with stress and syllable structure. He applies McCarthy's
(1979) version of autosegmental theory to account for some aspects of the non-concatenative
morphology of MA.
Later work on MA was undertaken with the purpose of reconsidering some of the
recalcitrant problems raised in the works above or treating other aspects that have not been dealt
with. Examples of this body of research include Bellout (1987), Ait Hammou (1988), Marsil
(1988), Hammoumi (1988), Rguibi (1990), Al Ghadi (1990), El Himer (1991), Imouzaz (1991),
Bennis (1992), Mawhoub (1992), Nejmi (1993), Boudlal (1993), Meliani (1994), Hammari
(1996), to cite but a few. The common denominator among these theses is that they all apply
recent approaches or theories in phonology to analyze aspects that relate to the phonology and/or
morphology of MA.
The theses that have been written on MA so far do not clearly state what is meant by MA.
This fact shows that the writers of these theses do not deal with the same variety. In fact the
divergence of these theses can be attributed to language variation. In this respect, Benhallam
(1998) presents data that relate to variation in MA and discusses three main problems that these
data pose: schwa strengthening, diphthongization and q/g variation.
Schwa strengthening refers to situations of variation between the schwa and the full
vowels [i, u, a] found in the northern and less urban central areas of Morocco. Benhallam (1998)

8
(see also Rguibi, forthcoming, and Hammari 1996, 2000) has shown that the variation related to
vowel quality is both an instance of dialectological variation and an illustration of historical
development. For illustration, consider some of the examples taken from Benhallam (1998: 28):

-1-

a. məqla maqla frying pan

lmuDəʕ lmuDaʕ the place
mxədda mxadda cushion

b. γəzlan γuzlan gazelles

qəTRa quTRa drop
nəSS nuSS half

c. waħəd waħid one

lqərd lqird the monkey
RRaʒəlRRaʒil the man

In 1a, the schwa alternates with the vowel /a/; in 1b, it alternates with the vowel /u/; and in 1c, it
alternates with the vowel /i/. Benhallam has shown that the alternation between the schwa and the
full vowels could be viewed either as a process of schwa strengthening, i.e. a process whereby
the schwa becomes a full vowel; or as a process of vowel reduction. Whichever the case is, the
alternation exists and reflects dialectal variation among the different varieties of MA. The
northern varieties of MA, such as the Tazi MA (Rguibi, forthcoming) or Rhafsai MA (Hammari,
1996, 2000), use a full vowel where other varieties use the schwa.
The second type of variation that constitutes a point of divergence among the different
works on MA relates to diphthongization. Diphthongization is a variation found in rural areas in
most cases. It is a phenomenon whereby high vowels alternate with the corresponding diphthongs
as in the items below taken from Benhallam (1998: 30):

-2-

lfuDa lfawDa disorder

9
DDu DDaw light
zzitun zzaytun olive
STila STayla small bucket

Here again, diphthongization is a characteristic of rural varieties, such as the variety spoken in El
Jadida (Benhallam and Dahbi 1990). In other varieties of MA, such as the southern variety of
MA spoken in Taroudant (Al Ghadi 1990), the diphthongs are monophthongized. This process
can be attributed to the influence of Berber where diphthongs are inexistent.
The third type of variation Benhallam (1998) cites is q/g variation. This variation
concerns the segments ʔ/q/g with the first two acting as mutually exclusive variants of each other
depending on the variety considered. Native speakers of the variety of MA spoken in Fès, for
example, tend to use [ʔ] instead of [q]. But the use of [ʔ] is in the process of fading away,
especially for the young generation. As to the q/g variation3, it seems that it is a way of
differentiating urban and northern from rural and southern varieties. Because native speakers may
not be consistent all the time in the use of either /q] or /g/, Benhallam assumes that speakers can
be sorted out into two classes: those who maximize the use of /q/ (qal speakers) and those who
maximize the use of /g/ (gal speakers).
In addition to what has been mentioned above, variation may also be due to the contact
between MA and other languages in Morocco such as CA, the native Berber varieties and foreign
languages. The influence of CA (the official national language of the country) on MA is
perspicuous in the speech of the young generation. The spread of education has led people to use
some intermediate forms between CA and MA thus resulting in what is referred to as MMA
(Youssi, 1986, 1992). This influence has led some Moroccan linguists to consider CA as the
reference language to the extent that one may even think that it is impossible to study MA
without ever referring to CA. As a matter of fact, we cannot neglect the fact that MA and CA are
related. However, there are significant differences between the two languages in their
phonologies, morphologies and lexicons. This makes it necessary to describe each language
separately without even referring to the other.

3
Benhallam (1998) (see also Moumine 1990, 1995) has pointed out that the substitution of /q/ for /g/ in certain
contexts leads to change in meaning: e.g. [qəlb] “heart” vs. [gəlb] “stomach”, [qərʕa] “bottle” vs. [gərʕa]
“pumpkin”. This type of variation is not relevant to the cases being discussed above.

10
 In addition to the influence of CA, there are other native languages that are in a contact
situation with MA. These include the three local varieties of Berber with their different regional
accents: Tarifit in the northern part of the country (the Rif mountains), Tamazight in the contral
part (the middle Atlas) and Tashlhit in the southern part (the High Atlas) (See Boukous 1979 for
a sociolinguistic profile of Morocco). Here again, the northern varieties of MA are influenced by
Tarifit, the central varieties are influenced by Tamazight, while the southern varieties are
influenced by Tashlhit.
The contact between Berber and MA is established at different linguistic levels. At the
phonological level, Boudlal (1998) has shown that a number of cases involving labialization of
the consonants /k, g, x, γ, q/ in what he calls southern varieties of MA are in fact cases of contact
between Tashlhit and these varieties of MA. By virtue of this contact, the corresponding
labialized consonants of /k, g, x, γ, q/ have been established as separate phonemes in certain
varieties of MA.
At the morphological level, a number of affixes have been incorporated into the
morphology of MA such as the Berber feminine discontinuous morpheme [ta-...-t] found in items
such as [tamγart] “woman” and [tafunast] “the cow”. However, such a morpheme is used in MA
not only to express gender but also to derive nouns of profession as seen in the items below:

-3-
Noun Noun of profession Noun Gloss
gəzzar tagəzzart butcher
xərraz taxərrazt shoe maker
ħəddad taħəddadt blacksmith
Səbbaγ taSəbbaγt painter

At the lexical level, for example, the MA lexicon has integrated a number of lexical items from
Berber such as [bħira] “vegetable garden” and [gaduma] “hoe”. It has also retained a number of
words of Berber origin characterized by the initial vowel such as [argan] “a kind of oil derived
from Argan tree” and [amlu] “a mixture of Argan and almonds”.
In addition to CA and the native varieties of Berber, foreign languages do undoubtedly
have an influence on MA. Their influence is characterized by cases of lexical borrowing from
French and Spanish. Here again, the northern varieties of MA are influenced by Spanish while

11
the other varieties are influenced by French. All the cases of borrowing have undergone the
phonological and morphological processes of MA in their way to being fully integrated.
Examples of lexical borrowing include words such as [RwiDa] “wheel”, [falTa] “fault” and
[Bala] “shovel”, from Spanish and words like [Tabla] “table”, [kaSiTa] “casette” and [Rubini]
'water tap', from French.
In short, the influence of the above languages and varieties on MA is really considerable.
It is this influence resulting from situations of language contact which is partly held responsible
for having different varieties of MA. In this respect, Boukous (1998: 9) distinguishes between
four major varieties of MA which are given below:

(i) The Urban Variety which is used in traditional Moroccan cities such as Fès, Rabat,
Salé and Tetuan and is marked by characteristics features from Andalusian varieties of Arabic.

(ii)The Mountain Variety (called Jebli) which is used in the north-west of the Morocco
and is characterized by the influence of Berber at the levels of phonology, morphology and the
lexicon.

(iii) The Bedouin Variety which is used by Arabophones of Bani Hilal and Bani Maaqil
origin who settled in the plains of Gharb, Chawiyya, Doukkala and the neighboring cities such as
Mohammedia, Casablanca, El Jadida, Marrakech, etc ...

(iv) The Hassani Variety used by speakers of Bani Maaqil origin now living in the Sahara
in the southern part of Morocco.

It should be noted that each of these varieties has its own specificities and at the same time each
shares the bulk of the grammar with the rest of the varieties, a fact which has led some linguists
to establish a MA where regional variations are not represented (Cf. Benhallam and Dahbi 1990
who establish the consonantal and vocalic inventories of what they call Average Moroccan
Arabic)
The corpus on which the present work is based is drawn from the variety of MA spoken
in Casablanca.

12
3. CASABLANCA MOROCCAN ARABIC

The variety of MA spoken in Casablanca (henceforth CMA) is characterized by certain

particularities attributed to Bedouin dialects4. Despite the heterogeneity of the inhabitants of the
city of Casablanca, we can, nonetheless, speak of a somehow homogeneous variety where
regional variations have been gradually neutralized to yield the variety referred to as CMA
described in works such as Khomsi (1979), Moumine (1990), Imouzaz (1991, forthcoming),
Mawhoub (1992), Boudlal (1993) and Nejmi (1993).

3.1 Geographical Situation

Casablanca is a coastal city located at the western side of Morocco on the Atlantic Ocean.
It is situated between Rabat in the north and El Jadida in the south. Casablanca is one of the
newly born cities, founded at the beginning of the twentieth century. Despite its recent
establishment, it has developed in such a short period that it is now considered the biggest
economic and industrial city in Morocco. The development is essentially due to its strategic port
on the Atlantic coast, a fact which makes the city of Casablanca a bridge between Morocco and
the rest of the world.
Because of its economic leadership, Casablanca has received millions of new settlers
since the beginning of the present century. At the moment, its population is estimated to be more
than four million. Most of the new settlers have come from the neighboring plains of Chawiyya,
Doukkala and Chyadma bringing with them their rural dialects. By the middle of the century,
especially after the independence of Morocco in 1956, other new settlers from the south of the
country (the Souss and the Dra Valleys) arrived in the city to contribute to the economic
development of Morocco. With urbanization and modernization taking place, great contact has
developed between Casablanca and the neighboring rural areas.

3.2 Dialectal Situation

4
On the different types of MA dialects spoken in Morocco, the reader is referred to Boukous (1998).

13
 As a result of the migration of the rural speakers of the neighboring areas to Casablanca
as well as migrants from other regions of the country, especially Berber-speaking areas,
Casablanca has become a melting pot of different rural and urban dialects. Undoubtedly, contact
between these varieties has taken place, causing dialect mixture and giving rise to new dialectal
forms.
Besides varieties of MA, foreign languages, namely French, Spanish and English, have
also established their presence in the city. Given the interaction of foreign languages as well as
the different regional varieties of MA and Berber, the linguistic situation in Casablanca seems
quite difficult to define. It should be pointed out here that our purpose is not to study language
variation or establish a diachronic study of MA but rather study some aspects of the phonology
and prosodic morphology of CMA.
Despite the heterogeneous dialectal situation in Casablanca, it is still possible to speak of
one specific variety of CMA. Moumine (1990), for example, describes the linguistic situation in
Casablanca and shows that the presence of so many languages in the city has led to the
emergence of an interdialect. As a matter of fact, it is this interdialect that later developed into
what came to be known as CMA whose native speakers could be identified throughout Morocco.
Of course this dialect shares most of the grammatical features of the other varieties of MA but at
the same time differs from them with respect to certain phonological and morphological aspects.
The difference and similarity between
CMA and other varieties of MA will be made clear, when relevant, as we proceed in this
dissertation.
Our concern with establishing the variety of MA dealt with in the present work stems
from our objective to achieve the greatest degree possible of homogeneity. This objective cannot
be achieved unless the corpus drawn from the variety under study shows some consistency to
enable us to capture a greater degree of regularity and to come up with significant generalizations

4. THE DATA

The analysis presented in this study is based on CMA, of which the writer is a native
speaker. The data were collected in Casablanca among family members and friends, in particular.
In collecting the data, certain variables have been taken into consideration. The informants I have

14
chosen were all born in Casablanca. Furthermore, their parents have been living there for a long
period. For the sake of homogeneity, I have disregarded the data collected from informants
whose parents speak Berber.
The corpus chosen is representative of the phonological and morphological aspects dealt
with in CMA; however, it can in no way be said to be exhaustive. The corpus has been enriched
by data from published work on the grammar of MA, especially Harrell (1962) and Abdelmassih
(1973). Only the data identified by native speakers of CMA were included in the corpus. Other
sources of material include the substantial body of data on MA found in Youssi (1986) and other
works on CMA, particularly works such as Imouzaz (1991), Mawhoub (1992), Boudlal (1993)
and Nejmi (1993).

5. BASIC PHONOLOGY AND MORPHOLOGY OF CASABLANCA

MOROCCAN ARABIC

This section sets the linguistic background of the variety of MA being studied. It
introduces the consonantal and vocalic system of CMA as well as the morphological aspects that
will be dealt with at length in the core chapters.

5.1 Consonant and Vowel Inventories

The consonantal system of MA has received various treatments from different linguists.
These linguists do not agree on the number of the consonant phonemes the language should have.
Some claim that MA has 31 consonant phonemes (Harrell, 1962); others (Benhallam 1980,
Benkirane 1982, Benkaddour 1982 and Youssi 1986) claim that there are only 27 consonant
phonemes in MA. The disagreement on the number of consonants is mainly due to the status of
the glottal stop and the primary emphatic consonants.

The consonantal inventory of CMA is summed up in the following chart:

-4-

15
 Labial Alveolar Alveo- Velar Uvular Pharyn- Glottal
palatal geal
t,d k, g
Stop b T, D kw, gw q, qw (ʔ)

s, z x, γ
Fricative f S, Z ʃ, ʒ xw, γw ħ, ʕ h

Nasal m n

l
Liquid r, R

Glide w y

Two remarks about the chart above are in order. First, the glottal stop [ʔ] is set between
parentheses to show that it is not part of the phonemic inventory of CMA (and all the varieties of
MA); it is prothesized before vowel-initial words for onset purposes. Second, the inventory
includes labialized dorsal consonants, a characteristic of Bedouin dialects of which CMA is one
member (See Aguade 1994 on the phenomenon of labialization). In previous work on MA,
labialization of dorsal consonants has always been considered as a case of dialectal variation (cf.
Harrell 1962, Youssi 1998 and Boudlal 1998 for the non-distinctiveness of rounding in certain
varieties of MA). While this is true, to some extent, in certain words in CMA, it is nonetheless
distinctive as will be seen below.
That labialized dorsals should figure in the phonemic inventory of CMA is corroborated
by items where labialization accompanies the word in its various realizations. Boudlal (1998)
refers to these as cases of lexical labialization. Consider the following examples for illustration:

-5-

Singular Plural Diminutive Gloss

kwərsi kwrasa kwrisi chair

gwəffa gwfaf gwfifa basket
qwənt qwnat qwniyyət corner
xwzana xwzayən xwzina/xwziyna tent
γwzala γwzalat γwzila/γwziyla gazelle

16
Given the data above, it would be inappropriate not to grant labialized dorsals a phonemic status.
Stating that labialized consonants in 5 are derived from their simple counterparts would simply
be ad hoc since we would have to claim that labialization is part of the morphemes denoting the
singular, the plural and the diminutive. In other words, for the morphological categories where
labialization is lexical, we would have to state that labialization is a morpheme associated with
that particular morphological category. Further evidence for considering labialized dorsals to be
underlying comes from doublets where a labialized dorsal stands in opposition to its non-
labialized counterpart. Consider the examples in 6 for illustration:

-6-

a. 3sg.Perfective 2sg.Imperative Gloss

xrəʒ xwrəʒ leave

dxəl dxwəl enter
tqəb tqwəb pierce
qtəl qwtəl murder
gʕəd gwʕəd sit down

b. Adj. (pl.) Verb

kwbar big kbar he grew up

qwdam old qdam he/it became old
rqwaq thin rqaq he/it became slim
rxwaS cheap rxaS it became cheap
γwlaD fat γlaD he/it became fat

c. kwħəl kohl kħəl black (mas. sg)

xwəDRa vegetables xəDRa5 green (fem. sg.)
gwəlsa site for selling st. gəlsa sitting
ʕgwər sterility ʕgər he became sterile

5
This doublet is taken from Youssi (1998: 207). In fact the author considers the non-labialization or the labialization
of [x] to be mere dialectal variation. While we believe that this is partly true of certain dialects, we do maintain that
in CMA, labialization may be distinctive.

17
In 6a, labialization contrasts the perfective and the imperative; in 6b, it distinguishes the plural
adjective and the verb; and finally in 6c, it contrasts the noun and the adjective in the first two
items, the noun and the noun in the third and the noun and the verb in the final. The substitution
of labialized consonants for the corresponding non-labialized will result in a change of meaning.
Thus both simple and labialized dorsals must be considered as separate phonemes in CMA .
At considering the data in 6, one may argue that labialization has a morphological status
since it serves to contrast different morphological classes such as the perfective and the
imperative (e.g. qtəl/qwtəl), the singular and the plural (e.g. kbir/kwbar, qdim/qwdam ...), the
adjective and the verb as in 6b, and therefore labialization should be a morpheme marking these
morphological categories as argued by El Medlaoui (1992) rather than lexically associated with
the dorsal consonants. If we assume that this is the case, we should expect, for example, the
plural morpheme to consist of the vocalic melody as well as the feature of rounding that
associates with dorsal consonants. However, this is not the case as shown in the examples below
taken from Boudlal (1998: 52):

-7-
Singular Plural Gloss

kəlb klab/*kwlab dog

gərba grabi/*gwrabi water container (made up out of goat skin)
qəhwa qhawi/*qwhawi cafe/coffee
xruf xərfan/*xwərfan sheep
γəmza γəmzat/*γwəmzat wink

If labialization were part of the plural morpheme, we would have first to explain why the
asterisked forms are ruled out even if they contain labialized dorsals. Second, we would have to
explain why it does not hook up to the initial dorsal consonants of the attested plural forms as is
the case with the diminutive forms where the feature causing labialization is assumed to be part
of the diminutive morpheme. (See chapter six for details on the labialization of dorsal and labial
consonants).
One additional remark about the consonantal inventory of CMA is that the labialized
dorsals make it look different from that established for AMA as described in Benhallam and

18
Dahbi (1990) and Youssi (1998) 6. This is no surprise given the fact that the consonantal system
Benhallam and Dahbi (1990) propose for AMA is intended for the purpose of describing all the
varieties of MA regardless of regional variation.

Having given the consonantal inventory of CMA, let us now turn to the vocalic inventory.
Generally, the previous works done on vowels agree on the fact that MA has three underlying
vowels which are /i, u, a/. The crucial issue, which is the point of divergence is related to the
status of the schwa. Should it be assigned a phonemic or a phonetic status ?

For most works on MA, the schwa is assumed to be purely phonetic and therefore
epenthetic (Benhallam 1980, 1988, 1990a, Marsil 1988, Al Ghadi 1990, El Himer 1991, Imouzaz
1991 and Boudlal 1993, among others). On the other hand, Benkaddour (1982) (see also Rguibi
1990, forthcoming), distinguishes two schwas: the phonemic schwa and the phonetic schwa.
According to him, the phonemic schwa serves as a morphological contrast between verbs of the
type /CCəC/ such as /DRəb/ “hit” and /lʕəb/ “play”, and nouns of the type /CəCC/ such as
/DəRb/ “hitting” and /ləʕb/ “play/game”. As to the phonetic schwa, its function is to break up a
three-consonant cluster that the language does not allow (e.g. /l-bnat/ ——> [ləbnat] “the girls”).
In the present work, we assume that the vocalic inventory of CMA consists of three
underlying vowels which are /i, u, a/ and an epenthetic schwa. This vowel system is given in 8
below. The schwa is enclosed between parentheses to denote its epenthetic status.

-8-

High i u

Mid (ə)

Low a

6
Youssi (1998) does not use the term AMA. But the consonantal inventory he gives corresponds exactly to that of
Benhallam and Dahbi (1990).

19
Unlike Benkaddour (1982) and Hammoumi (1988), we do not think that the language should
have long vowels since it is not distinctive. If length is used, it is simply for stylistic purposes
(Youssi 1986).
It should be noted that each of the four vowels in 8 has different phonetic variants
depending on the consonantal environments in which they occur. For this purpose, Youssi (1986)
groups the consonants of MA into three groups: simple consonants which are labials, alveolars
and palatals; back consonants which are velars, uvulars, pharyngeals and laryngeals; and finally
emphatic consonants. The phoneme /i/ appears as [i] in the environment of simple consonants, [ι]
next to back consonants and [ɨ] next to emphatics. The phoneme /u/ appears as [u] next to simple
consonants, [ɔ] next to back consonants and as [o] next to emphatics. The phoneme /a/ appears as
[æ] next to simple consonants, [ɑ] next to back consonants and [a] next to emphatics. Finally, the
schwa appears as [ə] next to simple consonants, [ɤ] next to back consonants and [ʌ] next to
emphatics.
In the absence of a truly experimental study of the phonetic variants of MA vowels, the
variants given above remain mere approximations and therefore will not be represented in the
present work.

5.2 Morphology

This subsection deals with some aspects of the prosodic morphology of CMA which we
judge need to be explored given the theoretical framework adopted in the present work. First, the
definition of the root-and-pattern morphology will be established. Second, examples of the
relevant derivational processes that will be dealt with in the core chapters will be given.

5.2.1 Root-and-Pattern Morphology

Most words in MA, and other Semitic languages, are built on a basic consonantal skeleton
called the root. This root occurs in patterns with different vowels to convey specific meanings.
For example, the root /ʃRb/, which has something to do with the concept of "drinking", has the
following patterns, among others:

20
-9-

ʃRəb he drunk
ʃRib drinking
ʃəRba one drink
ʃəRRəb cause to drink

Harrell (1962) distinguishes three basic root types in MA : triliteral, quadriliteral and atypical
roots 7. Triliteral roots are composed of three segments. The roots which have four constituent
elements are called quadriliteral. Roots with fewer than three or more than four segments are
called atypical. Triliteral and quadriliteral roots are further classified as strong or weak. Strong
roots are those composed entirely of consonants such as the root /ktb/ which is realized as [ktəb]
“he wrote”. On the other hand, weak verbs are those which have, at least, a glide as one of its
constituent elements such as /wSl/, /dwb/, /dwy/ realized respectively as [wSəl] “he arrived”,
[dəwwəb] “cause to dissolve”, and [dwi] “speak”8.
For the purpose of the present work, and following Al Ghadi (1990), the traditional terms
biliteral (or biconsonantal), triliteral (or triconsonantal) and quadriliteral (or quadriconsonantal)
will be replaced with the terms bisegmental, trisegmental and quadrisegmental (where segment
stands for vowels and consonants). The reason for such appelations lies in the fact that the roots
in MA may consist of both consonants and vowels, an issue that has been argued for in Al Ghadi
(1990), and other subsequent works such as Bennis (1992), Boudlal (1993) and Meliani (1994).

5.2.2 Relevant Morphological Processes

The central issue in the present work is to account, by means of constraints, for some
aspects of the phonology and prosodic morphology of CMA. The phonological aspects that will
be dealt with are mainly syllable structure and stress. As to morphology, the aspects that will be
dealt with include nisba adjectives derived from compound nouns, the causative, the passive

7
According to Harrell (1962), atypical verbs include verbs such as [dda] “he took” and [ʒa] “he arrived” which do
not conform to the general root-and-pattern system of the language.
8
It will be seen in chapters four and five below that verbs with high vocoids may be represented with a glide
underlyingly as is the case with /wSl/ or with a high vowel as is the case with /dub/.

21
participle and the diminutive. Representative examples of these aspects are given in the following
subsections.

5.2.2.1 The Nisba

The nisba formation is characterized by the suffixation of the morpheme [-i] to the base.
Consider the examples in 10 for illustration:

-10-

Base Nisba Gloss

a. fas fasi from Fès

məknas məknasi from Meknes
tunəs tunsi from Tunisia

b. ʕəbda ʕəbdi from the Plain Abda

xwribga xwribgi from Khouribga
taza tazi from Taza

c. sla slawi from Salé

sma smawi sky-blue
Səħra Səħrawi from the Sahara

When the base ends up in a consonant, nothing special happens after the affixation of the nisba
morpheme. However, when the base ends up in a vowel, two different processes could happen:
truncation or glide epenthesis. If the final vowel of the base is the feminine suffix, it gets
truncated after the suffixation of [-i] (10b). If this vowel is part of the base, the glide [w] is
epenthesized to serve as an onset to the nisba suffix (10c).
What interest us more in the present work are not items like those in 10 above but those
nisba adjectives derived from compound nouns and nouns with the affix [t-...-t]. Below we give
examples of these nisba adjectives:

-11-
Base Nisba Gloss
a. bni-məllal məllali from Beni-Mellal
DDar-(əl)biDa biDawi from Casablanca
wlad-ħəddu ħəddawi from Oulad Haddu

22
b. tarudant rudani from Taroudant
tafilalt filali from Tafilalt
taħənnawt ħənnawi from Tahannaout

The common characteristic among the nisba adjectives above is that they are all derived from
names of Moroccan localities. When the nisba affix is attached, part of the base is truncated: in
11a it is the left-hand member of the compound; in 11b it is the affix [ta-...-t].
This phenomenon of truncation is taken in chapter 4. In particular, we will try to answer
questions such as the following: What is it that conditions this truncation? Is the output of these
forms governed by some prosodic constraint delimiting its size and forcing deletion from the
base?

5.2.2.2 The Causative

The causative is formed by doubling the second segment of the base form as seen in 12
below:

-12-

Base Causative Base gloss

a. ktəb kəttəb write

wSəl wəSSəl arrive
dwa dəwwa speak
bka bəkka cry

b. dub dəwwəb dissolve

nuD nəwwəD stand up
fiq fəyyəq wake up
Tiħ Təyyəħ fall down

The causative has been analyzed as a case involving circumscription of a prosodic category
which is the minimal syllable (Bennis 1992). It involves the gemination of the second segment of
the base and operates at the left side of the minimal syllable as in 12a or at the right side thereof
as in 12b.

23
 In chapter four, this category will receive due consideration. In particular, we will point
out the limitations of a circumscriptive analysis and propose an alternative analysis couched in a
constraint-based framework.

5.2.2.3 The Passive Participle

The passive participle is obtained by the prefixation of [m-] and sometimes the infixation
of [-u-] to the base which is the verb as shown in the examples below:

-13-

Vb. Stem Passive P. Gloss

a. ktəb məktub written

DRəb məDRub hit
dir mədyur done
wləd məwlud born

b. kri məkri rent

səlləf msəlləf lent
kərkəb mkərkəb rolled
sali msali finished
surt msurət locked

The items in both 13a and 13b prefix the morpheme [m-] in the passive participle. The items in
13a proceed to further infixation of [-u-] before the final segment of the base.
In chapter four, we will decide on whether or not the prefinal vowel is part of the passive
participle morpheme. If it is part of the passive morpheme, we will have to explain why it does
not show up in items such as those in 13b. If it is not a morphemic vowel, we will have to explain
why it appears in 13a but not in 13b

5.2.2.4 The Diminutive

In most MA words, the diminutive is formed by the infixation of the segment [-i-] after
the second segment of the base, as seen in 14 below:

24
-14-

Base Diminutive Base gloss

a. kəlb kwliyyəb dog
qərd qwriyyəd monkey
bab bbwiyyəb door

b. wdən wdina ear

zit zwita oil
bənt bnita girl

c. kura kwira ball

TəbSil TbiSil plate
Siniyya Swiniyya tray

Three remarks need to be made about the items in 14. First, labial and dorsal consonants get
labialized in the diminutive, a fact which points out that labialization is part of the diminutive
morpheme. Second, the output of the diminutive consists minimally of two syllables and
maximally of three. Third, monosyllabic bases are augmented by the addition of [yyə] as in 14a
or the suffixation of [-a] as in 14b.
In chapter six, we will focus mainly on the augmented cases and show whether or not such
an augmentation is dictated by some prosodic requirement on the output. We will also try to
explain why in some cases augmentation is achieved by the addition of [yyə] while in others, it is
achieved by the suffixation of [-a].

6. THEORETICAL BACKGROUND

Research in Prosodic Morphology can be divided into two stages. The first stage started
with McCarthy (1979) who was the first to deal with languages having a non-concatenative
morphology within a non-linear theoretical framework. He proposed a prosodic theory of non-
concatenative morphology which was continuously revised until it developed into what came to
be known as Circumscriptive Prosodic Morphology. The second stage in the history of Prosodic
Morphology was characterized by a restatement of the principles of Prosodic Morphology within

25
Optimality Theory (McCarthy and Prince 1993a), thus paving the way to Correspondence Theory
(McCarthy and Prince 1995, 1999 and related work).
The theoretical framework we propose for the analysis of the phonological and
morphological problems raised above combines the proposals of two separate but closely related
theories: Optimality Theory and Correspondence Theory. In what follows, we present the
historical background that has led to the emergence of this Optimality Theory and
Correspondence Theory.

6.1 Prosodic Morphology Before Optimality Theory

Non-linear phonology came as a reaction against Chomsky and Halle’s (1968) position on
the linear nature of phonological representations. Prosodic Phonology, which is a sub-branch of
non-linear phonology, introduced notions such as the mora, the syllable and the foot. These
notions were later applied to the study of morphology and in the area where prosodic phonology
and morphology interact, thus giving rise to the theory of Prosodic Morphology.

6.1.1 Prosodic Morphology

With the emergence of Autosegmental Phonology (Goldsmith, 1976, 1979), the

characterization of non-concatenative morphology was made much easier than the boundary
apparatus utilized in Chomsky and Halle (1968). McCarthy (1979, 1981) applies the principles of
Autosegmental Phonology to the phenomena of discontinuous affixation in Semitic
Morphological systems, particularly that of CA. The adaptation of these principles to account for
languages having non-concatenative morphologies gave rise to the theory of Non-concatenative
Morphology (McCarthy, 1979, 1981, 1982), which later develops into Prosodic Morphology
(McCarthy and Prince 1986 et seq.).
The central concept involved in non-concatenative morphology is the ‘template’: a
skeletal morpheme that is segmentally unspecified except for information on the positions of
consonants and vowels. According to McCarthy (1979, 1981), morphemes are no longer
represented in a linear fashion as is the case with the standard theory of phonology embodied in
Chomsky and Halle (1968), but on different levels called tiers, all of which are linked to a

26
templatic skeleton. As CA is structured in such a way that the roots are made up solely of
consonants and the vocalism is provided by morphology, the mapping of different tiers to the
skeleton is ensured by a set of autosegmental principles (see McCarthy, 1981). In CA, the
consonants always associate with the C-slots while the vowels with the V-slots. For McCarthy
(1981), the skeleton in CA is found to be templatic in that it always corresponds to some
morphological meaning and is therefore morphemic. For example, the passive form [kutib] “it
was written” would be represented autosegmentally as in 15:

-15-
µ

Vocalic tier u i

Skeletal template CVCVC

| | |
Root tier k t b

The root provides the basic meaning of “write”, the template represents the perfective active and
the morpheme [u i] represents the passive.
Of close relevance to Templatic Morphology are two principles: the Obligatory Contour
Principle (OCP, Leben’s 1973) and Tier Conflation (TC, McCarthy 1986). The OCP prohibits
adjacent identical elements from occurring on the same tier. TC has the effect of conflating
different tiers into a linearized fashion.
In a later version of Prosodic Morphology, McCarthy and Prince (1986, 1990b) argue
against segmental templates in favor of prosodic ones. They claim that Prosodic Morphology
operates not with CV units but with authentic units of prosody such as the mora, the syllable, the
foot, and the prosodic word. They further argue that segmental templates are incapable of
specifying that certain elements in the template are obligatory while others may be optional. The
new version of Prosodic Morphology is later developed into the theory of Prosodic
Circumscription (McCarthy and Prince 1990a; Lombardi and McCarthy 1991).

6.1.2 Circumscription within Prosodic Morphology

27
 The theory of Prosodic Morphology, developed in McCarthy and Prince (1990a), is based
on three main theses which are stated as follows:

-16-

a. Prosodic Morphology Hypothesis (PMH)

Templates are defined in terms of authentic units of prosody : mora (µ), syllable (σ),
foot (Ft), prosodic word (PWd), and so on.

b. Template Satisfaction Condition (TSC)

Satisfaction of templatic constraints is obligatory and is determined by the principles
of prosody, both universal and language-specific.

c. Prosodic Circumscription of Domains (PCD)

The domain to which morphological operations apply may be circumscribed by
prosodic criteria as well as by the more familiar morphological ones. In particular,
the minimal word within a domain may be selected as the locus of morphological
transformation in lieu of the whole domain.

The first thesis states that templates in Prosodic Morphology are defined in terms of the authentic
units of prosody and not in terms of CV units. Following Selkirk (1980), McCarthy and Prince
(1993: 2) maintain that the authentic prosodic units are defined in a hierarchical way and in terms
of dominance. Thus the prosodic word dominates the foot, and the foot dominates the syllable
which, in turn dominates the mora. Each of these prosodic categories is defined in terms of the
lower one in the hierarchy. The mora, being the lowest unit, serves to determine syllable weight.
A light syllable consists of one mora [σµ], a heavy two moras [σµµ]. The syllable is the unit that
bridges two levels; the moraic level and the foot level. The foot according to McCarthy and
Prince (1986) is assumed to be governed by a constraint which requires that it be binary under
syllabic or moraic analysis. Together the prosodic hierarchy and the foot binarity constraint
derive the minimal word. The notion “minimal word” corresponds to “minimal foot” and is found
to play a major role in prosodic morphology. For example, the minimal word in CA is an iambic
foot, that is a sequence of light-heavy syllables, or light-light syllables or simply a heavy syllable.
The second thesis in 16 requires that all the elements of a template be obligatorily
satisfied by virtue of TSC. Under this condition, no morphological template is allowed to contain
excess material which presents serious mapping problems to segmental theories (for the
statements of these problems, see McCarthy and Prince, 1986). At the same time, the theory of

28
Prosodic Morphology allows the presence of optional material which is governed by both
universal and language-specific characterization of the prosodic units constituting the template.
The third thesis in 16 demands that morphological operations be circumscribed by
prosodic criteria as well as morphological ones. Central to prosodic circumscription is a parsing
function F which locates a prosodically-delimited domain of application of morphological rule
which is smaller than the base. According to McCarthy and Prince (1990a), circumscription can
be either negative or positive. In negative circumscription, some prosodic constituent C at the
edge E (left/right edges) of a form is disregarded (set as extrametrical) and the morphological
operation O applies to the remainder. The formula O/F (C, E) is written to denote the application
of O to a particular form minus the constituent C (the residue) parsed out at edge E by the
parsing function F.
In positive circumscription, the prosodically-delimited constituent at the edge serves itself
as the base of morphological operation. In this case, the formula O/F (C, E) is written to denote
the application of O to the constituent C parsed out at edge E by F.
In CA, for example, the domain of the plural is assumed to be a circumscribed minimal
word which consists of two moras (McCarthy and Prince 1990a). Adopting McCarthy and
Prince’s circumscriptive model, the derivation of the plural form [qanaafid] from the singular
[qunfud] ‘hedge hog’ proceeds by scanning the base [qunfud] into a minimal word which is [qun]
and a residue which is [fud]. The circumscribed domain which is [qun] is then mapped onto an
iambic foot. The whole picture is represented as follows:

-17-

a. Mapping of circumscribed [qun] onto an iambic foot

Ft

σ σ

µ µ µ

q n

29
b. Association of first vowel of the plural melody /a i/

Ft

σ σ

µ µ µ

q n

a i

c. Reestablishing the residue and association of the second vowel of the plural melody /a i/

Ft

σ σ σ

µ µ µ µ (µ)

q n f d

a i

As seen above, the morphological operation O proceeds by mapping the minimal word into an
iambic foot. The result of the mapping is concatenated with the residue, which is the final heavy
syllable of [qunfud]. The vowels that associate to the moraic positions are those of the plural
morpheme [a i]. They overwrite the vowels of the singular base.
However, with the recent developments in phonological theory, particularly, with the
emergence of constraints embodied in Optimality Theory (Prince and Smolensky 1993),
McCarthy and Prince (1993a) reformulate Prosodic Morphology in such a way that it is
conceived of as a theory of constraint interaction. In the following section, we expose the basic
tenets of Optimality Theory and see how the principles of the new version of Prosodic
Morphology will fit into the Optimality Program.

6.2 Optimality Theory

6.2.1 Introduction

30
 Optimality Theory differs from earlier phonological theories in at least two aspects. First,
previous works in phonology assume that the task of a phonological theory is to define the
underlying form (input) and the surface form (output) of a linguistic object. The matching
between the input and the output is achieved via phonological rules. However, with the rise of
Prosodic Morphology, the form of morphemes has been shown to be largely governed by
constraints on the well-formedness of the output. Parallel developments in phonological theory
were very important to the emergence of OT (Prince and Smolensky, 1991, 1992, 1993,
McCarthy and Prince 1993a). OT abandons the idea that the input-output matching is
accomplished by rules. In place of this is the function Gen (standing for Generator) which
produces for any given input a large number of candidate analyses.
The second distinguishing aspect of OT is that it abandons the widely held view that
constraints are language-particular statements of phonotactic truth. Instead of this is the claim
that constraints are essentially universal and of general formulation, with a great potential for
disagreement among the well-formedness of analysis. The redefinition of constraints on universal
grounds does not deny the role individual grammars play in phonological analysis. Prince and
Smolensky (1993) hold that the role of an individual grammar consists in the ranking of universal
constraints.
Four basic assumptions underlie OT. These are stated in 18 below:

-18-

a. violability
b. Ranking
c. Inclusiveness
d. Parallelism

The first assumption holds that constraints in OT are violable, but this violation should be
minimal. Minimal violation according to McCarthy and Prince (1993) is defined in terms of the
ranking of constraints. In OT, the optimal form is selected by a set of well-formedness constraints
ranked in a hierarchy of relevance, so that a lower-ranking constraint may be violated to secure a
higher-ranking one. Inclusiveness means that the candidate analyses are generated by general

31
considerations of structural well-formedness. No rules or repair strategies are admitted, contra to
other constraint-based theories (LaCharité and Paradis, 1993). The fourth assumption of OT,
parallelism, means that there is no serial derivation; all the possible candidate analyses produced
are evaluated according to the constraint hierarchy. The candidate that passes the higher-ranking
constraints is the output form.

6.2.2 The Construction of Grammar in OT

In OT, grammar must define a parsing of input and output forms. For any given input, the
function Gen which is part of Universal Grammar, produces a large set of candidate analyses
associated with that input. For example, Gen involves the construction of many different prosodic
parses or many different linear arrangements of morphemes. The function Eval (standing for
Evaluator) determines the relational harmony of the candidates, ordering them by how well they
best-satisfy the constraint system of the language. An optimal candidate is at the top of the
harmonic order of the candidate set. According to Prince and Smolensky (1993), grammar in OT
can be schematized as follows:

-19-

Gen (inputi) = {cand1, cand2...candn}

Eval ({cand1, cand2...candn}) --> candk (real output)

As could be seen in 19, the candidate analyses are supplied by Gen. For Prince and Smolensky,
three basic principles underlie the function Gen :

-20-

a. Freedom of analysis
b. Containment
c. Consistency of exponence

32
Freedom of analysis means that Gen can supply candidates with moraic, syllabic or other
prosodic structure, and with additional segmental material, ranging from empty nodes to fully-
specified vowels or consonants. No rules or repair strategies need to be posited.
Containment demands that the input must be present in any possible candidate form. For
example, containment means that segmental deletion phenomena such as [k] deletion in English
“know” should be looked at as a case of underparsing (“<k>now”, where < > enclose
underparsed material). Under containment, phonologically deleted segments are present in the
output but unparsed syllabically. In later development within the OT framework (McCarthy 1995
et seq.), the principle of containment was abandoned in favor of a more general way of
faithfulness regulating the relation between the input and the output.
Consistency of exponence is a hypothesis about the phonology-morphology interface. It
means that the lexical specifications of a morpheme can never be affected by Gen. In other
words, it demands that the phonological exponents of any given morpheme be identical in
underlying and surface forms.
The central proposal in OT is that constraints are violable and are ranked in a hierarchy of
relevance. The output candidates produced by Gen are evaluated according to a set of
hierarchically ranked constraints ( 1 >> 2 >> ... n, where the symbol “>>” shows domination
relation) each of which may eliminate some output candidates. The elimination process in OT is
schematized below:

-21-

Mapping of input to output in OT grammar (Kager 1999: 8)

1 >> 2 >> n
Candidate a ——> ——> ——>
Input Candidate b ——>
Candidate c ——> ——>
Candidate d ——> ——> ——> ——> Output
Candidate ... ——> ——>

The function Eval proceeds by evaluating all the possible candidates and then chooses the one
that is most harmonic with respect to the set of ranked constraints, i.e. the real output (optimal)
candidate.

33
 Closely related to constraint violability is the notion of constraint conflict. To best-
exemplify this notion, consider the following example taken from McCarthy and Prince (1993a).
Assume that a grammar consists of two constraints : constraint A and constraint B. Assume
further that Gen generates cand1 and cand2 from inputi. If A and B disagree, we say that we have
a constraint conflict. This conflict is represented in the constraint tableau in 22 below. Violations
are marked by an asterisk and a fatal violation is marked by !; the optimal candidate is marked by
).

-22-

Candidates A B
) Cand1 *
Cand2 *!

Cand1 meets constraint A but fails constraint B, whereas Cand2 meets constraint B but fails
constraint A. Assuming that Cand1 is the optimal output, the grammar requires that A dominates
B (written as A>>B).
The basic tenets of OT have been applied to the domain of Prosodic Morphology, firmly
establishing what is known as the OT-based Prosodic Morphology.

6.2.3 Prosodic Morphology within Optimality Theory

Prosodic Morphology within OT (PMOT) is a theory about prosody-morphology

interface. It is based on the idea that patterns of reduplication, root-and-pattern morphology, and
infixation and related phenomena result from a combination of independent and general
constraints, ranked under OT. The basic tenets of this theory are stated in McCarthy and Prince
(1993a: 103) as follows:

-23-

a. Prosodic Morphology Hypothesis (PMH)

Templates are constraints on the prosody/morphology interface, asserting the
coincidence of morphological and prosodic constraints.

34
b. Template Satisfaction Condition
Templatic constraints may be undominated, in which case they are fully satisfied, or they
may be dominated, in which case they are violated minimally, in accordance with the
general principles of Optimality Theory.

c. Ranking Schema
P>>M

The first thesis means that templates should be conceived of as constraints on the interaction
between prosody and morphology. In the standard version of the PMH (cf. 23a above), templates
are said to consist of authentic units of prosody. However, in PMOT, they are a particular kind of
constraint of the large ALIGN family, asserting the coincidence of morphological and prosodic
constituents or their edges.
The second thesis of PMOT is also a revision of the TSC expressed in the standard theory
of Prosodic Morphology (cf. 23b above). The new formulation of TSC states that templates may
be undominated, which means that they are obligatorily satisfied; or they may be dominated,
which means that they are violated. However, if they are ever violated, the violation should be
minimal, i.e. to secure higher-ranking constraints.
The ranking schema P>>M means that if some morphological domain is to be
prosodically conditioned, then in that domain P>>M, that is prosody dominates morphology.
In subsequent work within the OT framework, McCarthy and Prince (1993b) claim that
the constraints on the interface between prosody and morphology are of the general form
ALIGN, which requires that the edge of any grammatical category (G Cat) align with the
corresponding edge of some prosodic constituent (P Cat). In this view, templatic categories such
as the “Minimal Word” are no longer needed since their effect can be derived from constraint
interaction. McCarthy and Prince expand on Prince and Smolensky’s (1993) model of constraints
on faithfulness of the output to the input, and also on McCarthy and Prince’s (1993a) model of
constraints on identity between the base and reduplicant and propose a theory unifying both
faithfulness and identity - Correspondence Theory (1995, 1999).

6.2.4 Correspondence in Optimality Theory

35
 In OT, grammar is defined as a set of constraints belonging to UG and ranked on a
language-particular basis. These universal constraints involve two types: markedness constraints
and faithfulness constraints. Markedness constraints as McCarthy (1997) puts it, militate against
structural elaboration of various kinds while the antagonistic faithfulness constraints demand
identity of linguistically related forms.
Correspondence Theory treats identity between the base and the reduplicant like
faithfulness of the output to the input. Faithfulness and identity follow from the same kind of
formal constraints on the correspondence relation between representations. According to
McCarthy and Prince (1995: 15), correspondence is a function formally viewed as follows:

-24-

Correspondence
Given two strings S1 and S2, correspondence is a relation R from the elements of S1 to
those of S2. Elements α ∈ S1 and β ∈ S2 are referred to as correspondents of one another
when α R β.

Correspondence is at minimum a relationship between segments. For McCarthy (1995) it can be

extended to features and prosodic units such as moras, syllables, feet, heads of feet, as well as
tones and distinctive features and feature nodes. Furthermore, correspondence need not be
limited to base and reduplicant and input and output relations; it can also be extended to cover
cyclic and truncatory phenomena (Benua 1995, 1997), circumscriptional phenomena (McCarthy
1997) and relations between separate words (Benua 1995, 1997, Burzio 1996, Kenstowicz 1996,
1997, Kager 1996, Basri et al 1998, Selkirk 1999). (See chapter four for an extended version of
CT and an illustration of the areas it covers)
Some of the constraint families on correspondent elements are given in 25 below. These
constraints refer to a pair of representations (S1, S2) standing to each other as input/output (I-O),
base/reduplicant (B-R) and output/output (O-O). The constraints also refer to the correspondence
relation (R) defined for the representation being compared:

-25-

Constraints on the correspondent elements (McCarthy 1995: 123-125)

36
a. MAXIMALITY (MAX):
Every element of S1 has a correspondent in S2.
Domain (R)= S1

b. DEPENDENCE (DEP)
Every element of S2 has a correspondent in S1.
Range (R)= S2

c. IDENTITY-[F] (IDENT-F):
Correspondent segments in S1 and S2 have identical values for feature [F]
If x R y and x id [γF], then y is [γF]

d. CONTIGUITY (CONTIG)
i. I-CONTIG (No Skipping)
The portion of S1 standing in correspondence forms a contiguous string.
Domain (R) is a single contiguous string S1.

ii. O-CONTIG (No Intrusion)

The portion of S2 standing in correspondence forms a contiguous string.
Range (R) is a single contiguous string S2.

e. {RIGHT, LEFT}-ANCHOR (S1, S2)

Any element at the designated periphery of S1 has a correspondent at the designated
periphery of S2.

f. LINEARITY (No Metathesis)

S1 is consistent with the precedence structure of S2, and vice versa.
Let x, y ∈ S1 and x’, y’ ∈ S2.
If x R x’ and y R y’, then
x<y iff ¬ (y’ < x’)

g. UNIFORMITY (No Coalescence)

No element of S2 has multiple correspondents in S1.
For x, y ∈ S1 and z ∈ S2, if x R z and y R z, then x = y.

h. INTEGRITY (No Breaking)

No element of S1 has multiple correspondents in S2.
For x ∈ S1 and w, z ∈ S2, if x R w and x R z, then w = z.

The MAX constraint family is a reformulation of PARSE-seg connected to syllabification

and phonetic interpretation in Prince and Smolensky (1993). It prohibits phonological deletion,

37
demands complete copying of base in B-R relation and completes I-O mapping in root-and-
pattern morphology.
The DEP constraint family is also a reformulation of FILL in the containment model of
Prince and Smolensky (1993). It subsumes the anti-epenthesis effect of FILL without requiring
that epenthetic segments be literally unfilled positions whose content is specified by phonetics.
The IDENT constraint family replaces PARSE-feature and FILL-feature-node in Prince
and Smolensky (1993). It requires that correspondent segments be featurally identical to one
another. When undominated, IDENT requires complete featural identity between correspondent
elements. Alteration in featural identity arises only when this constraint is outranked by another.
To exemplify the constraints MAX, DEP, and IDENT, consider the following tableau
adapted from Kager and Zonneveld (1999: 15):

-26-

S1 S2 MAX DEP IDENT[high]

a. a1 b2 i3 a1 b2 i3
b. a1 b2 i3 a1 b2 *
c. a1 b2 i3 a1 b2 i3 m *
d. a1 b2 i3 a1 b2 e3 *

In 26a S1 is identical to S2 and therefore the three constraints are satisfied. In 26b, the final
segment of S1 is deleted in S2. This deletion results in a clear violation of MAX. In 26c, the
underlined segment in S2 does not figure in S1 and as such S2 incurs a violation mark of DEP.
Finally in 26c, the identity of the vowel [i] in S1 has been altered, causing a violation of the
constraint IDENT [high].
The constraint family CONTIG characterizes two types of contiguity: I-CONTIG and O-
CONTIG. I-CONTIG rules out deletion of elements internal to the input string (e.g. xyz ——>
xz). O-CONTIG rules out internal epenthesis (e.g. xz ——> xyz).

38
 The ANCHOR constraint family subsumes Generalized Alignment (McCarthy and Prince
1993b) and captures the effect of the constraint ALIGN (GCat, Left/Right, PCat, Left/Right). It
can be extended to cover two prosodic categories such as the foot and the head syllable in the
same foot. McCarthy and Prince (1999) cite the example of the foot (bí.ta) in which the left edge
of the foot anchors with the left edge of the head syllable.
LINEARITY simply preserves the linear order of elements in the input or the base.
McCarthy (1995) cites the example of Rotuman whereby the final two segments of the input
/pure/ metathesize in the incomplete phrase [puer], thus causing violation of the constraint
LINEARITY.

UNIFORMITY and INTEGRITY rule out two types of multiple correspondence.

UNIFORMITY is violated when two elements of S1 are fused in S2. INTEGRITY is violated
when one element of S1 is split or cloned in S2.

To sum up, this model of CT formulated within the OT framework seems to open up new
horizons for the understanding of the interaction of prosody and morphology. It is this general
framework that is adopted for the analysis of some aspects of the phonology and prosodic
morphology of CMA.

7. CONCLUSION

This chapter has reviewed some aspects of the phonology and morphology of CMA as
well as the theoretical framework that will be adopted for the analysis of these aspects.
After a brief review of previous works on MA, we have presented the phoneme
inventories of the variety under study. We have shown that while CMA shares common
characteristics with other varieties of MA, it differs from them by the number of consonant
phonemes it consists of. In particular, we have shown, based on the concept of opposition, that
the consonant inventory of CMA should also comprise the labialized dorsals /kw, gw, xw, γw, qw/.
In morphology, we have listed some representative examples of the morphological categories
which will be studied in depth in the core chapters of the present work. These categories include
the nisba adjective derived from compound nouns and names of localities with the affix [ta-...-t],

39
the causative, the passive participle and the diminutive. Finally, we have stated the basic
principles of Optimality Theory and Corresponding Theory which constitute the theoretical
framework that will be adopted for the analysis of the aspects mentioned above.

40
 Chapter Two

The Syllable Structure of Casablanca

Moroccan Arabic

1. INTRODUCTION

CMA has three underlying vowels which are [i, u, a] and an epenthetic schwa. The schwa
is epenthesized to break up consonantal clusters that the language does not allow. While
underlying vowels are not subject to any restrictions on syllable structure, schwas are problematic
in a number of respects. First, unlike full vowels which can occur in both open and closed
syllables, schwas never occur in open syllables. Second, the rule which epenthesizes a schwa has
to refer to the syntactic category of the base. Thus the way schwas behave in verbs and
adjectives, for example, is different from the way they behave in nouns. While the schwas
occurring in verbs and adjectives can be accounted for by a structure-building algorithm of
syllabification, nominal schwa epenthesis is dependent on the sonority of the consonants of the
base. Third, in order to derive the correct output, schwa epenthesis should be allowed to apply
cyclically in verbs and adjective and post-cyclically in nouns.
In this chapter, we argue that the schwa problems cited above, and consequently CMA
syllable structure can be accounted for adequately within OT as developed by Prince and
Smolensky (1993) and extended in CT by McCarthy and Prince (1995, 1999). In particular, we
will show that structural constraints such as the constraints requiring syllables to have onsets and
no codas, and faithfulness constraints regulating the relationship between the input and the output
along with other constraints, are what we need in order to account for CMA syllable structure.
We will also show that it is the ranking of these constraints determines syllabic well-formedness.
The chapter is organized into two major sections. In the first section, we present a critical
review of the previous accounts of MA syllable structure within a non-linear framework as
proposed in Benhallam (1990a) and Al Ghadi (1990). In the second section, we argue that a
constraint-based framework such as OT is far better than a rule-based one. Therein, we show,
following a proposal made in Al Ghadi (1994), the mechanism CMA resorts to in order to derive
the minimal prosodic word. Also in this section, the representation of geminates is raised in
relation to prosodic structure. Here we argue that prosodic minimality in non-derived words
containing geminates is achieved in the same way as other words which lack geminates.
Furthermore, we raise the question of cyclic syllabification in verbs and point to directions on
how to solve this problem. Finally we propose to reanalyze nominal schwa syllabification by
making use of a set of universal constraints which show that schwa syllables prefer a coda with a
high sonority. Throughout this chapter, we argue that CMA derives syllabic well-formedness
from the interaction of constraints pertaining to Universal Grammar.

2. PREVIOUS ACCOUNT OF MOROCCAN ARABIC SYLLABLE STRUCTURE

As it has been stated above, the vowel inventory comprises the three basic vowels [i, u, a]
and the epenthetic schwa [ə]. To understand the behavior of the schwa in CMA, one has to have
recourse to syllable structure. One of the most elaborate and frequently cited work on MA
syllable structure is that of Benhallam (1990a). Benhallam distinguishes two types of
syllabification in MA: full-vowel syllabification [i, u, a], and schwa syllabification. The author
proposes a Syllable Structure Assignment Algorithm (SSAA) which proceeds from right to left as
follows:

-1-
a. Onset and rime rule

σ σ

O RO R

N N

e.g. C V C V C ---------> C V C V C
| | | | | | | | | |
ʕ a z i z ʕ a z i z

42
 b. Assignment of a CvC syllable structure to every sequence of unsyllabified CC starting
from right to left. The v is to be interpreted as a schwa 1:

O R

N Cd
| |
e.g. C C C ---------> C C v C
| | | | | | |
k t b k t ə b

c. Assignment of a non-syllabified C to the coda position of a codaless syllable.

σ σ

O R O R
|
N N Cd
| | |
e.g. C V C −−−> CV C
| | | | | |
m a t m a t

d. Assignment of a stray C as premargin or postmargin to a following onset or a

preceding coda:

σ σ

O R O R

N Cd N Cd
| | |
e.g. C C v C C --------> C C v C C
| | | | | | | | | |
k t ə b t k t ə b t

1
I am using v instead of Benhallam’s dummy symbol ∆. The lower case v’s should be distinguished from the upper
case V’s. The former refer to the vocalic positions that are interpreted as schwas, whereas the latter refer to the full
vowels [a, u, i].

43
As we can see in 2 the SSAA proceeds by assigning CV core syllables (where V is one of the full
vowels [i, u, a]). It is only after this stage that schwa syllables are built. What the schwa
syllabification rule in 2b basically does is that it takes every unsyllabified CC sequence and
assigns to it the syllabic shape CvC. In other words, it creates what Selkirk (1981) calls
degenerate syllables whose nucleus we note as v. At a later stage, the v’s are filled with schwas
(Benhallam 1988, 1990a).
The SSAA accounts for a large number of items in MA. To start with consider the non-
derived trisegmental verbs in 2 below.

-2-
Vb root Vb stem Gloss

ktb ktəb write

ʃTħ ʃTəħ dance
DRb DRəb hit
gls gləs sit down

Basically all the non-derived trisegmental sound verbs are derived in the same way. What we
need are just rules 1b and 1d. The first rule creates a nucleus whose onset is the second consonant
of the root and coda is the third consonant. The second rule adjoins the first consonant of the root
as onset to the syllable created by the previous rule, thus creating a branching onset.
Non-derived adjectives and a large number of non-derived nouns can be obtained much in
the same way as the items in 2. Consider the examples in 3 below:

-3-

a. Adjectives
Root Stem Gloss

ʕrʒ ʕrəʒ lame

ħwl ħwəl cross-eyed
kħl kħəl black
SfR SfəR yellow
byD byəD white

44
b. Nouns
ktf ktəf shoulder
ʒml ʒməl camel
smn smən preserved butter
ʒbl ʒbəl mountain

What the examples in 2 and 3 show is that any /CCC/ sequence is syllabified as CCəC, exactly as
predicted by Benhallam’s SSAA. The algorithm in 1 also accounts and in a nice fashion for non-
derived quadrisegmental verbs (4a) and nouns (4b):

-4-

Root Stem Gloss

a. TRʒm TəRʒəm translate

frgʕ fərgəʕ explode
SRfq SəRfəq slap
krkb kərkəb roll

b. mslm məsləm Muslim

fndq fəndəq hotel
tnbr tənbər stamp

Given a sequence such /CCCC/, rule 1b will apply to give a disyllabic word of the type CəCCəC.
However, there are items that cannot be syllabified by the SSAA. Consider some representative
examples below:

-5-

a. Verbs
Root Stem Gloss

ʕDD ʕəDD bite

mdd mədd give
mss məss touch
ʒRR ʒəRR pull

b. Nouns
i. DRb DəRb hitting
lʕb ləʕb game
brd bərd cold

45
ii. srʒ sərʒ saddle
dnb dənb sin
frx fərx bird

The items in 5a represent the class of geminated verbs. Those in 5b represent the class of non-
derived trisegmental nouns, with the specification that the items in 5b.i have their corresponding
verbs where the schwa appears in a different environment (cf. [DRəb], [lʕəb] and [brəd]). These
items are counterexamples to Benhallam’s algorithm since the schwa is placed between the first
and second consonants and not between the second and third as predicted by 1b. To solve this
paradox, Benhallam (1990a), who was very much concerned with treating all the forms uniformly
regardless of their syntactic category, assumes that items having the structure CəCC as in 5b have
an underlying syllabic template which distinguishes them from the other forms syllabified by the
SSAA. The underlying template looks like the one in 6 below:

-6-
σ

O R

N Cd
|
Cv CC
| | | |
DəRb

This template accounts not only for forms on the pattern CəCC but also for roots whose second
consonant is geminated as could be seen from the structure of the verb [mədd].

-7-
σ

O R

N Cd
|
Cv CC
| |
mə d

46
Notice from the structure in 7 above that in the case of geminates, the only position of the schwa
is between the first consonant and the second one (that is the geminated consonant which is [d])
and this in conformity with the OCP (McCarthy 1986) which prohibits two identical segments
from occurring on the same tier. (See Benhallam 1991, and Rguibi 1990 on the OCP effect in the
treatment of geminates in MA)
Furthermore, the algorithm in 1 can adequately account for affixed items such as the
following:

-8-
a. DRb-at DəRbat she hit
lʕb-na lʕəbna we played

b. ktb-t ktəbt I wrote

ʒRʒR-u ʒəRʒRu they trailed

Assuming that syllabification applies after all the morphological rules have applied will yield the
correct output in 8a. First core syllables are formed, giving [DR.ba.t] and [lʕb.na.] (The periods
mark syllable edges). Only after that does schwa epenthesis apply to yield [DəRbat] and [lʕəbna].
The same assumption does not work for the items in 8b since a non-cyclic syllabification would
give the unattested forms *[ʒRəRʒru] and *[kətbət] 2. For Benhallam, syllabification should be
allowed to apply cyclically to yield the attested output. In the items in 8b above syllabification
applies as in 9 below. The brackets stand for cycles.

-9-

Input [[ktb]t] [[ʒRʒR]u]

First cycle
Syllabification ktəb ʒəR.ʒəR
Second Cycle
Affixation ktəb.-t ʒəR.ʒəR.-u

2
The form [kətbət] is attested in a number of varieties of MA but with the meaning “she wrote”. Following a
suggestion made to me by Selkirk (p.c.), I assume that the input of such a form is /ktb-Vt/ and that the V may be
realized as either [ə] or [a], depending on the variety of MA under study.

47
Syllabification ktəbt ʒəR.ʒə.Ru
Output [ktəbt] [ʒəRʒRu]

Syllabification first applies to the innermost bracketed items in the first cycle. It reapplies in the
second cycle after affixation to adjoin the stranded [t] as a postmargin to the preceding syllable,
thus giving the correct output [ktəbt]. In the remaining item, and after the suffixation of [-u] the
consonant [r] , syllabified as a coda in the first cycle, is assigned as an onset to this suffix and this
follows from the fact that MA does not allow onsetless syllables. Consequently, the schwa is left
in an open syllable, thus subject to deletion.
Although the SSAA seems to account for a large number of items in MA, it is
questionable on theoretical and empirical grounds. First, the extrinsic ordering of some
syllabification rules is established to avoid the generation of ungrammatical forms. If we order
rule 1c before rule 1b, we would, for example, get the ungrammatical item *[surt] instead of the
correct [surət] ‘to lock’. Moreover, an input form has to go through different stages before it
reaches the final stage of phonetic interpretation. (See for example the derivations in 9 above). In
section 3 a constraint-based analysis is proposed within the OT framework. Instead of the step-
by-step SSAA, OT accounts for schwa occurrences in terms of constraints pertaining to UG
which apply in a non-serialist way.
Second, the SSAA has to have recourse to the notion of directionality and the notion of
cyclicity to derive the correct output (see 9 above). The analysis proposed in this work derives
directionality and cyclicity of syllabification from the interaction of constraints requiring that
some edge of the output coincide with that of a prosodic category and other constraints requiring
that the derived output be faithful to the base (See chapter four for details).
Third, the SSAA does not give much detail about the nature of the rules themselves and
how they relate to other (similar) rules in natural languages. For example, it does not show that 1a
derives from the fact that CV core syllables are basic cross-linguistically, and that such syllables
have an obligatory onset. Nor does the SSAA make it clear that 1b derives from a universal
constraint, namely that segments must be parsed into some syllable. Within OT, 1a and 1b derive
from two different constraints: ONSET, which demands that syllables have an onset, and
PARSE-seg, which demands that all segments of the input belong to a syllable. The fact that all

48
MA syllables must have an onset and that consonants must belong to a syllable is achieved by
ranking the two constraints at the top of the ranking scale.
Fourth, the Algorithm distinguishes between two modes of syllabification: derived and
underlying. Derived syllabification is accounted for by the SSAA. As to underlying
syllabification, it is accounted for by the template in 6 and is needed especially for items having
the pattern CəCC which do not abide by the SSAA. Some of these items are given in 10a. The
items in 10b and 10c are intended for comparison; further examples are given in section 3.6
below:

-10-

a. dənb sin
DəRb hitting
bənt girl

b. rʒəl leg
qfəz cage
ʕsəl honey

c. smən preserved butter

nməl ants
gməl lice

According to Benhallam (1990a), the schwas in 10b and 10c are regular because they meet the
environment predicted by the SSAA. Contrariwise, the schwas in 10a are exceptions and
therefore should have an underlying syllable template like the one in 6 above. This means that
nouns on the pattern CəCC and CCəC belong to different classes, something which cannot be
justified on independent grounds. (See Al Ghadi 1990 for the arguments presented in favor of
considering the two patterns to belong to the same class, namely that of non-derived trisegmental
nouns).
To account for items like the ones in 10, Hammoumi (1988) proposes that the placement
of the schwa is determined by the degree of dissimilarity between the sonority of the second and
third consonants in a non-derived trisegmental noun or adjective. However, and as it has been
pointed out only nouns abide by the sonority constraints, whereas adjectives and verbs do not and

49
as such we are forced to abandon Hammoumi’s assumption in search of a more elaborate analysis
of the cases in 10.
A better solution is proposed by Al Ghadi (1990) who maintains that schwa epenthesis in
items such as those in 10a and consequently nominal schwa syllabification is to a large extent
dependent on the sonority of the consonants occupying the second and third positions in
trisegmental nouns. According to the author the schwa is epenthesized before the most sonorous
consonant. If the consonants in question have the same sonority3, the schwa is epenthesized
before the third consonant. Al Ghadi’s findings are reproduced below, where |C| stands for the
relative sonority of C.

-11-
In nonderived trisegmental nouns, a schwa is epenthesized in the following
environments

a. C1əC2C3, if |C2| > |C3|, e.g. 10.a.

b. C1C2əC3, if |C3| > |C2|, e.g. 10.b
c. C1C2əC3, if |C2| = |C3|, e.g. 10.c

It should be noted that there are some exceptional nouns that do not conform to the sonority
hierarchy (see Benhallam 1980, for a list of these items). Examples of such nouns include items
like [ʕməʃ] “sleep”, [ħəbs] “jail”, [ħməd] “Ahmed (proper noun)”and [ħnəʃ] “snake”.
Surprisingly enough, these items include a pharyngeal as one of their elements. All in all, we
believe, following Al Ghadi (1990), that a large number of nouns abide by the sonority principle.
Relative sonority derives from the universal theory of syllable structure. That the schwa
is epenthesized before the most sonorous consonant in a CCC sequence is not specific to MA but
it is found cross-linguistically, something that 11 above cannot predict.
OT offers a way to account for the regular cases of schwa epenthesis, that is cases which
can be handled by the SSAA in 1 by having recourse to universal constraints instead of language-
particular rules. It also offers a straightforward analysis for the exceptional items on the pattern

3
Al Ghadi (1990) assumes that the nasals and the liquids have the same sonority and that the class of glides
comprises, in addition to [w] and [y], the pharyngeals [ħ] and [ʕ].

50
CəCC to the effect that they are subject to a universal constraint, namely the Sonority Hierarchy
Constraint. (See section 3.6 for details)
To sum up, it has been pointed out that an OT approach, based on universal constraints
such as the one requiring that segments be grouped into syllables and that these syllables have
onsets obviates the need for a rule-based algorithm of syllabification. An OT approach has an
explanatory power since it derives syllabic well-formedness from constraints pertaining to UG.

3. CASABLANCA MOROCCAN ARABIC SYLLABLE STRUCTURE AND

OPTIMALITY THEORY
In this section we show how OT offers a better analysis of CMA syllabification based on
constraints that pertain to UG. More specifically, we intend to show that the structural constraints
ONSET and NO-CODA and the faithfulness constraints PARSE and FILL (revised by McCarthy
and Prince 1995) along with other constraints on sonority and directionality of syllabification will
allow us to capture significant generalizations about CMA syllable structure.
The following assumptions are maintained from previous analyses of MA syllable
structure:

(i) All cases of schwa in MA are epenthetic (Benhallam 1980, 1988, 1990a).
(ii) MA distinguishes between full vowel syllabification, which has the effect of forming
CV syllables; and schwa syllabification, which assigns to a non-syllabified CC sequence the
shape CəC in accordance with Benhallam’s (1990a) SSAA.
(iii) MA distinguishes between two modes of schwa syllabification: nominal schwa
syllabification and verb and adjective schwa syllabification. Nominal schwa syllabification is
argued to be dependent to a large extent on the sonority of the surrounding consonants (Al Ghadi
1990, Boudlal 1993, to appear a) while verb and adjective schwa syllabification is governed by
Benhallam’s SSAA 4.

4
That a language has two modes of syllabification is not unnatural. The way phonological rules apply to verbs and
nouns need not be the same. For example, Bobaljik (1997) notes that in Itelmen, a language spoken in the Northwest
coast of the Kamchatka peninsula of Russia, the rule which epenthesizes a schwa applies cyclically in the verbal
system but non-cyclically in the nominal system. Also, Smith (1997) proposes domain specific constraints for the
lexical category “noun”, which need not apply to other categories.

51
 These assumptions have been reformulated in terms of constraints on syllabic well-
formedness to fit in the theoretical framework adopted. (See subsections 3.2-3.6 below for more
detail)
In the following subsection we present the basic tenets of syllable theory in OT as
outlined in Prince and Smolensky (1993) and elaborated in McCarthy and prince (1994a, 1994b,
1995, 1999).

3.1 Syllable Sructure in Optimality Theory

According to Prince and Smolensky (1993) syllable structure in OT is generated in the

same way as any other grammatical property. The function Gen produces a set of candidates for
any unsyllabified input. The function Eval chooses the optimal candidate which should abide by
the constraints imposed by UG and ranked on a language-particular basis.
It is a widely held view that the basic syllable structure is of the type CV (Jakobson
(1962), Clements and Keyser (1983), among others). Two basic universal constraints emerge
from the above statement. Prince and Smolensky (1993: 85) state them as follows:

-12-

a. ONSET
Syllables must have an onset.

b. NO-CODA
Syllables must not have a coda.

Together ONSET and NO-CODA describe what is referred to as the universally unmarked
characteristic of the structures involved. Given an input with the shape /CVCV/, the function Gen
may supply the following candidates, among others:

-13-
a. CV. CV
b. CVC.V

52
Of the two parses, 13a is the optimal one since it satisfies the two constraints stated in 12. The
parse in 13b is suboptimal in two ways: the first syllable is closed and as such violates the NO-
CODA constraint; the second syllable violates the ONSET constraint.
Besides the ONSET and the NO-CODA constraints, Prince and Smolensky (1993) claim
that there is a second group of constraints on syllable structure, stated as follows:

-14-

a. PARSE
Underlying segments must be parsed into syllable structure.

b. FILL
Syllable positions must be filled with underlying segments.

Together PARSE and FILL constitute what is referred to as the “faithfulness family of
constraints”. They constrain the relation between structure and input. They also demand that
well-formed syllable structures are those in which input segments match the syllable positions in
a one-to-one fashion.
Later developments within the OT framework have given rise to CT (McCarthy and
Prince 1995, 1999) which extends the reduplicative copying relation of McCarthy and Prince
(1993a) to other domains where identity relations are imposed on pairs of related representations
such as input and output (and output and output in the extended version of CT (McCarthy, 1995,
1997, Benua 1995, 1997, Kenstowicz 1996, 1997, Kager 1996, Burzio 1996, Basri et al 1998,
Selkirk 1999). McCarthy and Prince (1995) reformulate Prince and Smolensky’s (1993)
faithfulness constraints in such a way as to liberate them from their connection with
syllabification and phonetic interpretation. They instead propose that the constraint FILL and
part of what the constraint PARSE does be replaced by DEP and MAX, respectively. The
domain-specific instantiations of MAX and DEP we will be using are mainly the ones that hold
between the input and the output. Under CT, the two constraints are formulated as follows:

-15-

a. MAX-IO
Every segment of the input has a correspondent in the output.

53
b. DEP-IO
Every segment of the output has a correspondent in the input.

To see how MAX and DEP work, consider an input of the shape /CVC/. The function Gen may
supply the candidates in 16. The lower case v stands for an epenthetic vowel:

-16-
a. CVC
b. CV
c. CV.Cv

In 16a, the whole input is parsed as one syllable, thus violating the NO-CODA constraint. In 16b
only the sequence CV is syllabified, satisfying the NO-CODA and at the same time violating
MAX-IO since the final consonant has been deleted. The sequence in 16c has resorted to final v
addition and as such satisfies both the NO-CODA and MAX-IO but violates DEP-IO which
demands that the segments of the output have correspondents in the input.
The optimal candidate cannot be determined from the structures in 16 above because each
of these structures violates one constraint. For Prince and Smolensky (1993), the optimal forms
are those that display minimal violation of universal constraints. Given the facts in 16, it follows
that the optimal candidate can only be determined after the ranking of these constraints. The
candidate that violates the higher-ranked constraints is suboptimal while the one that violates the
lower-ranked constraints is optimal. It should be noted here that individual grammars rank
universal constraints differently depending on the internal system of the language concerned. In a
language that allows codas, the optimal candidate would be 16a and as such the NO-CODA
constraint would be ranked low in the ranking scale. In a language where MAX-IO is ranked low,
the structure in 16b would be the optimal one. Finally, in a language where DEP-IO is ranked
low, the structure in 16c would be the optimal one.
To sum up, the constraints on syllable structure are of two types: the ONSET and NO-
CODA constraints, and the revised faithfulness constraints which comprise MAX-IO and DEP-
IO. In section 3.2 below, we turn to see how the ranking of MAX-IO and DEP-IO and their
interaction with other constraints can account for CMA syllable structure.

54
 3.2 Universal Constraints on CMA Syllable Structure

As stated in section 3.1 syllable structure in OT is generated in the same way as any
grammatical structure. The function Gen supplies a large number of candidate parses; Eval
chooses the optimal parse according to the constraint hierarchy. It has also been pointed out
above that in order to account for syllable structure in CMA, we need universal constraints like
ONSET, NO-CODA, DEP-IO and MAX-IO.
In this section, we consider first the interaction of ONSET and DEP-IO. It is an
established fact in MA that a syllable (whether it is at the beginning or within the prosodic word)
cannot start with a vowel. Whenever such a situation arises, recourse is made to epenthesis.
Consider the following items for illustration:

-17-

a. ʔatay tea ʃrib-atay drinking tea

ʔargan Argan (tree) zit-argan Argan oil
ʔana I bγit-ana I want
ʔaxwər Another waħd-axwər Another one

b. slawi from Salé sla Salé (a city)

smawi sky-blue sma sky
DRawi from the Plain Dra DRa The Plain Dra
tadlawi from the Plain Tadla tadla The Plain Tadla

In 17a the epenthetic element is the glottal stop; in 17b, it is the glide [w] which is epenthesized
between the suffix [i] and the stem final vowel. Both cases involve epenthesis and therefore
violation of DEP-IO. Consider the tableau below for illustration:
It is evident from the items in 17 that any form violating ONSET will be eliminated since
there are candidate parses that meet the constraint ONSET by epenthesizing a glottal stop or a
glide, thus forcing violation of DEP-IO. The behavior of the items in 17 points to the fact that
ONSET must be ranked above DEP-IO, that is ONSET must dominate DEP-IO. This domination
is shown in the following tableau for the input /atay/:

55
-18-

/atay/ ONSET DEP-IO

a. a.tay *!
)b. ʔa.tay *

If we reverse the ranking of ONSET and DEP-IO, the optimal candidate will be an item with an
onsetless syllable *[atay], a form CMA rules out.
To see if MAX-IO interacts with ONSET and DEP-IO, we add another candidate to the
ones in 18b above. The candidates we will examine are given below. The symbol 0 shows the
wrong optimal candidate according to the constraint ranking given:

-19-

/atay/ DEP-IO MAX-IO

a. ʔa.tay *!
0 b. tay *

It is to be noted here that Gen allows for the generation of candidates like the one in 19b, where a
segment is deleted. The deletion in 19 leads to the satisfaction of ONSET and forces violation of
MAX-IO. The domination relation established in 19 makes the wrong prediction since it posits as
the optimal parse the one where the vowel [a] of the input is deleted. This means that the two
constraints should not be ranked with respect to each other as in 20 below:

-20-

/atay/ MAX-IO DEP-IO

)a. ʔa.tay *
b. tay *!

The constraint tableau in 21 shows the interaction of ONSET, MAX-IO and DEP-IO. The dotted
line shows that ONSET and MAX-IO are not ranked with respect to each other:

56
-21-

/atay/ ONSET MAX-IO DEP-IO

) a. ʔa.tay *
b. tay *!
c. a.tay *!

As seen in 21, each of the candidates incurs one violation mark; but since violation of lower-
ranked constraints (DEP-IO) is allowed to secure higher-ranked constraints (MAX-IO and
ONSET), it follows that the optimal candidate is [ʔatay].
The constraints above could also account for the cases where a glide is epenthesized
instead of the glottal stop. In the constraint tableau below we consider three candidate parses for
the input /tadla-i/:

-22-

/tadla-i/ ONSET MAX-IO DEP-IO

) a. tad.la.wi *
b. tad.li *!
c. tad.la.i *!

The tableaux considered so far show that any form violating ONSET and MAX-IO will never be
optimal.
Having considered the interaction of the constraints in MAX-IO, ONSET and DEP-IO, let
us now examine the faithfulness constraints DEP-IO and MAX-IO and their interaction with the
NO-CODA constraint. Consider the parsed items below for illustration:

-23-

ʃaw.ya *ʃa.wə.ya roasting (fem.)

ʕay.ʃa *ʕa.yə.ʃa alive (fem.)
mat *ma.tə he died
sa.lat *sa.la.tə she finished

57
The items above show that it is more optimal to have codas than gratuitously violate DEP-IO.
They further show that NO-CODA must be ranked low in the scale, lower than DEP-IO as the
tableau in 24 below shows. Given the fact that MAX-IO dominates DEP-IO, it follows from this
that MAX-IO also dominates NO-CODA, by transitivity:

-24-

/ʃawya/ MAX-IO DEP-IO NO-CODA

)a. ʃaw.ya *
b. ʃa.ya *!
c. ʃa.wə.ya *!

The candidate in 24b has resorted to schwa epenthesis to avoid violation of the NO-CODA
constraint, thus resulting in a form that surfaces with a schwa in an open syllable. To block such
forms I have proposed a constraint dubbed *ə]σ which has the effect of ruling out open schwa
syllables. The effect of this constraint could be seen in forms inflected for the feminine and to
which the object clitic [-u] is added. If the feminine suffix is [at] ( as is the case with CMA),
nothing special happens; the constraint ONSET is satisfied by adjoining the [t] to constitute the
onset of a syllable whose nucleus is the object clitic. If, on the other hand, the feminine suffix is
[ət], another variant found in other varieties of MA, the final [t] gets geminated. The initial part
of the geminate functions as a coda, thus observing the constraint *ə]σ, whereas the second part
functions as an onset to the final syllable.

-25-

DəRbatu DəRbəttu she hit it/him

kətbatu kətbəttu she wrote it
sərqatu sərqəttu she stole it
qətlatu qətləttu she killed it/him

58
It will be seen that the effect of the constraint *ə]σ could be obtained from the NO-CODA
constraint. That the schwa does not occur in open syllables follows from ranking NO-CODA low
in the ranking scale and not from the constraint *ə]σ itself.5
The next cases that will be considered are cases of items that involve schwa epenthesis.
Consider the examples in 26 below for illustration.

-26-
a. xədma job
SəmTa belt
DəRba a hit
zəbda butter

b. DaRəb hitting
ʃarəb drinking
katəb writing
sakət mute

The schwa in the above items is not part of the underlying representation; it is epenthetic
(Benhallam, 1980, 1988, 1990a). In CT terms, epenthesis means violation of DEP-IO. In the
items above, we allow for the violation of DEP-IO to secure a higher-ranked constraint, namely
ONSET. A reasonable question that should be asked here is the following: What is it that forces
violation of DEP-IO in the items in 26?
As we have shown above, the first step in Benhallam’s (1990a) SSAA is to build core CV
syllables. Thus items such as those in 26a are syllabified as CC.CV. (cf. Sm.ta.) whereas items
such as those in 26b are syllabified as .CV.CC (cf. .ka.tb). Later syllabification rules assign the
first or final two unsyllabified consonants to a syllable whose nucleus is the schwa.
Within the theoretical framework adopted here, we assume that schwa epenthesis and
consequently DEP-IO violation is triggered by some dominating constraint labeled PARSE-seg
and which Prince and Smolensky (1993) state as follows:

-27-

PARSE-segment (henceforth PARSE-seg)

Every segment must belong to a syllable.

5
Thanks to Lisa Selkirk, Karim Bensoukas and Paul de Lacy for pointing out this to me.

59
We assume that the constraint PARSE-seg is also undominated since all the segments of an input
are parsed into some higher prosodic constituent, namely the syllable. Given the fact that all the
segments in the input must be realized in the output by virtue of undominated MAX-IO, the only
way to syllabify the two stray consonants in 26 is by epenthesizing a schwa. Consider the
different parses of the word [katəb] from the input /katb/. We assume that both MAX-IO and
PARSE-seg dominate DEP-IO 5.

-28-

/katb/ MAX-IO PARSE-seg DEP-IO

)a. ka.təb *
b. .ka.tb **!
c. ka **!
d. ka.tə.bə **!

The optimal candidate satisfies both MAX-IO and PARSE-seg but violates DEP-IO, a lower-
ranked constraint. The output in 28b is suboptimal because two of its segments (i.e. [t] and [b])
do not belong to a prosodic constituent, thus violating PARSE-seg. Even if we assume that the
sequence tb were in fact a syllable, we would have to assign the consonant [b] as a nucleus to
that syllable whose onset is [t], something that CMA does not resort to as will be argued below.
The parse in 28c is discarded because two of its input segments were deleted. Finally the parse in
28d is bad because it incurs two violation marks of DEP-IO.
What the tableau above does not include is the monosyllabic candidate [katb] which does
not violate any of the constraints in 28 and should therefore win over the real optimal candidate
[katəb], which violates DEP-IO. In order to rule out candidates such as [katb], we need to invoke
the constraint *COMPLEX-MARGIN (Prince and Smolensky 1993). The constraint is stated as
follows:

5
One might wonder whether we really need the constraint PARSE-seg since MAX-IO ensures that all the segments
in the input appear in the output. PARSE-seg triggers schwa epenthesis in words whose input consists exclusively of
consonants such as [ktəb]. An alternative constraint to PARSE-seg would be NUCLEUS, which Prince and
Smolensky (1993) assume to be universally undominated. However, the problem with NUCLEUS is that it cannot
force epenthesis in an input such as /CCC/. All it says is that if there are syllables, they have to have nuclei. Since
there are no syllables in /CCC/, it follows that NUCLEUS cannot trigger schwa epenthesis.

60
-29-

*COMPLEX-MARGIN (henceforth *COMPLEX)

Codas and onsets must not branch.

The constraint, if ranked above DEP-IO, would rule out forms such as [katb] with a complex
coda as the tableau below shows:

-30-

/katb/ *COMPLEX DEP-IO

)a. ka.təb *
b. katb *!

We assume that MAX-IO, PARSE-seg and *COMPLEX are not ranked with respect to each
other and that the three of them must dominate DEP-IO. This means that these three undominated
constraints are what triggers schwa epenthesis and therefore DEP-IO violation.
The constraints seen so far are of two types: (a) the undominated constraints, which are
ONSET and MAX-IO, PARSE-seg and *COMPLEX; and (b) the dominated ones which are
DEP-IO and NO-CODA.
Next, we consider items that begin with a cluster of consonants, which present a special
case that needs to be analyzed. An input such as /bka/ could have one of the following output
candidates.

-31-

PARSE-seg MAX-IO *COMPLEX DEP-IO

a. bka *!
0b. bə.ka *
c. ka *!

61
The constraints in 31 wrongly predict that the optimal candidate is 31b, i.e. the candidate that has
epenthesized a schwa to satisfy *COMPLEX. What facts about CMA point out to is that the
optimal form must be a form that preserves all of the input segments and does not incur a DEP-
IO violation. Such is the case with the candidate [bka], except that this form violates
*COMPLEX by allowing a complex onset in the output. Since deleting one of the two
consonants will result in violation of undominated MAX-IO in CMA (and in other Arabic
dialects, Abu Mansour 1995), it follows that the margin consonant [b] must be assigned as
extrasyllabic in the sense of Ito (1986, 1989) and McCarty and Prince (1988). The notion of
extrasyllabicity was first suggested for MA in Al Ghadi (1990) and adopted in subsequent work
such as Rguibi 1990, El Himer 1991, Imouzaz 1991, and Boudlal 1993, among others. In all
these works, extrasyllabicity is shown to operate at the edges of the word. In OT terms,
extrasyllabicity has been regarded as a lack of parsing and therefore violation of MAX-IO.
However and as we have seen above, the constraint MAX-IO is undominated in CMA, ensuring
that all the segments in the input appear in the output. Given a situation where an initial cluster of
consonants needs to be syllabified, the only way to do it is by assigning the first member of the
cluster to a degenerate syllable (Selkirk 1981) and the second member as an onset to the main
syllable. Thus the verb like [bka] “he cried” may be represented as in 32 below:

-32-

Ft

σ σ

µ
|
b k a

It should be noted that adjoining the consonant [b] to the syllable node does not constitute a
violation of PARSE-seg. In fact what is violated is Selkirk’s (1980) Strict Layer Hypothesis
(STRICT-LAYER) which demands that every prosodic constituent be dominated by a constituent
of the immediately superordinate type, that is the mora is dominated by the syllable, and the
syllable is dominated by the foot which is, in turn, dominated by the prosodic word.
The representation in 32 above shows that a distinction should be made between two
types of syllables: a degenerate syllable, which will be referred to as minor syllable, and a major

62
syllable. A minor syllable consists solely of a consonant, whereas a major syllable is one whose
nucleus is a schwa or one of the full vowels [i, u, a]. Since having a major syllable is better than
having a minor one, the grammar of CMA will have to incorporate a constraint against minor
syllables. This constraint is stated as follows:

-33-

*MINOR SYLLABLE (henceforth *Min-σ)

Minor syllables are prohibited.

This constraint will have to be dominated by DEP-IO so as to prevent epenthesis in cases such as
[bka]. In 34, we show how the output [b.ka] is obtained:

-34-

*COMPLEX DEP-IO *Min-σ

)a. b.ka *
b. bka *!
c. bə.ka *!

The structure in 32 satisfies *COMPLEX by assigning the first consonant of the optimal
candidate as a minor syllable but raises another issue related to word minimality. According to
Prince and Smolensky (1993), any member of the morphological category corresponds to a
prosodic word (LX ≈ PWd), which in turn corresponds to a foot. According to the authors, the
foot is subject to binarity which requires the PWd to have at least two moras if the language
under study is quantity-sensitive or two syllables if the language is quantity-insensitive. The
constraint FOOT-BINARITY is stated as follows:

-35-

FOOT-BINARITY (henceforth FT-BIN)

Feet are binary under syllabic or moraic analysis.

63
The word [bka], the way it is represented in 32, violates FT-BIN although it is a lexical word. In
the next section, we turn to examine the nature of the prosodic word and the mechanisms CMA
resorts to in order to satisfy FT-BIN.

3.3 The Prosodic Word in CMA

Under moraic theory (Hyman 1985, Zec 1988, Hayes 1989 and others), CMA
distinguishes between bimoraic CVC heavy syllables, where V is different from the schwa (36a);
and monomoraic light syllables, which, in turn fall into two types: one where the mora dominates
one segment (36b); the other where the mora dominates the schwa and another consonant (36c):

-36-
a. σ b. σ c. σ

µµ µ µ
| | |
CVC C V C ə C

There is yet another light syllable that is referred to as minor and whose status will be determined
down in this section.
The representation in 36c emanates from a proposal made in Jebbour (1996) and adopted
for Tashlhit Berber in Bensoukas (1994). These works assume that a closed syllable whose
nucleus is a consonant should be monomoraic. For the purpose of the present work, we assume
that the schwa in CMA is moraless and that it acquires a moraic structure only in combination
with a following consonant belonging to the same syllable. Along the same lines, Al Ghadi
(1994:5) assumes the moraic representations in 36, and this has led him to posit the following
equivalencies between syllables whose nucleus is a full vowel and syllables whose nucleus is a
schwa:

-37-
a. CV = CəC
b. CVC = CəCC
c. CVCV = CəCCəC
d. CCV = CCəC

64
Excluding the patterns in 37a which are not considered to be lexical words7, the patterns in 37b
and 37c occur as separate lexical words in CMA and therefore satisfy both the constraints LX ≈
PWd and FT-BIN by virtue of the fact that they are both bimoraic. The patterns in 42d are
monomoraic and therefore constitute a clear violation of FT-BIN. Notice that these patterns start
with a consonant cluster(CCV and CCəC). In section 3.2 we have posited the constraint
*COMPLEX which demands that syllable margins consist of only a single consonant. We have
shown that whenever there is a consonant cluster, one of the two members is assigned as an onset
or coda, depending on the position in the syllable, while the other member constitutes a minor
syllable on its own. To make things clearer, consider the structure of the verb [bka] in 32 above
which repeat in 38 below.

-38-

Ft

σ σ

b k a

The structure in 38 shows that the lexical word [bka] does not meet the requirement of a PWd. In
other words, it does not satisfy FT-BIN, a universal constraint observed cross-linguistically.
Given the state of affairs in 38, how is it possible to satisfy FT-BIN? In McCarthy and Prince
(1995), for example, it has been shown that languages resort to augmentation to satisfy this
constraint. Such is also the case for a very limited number of CMA words on the pattern CV (cf.
see footnote 7). One way of augmenting words on the pattern CCV or CCəC is by epenthesizing
a schwa between the cluster of consonants, thus resulting in the disyllabic patterns Cə.CV and
Cə.CəC, which satisfy FT-BIN. However this solution is undesirable since it results in forms
which are judged to be ill-formed (cf. *[bəka] and *[kətəb], for example which incur a fatal
violation of DEP-IO). The second solution, which is adopted in the present work, is proposed by

7
Only two words in MA have the form CV: [ʒa] “he came” and [ma] “water”, and these, Al Ghadi (1994) argues,
show augmentation when undergoing certain morphological processes: [ʔaʒi]/[ʒay] “come/coming” and
[mihan]/[myah]/[miman] “waters”.

65
Al Ghadi (1994:5) who considers the first member of an initial consonant cluster or the second
member of a final consonant cluster as part of a degenerate syllable, where the consonant is
dominated by a mora. He also proposes that this mora be adjoined directly to the foot instead of
projecting its own syllable, and this under some *STRUCTURE constraint which favors a
representation with less prosodic nodes and association lines. In the present work, we continue to
assume the Strict Layer Hypothesis to be able to encode the notions major and minor syllables.
Within Al Ghadi’s model, the words [bka] “he cried” and [kəlb] "dog" will have the structures in
39.

-39-

a. PWd b. PWd
| |
Ft Ft

σ σ σ σ

µ µ µ µ
|
bk a k ə l b
For simplification, moraic structure will be given only when it bears on the argument. Thus,
when moraic representation is relevant, structures such as those in 39a and 39b will be rewritten
as [bµ.kaµ] and [kəlµ.bµ], respectively; otherwise, they are written as [b.ka] and [kəl.b]. In both
cases the moraic consonant belongs to a minor syllable rather than to a major one.
The moraification of a consonant dominating a minor syllable leads to the recognition of
another type of light syllable in addition to CV and CəC. This syllable is represented as follows:

-40-

The minor syllable in CMA

σ
|
µ
|
C

66
Notice that the moraification of consonants is the result of the requirement that feet be binary.
Thus in a form such as [bµ.kaµ], satisfying the constraint FT-BIN forces the consonant [b] to be
moraic, thus violating a constraint Prince and Smolensky (1993) call Nuclear Harmony and
which we give in 41:

-41-

Nuclear Harmony (henceforth H-NUC)

A higher sonority nucleus is more harmonic than one with lower sonority.

H-NUC considers C-nuclei to be less harmonic than V-nuclei. However, with words on the
pattern CCV and CCəC, the only way to satisfy FT-BIN is by assigning a mora to the first
consonant, thus violating H-NUC and subsequently *Min-σ. This points out to the fact that FT-
BIN must outrank H-NUC and *Min-σ. We assume that *Min-σ and H-NUC are not ranked with
respect to each other:
-42-

/bka/ FT-BIN *Min-σ H-NUC

)a. bµ.kaµ * *
b. b.kaµ *! *

Contrary to Al Ghadi (1994) who assumes that words on the pattern CCV are derivable from
either the universal constraint FT-BIN or the constraint *COMPLEX, we maintain that the two
constraints, although not rankable with respect to each other, must be kept separate and that any
account of CMA non-derived trisegmental words which start or end up in a consonant cluster
should make use of both of them . What *COMPLEX basically does is force a member of a
consonant cluster to form a minor syllable without ever assigning a mora to it. FT-BIN, on the
other hand forces a member of a consonant cluster to be moraic.
The behavior of initial and final consonant clusters in non-derived trisegmental words
points out to the fact the only way to satisfy the constraint FT-BIN is by assigning a moraic status
to a member of the cluster, an assumption maintained throughout this work. Therefore and in
order not to be repetitive, we will be using only the constraint *Min-σ. Any form that incurs a
violation of *Min-σ automatically violates H-NUC.

67
 It should be noted that *Min-σ must outrank NO-CODA so that the final consonant of
words such as [mat] “he died” would not be dominated by a mora which is dominated by a minor
syllable. Consider the constraint tableau in 43 for illustration:
-43-
/mat/ FT-BIN *Min-σ NO-CODA
) a. Ft
|
σ

*
µ µ
| |
m a t

b. Ft
σ σ

*!
µ µ
| |
m a t

c. Ft
|
σ

*!
µ
|
m a t

In summary, the constraints at our disposal up to now are ranked below:

-44-
a. ONSET MAX-IO>>DEP-IO
b. MAX-IO, PARSE-seg>>DEP-IO>>NO-CODA
c. MAX-IO, PARSE-seg, *COMPLEX, >>DEP-IO>>*Min-σ, H-NUC
d. FT-BIN , *COMPLEX>> DEP-IO>>*Min-σ, H-NUC>>NO-CODA

Next, we consider non-derived items whose input is composed solely of consonants. Recall that
the general rule for verbs (and adjectives) is for the schwa to be epenthesized between the second
and third consonants of the root. Thus, given an input like /ktb/, Gen may provide the following
candidates:

68
-45-

a. kətb
b. ktəb
c. ktb
d. kə.təb
e. kə.tə.bə
The constraints developed so far will give the result in 46.

-46-

*COMPLEX DEP-IO *Min-σ NO-CODA

)b. kµ.təbµ * * *
(a. kətµ.bµ * * *
c. ktəbµ *! *
d. kəµ.təbµ **! *
e. kəµ.təµ.bəµ ***!

The tableau shows that candidate 46c is ruled out because it violates *COMPLEX by allowing a
complex onset. It also shows that candidates 46d and 46e are excluded because they incur too
many violations of DEP-IO. We are left with candidates 46a and 46b which tie in all the
constraints. If this is so, how is it possible to distinguish between [k.təb] where the schwa is
placed between the second and the third consonants of the root and [kət.b] where the schwa is
placed between the first and second consonants of the root?
To answer this question, it should be noted that the difference between the two forms
reflects the directionality of syllabification. Right-to-left syllabification gives the form [ktəb];
left-to-right gives the form [kətb]. As it has already been pointed out, CMA schwa
syllabification proceeds from right-to-left and assigns every unsyllabified CC sequence the shape
CəC. The question that should be asked at this stage is the following: how is it possible to capture
the sense of directionality within a constraint-based framework?
To account for directional syllabification in CMA, we make use of McCarthy and
Prince’s (1993b) Generalized Alignment, and more particularly the constraint ALIGN (stem, R,

69
σ, R) which has the effect of preventing epenthesis at the right edge of the root and ensuring that
it is flush against the right edge of the syllable:

-47-

ALIGN (stem, R, σ, R) (henceforth ALIGN-R)

The right edge of the root must be aligned with the right edge of the syllable.

To account for the difference between [k.təb] and [kət.b], we assume that no domination
relationship exists between *COMPLEX and ALIGN-R and that both constraints must dominate
DEP-IO as the constraint tableau below
shows 6:

-48-

/ktb/ *COMPLEX ALIGN-R DEP-IO *Min-σ

) a. kµ.təbµ * *
(b. kətµ.bµ * *

The two candidates tie once again in everything. They both satisfy ALIGN-R the way it is
formulated in 47: 48a satisfies ALIGN-R by virtue of the fact that the right edge of the stem
corresponds to the right edge of the syllable; 48b satisfies ALIGN-R although the right edge of
the stem corresponds to a minor syllable. How is it possible then to exclude the candidate in 48b
while at the same time establishing the candidate in 48a as the optimal one?
To answer this question, consider the structures in 49 below for both [k.təb] and [kət.b]:

6
ALIGN-R must dominate DEP-IO not only to account for the difference between [ktəb] and [kətb] but also to
prevent schwa epenthesis in stem-final position in forms such as [kətəbə] or [kətbə] for example.

70
-49-

a. PWd b. PWd
| |
Ft Ft

σ σ σ σ

µ µ µ µ

k t ə b k ə t b

Both structures contain a major syllable and a minor one. In 49a the minor syllable is at the left
edge of the major syllable; in 49b it is at its right. One way to get the optimal candidate in 49a is
by positing an alignment constraint, requiring that the right edge of the stem be aligned with a
major syllable (Maj-σ) as stated in 50:

-50-

ALIGN-R Maj-σ
The right edge of the stem aligns with the right edge of a major syllable.

This constraint will have to dominate the general version of ALIGN-R, an example where the
specific constraint dominates the general one (Beckman 1998). Thus [kət.b] is ruled out on the
ground that the right edge of the stem does not align with a major syllable.
However, the problem with this constraint is that it seems to weaken the Alignment
Theory by allowing it to look at the internal structure of the prosodic entity being aligned, i.e. it
has to see whether it is a major or a minor syllable. For this reason, we are led to abandon the
constraint ALIGN-R-Maj-σ in search for another constraint that has an explanatory power.
In order to distinguish [k.təb] from [kət.b], I assume, following a suggestion made to me
by Selkirk (p.c.), that epenthesizing a schwa before the third consonant of the root instead of the
second follows from the general requirement that the stem be iambic, a fact which is justified in
the stress system of the language (see chapter three below). Within the Alignment Theory,
iambicity could be expressed by positing a constraint requiring that the right edge of the stem be

71
aligned with the right edge of a prominent syllable in a foot. The notation σ′ refers to the
prominent syllable:

-51-

ALIGN-R (Stem, σ′) (ALIGN-R-σ′)

The right edge of the stem must be aligned with the right edge of the prominent
syllable.

The constraint is observed in a large number of items, all of which are non-derived trisegmental
adjectives and verbs (except verbs with final geminates) and a class of nouns as the examples
below show:

-52-

a. Verbs
Root stem Gloss
ktb ktəb write
DRb DRəb hit
gls gləs sit down

b. Adjectives
ħwl ħwəl cross-eyed
kħl kħəl black
SfR SfəR yellow

c. Nouns
ktf ktəf shoulder
ʒml ʒməl camel
smn smən preserved butter
sdr sdər chest

Assuming that the language does not allow complex margins ensures that the trisegmental items
in 52 are syllabified as C.CəC with the first consonant being dominated by a minor syllable.
Such a form satisfies ALIGN-σ′-R by virtue of the fact that the right edge of the stem coincides
with the right edge of the prominent syllable of the foot, i.e. the syllable which is susceptible to
bear the main-stress of the word. Such a function could be attributed only to a major syllable, i.e.
a syllable whose nucleus is one of the vowels [i, u, a, ə]. A minor syllable such as the one in 40

72
above, which is dominated by a consonantal mora, can never be the prominent syllable. The
constraint ensuring the non-prominence of a minor syllable is given below:

-53-

*Min-σ′
Prominent minor syllables are prohibited.

We assume that 53 is undominated and that it dominates ALIGN-R-σ′ as shown in the constraint
tableau in 54 below. Prominence is shown by an accent (′ ) over a vowel if the syllable in
question is a major syllable, and after a consonant if the syllable in question is a minor syllable.

-54-

/ktb/ *Min-σ′ ALIGN-R-σ′ DEP-IO

)a. k.tə́b *
b. kə́t.b *! *
c. kət.b́ *! *
Epenthesizing a schwa before the second consonant of the stem leads to a flagrant violation of
ALIGN-R-σ′. The form in 54c tries to satisfy ALIGN-R-σ′ by assigning prominence to a minor
syllable thus resulting in a fatal violation of *Min-σ. Thus the optimal candidate is the one that
satisfies both ALIGN-R-σ′ and *Min-σ′ by right-aligning a major syllable.
However, other items seem to argue against stem-prominent syllable right-alignment.
Such is the case with verbs with final geminates and a class of nouns as shown in the data below:

-55-

a. Verbs
Root Stem Gloss

ʕDD ʕəDD bite

mdd mədd give
mss məss touch
ʒRR ʒəRR pull

73
b. Nouns
DRb DəRb hitting
srʒ sərʒ saddle
lʕb ləʕb game
frx fərx bird

Unlike the items in 52, the ones in 55 are syllabified as CəC.C with the prominent syllable on
the left rather than the right edge of the foot, a clear violation of ALIGN-R-σ′. This violation
points out to the fact that ALIGN-R-σ′ must be outranked by some other constraints so that forms
such as those in 55 could be derived. As it has already been mentioned above, the syllabification
of nouns is governed by sonority and that when the sonority of the second consonant of the stem
is greater than that of the third, the schwa is epenthesized before the second consonant, resulting
in a stem whose right edge coincides with a minor syllable. Thus nouns such as those in 55
present ample evidence that constraints on sonority must be ranked above ALIGN-R-σ′, an issue
that constitutes the subject-matter of section 3.6 below. As to the verbs in 55a, we think that
satisfying ALIGN-R-σ′ by epenthesizing a schwa between the last two parts of the geminate
would split them up, a fact argued against in the relevant literature on geminates (Cf. Guerssel
1978, Benhallam 1980, 1991, Schein and Steriade 1986, Hayes 1986, and Keer 1998, 1999, to
cite a few). Here again, we think that the relevant constraint against splitting up geminates would
have to outrank ALIGN-R-σ′ as will be shown in section 3.5.1 below.
To recapitulate, it has been shown that CMA items on the pattern (C.CəC) abide by the
constraint ALIGN-R-σ′ which requires that the right edge of the stem be aligned with a
prominent syllable and that a minor syllable can never be in a prominent position by virtue of the
higher-ranking constraint *Min-σ′. It has also been pointed out to the fact that ALIGN-R-σ′ must
be dominated by some higher constraint in order to account for verbs with final geminates and a
class of nouns on the pattern (CəC.C).
As it has been referred to earlier in this chapter, CMA distinguishes two modes of schwa
syllabification: nominal schwa syllabification, which is dependent on the sonority of the
consonants constituting the stem, and verb and adjective schwa syllabification. In OT, the appeal
to the difference in morphological category in accounting for the differences in syllabic pattern
between verbs and adjectives, on the one hand, and nouns, on the other, could be expressed in
terms of an alignment constraint requiring that the right edge of the verb and adjective stem be

74
aligned with a prominent syllable. This verb-/adjective-specific constraint is stated as in 56
below:

-56-

ALIGN-R (verb/adjective, σ′) (henceforth ALIGN-R (Vb/Adj, σ′)

The right edge of the verb/adjective stem must be aligned with the right edge of the
prominent syllable.

For the time being, we assume that the verb-/adjective-specific stem-prominent syllable right
alignment ranks higher than the general stem-prominent syllable right alignment stated in 51
above. With this ranking, any trisegmental verb stem epenthesizing a schwa between the second
and third consonants satisfies both ALIGN-R (Vb/Adj, σ′) and
ALIGN-R-σ′. The problem with this ranking comes from words with final geminates which
epenthesize a schwa between the first and second parts of a geminate. Consider the following
illustration for an input such as /sdd/:

-57-

/sdd/Vb ALIGN-R (Vb/Adj, σ′) ALIGN-R-σ′ DEP-IO

0a. s.dəd *
b. səd.d *! *

As the tableau shows, the alignment constraints alone cannot derive the correct output. They
wrongly predict that the optimal candidate is [sdəd] instead of [sədd], a fact which calls for a
higher-ranked additional constraint that would block epenthesis in the case of geminates. In the
next section, we consider this constraint and see how the prosodic word minimality requirement
is achieved in non-derived words containing geminates.

3.4 Syllable Structure and Geminates

This subsection is meant to show why epenthesis is blocked in the case of verbs with final
geminates, leading to the violation of both ALIGN-R

75
(Vb/Adj, σ′) and ALIGN-R-σ′. It is also meant to present further evidence for considering the
first segment of an initial cluster and the last segment of a final cluster in non-derived forms to be
moraic. It is the way CMA resorts to in order to achieve the prosodic word minimality
requirements.
There are two different theories about the representation of geminates. The first is the
One-Root Theory of Length proposed in Hayes (1989) and McCarthy and Prince (1986).
According to this theory, geminates are linked to a single root node as shown in 58 below:

-58-

The One-Root Theory of length

a. Geminate consonant b. Geminate vowel

σ σ σ

... µ ... ... µ µ ...

RC RV
| |
Place Place

The proponents of the One-Root Theory of Length further assume that the root node consists of a
single mora and that the double linking is taken care of by general rules of syllabification.
The second view about geminates is expressed by the Two-Root Theory of Selkirk (1990,
1991). According to this theory, geminates are represented with two root nodes that share
stricture and place features as shown below:

-59-

The Two-Root Theory of length

a. Geminate consonant b. Geminate vowel

RC RC RV RV

Place Place

According to Selkirk, the representations above allow for a straightforward distinction between
full and partial geminates. Full geminates involve the sharing of all features; partial geminates, on

76
the other hand, are structures where specifications for laryngeal features or nasality may differ in
the two halves.
It is the Two-Root Theory of Length that will be adopted in the present work for the
analysis of the cases that involve geminates. The reason for this choice is that the Two-Root
Theory treats geminates as a cluster of consonants, something that points out to the possibility of
splitting geminates and consequently deriving words with final geminates (see chapter 5 and
section 3.5.1 in this chapter). With an underlying one-root representation, there is no reason why
we would actually derive a geminate in the case of words with final geminates
Within the Two-Root Theory, a word such as [dda] “he took away” will be represented as
follows:

-60-

Ft

σ σ

µ µ

RC RC RV

d a

To encode the Two-Root representation of geminates, a word such as [dda] will be represented
underlyingly as /dda/.
It is to be noted here that, contrary to the One-Root Theory, the Two-Root Theory does
not say anything about the moraic structure of geminates because this is a property of the
language under consideration. In CMA, and as it has already been pointed out, the initial segment
of the word in 60 is associated to two root nodes, thus producing initial geminate, the first of
which is associated to a mora to satisfy FT-BIN.
The analysis proposed in this section will cover both final and initial underlying
geminates.

77
 3.4.1 Final Geminates

In this subsection, we will consider both trisegmental words on the pattern CəCiCi and
quadrisegmental verbs on the pattern CəCCiəCi and show that their analysis goes along the lines
proposed for sound verbs of the type /ktb/ .
In 61 below, we present cases of trisegmental nouns and verbs whose final segment is
geminated:

-61-
a. Non-derived Nouns
bəqq bugs
w
m əxx brain
dəmm blood
fəkk jaw
nədd a kind of incense
w
f əmm mouth

b. Non-derived Verbs
sədd close
ħəll open
ʕəDD bite
ʒəRR pull
w
d əqq knock at
həzz lift
ħəTT put down

Given the prosodic organization adopted in the previous section, a word such as [sədd] could
have either of the two representations given in 62 below:

-62-

a. Ft b. Ft
|
σ σ σ
|
µ µ µ µ
RC RC RC RC

s ə d s ə d

78
The representation in 62a shows that the word [sədd] is monosyllabic, consisting of a heavy
syllable. The other representation shows that the word consists of two syllables, with the second
one being a minor syllable associated with the second part of the geminate. Both representations
satisfy FT-BIN. However, given the assumptions made above about syllable structure, 62a should
be excluded on the ground that it violates *COMPLEX. Note here that the Two-Root theory
treats geminates as clusters of consonants and as such they should abide by the constraint
*COMPLEX. However, One can spare violation of *COMPLEX by deleting a root consonant of
the input, thus resulting in a form such as [səd]. To prevent this deletion, we make recourse to a
constraint of the MAX family , namely MAX-RC, which demands that all root consonants of the
input be preserved in the output as shown in 63:

-63-

*COMPLEX MAX-RC DEP-IO *Min-σ

)a. sədµ.dµ * *
b. səddµµ *! *
c. sədµ *! *

Candidates 63b and 63c are ruled out for violating higher-ranked constraints: 63a is excluded
because it violates *COMPLEX by allowing the two parts of the geminate to occur as coda; 63c,
because it violates the constraint requiring that the root consonants of the input be preserved in
the output. 63c could also be excluded because it violates FT-BIN.
Returning back to the distinction between *[s.dəd] and [səd.d], we have pointed out that
although [səd.d] violates the specific instantiation of stem-prominent syllable right alignment, it
should be considered optimal. *[sdəd] is ruled out because the geminates are split up by schwa
epenthesis, a fact which has been argued against in the literature (Guerssel 1978, Benhallam
1980, 1991, Schein and Steriade 1986, Hayes 1986, Keer 1998, 1999, among others). This shows
that other relevant constraints ought to be incorporated into the grammar of CMA. In order to
derive the correct output, we introduce the constraint NO-SPLITTING which has the effect of
blocking schwa epenthesis from splitting geminates in words such as [sədd]. This constraint is
stated as follows:

79
-64-

NO-SPLITTING
Splitting up geminates is prohibited.

This constraint must rank higher than ALIGN-R(Vb/Adj, σ′) and ALIGN-R-σ′ in order to get the
optimal candidate [səd.d]:

-65-

/sdd/Vb NO-SPLITTING ALIGN-R ALIGN-R-σ′ DEP-IO

(Vb/Adj, σ′)
)a. səd.d * * *
d. s.dəd *! *

The effect of the NO-SPLITTING constraint is to prevent schwa epenthesis from applying in
final geminated verbs. The optimal candidate shows that it is more highly valued to violate the
alignment constraints than epenthesize a schwa between the two parts of the geminate.
Nouns such as the ones in 61a could be accounted for in the same way as verbs except
that it is only the constraint ALIGN-R-σ′ which is active and not ALIGN-R (Vb/Adj, σ′) as the
following tableau for the noun [bəqq] shows:

-66-

/bqq/N NO-SPLITTING ALIGN-R ALIGN-R-σ′ DEP-IO

(Vb/Adj, σ′)
)a. bəq.q * *
d. b.qəq *! *

The shading is meant to show that the concerned constraint is irrelevant since the target is a noun.
The final items we will consider in this subsection are cases of quadrisegmental verbs
whose final segment is geminated:

80
-67-

bərgəg he spied on
fərtət he broke into fritters
ʕəntət he showed stubbornness
ħənZəZ he gazed at
bəqləl he gazed at
kənZəZ he holds his jaws firmly

An input such as /brgg/ will have the candidates listed in the tableau in 68:

-68-

/brgg/Vb MAX-RC NO-SPLITTING ALIGN-R DEP-IO

(Vb/Adj, σ′)
a. bər.gəg *! **
b. b.rəg *! *
c. bər.g *! * *
0d. b.rəg.g * *

The constraints listed in this tableau wrongly predict that the optimal candidate is 68d instead of
68b. The other two candidates are excluded on the ground that they both violate MAX-RC by
deleting a root consonant. So, how is it possible to exclude the candidate in 68d while at the
same time establish 68a as optimal?
To answer this question, it is noteworthy to point out here that verbs in CMA, whether
they are trisegmental or quadrisegmental, are governed by a prosodic constraint which demand
that they consist exactly of two moras. This constraint is stated in 69 below:

-69-

VERB ROOT = [µ µ]
A verb root must correspond to two moras.

To get the correct output, the constraint Verb Root = [µ µ] has to dominate NO-SPLITTING to
allow schwa epenthesis to split the final geminates in quadrisegmental verbs as shown in 70:

81
-71-

/brgg/Vb Verb Root NO- ALIGN-R ALIGN-R-σ′ DEP-IO

= SPLITTING (Vb/Adj, σ′)
[µ µ]
)a. bərµ.gəgµ * **
b. bµ.rəgµ.gµ *! * * *

A second competing candidate to the one in 71a would be [bµ.rəggµ]. This form satisfies the
bimoraicity requirement at the expense of *COMPLEX.
To sum up, it has been shown that the Two-Root Theory of Length allows for a better
representation of geminates in that it treats them as a cluster of consionants that share common
features. It has also been shown that the fact that final geminates in trisegmental verbs are never
split by schwa epenthesis results from ranking NO-SPLITTING above ALIGN-R (Vb/Adj, σ′). In
quadrisegmentals, on the other hand, verbs with final geminates satisfy ALIGN-R (Vb/Adj, σ′) at
the expense of NO-SPLITTING. However, this violation is allowed to secure the higher ranked
constraint on verb bimoraicity.

3.4.2 Initial Geminates

Consider the examples with initial geminates given in 72 below:

-72-

a. DDaR < l-DaR the house

ssma < l-sma the sky
nnas < l-nas people
ddir < t-dir you do
DDRəb < t-DRəb you hit
nnʕəs < n-nʕəs we sleep
nnʕəs < n-nʕəs we engrave

b. dda he took
bbwa my father
mmwi my mother

82
The items in 72a are cases of heteromorphemic geminates; they arise whenever two coronal
segments come into contiguity. The items in 72b are cases of tautomorphemic geminates and
these are the only items that occur in the language. Our objective here is not to account for the
process of gemination but only to consider the cases that might pose a problem of word
minimality requirement. For a more detailed account of gemination in MA, the reader is referred
to works such as Benhallam 1980, 1991, Rguibi 1990 and El Himer 1993.
In order to account for words with initial geminates, we won’t make any recourse to the
alignment constraints since these are irrelevant in deciding about the optimal candidate. What we
need are in fact the constraints FT-BIN, MAX-RC, *COMPLEX, DEP-IO and *Min-σ. In the
tableau below, we show how a representative item such as [dda] from the list in 72b is obtained:

-73-

/dda/ FT-BIN *COMPLEX MAX-RC *Min-σ

)a. Ft

σ σ
| *
µ µ
| |
RC RC RV
|
d a

b. Ft

σ σ
*! *
µ µ
|
RC RV
| |
d a
c. Ft

σ
*(!) *(!)
µ
|
RC RC RV
|
d a

83
The Two-Root Theory, coupled with the relevant constraints considered above, allows us to
derive the optimal candidate in 73a which incurs a violation mark for *Min-σ to satisfy FT-BIN.
The candidate in 73b satisfies FT-BIN but is ruled out because it incurs a fatal violation of MAX-
RC by deleting a root consonant node of the input. Finally, the candidate in 73c associates both
root consonants to the syllable node and is excluded because it violates either FT-BIN or
*COMPLEX.
To sum up, it has been shown that initial geminates, and more particularly words on the
pattern CiCiV are treated in the same way as words on the pattern CCəC or CCV. The first part of
the geminate is always associated with a minor syllable to satisfy both *COMPLEX and FT-BIN.
In the next section, we will see if the constraints considered above can account for cyclic
syllabification in verbs and adjectives in CMA.

3.5 The Problem of Cyclic Syllabification

It has been pointed out in section 2 above that syllabification in CMA verbs and
adjectives should be allowed to apply cyclically in order to get the attested output. In this section,
we will reconsider this problem in the light of the OT constraints stated so far and see if their
ranking is capable of generating the correct output. But before we do that, let us consider the
adjective and the verb paradigms given in 74:

-74-
a. Adjectives
Masculine Feminine Gloss

ʕrəʒ ʕərʒ-a lame

Smək Səmk-in deaf
SfəR SəfR-a yellow
xDəR xəDR-in green
zRəq zəRqa blue
kħəl kəħla black

b. Perfective form of the verb [ktəb] “write”

1sg. ktəb-t 1pl. ktəb-na
2sg. ktəb-ti 2pl. ktəb-tu
3sg.mas. ktəb 3pl. kətb-u
3sg.fem. kətb-at

84
 c. Perfective form of the verb [ʒəRʒəR] “trail”
1sg. ʒəRʒəR-t 1pl. ʒəRʒəR-na
2sg. ʒəRʒəR-ti 2pl. ʒəRʒəR-tu
3sg.mas ʒəRʒəR 3pl. ʒəRʒR-u
3sg.fem. ʒəRʒR-at

The items in 74a represent a class of trisegmental adjectives on the pattern CCəC. When a vowel-
initial suffix is added, the final consonant of the stem (CCəC) ceases to function as a coda of the
schwa syllable and is adjoined as an onset to a syllable whose nucleus is the vowel of the suffix,
thus observing the ONSET constraint. The same thing can be said about the third person feminine
singular and the third person plural in the verb paradigms in 74b and 74c. These items point to
the fact that ONSET must dominate ALIGN-R since the satisfaction of ONSET leads to the
misalignment of the right edge of the stem and the right edge of the syllable and therefore
violation of ALIGN-R.
Within the derivational model of syllabification, the forms that have received special
treatments are the 1sg. of trisegmental verbs and the 1sg., 3sg.fem., and 3pl. of quadrisegmental
verbs. For example Benhallam (1990a) assumes that syllabification has to apply cyclically in
order to derive the correct output. We give below the derivation of [kətbt] “I wrote” and
[ʒəRʒRu] “they trailed”.

-75-

Input [[ktb]t] [[ʒRʒR]u]

First cycle
Syllabification .k.təb. .ʒəR.ʒəR.
Second Cycle
Affixation .k.təb.-t .ʒəR.ʒəR.-u
Syllabification .k.təb.t. .ʒəR.ʒə.Ru
Syllabification ── .ʒəR.ʒ.Ru.
Output [kətbu] [ʒəRʒRu]

Recall that the schwa in MA does not occur in open syllables and that explains why it drops in
the stem [ʒəRʒəR] after the affixation of [-u].

85
 The analysis in 75 is operational in the sense that an input form has to go through
different intermediate stages before it reaches the final stage of output representation. Such a
stand is therefore incompatible with the principles of OT, namely that of parallel evaluation of
candidates. Moreover
Let us now see if the constraints developed in the previous section could account for the
problematic cases stated above. First, consider a trisegmental verb to which the first person
singular marker is affixed. The competing candidates for the input /ktb-t/ are listed in the tableau
below:

-76-

/ktb-t/ *COMPLEX ALIGN-R DEP-IO *Min-σ

(Vb/Adj, σ′)
a. k.təb.t * * **
0b. kət.bət **
c. ktəb.t *! (*!) * *
d. k.təbt *! * *

This ranking wrongly predicts that the optimal candidate is 76b where the right edge of the verb
aligns with a prominent syllable. 76a is excluded because it right-aligns a minor syllable which,
as has been argued above, can never be prominent because of undominated *Min- σ′. The
candidates in 76c and 76d are both excluded on the ground that they violate *COMPLEX: 76c
has a complex onset while 76b has a complex coda. The form in 76c could also be excluded
because it violates ALIGN-R (Vb/Adj, σ′).
Trisegmental verbs show that the constraints developed above are not sufficient enough to
derive the optimal form. The same thing could be said about quadrisegmental affixed verbs. For
example, an input such as /ʒRʒR-u/ may have the following output candidates:

86
-77-

/ʒRʒR-u/ *COMPLEX ALIGN-R DEP-IO *Min-σ

(Vb/Adj, σ′)
) a. ʒəR.ʒ.Ru * *
(b. ʒ.Rəʒ.Ru * *
c. ʒRəʒ.Ru *! *
d. ʒəR.ʒə.Ru **!

The candidates in 77c and 77d are both excluded for different reasons: 77c incurs a fatal violation
of *COMPLEX, whereas 77d is excluded because it has resorted to schwa epenthesis to avoid a
complex onset, thus incurring an additional violation mark of DEP-IO and resulting in an open
schwa syllable. The forms in 77a and 77b tie in everything and as such the optimal form 68a can
be determined neither on the basis of the constraints listed in this tableau nor on the other
constraints seen so far.
Both trisegmental and quadrisegmental affixed verbs show that cyclic syllabification in
CMA poses a problem for the theoretical framework in the version adopted so far. This calls for a
revision or extension of the this framework. Suffice it here to raise the problem; it will receive
due consideration in chapter four where tit will be shown that cyclic phenomena are cases that
necessitate reference to a different kind of faithfulness relation holding between the derived
output form and the simple base form. In particular, it will be shown that in order to derive the
correct output, reference must be made to an output-output constraint requiring that the syllable
initial segments of output correspond to the syllable initial segments in the base form.

3.6 Sonority and Syllabification

The final case we will consider in this chapter is that of nominal schwa syllabification.
Consider some of the nonderived nouns given in 78 below:

-78-

a. kəlb dog
dənb sin
bərd wind

87
 fərx bird
ʃəmʃ sun
DəRb hitting
ləʕb game
qərd monkey
γərs plant
ħəRb war
qəlb heart
nəħs bad luck
bənt girl
kərʃ stomach

b. rʒəl leg
ktəf shoulder
STəl bucket
ħbəl rope
bγəl mule
wtəd peg
zbəl rubbish
qbəR tomb
Dbəʕ hyena
bħəR sea
sbəʕ lion
qfəz cage
ħTəb firewood
wdən ear

c. γnəm sheep
tmən hang-over
gməl lice
ʒməl camel
qməR gambling
tməR dates
wsəx dirt
ʃdəg loaf (of bread)
ftəq hernia
smən preserved butter
nmər tiger

The schwa in the above items is dependent on the sonority of the second and third consonants of
the root. It is epenthesized before the second consonant of the root if its sonority is greater than
that of the third consonant (78a). If the sonority of the third consonant is greater than that of the

88
second consonant, the schwa is epenthesized before the third consonant (78b). Also, the schwa is
epenthesized before the third consonant if its sonority equals that of the second consonant (78c).
What the items in 78b and 78c show is that right-to-left directionality of schwa syllabification is
also observed in nouns and this through satisfaction of the constraint ALIGN-R-σ′ demanding
coincidence of the right edge of the stem with the right edge of a prominent syllable. The only
cases where ALIGN-R-σ′ is violated is when sonority is at stake. This points out to the fact that
ALIGN-R-σ′ must rank below sonority. The question that should be asked here is the following:
how is it possible to express the relative sonority of consonants in the theoretical framework
adopted in the present work?
To answer this question, it should be noted that the schwa in CMA is moraless on its own
and that it acquires a moraic status only in combination with a following consonant in the same
syllable. Such an assumption excludes the possibility of having schwas in open syllables,
something which is true about MA (Benhallam 1980, 1988, 1990a; Hammoumi 1988, Al Ghadi
1990, Boudlal 1993, 1998 and others). In other words, all schwa syllables have a coda, and it is
the coda which determines the epenthesis of the schwa. The behavior of the schwa in 78 is
reminiscent of what Clements (1988:68) calls the Dispersion Principle which he states as follows:

-79-
The Dispersion Principle:
a. The preferred initial demisyllable maximizes sonority dispersion.
b. The preferred final demisyllable minimizes sonority dispersion.

Demisyllables according to Clements are overlapping portions of a syllable sharing the peak. For
example CV is an initial demisyllable while VC is a final demisyllable 9.
What interests us here are final demisyllables which Clements (1988:69) ranks as follows:

-80-
Final demisyllables
V ¾ VG ¾ VL ¾ VN ¾ VO
(G=glide, L=liquid, N=nasal and O=obstruent, and ¾ means better than)

9
According to Clements (1988) V is both an initial and final demisyllable. Syllables on this pattern are called one-
member demisyllables.

89
What 80 basically states is that codaless syllables rank high and that if there has to be a coda, the
difference between the sonority of the nucleus and that of the coda in a syllable should not be
significant. In other words, the closer the sonority of the coda is to that of the nucleus the better.
The CMA data in 78 seem to abide by the ranking in 80 except that final demisyllables of
the type V do not occur if V is a schwa. On a parallel basis, the constraints on CMA final ə-
demisyllables can be stated as in 81. We assume that a ranking should be established within the
class of obstruents whereby fricatives (F) dominate stops (S):

-81-

Final ə-demisyllables in CMA

əG >> əL >> əN >> əF >> əS

Recall from our analysis that both the schwa and the following consonant, i.e. the coda, are
associated with a single mora. If this is the case, the ranking in 81 could well be expressed in
terms of negative constraints on CMA ə-demisyllables. The ranking of these negative sonority
constraints is given in 82 below:

-82-

SONORITY (in nouns)

*µ >> *µ >> *µ >> *µ >> *µ

ə S ə F ə N ə L ə G

Note that the sonority constraints in 82 are to be distinguished from

10
H-NUC . They reflect the idea that the optimal coda of schwa syllables is one with a higher
sonority value.

10
My interpretation of H-NUC differs from that of Al Ghadi (1994) who assumes that a C occupies a nucleus
position if it is exclusively dominated by a mora in word-initial or coda positions (cf. C.CV and CəC.C, where the
moraic consonant is underlined) or if it is jointly with a schwa dominated by a mora (as in CəC.CəC, where both əC
are associated with a single mora. While we mainatin, following Al Ghadi, that H-NUC is incurred when a mora
dominates C, we believe that sequences such as əC should be explained by sonority constraints of the types proposed
in 82 where the schwa is placed before the most harmonic coda in terms of sonority.

90
 The sonority constraints in 82 along with the constraints developed so far can adequately
account for the nominal items in 78. Recall from our discussion above that nouns on the pattern
(CəC.C) violate ALIGN-R-σ′ because the right edge of the stem aligns with a minor syllable
which cannot be prominent. This points out to the fact that the sonority constraints must rank
higher than ALIGN-R-σ′. As to the constraint ALIGN-R(Vb/Adj, σ′), it does not have any visible
effect on nouns and as such will not be included in the analysis. In the tableau below, we consider
the different parses of the input /klb/N, where the sonority constraints outrank ALIGN-R-σ′.

-83-
*µ *µ ALIGN-R-σ′.
/klb/N
ə S ə L

)a. Ft

σ σ
|
µ µ * *
|
kəl b

b. Ft

σ σ
|
µ µ *!
|
k ləb

Now consider a case where the schwa is epenthesized between the second and third consonants of
the root and where the sonority of the third consonant is greater than that of the second. An input
noun such as /ktf/N would have the candidate parses represented in 84.

91
-84-
*µ *µ ALIGN-R-σ′
/ktf/N
ə S ə F

a. Ft

σ σ
|
µ µ *! *
|
kət f

)b. Ft

σ σ
|
µ µ *
|
k təf

The last case of trisegmental nouns we will consider is one where the sonority of the second
consonant equals that of the third. Here the schwa is epenthesized between the two consonants
and it is the constraint ALIGN-R-σ′ which is decisive. Consider the two parses of the input /smn/
given in 85 below:

92
-85-

*µ ALIGN-R-σ′
/smn/N
ə N

)a. Ft

σ σ
| *
µ µ
|
s mən

b. Ft

σ σ
| * *!
µ µ
|
səmn

Wherever the schwa is placed (before [n] or before [m]), the constraint *µ/əN is violated.
Although ALIGN-R-σ′ is dominated, it is still active in the language in that it enables us to
determine the appropriate placement of the schwa in trisegmental nouns whose second and third
consonants have equal sonority.
The syllabification of quadrisegmental nouns on the pattern CCCC is generally CəC.CəC
as shown in 86 below:

-86-

Root Stem Gloss

a. ʃrʒm ʃərʒəm window

mdfʕ mədfəʕ canon
tnbr tənbər stamp

b fndq fəndəq hotel

mħbq məħbəq flower pot
mslm məsləm Muslim

93
In non-derived quadrisegmental nouns such as the ones given in 86, one wonders whether it is
necessary to refer to the sonority constraints since there is no other way for the sequence CCCC
to syllabify except as CəC.CəC. Syllabifying the sequence as CCəCC would violate
*COMPLEX. Trying to avoid having complex margins by syllabifying the sequence as C.CəC.C
would constitute a violation of ALIGN-R-σ′ as shown in the constraint tableau below:

-87-

/mħbq/N *COMPLEX ALIGN-R-σ′. DEP-IO *Min-σ

)a. məħ.bəq **
b. mħəbq *! *
c. m. ħəb.q *! * **

Now let us consider quadrisegmental suffixed nouns to see if the constraints developed so far can
generate the correct output. First consider the examples in 88:

-88-

a. snsl-a sənsla zip

ʃrʒm-u ʃərʒmu his window
tnbr-i tənbri my stamp
brdʕ-a bərdʕa saddle-bag

b. mslm-in msəlmin Muslims

ʃfnʒ-at ʃfənʒat doughnuts
mTRq-a mTəRqa hammer
mγwrf-a mγwərfa ladle
mʕlq-a mʕəlqa spoon
mbxr-a mbəxra censer

In these items, the final consonant of the root is syllabified as the onset of a syllable whose
nucleus is the vowel of the suffix; the remainder is syllabified much in the manner of the
trisegmental nouns considered above. The schwa is epenthesized before the second consonant of
the root if its sonority is greater than that of the third consonant (88a), and before the third
consonant if its sonority is greater than that of the second (88b). The constraint tableau in 89
gives some of the candidate parses of the input /mslm-in/.

94
-89-

*µ *µ *µ
/mslm-in/N *COMPLEX ALIGN-R-σ′
ə S ə N ə L
)a. m.səl.min *
b. msəl.min *! *
c. məs.ləm.in *! *

The candidate in 89c can further be excluded because the right-hand syllable is left without an
onset, something which constitutes a fatal violation of undominated ONSET. Notice the
irrelevance of ALIGN-R-σ′ in determining the optimal parse. Once again, the constraints
developed in this paper can adequately account for noun as well as adjective and verb
syllabification.
To sum up, the constraints needed to account for CMA syllable structure are given below:

-90-

a. Undominated constraints: FT-BIN, *COMPLEX, MAX-IO, PARSE-seg, ONSET,

SONORITY (in nouns), *Min-σ′, ALIGN-R (Vb/Adj, σ′) and VERB ROOT =[µ µ].
b. Dominated constraints: DEP-IO, NO-CODA, H-NUC, *Min-σ, ALIGN-R,
ALIGN-R-σ′ and NO-SPLITTING.
The domination relation among these constraints is given in 91:

-91-

a. ONSET, MAX-IO, PARSE-seg>>DEP-IO>>NO-CODA

b. FT-BIN, *COMPLEX>>DEP-IO>> *Min-σ, H-NUC>>NO-CODA
c. *Min-σ′>>ALIGN-R (Vb/Adj, σ′)>>ALIGN-R-σ′>> ALIGN-R>>DEP-IO
d. VERB ROOT=[µ µ]>>NO-SPLITTING>> ALIGN-R (Vb/Adj, σ′)>>ALIGN-R-σ′
e. ALIGN-R (Vb/Adj, σ′), SONORITYN>> ALIGN-R-σ′

The constraints in 91 are reproduced in the hierarchical structure in 92:

95
-92-

Constraint hierarchy in CMA

FT-BIN *COMPLEX MAX-IO PARSE-seg ONSET SONORITY VERB ROOT = *Min-σ′

(in nouns) [µ µ]

NO-SPLITTING

ALIGN-R (Vb/Adj, σ′)

ALIGN-R-σ′

ALIGN-R

DEP-IO

*Min-σ, H-NUC

NO-CODA

4. CONCLUSION
This chapter has tried to account for CMA syllable structure within the OT constraint-
based framework. It has been shown that such a framework, which derives syllabic well-
formedness from the interaction of constraints belonging to UG, is far better than a step-by-step
syllable structure building algorithm, especially in the problematic cases of schwa syllabification.

96
 While we maintain the previous scholars’ assumptions that schwas are epenthetic and
dependent on syllable structure as well as on the sonority of the consonants of the base if this
happens to be a noun, the analysis offered in this chapter has an explanatory power since it shows
that prosodic structure assignment in CMA is governed not by rules but by constraints such as the
ones listed in 92. A constraint-based analysis offers a straightforward analysis to some of the
recalcitrant problems like directionality of syllabification and the representation of geminates and
their contribution to the achievement of prosodic word minimality requirement. Directionality of
syllabification has been shown to derive from alignment constraints such as ALIGN-R (Vb/Adj,
σ′), ALIGN-R-σ′ or else from ALIGN-R. It has been shown that in a large number of
trisegmental items, the schwa is epenthesized before the third consonant of the root and this
follows from the constraint requiring that the stem be iambic. It has also been shown that the
difference between verb and adjective schwa syllabification, on the one hand and noun schwa
syllabification on the other could be accounted for by ranking the verb-/adjective stem-prominent
syllable alignment above the general stem-prominent syllable right alignment. In both cases, it
has been shown that a minor syllable can never be in prominent position, a prohibition ensured by
the constraint *Min-σ′. As to minimality requirement it has been shown that in the case of non-
derived words, the first segment of an initial cluster or the second segment of a final cluster must
be moraic and therefore form a minor syllable on its own. By adopting the Two-Root Theory of
length, the proposed analysis also nicely accounts for words with initial and final geminates. It
has been shown that the fact that final geminates in trisegmental stems are never split up by
schwa epenthesis follows from ranking ALIGN-R (Vb/Adj, σ′) and ALIGN-R-σ′ immediately
below NO-SPLITTING. This constraint is violated only when the constraint on verb bimoraicity
is at stake as is the case with quadrisegmental verbs with final geminates.
The chapter has also shown that nominal cases, where schwa syllabification depends on
the sonority of the consonants in the input, can adequately be accounted for in terms of universal
constraints demanding that the sonority of the consonant serving as the coda of schwa syllables
be as close as possible to that of the nucleus. It has also been shown that directionality plays an
important role not only in verb and adjective schwa syllabification but also in nominal schwa
syllabification, especially in trisegmental roots whose second and third segments have the same
sonority value. Here, we have argued that ALIGN-R-σ′ decides in favor of the candidate that
best satisfies the constraint.

97
 In sum, two types of constraints have been distinguished: dominated constraints which are
DEP-IO, NO-CODA, H-NUC, ALIGN-R (Vb/Adj, σ′), ALIGN-R-σ′, ALIGN-R, NO-
SPLITTING and *Min-σ; and undominated constraints which are ONSET, MAX-IO, PARSE-
seg, *COMPLEX, SONORITY in nouns, *Min-σ′, VERV=[µ µ] and FT-BIN which is satisfied
in CMA by associating the first consonant of CCV or CCəC to a mora as proposed by Al Ghadi
(1994). We have argued that the undominated constraints are never violated and as such they are
ranked at the top of the ranking scale. Throughout this chapter, it has been argued that the relative
ranking of these constraints is what determines the right syllabic output.

98
 Chapter Three

The Stress System of Casablanca Moroccan Arabic

1. INTRODUCTION
One of the least studied linguistic phenomena in MA phonology is stress. Except for
the impressionistic study of Abdelmassih (1973), it was not until the beginning of the 80’s
that Moroccan scholars started studying stress, especially in the work of Benkirane (1982).
This work was subsequently followed by other works such as Benkaddour (1982),
Hammoumi (1988), Benhallam (1990b), El Hadri (1993), Fares (1993), and Nejmi (1993,
1995). The objective of this chapter is to enrich the research on MA stress both from an
empirical side by doing instrumental work and also from the theoretical side by applying the
OT principles to account for stress assignment.
Any analysis of MA stress has to distinguish between the epenthetic schwa [ə] and the
underlying full vowels /i, u, a/. Such a distinction is very important in the sense that it helps
characterize syllable weight which is a decisive factor in a number of stress systems. Works
such as Benkirane (1982), Bennis (1992), Al Ghadi (1994) and Boudlal (to appear a) maintain
that, in MA, a light syllable of the type CV (where V is a full vowel) is equivalent to CəC,
which should also be considered as light. If this is so, it follows that the weight distinction
needed to account for MA stress is one between the heavy CVC syllable and the light CV and
CəC syllables. Further support to this claim will be given in this chapter.
In the present chapter we will show that the location of stress depends on whether or
not the items considered occur in isolation or in context. The stress patterns obtained from
words in isolation show that CMA is a quantity sensitive system which favors trochaic feet.
we will also show that the fact that stress falls on one of the last two syllables of a word
follows from the constraint requiring the alignment of the right edge of the foot containing the
stressed syllable with the right edge of the prosodic word. When the word occurs in context,
stress falls consistently on the final syllable, a fact that calls for an iambic type of analysis.
 The chapter is organized as follows. In section 2, we present a review of the literature
on MA stress. In section 3, we lay down an empirical basis of stress in CMA. The objective
of this section is to quantify the native speakers’ intuition about the location of stress in
CMA. In section 4, we undertake an instrumental test to see whether the results obtained here
match up with those obtained from the quantitative test. Finally, in section 5, we offer an OT
based analysis.

2. REVIEW OF THE LITERATURE ON MOROCCAN ARABIC STRESS

The last two decades have witnessed a growing body of research on MA stress. A
common feature of the literature written on MA is that stress is phonetic and therefore does
not interact with word-formation rules.
The body of work on MA stress falls into one of the two categories given below:
(i) Impressionistic 1: Abdelmassih (1973), Benkaddour (1982), Fares (1993), El Hadri
(1993), and Benhallam (1990b). The works of the last three scholars include a quantitative
analysis of stress.
(ii) Instrumental: Benkirane (1982), Hammoumi (1988), and Nejmi (1993, 1995).
Abdelmassih (1973) and Benkaddour (1982) try to analyze MA stress on the basis of
corpuses that need reconsideration. The former did not include schwas where they should
actually be; the latter assumes that MA has long vowels, something which is not true about
the language. (For a detailed critical review of Abdelmassih and Benkaddour, the reader is
referred to Benhallam, 1990b).
Fares (1993) analyzes stress in nouns and adjectives in Tetuan MA (TMA) which is
one of the varieties spoken in the north of Morocco. This variety is characterized by the
occurrence of full vowels in places where other varieties of MA have schwas. This means that
any conclusion obtained from this study cannot be generalized to varieties such as the one
described in the present work.
The subjects, all of whom have a linguistic background, were given a list of more than
400 adjectives and nouns and were asked to locate stress in these words. The results obtained

1
The appellation is due to Benhallam (1990) who classifies studies on MA stress into two categories:
impressionistic and experimental (i.e. instrumental).

100
show that more than 75 % of the total number of items receive stress on the penultimate
syllable. Fares’ main findings can be summarized as follows:
(i) Stress assignment is postlexical, that is it applies after all morphological and
phonological rules have applied.
(ii) Stress in TMA nouns is not different from stress in adjectives, that is both
categories have identical stress patterns.
(iii) Stress is not affected by syllable weight (Cf. [msámaħ] ‘forgiven’ and [γúlal]
‘snails’ where the penultimate light syllable is stressed in spite of the fact that the final
syllable is heavy).
To account for stress placement in TMA, Fares (1993: 282) proposes the following
stress assignment rules:

-1-

Stress the penultimate syllable, except:

a. If the nucleus of the penultimate syllable is a schwa and the item contains more than
three syllables, stress the antepenultimate syllable; otherwise stress the penult (e.g. [xnáfərna]
‘our nostrils’ and [swáyəʕkum] ‘your (pl.) watches’)
b. If the penultimate syllable is a prefix, stress the final syllable (e.g. [məsrúq] ‘stolen’
and [məbní] ‘built’)

Within the theoretical framework she adopted (Metrical Stress Theory), Fares (1993) has
shown that stress is represented in terms of left-headed binary branching trees which are
constructed from right to left. In order to account for cases with antepenultimate stress, which
she considers as exceptional, she assumes ternary branching feet which are assigned by a rule
that applies before the binary foot assignment rule. As to items with final stress, Fares has
argued that these items are assigned binary branching feet which dominate the final syllable
in the head position and a zero syllable node in the weak position.
In the same variety of MA, El Hadri (1993) analyzes stress in verbs within Metrical
Theory. Two procedures were followed in the analysis of this phonological phenomenon.
First, the subjects, who were all linguistics students at the English department, were given a
list of more than 387 verbs and were asked to locate stress in these words. Second, the author
proceeded by recording native speakers with no linguistic background. He then listened to

101
these recordings and assigned stress on the basis of his intuition about the repetitions made by
the subjects.
Out of a total number of 387 subjects, El Hadri (1993: 236) found out that 208 stress
the penultimate syllable (about 54%), 126 stress the final syllable (about 33%), 31 stress the
antepenultimate syllable (about 8%) and finally for more than three-syllable words, 22
subjects stress the initial syllable(about 6%). The findings obtained from his experiment do
not differ much from those of Fares (1993). He has also found out that stress assignment rules
are insensitive to syllable weight; it is the position of the syllable which is the determining
factor. Further he has shown that these rules are not sensitive to vowel quality, that is both
schwas and full vowels can be stressed. The only exception relates to the schwa and the
morphological affiliation of the category within which it occurs. If the schwa is part of the
stem, it gets stress as in the case of words such as [nkə́tbu] “we write” and [TaRʒəmli]
“translate for me”; if it is part of an affix, it does not get stress as in words such as [nəbRá]
“we recover” and [maTarʒmáləkʃi] “he did not translate it (fem.) for you”.
El Hadri (1993) further assumes that stress may be sensitive to the syntactic
information of verbs, namely tense and aspect. Thus trisyllabic verbs in the perfective and
imperfective aspect are generally stressed on the penultimate syllable as in [kətbúlum] ‘they
wrote to them’ and [ʒawəbni] ‘he answered me’ while verbs in the imperative receive initial
stress (cf. [kə́tbulum] and [ʒáwəbni]).
Benhallam (1990b) tries to quantify data on the intuitions of the native speaker of MA
about the location of stress. His corpus was chosen in such a way that both full vowels [i, u,
a] and the schwa [ə] would be tested in all possible environments. The items selected were
disyllabic, trisyllabic and quadrisyllabic. His subjects, who were all linguistics students, were
given a list of items and were asked to mark stress on the appropriate spot, relying on their
intuition and on their prior knowledge of English word stress.

Benhallam’s (1990b) findings are reproduced in 2 below:

-2-

Stress the penultimate except,

102
 A. if the word starts by a sequence of CV syllables, stress the one in initial
position: krínahum “we rented them.”

B. In disyllabics
a. in a structure such as CəCCəC, stress falls on the stem vowel:
nəxdə́m “I work.”

b. in a structure CəCCVC, stress falls on the final syllable, that is the

syllable which contains a full vowel: Trəztíh “you embroidered
it (masc.)

C. If the penultimate syllable is an object clitic, stress the preceding syllable:

kərkbíhalhum “roll (2 fem.sg.) it for them.”

The advantage of Benhallam’s experiment is that it is one of the fewest empirical studies of
MA stress that tries to quantify the results. The generalizations obtained from this study seem
to partially reflect the stress system of the language and therefore the tendency made in
previous works, namely that stress falls on the penultimate syllable.
The common point among the works of Benhallam (1990b), El Hadri (1993) and
Fares (1993) is that they all try to capture generalizations about stress by relying only on
intuitions of native speakers. However, in order for these generalizations to hold for MA, they
need to be corroborated by an instrumental study that should refute or confirm the intuition of
the MA speaker about the placement of stress. Also, an instrumental analysis would have to
examine the effect of clitics on word stress, that is whether or not they are counted in the
assignment of stress.
The first instrumental work undertaken on MA stress is that of Benkirane (1982). The
author submitted a corpus which consists of monosyllabic, disyllabic and trisyllabic words to
34 subjects who were asked in an auditory test to identify the stressed syllable. The patterns
he considered are given in 2 below:

-3-

a. Monosyllabics
CVC bál he urinated

b. Disyllabics
CVCV bála cunning (sg.)
CV-CVC mazál it (masc.)/he is still …

103
 CVCVC bánan bananas
CV-CVC-V mazála it (fem.)/he is still …
CCVCVC lħánut store

c. Trisyllabics
CVCVCV taráza turban
CV-CVC-VC mazálin they are still …
CCVCVCV zzitúna olive
CVCVCVC tarázat turbans
CCVCVCVC lbaláwat cunning (pl.)

The results of the auditory test were confirmed by the instrumental one where 351
realizations were subjected to a study of fundamental frequency, intensity and duration. The
conclusion Benkirane draws from both tests is that stress in MA falls on the ultimate syllable
if it is heavy; otherwise it is on the penultimate syllable. The rule responsible for stress is
formulated by Benkirane (1982: 78) as follows:

-4-

V ----------> [+accent] / ── C (V) # #

In spite of the instrumental nature of Benkirane (1982), the work is questionable from the
empirical point of view. First, the rule in 4 does not reflect the stress patterns in 3. For
example, the word [banan] receives penultimate stress instead of final stress as Benkirane’s
rule predicts. Second, the patterns considered do not reflect all the syllable types of the
language. For example, schwa syllables were neglected in the analysis in spite of the fact that
they are also taken into account in the placement of stress. Third, not all syllable types are
considered in different environments. That is, the patterns in 3 do not measure the effect of
syllable weight on stress placement, an element which has been proved to be essential to the
understanding of stress in MA (Benkaddour 1982, Bohas et al 1989, Bouziri 1991, and Nejmi
1993, 1995) and other languages (Hyman 1985, Prince and Smolensky 1993, McCarthy and
Prince 1993a, Hung 1995, Hayes 1995, Pater 1995, Alber 1997, Green 1997, among others).
Third, the patterns given above seem to point to the fact that stress in MA takes into
consideration morphological boundaries, something which has been found not to be true in
later experimental works (Hammoumi 1988 and Nejmi 1993, 1995).
The second instrumental work on MA is that of Hammoumi (1988). Hammoumi’s
analysis is based on a corpus which contains 61 words ranging from two-syllable to five-

104
syllable cliticized words in verbs and from two-syllable to seven-syllable cliticized words in
nouns. The subjects were asked to identify the most prominent syllable and the results of the
auditory test were subjected to an instrumental test which confirmed that fundamental
frequency and duration are the most important parameters in stress.
The conclusion one can draw from Hammoumi’s analysis is that stress falls on one of
the last two syllables of the word, a result confirmed by Nejmi (1993, 1995) and the work
undertaken in this chapter. Furthermore, the analysis implicitly shows that one has to
distinguish between two types of heavy syllables: CəC and CVC, and one type of superheavy
syllable CV:C 2. The stress rules themselves scan the last syllable of word to assign stress first
to a final superheavy syllable as in [ʕəgzá:n] “lazy”, then to a CVC as in [babkúm] “your (pl.)
door”, and finally to a CəC heavy syllable as in [Sa:ħə́b] “friend” and [bəddə́l] “he changed”.
If the last syllable is neither superheavy nor heavy, stress falls on the penultimate syllable as
in [DRə́bti] “you (sing.) hit” and [qbí:la] “a tribe”.
Hammoumi’s analysis could be questioned for two reasons. First, unlike most of the
work done on Moroccan Arabic, he assumes that the language has long vowels that shorten in
specific environments without presenting evidence for that. Such an assumption about vowel
length has serious implications on the analysis since work on stress in natural languages has
shown that heavy syllables attract stress and that light syllables receive stress only in the
absence of heavy ones. Second, his analysis implicitly equates CVC with CəC, a fact which is
not true about MA (see Benkirane 1982, Al Ghadi 1990, 1994, Nejmi 1993 and Boudlal to
appear a).
Perhaps the most elaborate instrumental work on MA stress (and more particularly on
CMA stress) is that of Nejmi (1993), which analyzes stress in the variety spoken in
Casablanca. It is so in the sense that the corpus studied is varied and takes into consideration
the schwa and full vowels as well as the different syllables in different environments. The
strategy followed was to undertake an auditory test where the subjects were asked to locate
the most prominent syllable in a corpus which consists of 104 disyllabic and trisyllabic verbs
and nouns. About 20 subjects with a linguistic background, aged between 24 and 30, were
selected for this test and asked to provide three repetitions for the target words. Prior to that

2
In fact Hammoumi (1988) did not distinguish between light/heavy and superheavy but the analysis he
presented implicitly makes this distinction. A super heavy syllable is represented here as CV:C where V: stands
for a long vowel.

105
task, the subjects were explained the task they had to do. The results of this test were
quantified much in Benhallam’s (1990b) sense, and then subjected to a statistical analysis of
the mean, the standard deviation and the probability to see their significance for any
generalization to be made. The conclusion drawn is that stress is affected by syllable weight,
i.e. whether the syllable is light or heavy; and by syllable position, i.e final or penultimate.
The perceptive test was followed by the acoustic test to see to what extent they match
up. The recordings of three subjects underwent the instrumental analysis where the prosodic
parameters of fundamental frequency, duration and intensity were considered. The analyzed
syllabic patterns are: CVCCVC, CVCVC, CVCCV, CVCəC, CəCCV, CVCVCV and finally
CVCCVCV.
The results obtained from the instrumental test confirm the perceptive test. The
conclusion Nejmi (1993) draws is that the placement of stress in CMA depends on three
factors: first, syllable weight, i.e. light versus heavy. Second, the nature of the heavy syllable.
In this respect, Nejmi (1993) distinguishes between first degree CVC heavy syllables and
second degree CəC heavy syllables 3. Third, stress placement takes into consideration the
position of the syllable, i.e. final or penultimate syllable. According to him, the domain of
stress in CMA is restricted to the final two syllables and this in contradistinction with the
other Arabic dialects where stress may be on the antepenultimate syllable (Farwaneh 1996).
According to Nejmi, the generalization governing stress in CMA could be stated as follows:

-5-

Stress the final syllable if it is a first degree heavy syllable; otherwise stress the
penultimate.

This statement accounts for final stress in words such as [fəkrún] “turtle” and [məktabát]
“book stores” and penultimate stress in words such as [xúxa] “a peach”, [wáʕdək] “he
promised you”.
The only criticism that could be leveled at Nejmi (1993) is that it did not consider
polysyllabic words. Nor did it consider cliticized words to see whether or not clitics affect

3
Nejmi (1993) uses the expressions “syllabe lourde du premier niveau” to refer to CVC heavy syllables, and
“syllabe lourde du dexième niveau” to refer to CəC syllables, which he qualifies as second degree heavy
syllables. (See chapter 2 and also the sections below for an argument against this distinction).

106
word stress. In the sections to follow in this chapter, we will try to overcome this handicap
and consider a varied corpus that ranges from two- to five-syllable words both cliticized and
non-cliticized.

3. ESTABLISHING AN EMPIRICAL BASIS FOR STRESS IN

CASABLANCA MOROCCAN ARABIC
This section is intended to lay down the background for the analysis of stress
presented within the OT framework in section 4 below. It is divided into two major
subsections. In the first subsection, a quantitative test is undertaken and a corpus of 39 items
is presented to a group of 60 subjects who were asked to place stress on the syllable they
perceive as prominent. In the second subsection, the recordings of 5 selected subjects out of
the 60 subjects who participated in the quantitative test, were subjected to a laboratory
analysis to measure fundamental frequency, intensity and duration.

3.1 The Quantitative Test

The idea of establishing a quantitative analysis of stress is inspired from a similar

work undertaken by Benhallam (1990b) whose objective is to quantify the intuitions of native
speakers of MA about the placement of stress. The objective of the test undertaken here is to
try to set the basis for an instrumental study and see whether or not such a study would match
up the intuitions of the native speaker of CMA.

3.1.1 The Hypothesis

Following Benhallam (1990b), we will assume that the working hypothesis in the
present work is that stress is placed on the penultimate syllable. Such a hypothesis is to a
large extent confirmed by most of the works undertaken on MA stress. Thus our objective in
the present research will be to consider why words with final or antepenultimate stress, if
there are any, do not receive penultimate stress.

107
 3.1.2 The Questionnaire

The questionnaire consists of two parts: general information about the subjects, and the test
items 4. In the first part, the subjects were asked to respond to the following:

-6-

a. First name:
b. Last name:
c. Sex: † male † female
d. Date of birth:
e. Place of birth:
f. Occupation:
g. Have you taken any linguistics courses?
† Yes † No
h. If yes, for how long?
i. Do you speak Berber?
†. Yes † No

The questionnaire was prepared in such a way that the subjects would first proceed by giving
personal information about themselves and then about activities that relate to their field of
interest, that is linguistics. While we think that students’ names and their gender are not
determinant in the activity undertaken, we strongly believe that information about the place of
birth and whether or not the subject is a native speaker of Berber play a key role in the
research undertaken. Given the fact that the variety of MA dealt with is the one spoken in
Casablanca, it follows that choosing the subjects who were born in Casablanca would ensure
a high degree of homogeneity. Also homogeneity would be preserved if the subjects speaking
Berber were discarded.
The second part of the questionnaire is the list of test items which we give in the
following subsection.

4
I would like to thank Said Bennis for helping out in the designing of the questionnaire.

108
 3.1.3 The Data

The test items were chosen in such a way as to include, whenever possible, only
sonorant segments and this to neutralize differences that may be due to the nature of the
consonants. It is assumed that a vowel preceding a voiced consonant has a longer duration
than one preceding a voiceless consonant (Peterson and Lehiste 1960, Chen 1970, Mack
1982, Kluender et al. 1988, Laefur 1992, and de Lacy 1998, to cite a few). Also in choosing
the test items, we tried to include all syllable types (that is CVC, CəC, and CV) in different
environments to see whether stress is sensitive to syllable weight as claimed by Hammoumi
(1988) and Nejmi (1993) or syllable position as shown by Nejmi (1993). Walker (1996) has
shown that languages vary on the way they assign stress. Thus in some languages, stress is
sensitive to syllable weight (e.g. Classical Arabic and Hindi); or, it is sensitive to the edge,
whether it is the right edge or the left edge (e.g. Tinrin and Uzbek). Finally, in some other
languages, stress is prominence-driven, that is to say sensitive to the nature of the vowel that
constitutes the nucleus of the syllable (e.g. Koya and Kuuku-Yaʔu).
The test items considered in this work include both simple and affixed forms as well
as cliticized forms. Considering this array of forms will allow us to see whether cliticization
or affixation, in general, plays any major role in stress assignment. The list of test items
includes disyllabic, trisyllabic and polysyllabic words and is listed the way it has been
presented to the subjects:

-7-

a. Disyllabic words

law.yin wilted (pl.)

kal.kum he ate you (pl.)
məl.yun a million
DRəb.hum he hit them
wal.mək it (mas.) suits you
naw.ya intending (fem.)
bab.ha her/its (fem.) door
li.mun oranges
da.ruh they did it (mas.)
li.kum for/to you
məl.məl he shook (sth.)
wəl.dək your (sg.) son

109
 ma.yəl bent (mas.)
ʕu.dək your (sg.) stick
Rəm.la sand
Səm.na we fasted
wa.lu nothing
Da.ru his house
li.na for/to us

b. Trisyllabic words

law.yin.hum they are twisting them

man.Da.Rin clementine
wal.mu.kum they fit you (pl.)
bəZ.Tam.kum your (pl.) wallet
mər.məd.nak we trailed you (in dust)
ru.bəl.kum he disturbed you (pl.)
ʕəR.ga.nin they are sweaty
mqul.bi.nək they are deceiving you
wəl.dat.ni she gave birth to me
min.ʒa.ra sharpener
ban.ya.ha she built it (fem.)
rub.lu.ni they disturbed me
gən.Du.Ra a Moroccan gown
yəD.Rəb.na he hits us
məl.mlu.na they shook us
li.mu.na an orange
ʕi.ni.na our eyes

c. Polysyllabic words

ban.ya.ha.lih she is building it (fem.) for him

di.ri.ha.li.ha do it (fem.) for her
gəl.bu.ha.li.kum they reversed it (fem.) for/to you (pl.)

A remark that should be made here concerns the polysyllabic words above. It concerns the
dative [-l-] which is always followed by the vowel [i] in CMA, whereas in other varieties of
MA such as the one described in Benhallam (1990b), this [i] never appears. Thus [dirihaliha]
in CMA is realized as [dirihalha] in other varieties. Also the object clitic [-h] found at the end
of words such as [banyahalih] is replaced by [u] in other varieties (cf. [banyahalu]). The
difference reflects the variation that exists among the varieties of MA. This justifies our
choice of excluding non-native speakers of CMA from the list of the subjects who have
completed the questionnaire.

110
 3.1.4 The Subjects

The initial number of subjects was 60. 8 of them were researchers in linguistics 5 ; the
rest were all linguistics students in the English department at the Faculté des Lettres, Ain
chok, Casablanca6. These students had at least two years of linguistics studies. All the
subjects were given the list of test items listed in the subsection above and were asked to
mark stress on the syllable they judge to be most prominent. Prior to this, we made it clear to
the students that the questionnaire was undertaken in the framework of a research project
whose objective was to determine some phonological aspects of CMA without ever
specifying these aspects. In order not to give the subjects two difficulties at a time, we chose
to parse all the test items into syllables. The students were given a brief survey of stress in
English, and then explained the task they had to do. They were asked to identify the stressed
syllable in the items concerned by pronouncing the word several times and emphasizing at
every syllable until they found the alternative pronunciation which best matched their
intuition.
Before processing the data, the first task we had do was to select the subjects that meet
the criteria advanced above. Out of the 60 subjects, only 35 were retained: 17 males and 18
females. The others were excluded on the ground that their results might threaten the validity
of the test either because they were not native to Casablanca or because they had another
mother tongue, namely one of the varieties of Berber.

3.1.5 Results And Analysis

Let us first give the numbers that will be used as a base to the interpretation of the results.
The total number in 8a corresponds to the maximum possible number of answers which is
obtained by multiplying the total number of subjects (35) and the total number of test items
(39). The number in 8b corresponds to the total non-respondents (NR), that is the subjects,
who for one reason or another, were unable to place stress on one of the syllables. Third, the

5
Special thanks to Mohamed Sghir Syad, Rachid Hamdi, Samir Diouny, Brahim Taoukkoul, Mourad Mawhoub
for accepting to respond to the questionnaire about CMA stress.
6
Thanks to Professor Mohamed Es-Safi and all his 1998-99 Linguistics students for hosting me at the Faculté
des Lettres, Ain Chok, Casablanca and also for accepting to respond to the questionnnaire.

111
number in 8c corresponds to the total number of responses obtained by substructing the total
number of NR from the maximum possible number of answers.

-8-

a Maximum possible number of answers: 35 x 39 = 1365

b. Total NR = 37
c. Total number of responses (a - b = c) = 1328

Since the working hypothesis in this work is that stress falls on the penultimate syllable, we
judge it necessary to give the total number of times penultimate syllables in the test items
above receive stress.

-9-

a. Disyllabic words: 438

b. Trisyllabic words: 339
c. Polysyllabic words: 12

The total number of penultimate stresses is 789, a number which corresponds to more than
half of the number of total responses. The percentage of responses totaled by penultimate
stress is given in 10 below:

-10-

789
─── = 59 %
1328

In what follows is a listing of the number of stresses by position. The position includes
antepenultimate, penultimate and ultimate stress in all words, whether they are disyllabic,
trisyllabic or polysyllabic; and the first and second syllables in polysyllabic words that consist
of 4 or 5 syllables:

112
-11-

-Disyllabic -Trisyllabic -Four-syllable -Five-syllable

words words words words

-First syllable - - - 5

-Second - - 4 17
syllable

- - 133 8 17
Antepenultim
ate syllable

-Penultimate 438 339 3 9

syllable

-Ultimate 218 112 14 11

syllable

In 12 below, we give the number of items totaling the maximum number of stresses on
penultimate syllable:

-12-

a. Disyllabic words: 14/19

b. Trisyllabic words: 13/17
c. Polysyllabic words: 0/3

In disyllabic words, 14 out of 19 receive penultimate stress, i.e. about 76%. In trisyllabic
words 13 out of 17 receive penultimate stress, i.e. about 74%. Finally none of the three
polysyllabic words considered receive final stress, a fact that shows that the more syllables a
word has the more difficult it is for the subjects to locate stress.

113
 3.1.5.1 Disyllabic Words

For the sake of clarity of presentation, the results of the questionnaire have been
tabulated. The first column contains all the test items; the second contains the total number of
no response. The other columns numbered from 1 to 5 correspond to the number of syllables
in the word.
Recall from our assumption above that stress falls on the penultimate syllable, an
assumption that follows from Benhallam (1990b). If this is so, an explanation of why some
disyllabic items do not take penultimate stress is in order. But before we do that, consider the
score of the subjects in disyllabic words:

114
-13-

ITEMS NR σ1 σ2
a. law.yin 1 15 19
b. kal.kum 0 21 14
c. məl.yun 1 5 29
d. DRəb.kum 0 27 8
e. wal.mək 0 27 8
f. naw.ya 1 28 6
g. bab.ha 0 26 9
h. li.mun 0 23 12
i. da.ruh 1 13 21
j. li.kum 1 16 18
k. məl.məl 1 23 11
l. wəl.dək 1 14 20
m. ma.yəl 1 23 11
n. ʕu.dək 0 26 9
o. Rəm.la 0 34 1
p. Səm.na 0 26 9
q. wa.lu 1 28 6
r. Da.Ru 0 31 4
s. li.na 0 32 3

The first remark that could be made about the results in the tableau above is that the subjects
did not hesitate in stressing the penultimate syllable if the word consists of a succession of
two open syllables of the type CVCV. Thus words such as [dáru], [wálu] and [lína] take
penultimate stress. Such is also the case when the word consists of a succession of a closed
syllable of the type CəC and an open syllable of the type CV (cf. [Rəmla] and [Səmna], for
example), or a succession of a CV syllable and a CəC syllable (cf. [máyəl], [ʕúdək]). This
points out to the fact that CV behaves like CəC, a fact that is phonetically justified (Benkirane
1982, Nejmi 1993). Words on the pattern CəCCəC show a variation between penultimate and

115
ultimate stress. Here again, the instrumental test undertaken in Nejmi (1993) confirms this
and shows that words on the pattern CəCCəC may receive peultimate stress or final stress
([mə́lməl]/ [məlmə́l]).
The words which do not conform to penultimate stress are [lawyín], [məlyún] and
[darúh] in which stress falls on the final syllable. The final syllable of these words is closed,
that is heavy, something that shows that syllable weight might be the reason why stress fails
to apply to the penultimate syllable. Does that mean that stress in CMA prefers heavy
syllables and that these syllables must be word final? The answer could only be positive at
considering the three examples above. However, words such as [kálkum], [líkum] and [límun]
seem to contradict this statement. Given that the final syllable in these words is heavy, one
should expect it to bear stress. Take for example the words [kalkum] and [likum]. The final
syllable in these words is the object clitic
[-kum]. As it has been shown in the literature on stress, clitics tend to show irregularity as far
as the stress pattern of the language is concerned (see Benhallam 1990b). Assuming this to be
correct and that words such as [kalkum] and [likum] receive penultimate stress because [kum]
is a stress neutral clitic, we are left with the word [límun] which does not stress a final heavy
syllable. The reason might be that the subjects might have misidentified it with the words that
contain clitics in them such as [kalkum] or that are themselves clitics such as [likum]. It is
only the instrumental test that will determine the exact location of stress in this word.
The generalization that we could come up with at considering the tableau in 13 and
the discussion that follows is that stress falls on the ultimate syllable if it is heavy; otherwise
on the penultimate. The exception to this generalization are the words [likum], [kalkum] and
[limun] which fail to stress a final heavy syllable.
Now let us consider the results obtained from trisyllabic words to see if they go along
the generalization stated about disyllabic words or whether they form a special pattern.

3.1.5.2 Trisyllabic Words

Trisyllabic words seem to confirm, to a large extent, the generalization made in the
previous section, namely that stress falls on the penultimate syllable if the final is not heavy.
Consider the results listed in the tableau below:

116
-14-

ITEMS NR σ1 σ2 σ3
a. law.yin.hum 2 0 22 11
b. man.Da.Rin 1 7 16 11
c. wal.mu.kum 1 9 17 8
d. bəZ.Tam.kum 0 0 30 5
e. mər.məd.nak 0 2 7 26
f. ru.bəl.kum 0 11 11 13
g. ʕəR.ga.nin 2 5 20 8
h. mqul.bi.nək 2 15 14 4
i. wəl.dat.ni 0 3 31 1
j. min.ʒa.ra 0 9 23 3
k. ban.ya.ha 0 11 20 4
l. rub.lu.ni 1 10 19 5
m. gən.Du.Ra 1 14 20 0
n. yəD.Rəb.na 0 7 24 4
o. məl.mlu.na 0 4 29 2
p. li.mu.na 0 8 24 3
q. ʕi.ni.na 1 18 12 4

Out of 17 items, 12 are stressed on the penultimate syllable by a maximum total of 28

subjects as shown by words such as [wəldátni] in 14i and a minimum total of 6 subjects as in
the word [gənDúra] in 14m. The exceptional cases are given below where the letters
correspond to the items given in 15:

-15-
b. maDáRin
f. rubəlkúm
g. ʕəgánin
h. mqúlbinək
q. ʕínina

117
It should be noted that 15h and 15q are the only items that show antepenultimate stress, where
penultimate stress is expected. In 15h, it could be argued that the difference in the number of
responses between penultimate and antepenultimate positions is not significant (1 response)
and that the failure of subjects to score high in penultimate position may be attributed to the
fact that the syllable is closed and contains a full vowel as opposed to the penultimate and
ultimate syllables which are light. In 15q, there is no way whatsoever to show why stress
falls on the antepenultimate syllable instead of the penultimate. Is it because of the existence
of 3 light syllables in a row? If this is so, then we should expect stress to shift to
antepenultimate syllable in words such as [limuna], [məlmluna] and [gənDura], which show
regular penultimate stress (recall that CəC is equivalent to CV). Because it is the only
counterexample, it should be discarded and the generalization be made on the basis of words
with similar syllabic composition. The same thing could be said about [rubəlkúm] which is
the only item where the clitic [-kum] is stressed. We are left with 15b and 15g which have
penultimate stress instead of ultimate. These words should be added to [limun] which are the
remaining cases that do not stress a final heavy syllable. These are left aside for the moment
until the instrumental test is carried out to affirm whether or not they should abide by the
general stress pattern of the language.

3.1.5.3 Polysyllabic Words

These include four- and five-syllable cliticized words. The scores obtained from the
test are listed in the tableau in 16:

-16-

ITEMS NR σ1 σ2 σ3 σ4 σ5
a. ban.ya.ha.lih 6 4 8 3 14 ─
b. di.ri.ha.li.ha 6 1 9 8 5 6
c. gəl.bu.ha.li.kum 5 4 8 9 4 5

These items are characterized by the highest number of no responses, a fact that explains why
the subjects are hesitant about which syllable should be considered as most prominent. The

118
item in 16a stresses a final heavy syllable and thus conforms to the general tendency. The
remaining items are problematic because stress fails to fall on the penultimate syllable. One
possible way of explaining this is to say that subjects conceive of these words as phonological
phrases that consist of independent words, each of which should be stressed on its own 7.
Thus [banyahalih] consists of [banya] and the clitic [lih], [dirihaliha] consists of [diriha] and
[liha] and finally [gəlbuhalikum] consists of [gəlbuha] and [likum]. If this decomposition is
correct, the stems in 16a and 16b have stress on the appropriate location while the stem in 16c
has stress on the object clitic [ha] instead of the preceding syllable. Here again this example
might be neglected on the basis of the total number of responses obtained (9 to the actual
stressed syllable as opposed 8 obtained for the syllable which ought to have been stressed).
To sum up, the tendencies that seem to account for stress in the 39 items chosen in
this quantitative test could be stated as follows:

-17-

a. In disyllabic and trisyllabic words, stress falls on the penultimate syllable if the
final syllable is not heavy or is not one of the object clitics [kum] and [hum].

b. In polysyllabic words, if the word contains independent clitics such as [likum],

[lihum] or [liha], then stress the stem according to 17a.

Having quantified the intuition of the native speaker about the location of stress, let us
now turn in the next section to see whether or not the instrumental test confirms the results
obtained in the quantitative test.

3.2 The Instrumental Test

Prior to this test, we conducted an instrumental analysis whose purpose was to

replicate the results of Nejmi (1993)8. The results obtained match up to a large extent those of
Nejmi. However they cannot be used as a reference since they rely on a single subject (the

7
Clitics in MA fall into two categories: independent clitics such as [liya] “to me”, [lik] “to you” and [lih] “to
him” versus dependent clitics such as [-li], [-lək] and [-lu] which occur mostly with verbs (see Harrell (1962) for
further details about the distribution of clitics)
8
I would like to thank Lisa Selkirk for helping me carry out this test and also for initiating me into the domain
of instrumental phonetics at the University of Massachusetts laboratory. Thanks also go to John Kingston for
helping me out with laboratory work.

119
author of this dissertation) and as such might threaten the external validity of the test as a
whole.
The objective of the present test is to try and validate or invalidate the results obtained
in the quantitative test and this by comparing the values for fundamental frequency (F0),
intensity (IT) and duration (D) to find out which of these three parameters is the determining
factor in stress.

3.2.1 The Corpus

The corpus consists of 28 items taken out of the data presented in section 3.1.3 above.
It was organized in such a way to include disyllabic, trisyllabic and polysyllabic items of both
simple and affixed forms. Such a distinction aims at considering the effect of affixation
and/or clicticization on word stress.
The corpus subjected to the instrumental analysis is given in 7 above and repeated in
18 below:

-18-

A. Disyllabic words

a. CVC.CVC lawyin
kalkum
b. (C)CəC.CVC məlyun
DRəbhum
c. CVC.CəC walmək
d. CVC.CV babha
e. CV.CVC limun
daruh
f. CəC.CəC məlməl
wəldək
g. CV.CəC mayəl
h. CəC.CV Rəmla
i. CV.CV walu
lina

B. Trisyllabic words

a. CVC.CVC.CVC lawyinhum
b. CVC.CV.CVC manDaRin
c. CəC.CəC.CVC mərmədnak

120
 d. CəC.CV.CVC ʕəRganin
e. CCVC.CV.CəC mqulbinək
f. CəC.CVC.CV wəldatni
g. CVC.CV.CV banyaha
h. CəC.CV.CV gənDuRa
i. CəC.CəC.CV yəDRəbna
j. CəC.CCV.CV məlmluna
k. CV.CV.CV limuna

C. Polysyllabic words

a. CVC.CV.CV.CVC banyahalih
b. CV.CV.CV.CV.CV dirihaliha
c. CəC.CV.CV.CV.CVC gəlbuhalikum

3.2.1 The Subjects

Four among the 35 subjects retained from the quantitative test, were chosen for this
test: three males and one female. As it has already been stated in section 3.1.4 above, all the
subjects were born in Casablanca and are still living there. The reason behind choosing
Casablanca subjects is to ensure a maximum degree of homogeneity. The subjects were given
the target words in 18 above and were asked to make three repetitions of the target words
first in isolation, and then in a context: (gal [target word] ʒuʒ məRRat) “he said [...] two
times/he said [...] twice”. The objective of following this procedure is to see whether or not
stress placement varies according to whether the word is in isolation or in context.

3.2.2 Words in Isolation

The items considered in the test have been pitch-tracked using a program called Win-
Pitch developed by Philip Martin in (1996). The values for F0, IT and D are given in the
charts in Appendix C; the F0 pitch-tracks themselves along and IT curves as well as the
sound waves are given in Appendix D.
The stress patterns of CMA obtained from test items in isolation are represented by
the following items:

121
-19-

a. Final stress

lawyín
limún
məlyún
manDaRín
mərmədnák
banyahalíh

b. Penultimate stress

kálkum
bábha
mə́lməl
máyəl
Rə́mla
lawyínhum
mqulbínək
wəldátni
limúna
dirihalíha

The data above show that CMA makes recourse to syllable weight in the assignment of stress.
In terms of syllable weight, it has been shown in chapter two that CMA distinguishes between
bimoraic heavy syllables (CVC) and monomoraic light syllables, which fall into two types:
one where the mora dominates one segment (CV); the other where the mora dominates the
schwa and another consonant (CəC). The minor syllable is not included since it does not
receive stress by virtue of its being dominated by a mora (see section 4 below) The examples
also show that the domain of stress is restricted to one of the last two syllables of the word.
Thus a word receives final stress if it ends up in a heavy syllable (e.g. 19a). If the final
syllable is light, stress falls on the penultimate syllable, be it light or heavy (e.g. 19b).
The generalization governing stress in words in isolation could be stated as follows:

-20-

When the word occurs in isolation, stress falls on the final syllable of that word if it is
heavy; otherwise, on the penultimate.

122
This statement is generalizable to all the items except those that end up in the object clitics [-
kum] and [-hum] which have the shape CVC and yet do not receive stress in spite of their
being in final position. As has already been pointed out in the quantitative test, two out of
three items that end up in these clitics have their penultimate syllable stressed instead of the
final. This shows that the clitics [-kum] and [-hum] could either be stress neutral and
therefore lie outside the stress domain or behave as if they were light syllables. The first
thesis cannot be defended since, as it will be seen in the next section, it can receive stress
when the word occurs in context. We are left with the second, namely that the two clitics
should be treated as light syllables. It should be noted here that in other varieties of MA,
namely the ones spoken in the north of Morocco (Hammari 2000), the two clitics appear as
[kəm] and [həm], i.e. light syllables.
The instrumental test undertaken on words in isolation confirms the native speakers’
intuitions about stress and therefore the results obtained from the quantitative test. The
doubtful cases, which were found not to correspond to the generalizations in 17, should now
be fully integrated on the basis of this instrumental test.
Having shown the stress pattern of isolated words, let now now consider which of the
three parameters of F0, IT and D is the determining factor in stress placement. Previous
instrumental work on MA stress such as Hammoumi (1988) and Nejmi (1993) ascertain that
it is always F0 in combination with either IT or D that determines stress.
In the present work, the analysis of stress offered is based essentially on the
distribution of the F0 peaks in spite of the fact that some of the items analyzed in the
instrumental test confirm to some extent the results of our predecessors concerning the
determining factors of stress. A look at the charts in Appendix D shows that stressed syllables
in some words are generally characterized by higher values for F0, IT and D compared with
unstressed syllables. Such is the case in most of the words in isolation (see Appendix D for
the results). The values of the word [Rəmla] (corresponding to #12 in Appendix D) for the
three parameters are given in the chart in 21. The numbers from 1 to 4 correspond to the
subjects. The first three are males; the last is a female. F0 is measured in terms of Hertz (Hz),
IT is measured in terms of decibel (db) and D is measured in terms of millisecond (ms).

123
-21-

F0 Rəm la IT Rəm la D Rəm la

S1 144 113 S1 34 32 S1 232 198

S2 144 85 S2 33 28 S2 190 144
S3 174 123 S3 27 21 S3 196 154
S4 265 179 S4 27 21 S4 220 144

The values for F0, IT and D, obtained from the four subjects, show that it is the first syllable
in [Rəmla] which is stressed. The minimum difference between the stressed and unstressed
syllables for the three parameters is significant: 31 in F0, 2 in IT and 34 in D (the numbers are
obtained by substracting the values for the unstressed syllable from those of stressed
syllables). The maximum difference between the two syllables, scored by S5, the female
subject, is also significant: 77 in F0, 6 in IT and 76 in D.
Further support for the interaction of the three parameters in stress comes from words
such as [kalkum]. The chart below gives the values for F0, IT and D:

-22-

F0 kal kum IT kal kum D kal kum

S1 128 93 S1 37 26 S1 304 187
S2 141 121 S2 33 27 S2 213 236
S3 167 117 S3 28 24 S3 262 255
S4 240 242 S4 31 29 S4 289 278

The conclusion we can draw from these results about items in isolation is that the
combination of F0 with IT and/or D is important in deciding about stress, an assumption
which is hard to defend given the patterns obtained from stress of words in isolation to which
we turn in the next section.

124
 3.2.2 Words in Context

In order to determine whether or not stress varies according to context, we carried out
an experiment parallel to that undertaken for words in isolation. The items analyzed in this
experiment are the ones listed in 18 in the previous subsection. The subjects used for the
recordings are also the same, except for S5 who figures in this experiment but not in the
previous one 9. The subjects were asked to go through the target words and read them within
a predetermined context which is: [gal [target word] ʒuʒ məRRat] “he said [...] two times/he
said [...] twice”. The recordings have been submitted first to a process whereby the target
words decomposed out of their context and then pitch-tracked, as we did for isolated words in
the previous experiment. The values for F0, IT and D are given in Appendix E; the F0 pitch-
tracks, the IT curves and the sound waves are given in Appendix F.
The results obtained from the experiment on words in context are different from those
obtained from words in isolation in that stress consistently falls on the final syllable of the
word as the representative items below show:

-23-

a. Final heavy
lawyín
məlyún
ʕəRganín
banyahalíh

b. Final light
babhá
mqulbinə́k
wəldatní
kalkúm
lawyinhúm

The set in 23a confirms the generalization made in 20, namely that a final heavy syllable
receives stress. However, the items in 23b show that even if the final syllable is light, it is
stressed instead of the penultimate syllable despite being heavy. Notice that the clitics [kum]

9
In fact this subject was consistent in as far as stress is concerned. The patterns obtained from this subject are
consistently stressed on the final syllable and this explains why her values were not included in the charts for
words in isolation.

125
and [hum] are considered as light, a proposal that emanates from their behavior when the
word occurs in isolation.
The stress generalization about words such as those in 23 could be stated as follows:

-24-
When the word occurs in context, stress falls on the final syllable.

Having determined the location of stress in words in context, let us now consider the
interaction of the three parameters of F0, IT and D. For words in isolation, it has been found
out that a stressed syllable, whether light or heavy, may combine all the three parameters.
Such is also the case for a certain number of words in context which end up in a heavy
syllable. Consider the values obtained from the word [məlyún] in the chart below:

-25-

F0 li mun IT li mun D li mun

S1 128 154 S1 22 26 S1 187 322
S2 137 173 S2 25 29 S2 226 315
S3 131 164 S3 24 27 S3 210 318
S4 216 262 S4 31 33 S4 176 321
S5 198 242 S5 23 34 S5 256 319

The stressed syllable in this word is characterized by higher values for F0, IT, and D. The
minimum difference in value between stressed and unstressed syllable is 26 in F0 (scored by
S1), 2 in IT (scored by S4) and 63 in D (scored by S5). The maximum difference is 46 in F0
(scored by S4), 11 in IT (scored by S5) and 145 in D (scored by S5). These numbers confirm
our previous conclusion, namely that stress involves an interaction among all the three
parameters.
This conclusion seems to be contradicted by other items where unstressed syllables
get higher values for intensity than stressed ones. Consider the values for the item [lawyín]
(corresponding to #1 in Appendix E) given in the chart below:

126
-26-

F0 law yin IT law yin D law yin

S1 120 164 S1 28 24 S1 247 283
S2 131 176 S2 35 27 S2 317 347
S3 122 170 S3 30 28 S3 225 312
S4 201 268 S4 27 25 S4 256 320
S5 204 260 S5 36 32 S5 286 340

For ll of the five subjects, the values of F0 and D for stressed syllables are higher than those
for unstressed syllables. As to IT, it is higher in the unstressed syllables. The reason may be
attributed to the quality of the vowel: [a], being open, is more sonorous than the closed vowel
[i]. It seems that the distinction between vowel quality works only when the word is in
context; when the word is stressed in isolation, vowel quality does not seem to play any major
roles as the results in Appendix C show.
Other items in context show that some stressed syllables may get higher values for F0
but not for IT and D. In fact most of the items of words in context show that it is only F0 that
is determinant in stress (e.g. [mqulbinək], [wəldatni], [banyaha] and [gənDura]). Compare the
following values for IT and D of the stressed syllables in words such as [walmək]:

-27-

F0 wal mək IT wal mək D wal mək

S1 127 170 S1 29 23 S1 258 233

S2 139 179 S2 32 25 S2 336 250
S3 132 178 S3 30 24 S3 279 240
S4 207 264 S4 38 33 S4 285 269
S5 205 233 S5 35 31 S5 293 270

On the basis of cases such as [walmək], we conclude that it is F0 which is the determining
factor in stress assignment. Even when we take into account the intrinsic effect of vowels (/a/
is lower and /i/ and /u/ are higher) there is a clear pattern that it is F0 (and not IT or D or a

127
combination of the two with F0) which is consistently high in all the stressed syllable. In fact,
D is much more variable in the data analyzed and as such there are real difficulties in
interpreting the results because of the differing intrinsic durations of the consonants and
vowels in the syllables being compared. Even in cases where the syllables being compared
have the same constituents, we cannot build any real pattern if we have to do it on the basis of
D. The word [məlməl] is a clear-cut case for discarding the parameter of D; the second
syllable of the word gets a higher value for D whether it is stressed as is the case when the
word is in context, or unstressed when the word is in isolation. (Compare # 9 in Appendix F
and # 9 in Appendix D).
To sum up, the instrumental test we have carried out has allowed us to find out about
the location of stress in CMA. Both words in isolation and words in context have been
analyzed. The experiment on words in isolation has generally confirmed the major results
obtained from the quantitative test, namely that stress is triggered by two factors: syllable
weight and syllable position. It has been shown that a final heavy syllable gets stress. In the
absence of a final heavy syllable, stress falls on the penultimate syllable. It has also been
shown that the clitics [kum] and [hum] should be considered as light and are therefore not
stressed in final position. The subsection has also considered the interaction of F0, IT and D
as far as stress is concerned. We have come to the conclusion that it is the distribution of the
F0 peak which is decisive in stress. The F0 patterns may not simply be a reflex of stress. In
English, for example Pierrehumbert (1980) has shown that it is the main-stressed syllable of
the word which is assigned a high pitch accent. Similarly, we think that the F0 peaks in CMA
may well indicate the presence of a high pitch accent that associates with the main stressed
syllable of the word. This conclusion is further corroborated by a systematic study of
intonation in CMA undertaken in Mawhoub (1990).
The experiment on words in context has shown that stress falls consistently on the
final syllable of a word regardless of whether or not it is heavy. As to the three parameters
associated with stress, it has been shown that some stressed syllables may combine these
three prameters or combine F0 with either IT or D; whereas others show only higher values
for F0, a fact which has led us to conclude that F0 is the determining factor in stress in
context words.

128
 Having established an experimental basis for stress assignment in CMA, let us turn in
the following section to see how the variation in stress could be accounted for within the OT
framework.

4. WORDS IN ISOLATION AND WORDS IN CONTEXT: A UNIFIED

ACCOUNT OF THE STRESS SYSTEM OF CASABLANCA
MOROCCAN ARABIC

The results obtained from the instrumental test about words in isolation show that
CMA is a quantity sensitive language with trochaic feet. Those obtained about words in
context show that the stress pattern of the language is basically iambic. According to Prince
and Smolensky (1993) and McCarthy and Prince (1993a), feet are subject to the constraint
FT-BIN which demands that they be binary under syllabic analysis if the language in question
is quantity-insensitive, or moraic if it is quantity-sensitive. The constraint, given in chapter
two, is repeated in 29 below:

-28-

FT-BIN
Feet are binary under syllabic or moraic analysis.

The constraint in 28 is generally in conflict with the constraint PARSE-σ which demands that
syllables be parsed into feet:

-29-
PARSE-σ
Syllables must belong to a foot.

The conflict between PARSE-σ and FTBIN arises only when the word concerned contains an
odd number of syllables. In CMA this conflict is resolved by ranking FTBIN above PARSE-σ
as shown by the foot structure of [limúna] in the tableau below (Feet are enclosed between
parentheses):

129
-30-

FTBIN PARSE-σ
) a. li.(mu.na) *
b. (li.)(mu.na) *!
The tableau shows that any form violating FTBIN is ruled out. In chapter two, section 3.3, we
have presented evidence for the non-foot status of a light syllable of the type CV or CəC.
Therein, and with the exception of the two words [ʒa] “he came” and [ma] “water” (which
themselves show augmentation to achieve FTBIN, cf. [ʒay] and [lma], “the water”
[myah]/[mihan] “waters”) and other function words, it has been shown that no lexical word in
CMA consists of CV or CəC. Thus the parse in 30b is ruled out exactly because the first
syllable of the word, which happens to be light, cannot form a foot on its own.
The results obtained from the instrumental test confirm Nejmi’s (1993, 1995) finding,
namely that CMA stresses one of the last two syllables of a word. This means that the foot
which contains the stressed syllable must be at the right edge of the prosodic word. It also
shows that the directionality of footing is from right to left. Within the OT framework,
directionality is captured by making recourse to McCarthy and Prince’s (1993b) Generalized
Alignment Theory. In particular, the constraint we will be using is the one requiring
alignment of the right edge of every foot with the right edge of the prosodic word:

-31-

ALIGN-R (Ft, PWd)

The right edge of every foot must be aligned with the right edge of the prosodic
word.

Violations of ALIGN-Ft-R are reckoned gradiently in terms of lower constituents, i.e.

syllables. Thus a foot structure such as [(σ σ) (σ σ) (σ σ)]PWd incurs 6 violation marks of
ALIGN-R (Ft, PWd). Only the rightmost foot obeys the constraint. In CMA, PARSE-σ
should be allowed to dominate ALIGN-R (Ft, PWd) to derive the correct foot structure as
shown in 33 by the parsing of the word [gulihaliha] which receives penultimate stress when it
occurs in isolation:

130
-32-

FTBIN PARSE-σ ALIGN-R (Ft, PWd)

) a. gu.(li.ha)(lí.ha) * **
b. gu.li.ha.(lí.ha) **!*
c. (gu)(li.ha)(lí.ha) *! ******

The parse in 32b, although it observes ALIGN-Ft-R, is not optimal because three of the
syllables are not parsed. 32c is also excluded because a monomoraic syllable is parsed into a
foot, something that violates FTBIN.
The universal repertoire of feet includes two types: Trochaic feet and Iambic feet.
These types are distinguished in terms of headedness. Iambic feet are right headed, that is the
stressed syllable is located at the right periphery of the foot (Prince and Smolensky, 1993,
McCarthy and Prince, 1993a, Hayes, 1995, among others). Trochaic feet are leftheaded and
are further distinguished into moraic trochees which consist of heavy bimoraic syllable or a
sequence of two light syllables where the one to the left bears stress, and syllabic trochees in
which the foot counts only syllables regardless of their internal structure. According to Hayes
(1995) the universal foot types are given in 33 below. Headedness is shown by the accent (′)
after the designated element.

-33-

Foot Types
a. Iamb: L H′, LL′ or H′
b. Moraic Trochee: L′ L or H′
c. Syllabic Trochee: σ′ σ

The CMA patterns about isolated words show that feet must be trochaic and therefore
the system is quantity sensitive since it is the heavy syllable of the word which receives
stress. If there is no heavy syllable, it is the penultimate syllable of the foot which is stressed.
This could be captured in terms of a constraint which demands that the foot be trochaic. The
constraint, labeled TROCHEE, is stated as follows:

131
-34-
TROCHEE
Feet are left-headed.

In a disyllabic word such as [sala], where both syllables are light, initial stress is determined
by the constraint TROCHEE as shown in the tableau below:

-35-

/sala/ TROCHEE
) a. (sá.la) √
b. (sa.lá) *!

The same word receives final stress when it is put in context. This shows that the foot must
instead be iambic. The relevant constraint is given in 36 below:

-36-

IAMB
Feet are right-headed.

Thus a sequence of a light syllable and a heavy syllable will have to be footed as (LH′) and
not as L(H′) where stress falls on the syllable on the right. In the same way, a sequence of two
light syllables is footed as (LL′) but not (L′L) as the tableau below shows:

-37-

/sala/ IAMB
a. (sá.la) *!
b. (sa.lá) √

The constraint in 37, if satisfied will always stress the rightmost element in LL and LH, thus
giving rise to iambic feet rather than trochaic ones. However, and as it has been pointed out,
trochaic feet do exist in the language but arise only under certain conditions as will be shown
as we proceed down in the analysis. This means that the results obtained from the

132
instrumental test lend themselves to two possible analyses in terms of foot structure: one
where the foot is trochaic (i.e. words in isolation); the other where the foot is iambic (i.e.
words in context). The challenging question we will try to answer in this section is the
following: how could the OT apparatus be made use of to analyze the conflictual stress
patterns exhibited in isolated words and words in context? Put differently, is there a way out
for a fixed ranking of constraints that would allow us to account for stress in both words in
isolation and in context?
The questions raised here relate to the apparent problems a non-OT analysis faces
given the stress patterns that the CMA data exhibit: an iambic pattern and a trochaic pattern.
This apparent paradox is hard to account for given a parametrized theory which forces a
language to choose between iambs and trochees. At this point, it should be made clear that
OT differs from earlier treatments within parametrized theories in allowing for both iambs
and trochees in the same language, in conditions where higher order constraints end up in
determining whether one or the other will appear. McCarthy and Prince (1993a:150), for
example, shows that the basic stress pattern of Axininca Campa, an Arawakan language of
Peru, is both iambic and trochaic as the examples below show:

-38-

a. Iambic feet
(hinó)ki arriba (por el río)
(iráa)(wanì)ti su caoba
(apà)(nirói)ni solo

b. Trochaic feet
(círi) brea de àrbol
(máto) polilla
(chími) colpa

According to McCarthy and Prince (1993a) Trochaicity is derived from the fact that final
syllables, except those containing diphthongs, are always unstressed as is the case with the
disyllabic items in 38b. The authors, citing Prince and Smolensky (1992, 1993), show that the
basic pattern of iambic stress with final stresslessness can be obtained from the interaction of
FT-BIN with three other constraints which are FT-FORM (= IAMB), PARSE-σ and NON-
FINALTY. (See McCarthy and Prince 1993a for a more detailed analysis of this seemingly
paradoxical stress pattern of Axininca Campa).

133
 In order to account for the stress system of CMA, we assume that the basic stress
pattern of the language is iambic and that trochaic feet arise under certain conditions. That
iambic feet are basic can be obtained by ranking IAMB above TROCHEE. The effect could
be seen in the tableau below for the parses of the input [sala]:

-39-

IAMB TROCHEE
)a. (sa.lá) *
b. (sá.la) *!

However and as it has already been shown in the previous section, trochaic feet of the type in
39b do exist in CMA. In fact, the two parses in 39 should be allowed to be optimal. In order
to allow for both parses to be optimal, we have to define the various levels of phrasing at
which trochaic or iambic feet are obtained, a fact which calls for a theory of Prosodic
Structure. Works which argue for prosodic structure as the basis for a theory of sentence
phonology include Selkirk (1978, 1980, 1984, 1986, 1995a); Nespor and Vogel (1982,
1986); Beckman and Pierrehumbert (1986), etc ... According to Prosodic Structure Theory,
sentences are organized into a structure whose categories are defined by Selkirk (1978) as
follows:

-40-

The Prosodic Hierarchy

Utt Utterance
IP Intonational Phrase
PPh Phonological Phrase
PWd Prosodic Word
Ft Foot
σ Syllable

This hierarchy of prosodic categories forms the essence of the theory of phonological
constraints on prosodic structure. Selkirk (1995:5) states the constraints on prosodic
domination as follows:

134
-41-

(i) Layeredness: No Ci dominates Cj, j>i,

e.g. No σ dominates a Ft.

(ii) Headedness: Any Ci must dominate a Ci-1 (except if Ci = σ),

e.g. A PWd must dominate a Ft.

(iii) Exhaustivity: No Ci immediately dominates a constituent Cj, j<i-1,

e.g. No PWd immediately dominates a σ.

(iv) Nonrecursivity: No Ci dominates Cj, j = i,

e.g. No Ft dominates a Ft.

It is the constraints on prosodic domination along with the prosodic hierarchy and their
interaction with other constraints on the prosodic hierarchy that will allow us to derive both
iambic and trochaic feet. In the CMA case, the parse in 39b is possible only when the word is
in isolation; or to put it in Selkirk’s (1978) terms when the word is at the end of a PPh. On the
other hand, the parse in 39a is possible when the word is in context, i.e. when it is non-final in
a PPh. Thus a word such as [lina] will have either one of the structures in 42 below:

-42-

a. In isolation: PPh{[salá]PWd}PPh
b. In context: PPh{[gal]PWd [salá]PWd [ʒuʒ]PWd [məRRat]PWd}PPh

In phrase final position, [lina] will be assigned a trochaic foot; in context, i.e. when the same
word is not in phrase final position, it will be assigned an iambic foot. To keep stress off
word-final syllables, Prince and Smolensky (1993:40) posit the constraint NON-FINALTY
which they formulate as follows:

-43-

NON-FINALTY
The prosodic head of the word does not fall on the word-final syllable.

However, the problem with this version of the constraint NON-FINALTY is that it can derive
only trochaic feet in words such as [lina] but never iambic ones, a fact which calls for

135
reconsidering this constraint in such a way that it refers to main-stress at a level of
phonological phrasing above the PWd.
The desired version of NON-FINALTY we propose is the one that refers to non-
finalty in a PPh. This constraint could be expressed as follows:

-44-

NON-FINALTY (σ′, PPh)

The main-stressed syllable of a word is not in phrase final position.

The constraint NON-FINALTY (σ′, PPh) will have to dominate IAMB to be able to derive
phonological phrases with penultimate stress. It will also have to dominate TROCHEE
because, as it will be shown below, satisfying NON-FINALTY (σ′, PPh) does not guarantee a
trochaic foot. In the tableau in 45 we show how a word such as /lina/ is assigned stress
depending on whether it is in isolation or in context. (The straight bars mark PWd boundary).

-45-

NON-FINALTY IAMB TROCHEE

(σ′, PPh)
a. {|(li.ná)|PWd}PPh *! *
)b. {|(lí.na)|PWd} PPh *

The candidate in 45a which appeals to the phrasal organization in 40 has stress on the final
syllable and therefore conforms to an iambic rather than a trochaic foot. This leads to a fatal
violation of the constraint requiring the main-stressed syllable to be nonfinal within a PPh. In
the optimal candidate, stress falls on the penultimate syllable which happens to be nonfinal
within the PWd and also within the PPh.
The same word in context would get stress on the final syllable and consequently an
iambic foot rather than a trochaic one as the tableau in 46 shows. The dots show that the word
is put in context.

136
-46-

NON-FINALTY IAMB TROCHEE

(σ′, PPh)
)a. {...|(li.ná)|PWd...}PPh *
b. {...|(lí.na)|PWd...} PPh *!

The constraint ranking given in 46 allows for a straightforward account of the contradictory
stress system of CMA. It also shows that the prosodic category PPh emanating from the
hierarchy in 40 above is very determinant in stress assignment.
Other disyllabic words that are of interest to us here are words that consist of a
sequence of the type CVCVC, that is to say a light syllable followed by a heavy syllable. The
foot structure of such words may be either (CVCVC) or CV(CVC). In (CVCVC) both
syllables are parsed into a single iambic foot; in CV(CVC) only the second syllable is parsed
into a foot, a fact which constitutes a violation of PARSE-σ; the remaining syllable is
adjoined directly to the PWd. The unparsed syllable cannot form a foot on its own because
this would constitute a violation of FT-BIN. Given these two parses, one has to determine
which one is the most optimal. Is it (CVCVC) or CV(CVC)?
Ranking PARSE-σ above TROCHEE and below IAMB would ensure that all
syllables are parsed and that in the case of conflicts between these two constraints, one has to
satisfy the first one. In the tableau below, we show how a disyllabic word such as [limun]
would be stressed when it is in phrase final position:

-47-

NON- IAMB PARSE-σ TROCHEE

FINALTY (σ′,
PPh)
a. {|(li.mún)|PWd}PPh *! *
0b. {|(lí.mun)| PWd}PPh *
c. {|li(mún)| PWd}PPh *! *

This tableau shows that satisfying both PARSE-σ and TROCHEE gives the wrong output
with stress on the light syllable rather than the heavy. This means that we need other

137
constraints to be able to get the optimal output. The relevant constraint here is Weight-to-
Stress Principle (Prince and Smolensky 1993):

-48-

Weight-to-Stress Principle (henceforth WSP)

A heavy syllable is stressed in foot structure.

WSP applies only at the foot level in order to rule out structures such as (CVC)(CV′.CVC) or
(CVC.CV′)(CVC), where the light instead of the heavy syllable of the foot which is stressed.
If a foot contains syllables that have equal weight as in (CV.CV), for example, either one
could be stressed without there being violation of WSP. By WSP, a bad trochee of the type
L′H represented by the candidate *[límun] will be avoided in favor of a parse that gives
priority to an iambic foot. To do so, we need to rank WSP above NON-FINALTY (σ′, PPh)
to make sure that it is the final heavy syllable that gets main stress. The following tableau
shows how the optimal parse is obtained from the input /limun/ when it is a PPh.

-49-

WSP NON-FINALTY IAMB PARSE-σ

(σ′, PPh)
)a. {|(li.mún)|PWd}PPh *
b. {|(lí.mun)| *! *
PWd}PPh

c. {|li(mún)| PWd}PPh * *!

Candidate 49b is excluded on the ground that it violates a higher-ranked constraint, namely
WSP. The remaining candidates both incur a violation mark for the constraint NON-
FINALTY (σ′, Ft) but it is 49a which is optimal although it violates TROCHEE, a constraint
ranked low in the ranking scale. By parsing the final heavy syllable as a foot, 49c incurs a
fatal violation of PARSE-σ.
When the word in 49 occurs in context, we would still get final stress with the
difference here that the optimal candidate 49a would spare a violation mark for the constraint

138
NON-FINALTY (σ′, Ft) since the stressed syllable would no longer be in phrase final
position as shown in 50 below:

-50-

WSP NON-FINALTY IAMB PARSE-σ

(σ′, PPh)
)a. {...|(li.mún)|PWd...}PPh

b. {...|(lí.mun)|PWd...}PPh *! *
c. {...|li(mún)|PWd...}PPh *!

Although PARSE-σ is ranked low, it is still relevant in determining the optimal output.
What the tableau in 50 shows is that stress of words in isolation does not always lead
to a trochaic foot. Given a sequence CVCVC, an exhaustive parse of this sequence into an
iambic LH foot is more highly valued than one where the initial light syllable would be left
unparsed. Further support for this claim comes from trisyllabic words such as [manDarín]
where it is the final heavy syllable which is stressed whether the word is in isolation or in
context. As a PPh, the input /manDarin/ may have the following possible parses:

-51-

WSP NON-FINALTY IAMB PARSE-σ

(σ′, PPh)
)a. {|(man)(Da.rín)|PWd}PPh *
b. {|(man)Da(rín)| PWd}PPh * *!
c. {|(man)(Dá.rin)| PWd}PPh *!

Candidate 51c is ruled out because it violates WSP by assigning stress to the light rather than
the heavy syllable of the foot. Notice here that WSP is invisible to the first heavy syllable in
all the candidates since it is the only syllable of the foot. Both candidates 51a and 51b incur a
violation mark for the constraint NON-FINALTY
(σ′, PPh) but 51b is suboptimal because it fails to parse the light syllable of the word, thus
incurring a fatal violation of PARSE-σ.
Another distinguishing characteristic about CMA stress is that only one of the last two
syllables gets stress (Nejmi, 1993, 1995). Translated into OT terms this amounts to saying

139
that the prominent foot which contains the main-stressed syllable has to be right-aligned with
the prosodic word. The constraint responsible for this is formulated within McCarthy and
Prince’s (1993b) Alignment Theory as in 52.
(Ft′ stands for the prominent foot, i.e. the foot that contains the main-stressed syllable)

-52-
ALIGN-R (Ft′, PWd)
The right edge of the prominent foot must be aligned with the right edge of
the PWd.

The constraint ALIGN-R (Ft′, PWd) has to dominate WSP in order to rule out forms such as
[(Sán)(Da.la)] where stress falls on the antepenultimate heavy syllable instead of the
penultimate syllable when the word is a PPh. Consider two of the possible output candidates
from the input /SanDala/:

-53-

ALIGN-R (Ft′, PWd) WSP

)a. (San)(Dá.la) *
b. (Sán)(Da.la) *!

The losing candidate stresses the only heavy syllable of the word which belongs to a foot
other than the final, thus fatally violating ALIGN-R (Ft′, PWd).
Since ALIGN-R (Ft′, PWd) dominates WSP which in turn dominates NON-
FINALTY (σ′, PPh), it follows that ALIGN-R (Ft′, PWd) must dominate NON-FINALTY
(σ′, PPh), by transitivity. To illustrate this domination relation, consider a sequence such as
CVCVCV which is footed as CV(CVCV) with stress on either the penultimate if the target
word is a PPh, or else on the final syllable if the target word is in the middle of a PPh but
never (CVCV′)CV or (CV′CV)CV as shown in the tableau in 54 for the different parses of a
word in isolation such as [banana]:

140
-54-

ALIGN-R NON-FINALTY IAMB TROCHEE

(Ft′, PWd) (σ′, PPh)
)a. {|ba(ná.na)|PWd}PPh *
b. {|ba(na.ná)|PWd}PPh *! *
c. {|(ba.ná)na|PWd}PPh *! *

Both Candidates 54b and 54c fail to conform to a trochee; however, each is excluded for
different reasons. 54b fails because it is assigned final stress at the end of a PPh, a fatal
violation of NON-FINALTY (σ′, PPh). 54c is excluded because it fails to parse the rightmost
syllable thus incurring a fatal violation of the constraint requiring the alignment of the
prominent foot and the PWd.
When in context, the same word would get final stress instead of penultimate. Witness
the tableau below for illustration.

-55-

ALIGN-R NON-FINALTY IAMB TROCHEE

(Ft′, PWd) (σ′, PPh)
a. {...|ba(ná.na)|PWd...}PPh *!
)b. {...|ba(na.ná)|PWd...}PPh *
c. {...|(ba.ná)na|PWd...}PPh *! *

The optimal candidate is the one where main stress falls on the final syllable, giving rise to an
iambic foot in which the prominent foot is right-aligned with the PWd. Once again, the
constraints at our disposal allow us to account adequately for stress cases involving both
words in isolation and words in context.
Next, consider another trisyllabic word that consists of a sequence of two light
syllables followed by a final heavy syllable. The tableau in 56 lists the different candidate
parses for the word [ʕəRganin] in isolation:

141
-56-

ALIGN-R WSP NON-FINALTY IAMB PARSE-σ

(Ft′, PWd) (σ′, PPh)
)a. {|(ʕəR.ga)(nín)|PWd }PPh *
b. {|ʕəR(ga.nín)|PWd }PPh * *!
c. {|ʕəR(gá.nin)|PWd }PPh *! * *
d. {|(ʕəR.gá)(nin)|PWd }PPh *! * *

Both candidates 56a and 56b assign stress to the final heavy syllable, thus violating NON-
FINALTY (σ′, PPh). But because candidate 56a has parsed all of its syllables into feet, it is
optimal given the fact that PARSE-σ dominates ALIGN-R (Ft, PWd) in 32 above. 56c is
excluded because it stresses the light instead of the heavy syllable of the foot; and finally 56d
is also excluded because the right-hand foot does not contain the main-stressed syllable.
In context, the word [ʕəRganín] is assigned final stress. The optimal foot structure
corresponds exactly to that in 56a except that here the concerned word spares a violation
mark for the constraint NON-FINALTY (σ′, Ft) because it is no longer in phrase final
position.
The final cases of words that deserve special attention are words that end up in or
have the shape HL, i.e. words that consist of a heavy syllable followed by a light one. Prince
and Smolensky (1993: 59) (see also McCarthy and Prince 1993a, Hayes 1995 among others)
show that the sequence HL is absent from trochaic systems and as such feet of the type HL
are banned on the basis of rhythmic structure which favors heavy syllables at the end of
constituents. To rule out HL foot types, Prince and Smolensky (1993) propose a constraint
dubbed RHYTHMIC-HARMONY (RH-HARM), also identified as *HL (Prince and
Smolensky 1993, Cabré and Kenstowicz 1995).
One could possibly argue against the constraint *HL if FTBIN were stated in such a
way that it allows just two moras in a foot. While this solution seems plausible theoretically,
it cannot be adopted in the analysis of CMA words that end up in HL for reasons that will
become clear down in this chapter. Although the HL foot type does not figure in the inventory
in 33, it does not mean that they don’t occur at all. Works which have reported HL trochaic
foot types to exist include Kenstowicz 1994 for Bani Hassan Arabic, Cabré and Kenstowicz
1995 for Catalan, and Buckley 1994 for Manam. For CMA, we maintain that trochaic feet of

142
the type (HL) do arise for two reasons. First, to secure the undominated constraint ALIGN-R
(Ft′, PWd), we need to incorporate the final light syllable into foot structure and at the same
time avoid stressing this syllable in order not to violate WSP and NON-FINALTY (σ′,PPh) in
case we have a phrase such as [babha]. Second, given the fact that the same word in context
would get final stress, there is no way whatsoever to derive the optimal candidate since WSP
is ranked above IAMB. More than that, even if WSP were not there, we would get an iambic
foot of the type HL′ which is not reported to be part of the foot inventory (cf. 33 above).
Before we see how final stress is obtained in disyllabic words such as [babha], let us
consider how the word gets penultimate stress when it is a PPh. In order not to leave the final
light syllable unparsed, the constraint *HL has to be ranked below ALIGN-R (Ft′, PWd) and
above WSP as the three competing parses for the input /babha/ show:

-57-

ALIGN-R *HL WSP

(Ft′, PWd)
)a. {|(báb.ha)|PWd} PPh *
b. {|(bab.há)|PWd} PPh * *!
c. {|(báb)ha|PWd}PPh *!

Candidate 57c tries to satisfy *HL by unparsing the final light syllable, thus causing fatal
violation of right alignment of the prominent foot. Both candidates 57a and 57b violate *HL
and tie in everything except the decisive lower-ranked WSP which favors 57a.
This line of analysis seems to be compatible with that offered by Cabré and
Kenstowicz (1995) for certain cases of Catalan Hypocoristics Where a LHL in words such
“Domingo” is analyzed as a foot of the type (HL) by virtue of an alignment constraint
requiring coincidence of the right edge of the PWd with that of the foot. Cases like these are
referred to as prosodic trapping and are accounted for by an undominated constraint ranked
above *HL, which the authors refer to as ALIGN (PWd, Ft). This constraint has the same
effect as ALIGN-R (Ft′, PWd) for CMA cases such as [babha].
The same word in context gets final stress. As it has been pointed out above, given the
impossibility of deriving iambic feet of the type (HL′), the only and possible way of getting
the optimal output is by regarding the final light syllable as constituting a degenerate foot on

143
its own, thus violating the constraint FT-BIN. FT-BIN in this case needs to be ranked
immediately below IAMB as the tableau in 58 shows. Only the relevant constraints are
included:

-58-

ALIGN-R *HL WSP IAMB FT-BIN

(Ft′, PWd)
)a. {...|(bab)(há)|PWd...}PPh * *
b. {...|(báb.ha)|PWd...}PPh *! *
c. {...|(bab.há)|PWd...}PPh *! *
d. {...|(báb)ha|PWd...}PPh *!

The optimal candidate in 58a shows that the best we can get out of a sequence of a heavy
syllable followed by a light syllable is a sequence of two iambic feet, the second of which is
degenerate. This foot, which we will refer to as the worst iamb, is obtained by ranking FT-
BIN below WSP and IAMB. The remaining candidates are excluded for different reasons:
58b and 58c are ruled out because of *HL, whereas 58d is ruled out because the prominent
foot is not right-aligned with the PWd. Notice here that FT-BIN is violated only to secure
higher-ranked constraints. In cases where these constraints are not at stake, FT-BIN is
observed as the different parses of the word [limuna] with final stress when used in context
show:

-59-

ALIGN-R WSP IAMB FT-BIN PARSE-σ

(Ft′, PWd)
)a. li(mu.ná) *
b. (li)(mu.ná) *!
c. (li.mu)(ná) *!

Next, consider trisyllabic words of the type CVCVCCV. The foot structure of such a word
when it is in isolation (i.e. a PPh) could be one of the following:

144
-60-

a. {|CV(CV′C.CV)|PWd}PPh
b. {|CV(CV′C)CV|PWd}PPh
c. {|(CV.CV′C)CV|PWd}PPh

In the parse in 60a the sequence forms an even HL trochaic foot which we have shown to
arise in CMA only when ALIGN-R (Ft′, PWd) is at stake. The parse in 60b tries to satisfy
*HL by unparsing the final light syllable while at the same time leaving the initial syllable
unparsed, thus giving rise to a moraic trochee of the type H. The remaining candidate satisfies
*HL by unparsing the final light syllable while grouping the remaining sequence of light and
heavy syllables into an iambic foot of the type LH. The result is shown in the constraint
tableau for the different parses of the word [wəldatni] in isolation:

-61-

ALIGN-R *HL WSP NONFINALTY IAMB

(Ft′, PWd) (σ′, PPh)
)a. {|wəl(dát.ni)|PWd}PPh * *
b. {|wəl(dat.ní)|PWd}PPh * *! *
c. {|(wəl.dát)ni|PWd}PPh *!

The tableau in 61 shows that whenever it is possible, priority is given to a parse that would
lead to the right alignment of the prominent foot and the PWd even at the expense of the
constraint *HL. This is the case with candidates 61a and 61b but 61b is ruled out on the
ground that it violates WSP by assigning stress to the light rather than the heavy syllable. In
the case of 61c, it is shown that because the final light syllable cannot be integrated in foot
structure, the candidate is ruled out.
In context, the word [wəldatni] would receive final stress. Since the final stressed
syllable is light and the penultimate is heavy, one would have to expect the foot to be of the
type (HL′), where it is the light rather than the heavy syllable which is stressed. Stressing the
light instead of the heavy syllable constitutes a flagrant violation of WSP. The only way to

145
get final stress is for the final syllable to form a foot on its own as shown in the tableau below
for some candidates of the word [wəldatni]:

-62-

ALIGN-R *HL WSP IAMB FT-BIN

(Ft′, PWd)
)a. {...|(wəl.dat)(ní)|PWd...}PPh * *
b. {...|wəl(dat.ní)|PWd...}PPh *! *
c. {...|wəl(dát.ni)|PWd...}PPh *! *
d. {...|(wəl.dát)ni|PWd...}PPh *!

The tableau shows that it is more optimal to violate FT-BIN than have a foot that violates
both the anti-iambic foot of the type HL and WSP. When the word is stressed in context, *HL
is satisfied by unparsing the final light syllable and reanalyzing the HL foot type as (H)(L′) .
Such is the case with the optimal candidate 62a. 62b is ruled out because it fatally violates
*HL in addition to WSP by virtue of stressing the final light syllable. The same thing could
be said about candidate 62c, except that it spares violation of WSP by stressing the heavy
instead of the light syllable. Finally candidate 62d is ruled out because it parses the sequence
LHL as (LH′)L and assigns penultimate stress to the heavy syllable thus satisfying *HL, WSP
and IAMB but falling afoul of the constraint requiring right alignment of the prominent foot
and the PWd.
To sum up, the constraint hierarchy needed to account for CMA is given in 63 below:

146
-63-

Constraint hierarchy for CMA stress cases

ALIGN-R (Ft′, PWd)

|
*HL (RH-HARM)
|
WSP
|
NON-FINALTY (σ′, PPh)
|
IAMB
|
FT-BIN
|
PARSE-

ALIGN-Ft-R TROCHEE

The OT analysis of stress we have offered in this section has allowed us to account for
variation in the stress pattern of CMA and determine the possible foot types of the language.
When words are assigned stress in isolation, we have shown that the foot types that arise are
the universal trochaic L′L and H′ and the anti-iambic foot H′L we were forced to recognize.
The three foot types are illustrated by the words [lawyín], [limúna] and [náwya]

-64-

Trochaic feet in CMA

a. PPh b. PPh
| |
PWd PWd

F F′ F
| |
σ σ′ σ σ′ σ

µ µ µ µ µ µ µ
| | | | | | |
l a wy i n l i mu n a

147
 c. PPh
|
PWd

σ′ σ

µ µ µ
| | |
n a wy a

On the other hand, when words are stressed in context, the foot types obtained are iambic
LH′, LL′, H′and another degenerate foot of the type L′ that arises when a word end up in HL.
In 65 below, we exemplify these types by the structures of words such as [manDarín],
[limuná], [lawyín] and [babhá]:

-65-

Iambic feet in CMA

a. PPh b. PPh

... PWd PWd PWd ... ... PWd PWd PWd ...

F F′ F′
|
σ σ σ′ σ σ σ′

µ µ µ µ µ µ µ µ
| | | | | | | |
m a nD a r i n l i mu n a

148
 c. PPh

... PWd PWd PWd ...

F F′
| |
σ σ′

µ µ µ µ
| | | |
l a wy i n

d. PPh

... PWd PWd PWd ...

F *F′
| |
σ σ′

µ µ µ
| | |
b a bh a

It is noteworthy here that the degenerate foot *Ft′ can never form a lexical word on its own; it
occurs only in combination with other iambic feet of the type H as in [(bab)(há)] or LH as in
[(wəldat)(ní)]. It should also be noted that these are not the only iambic feet that occur in the
language. There is yet another type labeled minor iamb and is called so because it consists of
a major syllable preceded by a minor syllable (See chapters five and six below for details).

5. CONCLUSION
We have argued that the stress system of CMA can be accounted for by the ranking of
a small set of universal constraints. The analysis we have offered shows that in both isolation

149
and context cases, the location of stress and consequently the foot types depend on the nature
of the organization of prosodic words into phrases. The unitary account of stress in CMA
proposed in this chapter is based on the idea that both iambic and trochaic feet do occur in the
language and this depending on whether or not the word is in isolation or in context, i.e.
whether it is a PPh or a PWd within a PPh.
In a phrase with a single word, it has been shown that stress falls on the final syllable
if it is heavy; otherwise, on the penultimate. This conclusion has been reached after carrying
out both a quantitative and an instrumental test of stress. The instrumental test has provided
further support for the quantitative test whose objective was to reflect the native speakers’
intuitions about stress. The results obtained show that CMA is a quantity sensitive system
which favors trochaic feet. The OT analysis has also shown that restricting stress to the last
two syllables of a word is a result of the constraint ALIGN-R (F′, PWd) demanding
coincidence of the prosodic word and its head, i.e. the prominent foot. It has also been shown
that stressing the final heavy syllable of a word at the end of a PPh is the result of the
constraint WSP which is observed in all words except when constraint ALIGN-R (F′, PWd) is
at stake. Cases with penultimate stress are obtained by modifying Prince and Smolensky’s
(1993) NON-FINALTY in such a way as to make it refer to a level of phonological phrasing
Selkirk (1978) calls the PPh level. The constraint, dubbed NON-FINALTY (σ′, PPh) and
which appeals to the phrasal organization in 40 above, needs to be ranked below WSP in
order to derive PPhs with final stress.
In a PPh with more than one member, one word will receive final stress and the others
will receive final or no-final stress, depending on whether or not the final syllable is heavy.
The arguments presented show that iambic feet take priority over trochaic ones and that this
follows from ranking IAMB above TROCHEE. The analysis has also shown that CMA is a
language which favors exhaustive parsing of syllables into feet if this parsing does not lead to
the violation of higher-ranked constraints. Thus, because the constraint ALIGN-R (Ft′, PWd)
is undominated in CMA, we were led to recognize a trochaic foot of the type (H′L) that
violates RH-HARM (i.e. the constraint *HL) and an iambic degenerate foot of the type (L′)
which is the worst iamb we can get out of the ranking proposed in 63 above.

150
 The analysis about stress and foot structure proposed in this chapter finds support in
the prosodic morphology of CMA, namely in the passive participle and the diminutive forms
where the foot is basically iambic. Before we show how these morphological classes could be
obtained within the OT framework, we turn in the next chapter to consider cases that involve
some transderivational relations.

151
 Chapter Four

Output-Output Correspondence in Casablanca

Moroccan Arabic

1. INTRODUCTION
In Optimality Theory (Prince and Smolensky 1993), grammar is conceived of as a set
of ranked and violable structural and faithfulness universal constraints regulating the
relationship between an input and an output. The development of this theoretical model with
McCarthy and Prince (1995, 1999) has led to a new version of faithfulness constraints
formulated within CT. In this theory, correspondence is assumed to regulate not only the
relation between an input and an output but also the relation between a base form and its
reduplicative copy. In the extended version of CT, Benua (1995, 1997), McCarthy (1995,
1997), Kenstowicz (1996, 1997), Kager (1996), Burzio (1996), Basri et al (1998) and Selkirk
(1999), among others, have shown that correspondence should relate, in addition to an input
and an output and a base and its reduplicant, a derived output form and its morphologically
simple base output form. This new version of correspondence is dubbed in the CT literature
as output-output (O-O) correspondence.
The present chapter is an attempt to reanalyze the three linguistic issues of cyclicity
effects of syllabification, truncation and prosodic circumscription in CMA by making use of
the extended version of correspondence whereby correspondence can be established between
words standing in a transderivational relationship.
The first issue that will be tackled relates to the notion of cyclic syllabification. For
illustration, consider the trisegmental and quadrisegmental verb paradigms given in 1 below:
-1-

a. The perfective form of the verb [ktəb] “write”

1sg ktəb-t 1pl ktəb-na
2sg ktəb-ti 2pl ktəb-tu
3sg ktəb 3pl kətb-u
3sg.fem. kətb-at

b. The perfective form of the verb [kərkəb] “roll”

1sg kərkəb-t 1pl kərkəb-na
2sg kərkəb-ti 2pl kərkəb-tu
3sg kərkəb 3pl kərkb-u
3sg.fem kərkb-at

The schwa in the above cases is not part of the base; it is epenthesized for syllabic purposes in
accordance with Benhallam’s (1990a) SSAA (see chapter two for details). According to the
SSAA, schwa epenthesis in items like the ones in 1 operates from right to left. What it
basically does is to take every CC sequence and assign to it the canonical shape CəəC.
Accordingly, the schwa is epenthesized between the second and third consonants in
trisegmental roots as in the stem [ktəb] and between each cluster of two consonants in
quadrisegmental roots as in the stem [kərkəb]. Of particular interest to us here are the verbs
marked for the first person singular and the verbs whose affixes begin with a vowel. Given an
input like /krkb-t/, the SSAA proceeds from right-to-left and epenthesizes a schwa between
each pair of consonants. The result is the ungrammatical form *[krəkbət]. On the basis of
items like these, Benhallam (1990a) concludes that schwa epenthesis should be allowed to
apply cyclically if we are to derive the correct output. With the notion of cyclicity included,
schwa epenthesis will apply first to the inner-most bracketed item in [[krkb]t] to yield
[kərkəb]. Only then is the suffix added to give [kərkəbt], an output form wherein the schwa is
not epenthesized between the affixal and the stem-final consonants. However, this account
fails to account for items like [kərkbək] “he rolled you”, where the schwa is epenthesized
between the object affix [-k] and the stem-final consonant.
In section 3 below, we reconsider the phenomenon of cyclicity of syllabification in the
light of the extended version of correspondence based on a morphologically-grounded theory
of O-O correspondence proposed in Basri et al (1998) and Selkirk (1999). In order to provide
a systematic treatment of cyclicity and get a full picture of how O-O faithfulness operates in
CMA, we will contrast the patterns of syllabification with inflectional endings such as the

153
paradigms in 1 above with the patterns of syllabification with pronominal enclitics of the
form C, CV and V.
The second issue analyzed in this chapter relates to truncation. In the formation of a
morphological class called the nisba derived from certain nouns, part of the base is deleted as
shown in the examples given below:

-2-

a. bni-məllal məllali from Beni-Mellal

nas-əl-γiwan γiwani of Ghiwane (a musical band)
DDaR-əl-biDa biDawi from Casablanca

b. ʔa-məzzru məzzriwi from Amezzrou

ta-rudan-t rudani from Taroudant

The items in 2a are derived from compound nouns; those in 2b are derived from nouns whose
affixes are of Berber origin ([ta-...-t] and [ʔa-] where the glottal stop is prothesized for onset
purposes). Both cases involve deletion of some material from the base. In 2a it is the left-hand
member of the compound which is deleted; in 2b it is [ʔa-] and [ta-...-t] which are deleted. In
section 3 below, we will consider what motivates this deletion: is it some prosodic constraint
on the output that forces truncation, or is truncation morphologically conditioned?
The final issue addressed in this chapter is the causative. Consider the following cases
for illustration.

-3-
Vb stem Causative Gloss

a. ktəb kəttəb write

wSəl wəSSəl arrive
bka bəkka cry

b. fiq fəyyəq wake up

ʃuf ʃəwwəf see

Bennis (1992) has argued that the base for the causative derivation is the stem, that is an
output form, or more specifically a prosodic category which forms part of the stem.
According to the author, the causative is formed by the affixation of a bimoraic syllable to a
circumscribed prosodic domain which is a minimal syllable of the type CəC or CV. In 3a, the
bimoraic syllable is prefixed to CəC; in 3b it is suffixed to CV. The gemination obtained is

154
the result of the spreading of the second segment of the base. In O-O correspondence,
circumscription has been analyzed as a case involving positional faithfulness to some
designated edge of a base output (McCarthy 1997). In section 5 we will show that a
circumscriptive analysis cannot account for the causative forms in 4b and propose that all
causative forms be better analyzed as cases involving reduplication.
The present chapter is organized into five major sections. In section 2, we lay down
the basic tenets of the extended version of CT as well as the definition and nature of the base
and the output forms and how they relate to each other. In section 3, we discuss cyclicity of
syllabification in CMA. In particular, we will show that cyclicity can be accounted for by an
O-O constraint where the derived form is related to its morphologically related output form.
In section 4, we will consider morphological truncation and argue that the fact that part of the
base deletes cannot be attributed to a prosodic constraint. In particular, we will show that the
nisba morpheme attaches to bases that do not have a complex morphological nature. Finally,
in section 5, we reanalyze the causative as a case involving reduplication as proposed in
Imouzaz (forthcoming). We will especially show that the causative could best be understood
if we assume output constraints regulating the relationship between the base form and its
reduplicant.

2. TRANSDERIVATIONAL RELATIONS IN OPTIMALITY

THEORY
In recent developments within the OT framework, McCarthy and Prince (1995, 1999)
reformulate Prince and Smolensky’s (1993) faithfulness constraints under CT. According to
McCarthy and Prince (1999), correspondence is at minimum a relationship between segments
and can also be extended to features and prosodic units such as moras. Correspondence was
originally conceived of as a relation between the base and the reduplicant (B-R
correspondence) but it was soon extended to cover input-output faithfulness (I-O
correspondence). However, in works such as McCarthy (1995, 1997), Benua (1995, 1997),
Kenstowicz (1996, 1997), Burzio (1996) and Kager (1996), Basri et al (1998) and Selkirk
(1999), the notion of correspondence has been extended to cover O-O relations (O-O
correspondence) to account for morphologically related output forms which stand in a
transderivational relation.

155
 In dealing with truncatory phenomena, Benua (1995) proposes that truncation
involves two correspondence relations: one that holds between an input and the base
(I-B correspondence) and another that holds between the base and the truncated form (B-T
correspondence). The truncation model proposed by Benua (1995:6) is as follows:

-4-

Truncation

B-T-Identity
Base ──────────> Truncated form
^

IO-Faith
Input

As shown in 4 the input is related to the base by I-O correspondence while the base is related
to the truncated form by B-T correspondence.
Benua (1995:1, 2) argues that truncated hypocoristics in certain dialects of American
English (e.g. New York and Philadelphia American English) behave differently from their
non-truncated counterparts vis-à-vis [æ] ≈ [ɑ] alternation found in syllables closed by [r].
Orthographic “a” is realized as [ɑ] before a tautosyllabic [r] as in [mɑr], [kɑr]; and as [æ]
when it occurs in syllable final position as in [læ.ri] and [hæ.ri]. However, in truncated forms,
the constraint against tautosyllabic [ær] is not respected as we find items such as [hær] and
[lær], which are the truncated forms of [læ.ri] and [hæ.ri]. Benua argues that the
underapplication of [ɑ] in [lær] and [hær] is attributed to the fact that the truncated items are
related to the initial string of their source words as shown below:

-5-

B-T-Identity
[læ.ri] ────────> [lær]
^

IO-Faith
/læri/

Notice that the base in 5 is a fully prosodized word and that the base and the truncated forms
are separate words.

156
 Transderivational relations of the type that exists between the base [læ.ri] and the
output [lær] could also be extended to account for cyclic phenomena. In this respect, Kager
(1996) has shown that an O-O correspondence relation is needed to account for phenomena
that Brame (1974) attributed to cyclic rule application in Palestinian Arabic. Brame maintains
that the stress rule interacts with a rule of syncope that deletes an /i/ in an open unstressed
nonfinal syllable. To understand this interaction, one has to understand the distinction
between Class I affixes, which are markers indicating agreement with the subject for person,
number and gender; and Class II affixes which are object markers. Consider the following
examples taken from Kager (1996: 5,6):

-6-

a. stem /fihim/ + class I affixes

/fihim/ fíhim he understood

/fihim-na/ fhímna we understood
/fihim-u/ fíhmu they understood

b. stem /fihim/ + class II affixes

/fihim/ fíhim he understood

/fihim-ni/ fihímni *fhímni he understood me
/fihim-ha/ fihímha *fhímha he understood her
/fihim-na/ fihímna *fhímna he understood us

The behavior of the verbal forms in 6 above vis-à-vis i-syncope can be explained by reference
to the base which is defined by Kager (1996: 7) as follows:

-7-

... a form that is compositionally related to the affixed word in a morphological and a
semantic sense. (The meaning of the affixed form must contain all the grammatical
features of its base.) Moreover, the base is a free form, i.e. a word. This second
criterion implies that a base is always an output itself.

Kager shows that the failure of /i/ to delete in the object forms in 6b can be attributed to the
fact that these have a morphological base form (i.e. [fíhim]) in which [i] is stressed. The base
[fíhim] is related both morphologically and semantically to the affixed form in 6b. In contrast,
the same /i/ deletes in the subject forms because these are assumed not to have a base; or
more accurately the verb stem /fihim/, which is morphologically related to the affixed forms
in 6a, does not fulfill the second requirement, namely semantic compositionality.

157
 An account of the cases of affixation witnessed in the Palestinian Arabic cases based
on the cyclic lexical phonology and morphology in OT is offered in Kiparsky (1998).
According to Selkirk (1999), Kiparsky’s theory is based on four properties. First, faithfulness
is defined only on input-output relations. Second, input-output relations are defined within a
Stem-grammar and a Word-grammar. Third, the output of stem grammar is the input of the
word grammar. Finally, the constraints of Stem-grammar and Word-grammar are identical
but the ranking of the constraints may be different. Kiparsky (1998) holds that the asymmetry
exhibited in Palestinian Arabic between subject morphology and object morphology reflects
the distinction between stem-level constraints and word-level constraints. Although,
Kiparsky’s analysis proves to work well for the Palestinian Arabic paradigm asymmetry and
other cases of opacity cross-linguistically, it cannot be adopted in this work given its serialist
nature which is incompatible with the principles of the OT model.
A better solution to the Palestinian Arabic problem is offered in Basri et al (1998) and
Selkirk (1999). Basri et al point out that Kager’s definition of the base and consequently his
theory of O-O correspondence is too restrictive in the sense that it requires the base to have
all the grammatical features of a related independent word, and in so doing confines O-O
correspondence relations to cases of syntactic affixation to words. The aforementioned
authors instead develop a morphologically-grounded theory of O-O correspondence which
distinguishes between two sets of O-O constraints: O-Ostem Faith and O-Oword Faith.
Basri et al (1998) propose that O-O correspondence be defined as follows:

-8-

Two output strings S1 and S2 are in a correspondence relation if:

i- the input S1’ and S2’ to which S1 and S2 respectively correspond are dominated by
the morphological constituents ϕ and Ψ, respectively,
ii- ϕ is an independent word
iii- Ψ is an immediate daughter of an independent word, and
iv- Ψ is morphologically nondistinct from ϕ (in the sense that the morphological
properties of Ψ are a proper subset of those of ϕ)

The definition of O-O correspondence established in Basri et al differs from that of Kager in
the sense that the affixed form is not necessarily required to share all the grammatical features

158
of the related simple base form but rather only a proper subset of those properties. Under this
definition of O-O correspondence, the Palestinian Arabic data in 6b could be viewed as a case
involving O-Oword faithfulness constraints. Consider the embedded structure of [fíhim] and
[fihímna] taken from Basri and others (1998):

-9-

Output-output correspondence under affixation-to-word (O-Oword)

Input: word[stem[fihim] stem-ø] word word[word[stem[fihim] stem-ø] word-na] word

└—————--┘ └——-———┘
S1’ S2’

Output: fíhim fihím-na

└----┘ └----┘
S1 S2
“he understood” “he understood us”

On the other hand, the data in 6a involve O-Ostem faithfulness constraints. The
structures of words such as [fíhim] and [fhímna], taken from Basri et al (1998), are
reproduced in 10 below:

-10-

Output-output correspondence under affixation-to-stem (O-Ostem)

Input: word[stem[fihim] stem-ø] word word[stem[fihim] stem-na] word

└—————--┘ └——┘
S1’ S2’

Output: fíhim fhím-na

└----┘ └---┘
S1 S2
‘he understood’ ‘we understood’

According to Basri et al (1998) correspondence in the way defined in 9 holds for both
9 and 10. In the case of 9, the embedded morphological constituent shares all the properties of
the independent word. In 10, the embedded morphological constituent is the stem. It is not
identical to the independent word. However both of them are morphologically nondistinct and
are related to the stem /fihim/.

159
 This new theory of O-O correspondence, as Selkirk (1999) puts it, is superior to both
the Stem-grammar and Word-grammar theory of Kiparsky (1998) and to the earlier versions
of O-O Faith theory (Burzio 1996, Kenstowicz 1996, 1997, Kager 1996, Benua 1995, 1997).
Kiparsky’s model is serialist in nature whereas the earlier versions of O-O Faith make no
appeal to the internal morphological properties of words in an O-O correspondence relation.
Within the morphologically-grounded theory of O-O correspondence, correspondence is
defined by Selkirk (1999) as follows:

-11-

a. Definition of O-Ostem correspondence

An O-Ostem correspondence relation holds between two output strings S1’ and S2’
when S1’ and S2’ are in an Output-Output correspondence relation and the input
strings corresponding to S1’ and S2’ are terminal strings of morphological
constituents of the type Stem.

b. O-Oword correspondence

An O-Oword correspondence relation holds between two output strings S1’ and S2’
when S1’ and S2’ are in an Output-Output correspondence relation and the input
strings corresponding to S1’ and S2’ are terminal strings of morphological
constituents of the type Word.

Consistent with the general theory of correspondence of McCarthy and Prince (1995), Selkirk
further assumes that there are two different families of O-O faithfulness constraints: O-O
Faithstem and O-O Faithword. The first holds in cases of O-Ostem correspondence while the
second holds in cases of O-Oword correspondence.
It is the morphologically-grounded theory of output-output correspondence proposed
in Basri et al (1998) and Selkirk (1999) that will be adopted in the present work for the
analysis of the CMA cyclic and truncatory phenomena as well as circumscription. We will
start first with the cyclic effect of syllabification in verbs.

3. CYCLIC SYLLABIFICATION
3.1 Introduction

In chapter two, we have shown how an OT account of CMA syllable structure is far
better than a rule-based analysis which relies on syllable structure building rules,

160
directionality and cyclicity. In particular, we have shown that an analysis in terms of the
ranking of a subset of constraints pertaining to UG provides a straightforward account of all
cases of syllabification except those that require recourse to the cycle.
As noted earlier in this chapter, the more problematic cases are those quadrisegmental
verb bases attached to vowel-initial affixes and the first person suffix [t] as well as
trisegmental verb bases to which the first person singular [-t] is attached. For example, words
such as [kərkb-u] “they rolled”, [kərkəb-t] “I rolled” and [DRəb-t] “I hit” need a cyclic
account. Assuming that syllabification applies after all the morphological rules have applied
will yield the incorrect output *[krəkbu], *[krəkbət] and *[DəRbət]. On the basis of items like
these, Benhallam (1990a) assumes that syllabification should be allowed to apply cyclically
to yield the attested output. In the items [DRəbt] and [kərkbu], syllabification applies as in 12
below. The brackets stand for cycles.

-12-

Input [[DRb]t] [[krkb]u]

First cycle
Syllabification D.Rəb kər.kəb
Second Cycle
Affixation D.Rəb-t kər.kəb-u
Syllabification D.Rəb.t kər.kə.bu.
Output [DRəbt] [kərkbu]
“I hit” “they rolled”

Syllabification first applies to the innermost bracketed items in the first cycle. It reapplies in
the second cycle after affixation to adjoin the stranded [t] as a postmargin to the preceding
syllable, thus giving the correct output [DRəbt]. In the remaining item, and after the
suffixation of [u], the segment [b] , syllabified as a coda in the first cycle, is assigned as an
onset to this suffix and this follows from the fact that CMA does not allow onsetless syllables.
Consequently, the schwa is left in an open syllable, thus subject to deletion.
The question we will try to answer here is the following: how is it possible to capture
the sense of cyclicity of right-to-left syllabification in CMA within a non-derivational

161
constraint-based framework? To answer this question, we will explore an analysis in terms of
O-O correspondence along the lines suggested in Basri et al (1998) and Selkirk (1999).

3.2 Cyclicity as Output-Output Correspondence

In this subsection we show how cyclic phenomena could be accounted for by

reference to constraints requiring phonological identity between separate output forms that
stand in a transderivational relation as defined in 8 above.
To solve problems of cyclic syllabification in CMA cases like the ones brought up in
section 1 above, Boudlal (to appear a) proposes, following McCarthy (1995, 1997) and other
work on transderivational correspondence (Benua 1995, 1997, Kenstowicz 1996, 1997,
McCarthy 1995, 1997, Burzio 1996, Kager 1996), that cyclic syllabification in CMA be
accounted for by invoking a constraint establishing correspondence between an affixed form
and a simple base form. Thus a word such as [DRəbt] could be derived by comparing it to the
simple base form [DRəb], which is itself an independent form, and not by comparing it to the
input form /DRb/. The whole picture is schematized in 13 below:

-13-

O/B-O Identity
[D.Rəb] ——————> [D.Rəb.t]
^
I-O Faith

/DRb/

However, and as it has been pointed out above this model does not make any reference to the
internal morphological properties of the words standing in an O-O correspondence. For this
reason, Basri et al (1998) and Selkirk (1999) propose that the theory of O-O correspondence
be developed in such a way that it is grounded in the morphosyntactic representation and in
particular the stem-word distinction. The authors distinguish between two sets of O-O
faithfulness constraints: O-Ostem faithfulness constraints and O-Oword faithfulness constraints.
The first govern alternations at the stem level while the second account for alternation at the
word level.
In order to account for cyclicity of syllabification and get the full picture of how O-O
faithfulness works in CMA, we will consider two different types of affixation reflecting the
stem-word distinction: affixation to the stem which involves the subject affixes [-t] “I”, [-na]

162
“we” and [-u] “they”; and affixation to the word which involves the object clitics [-k] “you”,
[-u] “him” and [-na] “us”.
First consider the stems [DRəb] and [kərkəb] to which the inflectional affixes marking
person are added:

-14-

Verb stem + subject marker

a. DRb DRəb he hit

DRb-t DRəbt I hit
DRb-na DRəbna we hit
DRb-u DəRbu they hit

b. krkb kərkəb he rolled

krkb-t kərkəbt I rolled
krkb-na kərkəbna we rolled
krkb-u kərkbu they rolled

Two generalizations could be made about the data above. First, when consonant-initial affixes
(i.e. C and CV) are introduced, the stem remains intact, i.e. the schwa retains its position
between the second and third consonants of the stem in trisegmental verbs and between the
third and fourth consonants in quadrisegmental verbs. Second, when vowel-initial affixes are
introduced (i.e. V), the schwa, originally placed before the final segment of the stem deletes
in quadrisegmental verbs or is placed before the second segment of the stem in trisegmental
verbs. The deletion of the schwa leads to a change in the syllabic configuration of the stem.
When a pronominal enclitic of the form C, V and CV attaches to the same verb forms,
the result obtained is the following:

-15-

Verb form (3 sg. mas) + object marker

a. DRb DRəb he hit

DRb-k DəRbək he hit you
DRb-u DəRbu he hit him
DRb-na DRəbna he hit us

163
b. krkb kərkəb he rolled
krkb-k kərkbək he rolled you
krkb-u kərkbu he rolled him
krkb-na kərkəbna he rolled us

The forms in 15 look much like those in 14 except for affixes that consist of a single
consonant. In 15 the schwa is epenthesized before the consonantal affix [-k] whereas in 14
such epenthesis does not take place. The behavior of verbal forms in 15 is identical to nouns
to which the pronominal enclitics of the form C, V and CV (marking the possessive) are
added. Consider the following data for illustration:

-16-
Noun + possessive marker

a ktf ktəf shoulder

ktf-k kətfək your shoulder
ktf-u kətfu his shoulder
ktf-na ktəfna our shoulder

b. ʃrʒm ʃərʒəm window

ʃrʒm-k ʃərʒmək your window
ʃrʒm-u ʃərʒmu his window
ʃrʒm-na ʃərʒəmna our window

In these nominal forms, the words [ktəf] and [ʃərʒəm] retain their syllabic configuration only
when they attach to the clitic [-na]. When they attach to the clitics
[-u] and [-k], their syllabic configuration changes. Given that the clitic [-u] is a vowel and that
the onset is obligatory in CMA, it follows that the change in syllabic configuration is dictated
by the constraint ONSET requiring that every syllable have an onset. With respect to the clitic
[-k], one should expect it to behave like the subject prefix [-t] which does not cause the stem
to change its syllabic configuration. This points out to the fact that the two affixes are
attached to two distinct morphological categories. The fact that the schwa is epenthesized
before the subject affix [-t] in [DRəbt] but not before the object affix [-k] in [DəRbək]
reflects a distinction between affixation to the stem and affixation to the word. In affixation to
the stem, the correspondence between the verb stem [DRəb] and the affixed verb form
[DəRbt] is shown in 17 below:

164
-17-

Affixation to stem in CMA

Input: word[stem
[DRəb] stem ø] word word[stem [DRəb]stem -t] word
└—————-—┘ └--—-┘
S1’ S2’

Output: [D.Rəb] [D.Rəb.t]

└-------┘ └------┘
S1 S2

“he hit” “I hit”

The base in 17 conforms to the definition set by Basri et al (1998). Correspondence here is
established between S1 and S2 which are both stems. On the other hand, the correspondence
between [DRəb] and [DəRbək] is one that relates a word to another word as in 18:

-18-

Affixation to word in CMA

Input: word[stem
[DRəb] stem ø] word word [word[stem [DRəb]stem
-ø] word -k]word
└—————-—┘ └-——————┘
S1’ S2’

Output: [D.Rəb] [DəR.bək]

└-------┘ └------┘
S1 S2

“he hit” “he hit you”

The hierarchical morphological structures of the affixation-to-stem and affixation-to-word

instances in 17 and 18 are given in 19 below:

165
-19-

a. Word b. Word c. Word

Stem Stem Subj. Af. Word Obj. Af.

Stem

DRəb DRəb t DRəb k

[DRəb] [DRəbt] [DəRbək]

“he hit” “I hit” “he hit you”

According to the definition set in 8 above, a correspondence relation holds between 19a and
19b and between 19a and 19c. In the case of affixation to stem (i.e. 19b), the stem ϕ, which
stands for the output string S1 [DRəb], is an independent word. Moreover, ψ , which stands
for the stem part in the affixed form [DRəbt], is an immediate daughter of an independent
word, whose morphological properties are a subset of ϕ , which is the stem [DRəb]. The same
thing could be said about the embedded morphological constituent standing for the word part
in the affixed form [DəRbək], which is a morphosyntactic word, and shares all the
morphological properties of the independent word. Thus and according to Basri et all (1998)
and Selkirk (1999), an O-O correspondence holds in both 17 and 18 . So, the difference
between [DRəbt], where schwa epenthesis applies and [DəRbək], where it does not, finds its
explanation in the distinction between O-Ostem correspondence versus O-Oword
correspondence. It is evident enough that the output [D.Rəb.t] is closer to the base [D.Rəb]
than [DəRbək] is. In particular [D.Rəb.t] is identical to [D.Rəb] but not to [DəRbək] in as far
as the initial segments of the stem syllables are concerned. Faithfulness to some designated
periphery of a prosodic category has been captured in the literature by invoking an O-O
constraint, dubbed ANCHOR (McCarthy 1997). McCarthy assumes the existence of distinct
Anchoring constraints from S1 to S2 and from S2 to S1, referred to as I-ANCHOR and O-
ANCHOR. He also assumes that there are two senses of Anchoring: ANCHOR-POS which is
satisfied when a segment’s position as head, initial or final is preserved under
correspondence; and ANCHOR-SEG, which demands that the segment itself be conserved in

166
the designated position (i.e. the Beckman (1995, 1998) positional faithfulness). Within each
of the two families of constraints, a token must also specify the constituents involved, the
type of correspondence relation between them (I-O, B-R, O-O) and the position anchored to
(head, initial, final).
For the purpose of the syllabification cases considered in this section, the constraint
needed is the one that anchors position and is formulated by McCarthy (1997: 12) as follows:

-20-

O-ANCHOR-POS S1-S2 (Cat1, Cat2, P)

If ς1 Cat1 ∈ S1
ς2 Cat2 ∈ S2
ς1 R ς2, and
ς2 stands in position P of Cat2
then ς1 stands in position P of Cat1.

According to McCarthy, when Cat1=Cat2, the result is prosodic faithfulness per se. An
example of this type is I-ANCHOR-POSIO (Ft, Ft, Head) which states that the locus of stress
must not change in the input/output mapping. When Cat1=Base and Cat2=Reduplicant, the
result is a typical BR- Anchoring. Finally when Cat1= stem and Cat2= σ, the result is the
alignment of a morphological category and a prosodic one. It should be noted that the
constraint in 20 is irrelevant when a segment is deleted or epenthesized at the designated
edge.
In order to account for the difference between [DRəbt] and [DəRbək], correspondence
has to refer to the initial position of the syllable in the derived output form and its related base
output form. Given the distinctions made between stem and word, we will have to distinguish
between O-Ostem ANCHORPOS and O-Oword ANCHORPOS. These constraints are formulated,
after Selkirk (1999) as in 21 below:

-21-

a. O-Ostem ANCHOR (σ, σ, Initial)

Where two strings S1 and S2 are in an O-Ostem correspondence relation and S1 is the
base and S2 the affiliate of that correspondence relation, a syllable-initial segment
belonging to S2 must correspond to a syllable-initial segment belonging to S1.

167
b. O-Oword ANCHOR (σ, σ, Initial)

Where two strings S1 and S2 are in an O-Oword correspondence relation and S1 is the
base and S2 the affiliate of that correspondence relation, a syllable-initial segment
belonging to S2 must correspond to a syllable-initial segment belonging to S1.

[DRəb] is related to [DRəbt] in an O-Ostem correspondence whereas [DRəb] is related to

[DəRbək] in an O-Oword correspondence relation. The word [DRəbt] shows that it is more
important to keep the left edges of the stem when the suffix is added than to avoid a minor
syllable. Thus in order to rule out a potential output candidate such as *[DəRbət], O-Ostem
ANCHOR (σ, σ, Initial) must dominate the markedness constraint *Min-σ which penalizes
minor syllables as shown for the derivation of [ktəbt], which stands in correspondence with
the base output form [k.təb] in the constraint tableau
in 22:

-22-

/DRb-t/ O-Ostem ANCHOR *Min-σ

Base: [D.Rəb]stem (σ, σ, Initial)
)a. D.Rəb.t *
b. DəR.bət *! *
c. D.Rə.bət *!
d. Də.Rə.bət *! *

In the optimal candidate, the initial segment of the first syllable in the affiliate (derived) form
corresponds to the initial segment of the first syllable in the simple output base form (i.e. the
segment [D]). The same thing could be said about the initial segment of the second syllable
which is the segment [R]. Candidates 22b-d incur a single violation mark of the constraint
requiring left anchoring of the initial segments in the derived output and the simple base
form. In 22b, the initial segment of the second syllable in the affixed forms (i.e. the segment
[b]) does not correspond to the initial segment of the second syllable in the simple base form
(i.e. the segment [R]). In 22c and 22d, the initial segment of the third syllable in the derived
form does not have a correspondent in the simple base form.

168
 In the instance of affixation to word [DəRbək], satisfying the constraint *Min-σ is
more important than conserving the initial position of the base syllables. This points out to the
fact that *Min-σ must rank higher than O-Oword ANCHOR (σ, σ, Initial) as the tableau below
shows:

-23-

/DRb-k/ *Min-σ O-Oword ANCHOR (σ, σ,

Base: [[D.Rəb]stem]word Initial)
)a. DəR.bək *
b. D.Rəb.k **!
c. D.Rə.bək *! *

The tableau shows that any candidate violating *Min-σ is ruled out. This is the case with 23b
and 23c. The optimal candidate spares that markedness constraint but instead violates the
lower-ranking O-Oword ANCHOR (σ, σ, Initial) because the initial segment of the second
syllable in the affixed form (i.e. the segment [b]) does not correspond to the initial segment of
the second syllable in the base form (i.e. the segment [R]). Since O-Ostem ANCHOR (σ, σ,
Initial) dominates *Min-σ and *Min-σ dominates O-Oword ANCHOR (σ, σ, Initial), we
therefore conclude that O-Ostem ANCHOR (σ, σ, Initial) dominates O-Oword ANCHOR (σ, σ,
Initial), by transitivity. This ranking predicts that faithfulness should appear in cases of
affixation to stem but not in cases of affixation to words. While this prediction is partially true
and allows for a clear distinction between cases like [DRəbt], where stem faithfulness is
satisfied, and [DəRbək], where word faithfulness is sacrificed, it cannot be generalized to
account for all the paradigms, especially the cases involving vowel-initial affixes. When a
vowel-initial affix is attached to a verbal form, be it a stem or a word, its syllabic
configuration changes and as such both O-Ostem faithfulness and O-Oword faithfulness
constraints are violated. This points to the fact that ONSET must dominate both
O-Ostem ANCHOR (σ, σ, Initial) and O-Oword ANCHOR (σ, σ, Initial). In the tableau below,
we show how the candidate [DəR.bu] which violates O-Ostem ANCHOR (σ, σ, Initial) wins
over *[DRəbu]:

169
-24-

/DRb-u/ ONSET O-Ostem ANCHOR (σ, σ, Initial)

Base: [D.Rəb]stem
)a. DəR.bu *
b. D.Rəb.u *!

In the optimal candidate, only the initial segment of the first syllable of [DəRbu] corresponds
to the syllable initial segment of the base [DRəb]. The initial segment of the second syllable
in the derived output form which is [b] does not correspond to the initial segment of the
second syllable in the base which is [R] and this mismatch leads to the violation of O-Ostem
ANCHOR (σ, σ, Initial). However, this violation is not fatal since the constraint O-Ostem
ANCHOR (σ, σ, Initial) is violated only for the purpose of securing a higher-ranking
constraint, namely ONSET. The candidate in 24b preserves the initial segments of the first
and second syllables of the base form but fails because it incurs a fatal violation of ONSET.
What the tableau above shows is that it is more optimal to violate O-Ostem ANCHOR (σ, σ,
Initial) than ONSET.
Quadrisegmental verbs to which the subject affix [-t] is added could be obtained much
in the same way as trisegmental ones. Thus an input such as /krkb-t/ surfaces as [kərkəbt] (cf.
the stem [kərkəb]) and not as *[krəkbət] by virtue of ranking O-Ostem ANCHOR (σ, σ, L)
higher than *Min-σ. On the other hand, an input such as /krkb-k/, where the object clitic is
suffixed to the verb, surfaces as [kərkbək] and not as *[kərkəbk] because of ranking O-Oword
ANCHOR (σ, σ, Initial) lower than *Min-σ. As to verbs to which vowel-initial suffixes are
attached, they are expected to violate the anchoring constraints given that ONSET is
undominated. For illustration, consider the input /krkb-u/ to which the subject affix
[-u] is attached:

170
-25-

/krkb-u/ ONSET O-Ostem ANCHOR (σ, σ,

Base: [kər.kəb]stem Initial)
0a. kər.kbu

b. kər.k.bu *!
c. kər.kəb.u *!

The two constraints in 25 wrongly predict that the optimal candidate is 25a. This candidate
satisfies ONSET and O-Ostem ANCHOR (σ, σ, Initial) by virtue of the fact that the initial
segment of the first and second syllables of the derived output (i.e. the segment [k])
correspond to the initial segment of the first and second syllables of the base. Although
candidate 25a satisfies both constraints, it should be excluded on the ground that it incurs a
fatal violation of the undominated constraint *COMPLEX. Therefore, in order to derive the
correct output, *COMPLEX has to dominate O-Ostem ANCHOR (σ, σ, Initial) as the
following tableau shows:

-26-

/krkb-u/ ONSET *COMPLEX O-Ostem ANCHOR

Base: [kər.kəb]stem (σ, σ, Initial)
)a. kər.k.bu *
b. k.rək.bu **!
c. kər.kbu *!
d. kərk.bu *! *
e. kər.kəb.u *!

The optimal candidate comprises three syllables, one of which is minor and is associated with
the consonant [k]. This minor syllable arises in order to satisfy *COMPLEX. However this
satisfaction causes violation of O-Ostem ANCHOR
(σ, σ, Initial) by virtue of the fact that the initial segment of the third syllable of the derived
output which is [b] does not correspond to any syllable in the base form. In candidate 26b the
minor syllable arises at the left periphery, a fact which causes a fatal violation of O-Ostem
ANCHOR (σ, σ, Initial) because the initial segment of the second syllable of the derived

171
output does not correspond to the initial segment of the second syllable of the base and the
initial segment of the third syllable of the output candidate does not have any correspondent.
Candidates 26c and 26d are both excluded because of *COMPLEX. Finally candidate 26e
retains the initial segments of the base syllables in their position and thus incurs a fatal
violation of ONSET.
The cases so far seen have the shape /CCC/ or /CCCC/. A reasonable question one
should ask here is whether or not verbs with underlying full vowels behave like the patterns
already considered. More specifically, could the analysis undertaken for verbs on the pattern
/CCC/ and /CCCC/ be extended to verbs on the pattern /CVC/ and /CVCC/?
To answer this question, consider the following examples which include trisegmental
and quadrisegmental verbs and nouns to which affixes on the pattern C, V and CV are
attached:

-27-

a. Verb stem + subject marker

i. bas bas he kissed

bas-t bəst 1 I kissed
bas-na bəsna we kissed
bas-u basu they kissed

ii. samħ saməħ he forgave

samħ-t saməħt I forgave
samħ-na saməħna we forgave
samħ-u samħu they forgave

b. Verb form (3 sg. mas) + object marker

i. bas bas he kissed

bas-k basək he kissed you
bas-u basu he kissed him
bas-na basna he kissed us

1
Notice here that in verbs on the pattern CVC (where V is a full vowel), the medial vowel appears consistently
as a schwa in the first and the second person singular and plural. This phenomenon of vowel reduction will not
be considered in this section, but for an account of such a phenomenon, the reader is referred to works such as El
Himer (1991), Boudlal (1993), Hammari (1996, 2000), Benhallam (1998) and Rguibi (forthcoming).

172
 ii. samħ saməħ he forgave
samħ-k samħək I forgave
samħ-u samħu he forgave him
samħ-na saməħna he forgave us

c. Noun + possessive marker

i. bab bab door

bab-k babək your door
bab-u babu his door
bab-na babna our door

ii. faxr faxər coal

farx-k faxrək your coal
faxr-u faxru his coal
faxr-na faxərna our coal

The data in 27 comprise both trisegmental and quadrisegmental verb and noun forms.
Trisegmental forms have the shape CVC while quadrisegmental forms have the shape CVCC.
In the affixation-to-stem cases (i.e. 27a), no schwa is epenthesized between the verb stem and
the first person marker [-t]. Thus we get output affixed forms such as [bəst] and [saməħt]
which satisfy O-Ostem ANCHOR (σ, σ, Initial) but not *[bəsət] and *[samħət] which violate
it. An output candidate such as [basət] meaning “she kissed” is allowed in CMA. The subject
suffix here is the morpheme [-at] whose low vowel is reduced to a schwa because the stem
contains another low vowel2.In the affixation-to-word cases, the schwa is epenthesized before
the affix [-k] in the verbal as well as the nominal forms. Thus inputs such as /bas-k/, /DaR-k/,
/samħ-k/, and /faxr-k/ violate the constraint O-Oword ANCHOR (σ, σ, Initial) and surface
respectively as [basək], [DaRək], [samħək], and [faxrək] but not as *[bask], *[DaRk],
*[saməħk], and *[faxərk]. It should be noted that when the affix added is vowel-initial, O-O
faithfulness in the stem and word levels is sacrificed to secure the undominated constraint
ONSET.
In the following tableau, we consider some possible candidate output forms obtained
from the input /samħ-u/ where the suffix [-u] marks the third person plural:

2 Items such as [salat] do not constitute any counterexamples to vowel reduction. Given the fact that the input is
/sala-at/ and not /sala-t/, one should expect the vowel of the suffix to truncate in order to avoid a cluster of two
vowels.

173
-28-

/samħ-u/ ONSET *COMPLEX O-Ostem ANCHOR

Base: [sa.məħ]stem (σ, σ, Initial)
)a. sam.ħu *
b. sa.mħu *!
c. sa.məħ.u *!

The optimal candidate wins although it incurs a violation mark of O-Ostem ANCHOR (σ, σ,
Initial). In this candidate, only the initial segment of the first syllable corresponds to the initial
segment of the first syllable in the base form. The initial segment of the second syllable in
[sam.ħu] does not correspond to the initial segment of the second syllable in the base form
(i.e. the segment [m]). Candidates 28b and 28c are ruled out although they satisfy O-Ostem
ANCHOR (σ, σ, Initial): 28b is excluded because of *COMPLEX, and 28c, because of
ONSET.
To sum up, the examples considered in this subsection show that cyclicity in CMA
can be explained in the extended version of correspondence where a distinction should be
made between O-Ostem and O-Oword. This distinction allows for a straightforward explanation
of schwa epenthesis and non-epenthesis when a suffix that consists of a single consonant is
added to the verb form. It has been shown that in the case of affixation to stem, no schwa is
epenthesized between the stem and the subject suffix [-t], a fact which follows from high
ranking O-Ostem ANCHOR (σ, σ, Initial) above the markedness constraint *Min-σ. In the
case of affixation to the word, schwa epenthesis applies before the object suffix [-k] and leads
to a mismatch in position between the syllable initial segments of the affixed and the base
forms. We have also shown that O-Ostem ANCHOR (σ, σ, Initial) is satisfied only when
ONSET and *COMPLEX are not at stake. Evidence for high ranking ONSET comes from
vowel-initial suffixes that satisfy the constraint by changing the syllabic configuration of the
base thus leading to the violation of O-Ostem ANCHOR (σ, σ, Initial) and O-Oword ANCHOR
(σ, σ, Initial). As to the constraint *COMPLEX, we have shown that the initial segment of
the third syllable of the affixed form in quadrisegmental verbs and nouns does not correspond
to any segment in the base form.

174
 In what follows, we will try to show if this line of analysis, which distinguishes
between O-Ostem Faith and O-Oword Faith, can account for truncation observed in certain items
to which the nisba morpheme is attached.

4. TRUNCATION
In this section, we will consider how O-O correspondence could be made use of to
account for a class of nisba adjectives derived from compound nouns. Consider the following
examples for illustration:

-29-

a. nas-əl-γiwan γiwani (of) Ghiwane (a musical band)

wlad-ħriz ħrizi from Oulad Hriz
bni-məllal məllali from Beni-Mellal
bni-zərwal zərwali from Beni Zerwal
bni-məskin məskini from Beni-Meskine

b. DDaR-(əl)-biDa3 biDawi from Casablanca

qəlʕət-əs-sraγna sərgini from Qalaat Seraghna
wlad-ʕəbbu ʕəbbawi/ʕəbbubi from Ouled Abbou
wlad-ħəddu4 ħəddawi from Ouled Haddou

Most of these toponyms are formed by compounding the word [wlad] or [bni] (both meaning
“sons of”) and another noun. However the meaning of the compound is not compositional.
For example, the meaning of a word such as [bniməskin] is not predictable from the
constituent elements [bni] meaning “sons” and [məskin] meaning “beggar” but it refers to a
geographical area called “Beni-Meskine”. In 29, the base to the derivation of the nisba
adjective is the compound noun. When the nisba suffix is added, part of the base is deleted, a
fact which shows that the output might be governed by some prosodic constraint that limits its
output size. A consideration of the forms in 29 shows that the output does not exceed three

3
the first [D] in [DDaRəlbiDa] and the first [s] in [qəlʕətəssraγna] result from the assimilation of the definite
article [l-] to the following coronal segment of the right-hand member of the compound.
4
Notice here that the base final vowel [u] is realized as [a] in the nisba. The contiguity of the nisba suffix and
the base final vowel gives rise to a hiatus that the language resolves by epenthesizing the glide [w] which agrees
with the preceding vowel. Given the fact that CMA does not allow a sequence of two rounded segments (as will
be seen in chapters five and six), the [u] of the base dissimilates to [a] to avoid output forms such as *[ʕəbbuwi]
and *[ħədduwi].

175
major syllables. However, other nisba adjectives show that the output may also consist of two
major syllables as the examples below:

-30-

sla slawi from Salé

sma smawi sky blue
wəʒda wəʒdi from Oujda
ʕəbda ʕəbdi from the Plain Abda
TaTa TaTawi from Tata
fas fasi from Fes
tazi tazi from Taza

Thus, it might be the case that part of the base in the compound items in 29 is
truncated in order not to end up in forms that consist of more than three syllables. If this is the
case, one might be tempted to assume that the deletion in the items in 29 is the result of a
prosodic constraint requiring that the output consist of a minimum of two syllables and a
maximum of three. The minimality requirement could be obtained by FT-BIN, which
requires that the foot be binary at some level of analysis. As to the maximality requirement, it
could be obtained by invoking the constraint ALIGN-Ft-R, demanding the alignment of the
right edge of every foot with the right edge of the prosodic word. In this way, the constraint
ALIGN-Ft-R may act as a prosodic delimiter of the output to the nisba. Thus the more feet a
form has the less optimal it is. This constraint will have to dominate PARSE-σ demanding
that every syllable be parsed into foot structure unless FT-BIN is at stake. It will also have to
dominate the faithfulness constraint O-Oword MAX (corresponding to MAX-Base/Truncated
in Benua 1995, 1997) to account for output-output correspondence between two words: the
base output and its affiliate truncated form. In the tableau below, we show if the
aforementioned constraints could account for the relation between the base [nasəlγiwan] and
the truncated form [γiwani].

176
-31-

Input: /nas-l-γiwan-i/ FT-BIN PARSE-σ ALIGN-FT-R O-Oword

Base: [nasəlγiwan]word MAX
0a. na(səl.γi)(wa.ni) **

b. (γi)(wa.ni) *! ** *****

c. γi.(wa.ni) *! *****

The optimal candidate according to the ranking in 31 is the wrong output *[nasəlγiwani]
which contains only one foot that right-aligns with the PWd. The candidate in 31b is excluded
because it violates FT-BIN and in so doing incurs two violation marks of the constraint
ALIGN-Ft-R. The final candidate, which should be the actual optimal candidate, is ruled out
because it incurs a fatal violation of PARSE-σ. One possible way of obtaining the optimal
candidate [γiwani] is by ranking ALIGN-Ft-R above PARSE-σ thus allowing monomoraic
syllables to be left unparsed as the following tableau shows:

-32-

Input: /nas-l-γiwan-i/ FT-BIN ALIGN-FT-R PARSE-σ O-Oword

Base: [nasəlγiwan]word MAX
)a. γi.(wa.ni) * *****

b. na(səl.γi)(wa.ni) **! *

This ranking seems to yield the correct output. However, it cannot be accepted for three
different reasons. First, and as it has already been established in chapter three, CMA is a
language with exhaustive parsing, a fact which explains why PARSE-σ has to dominate
ALIGN-Ft-R and not the other way round. Second, the ranking above would give the wrong
output as to trisyllabic outputs not derived from compound nouns and which consist of a
heavy syllable followed by two light syllables as in 33:

177
-33-
Base Nisba Gloss
tadla tadlawi from Tadla
TiTwan TiTwani from Tetuan
warzazat warzazi from Ouarzazate
bulman bulmani from Boulmane
wazzan wazzani from Ouazzane
tiznit tizniti from Tiznit

The foot structure of the output form should in fact be (H)(LL). But given the wrong ranking
in 32, we should expect a word such as [tadlawi], for example, to be footed as tad.(la.wi) or
even see its heavy syllable delete to surface as (la.wi). These two candidates fare better than
the optimal (tad)(la.wi) in-as-far as the constraints in 32 are concerned. Third, the ranking of
the constraints 32 predicts that words which consist of more than three syllables would never
arise because this would incur additional violations of ALIGN-Ft-R. Such is not always the
case as there are other nisba adjectives whose output consist of four syllables as shown in the
examples below:

-34-

Base Nisba Nisba Gloss

filiSTin filiSTini from Palestine

muriTanya muriTani from Mauritania
mirikan mirikani from America
sinigal sinigali from Senegal
pakistan pakistani from Pakistan
l-kamirun kamiruni from Cameroon
l-filippin filippini from the Philippines

Truncation does not apply in these simple words even if the output consists of four syllables
constituting two feet. Does that mean that the first two syllables to the left should be left
unfooted? Or should the first be truncated and the second unfooted? Or should the first and
second be truncated? Neither of these would work. If this is so, how is it then possible to
account for deletion of part of the base in compound nouns when the nisba suffix is added?
The answer to this question comes from morphology and more particularly from the
distinction between the categories ‘stem’ and ‘word’5. A look at the compound words in 29
shows that the nature of the base of affixation, i.e. the constituent to which the nisba affix

5
See Selkirk (1982) on the theory about the syntax of words and the distinction she makes between the
categories Affix, Root, Stem and Word.

178
attaches, is defined in morphological terms. The morphological structures of the compound
forms in 29 are given in 35 below:

-35-

a. word[[[nas]stem]word-[əl-[γiwan]stem]word]word
word[[[wlad]stem]word-[[ħriz]stem]word]word

word[[[bni]stem]word-[[məllal]stem]word]word

word[[[bni]stem]word-[[zərwal]stem]word]word

b. word[[D[DaR]stem]word-[əl-[biDa]stem]word]word

word[[[qəlʕət]stem]word-[əs-[sraγna]stem]word]word

word[[[wlad]stem]word-[[ʕəbbu]stem]word]word

word[[[wlad]stem]word-[[ħəddu]stem]word]word

In the case of the compounds, the nisba morpheme has to attach to a base which
should not have a morphologically complex shape. More specifically, it has to attach to the
rightmost stem of the base. The input to the nisba cannot be the rightmost stem alone but it is
the rightmost stem part of the whole compound word. In other words, nisba forms such as
[γiwani] and [ħrizi] are derived not from the stems [γiwan] and [ħriz] but from the compound
words [nasəlγiwan] and [wladəħriz], respectively. Thus nisba adjectives such as
*[nasəlγiwani] and *[wladəħrizi] are failed candidates not because they consist of more than
three syllables but because the nisba morpheme has been attached to the category ‘word’ and
not to the category ‘stem’. Within the theory of word syntax (Selkirk 1982), the structures of
*[nasəlγiwani] and *[wladəħrizi] are shown in 36:

179
-36-

a. *Word b. *Word

Word Aff Word Aff

word word word word

stem Aff stem stem stem

nas l γiwan i wlad ħriz i

The question we will try to answer below relates to the nature of the constraint(s) ruling out
structures where the nisba affix is sister to the category ‘word’ instead of the category ‘stem’.
In order to exclude structures like the ones in 36, we could possibly posit, following
McCarthy (1993b), an alignment constraint of the type ALIGN (Nisba, L, Rightmost Stem,
R) requiring that the left edge of the nisba morpheme coincide with the right edge of the
rightmost stem. However, the problem with alignment constraints is that they do not say
anything about hierarchical morphological structural relations since these constraints govern
the relation of terminal strings. Therefore, and in order to account for the case of affixation to
stem observed in the nisba adjectives derived from compounds, we need to invoke a specific
class of constraints, which we call affixation constraints. These constraints account for the
place of affixes in the hierarchical morphological structure, i.e. they specify the
morphological category to which a particular affix attaches. In the case of the CMA nisba
adjectives, the affixational constraint needed is stated in 37 below:

-37-

AFFIX (Nisba, R, NOUNstem, R) ( henceforth AFFIX-TO-Nstem)

The nisba morpheme is suffixed to a noun stem.

Following Selkirk (1982), satisfying this constraint will give a derived structure where a stem
dominates the nisba suffix and the noun stem. An additional principle from Selkirk calls for
the projection of the features associated with the head of a morphological constituent. Since
the nisba is an adjective-creating suffix in that it changes the syntactic category of the base to

180
which it attaches, we assume that this suffix is specified as [+adjective] and that, being the
head, it projects its adjectival feature, thus producing an adjectival stem as the dominating
node. In addition to the feature [+adjective], the nisba suffix projects other features such as [-
feminine] and [-plural] since the resulting word is always masculine singular. Given the
affixation constraint in 37 and the projection of the adjectival feature of the nisba, the
structures in 36 will look like the ones in 38 below. (Noun and Adj stand for the major lexical
category corresponding to ‘word’, Naff stands for ‘noun affix’ and Adjaff stands for ‘adjectival
affix’)

-38-

a. ?? b. ??

Noun Adj Noun Adj

Nstem Naff Adj Nstem Nstem Adjaff

Nstem Adjaff

nas l γiwan i wlad ħriz i

As to the syntactic category of the whole compound, it is yet to be determined. It will be

shown below that whenever a situation such as the one in 38 arises, the leftmost member of
the compound is truncated. Truncation of this type could be understood if we know that the
only compound structure allowed in CMA is one where both members are nouns.
Given that the compound nouns in 35 consist of two noun stems, attaching the nisba
morpheme to the leftmost or to the rightmost stem would satisfy the constraint AFFIX-TO-
Nstem. If this is the case, how is it possible to exclude forms such as *[nasilγiwan] and
*[wladiħriz], where the nisba morpheme (written in bold) attaches to the leftmost stem and

181
retain forms such as those in 38, where the nisba attaches to the rightmost stem?6
The answer to this question comes from the Generalized Alignment Theory of
McCarthy and Prince (1993b), and more particularly from a constraint requiring right
alignment of the nisba affix and the PWd. This constraint is stated in 39 below:

-39-

ALIGN (Nisba, R, PWd, R) (henceforth ALIGN-Nisba-R)

The right edge of the nisba affix must be aligned with the right edge of
the PWd.

This constraint, if satisfied, rules out any output form where the nisba is attached to the
leftmost noun stem, a fact which shows that ALIGN-Nisba-R has to dominate
AFFIX-TO-Nstem.
In compound nouns, suffixation of the nisba morpheme to the stem causes truncation
of part of the base, i.e. any segment occurring to the left of the rightmost stem of the
compound. In other words, the definite article (if any) and the leftmost member of the
compound, which happens to be the head (Al Ghadi 1990, Boudlal 1993), are deleted, thus
causing violation of the constraint Morpheme Realization (Samek-Lodovici 1993, Rose 1997,
Gnanadesikan 1997) which is stated as follows:

-40-

Morpheme Realization (henceforth MORPH-REAL)

An input morpheme must be realized in the output.

In order to get the truncation of any morpheme occurring to the left of the rightmost
stem, we need a constraint of the type in 41 below, ruling out compound constructions that
consist of a noun and a nisba adjective.

6
Further support for choosing the rightmost stem of a compound as a base for the nisba comes from abbreviated
names whose leftmost stem, which is consistently the stem [ʕəbd] “servant of”, is truncated as shown in the
following items:
ʕəbd-lə-ʕziz ʕaziz
ʕəbd-lə-krim karim
ʕəbd-əl-ħakim ħakim
The rightmost adjectival noun which is retained is subject to a constraint requiring that it be disyllabic, a fact
which shows why stems such as [ʕaziz] and [ħakim] proceed to augmentation by a-affixation.

182
-41-

*[Noun + Adjective]PWd (henceforth *[N + A]PWd)

A sequence of a noun and an adjective is prohibited.

If this constraint together with the constraints ALIGN-Nisba-R and AFFIX-TO-Nstem

dominates MORPH-REAL, the morpheme(s) occurring to the left of the rightmost stem will
not be realized. In 42 below, we shown how the nisba adjective [γiwani], derived from the
base noun [nasəlγiwan] wins over other competing candidates. We assume that violation of
MORPH-REAL is categorical, i.e. a form incurs a single violation even if more than one
morpheme is deleted:

-42-

word[[[nas]stem]word-[əl-[γiwan]stem]word]word -i *[N + A]PWd ALIGN- AFFIX- MORPH-

Nisba-R TO-Nstem REAL

)a. γiwani *

b. nasəlγiwani *! *

c. nasilγiwan *!
d. nasi *! *

In the optimal candidate, the right edge of the nisba suffix is aligned with the right edge of the
PWd, which happens to be the base. The failure of the definite article and the leftmost stem of
the compound to be realized in 42a is dictated by the constraint banning structures where the
PWd contains compound structures that consist of a noun and an adjective. Such is not the
case with candidate 42b which is ruled out because of violating higher-ranked *[N + A]PWd.
One way of avoiding a sequence of a noun and an adjective and therefore violation of the
constraint *[N + A]PWd is for the nisba morpheme, held responsible for projecting its
adjectival feature to the mother node dominating this morpheme and the noun stem, to attach
to the leftmost stem of the compound. This is exactly the case with candidate 42c which
satisfies *[N + A]PWd but fails because the right edge of the nisba suffix does not correspond
to the right edge of the PWd. The same thing could be said about candidate 42d, except that it
incurs, in addition to *[N + A]PWd, a gratuitous violation of MORPH-REAL.

183
 Next, consider an input such as [DDaRəlbiDa] (from the list in 35b) where the base
ends up in a vowel. Since the nisba suffix is a vowel, we should expect glide epenthesis to
apply to provide an onset to the syllable whose nucleus is the suffix vowel. This is exactly
what happens. The optimal candidate [biDawi] incurs a violation of the lower-ranked DEP-IO
by epenthesizing the glide [w] and this in order to satisfy the undominated constraint ONSET.
It also incurs a violation of another lower-ranked constraint, namely MORPH-REAL, by
deleting the definite article and the leftmost stem of the compound in order to satisfy the
constraint *[N + Adj]PWd. A candidate such as *[biDai] is excluded on the ground that it
incurs a fatal violation of ONSET. Finally, other forms such as *[DDaRilbiDa] and
*[DDaRəlbiDawi] are failed candidates because of different reasons. *[DDaRilbiDa] fails
because the nisba morpheme is adjoined to the leftmost stem of the base compound, thus
causing a fatal violation of ALIGN-Nisba-R, whereas *[DDaRəlbiDawi] fails because it
consists of a sequence of a noun and an adjective which is ruled out by the constraint *[N +
Adj]PWd.
Morphologically conditioned truncation is not restricted to compound constructions; it
also applies in certain toponyms containing the discontinuous feminine affix [ta-…-t] and the
prefix [(ʔ)a-] as the examples in 43 for illustration:

-43-

Base Nisba Gloss

a. tafilalt filali from Tafilalt

tarudant rudani from Taroudant
taħənnawt ħənnawi from Tahannaout
taməSluħt məSluħi from Tameslouht

b. ʔazəmmur zəmmuri from Azemmour

ʔasfi sfiwi/məsfiwi from Safi
ʔaməzzru məzzriwi from Amezrou

The suffixation of the nisba morpheme results in the deletion of the discontinuous morpheme
[ta-...-t] and the prefix [(ʔ)a-]. The two affixes are of Berber origin: [(ʔ)a-] marks the singular

184
number, and [ta-...-t] shows that the word is feminine singular7.What is more important for us
here is that the nisba morpheme adjoins to a noun stem rather than to the major lexical
category noun, corresponding to the ‘word’. Adjoining the nisba morpheme to the ‘word’
would result in ungrammatical forms exemplified by the words *[tafilalti] and *[ʔazəmmuri]
whose internal hierarchical morphological structures we give in 44 below:

-44-

a. * Word b. * Word

word Aff word Aff

Aff stem Aff Stem

ta-...-t rudan i ʔa zəmmur i

In the structure in 44a, the discontinuous affix [ta-...-t] could well appear to the right of the
stem and the result would always be the same. It is the sister node that determines the
morphological category to which the nisba affix is attached. In both structures, the nisba
morpheme is sister to the category ‘word’, a state of affairs which is ruled out by the
constraint AFFIX-TO-Nstem. It has already been established above that ALIGN-Nisba-R must
dominate AFFIX-TO-Nstem to ensure that the nisba morpheme appears at the right edge of the
base in the case of compounds. With simple words such as those in 43, attaching the nisba
morpheme to a noun stem would lead to the violation of ALIGN-Nisba-R, especially in bases
with the discontinuous morpheme [ta-...-t] (cf. words such as *[tarudanit] and *[tafilalit],
where the nisba is placed before the second part of the discontinuous morpheme). This points
out to the fact that another constraint must dominate ALIGN-Nisba-R. In the case of the nisba

7
The same morpheme is used in MA to derive abstract nouns of profession such as the following:

Base Noun N. of Profession Base Gloss

bənnay tabənnayt mason
fəllaħ tafəllaħt farmer
ħəddad taħəddatt blacksmith
ʃəffar taʃəffart thief
Here the discontinuous morpheme [ta-...-t] has both a derivational and an inflectional status. It performs more
than the function it has in Berber which is marking the word for the feminine gender.

185
with compounds, it has been shown that it is the constraint *[N + Adj]PWd which forces the
deletion of the leftmost member of the compound. In the case of the nisba with bases
containing the affixes [ta-...-t] and [ʔa-], adjoining the nisba morpheme to the noun stem
produces an adjectival stem as the dominating node, thus creating a sequence of a noun and
an adjective as shown in 45:

-45-
a. ?? b. ??

Naff Adj Naff Adj

Nstem Adjaff Nstem Adjaff

ta-...-t rudan i ʔa zəmmur i

The structures in 45 are reminiscent of a situation which is déjà vu with the compounds in 38
above. Both the structures in 45 and the structures in 38 contain a sequence of a noun and an
adjective. The only difference is that the noun in 38 does not correspond to a major lexical
category; it is an affix which is specified as [+noun]. The juxtaposition of a noun (or a
nominal affix) and an adjective makes it impossible to determine which of the two categories
the mother node, dominating both the nominal affix and the nisba adjective, should be
specified for. Because the nominal affixes in 45 and the nisba adjectival affix bear conflicting
features, the language resolves this conflict by truncating the nominal affix in order to satisfy
the constraint *[N + Adj]. In the following tableau, we show how the nisba adjective [rudani],
derived from the base [tarudant], wins over two other competing candidates:

186
-46-

word[[ta-[rudan]stem -t] word]word -i *[N + A]PWd ALIGN- AFFIX- MORPH-

Nisba-R TO-Nstem REAL
a. tarudanti *! *
b. tarudanit *! *
)c. rudani *

Both candidates 46a and 46b are ruled out for violating the higher-ranked constraint *[N +
Adj]PWd by allowing the nominal affix to occur with the nisba adjectival stem. Candidate 46a
retains the discontinuous affix [ta-...-t] of the base and suffixes the nisba morpheme to the
word, thus incurring a violation of the constraint AFFIX-TO-Nstem. As to candidate 46b, it
satisfies AFFIX-TO-Nstem but incurs a violation of ALIGN-Nisba-R by retaining the suffixal
part of the same discontinuous morpheme which happens to mark the right edge of the base
word.
To sum up, this section has shown that if we are to account for the truncatory
phenomenon exhibited by the nisba adjective derived from compound nouns, we need to
make recourse to constraints governing morphological structure. We have shown the nisba
morpheme is introduced by an affixation constraint of the type AFFIX-TO-Nstem, requiring
that it attach to the noun stem and not to the major lexical category noun. we have also shown
that this constraint needs to be complemented by an alignment constraint requiring
coincidence of the right edge of the nisba suffix with the right edge of the PWd. We have
argued that these two constraints along with the undominated constraint *[N + Adj]PWd force
truncation of the nominal affixes in the toponyms in 43 and the leftmost stem of compound
bases
Having shown how the nisba adjectives with compound nouns and nouns with the
affixes [ta-...-t] and [ʔa-] are derived, we turn, in the following section, to show how
correspondence constraints, formulated within the OT framework, could account for the
causative forms in CMA.

5. PROSODIC CIRCUMSCRIPTION AS REDUPLICATION

As stated in the introduction to this chapter, Bennis (1992) has shown that the
causative could be analyzed as a case of prosodic circumscription. The domain circumscribed

187
prosodically is a minimal syllable which corresponds to CəC or CV (where V is one of the
full vowels [i, u, a]). Consider some causative forms derived from the base which corresponds
to the verb stem:

-47-

Base Causative Base Gloss

a. ktəb kəttəb write

dxəl dəxxəl enter
lʕəb ləʕʕəb play
tləf təlləf lose
wləd wəlləd give birth to
bki bəkki cry
dwi dəwwi speak
kma kəmma smoke

b. fiq fəyyəq wake up

Tiħ Təyyəħ fall down
TiR TəyyəR fly
ʃuf ʃəwwəf see
dub dəwwəb dissolve
nuD nəwwəD get up
bul bəwwəl urinate
gul gəwwəl say

According to Bennis (1992), the causative is obtained by the prefixation of σµµ to a

prosodically circumscribed syllable which is CV or CəC as in 47a, or the suffixation of σµµ to
a minimal CV syllable as in 47b. This affixation is obligatorily accompanied by left-to-right
or right-to-left spreading of the second segment of the base and this to satisfy a template
requiring that the causative consist of two syllables. Unlike Bennis, we assume that the
causative involves affixation of a monomoraic rather than a bimoraic syllable and this in
conformity with the claim made in Al Ghadi (1994) and Boudlal (to appear a), namely that
the schwa and a following consonant are dominated by a single mora. With this revision in
hand, let us see how words such as [bəkka] and [fəyyəq] could be derived

188
-48-

a. Input (= stem)

σ σ σ

µ µµ
| | |
bk a f I q
b. Consonant extrametricality and circumscription of a minimal syllable

σ σ

µ µ
| |
<b> k a f I <q>

c. Causative (Affixation of σµ)

σ σ σ σ
| |
µ + µ µ + µ
| |
k a f I

d. Right-to-Left/ Left-to-right spreading

σ σ σ σ
|
µ µ µ µ
|
k a f I

e. Restoring extrametrical consonants

σ σ σ σ

µ µ µ µ
|
b k a f I q

189
f. Schwa epenthesis

σ σ σ σ

µ µ µ µ
|
b ə ka f ə y ə q

The derivation of the causative in 48 is operational and is therefore incompatible with the
theoretical model adopted here. A circumscriptional analysis operates in successive steps: it
first scans for a prosodic constituent in the input, which is either CV or CəC. Then, it
performs an operation on that constituent by proceeding to the prefixation or suffixation of
the causative morpheme and the spreading of the base medial segment and schwa epenthesis.
Finally, the result is put together and mapped onto a disyllabic template.
As seen in section 3.2 above, McCarthy (1997) has shown that much of the burden of
operational prosodic circumscription could be taken on by ANCHOR-POS constraint family
which requires forms to match in specific aspects of prosodic constituency. It has been shown
that cyclic syllabification in CMA could be accounted for by invoking either the constraint O-
Ostem ANCHOR (σ, σ, Initial) requiring preservation of syllable-initial segments of the base
and derived output forms in the morphological constituent stem, or the constraint O-Oword
ANCHOR (σ, σ, Initial) requiring preservation of syllable-initial segments of the base and
derived output forms in the morphological constituent word.
For example, in the case of the causative form [kəttəb], we assume that the base is the
stem [ktəb] and that both [kəttəb] and [ktəb] are related in an O-Ostem correspondence relation
requiring that the syllable-initial segments of the base preserve their positions in the derived
output. Compare the candidates in 49 below to see to what extent they are related to the base.
For the sake of clarity, the geminate part indicating the causative is underlined.

-49-

/ktb, Caus/ O-Ostem ANCHOR ALIGN-L (Causative, PWd)

Base: [k.təb]word (σ, σ, Initial)
)a. kət.təb **
(b. kəb.təb **
c. kət.bəb *! ***

190
This tableau does not consider candidates such as *[kəktəb], where the first segment of the
root is geminated. The reason is that such a form is ruled out by an independently motivated
OCP constraint that prohibits geminates from occurring in initial position (see below for
details). Candidate 49c is excluded because it incurs a fatal violation of O-Ostem ANCHOR
(σ, σ, Initial): the initial segment of the second syllable, which is [b], does not correspond to
the initial segment of the same syllable in the base (i.e. the segment [t]). In 49a and 49b, both
candidates satisfy O- O-Ostem ANCHOR (σ, σ, Initial). The syllable-initial segment of the first
syllable (i.e. [k]) in the derived forms corresponds to the syllable-initial segment in the simple
form (i.e. the segment [k] that constitutes a minor syllable on its own). Similarly, the syllable-
initial segment of the second syllable (i.e. [t]) corresponds to the syllable-initial segment of
the second syllable in the simple base form. Note also that both 49a and 49b incur two
violation marks for the constraint requiring left alignment of the causative and the PWd.
One could possibly argue that given the fact that the causative formation yields a
geminate, an output form such as *[kəbtəb] could possibly be ruled out by invoking the NO-
CROSSING principle (Goldsmith 1976, 1979 and McCarthy 1979) as the representation in 50
shows. The lower case v stands for the schwa position:

-50-
*C v C C v C

k t b

The long distance consonant spreading is blocked in 50 exactly because it creates line-
crossing since the first part of the geminate denoting the causative is not contiguous to the
second part. Thus in order to derive the correct output, NO-CROSSING and O-Oword
ANCHOR (σ, σ, Initial) must dominate ALIGN-L (Causative, PWd) as the tableau below
shows:

-51-

/ktb, Caus/ NO-CROSSING O-Ostem ANCHOR ALIGN-L

Base: [k.təb]stem (σ, σ, Initial) (Causative, PWd)
)a. kət.təb *
b. kəb.təb *! *

191
Candidate 51 is ruled out because the underlined segment denoting the causative is the result
of a consonantal spreading that causes violation of NO-CROSSING.
Verb bases on the pattern CCV can be accounted for in the same way as bases on the
pattern CCVC considered in 51. For example, the causative form of a verb such as [b.ka] “he
cried” is [bək.ka] where the syllable-initial segments in the derived word correspond to the
syllable-initial segments in the base stem. The problematic cases are bases on the pattern
CVC whose causative form is realized as CəG.GəC, where G stands for glide. Thus, in an
example such as [fiq] “wake up!”, the causative form [fəy.yəq] violates the constraint O-Ostem
ANCHOR (σ, σ, Initial) because the initial segment of the second syllable in [fəy.yəq] (i.e.
the glide [y]) does not have a correspondent in the base [fiq]. This shows that the
circumscriptive analysis, reformulated within the correspondence model of McCarthy and
Prince (1995, 1999), is incapable of deriving the correct output causative form.
In what follows, we propose to analyze the causative within nonoperational CT
without reference to the O-Ostem ANCHOR (σ, σ, Initial) constraint. In particular, we will
make use of a proposal made by Imouzaz (forthcoming), namely that the causative formation
involves partial reduplication of the base. Stating that the causative morpheme is
reduplicative obviates the need for a template since the role played by this latter follows from
the interaction of universal constraints in the grammar of CMA. Moreover, we would not
have to resort to the NO-CROSSING principle to block long distance spreading in cases such
as *[kəbtəb] since this would derive from segmental copying as in reduplication (Gafos 1996,
1998).
It should be noted that the output of the causative is a disyllabic word which satisfies
FT-BIN (McCarthy and Prince 1993a and Prince and Smolensky 1993) and conforms to an
iambic foot of the type LL 8. Note also that the reduplicant in CMA is always the second
segment of the base. The causative reduplicative affix itself is not specified for any segmental
content; its realization depends on constraint interaction.
One of the constraints needed for the derivation of the causative is E-ALIGN (Root,
PWd) formulated by Nelson (1998) within McCarthy and Prince’s (1993b) Alignment
Theory. This constraint is stated as follows:

8
This is a piece of evidence for the line of analysis suggested in chapter three for the stress system of the
language, i.e. that the foot is iambic. It might even be said that the default foot of the language is of the type LL.
(See Boudlal to appear b, and chapter six below for examples supporting this assumption)

192
-52-

ALIGN-E (Root, PWd)

The left edge of the root must correspond to the left edge of the PWd and the
right edge of the root must correspond to the right edge of the PWd.

As predicted by Nelson (1998), the constraint in 52 forces infixation of a reduplicative

morpheme, exactly as is the case in CMA causative forms. This constraint also prevents total
reduplication of the base, thus forcing violation of MAX-Rt-BR:

-53-

MAX-Rt-BR
Every root segment of the base has a correspondent in the reduplicant.

The constraint ALIGN-E (Root, PWd) must dominate MAX-Rt-BR since only a single
segment of the base is reduplicated in the data in 47 above. It has also to dominate another
constraint, proposed in McCarthy and Prince (1995), demanding that the left edge of the
reduplicant correspond to the left edge of the base:

-54-

ANCHOR (Base, L, RED, L) (henceforth L-ANCHOR-BR)

The left edge of the base must correspond to the left edge of the reduplicant.

Violations of ALIGN-E (Root, PWd), MAX-Rt-BR and L-ANCHOR (RED, Base) are
assessed gradiently; each failed candidate receives one violation for every segment violating
ALIGN-E (Root, PWd) or L-ANCHOR (RED, Base).
With the constraints in 52, 53 and 54 in hand, let us proceed by showing how the
causative form is obtained. The constraint tableau in 55 shows how the constraints developed
above interact to give the output form [kəttəb] from the input /RED, ktb/. For clear
exposition, the reduplicant is underlined:

193
-55-

/RED, ktb/ ALIGN-E L-ANCHOR-BR MAX-Rt-BR

(Root, PWd)
) a. kət.təb * **
b. kək.təb **! **
c. kət.bəb **! **
d. kəb.təb **! **
e . k.təb.k.təb ****!

Two observations need to be made here: first, is that only consonants (including glides) and
high vocoids are copied. The low vowel [a] and the epenthetic schwa are never copied.
Second, all the causative forms resort to schwa epenthesis to yield a disyllabic output. This
epenthesis leads to the violation of DEP-IO which will not be shown in the tableaux here
since the constraint ranks low in the constraint hierarchy.
The optimal candidate in 55 satisfies higher-ranked ALIGN-E (Root, PWd) but incurs
one violation mark of L-ANCHOR-BR because the reduplicant copies the second segment of
the base and not the first. It also violates MAX-Rt-BR because the reduplicant is only one
segment of the base. Although the reduplicant in candidate 55b is left-anchored to the base, it
is excluded because it violates an undominated constraint, namely ALIGN-E (Root, PWd).
Both 55c and 55d are failed candidates because the reduplicant copies the third segment of
the base and not the second thus causing a fatal violation of L-ANCHOR-BR. Finally,
candidate 55e is ruled out because the reduplicant copies all of the base segments, thus fatally
violating ALIGN-E (Root, PWd).
Now let us return to the candidate where the reduplicant copies the first segment of
the base. A form such as *[kəktəb], where both the left and right edges of the root and the
PWd are aligned, would be hard to defend given the constraints in 55. It should win over the
optimal candidate since the left edge of the base anchors with the left edge of the reduplicant.
But as we have already pointed out, forms such as *[kəktəb] never arise since they violate the
OCP which prohibits words in CMA to start with initial geminates. In fact, initial geminates
do occur in CMA as the following examples show:

194
-56-

a. bbwa my father
mmwi my mother
xxwa brother/friend (colloquial)

b. l-DaR < DDaR the house

l-tub < ttub the fabric
l-ʃəmʃ < ʃʃəmʃ the sun
l-DəRb < DDəRb the hitting

The geminates in 56a are referred to as underlying geminates. Those in 56b are derived
geminates; they are the result of the assimilation of the definite article [l-] to a word-initial
coronal consonant. As we have already pointed out in chapter two, the first and second
members of the geminate belong to two different syllables, something that follows from the
constraint *COMPLEX. If this is so, then the prohibition against the occurrence of initial
geminates in cases such as *[kəktəb] could be seen as one referring to the first syllable of a
word9. The constraint is formulated within the Beckman (1998) Positional Faithfulness
Theory as follows 10:

-57-

*GEMINATES-σ1 (henceforth *GEM-σ1)

Geminates are prohibited in the first syllable.

This constraint will have to dominate the more general version of the markedness constraint
banning the occurrence of geminates (i.e. *GEM) as the tableau below expounds:

9
McCarthy (1997) has pointed out that imperfective initial gemination cannot apply in Berber words such
as[bb.xl] (where underlined segments are syllabic) because of an undominated constraint against syllabifying a
geminate as an onset and nucleus of a single syllable.
10
Beckman (1998) assumes that there are a variety of phonological asymmetries exhibited by segments in
certain perceptually or psycholinguistically prominent positions. Such positions include stressed syllables,
syllable onsets, roots, and root-initial syllables. The prohibition against initial geminates in CMA could be seen
as an example of positional faithfulness.

195
-58-

*GEM-σ1 *GEM

)a. kət.təb *
b. kək.təb *! *

This tableau shows that if geminates are to occur in CMA, they have to be heterosyllabic in
order to avoid violating *GEM-σ1.
Let us next see how candidates such as *[kək.təb] and *[kək.təb] are suboptimal. We
assume that *GEM-σ1 should outrank L-ANCHOR-BR and that *GEM-σ1 is not ranked
with respect to ALIGN-E (Root, PWd) as shown in 59:
-59-

/RED, ktb/ *GEM-σ1 ALIGN-E L-ANCHOR-BR

(Root, PWd)
) a. kət.təb *
b. kək.təb *(!) **(!)
c. kək.təb *!

Note that in this tableau, candidate 59b could be excluded either because the initial syllable
contains a geminate or because the left edge of the root is not aligned with the left edge of the
PWd.
Roots whose medial segment is a high vocoid need reference to the constraints in 55
along with ONSET, an undominated constraint in CMA. Verb bases of this type always
geminate their second segment and surface as CəyyəC or CəwwəC. Consider how [fəyyəq] in
60 wins over any of the other candidates. Here, we represent a high vocoid with the
underspecified segment /I/, but see the tableau below for a different assumption that will be
adopted in this work.

196
-60-

/RED, fIq/ ONSET ALIGN-E L-ANCHOR- MAX-Rt-BR

(Root, PWd) BR
)a.fəy.yəq * **
b. fi.iq *! * **
c. fiq.fiq ***!
d. fi.qəq **! **

It has now become clear that any output form violating the undominated ONSET will be
rejected and so will any form resorting to total reduplication. Such is the case with candidates
60b and 60c. The candidate in 60d is suboptimal because instead of copying the second
segment of the base, it copies the third and in so doing incurs two violations of L-ANCHOR-
BR which prove fatal. The optimal candidate incurs only a single violation mark of L-
ANCHOR-BR compared to candidate 60d. It also incurs two violation marks of DEP-IO
since it has resorted twice to schwa epenthesis.
Note that there are other candidates that have not been included in the tableau above
and these are candidates that tightly compete with the optimal one and seem to incur less
violations as to the constraints given in tableau 60. Take for example a possible candidate
such as *[fi.yəq] where the reduplicant is the high vowel [i] which corresponds to the glide
[y] in the base. Both *[fi.yəq] and [fəy.yəq] incur a single violation mark of L-ANCHOR-BR
by virtue of the fact that they copy the second segment of the base, and two violation marks
of MAX-Rt-BR by virtue of the fact that two base segments are not copied. However,
[fəy.yəq] incurs two violations of DEP-IO by epenthesizing two schwas whereas *[fi.yəq]
incurs only one. This points to the fact that other constraints are needed to distinguish the two
candidates.
To account for roots whose medial segment is a high vocoid, we assume, following
Rosenthall (1994), that vowel/glide alternation follows from constraint interaction and that
the difference between a vowel and a glide corresponds to association to a mora. In other
words, a glide is a high vocoid linked directly to a syllable node and a vowel is a high vocoid
associated to a mora. This way of viewing high vocoids eliminates underspecification for the
feature [consonantal] and as such the underspecified high vocoids /I/ and /U/, representing
both high vowels and their corresponding glides, will be simply represented as the vowels /i/

197
and /u/. The alternation between high vowels and their corresponding glides has been
accounted for in feature geometry by assuming that glides have a [+consonantal] root note
dominating a V-Place node while high vowels have a [-consonantal] root node dominating a
V-Place node (Clements and Hume 1995). The assumption underlying the present work about
high vocoids is that underlying high vowels are specified as [-consonantal] whereas
underlying glides are specified as [+consonantal] (Hyman 1985, Waksler 1990, Hume 1992
and Clements and Hume 1995). The realization of an underlying high vowel as a glide or an
underlying glide as a high vowel results in a change of featural specification of the input.
Thus an output such as *[fi.yəq] is a failed candidate because there is a lack of identity
between the base and the reduplicant in terms of featural correspondence. The constraint
needed to account for this mismatch is formulated along the lines suggested in McCarthy and
Prince (1995) and which demands featural identity of the base and the reduplicant:

-61-

IDENT-BR [cons]
The base featural specification for [cons] must be preserved in the reduplicant.

We assume that this constraint outranks another IDENT constraint, IDENT-IO [cons]
constraint demanding preservation of featural identity in the input/output mapping:
-62-

IDENT-IO [cons]
Featural specification for [cons] must be preserved in the input/ output mapping.

Let us see how the interaction of the two IDENT-IO [cons] constraints would favor
[fəyyəq] over other competing candidates:

-63-

/RED, fiq/ IDENT-BR [cons] IDENT-IO [cons]

)a.fəy.yəq *
b. fi.yəq *! *
c. fi.yəq *!

198
Both candidates 63b and 63c incur a fatal violation of higher-ranked IDENT-BR [cons] either
because the reduplicant is a high vowel and the base is a glide or vice versa. 63b fails exactly
because the reduplicant which is [i] in this case corresponds to [y] in the base; candidate 63c
is also excluded because the reduplicant [y] corresponds to the base [i]. The optimal candidate
63a satisfies IDENT-BR [cons] at the expense of low-ranked IDENT-IO [cons]. Here, the
input vowel associated to a mora is realized as a glide in the output, thus losing its moraic
status. Of a particular interest in tableau 63 is that the base copies the reduplicant, a state of
affairs which leads to the violation of input-output faithfulness. Given that the second
segment of the input is /i/, its realization as the glide [y] will automatically lead to the
violation of IDENT-IO [cons]. Such account is available only under CT where an identity
relation holds between the base and the reduplicant, on the one hand, and between the input
and the output, on the other.
One may wonder why a form such as *[fiyiq] is ruled out although it satisfies both
IDENT-BR [cons] and IDENT-IO [cons] by virtue of the fact that the reduplicant has a
correspondent in the base (the glide being epenthesized to satisfy ONSET). It should be noted
that the causative form is governed by a prosodic constraint which requires that the output
form consist exactly of an iambic foot of the type LL. The constraint LL will have to
dominate MAX-Rt-BR so as to exclude cases that resort to total reduplication or any other
type of feet as the tableau below shows:

-64-

LL MAX-Rt-BR DEP-IO
a. fəy.yəq ** **
b. fi.yiq *! ** *
c. fiq.fiq *!
d. fi.fiq *! *

All of the candidates, except 64a, are ruled out because they fail to conform to an iambic foot
of the type LL. The optimal candidate has proceeded to a double epenthesis of a schwa to
achieve the desired foot type. Note also that this candidate violates IDENT-IO [cons] but this
violation is achieved for the purpose of establishing featural correspondence between the base
and the reduplicant.

199
 Next, consider another causative case obtained from verbs whose second segment is
the high vocoid /u/. The tableau below lists some possible candidates from the input /RED,
dub/:

-65-

/RED, dub/ ONSET ALIGN-E L-ANCHOR- MAX-Rt-BR

(Root, PWd) BR
)a.dəw.wəb * **
b. du.ub *! * **
c. dub.dub ***!
d. du.bəb **! **

Once again, the constraints developed above predict that it is always the form that copies the
second segment of the base which is optimal. Of course, this form has to satisfy higher-
ranked constraints such as ONSET and ALIGN-E (Root, PWd). Note that the constraint LL
considered in 64 blocks total reduplication and in so doing does part of the functions
performed by ALIGN-E (Root, PWd). In the optimal candidate in 65, the input vowel, which
is /u/ loses the mora associated with it and surfaces as the glide [w] due to undominated
IDENT-BR [cons] requiring identity between the base and the reduplicant.
Finally, let us consider a causative form obtained from verbs whose final segment is a
vocoid. From the input /RED, bki/, Gen would allow the generation of output candidates such
as the following:
-66-

/RED, bki/ ALIGN-E L-ANCHOR-BR MAX-Rt-BR

(Root, PWd)
)a. bək.ki * **
b. bəy.ki **! **
c. bək.yi **! **

Both candidate 66b and 66c are excluded because they incur two violations of L-ANCHOR-
BR. The optimal candidate itself incurs only a single violation of the same constraint. As to
the IDENT constraint family, candidates 66b and 66c violate IDENT-BR [cons] because the
reduplicant, which is a glide, stands in correspondence with the base vowel.

200
 To conclude, we have argued that the causative in CMA could be accounted for
adequately by assuming a set of universal constraints ranked on a language-particular basis.
In essence we have shown that operational circumscription is not warranted for the different
steps it involves. We have also shown that the reanalysis of circumscription as prosodic
faithfulness to some designated syllable edge can neither account for verb bases having the
shape /CVC/ nor block long distance consonantal spreading. We have instead proposed an
output-output analysis based on the idea that the causative involves partial reduplication of
the second segment of the base. The output-output relation dealt with in this section involves
two strings produced simultaneously (i.e. the base and the reduplicant) which do not exist as
separate words. This relation differs from the output-output relation encountered in section 3
above involving separate words that are not produced simultaneously. (For a comparison
between the two output-output relations, the reader is referred to Benua 1995, 1997).

6. CONCLUSION
This chapter has tried to analyze phonological similarities between words that stand in
a transderivational relationship. It has been shown that cases originally attributed to cyclicity,
truncation and circumscription could be accounted for by output-output constraints
demanding correspondence between a derived form and its morphologically related form. The
cases considered were cyclic syllabification, truncation and finally causative formation.
In dealing with cyclic syllabification in CMA, we have proposed an analysis in terms
of a small subset of constraints, one of which demands faithfulness of the derived form to the
simple base form. Our notion of the base rests essentially on the definition set up in Basri et al
(1998) and Selkirk (1999). These authors have proposed a morphologically grounded theory
of O-O correspondence that distinguishes two different O-O faithfulness constraints: O-Oword
Faith and O-Ostem Faith. In order to account for cyclic syllabification in CMA, we have
proposed two different O-O constraints formulated after Selkirk (1999) and consistent with
the correspondence model of McCarthy and Prince (1995, 1999): O-Ostem ANCHOR (σ, σ,
Initial) and O-Oword ANCHOR (σ, σ, Initial). This distinction is based on the distribution of
subject and object suffixes. In particular, we have shown that O-Ostem ANCHOR (σ, σ,
Initial) must dominate O-Oword ANCHOR (σ, σ, Initial) based on items such as [DRəb-t]
which does not epenthesize a schwa before the subject suffix and [DəRbək] which

201
epenthesizes a schwa before the object suffix and hence violates O-Oword ANCHOR (σ, σ,
Initial). We have also shown that O-Ostem ANCHOR (σ, σ, Initial) is never violated except
when the markedness constraints ONSET and *COMPLEX are at stake. Resorting to a
morphologically-grounded theory of O-O correspondence to account for cyclic syllabification
obviates the need for any mechanisms, such as those used in rule-based systems, which try to
explain cyclicity by reference to intermediate stages in the input-output mapping process.
The chapter has also dealt with cases where part of the base is truncated in the
formation of some nisba adjectives derived from compound nouns and from nouns with the
affixes [ta-...-t] and [ʔa-]. We have argued that in order to adequately account for these nisba
adjectives, recourse needs to be made to morphology, and especially to the distinction
between the stem and the word. We have shown that truncation follows from ranking *[N +
Adj]PWd along with the affixation constraint AFFIX-TO-Nstem and the alignment constraint
ALIGN-Nisba-R above the constraint MORPH-REAL, thus forcing deletion of the leftmost
stem of a compound and the nominal affixes [ta-...-t] and [ʔa-] in toponyms.
Finally, following Imouzaz (forthcoming) we have argued that the causative could
adequately be accounted for in terms of O-O correspondence relating the base and its
reduplicant. We have shown that the constraint ALIGN-E (Root, PWd) proposed in Nelson
(1998) is very active in CMA since it prevents total reduplication and forces the reduplicant
to be an infix. We have also shown that the causative form of words such as [fəyyəq] “wake
up” and [dəwwəb] “dissolve”, derived from the roots /fiq/ and /dub/, exhibits a special
behavior in that the base in both forms copies the reduplicant which is [y] in the first item and
[w] in the second. Such an explanation could only be achieved if we posit constraints on the
output such as the one between the base and the reduplicant.

202
 Chapter Five

The Prosodic Nature of

the Passive Participle in Casablanca Moroccan Arabic

1. INTRODUCTION
One of the most productive morphological operations in MA is the derivation of the
passive participle (PP). It has received different treatments according to different linguists. The
difference resides in the nature of the PP morpheme. Some linguists (see Marsil 1988, for
example) assume that the PP is obtained via the affixation of the discontinuous morpheme
[m...u...] to the base, which is assumed to be the verb root. [m-] is prefixed to the root while [-u-]
is infixed after the second segment of the base. Others (Youssi 1986, and Boudlal 1993) assume
that the PP morpheme is simply the morpheme [m-] which is prefixed to the base.
This chapter gives further support to the stress pattern of CMA and consequently the foot
types proposed in chapter 3. The idea defended here is that the PP marker is [m-] and that the
prefinal vowel that appears in some forms is the lexical default segment /u/ which is epenthesized
to conform to an iambic foot of the type LH. It will be shown that the LL foot type arises only in
cases where the epenthesis of /u/ does not take place, something that suggests which there are
other constraints that outrank those governing foot types. In particular it will be shown that the
PP forms that do not end up in a heavy syllable can only be obtained by making recourse to
constraints on output-output correspondence much in the spirit undertaken in chapter four.
The chapter is divided into three major sections. In section 2, a taxonomic survey of the
PP forms is offered. The verb bases considered include: sound verbs, geminated verbs,
assimilated verbs and verbs whose segments contain high vocoids. Both derived and non-derived
trisegmental and quadrisegmental verbs are considered. Section 3 deals with previous account of
the PP. Therein we expose two different approaches and show the limitations of each. In section
4, we present an alternative analysis couched within the OT framework. Here two hypotheses
will be advanced to explain the behavior of the PP: the first says that there are prosodic
constraints on the PP verb stem; the second, which is adopted in the present work, maintains that
these constraints hold on the PP word rather than the verb stem. Throughout the chapter, it will be
argued that the analysis that works best for CMA is one where the output of the PP corresponds
to an iambic foot of the type LH or else LL.

2. A TAXONOMIC SURVEY OF THE PASSIVE PARTICIPLE

This subsection will give a survey of all PP forms obtained from both derived and non-
derived transitive verb bases. The word base is taken here to be a root, a stem or a word.

2.1 The PP of Trisegmental Verb Bases

2.1.1 Non-Derived Verbs

Consider the sets in 1 below. The base in 1a represents the class of sound verbs; 1b
represents the class of verbs whose medial segment is one of the high vocoids /i/ or /u/; 1c
represents the class of assimilated verbs; 1d is that of final geminated verbs, and finally 1e
represents the class of verbs whose second and/or third segment is a glide or a high vocoid.
-1-
Base PP PP Gloss

a. ktb məktub written

lʕb məlʕub played
ʃrb məʃrub drunk

b. ʃuf məʃyuf seen

biʕ məbyuʕ sold
dir mədyur done

c. wqf məwquf stopped

wzn məwzun weighed
wld məwlud born

d. sdd 1 məsdud locked

ʕDD məʕDuD bitten
hzz məhzuz taken

1
The final consonant in these items is doubled to show they behave like the rest of trisegmental verbs and also to
encode the notion of the Two-Root Theory of geminates proposed in Selkirk (1990, 1991). See chapter two and
section three in this chapter for details about the representation of geminates.

204
e. ʃri məʃri bought
kri məkri rented
ʃwi məʃwi roasted

A comment about roots containing a high vocoid is in order. In chapter four, we assume,
following Rosenthall (1994), that verbs with medial high vocoids such as /fiq/ and /dub/ are
represented underlyingly with the vowels /i/ and /u/ and that the non-moraic realization of these
vowels as glides in words such as [fəyyəq] and [dəwwəb] is the result of constraint interaction.
This assumption works for verbs with medial vocoids as has already been shown in chapter four
and also with verbs with final high vocoids such as /ʃri/, and /kri/ (Cf. [ʃərray] “buyer”and
[kərray] “tenant”, where the input high vowel is realized as the glide [y]). Other examples that
exhibit a vowel/glide alternation are given below:

-2-

Verb Causative Agent Noun Gloss

a. dub dəwwəb — dissolve

ʃuf ʃəwwəf ʃəwwaf see
bul bəwwəl bəwwal urinate
gul gəwwəl gəwwal say
nuD nəwwəD — get up
fiq fəyyəq fəyyaq wake up
Tiħ Təyyəħ — fall down
biʕ bəyyəʕ bəyyaʕ sell

b. ʃri — ʃərray buy

kri — kərray rent
bki bəkki bəkkay cry
ʒri ʒərri ʒərray run

For all these cases where there is a vowel/glide alternation, we will continue to assume that high
vocoids are underlyingly represented with /i/ and /u/.
As to the cases where the distribution of high vowels and glides cannot be attributed to an
alternation between the two, Rosenthall (1994) posits glides in the underlying representation.
This could be the case of trisegmental verbs whose initial segment is a high vocoid and verbs
whose second and third segments are vocoids as shown in the examples below:

205
-3-

Vb Root Vb Stem Gloss

a. wzn wzən *ʔuzən weigh

wld wləd *?uləd give birth to
wqf wqəf *?uqəf stand
wSl wSəl *?uSəl arrive
wħl wħəl *?uħəl get stuck
ybs ybəs *?ibəs get dry

b. lwi lwi *luy twist

kwi kwi *kuy burn
dwi dwi *duy speak
Twi Twi *Tuy fold
nwi nwi *nuy intend

The asterisked forms show that the initial glide in 3a and the medial glide in 3b never alternates
with the corresponding high vowel, at least in CMA2. In 3a the initial glide can never surface as a
vowel (the glottal stop is prothetic and serves as an onset). Similarly, the medial glide in 3b is
always realized as non-moraic. For verbs like these ones, we assume that the glide is part of the
underlying form.
The derived verb bases of the forms corresponding to the ones in 1a-e form their PP in the
same way as the non-derived counterparts. Thus derived trisegmental verb bases such as
[təktəb], [tʃədd] and [təwzən] have the forms [məktub], [məʃdud] and [məwzun] as their
respective PP.
The patterns of the PP forms of non-derived verbs are listed in 4 below. The period marks
syllable boundary; G stands for glide; the subscript means that the two segments are identical and
finally the letters correspond to those of the sets given in 1 above:

-4-

a. CəC.CuC
b. CəC.GuC

2
Benkaddour (1982) assumes that in Rabati MA, high vocoids are allowed to surface as vowels in initial position.
Thus, for the author, forms such as [ʔuSəl] and [ʔibəs] are attested in this variety.

206
 c. CəG.CuC
d. CəC.CiuCi
e. CəC.Ci / CəC.Gi

2.1.2 Derived Verb Bases

The forms in 5 below are derived either by medial gemination of the second segment of
the root or by the prefixation of the medio-passive morpheme [t-] to the geminated form:

-5-

Vb Base PP PP Gloss

a. kəttəb mkəttəb written

tkəttəb mkəttəb
ləʕʕəb mləʕʕəb played
tləʕʕəb mləʕʕəb

b. ʃəwwəf mʃəwwəf seen

tʃəwwəf mʃəwwəf
dəwwəz mdəwwəz examined
ddəwwəz mdəwwəz

c. wəqqəf mwəqqəf stopped

twəqqəf mwəqqəf
wəlləf mwəlləf be accustomed to
twəlləf mwəlləf

d. rəqqəq mrəqqəq made thin

trəqqəq mrəqqəq
nəZZəZ mnəZZəZ toasted
tnəZZəZ mnəZZəZ

e. ʕərri mʕərri uncovered

tʕərra 3 mʕərri
γəTTi mγəTTi covered
TγəTTa mγəTTi

3
Verbs of this type show an alternation between the vowel [i] and the vowel [a]. Thus they have two allomorphs;
one with [i] as in the passive [mʕərri], the other with [a] as in [tʕərra]. (For i/a and u/a allomorphy in MA, the reader
may refer to works such as Al Ghadi (1990), Bennis (1992), Boudlal (1993) and Meliani (1995).

207
Notice that the medio-passive morpheme [t-] takes on the voicing feature of the root initial
consonant if it is coronal as is the case with the verb [ddəwwəz]. Note also that whether or not the
mediopassive morpheme is prefixed to the base, the PP obtained is always the same. The patterns
of the PP forms in 5 are listed below:

-6-
a. CCəCi.CiəC
b. CCəGi.GiəC
c. CGəCi.CiəC
d. CCəCi.CiəCi
e. CCəCi.Cii

There is a second category of the PP obtained from derived verb bases. These verb bases
are themselves derived by the infixation of the consonant [-t-] (and possibly the vowel [-a-] that
follows it) after the first segment as shown below:

-7-

Vb Base PP PP Gloss

a. htəmm məhtəmm be interested in

ħtəʒʒ məħtəʒʒ protesting

b. ħtaʒ məħtaʒ be in need of

rtaħ mərtaħ relaxed

c. xtarəʕ məxtarəʕ invented

ħtarəm məħtarəm respected

Most of the forms in 7 are classicized forms, reported to be used in MMA (Youssi, 1992). The
patterns of the items in 7 are given in 8:

-8-
a. CəC.CəCiCi
b. CəC.CaC
c. CəC.Ca. CəC

208
 One final category of derived verbs which deserves special attention is the one given
below:

-9-

zyan to become nice *məzyan

sman to become fat *məsman
byaD to become white *məbyaD
ħmaR to become red *məħmaR
kħal to become black *məkħal

These forms do not have a PP simply because they are all intransitive, and as it has already been
pointed out, only transitive verbs allow the PP.

2.2 The PP of Quadrisegmental Verb Bases

Again here both non-derived and derived verbs will be considered. The only type of
derived verbs obtained is by the prefixation of the medio-passive prefix [t-] or [tt-], depending on
the variety under study.

2.2.1 Non-derived verb bases

The PP is obtained by simple prefixation of the morpheme [m] to the verb base as shown in 10
below:

-10-

Base PP PP Gloss

a. TəRʒəm mTəRʒəm translated

kərkəb mkərkəb rolled
fərgəʕ mfərgəʕ exploded

b. ʃawər mʃawər consulting

SafəR mSafəR Traveling
sayən msayən waiting for

c. quləb mquləb deceived

surət msurət locked
SifəT mSifəT sent

209
d. sərbi msərbi quickly done
SəqSi mSəqSi asking for

e. sali msali finished

qadi mqadi finished

The patterns that could be obtained from the items in 10 are the following:

-11-
a. CCəC.CəC
b. CCa.CəC
c. CCu.CəC/CCi.CəC
d. CCəC.Ci
e. CCa.Ci

2.2.2 Derived Verb Bases

These verb bases behave much like those in 10 with respect to the formation of the PP.
Consider the items in 12 for illustration

-12-

Vb Base PP PP Gloss

a. TTəRʒəm mTəRʒəm translated

tfərgəʕ mfərgəʕ exploded
tquləb mquləb deceived
tsurət msurət locked

b. TSəqSa mSəqSi asked

tsərba msərbi quickly done
tsala msali finished
tqada mqadi finished

Since the derived verbs in 12 have the same PP as their corresponding non-derived forms, it
follows that their patterns do not differ from those given in 11 above.
It should be noted that the PP of some medio-passive verbs may be realized differently if
[tt] is prefixed instead of [t]. Consider the items below for illustration:

210
-13-
Vb Base PP PP Gloss

a. ttwəqqəf m(ət)wəqqəf stopped

ttrəqqəq m(ət)rəqqəq made thin
ttnəZZəZ m(ət)nəZZəZ made crunchy

b. TTγəTTa m(əT)γəTTi covered

ttʕərra m(ət)ʕərri uncovered

c. ttquləb m(ət)quləb cheated

ttsurət m(ət)surət locked

These examples show that if the medio-passive prefix is [tt], the PP will have two realizations:
one with two major syllables, the one which is widely used in CMA, the variety under
consideration; the other with three. Notice also the degemination of [tt-] in the PP. The patterns
of the trisyllabic realizations are given below:

-14-
a. CəC.CəC.CəC
b. CəC.CəC.Ci
c. CəC.CV.CəC

To sum up the different patterns of the PP in CMA have been reduced to the ones in 15. C stands
for consonant (geminate or non geminate) and V stands for any of the full vowels [a, u, i].

-15-
Pattern Examples

a. CəC.CVC məktub, məʕDuD

b. CəC.CV mək.ri, məʃri
c. C.CV.CV msali, mqa.Di
d. C.CəC.CəC mdəwwəb, mrəqqəq
e. C.CəC.CV mγəTTi, mʕərri
f. C.CV.CəC mquləb, mSifəT
g. CəC.CəC.C məhtəmm, məħtəʒʒ
h. CəC.CəC.CəC məTTəRʒəm, mətwəqqəf
i. CəC.CəC.CV mətγəTTi, mətʕərri
j. CəC.CV.CəC mətquləb, məxtarəʕ, məħtarəm

211
Note here that the more regular and frequently used pattern in CMA is the disyllabic one (15a-g)
and that the trisyllabic one is generally another variant of the first type which is used instead
(15h-j).

3. PREVIOUS ACCOUNT OF THE PASSIVE PARTICIPLE

In this section we expose two different pre-OT analyses undertaken within two different
theoretical frameworks and disclose the limitations of each. The first is represented by Marsil
(1988). Marsil assumes that the PP morpheme has two allomorphs: [m] and [m...u...]. The
allomorph [m...u...] is attached to non-derived trisegmental verb bases, whose final segment is
not a glide while the allomorph [m] is affixed to non-derived quadrisegmental bases, derived
trisegmental bases and non-derived trisegmental bases whose final segment is a glide. Marsil has
to posit a segmental template with five positions for trisegmental roots and assume a
morphologically restricted rule that applies to verbs whose final segment is a vocoid. This rule
has the effect of deleting a segmental position from the template so as to block the epenthesis of
the prefinal [u].
The analysis undertaken in Marsil (1988) does not have an explanatory power and
therefore cannot be considered adequate. Apart from the criticism leveled at the nature of the
template chosen (i.e. a segmental template), the analysis fails to explain the presence of [-u] in
some forms and its absence in some others. In so doing, it treats the PP of non-derived
trisegmental verb bases differently.
Boudlal (1993) offers a different analysis to the effect that the PP morpheme is
systematically [m] in all the forms. According to the author, the segment [-u-] ( which he
represents with the archisegment /U/), appearing before the final segment of the base, is not part
of the PP morpheme; rather, it is the lexical default segment which is realized as [u] in the
nucleus position of a syllable and as [w] in the margin position. According to Boudlal, the
presence of [u] in the PP forms is dictated by the Template Satisfaction Condition (McCarthy and
Prince, 1986) which requires that all the positions of a template be satisfied. For example, the
items [məʃri], [məktub] and [msali] would be derived in Boudlal’s (1993) model as follows:

212
-16-

Stratum 1

Input: XX[X]X XX[X]X XX[X]X

ʃ r I k t b s a l I

The mapping of the melodic elements on to the template positions proceeds from the edges
inward by associating the non-marked positions first (the marked positions, being the bracketed
X’s). The bracketed X is associated only when it finds an available melodic element; otherwise it
is filled by the lexical default segment as a result of the Template Satisfaction Condition (TSC):

-17-

Assoc: XX[X]X XX[X]X XX[X]X

| | | | | | | | | |
ʃ r I kt b s a l I

U U
| |
TSC XX[X]X XX[X]X __
| | | | | |
ʃr I kt b

TC XXXX XXXX XXXX

| | | | | | | | | | | |
ʃ r UI k t Ub s alI

In order not to generate an ungrammatical form, Boudlal (1993) has to posit a rule which
assimilates /U/ to /I/ before another /I/, and then another rule which deletes one of the two
contiguous I’s as a result of a dissimilation process which is independently justified. The forms
obtained are then subject to CV-syllabification. After the prefixation of the PP marker at stratum
2, and schwa epenthesis at the post-lexical stratum, we obtain the phonetic representations:

213
-18-
σ σ σ σ

CV-syll. XXX XXXX XXXX

| | | | | | | | | | |
s‡r I k tU b sa l I

σ σ σ σ Ap σ σ

-Aff.& syll. XXXXX XXXXXX XXXXX

| | | | | | | | | | | | | | | |
məʃrI mək tUb m sa l I

The analysis provided in Boudlal (1993) is advantageous in many respects. First, it gives a
unitary morpheme to all the PP forms. Second, it provides a unitary analysis for all the
participles, regardless of whether they are derived or non-derived, trisegmental or
quadrisegmental. Third, it shows that the segment [-u-], found before the final segment of the
verb base, is no more than the lexical default segment which fills positions that would otherwise
be subject to Stray Erasure. However nice this analysis seems to be, it is not unquestionable. First
it posits templates that consist of timing units (X’s) whose prefinal slot is marked. Development
in Prosodic Morphology (McCarthy and Prince 1986) has shown that morphological rules have
access not to segmental skeleta but to prosodic categories such as the syllable and the foot.
Second, an input form has to go through different stages before reaching the phonetic
representation. This complicates the grammar of CMA since the lexical default segment has to be
epenthesized to obey the TSC, and then deleted to attain the attested output. Third and more
importantly, the analysis presented above fails to capture the relationship between prosody and
morphology, namely that the PP as a morphological operation is to a large extent governed by
constraints on the prosodic structure of CMA.
In a later work, Boudlal (1996) reanalyzes the PP in the framework of Prosodic
Morphology conceived within early OT (McCarthy and Prince, 1993). The core of the analysis
rests on the fact that Prosody dominates Morphology (P>>M). Boudlal proposes that the PP in
CMA should consist of exactly an iamb of the type LL and that the possible PP patterns of the
language are the ones grouped in 16 below:

214
-19-
a. CəC.CVC
b. (C)CəC.CV
c. CCV.CəC
d. CCV.CV

Following McCarthy and Prince (1990b), Boudlal (1996) assumes that the number of consonants
the onset consists of does not bear upon syllable weight, something that finds its justification in
the stress system. He further assumes that the final consonant in a final CVC syllable is
associated with an extrasyllabic mora as shown below:

-20-

Extrasyllabicity in final position:

σ

µ(µ)
| |
CVC

Assuming the representation in 20 has led Boudlal (1996) to conclude that all the PP’s in CMA
must conform to an LL iamb as shown in 21 (the brackets indicate extrasyllabic material):

-21-

LL iamb in MA:

a. Ft b. Ft c. Ft

σ σ σ σ σ σ

(µ) µ µ (µ) (µ) µ µ (µ) µ µ

| | | | | | | |
(C) C ə C CV (C) (C) C V C ə C (C) C V C V

The foot in 21a corresponds to the items which have the pattern in 19a and 19b; the feet in 21b
and 21c correspond respectively to the items which have the pattern in 19c and 19d. The

215
extrasyllabic consonants in 21a are mutually exclusive; the presence of one entails the absence of
the other.
The analysis above tries to give a unified account of the PP formation by making recourse
to the foot. However, assuming that heavy syllables do not occur word finally would neutralize
weight distinctions in final position, thus wrongly predicting that only monomoraic light syllables
occur in this position. Facts about word stress at the end of a PPh show that final heavy syllables
do occur and because they are heavy, they bear stress.

4. ALTERNATIVE OT ANALYSIS
4.1 Introduction

In this section, we attempt to reanalyze the PP in CMA within the OT framework. In

particular, we will show that an adequate analysis of this morphological category could be
attained by the ranking of faithfulness constraints requiring that the input be identical to the
output and prosodic constraints requiring that the output of this category conforms to an iambic
foot of the type LH in the ideal cases or an iamb of the type LL in the worst cases but never an
iamb of the type H or a minor LH iamb whose light syllable is minor.
The assumption underlying the analysis of the passive forms is that the morpheme marker
is [m-] and that the prefinal vowel [u] that shows up in forms such as [məktub] is epenthesized
for prosodic purposes. Boudlal (1993) has argued that this epenthesized vowel is no more than
the default segment transcribed as /U/ which shows up as either [u] or [w] depending on its
position in the syllable. Within a lexical-phonology framework, the author assumes that MA
distinguishes between two default segments; the lexical default segment which is /U/ and the
postlexical default segment which is the schwa [ə]. That /U/ is the lexical default segment is
corroborated by data from CMA as the cases below show:

-22-

a. Base Diminutive Gloss

Dar dwira house
sarut swirit key
ħanut ħwinit shop
RaS RwiyyəS head

216
b. Singluar plural Gloss
sarut swarət key
ħanut ħwanət shop
saʕa swayəʕ hour
Tabla Twabəl table

From the descriptive point of view, the data above show that both the diminutive and the broken
plural start with an initial CC cluster, where the second C is [w]. Given the fact that the base
starts with a single consonant, recourse is made to the lexical default segment of the language to
fill in the gap and satisfy the required prosodic shape. (See chapter 6 for a detailed analysis of the
diminutive and the need for positing a constraint forcing epenthesis of the default segment of the
language).
In the present work, we assume that CMA has two default segments: the schwa ([ə]),
which is epenthesized for syllabic purposes, and the lexical default segment, which is represented
underlyingly as /u/. In the passive participle, /u/ is always realized as a vowel, whereas in cases
such as the diminutive and the plural, it loses its moraic status and gets realized as a glide.
Assuming the PP maker to be [m-] and /u/ to be epenthesized for prosodic requirement, it remains
to explain the nature of the verb base to which the PP prefix is attached. The data presented in
section 2 above show that it is the verb stem which serves as a base to the PP. Thus the structure
of the PP form [məktub] is the one given in 23 below:

-23-

Word

Af Verb stem

m ktub

This structure shows that the constituent which is sister to the PP morpheme is the verb stem and
that that stem has undergone u-epenthesis. In OT terms, u-epenthesis means violation of DEP-IO.
At this stage, it is worthwhile to ask the following question: what is it that forces the epenthesis
of /u/ in items such as [məktub] rather than [ə]? In other words how should we distinguish

217
between a trisegmental stem such as /ktub/ in 23, where /u/ is epenthesized and a trisegmental
verb root such as [ktəb], where it is the schwa [ə] rather than /u/ which is epenthesized?
To answer this question, one has to know how the language distinguishes syllable weight.
In chapters two and three, we have shown that CVC syllables formed with a schwa are light,
whereas those with full vowels like /u/ are heavy. This points out to the fact that weight is the
crucial factor in deciding which vowel to epenthesize. That it is a schwa rather than /u/ which is
epenthesized in [ktəb] suggests that there must be a domination relation between the constraints
that penalize epenthesis. Here there should be a distinction between DEP-ə and DEP-u. Consider
the following tableau for the derivation of the verb root [ktəb]. Recall from chapter 2 that the
constraint PARSE-seg, which requires segments to be organized into syllables along with
constraints on possible syllables is what forces epenthesis.

-24-

/ktb/ PARSE-seg DEP-ə DEP-u

a. ktb ****!
b. k.təb **!
0c. k.tub **

The ranking in 24 wrongly predicts that the correct output is [ktub] rather than [ktəb]. This result
calls for a reranking of the constraints; the correct ranking would be one where
DEP-u outranks DEP-ə as shown in the tableau below:

-25-

/ktb/ PARSE-seg DEP-u DEP-ə

)a. ktəb *
b. k.tub *!

Having shown that DEP-u must dominate DEP-ə and that an input such as /ktb/ can never surface
as [ktub], an explanation of why it surfaces as such in the PP is in order. As a matter of fact, the

218
constraints in 25 predict that the PP of an input form like /m-ktb/ would be [məktəb] rather than
[məktub] as the tableau below shows:

-26-

/m-ktb/ PARSE-seg DEP-u DEP-ə

a. mk.tub **! *
b. mək.tub *! *
0c. mək.təb **

The account given in 26 shows that the passive forms need more than just these three constraints
which wrongly predict that the output is [məktəb] and not the correct form [məktub] 4. Although
these constraints predict that it is the schwa rather than the default vowel /u/ which is
epenthesized in trisegmental verb roots such as /ktb/ and /krkb/, they fail to explain why in
quadrisegmental and derived trisegmental verb bases, u-epenthesis is blocked from applying in
the PP (cf. [mkərkəb] “rolled” but not *[mkərkub] and [mkəttəb] “written” but not *[mkəttub])
while it is allowed to apply in certain trisegmental verb bases (cf. [məktub] “written” but not
*[məktəb]). Perhaps the explanation is to be sought in the morphology, and more particularly in
its interaction with prosody.

4.2 Verb Prosodization

Recent work on MA (Al Ghadi 1990, Imouzaz 1991, El Himer 1991, Bennis 1992, among
others) shows that verb syllabification is governed by prosodic templates instead of structure
building syllabification rules of the type proposed in Benhallam (1990a). Non-derived
trisegmental verbs such as [ktəb] “write” and [wləd] “give birth to” would have the
representations in 27a.i whereas non-derived trisegmental verbs such as [dir] “do”, [ʃuf] “see”
and [ʕəDD] “bite” would have the representation in 27b.ii. On the other hand all non-derived

4
The item [məktəb] exists but as a noun with the meaning “desk/office” and is used in MMA (Youssi 1992). Instead
of [məktəb], the word used in CMA is [biru].

219
quadrisegmental verbs will have the structure in 27b . Examples of such verbs include [kərkəb]
“roll”, [SifəT] “send” and [sala] “finish”.

-27-

a. Non-derived Trisegmental verbs

(i) PWd 5 (ii) PWd

│ │
F F
│
σ σ σ
│ │
µ µ µ µ

b. Non-derived quadrisegmental verbs

PWd
│
F

σ σ
│ │
µ µ

The first generalization that could be made about the representations in 27 is that they are all
bimoraic. This derives from the constraint FT-BIN which requires that a foot be binary under
syllabic or moraic analysis (McCarthy and Prince 1993a, and Prince and Smolensky 1993). The
second generalization that could be made about the structures in 27 is that non-derived
trisegmental verbs, unlike quadrisegmental ones, always surface with one major syllabe, a fact
which is determinant in the derivation of the PP.
The fact that non-derived verb roots in CMA proceed to the epenthesis of a schwa rather
than a full vowel is a result of the requirement that a verb be exactly bimoraic. Thus the grammar
of CMA should incorporate a constraint of the type VERB ROOT = [µ µ], already seen in
chapter two and repeated in 28 below:

5
This template looks like that of non-derived quadrisegmental verbs. However, this is not true because in
trisegmental verbs, one of the two monomoraic syllables (either the left or the right syllable) is dominated by a minor
syllable.

220
-28-

VERB ROOT = [µ µ]
A verb root must correspond to two moras.

In 29 below, we show how a trisegmental verb is obtained. The bold face V stands for any of the
full vowels [i, u, a]:

-29-

/ktb/ VERB ROOT = [µ µ] DEP-V DEP-ə

)a. kµ.təbµ *
b. kəµ.təbµ **!
c. kµ.tVbµµ *! *
d. k.tVbµµ *!

Candidate 29b is ruled out because it incurs two violations of the constraint DEP-ə. Candidate
29c is also ruled out because it makes recourse to a full-vowel epenthesis, thus violating
bimoraicity. Finally candidate 29d epenthesizes a full vowel instead of a schwa, a fact that causes
a fatal violation of DEP-V.
The same analysis could be extended to non-derived quadrisegmental verbs. Consider the
tableau below for the derivation of [kərkəb].

-30-

/krkb/ VERB ROOT = [µ µ] DEP-V DEP-ə

)a. kərµ.kəbµ **
b. kəµ.rəkµ.bµ *! *
c. kVrµµ.kVbµµ *! **

It is clear that any analysis that does not take into consideration the moraic quantity of the root
will fail to account for the verbal morphology of CMA.

221
 The effect of the bimoraicity constraint could be seen even if we adopt Prince and
Smolensky’s (1996) notion of richness of the base and posit verbs with underlying full vowels in
the input. Thus, any proposed system of constraint interaction must rule out forms that surface
with a full vowel and allow only forms that epenthesize a schwa. Consider an input with an
underlying full vowel such as /ktVb/. To get the correct output [ktəb], the constraint VERB
ROOT = [µ µ] must dominate MAX-V to make sure that the vowel will delete:

-31-

/ktVb/ PARSE-seg *COMPLE VERB ROOT MAX-V DEP-ə

X = [µ µ]
)a. kµ.təbµ * *
b. kµ.tVbµµ *!
c. k.tVbµµ *!
d. ktVbµµ *!

Candidate 31b has maintained the vowel of the input, something that causes a fatal violation of
bimoraicity. Candidate 31c satisfies bimoraicity at the expense of PARSE-seg, another
undominated constraint in the language. Finally candidate 31d satisfies both PARSE-seg and
bimoraicity but is ruled out on the ground that it violates *COMPLEX.
Next, consider a quadrisegmental verb such as /kVrkVb/ with underlying full vowels. The
constraints in 32 would derive the correct output [kərkəb]:

-32-

/kVrkVb/ PARSE-seg *COMPLEX VERB ROOT MAX-V DEP-ə

= [µ µ]
)a. kərµ.kəbµ ** **
b. kVrµµ.kVbµµ *!
c. kərµ.kVbµµ *! *
d. k.rkVbµµ *(!) *(!) *
e. kµ.rəkµµ.b µ *! ** *

222
Candidates 32b, 33c and 32e are all ruled out because they fatally violate bimoraicity. Candidate
32d could be ruled out either because it violates PARSE-seg or *COMPLEX. The winner
candidate spares all the higher-ranked constraints although this has been done at the expense of
MAX-V and DEP-ə, two lower-ranked constraints.
The constraints developed above could not only account for verbs with underlying full
vowel but also for hypothetical five-consonant verbs as could be seen from the tableau in 33. For
such verbs we need to assume that bimoraicity also dominates MAX-C:

-33-

/CCCCC/ *COMPLEX VERB ROOT MAX-C MAX-V DEP-ə

= [µ µ]
)a. CəCµ.CəCµ * **
b. Cµ.CVCµµ.CVCµµ *!
c. Cµ.CəCµ.CVCµµ *! *
d. C.Cµ.CCVCµµ *(!) *(!)
e. CəCµ.CəCµµ.Cµ *! **

To recapitulate, it has been shown that verbs in CMA abide by a prosodic constraint
which requires that the output be exactly bimoraic. This bimoraicity requirement explains why a
sequence such as /ktb/ surfaces as [ktəb] while the same sequence in a form such as /m-ktb/
surfaces as [məktub], with an epenthesized /u/ rather than a schwa. The first is a root and is
subject to the bimoraicity constraint; the second is an affixed stem which together with the PP
morpheme must conform to a prosodic constraint requiring that the foot be iambic as will be
shown in the next section.

4.3 An Iamb-Based Analysis of the Passive Participle

The assumption defended here is based on the idea that the output of the PP is governed
by a prosodic constraint which demands that it conform to an iambic foot. Prince and Smolensky
(1993) and McCarthy and Prince (1993a) maintain that there should be a hierarchy among foot
types in the grammar of languages. This means that a domination relation should be established

223
among iambic feet in CMA. Since the epenthesis of /u/ creates a final heavy syllable, we assume
that the most harmonic type of iambic foot is LH and as such should dominate LL. Incorporating
such constraints (i.e. LH and LL) in the grammar of CMA along with other markedness and
faithfulness constraints will allow us to explain u-epenthesis in the PP’s obtained from non-
derived trisegmental verb bases and its absence in PP’s obtained from derived trisegmental verb
bases and quadrisegmental verb bases. It will be shown that when LH is dominated, the foot
obtained is of the type LL but never H or a minor LH iamb where the first syllable is minor. It
will also be shown that these constraints operate on the output of the PP rather than on the verb
stem which serves as a base for derivation.
At a first stage, let us explore the possibility that the constituent which is the sister to the
PP prefix is what needs to be counted as the stem and that this stem is governed by a prosodic
constraint requiring that it conform to an iambic foot. To see whether this is actually the case,
consider representative items for the possible PP patterns in CMA.6

-34-

i. PP of trisegmental verb stems

a. m-C.CuC mə-k.tub
b. m-C.CaC mə-r.taħ
c. m-CəCi.Ci əC m-kət.təb
d. m-C.CV.CəC mə-x.ta.rəʕ
e. m-C.CəCi.Ci mə-h.təm.m
f. m-C.CV mə-k.ri

ii. PP of quadrisegmental verb stems

a. m-CəC.CəC m-kər.kəb
b. m-CV.CəC m-su.rət
c. m-CV.CV m-sa.li

The first generalization that could be made about the items in 34 is that all quadrisegmental verb
stems and few cases of trisegmental stems (34i.c and 34i.d) conform to a bisyllabic iambic foot of
the type LL, where both syllables are major syllables. The second generalization concerns some

6
The list does not include verb stems such as [tkəttəb], [tkərkəb], [tsurət] and [tsala] which are derived by the
prefixation of the medio-passive morpheme [t-].

224
of the trisegmental verb stems such as 34i.a and 34i.b which conform to an iambic foot of the
type LH, where the light syllable is minor. In 34i.a the heavy syllable is obtained by u-epenthesis;
in 34i.b the heavy syllable is part of the base. The rest of trisegmental verb stems, i.e. 34i.e and
34i.f conform to an iambic LL foot where one of the two syllables is minor.
At this stage, it is reasonable to ask why u-epenthesis applies in 34i.a but not in the rest of
the items. In 34i.b, this epenthesis does not apply because the final syllable is already heavy and
as such does not need to be augmented. In 34a. u-epenthesis applies to avoid output forms such as
*[məktəb], where the verb stem is still iambic but consists of a minor syllable. This points out to
the fact that the language distinguishes between bad iambs such as C.CəC that are improved by
u-epenthesis and good iambs which consist of two light major syllables. If this is the case, one
may wonder why verb stems on the pattern C.CV, and consequently all vowel-final verbs, do not
make recourse to u-epenthesis to achieve a good iamb. Why is it the case that u-epenthesis is
blocked in certain trisegmental verbs such as 34i.d and 34i.e and all quadrisegmental verbs
ending in consonant (34ii.a and 34ii.b)?
One can argue that u-epenthesis is blocked from applying to vowel-final verbs because of
the various constraint violations that would be incurred if the vowel of these verbs were
juxtaposed with /u/ (see section 4.3.1 below for these constraints). In the trisegmental verbs in
34i.d and 34i.e and quadrisegmental verbs ending in a consonant, there is nothing that would
block this epenthesis from applying, thus leading to the most harmonic iambic LH foot type and
resulting in forms such as *[mhətmum], *[məxtaruʕ], *[mkərkub] and *[msurut], which do not
correspond to the optimal output forms. However, if we assume that iambic requirement is a
constraint that holds on the derived PP word rather than on its verb stem, the non-epenthesis of
/u/ in quadrisegmental verbs and certain trisegmental verbs could be accounted for in a
straightforward manner. Such is the hypothesis that will be pursued below in the alternative
analysis of the prosodic shape of the PP.

4.3.1 Augmented Passive Participle Forms

This subsection will consider representative items from all the possible trisegmental verb
bases of the language: sound verbs, medial weak and final weak verbs (i.e. verbs that contain
vowels and/or glides), assimilated verbs, and geminated verbs.

225
 To start with, the category of non-derived trisegmental verbs exhibits a special behavior
which resides in the fact that the PP form shows a prefinal vowel [u]. The passive marker, being
the prefix [m-], is attached to the base which is the verb root here. Given the fact that a foot of the
type LH is more harmonic than LL, it follows that LH must dominate LL. The two constraints on
foot type must dominate DEP-u which in turn dominates
DEP-ə. Thus an input form like /m-ktb/ would be obtained as in 35 below:

-35-

/m-ktb/ LH LL DEP-u DEP-ə

)a. (mək.tub) * * *
b. mək.(tu.bu) *! ** *
c. (mək.təb) *! **

The analysis above could well be extended to passives obtained from other non-derived
trisegmental verbs, namely glide-medial verbs, that is verbs whose second segment is the vocoid
/i/ or /u/. Thus a word such as [məbyuʕ] is derived as in 36. Since all augmented forms must
conform to an LH iamb, the constraint LL won’t play any role in deciding about the optimal
candidate and as such it won’t be included in tableaux unless it bears on the argument.

-36-

/m-biʕ/ LH DEP-u DEP-ə

a. (məb.yuʕ) * *
b. m.(bu.yəʕ) *! * *
c. (məb.yəʕ) *!
d. məb.(yu.ʕu) *!
0e. (m.biʕ)

The constraints in 36 wrongly predict that the optimal candidate is *[mbiʕ] which conforms to an
iambic foot of the type LH, where the light syllable is a minor syllable attached to a consonant.

226
Given the assumption made about syllable structure in chapter 2, such a word would have the
representation in 37 below:

-37-

Ft

σ σ

µ µ µ

m b i ʕ

The actual optimal candidate [məbyuʕ] has the structure in 38:

-38-

Ft

σ σ

µ µ µ

m əb y u ʕ

To exclude a candidate such as [mbiʕ], there should be a way of distinguishing between the iamb
in 37 and the iamb in 38. It has been shown in chapter 2 that CMA distinguishes between a major
syllable, whose nucleus is a full vowel or a schwa and a minor syllable which consists solely of a
consonant. In the same spirit, the language should also be allowed to carry out this distinction to
the foot level. Thus we should distinguish between a true LH iamb of the type in 38 and a minor
LH iamb of the type in 37. Also a distinction, which is irrelevant for the PP cases here, should be
made between an iamb of the type LL and a minor iamb of the same type but where one of the
two syllables is associated with a minor syllable.
To get the optimal output, we need to include a constraint of the type *Min-LH which
must dominate LH (the true iamb) to rule out a candidate such as [mbiʕ]:

227
-39-
/m-biʕ/ *Min-LH LH DEP-u DEP-ə

)a. (məb.yuʕ) * *
b. (m.biʕ) *!
c. m.(biʕ) *!

The constraint *Min-LH will be included below only when it is relevant. We assume that any PP
form that violates this constraint is excluded. (See chapter 6 for the relevance of the constraint
*Min-LH in the analysis of the diminutive). Note here that the base vowel loses its moraic status
and is realized as a glide to serve as an onset to the epenthesized [u]. Assuming that high vowels
are [-consonantal] and that glides are [+consonantal], the realization of the high vowel in 39 as a
glide causes violation of the constraint IDENT-IO [cons], given in chapter four, and repeated in
40 below:

-40-
IDENT-IO [cons]
Featural specification for [cons] must be preserved in the input/output mapping.

In order to derive the optimal form, ONSET must be allowed to dominate IDENT-IO [cons] to
avoid output forms such as *[mbiuʕ] or *[mbiwəʕ].
The other type of forms that need additional constraints are verb bases whose second
segment is the vocoid /u/. Assuming the prefinal epenthesis of /u/ gives rise to two u’s, the
second of which is the root segment. Thus, from an input form such as /m-ʃuf/, Gen supplies the
competitive candidates listed in the following tableau:

-41-

/m-ʃuf/ *Min-LH LH DEP-u DEP-ə

0a. (məʃ.wuf) * *
b.m.(ʃəw.wəf) *! * **
c. m.(ʃuf) *!
d. (m.ʃuf) *!

228
Both candidates 41c and 41d are excluded because they do not conform to the most harmonic
iamb. Candidates 41a and 41b have resorted to u-epenthesis to create that final heavy syllable
required for the iamb. However, this epenthesis results in a sequence of two vowels which need
to be syllabified as one syllable. In 41b, both vowels lose their moraic status and get realized as
glides, a fact that incurs a double violation of DEP-IO and a fatal violation of LH. In 41a, the first
vowel is realized as non-moraic and thus surfaces as a glide that serves as an onset to the
epenthetic [u].
Although candidate 41a conforms to an iambic foot of the type LH, it is not optimal not
because the constraint hierarchy established above is wrong but because the form violates a
constraint which prohibits two contiguous rounded segments. This markedness constraint,
dubbed *ROUND ROUND after Bensoukas (1999), is formulated as follows:

-42-
*ROUND ROUND (henceforth *RdRd)
Sequences of round segments are prohibited.

This constraint is observed not just in the PP but also in other morphological categories as the
following examples show:

-43-

a. Base N Unattested Nisba Nisba Nisba Gloss

?aməzzru *məzzruwi məzzriwi from Amezrou

SəfRu *SəfRuwi SəfRiwi from Sefrou
bzu *bzuwi bziwi from Bzou

b. Sg Unattested Pl. Pl. Gloss

raS *ruuS/*rwuS ryuS head

DaR *DuuR/*Dwur DyuR house

In 43a the addition of the nisba suffix gives rise to a hiatus that the language resolves by inserting
a glide which shares the features of the preceding vowel. Since the constraint *RdRd bans
sequences of identical rounded segments, the final vowel of the stem dissimilates to [i]. In 43b,
the plural morpheme [u] is juxtaposed with another [u] which we assume to be the lexical default
segment of the language needed for a constraint requiring the plural (and other morphological

229
categories such as the diminutive) to start with a sequence of two consonants. The realization of
the first [u] as non-moraic results in forms that violate *RdRd. This state of affairs is remedied by
dissimilating the first rounded segment, thus getting the correct output.
Now, let us return to the passive forms of verbs with a medial vocoid to see how items
like [məʃyuf] could be derived from the input /m-ʃuf/. Such a form necessitates the constraints on
foot types developed above along with the markedness constraint *RdRd to account for the
dissimilation of the root /u/ into the glide [y]. The dissimilation process causes violation of the
constraint IDENT-IO [round] which penalizes an output form that alters the input specification
for the feature [round]. Therefore, in order to get the optimal output, *RdRd must be allowed to
dominate IDENT-IO [round] for dissimilation to apply:

-44-

/m-ʃuf/ *RdRd IDENT-IO [round]

)a. məʃ.yuf *
b. məʃ.wuf *!
c. m.ʃu.wəf *!
d. m.ʃu.wuf *!

Although the optimal candidate in 43a violates IDENT-IO [round] by virtue of the fact that
the medial segment of the verb base [u] is realized as the glide [y], it is the winner because it
conforms to the most preferred iamb. The other candidates satisfy IDENT-IO [round] at the
expense of a higher-ranked constraint, namely *RdRd. Apart from the constraint *RdRd, the third
candidate is suboptimal because it violates LH, a higher-ranked constraint. Note further that all
the candidates in 44 violate the constraint IDENT-IO [cons] by virtue of the fact that the input
high vowel is realized a glide to serve as an onset to the epenthetic [u]. Next, we consider the PP
of assimilated verbs, that is verbs whose first segment is essentially the glide [w] 7. The
assumption underlying this work is that the initial segment in such verbs is a glide underlyingly.
Thus a form such as [məwlud] from the input /m-wld/ could be obtained using the same
constraints developed for the analysis of the PP forms derived from sound bases.

230
-45-

/m-wld/ *Min-LH LH DEP-u DEP-ə

)a. (məw.lud) * *
b. (məw.ləd) *! **
c. mə.(wə.lud) * **!
d. m.(w.lud) *!

Candidate 45b is excluded because it violates LH. As to candidate 45c, it is ruled out for having
made recourse to many ə-epenthesis instances, thus violating DEP-ə. Finally, 45d is ruled out
because of the constraint *Min-LH. It can also be ruled out on the ground that it incurs two
violation marks of *Min-σ by allowing the segments [m] and [w] to attach to a minor syllable.
Another candidate that deserves special consideration is [mulud]. This form corresponds
to an LH iamb and spares DEP-ə and leads one to wonder why it cannot be the optimal candidate
rather than [məwlud]. The form [mulud] is attested in other varieties of MA, especially the
southern varieties. In the variety under study, this form may frequently be heard in rapid speech,
but the form used in normal speech is the one that preserves the initial underlying glide and
epenthesizes a schwa between this glide and the

PP prefix. The relevant constraint penalizing featural change of the underlying glide is IDENT-
IO [cons]. To derive [məwlud], the constraint IDENT-IO [cons] will have to be dominate
constraint DEP-ə. The tableau below shows how [məwlud] is chosen over [mulud].

7
The only verb whose first segment is [y] is [ybəs]. All other glide initial verbs in MA start with the glide [w] in
most of the varieties.

231
-46-

/m-wld/ IDENT-IO [cons] DEP-ə

a. məwlud *
b. mulud *!

In varieties where [mulud] is optimal, DEP-ə must outrank IDENT-IO [cons]. (For a more
detailed analysis of variation in MA, the reader is referred to Rguibi, forthcoming).
The final case that shows a prefinal [u] is that of geminated verbs such as [ʃədd] “hold”
and [həzz] “lift” which get their PP form by epenthesizing [u] between the two parts of the
geminate to get [məʃdud] and [məhzuz], respectively. Given that the verb base is the root which
is bisegmental by virtue of the fact that geminates are represented by a single melodic element
(McCarthy’s 1986 OCP), the constraints developed so far in this chapter are incapable of
generating the correct output. The constraints we have would allow candidates such as the ones
given in 47 below:

-47-

/ m-ʃd/ LH DEP-u DEP-ə

0a. (məʃ.du) * * *
b. mə.(ʃu.du) * **! *

This constraint tableau wrongly predicts that the optimal candidate is 46a. Candidate 47b is
excluded because it makes recourse to u-epenthesis twice and allows the schwa to occur in open
syllables, something the language does not allow.
Confronted with such a situation like the one in 47, two solutions emerge in order to
satisfy the optimal foot. First, epenthesize the segment /u/ and then lengthen it to have a heavy
syllable that conforms to the desired iamb. The result is the form *[məʃduu] which is suboptimal
because it contains long vowels that the language bans. The constraint ruling out long vowels is
stated as follows:

232
-48-

NO-LONG-V (McCarthy and Prince 1993a, Rosenthal 1994)

Long vowels are prohibited.

The second solution to achieve an LH iamb is to close the syllable whose nucleus is the
epenthesized /u/ by the labiovelar glide [w], thus getting the output [məʃduw]. This form is ruled
out because it incurs a fatal violation of the constraint DEP-C:

-49-

DEP-C
Every consonant in the output must be in the input.

We assume that this constraint should be allowed to dominate LH so that no epenthesis could
take place to satisfy the iamb requirement.
The third solution which will be adopted in the analysis of verbs like [ʃədd] and [həzz] is
the one suggested in chapter 2 and whereby geminates have two root nodes (Selkirk 1990, 1991).
Adopting the Two-Root Theory of geminates, verbs such as [ʃədd] would have the structure in 50
below:

-50-

Ft

σ σ

µ µ
|
RC RC

ʃ ə d

The structure in 50 is written as /ʃdd/ to encode the notion of the Two-Root Theory of geminates.
The addition of another root node and the spreading of the final consonant results in the violation
of the anti-structure constraint *STRUCTURE, but since this constraint is violated in all
candidates, it should be ranked low in the constraint hierarchy.

233
 In the constraint tableau below we expose the three possibilities suggested for the PP of
geminated verbs and show how the form [məʃdud] is the only optimal candidate of the three.

-51-

/m-ʃdd/ DEP-C NO-LONG-V LH DEP-u

)a. (məʃ.dud) *
b. (məʃ.duu) *! **
c. (məʃ.duw) *! **

What is not shown in the tableau above is that candidates 51b and 51c fail to realize the two-root
nodes associated with the geminate, thus incurring another fatal violation of the constraint MAX-
RC, requiring the preservation of the root consonants of the input (see chapter two). Notice also
that the optimal candidate splits up the geminate, a fact which points out that LH has to dominate
the constraint NO-SPLITTING, which in turn has to dominate DEP-u as shown in the tableau
below:

-52-

LH NO-SPLITTING DEP-u
)a. (məʃ.dud) * *
b. m.(ʃəd.d) *!

In chapter two, it has been shown that NO-SPLITTING is violated in quadrisegmental verbs to
secure the higher-ranked constraint on verb root bimoraicity. Similarly, items such as [məʃ.dud]
show that the NO-SPLITTING constraint can be violated if the output obtained conforms to an
iambic foot of the type LH.
To sum up, it has been shown that there are two different ways through which CMA PP
forms achieve an iambic foot of the type LH: first by epenthesizing /u/ in sound verbs,
assimilated verbs and verbs whose medial segment is a vocoid; second, by epenthesizing /u/
between the two parts of a final geminate in geminated verbs, something that is possible only
under the Two-Root Theory of geminates.

234
 Next, we consider the class of passives derived from verbs whose final segment is a
vocoid. The output form does not show the epenthesis of a prefinal /u/. Does that mean that these
bases (along with bases such as /kwi/, i.e. bases whose second and/or final segments are vocoids)
behave like derived trisegmental and quadrisegmental verbs which do not show any
augmentation? Or is it some other higher-ranked constraint that prevents the prefinal /u/ from
surfacing?
At considering a passive form such as [məʃri] we notice that it does not conform to the
most harmonic LH iamb. It has already been shown that making recourse to final-vowel
lengthening would not work as the language bans any structure with long vowels. Since the verb
base ends up in /i/, the only way to make the final syllable heavy is by closing that syllable with a
glide sharing the same features of the base final segment, thus getting the word *[məʃriy], or by
epenthesizing the prefinal /u/ and realizing the final base segment as a glide, thus getting the
output *[məʃruy]. *[məʃriy] could be excluded because it violates DEP-C, a constraint that
prohibits consonant epenthesis. As to *[məʃruy], it could be argued that it violates IDENT-IO
[cons] since the input vowel /i/ is realized as the consonantal [y].
To see how a form such as [məʃri] is obtained, consider the constraint tableau below
where DEP-C, NO-LONG-V and IDENT-IO [cons] dominate LH. Here the optimal foot is of the
type LL.

-52-

/m-ʃri/ DEP-C NO-LONG-V IDENT-IO LH LL DEP-u

[cons]
)a. (məʃ.ri) *
b. (məʃ.riy) *! *
c. (məʃ.ruy) *! * *
d. (məʃ.rəy) *! *

Although the candidates in 52b and 52c end up in an LH iamb, they are excluded because the first
violates the constraint in DEP-C while the second violates IDENT-IO [cons]. Candidate 52c is
also excluded because it fails to preserve the identity of the input vocoid. The optimal candidate

235
does not conform to an ideal iamb and yet it is the winner simply because it satisfies all the three
higher-ranked constraints.
Bases whose second and third segments are vocoids can be treated much in the same way.
The second segment of these verbs is a glide underlyingly while the third is a vowel. The tableau
below lists some of the candidates obtained from the input /m-ʃwi/:

-53-

/m-ʃwi/ DEP-C NO-LONG-V IDENT-IO LH LL DEP-u

[cons]
)a. (məʃ.wi) *
b. məʃ.(wiy) *! *
c. məʃ.(wuy) *! * *
d. (məʃ.wəy) *! *

It has become clear by now that any candidate resorting to consonantal epenthesis or to a change
in the identity of the base vocoid to achieve the LH iamb will be excluded.
To conclude this subsection, it has been shown that the PP morpheme in CMA is [m-] and
the prefinal vowel /u/ is the lexical default segment that is epenthesized to achieve the ideal LH
iambic foot. In contexts where this epenthesis would violate other higher-ranked constraints such
as DEP-C, NO-LONG-V or IDENT-IO [cons], for example, the foot that emerges as the optimal
one is of the type LL.

4.3.2 Unaugmented Passive Participle Forms

There is a class of passive forms derived from verb bases that does not show
augmentation by the epenthesis of /u/. This class comprises the passive form of medial geminated
verb bases such as [mkəttəb] and the passive of all quadrisegmental verbs such as [mTəRʒəm]
“translated”, [msali] “finished”, [mSifəT] “sent” and [mSəqSi] “asking”

236
 4.3.2.1 Derived Trisegmental Verb Bases

These forms along with the other quadrisegmental verb bases constitute a testing ground for the
constraints we have advanced so far. As the passive forms do not show augmentation by u-
epenthesis, they all violate LH because they end up in a vowel or a schwa and consonant, which
are both dominated by a single mora.
Given a PP form such as [mkəttəb] from the input /m-ktb/, one should wonder why
[məktub] is not the optimal output. Put differently, what is the base of derivation for [mkəttəb]?
Is it /ktb/ or [kəttəb]? Assuming that it is /ktb/ would allow us to derive the passive via u-
epenthesis and therefore get a form that is similar to those in 1a-d above. However the target this
time is different; it is the PP of medial geminated verb bases, that is to say forms that have
already undergone some derivation. In OT terms this means an output form.
In chapter four above we have shown that forms like [kəttəb] are forms which involve
what Imouzaz (forthcoming) calls partial reduplication. We have argued that in order to better
understand these forms, an output-output relation of the type Base/Reduplicant should be
invoked. Such a relation relates two output forms produced simultaneously: the base and the
reduplicant. (See chapter 4, for details about the base form).
Assuming that the base can be an output form would allow us to account for PP forms
obtained not only from derived trisegmental bases but also quadrisegmental verb bases. To derive
the correct output, We need a morphologically-grounded output-output constraint (Basri et al
1998 and Selkirk 1999) preserving the initial edge of the base foot in the derived form. This
constraint is formulated in the correspondence model of McCarthy and Prince (1995) as in 54:

-54-

OOStem ANCHOR (Ft, Ft, Initial)

A foot-initial segment in the affiliate output form must correspond to a foot-initial
segment in the base output form.

We also need an output-output constraint requiring the preservation of weight identity in the base
syllables of the stem. This constraint is given in 55:

237
-55-

OOStem IDENT-σ
Light/heavy syllables in the affiliate output form must correspond to light/heavy
syllables in the base output.

The constraint in 55 makes sure that the left edge of the base foot remains intact while the
constraint in 55 makes sure that the stem syllables of the base and those of the output correspond
to each other in terms of weight. Both constraints must dominate the constraints on foot type. In
the constraint tableau in 56, we show how the output [mkəttəb] is chosen as the optimal
candidate:

-56-

/m-ktb/ OOStem OOStem LH LL DEP-u DEP-ə

base:(kət.təb) ANCHOR
IDENT-σ
(Ft, Ft, Initial)
a. mə.(kət.təb) * ***!
b. m.(kət.tub) *! * * *
c. (mək.tub) *! * * *
)d. m.(kət.təb) * **

The base [kəttəb] consists of one foot and since the constraint OOStem ANCHOR
(Ft, Ft, Initial) is undominated, it follows that the optimal candidate has to be identical to the
base. Such is the case with candidate 56d where the foot initial segment of the base (that is the
segment [k]) corresponds to the foot initial segment of the output [mkəttəb]. Notice that the
passive marker in the output candidate lies beyond the domain of the foot; it is dominated by a
minor syllable which is directly associated with the prosodic word. Candidate 56a fails although
it perfectly matches the foot initial segment of the base. The reason is that it incurs too many
violations of DEP-ə. Finally candidate 56b is ruled out because it satisfies LH at the expense of
higher-ranked OOStem IDENT-σ and 55c is also excluded on the ground that it violates both
OOStem IDENT-σ and OOStem ANCHOR (Ft, Ft, Initial).

238
 Although the PP forms of derived trisegmental verbs do not conform to the prototypic
iamb in CMA, they are nonetheless optimal because they their foot-initial segments correspond to
the foot-initial segments of the base, and this explains why OOStem ANCHOR (Ft, Ft, Initial)
must outrank the constraint LH on foot type.
Verb bases whose final segment is a geminate are derived much in the same way.
Consider the candidates that could be derived from the input
/m-nZZ/. Each candidate is evaluated on how good it matches the base [nəZZəZ]:

-57-

/m-nZZ/ OOStem OOStem LH LL DEP-ə

Base: (nəZ.ZəZ) IDENT-σ ANCHOR
(Ft, Ft, Initial)
)a. m.(nəZ.ZəZ) * **
b. (mən.ZəZ) *! * **
c. (mən.ZuZ) *(!) *(!) * *

Candidate 57a is the winner because the foot-initial segment corresponds to that of the base.
Candidates 57b and 57c are excluded because they don’t perfectly match the foot-initial segment
of the base, which is the stem [nəZZəZ]. In 57b, for example, the initial segment of foot (i.e. the
segment [m] of the passive) does not correspond to the initial segment of the base (i.e. the
segment [n]).
There is yet another output form which might compete with the optimal candidate. The
output *[mnəZZ], footed as m(nəZ.Z), incurs a single violation of both LH and DEP-ə and
therefore should win over the optimal candidate. To exclude this possibility, it should be noted
that there is a domination relationship that holds among iambic feet. Thus for example, LH is
better than LL. Given that CMA contain both major and minor syllables, it follows that feet have
to be divided into major and minor. Thus we have already shown that *Min-LH must dominate
LH to distinguish *[mbiʕ] and [məbyuʕ]. Similarly, with a constraint such as *Min-LL, being
outranked by LH, one could argue that the form *[mnəZZ], consisting of a foot of the type LL
where the right-hand L is a minor syllable associated with the second part of the geminate, is
ruled out exactly because it violates * Min-LL.

239
 The PP of verb bases derived by the infixation of [t] such as [məhtəmm] and [məħtaʒ] (cf.
roots /hmm/ and /ħaʒ/) or by [t] and [a] infixation such as [məxtarəʕ] (cf. root /xrʕ/) is obtained
by making recourse to the base, i.e. an output form. The tableau below presents different
candidates for the input /m-htmm/:

-58-

/m-htmm/ OOStem OOStem LH LL DEP-u

base: h.(təm.m) IDENT-σ ANCHOR
(Ft, Ft, Initial)
)a. məh.(təm.m) *
b. mə.(hət.mum) *(!) *(!) * *
c. (məh.tum)m *(!) *(!) * *
d. m.(hət.məm) *! *

In 58a, the foot initial segment of the derived form (i.e. [t]; the segment [h] being associated with
a minor syllable attached directly to the PWd) corresponds to the foot initial segment in the base.
In 58b and 58c both candidates are excluded because they have made recourse to u-epenthesis
and therefore fail to match the base syllable identity. Finally, 58d is ruled out because of a
mismatch in foot between the initial segments of the base and the derived output.
A final case which deserves special attention is that of PP forms whose verb bases contain
the prefinal vowel /a/. These forms end up in a heavy syllable, something that obviates the need
for u-epenthesis. However, they present a testing ground for the constraint OOStem ANCHOR
(Ft, Ft, Initial). Take for example the form [mərtaħ]. This form consists of a single iambic LH
foot whose initial segment is the PP prefix [m-] but the initial segment of the foot containing the
verb stem serving as a base for derivation is the segment [r] (cf. the foot (r.taħ)). Therefore the
optimal candidate [mərtaħ] violates the constraint OOStem ANCHOR (Ft, Ft, Initial) by virtue of
the fact that the initial segment of the affiliate output form does not correspond to the initial
segment of the base output. Notice here that the foot of the verb stem corresponds to a minor LH
foot which has been argued to be dominated by the true LH. Thus in order for the candidate
[mərtaħ] to be optimal, *Min-LH has to outrank the constraint OOStem ANCHOR (Ft, Ft, Initial)
as the tableau below shows:

240
-59-

/m-rtaħ/ *Min-LH OOStem ANCHOR LH DEP-IO

Base: (r.taħ) (Ft, Ft, Initial)
)a. (mər.taħ) * *
b. mə.(r.taħ) *! *
c. m.(r.taħ) *!

In the optimal candidate, the foot initial segment does not correspond to the foot initial segment
in the simple base output form. But the ranking argued for in 59 shows that avoiding minor LH
iambs is better than achieving total correspondence between the foot initial segments in the
derived and simple output forms.
Further support of the analysis undertaken for the PP of derived trisegmental verb bases
comes from participle forms derived from quadrisegmental verbs to which we turn in the
following subsection.

4.3.2.2 Quadrisegmental Verb Bases

The forms considered in this subsection include passive forms of sound verbs such as
[TəRʒəm] “translate” and [bərgəg] “he spied (on someone)”, verbs whose second segment is i/u
such as [SifəT] “he sent” and [surət] “he locked”, verbs whose fourth segment is [i] such as
[SəqSi] “(you) ask” and finally verbs whose second segment is [a] and fourth is [i].
A common feature among all these verb bases is that they are all disyllabic. The base foot
is of the type LL which means that the constraint LH is violated in all the passive forms of these
verb bases. Given this fact, the only decisive constraints are higher-ranked constraints on output-
output-correspondence between the foot-initial segment of the base and the foot-initial segment
of the derived output, and also the constraint on the identity of the base syllables and those of the
derived output.
In tableau 60 below, we evaluate the different candidates for the input /m-TRʒm/.

241
-60-

/m-TRʒm/ OOStem OOStem LH LL DEP-u

Base: (TəR.ʒəm) IDENT- ANCHOR
σ (Ft, Ft, Initial)
a. mə.(TəR.ʒum) *! * *
b. (məT.Rəʒ)m *! *
c. mu.(TəR.ʒəm) * *!
)d. m.(TəR.ʒəm) *

The winner is candidate 60d which consists of a foot whose initial segment perfectly matches that
of the base. Candidate 60c corresponds perfectly to the base except that it incurs a fatal violation
of DEP-u. Candidate 60b fails to satisfy OOStem ANCHOR (Ft, Ft, Initial). Candidate 60a is
excluded because it resorts to u-epenthesis and therefore incurs a violation of OOStem IDENT-σ.
It should be noted here that an output such as (mTəR.ʒəm) where the consonant [m] belongs to
the first syllable is ruled out for violating either of the two constraints: first, it violates the
constraint OOStem ANCHOR (Ft, Ft, Initial) because the initial segment of the foot in the output
form (that is [m]) does not correspond to the initial segment of the base foot; second it violates
the constraint *COMPLEX which has been shown to be dominated.
Not surprisingly enough, the constraints developed above could also account for final
geminated verbs such as [bərgəg] and [fərtət]. Notice that with these verbs the geminates are split
up under pressure from the constraint on the foot type, a fact which has led us to rank NO-
SPLITTING below LH and LL. Consider an input such as /m-frtt/ for illustration.

-61-

/m-frtt/ OOStem OOStem LH NO- DEP-ə

Base: (fər.tət) IDENT-σ ANCHOR SPLITTING
(Ft, Ft, Initial)
)a. m.(fər.tət) * * **
b. mə.(fər.tət) * * ***!
c. (məf.rət).t *! * **
d. (məf.rətt) *! **

242
Although candidate 61b satisfies the two higher-ranked constraints, it is ruled out because it
incurs too many violations of DEP-ə. Candidates 59c and 59d are both excluded because the foot
initial segment in each does not correspond to the foot initial segment of the base. 61d could also
be excluded on the ground that it violates *COMPLEX by virtue of the fact that the final
geminate belong to the same syllable.
Once again, showing complete correspondence of the foot initial segments of the derived
output and the base as well as preserving the nature of syllables constituting this foot is the only
way of deriving the correct output. This is established by ranking OOStem IDENT-σ and OOStem
ANCHOR (Ft, Ft, Initial) on the top of the other constraints, namely the ones on the foot types of
the output.
Verb bases of the type [sali] and [SəqSi] do not need any additional constraints. Their PP
forms can be derived much in the same manner as shown in the tableau below where we list some
of the candidates obtained from the input /m-SqSi/.

-62-

/m-SqSi/ OOStem OOStem LH LL DEP-ə

Base:(Səq.Si) IDENT-σ ANCHOR
(Ft, Ft, Initial)
)a. m.(Səq.Si) * *
b. mə.(Səq.Si) * **!
c. məS.(qəS.wi) *! * **

Candidate 62b is suboptimal because it incurs one violation mark of DEP-ə than the optimal
candidate. Candidate 62c is excluded because it has resorted to the epenthesis of a prefinal /u/
that is realized as the glide [w] to serve as an onset to the base final vocoid. This epenthesis leads
to a gratuitous violation of the constraint DEP-u and OOStem ANCHOR (Ft, Ft, Initial).
Verbs whose second segment is /i/ or /u/ such as [SifəT] and [surət] are somehow
different from the previous quadrisegmental bases because they end up in two consonants, a
potential environment for epenthesizing the segment /u/ and therefore establishing an iamb of the
type LH. Consider the competing candidates given in tableau 63 for the input /m-SifT/.

243
-63-

/m-SifT/ OOStem OOStem ANCHOR LH LL DEP-ə

base: (Si.fəT) IDENT-σ (Ft, Ft, Initial)
)a. m.(Si.fəT) * *
b. mə.(Si.fəT) * **!
c. (məS.yəf)T *! * **
d. m.(Si.fuT) *! *

Candidate 63b is suboptimal for a reason that has now become obvious. Although the constraint
DEP-ə is ranked low in the ranking hierarchy, it is still decisive. The candidate in 63c is excluded
because it violates OOStem ANCHOR (Ft, Ft, Initial). The constraint IDENT-IO [cons] is also
violated in 63c because the base [i] loses its moraic status and gets realized as the corresponding
glide in the output. Finally, candidate 63d is excluded because of the weight mismatch between
the final syllable of the base and that of the derived output.
Now, consider a quadrisegmental verb base whose second segment is the vocoid /u/. The
following tableau lists candidates derived from the input /m-surt/:

-64-

/m-surt/ OOStem OOStem ANCHOR LH LL DEP-ə

Base: (su.rət) IDENT- (Ft, Ft, Initial)
σ
)a. m.(su.rət) * *
b. mə.(su.rət) * **!
c. m.su.rut) *! *

The constraints in 64 correctly predict that the optimal candidate is the one in 64a which
preserves weight identity of the base syllables as well as the position of the initial segment of the
base foot.
The final case of PP we will consider is that of forms derived from verb bases which are
themselves derived by the prefixation of the medio-passive morpheme [t-]. These items should
not pose any problem to the analysis presented in this chapter. The stem consists of three

244
syllables the first of which is minor and is adjoined directly to the PWd. Only the major syllables
in such items are part of the only existing foot. The tableau in 65 below lists some competing
candidates from the input /m-t-qulb/:

-65-

/m-t-qulb/ OOStem OOStem IDENT- LH DEP-ə

Base: t.(qu.ləb) IDENT-σ ANCHOR IO [cons]
(Ft, Ft, Initial)
)a. mət.(qu.ləb) * **
b. (mət.qu).ləb *! * **
c. mət.(qəw.ləb) *! * ***
d. mət.(qu.lub) *! *

Clearly candidate 65b is excluded because it violates OOStem ANCHOR (Ft, Ft, Initial): the initial
segment of the foot in the derived output form does not correspond to the initial segment in the
base foot. Candidates 65c and 65d are both suboptimal because the first violates IDENT-IO
[cons] by virtue of a change in the identity of the base input vocoid, while the second violates
OOStem IDENT-σ by virtue of epenthesizing /u/ to achieve the LH iamb.
Once again, grouping both derived trisegmental and quadrisegmental verb bases under the
same rubric proves to be well-founded since their PP forms do not show augmentation by u-
epenthesis. They show that in case there is a conflict between a prosodic constraint and a
faithfulness constraint, it is the latter which is worth satisfying.
The constraint hierarchy needed to account for the PP in CMA is given in 66 below:

245
-66-

OOStem IDENT-σ *Min-LH ONSET NO-LONG-V DEP-C

OOStem ANCHOR IDENT-IO

(Ft, Ft, Initial) [cons]

LH

LL

NO-SPLITTING

DEP-u

DEP-ə

The constraints *RdRd and IDENT-IO [round] are not included in 66 because they don’t interact
with the rest of the constraints.

5. CONCLUSION
This chapter has offered an OT analysis of CMA PP forms. The idea maintained
throughout is that the PP morpheme is the prefix [m-] and that the PP forms fall into two classes
based on their behavior vis-à-vis /u/ epenthesis. We have argued that the segment /u/, found
prefinally in some forms is the lexical default segment of the language and that its epenthesis is
dictated by prosodic constraints, namely the need for the output to conform to an iambic foot of
the type LH. Forms that make recourse to u-epenthesis include the class of non-derived
trisegmental verb bases with the exception of verb bases whose second and/or final segments are
vocoids. For these forms, we have shown that if u-epenthesis were to apply, we would end up

246
with forms that violate the constraint DEP-C, banning the epenthesis of a consonant or the
constraint NO-LONG-V, prohibiting long vowels or else IDENT-IO [cons], requiring featural
identity between the input and the output.
Other classes that do not show u-epenthesis include PP forms derived from a class of
trisegmental verb bases and all the quadrisegmental verb bases. The foot structure of such forms
still conforms to an iamb but of the type LL, a clear violation of the constraint demanding that the
most harmonic PP foot be of the type LH. These forms constitute a strong evidence for the
account proposed in terms of output-output correspondence. In particular, we have shown that if
we are to derive the optimal output, the constraint OOStem ANCHOR (Ft, Ft, Initial), demanding
left-anchoring of the initial segment of the foot in the derived output and the initial segment of
the base foot, and OOStem IDENT-σ, demanding the conservation of weight identity, have to
outrank the constraints on the optimal foot types.

247
 Chapter Six

Augmentation in the Prosodic Morphology of the Diminutive

1. INTRODUCTION
In chapter 5, we have shown the need to distinguish between two types of PP’s: those that
appeal to the epenthesis of /u/ to achieve an iambic foot and those that do not. We have shown
that the non-epenthesis or epenthesis of /u/ follows from the interaction of prosodic constraints
requiring that the output be an iamb and faithfulness constraints requiring identity of the input
and the output. In this chapter we present further evidence for foot ianbicity based on another
aspect of the prosodic morphology of CMA which is the diminutive.
Like the PP, the output of the diminutive is also governed by a prosodic constraint
demanding that the output be an iambic foot. Also, the output of the PP and the diminutive must
be an iambic foot of the type LL or in the ideal case LH but never a minor LH iamb, i.e. an iamb
whose light syllable consists solely of a consonant. Both classes resort to augmentation to achieve
this iambic requirement. However, they differ in the way this augmentation is achieved as well as
in the result it leads to. The PP appeals to the language default segment /u/ to form an iambic foot
of the type LH, whereas the diminutive may add a whole default syllable or the feminine suffix [-
a], depending on whether the base is masculine or feminine and this in order to achieve an iambic
foot of the type LL. It will be shown that achieving this kind of iambicity is the result of
satisfying a constraint referred to as INITIAL-CC which requires that the PWd start with a cluster
of two consonants, thus forcing the diminutive morpheme [-i] to be placed after this cluster and
resulting in the required prosodic shape.
The chapter is divided into 5 major sections. Section 2 lists the possible diminutive forms
of the language. Section 3 shows how a derivational approach such as circumscriptive
morphology fails to adequately account for diminutive formation. Section 4 argues for the
necessity of incorporating in the analysis of the diminutive a constraint referred to as INITIAL-
CC which has the effect of forcing the diminutive morpheme to be placed after two consonants.
Support for this constraint comes from cases involving labialization and gemination of labial
consonants. Here, it will be pointed out that the gemination of labial consonants in the diminutive
follows from the constraint INITIAL-CC and the interaction of markedness constraints on
labialized consonants. Sections 5, 6 and 7 offer an OT account of augmented and unaugmented
diminutive forms. In particular, it will be argued that a fair analysis of the diminutive is derived
from the interaction of prosodic constraints on foot types with markedness and faithfulness
constraints as well as their respective ranking.

2. THE DATA
The diminutive is formed by the affixation of the morpheme [-i-] after the second segment
of the base. This process is often accompanied by the labialization of the first segment of the base
if it happens to be one of the dorsal consonants [k, g, x, γ, q] or labial consonants [b, f, m]. The
body of relevant data is listed in 1 below:

-1-

Base Diminutive Gloss

a. wəld wliyyəd boy

ʒməl ʒmiyyəl camel
sbəʕ sbiyyəʕ lion
nmər nmiyyər tiger
sdər sdiyyər chest

b. fərx friyyəx bird

bərd briyyəd wind
fərħ friyyəħ feast of rejoicing
bγəl bγiyyəl mule

c. kəlb kwliyyəb dog

ktəf kwtiyyəf shoulder
qərd qwriyyəd monkey
qəlb qwliyyəb heart
qfəz qwfiyyəz cage
gdəm gwdiyyəm heel
xwəbz xwbiyyəz bread
γwrab γwriyyəb crow

249
d. bit bbwiyyət room
bir bbwiyyər well
mus mmwiyyəs knife
buq bbwiyyəq loud sppeaker
BuT BBwiyyəT rubber boots
fwəmm ffwiyyəm mouth

e. far ffwiyyər mouse

kas kwiyyəs A (drinking) glass
ʒib ʒwiyyəb pocket
BaR BBwiyyəR pub
RaS RwiyyəS head
bab bbwiyyəb door
γaR γwiyyər cave

f. ʃəmʃ ʃmiʃa sun

zit zwita oil
bənt bnita girl
Dar dwira house
kəréʃ kwriéʃa belly
ʕin ʕwina eye
wdən wdina ear

g. ʕʃa ʕʃiwa dinner

γda γdiwa lunch
mRa mRiyya / mRiwa woman
ʕSa ʕSiyya / ʕSiwa a stick
bRa bRiyya / bRiwa letter
ʃta ʃtiwa rain

h. bəlγa bliγa (oriental) slippers

dəmʕa dmiʕa a tear
kura kwira ball
gwəRSa gwRiSa a (small, circular) loaf of bread
wəRDa wRiDa flower
xuxa xwixa peach

i. TəbSil TbiSil plate

sərwal sriwil pants
ʃərbil ʃribil women’s (oriental) slippers
dəbliʒ dbiliʒ bracelet
sarut swirit key
ħanut ħwinit shop

250
j. SəBBaT SBiBiT shoes
gwəffa gwfifa basket
gwəlla gwlila earthware jar
kəBBuT kwBiBiT jacket
səllum slilim ladder
bərrad bririd tea pot

k. limuna lwimina orange

banana bbwinina banana
SənDala SniDila sandals
gənDuRa gwniDiRa a Moroccan gown
baliza bbwiliza suitcase
hiDuRa hwiDiRa sheepskin

l. ʒəllaba ʒliliba jellaba (a gown with a hood)

bərraka bririka a slum hut
ʃəkkwara ʃkwikira satchel
nəffaxa nfifixa baloon
nəwwara nwiwira flower
ZəmmaRa ZmimiRa horn

The generalization that could be made about the items in 1 is that disyllabic 1h-j as well as
trisyllabic 1k and 1l bases surface with the same number of syllables in the diminutive. The more
interesting cases are those diminutive items in 1a-g which surface as disyllabic in spite of the fact
that their bases consist of only one major syllable. The data above also show that there are two
ways a base can be augmented: first by the building of what Al Ghadi (1990) refers to as the
default syllable of the language, i.e. a syllable whose nucleus is the schwa and whose coda is the
final consonant of the base as in 1a-e. Second, the base can be augmented by the suffixation of
the feminine morpheme [-a] as is in 1f.
The questions we should ask here are the following: why is it the case that the diminutive
forms in 1a-g are augmented by one syllable while the disyllabic and trisyllabic ones are not?
Why is it the case that only very specific consonants are labialized in the diminutive?
In what follows, we will attempt to provide answers for these two questions within the
theoretical framework adopted. First, we will show how a non-constraint based framework such
as operational circumscriptive theory is incapable of accounting for the diminutive forms in
CMA.

251
3. AGAINST A CIRCUMSCRIPTIVE ACCOUNT OF THE
DIMINUTIVE
McCarthy and Prince (1990a) assume that templates in Prosodic Morphology are defined
in terms of prosodic categories such as the mora, the syllable, the foot and the prosodic word.
They further assume that the domain to which a morphological operation applies is circumscribed
by prosodic as well as morphological criteria. In particular, a base, which corresponds to the
minimal word, is selected as the locus of the morphological operation. In MA, the base has been
found to correspond to a minimal syllable of the type CV or CəC (Al Ghadi 1990, Bennis 1992).
In what follows, we show how the diminutive in CMA could be derived within the theory of
prosodic circumscription. Thus, an item having the shape CəCC, such as [bərd] “wind”, could be
derived as in 9 below:

-2-

a. Input
PWd

Ft

σ σ

µ µ

b ə r d

b. Circumscription of a CəC and final C extrametricality

PWd

Ft

b ə r <d>

252
c. Suffixation of the diminutive to the circumscribed CəC and restoration of <b>

PWd

Ft

σ σ

µ µ µ

b r i d

d. Augmentation to satisfy an LL iambic foot

PWd

Ft

σ σ σ

µ µ µ

b r i y ə d

In this representation, we ignore the gemination of the glide that serves as an onset to the schwa
syllable. We take this problem down in section 6.
Leaving aside the critique that might be leveled at such an approach, we turn now to see
whether the circumscriptive analysis could adequately account for all the diminutive cases. In
fact any analysis based on circumscription of a prosodic unit will fail given an input of the shape
CCəC. Given that the diminutive is suffixed to a minimal syllable, words like [ktəf] “shoulder”
and [Smək] “deaf” would be derived as in 3 below:

-3-

a. Input k.təf S.mək

b. Circumscription of a CəC and Initial C extrametricality
<k> təf <S> mək
c. Suffixation of the diminutive to the circumscribed CəC.
təf-i mək-i

253
d. Restoration of extrametrical consonant
ktəf-i Smək-i
e. Syllabification
kət.fi Səm.ki
f. Augmentation to satisfy an LL iambic foot
“satisfied” “satisfied”
g. Output *[kətfi] *[Səmki]

As shown above the operational circumscriptive analysis does not work and therefore should be
abandoned because it is incapable of giving the correct output.
In an attempt to come up with a unified analysis for the diminutive in MA, Lasri (1989),
who assumes that the diminutive morpheme is placed after the first syllable of the base, maintains
that forms on the pattern CCəC should be treated like the rest of the items of the language. He
assumes that items such as [ktəf], i.e. items on the pattern CCəC, have the template CəCC
underlyingly and as such the diminutive would be placed in the right location. Within the
theoretical framework he adopted, a form like [bγiyyəl] “mule” would be derived as in 4 below:

-4-

a. O R Ap b. O R A R Ap
│ │
N C N N
│ │ │
X X X X X X X X X X
│ │ │ │ │ │
b γ l b γ i l

c. O R A R d. O R A R A R
│ │ │
N N N N N C
│ │
X X X XXX XX X X X X XXX X X
│ │ │ │ │ │
b γ l b γ i y l

For Lasri (1989) representations such as the ones above reflect the steps below. The first step is
to place the diminutive morpheme after the first syllable of the base and the syllabification of the

254
coda of this syllable as an onset of the following syllable. The second step is to add two positions
after the diminutive morpheme. At the final step, resyllabification applies, this time inserting
another X-position which is assumed to be the nucleus of a syllable whose onset is the glide [y],
resulting from the spreading of the diminutive vowel [-i-].
The analysis in 4 could be refuted on both theoretical and empirical grounds. On the
theoretical level, the framework adopted by Lasri appeals to the syllable to make generalizations
about the placement of the diminutive morpheme but at the same time makes recourse to skeleta
that consist of timing units, something that goes against McCarthy and Prince’s (1986) Prosodic
Morphology Hypothesis. Templates are defined in terms of the authentic units of prosody and not
the CV or the timing (X) units. Moreover, Lasri’s analysis and the analysis that appeals to
prosodic circumscription are operational in the sense that they have to scan an input form to look
for a minimal syllable, then place the diminutive morpheme and make the necessary adjustment
to get the output. In OT, this operational analysis is not needed since constraints apply on output
forms in a non-serialist way. On the empirical level, Lasri assumes, without proving it, that the
language has long vowels. In addition, saying that the diminutive is placed after the first syllable
of the base gives the wrong output form as all bases on the pattern CCəC will surface as [CəCCi],
unless recourse is made to some ad hoc mechanism to avoid that, a fact which is conspicuous
from Lasri’s analysis.
In the rest of this chapter, we offer an alternative analysis couched within the OT
framework and show to what extent it is far better than the approaches undertaken in 3 and 4
above. But before doing that, let us consider the issue related to the cluster of consonants that
diminutive forms start with as well as the process of labialization because of their key role in the
understanding of how the diminutive formation works in CMA.

4. ON THE NEED FOR THE CONSTRAINT INITIAL-CC

The assumption underlying the present work is that the diminutive morpheme consists of
the vowel [-i-] and the feature [+round]. The vowel is affixed after the second segment of the
base while the feature [+round] hooks up to a word-initial dorsal consonant or a geminate labial.
The legitimate question one could ask is why the diminutive affix is placed after the second
segment of the base. To answer this question, recall that the regular cases of schwa epenthesis is

255
for the schwa to be placed after the second segment of a CCC sequence in order to align the right
edge of the stem with the right edge of a prominent syllable (see chapter two for details). The
exception comes from nouns where the schwa, in some cases, shows up before the second
consonant if its sonority value is greater than that of the third consonant.
Given the fact that the diminutive morpheme is a full vowel, it can be placed only after
the second consonant to yield the correct output. Consider the possible candidates of an input
such as /blγ-a, i/. The /i/ stands for the diminutive affix which comprises also the feature [+rd],
which is not germane to the argument here. The diminutive morpheme is introduced by an
alignment constraint requiring coincidence of the left edge of this morpheme and the left edge of
the PWd:

-5-

/blγ−a, {i, [+rd]}/ *COMPLEX ALIGN DEP-IO *Min-σ

(Dim, L, PWd, L)
a. b.li.γa **! *
b. bli.γa *! **
0c. bil.γa *

The alignment constraint can never be satisfied because if we place the diminutive /i/ in initial
position, we will fatally violate the constraint ONSET. Placing the morpheme after the initial
segment of the base seems to be the closest location possible to the left edge of the PWd. This is
exactly the case with candidate 5c which is wrongly predicted to be the optimal candidate. The
real optimal candidate is 5a and is ruled out in 5 because it incurs a fatal violation of ALIGN
(Dim, L, PWd, L).
Assuming that the diminutive morpheme must be aligned with the right of the PWd would
give the correct output candidate:

256
-6-

/blγ-a, {i, [+rd]/ *COMPLEX ALIGN DEP-IO *Min-σ

(Dim, R, PWd, R)
)a. b.li.γa ** *
b. bli.γa *! **
c. bil.γa ***!

The candidates that best satisfy ALIGN (Dim, R, PWd, R) are 6a and 6b. But only 6a is retained;
6b is excluded on the ground that it violates *COMPLEX. Note also that the only candidate
whereby ALIGN (Dim, R, PWd, R) is satisfied is one that suffixes the diminutive morpheme,
giving rise to a form such as *[bəlγay] which is excluded because it violates IDENT-IO [cons]
(i.e. the input vowel morpheme loses its moraic status and is realized as a glide).
With a quadrisegmental root, the alignment constraint in 6 would decide in favor of the
form that suffixes the diminutive morpheme and epenthesizes a schwa between the second and
third consonants of the base. Consider some output candidates from an input such as /mħbq, i/:

-7-

/mħbq, {i, [+rd}/ *COMPLEX ALIGN DEP-IO *Min-σ

(Dim, R, PWd, R)
a. m.ħi.bəq ***! * *
b. miħ.bəq ****! *
c. məħ.biq *! *
d. mħi.bəq *! *** *
0e. m.ħəb.qi * *

Given the constraints in this tableau, the optimal output is candidate 7e, a form that does not
correspond to the actual output [m.ħi.bəq]. This form incurs two violation marks of ALIGN
(Dim, L, PWd, L) and three of ALIGN (Dim, R, PWd, R). This points to the fact that other
constraints, which play a major role in deciding among the output candidates, are missing.
Nevertheless, we will assume ALIGN (Dim, L, PWd, L) to be the right constraint since it allows
for a better analysis for both trisegmental and quadrisegmental bases.

257
 To account for diminutive formation in CMA, we propose a constraint dubbed INITIAL-
CC which has the effect of forcing the diminutive affix in 7 (as well as other affixes as will be
seen below) to be placed after the initial CC sequence of the base. This constraint is given in 8
below:

-8-

INITIAL-CC
Words must begin with two consonants.

This constraint predicts that words must begin with complex onsets, a fact which points out that
this constraint must be ranked below *COMPLEX. The constraint INITIAL-CC is operative in a
number of morphological categories in CMA other than the diminutive as could be seen in the
examples below:

-9-

a. The plural
Singular Plural Gloss

kəlb klab dog

ktəf ktaf shoulder
STəl STula bucket
qərd qruda monkey
ktab ktuba book

b. Adjective formation
Noun Adjective Adj. Gloss

kwbəR kbir old

SγwəR Sγir young
DRafa DRiyyəf nice
smuniyya smin fat
lTafa lTif nice

c. Verb derived from adjective

Adjective Verb Vb Gloss

kħəl kħal darken

byəD byaD whiten
smin sman become fat

258
 zRəq zRaq become blue
Smək Smak become deaf

d. Deverbal nouns
Verb Noun Vb Gloss

DRəb DRib hit

qtəl qtil murder
nʕəs nʕas sleep
fiq fyaq wake up
nuD nwad get up

The vowel in the items in the middle column is morphemic since it marks the morphological
category. It is placed after the second segment of the base, exactly as is the case with the vowel
of the diminutive. The items in 9 as well as all the diminutive forms without exception start with
a cluster of consonants and therefore satisfy the constraint
INITIAL-CC.
One could possibly argue that the constraint INITIAL-CC could be dispensed with for the
typological consequences it has and that its effects follow from the interaction of other universal
constraints. Thus, one could say that trisegmental forms such as the ones in 9 are governed by a
prosodic constraint requiring the output to consist of an iambic foot of the type LH (where L is a
minor syllable) and that it is this requirement that determines the location of the morphemic
vowel. Given the constraint LH, there is only one place for the plural morpheme [a] to be placed
in the nominal input /klb/; it is after the second segment of the base to get the correct output
[klab]. If on the other hand, the plural morpheme is placed after the first segment of the base, the
output obtained is the ungrammatical *[kaləb], a form that violates LH. Finally another
alternative position would be for the plural morpheme to be suffixed to the base; the result that
would be obtained is the ungrammatical plural *[kəlba]1, a form that also violates both LH and
INITIAL-CC.
Assuming a constraint of the type LH does not solve the problem of the morpheme
location in other morphological categories such as the diminutive which is the main focus in this
chapter. As shown in the data in 1 above, bases with one major syllable always surface with two
major syllables. Thus, for example, the input /klb, i/ never surfaces as *[kwlib], that is a minor LH

1
The ungrammatical plural form *[kəlba] should be kept different from the output [kəlba] ‘bitch’, where the final
vowel stands for the feminine suffix.

259
foot. Such a form, as it will be shown in the sections to come, does not conform to the required
foot of the diminutive which is either a true iambic foot of the type LH as in [TbiSil] and [bririd],
or an iambic foot of the type LL as in [kwliyyəb] and [kwira]. Since LL is a possible foot type in
bases with one major syllable, one could wonder why a form such as *[kwiləb] is ruled out
despite the fact that it is a foot of the type LL. The answer comes from the constraint INITIAL-
CC. The form *[kwiləb] is ruled out not because of the prosodic requirement on foot type but
because the diminutive morpheme is not placed after an initial CC sequence.
Quadrisegmental bases show the same behavior as trisegmental bases such as [kəlb]. If it
were only a question of foot structure, we would expect an input such as /mħbq, i/ to surface as
[məħbiq], a form that corresponds to a true iambic foot of the type LH. But since this output form
does not place the diminutive morpheme after an initial CC sequence, it fails exactly because of
the constraint INITIAL-CC.
Assuming that the output of the diminutive forms is governed by a prosodic constraint
requiring that they conform to an iambic foot of the type LH or LL forces the placement of the
diminutive morpheme to be after two consonants and hence satisfaction of INITIAL-CC. Take
for example the nouns [bəlγa] and [məħbəq]. If the diminutive morpheme is placed after the
initial consonant of the base, the results obtained are the forms *[bilγa] and *[miħbəq] which are
ruled out because they correspond neither to LH nor to LL. Given the fact that the diminutive
morpheme is a full vowel, placing it after the initial consonant of quadrisegmental and suffixed
trisegmental bases such as [bəlγa] and [məħbəq] would result in an anti-iambic foot of the form
HL. However, placing this morpheme after an initial cluster and therefore satisfying INITIAL-
CC ensures that medial CC clusters would never arise and that the output obtained would always
conform to a foot of the type LH or LL.
With the constraint INITIAL-CC in hand, we show how the diminutive output candidate
[mħibəq] is selected among other candidates. As already mentioned, we assume that the
diminutive morpheme must be left aligned with the PWd. This constraint must be outranked by
INITIAL-CC, which in turn must be outranked by *COMPLEX. The target word must conform
to a foot of the type LH in the ideal cases or else to a foot of the type LL as shown in 10:

260
-10-

/mħbq, i/ INITIAL-CC ALIGN LH LL

(Dim, L, PWd, L)
)a. m.(ħi.bəq) ** *
b. (miħ.bəq) *! * * *
c. (məħ.biq) ****! *
d. m.(ħəb.qi) *****! *

Candidates 10b is excluded because it violates the constraint INITIAL-CC. Candidates 18c and
18d are also eliminated because the first places the diminutive morpheme before the final
segment of the base while the second suffixes the same morpheme to the base and in so doing
they incur a fatal violation of ALIGN (Dim, L, PWd, L).
To sum up, this section has tried to show the need for a constraint of the type INITIAL-
CC which has the effect of forcing the diminutive morpheme (as well as other morphemes that
show a similar behavior) to be placed after an initial CC sequence in order for the output to
conform to an iamb of the type LH or LL. We have shown that neither the alignment constraint
alone nor the constraints on foot types allow us to come up with the correct output. It is only
these constraints in combination with INITIAL-CC and other constraints, which we will consider
as we proceed further, that could account for the diminutive in CMA. The need for INITIAL-CC
finds its justification in diminutive cases where an initial labial consonant is geminated. To these
cases we shift in the next section.

5. EVIDENCE FOR INITIAL-CC

Further support for the constraint INITIAL-CC comes from cases involving labialization
and gemination of the consonants [f, b, m] before [w]. In order to fully understand the process of
labial gemination, prior knowledge of how labialization works in CMA is required.

5.1 Labialization

CMA is characterized by a set of labialized consonants which include the labials [b, f, m]
and the dorsals [k, g, x, γ, q]. These consonants are subject to three types of labialization. The

261
first type is lexical labialization and is called so because it accompanies the word in its various
realizations. The second type is morphological labialization. It serves to contrast different
morphological classes. The third type of labialization is referred to as phonological labialization.
It is the result of the contiguity of a labial consonant and the velar [w].
If both morphological and phonological labialization are to a large extent predictable,
lexical labialization is not as could be shown in the examples below:

-11-

Base Plural Diminutive Gloss

kwəmm kwmayəm kwmiyyəm sleeve

kwərsi kwrisi kwrasa chairs
gwəRSa gwRaSi gwRiSa small (circular) loaf of bread
gwəffa gwfaf gwfifa basket
xwzana xwzayən xwzina tent
γwTa γwTawat γwTiwa cover
qwənt qwnat qwniyyət corner

Boudlal (1998) assumes that this kind of labialization is attributed to the influence of Tashlhit
Berber, and as such the phoneme inventory of CMA should incorporate both the labialized
dorsals as well as their non-labialized counterparts. (For more details, see section 5.1 in chapter 1
above)
Morphological labialization, on the other hand, is associated only with the dorsal
consonants and serves to mark certain morphological classes such as the diminutive and the
imperative as shown in the examples below:

-12-

a. Base Diminutive Gloss

kəlb kwliyyəb dog

qərd qwriyyəd monkey
xruf xwriyyəf a lamb
gəmla gwmila louse
γliD γwliyyəD fat

262
b. 3 sg. Perfective 2 sg. Imperative Gloss

dxəl dxwəl come in

xrəʒ xwrəʒ go out
gʕəd gwʕəd sit down
qtəl qwtəl kill
tqəb tqwəb pierce

The labialization in 12b serves to contrast the perfective and the imperative forms; the
labialization in 12a indicates that the word is in the diminutive form. Both types of labialization
differ from each other in that the one associated with the diminutive is unbounded and as such
applies to any dorsal consonants whereas the one associated with the imperative applies only to
some words with dorsal consonants. In other words, there are some words whose dorsal
consonants are not labialized in the imperative. We will not pursue the argument here but for a
detailed account of this kind of labialization, the reader is referred to Boudlal (1998). What is of
relevance to us here is that the labialization associated with the diminutive is no more than the
full realization of the diminutive morpheme which is assumed to be the vowel [-i-] and the
feature [+round] (Al Ghadi 1990, Boudlal 1993). It will be seen below that labialization in the
diminutive case is the result of an alignment constraint requiring that the feature [+round] be left
aligned with the prosodic word, thus ensuring that any dorsal consonant occurring in word initial
position gets labialized.
Also of relevance to us in this chapter is what we refer to as phonological labialization:

-13-

Singular Plural Gloss

baliza *bwaləz < bbwaləz suitcases

fuTa *fwaTi < ffwaTi towels
manTa *mwanəT < mmwanəT blanket
musəm *mwasəm < mmwasəm annual festival
musiqa *mwasəq < mmwasəq music
fuqiyya *fwaqi < ffwaqi a Moroccan gown
bulisi *bwaləs < bbwaləs policeman

The examples in 13 show that the sequences fw, bw, mw are not allowed in CMA. Whenever
such sequences arise, labialization applies to give a labialized consonant. Previous approaches to
phonological labialization (see Al Ghadi 1990, El Himer 1991, Boudlal 1993, 1998) assume that

263
words such as those in 13 are derived in two stages. After the affixation of the plural morpheme
[a], the [u] of the singular forms changes into [w] to serve as an onset to the plural morpheme.
The initial labial consonant then gets geminated and the [w] is realized as a secondary labial on
the geminate.
It should be noted here that not all cases of a labial and [w] result in labialization.
Consider the following examples where the prefix [m-] in 14a, denoting the passive participle
and the preposition consonants in 14b, are attached to a verb beginning with the glide [w] without
there being a labialization process:

-14-

a. Vb base PP Vb Gloss

wəlləf m-wəlləf get accustomed to

wəlləd m-wəlləd assist in childbirth
wəkkəl m-wəkkəl feed
wəSSəl m-wəSSəl walk (someone)
wəDDəR m-wəDDəR lose (something)
wənnəs m-wənnəs accompany (someone)

b. Noun Prep + Noun Gloss

waħəd b-waħəd by one

walu b-walu with nothing
wad f-wad in a river

Within a Lexical-Phonology framework, Boudlal (1993) has shown that the domain of
labialization is the first stratum. This shows why labialization fails to apply to the passive forms
in stratum 2 and to the words in 14b in the postlexical stratum.
Within the OT framework, the mismatch between the data in 13 and the data in 14 could
be explained by reference to domain-specific instantiations of OCP (lab) much in the spirit
undertaken by Selkirk (1995b) for the analysis of Berber. Thus it could be argued that the OCP
(lab) applies at the stem level only, and that at the word or phrase level, this constraint is blind to
any sequence of a labial consonant and [w]. This shows that OCPstem (lab) must dominate
OCPword (lab). The morphological composition of representative items from 13 and 14 is given in
15 below:

264
-15-

a. [[bwaləz]stem]word
b. [m[wəlləf]stem]word
c. b[[waħəd]stem]word

Since [b] and [w] are juxtaposed within the stem in 15a, the form [bwaləz] is ruled out because it
violates OCPstem (lab). In 15b, the sequence mw occurs at the word level and as such avoids
violation of higher-ranked OCPstem (lab). Finally the sequence bw violates none of the OCP
constraints since the preposition [b-] is introduced at the phrase level.
The case of OCP constraint dealt with in the present work is the one that applies at the
stem level, i.e. OCPstem (lab). It should be noted that this labialization does not apply in cases
where a labial consonant is followed by the vowel [u], something that points out to the necessity
of distinguishing [u] and [w]. Like Hammari (1996), we assume that [u] and [w] have the
structure in 10 below:

-16-

a. u b. w
| |
Dors Lab
| |
Lab Dors

The labial consonants themselves have the structures in 17:

-17-

a. b b. f c. m

Lab Lab Lab

The representations in 16 and 17 allow us to explain why the sequence of a labial consonant and
[w] are not allowed. Clearly such a sequence violates the OCP and that explains why the
consonant is labialized, i.e. getting the secondary dorsal articulation. The whole picture is
represented below:

265
-18-
Input OCP (lab) Dorsal Spread Output

b w b w b w bw
│ │ │ │ │ │ │
Lab Lab Lab Lab Lab Lab Lab
│ │ │ │
Dors Dors Dors Dors

The representations in 18 fairly explain how a sequence of labial consonant and [w] leads to a
labialized consonant but does not explain how the resulting geminate in 13 above is obtained. In
fact feature geometry cannot explain that. Suffice it to raise the problem here. It will be shown in
the following subsection that gemination is the result of the constraint requiring the diminutive
forms to start with two consonants.
To sum up, This subsection has raised questions that relate directly to the diminutive in
CMA. It has shown that labialization in the diminutive is a consequence of realizing the feature
[round], which is part of the diminutive morpheme, on an initial dorsal consonant. As to the
labialization of the labial consonants, we have shown that it is the result of an OCP constraint,
dubbed OCP (lab), which prohibits a sequence of a labial consonant and [w].
In the next subsection, we will consider how the labialization and gemination of the
consonants [b, f, m] are achieved to satisfy the constraint INITIAL-CC.

5.2 Labial Consonant Gemination

As it has already been mentioned above, the labial consonants [b, f, m] are labialized and
geminated before [w] in some morphological categories such as the plural and the diminutive.
Consider some examples from both classes for illustration:

-19-

a. Singluar Unattested Pl. Actual Pl. Gloss

in CMA in CMA

fuTa *fwaTi ffwaTi towel

fasi *fwasa ffwasa native to Fes
muTuR *mwaTəR mmwaTəR motorcycle
baliza *bwaləz bbwaləz suitcase
manTa *mwanəT mmwanəT blanket

266
b. Base Unattested Dim. Actual Dim
in CMA in CMA

fuTa *fwiTa ffwiTa

fasi *fwisi ffwisi
muTuR *mwiTiR mmwiTiR
baliza *bwiliza bbwiliza
manTa *mwiTa mmwinTa

In derivational terms, the data above show that after the affixation of the plural morpheme in 19a
and the diminutive in 19b, the labial segment and the dorsal [w] are juxtaposed, giving rise to
intermediate forms (the asterisked items) which are not attested in what Boudlal (1998) refers to
as Southern Varieties of MA. In CMA, these forms are not attested because they violate the
constraint OCP (lab) by juxtaposing two labials of the same rank (see section 5.1 above).
In the Northern Varieties of MA (cf. the variety of MA spoken in Fès, for example),
labialization is almost absent, giving rise to forms such as the following:

-20-

Base Diminutive Gloss

faR fwir mouse

bab bwiba door
mus mwis knife
fanida fwinida candy
ma mwiha water

The data in 20 reveal two things. First, the constraint INITIAL-CC is not particular to a specific
variety; it is observed in all the varieties of MA. Second, in varieties such as the ones in 20 the
feature [+round], which is held responsible for the labialization of geminate labial consonants and
dorsal consonants, does not show up; it shows up only in what is referred to as Southern Varieties
of MA. (See Boudlal 1998, for details about labialization in Southern Varieties of MA)
In the present work, we assume that the diminutive morpheme consists of the vowel [-i-]
and the feature [+round]. The vowel [-i-] is placed after the initial CC sequence of the base,
whereas the feature [round] attaches to the initial consonant of the base if it is dorsal or labial. We
assume that the feature [+round] is not realized on the labial consonants [b, f, m] when they are
not geminated. That this is true is shown by cases such as [mriwa], [bniyya] and [friyyəx] whose
labial consonants have not undergone labialization. The diminutive cases that show the

267
labialization of these consonants (cf. [mmwiyyəs] “knife”, [bbwiyyət] “room” and [ffwiyyər]
“mouse”) are cases that result from the juxtaposition of the labial consonants and [w]. The
gemination of the labial consonant is the result of the constraint INITIAL-CC.
The labialization of consonants to mark certain morphological classes is reminiscent of a
similar phenomena in Chaha treated in McCarthy (1983) and Gafos (1998). In this language,
certain morphological categories in verbs are marked by assigning the feature round to the
rightmost labializable velar or labial consonant as shown in the examples below taken from
McCarthy (1983:3):

-21-

Perfective 3 mas.sg.
Without object With 3 mas.sg. Object Gloss

dænæg dænægw hit

nædæf nædæfw sting
nækæb nækæbw find
nækæs nækwæs bite
kæfæt kæfwæt open
mæsær mwæsær seem
qætær qwætær kill

Labialization in this language applies regardless of the distance that separates the labializable
consonant from the end of the root. The scanning starts from right to left and the rightmost
consonant is labialized even if it is initial. In case the word has more than one potential
labializable consonants, it is the rightmost one that undergoes the process (cf. [nækæbw], for
example).
Within an OT framework, Gafos (1998) assumes that labialization in Chaha could be
accounted for by assuming an alignment constraint which requires that the [round] feature-
morpheme be aligned with the right edge of the output. For the diminutive cases in CMA, we also
assume that the same constraint holds except that it applies at the left edge of the output. This
constraint is stated in 22 below:

-22-

LABIALIZE (C, L, PWd, L)

The left edge of the prosodic word must be aligned with a labialized consonant.

268
This constraint must dominate the IDENT-IO [rd] faithfulness constraint which prohibits
changing the [round] feature that exists in the input.
Having said this, an explanation of the phenomenon of labialization of dorsals and
geminnate labials in the diminutives is in order. Given a base form such as [kəlba] “bitch”, Gen
could produce the following output candidates:

-23-

/klb-a, {i, [+rd]}Af/ LABIALIZE IDENT-IO [rd]

) a. kwliba *
b. kliba *!

The tableau above rules out candidate 23b because it fails to realize the feature [round] which is
part of the diminutive, thus incurring a fatal violation of LABIALIZE. Since [round] is part of the
diminutive affix, it could also be claimed that the non-realization of this feature constitutes a
violation of another constraint we state below:

-24-
MAX-IO [rd]
The feature [round] must be preserved in the input/output mapping.

Thus a form such as *[kliba] is suboptimal not only because it incurs a violation of the constraint
LABIALIZE but also because it violates MAX-IO [rd], a fact which suggests that this constraint
must dominate IDENT-IO [rd] as shown below:

-25-

/klb-a, {i, [+rd]}Af/ MAX-IO [rd] IDENT-IO [rd]

) a. kwliba *
b. kliba *!

However, such constraints are incapable of explaining why labialization affects only dorsal and
geminate labial consonants. What would happen in a form that does not consist of a dorsal
consonant? The data in 1 show that all the consonants, except the dorsal and geminate labial

269
ones, do not labialize and as such their failure to labialize would constitute a clear violation of
LABIALIZE and MAX-IO [rd]. Given an input form such as /dmʕ-a/ “tear” how do we get to the
correct output [dmiʕa] without ever labializing the consonant [d]? To account for the non-
labialization of [d] or [m], we assume the following markedness constraints on complex feature
combination formulated à la Selkirk (1993):

-26-

*Tw, *Hw, *Bw >>*Kw

Each of these capital symbols in 26 stands for a whole class. Thus Tw stands for the class of
coronals, Hw for pharyngeals, Bw for labials and finally Kw for dorsals. Since we do not have
evidence for ranking the first three constraints in 26, we assume that the three of these dominate
*Kw. In order to ensure that only dorsal consonants are labialized in the diminutive, we need to
assume that both LABIALIZE and MAX-IO [rd] must be ranked on top of *Kw and below the
rest of the constraints. In the tableau below we show how the ranking *Tw, *Hw and *Bw above
the constraint LABIALIZE and MAX-IO [rd] yields the correct output. Since the effect of
LABIALIZE could be achieved by MAX-IO [rd], only the latter constraint will appear in the rest
of the tableaux presented in this chapter:

-27-

/dmʕ-a, *Tw *Hw *Bw MAX-IO *Kw IDENT-

{i, [+rd]}Af / [rd] IO [rd]
a. dwmiʕa *! *
b. dmwiʕa *! *
)c. dmiʕa *

This tableau shows that it is more optimal not to realize the [round] feature than to violate the
higher-ranked markedness constraints. Thus, a candidate labializing any consonant other than
dorsal is doomed to be ruled out
Going back to the data in 19 above, the asterisked forms are ruled out because they allow
the juxtaposition of two labial consonants. We have shown that this is an instance of the OCP
violation and that it is resolved by realizing the [w] on the labial consonant, thus giving rise to a

270
labialized consonant. If this is the case, how can the gemination be explained? Clearly it can only
be attributed to the constraint INITIAL-CC which demands that the diminutive morpheme (as
well as other morphemes such as the plural morpheme) be placed after two initial consonants. If
no consonant is available, recourse is made to the lexical default segment of the language noted
as /u/.
Consider a diminutive form such as [ffwiTa] from the input /fuT-a/. Within a derivational
framework, the input/output mapping would possibly look something like 28:

-28-

Input: /fuT-a/
Dim. Af.: fuiTa
INITIAL-CC: fwiTa
OCP (lab): fwiTa
INITIAL-CC: fwwita
OCP (lab): fwwiTa
Gemination: ffwiTa
Output: [ffwiTa]

The affixation of the plural morpheme [a] gives rise to intermediate [fuiTa] in which the base [u]
loses its moraic status to serve as an onset to the diminutive vowel. The form [fwiTa], in turn,
changes into [fwiTa] to satisfy the constraint OCP (lab). Given the fact that the constraint
INITIAL-CC is not satisfied in [fwiTa], the default segment of the language (i.e. /u/) is introduced
to provide a second consonant, thus resulting in a form that violates OCP (lab) (cf. *[fwwiTa]).
Because of the constraint OCP (lab), the language does not epenthesize /u/ to satisfy INITIAL-
CC; rather it geminates the initial consonant to give the correct output form [ffwiTa].
However, the analysis proposed in 28 cannot be accepted because it is operational and is
therefore incompatible with the principles laid down by the OT framework. It further recognizes
intermediate forms that are impossible to predict from either the input or the output.
The behavior of diminutive forms with initial geminates indicates two things: first, the
constraint INITIAL-CC is essential if we are to get the optimal output. It is the constraint
INITIAL-CC that forces the gemination of the base initial consonant and not the OCP (lab), a fact
which indicates that the two constraints are not ranked with respect to each other. Second,

271
examples such as the ungrammatical form *[fwiTa] present further evidence to the ranking
established in 26 which predicts that only simple dorsal consonants (i.e. Kw) are labialized. The
grammatical form [ffwiTa], on the other hand, shows that labialization of geminates does in fact
occur but is restricted only to labial consonants. If we assume that the ranking concerning
labialization is the same whether segments are geminates or not, then there is no way to account
for initial labial geminates. For illustration, consider the competing candidates for the diminutive
form of the word [fuTa]:

-29-

/fuT-a, *Tw *Hw *Bw MAX-IO *Kw IDENT-

{i, [+rd]}Af / [rd] IO [rd]
a.ffwiTa *! *
0b. ffiTa *

It is clear that the constraints in 29 alone cannot derive the correct output. Instead they wrongly
predict that the optimal candidate is [ffiTa]. The question that needs to be addressed at this stage
is the following: why is labialization allowed when the labial consonant is a geminate and
disallowed when the labial consonant precedes another consonant?
To answer this question, let us assume that the morphological labialization of the
diminutive is in principle available to all forms regardless of their beginning consonants and that
the labialization or non-labialization of initial dorsal and labial consonants follows from the
interaction of the markedness constraints in 26 above and other faithfulness constraints of the
MAX family. Let us further assume that labialization is permitted with the geminate labial when
this labial happens to be in onset position as it is the case with BBw, and disallowed when the
labial consonant is followed by another consonant, i.e. when it is not in onset position. Thinking
about labialization this way, it is reminiscent of the Beckman (1998) positional faithfulness
whereby certain positions are more privileged than others. In the case considered here, syllable
onset positions are more privileged than other positions. This privilege can be made conspicuous
by ranking MAX-Onset [rd] above MAX-IO [rd]. The whole scenario is given in 30 below:

272
-30-

Constraints on labialized consonants (to be revised)

*Tw , *Hw , OCP (lab) >> MAX-Onset [rd] >> *Bw >> MAX-IO [rd] >> *Kw

Realizing the feature [round] in a surface onset must dominate the constraint against rounded
labials which, in turn, must dominate the general MAX-IO [rd] constraint. With the ranking in 30
above, a labial geminate could be obtained in a straightforward manner. Thus the competing
candidates for diminutive of [fuTa] are given in the following tableau:

-31-

/fuT-a, OCP (lab) MAX-Onset *Bw MAX-IO

{i, [+rd]}Af/ [rd] [rd]
)a. ffwiTa *
b. fwiTa *! * *
c. ffiTa *! *

Notice that a form such as [fwiTa] is ruled out because it does not satisfy the constraint INITIAL-
CC, something that argues for ranking this constraint above MAX-Onset [rd]. The pair
*[fwiTa]/[ffwiTa] presents the first piece of evidence that INITIAL-CC must be undominated. It
also shows that the labialization of labial consonants in the diminutive is of course driven by the
constraint OCP (lab) forcing any sequence Bw to be realized as Bw.
The ranking in 30 can also account for the non-labialization of a labial consonant when it
is not in onset position. Thus a word such as [bəSla] is realized in the diminutive as [bSila] and
not *[bwSila]. Both candidates violate MAX-Onset [rd] and MAX-IO [rd] but *[bwSila] is ruled
out because the labialization of the labial consonant incurs a fatal violation of *Bw. Trying to
keep [round] in a surface onset would incur the undominated constraint *Tw. So the optimal
candidate is the one that does not realize the [round] feature on any of the initial consonants.
Next, let us see if the ranking in 30 could account for the non-labialization of a simple
labial consonant in onset position. Thus, in the diminutive form of a word such as [dəmʕa], the
consonant [m], which serves as an onset to the diminutive vowel [-i-], must be labialized and as
such should surface as *[dmwiʕa] given the constraints already stated:

273
-32-

/dmʕ-a, *Tw *Hw OCP (lab) MAX- *Bw MAX-IO

{i, [+rd]}Af / Onset [rd]
[rd]
a. dmiʕa *! *
w
0b. dm iʕa *
c. dwmiʕa *! *

It is evident that these constraints, and consequently their ranking, cannot generate the correct
output. By virtue of higher-ranking MAX-Onset [rd], they predict that any labial consonant,
whether it is simple or geminate, should be labialized. The example in 32 shows that in order to
account for the non-labialization of a labial consonant and its labialization when it is a geminate,
we need in addition to the constraints in 26 other markedness constraints that apply to labialized
geminates. These constraints are given in 33 below:

-33-

*TTw, *HHw, *KKw >> *BBw

We do not have evidence as to the ranking of the first three constraints and as such we assume
that they must not be ranked with respect to each other and that the three of them must dominate
the constraint *BBw, given the fact that a labial geminate occurs in CMA. Incorporating
markedness constraints on labialized geminates obviates the need to make recourse to the MAX-
Onset [rd] constraint. All we need to account for the (non)labialization of dorsal and labial
consonants (geminates and non-geminates) are the markedness constraints in 26 and 33 along
with the OCP and the MAX-IO [rd] faithfulness constraint. Since both Bw and KKw never surface
in the diminutive while BBw and Kw do, it follows that the constraint *Bw and KKw must
dominate MAX-IO [rd] which in turn must dominate *BBw and Kw as shown in 34:

-34-

Constraints on the labialized consonants in CMA

*Tw , *Hw , OCP (lab) >> KKw, *Bw >> MAX-IO [rd] >> BBw, *Kw >> IDENT-IO [rd]

With this ranking, let us see how the diminutive form of [dəmʕa] is obtained:

274
-35-

/dmʕ-a, OCP (lab) *Bw MAX-IO [rd] BBw IDENT-IO

{i, [+rd]}Af/ [rd]
)a. dmiʕa *
b. dmwiʕa *! *

The ranking shows that a non-geminate labial consonant never surfaces even if it is in onset
position as wrongly predicted in 30 above. When this labial consonant is a geminate, the [round]
feature reemerges as a result of lower-ranking BBw. Consider the different candidates of the
diminutive form of [fuTa] for illustration:

-36-

/fuT-a, OCP (lab) *Bw MAX-IO [rd] BBw IDENT-IO

{i, [+rd]}Af/ [rd]
)a. ffwiTa * *
b. fwiTa *! *
c. fwiTa *! *
d. ffiTa *!

The realization of the [round] feature on simple labial consonants in 36b leads to the violation of
*Bw (and INITIAL-CC). This violation eliminates candidate 36b from the race to optimality. 36c
is excluded because it violates OCP (lab) by juxtaposing a labial consonant and the glide [w].
Finally 36d is excluded because the feature [+rd] of the input fails to surface. Note that the
optimal candidate violates another constraint against having geminates, a fact which shows that
*BBw must dominate *GEM.
A question that raises itself when dealing with cases presenting initial geminates is why
for example dorsal consonants are not geminated in the diminutive while labial consonants are. In
other words, why is [kwliba] optimal, whereas *[kkilba] and *[kkwilba] are not?
The answer to this question comes from ranking KKw above MAX-IO [rd] and Kw as the
tableau below shows:

275
-37-

/klb-a, OCP (lab) *KKw MAX-IO [rd] Kw IDENT-IO

{i, [+rd]}Af/ [rd]
)a. kwliba * *
b. kkilba *!
c. kkwilba *! *

This tableau shows that it is more optimal to realize rounding on a simple dorsal consonant than
geminate this consonant and/or labialize it as is the case with 37b and 37c. Note also that the
suboptimal candidates in 31b and 37c could also be excluded because they don’t conform to an
iambic foot of the type LH or LL.
The final ranking responsible for the labialization or the non-labialization of dorsal and
labial consonants is given in 38 below:

-38-

*Tw *Hw OCP *TTw *HHw

Bw KKw

MAX-IO [rd]

Kw BBw

IDENT-IO [rd]

To sum up, this section has shown that the gemination exhibited by bases whose initial
segment is labial provides support for the constraint INITIAL-CC. It has shown that words
beginning with a CV, where the C stands for a labial consonant, make recourse to the default
segment of the language [w] to satisfy the constraint INITIAL-CC. The juxtaposition of the labial

276
consonant and [w] leads to the violation of OCP (lab). This violation is avoided not by the
deletion of [w] but by the gemination of the labial consonant, hence the satisfaction of INITIAL-
CC. It has been further shown that INITIAL-CC needs to be supplemented by markedness
constraints on both simple and geminated labialized consonants as well as faithfulness constraints
militating against any change that occurs in the input-output matching.
Having justified the need for the constraint INITIAL-CC, we turn in the next sections to
consider diminutive formation in CMA and show how it could best be accounted for by a set of
markedness constraints and their interaction or non-interaction with prosodic and faithfulness
constraints. First, we consider how certain diminutive forms which do not fulfill prosodic
requirements on the output are augmented to achieve this ideal shape.

6. AUGMENTED DIMINUTIVE FORMS

In this section we show how the OT framework, in general and CT in particular could
account for the diminutive in CMA. It is divided into two major subsections. The first deals with
the diminutive cases that are augmented by the addition of a whole syllable whose nucleus is the
schwa, i.e. what is referred to as the default syllable in the language (Al Ghadi 1990). The second
subsection deals with the diminutive forms that are augmented by the suffixation of the feminine
morpheme [-a].
Throughout this section, it will be shown that the diminutive forms in CMA abide by a
prosodic constraint which requires that the output consist of an iamb of the type LH or LL. In
particular, It will be shown that bases containing one major syllable never achieve the true LH
iamb; instead they proceed to augmentation to avoid what we refer to as a minor LH iamb, that is
an iamb whose light syllable dominates a consonant.

6.1 Augmentation as the Addition of the Default Syllable

Having shown the utility of the constraint INITIAL-CC for an adequate account of the
diminutive, let us now return to the data in 1a-e to see how the diminutive forms can be derived
given the theoretical framework adopted in the present work. Representative items are given in
39 below:

277
-39-

Base Diminutive Gloss

fərx *frix / friyyəx bird

kəlb *kwlib / kwliyyəb dog
kas *kwis / kwiyyəs a (drinking) glass
wəld *wlid / wliyyəd boy
bit *bbwit / bbwiyyət room
sdər *sdir / sdiyyər chest

Recall that all the cases with one major syllable we are considering surface with two major
syllables and this by building a syllable whose nucleus is the schwa. In this subsection, we will
try to answer the following question: what is it that forces this augmentation? And why is
augmentation located word-internally?
The answer to the above questions comes from prosody and its interaction with other
constraints in the grammar. We have seen in the previous chapter that the passive participle
makes recourse to u-epenthesis to achieve an iambic target of the type LH. In the case of the
diminutive, augmentation targets a foot of the type LL. If this is the case, why isn’t the
diminutive of the bases in 39 simply [C.CVCu] with simple addition of the lexical default
segment /u/. Forms having this pattern could be ruled out on the ground that they violate ALIGN-
R since the right edge of the root does not correspond to the right edge of the syllable. u-
epenthesis cannot apply at the left edge of the root because of ALIGN-L requiring that the left
edge of the root correspond to the left edge of the PWd. Therefore, the only location for
augmentation to apply is word-internally. The diminutive, like the PP, makes recourse to
augmentation for prosodic purposes. While the target in the PP is achieving the ideal LH iamb,
the target in the diminutive is an iamb of the type LL. This does not mean that LH iambs do not
occur in the diminutive; they do but only in disyllabic and trisyllabic bases. As a matter of fact, it
might be argued that the default iamb of the language is of the type LL as could be shown in the
following examples. (AP stands for active participle and N. Ins for noun of instance):

278
-40-

Verb A.P N. Ins Vb Gloss

ktəb katəb kətba write

DRəb DaRəb DəRba hit
lbs labəs ləbsa dress up
dxəl daxəl dəxla come in
nʕəs naʕəs nəʕsa sleep
lʕb laʕəb ləʕba play
qtl qatəl qətla kill
gls galəs gəlsa sit down

In chapter five, we have shown that an input form such as /m-biʕ/ could give rise to the forms
[məbyuʕ] and [mbiʕ],both of which correspond to an LH iamb. However, only [məbyuʕ] is a
correct output. To exclude [mbiʕ], we have argued that CMA should incorporate a constraint
which distinguishes between a true LH iamb and a minor LH iamb, in the same way as we have
distinguished major from minor syllables.
Similarly, augmentation in the diminutive forms is achieved to avoid a minor LH iamb of
the type given in 41:

-41-

Minor LH iamb

Ft

σ σ
│
µ µ µ
│ │ │
f r i x

Before we show how a form such as [friyyəx] is chosen as the optimal candidate, it should be
noted that the diminutive forms in 39 (and the rest of the diminutives in 1a-e above) do conform
to an LL iamb in spite of the fact that the first major syllable is closed. Al Ghadi (1990), for
example, assumes that the diminutive of the base [fərx] is derived as follows:

279
-42-

Input frx
Dim. Affix frix
Core syllabification f.ri.x
Dim Prosodic Constraint f.ri.əx
Onset filling f.ri.yəx
Output [friyəx]

After the affixation of the diminutive morpheme [-i-], syllabification applies to give the
monosyllabic output [frix] (not counting the minor syllable of course), a form that does not
satisfy the diminutive prosodic constraint which requires that the output be minimally constituted
of two syllables. So another syllable has to be built to derive the correct output. For Al Ghadi
(1990) the building of this syllable proceeds as follows: first a schwa is epenthesized to serve as
the nucleus. Second, since onsetless syllables are not allowed in CMA, a glide is added to serve
as the onset to this syllable whose nucleus is the epenthesized schwa and whose coda is the final
consonant of the base.
Boudlal (1993) has adopted the same analysis and assumes that [friyəx] is the form
obtained at the lexical level and that the gemination of [y] to give [friyyəx] is a matter that takes
place at the postlexical level in the module of phonetic implementation. The same process takes
place in some cases where a glide is epenthesized for onset purposes as could be seen from 43
below:

-43-
Singular Plural Gloss

a. biru biruwwat office

kilu kiluwwat kilo
trikku trikkuwwat sweater
ʒili ʒiliyyat vest

b. Siniyya Siniyyat tray

fuqiyya fuqiyyat a traditional gown
ħəwliyya ħəwliyyat a ewe

280
The data above show that the glide is epenthesized to serve as an onset to the vowel-initial plural
suffix [-at] or the feminine suffix [-a] in the singular form. The gemination of the epenthetic glide
in 43 does not have a morphological status; it is there to regularize the pronunciation. One could
wonder whether gemination is a case of off-gliding that is realized on the vowel and as such
could be represented as [iy] and [uw]. This off-gliding is phonetically justified since it anticipates
the next sound which is a glide itself. For this reason, we assume that the diminutive forms in 1a-
e have the shape of an LL iamb. This could be made possible if we assume that the coda part of
the geminate is not moraic and that the glide is ambisyllabic as the representation of the
diminutive word [fwriyyəx] “bird” below shows:

-44-

PWd

Ft

σ σ σ

µ µ µ

f r i yə x

Having shown the need to distinguish between a minor iamb and a true iamb of the type LH as
well as the need to consider the first syllable in 44 as light, let us now see how the diminutive
forms in 1 could be obtained within a constraint-based framework. The analysis will proceed in
such a way that only a representative item will be dealt with from each set of the items in 1
above. First, let’s consider the possible candidates for an input such as /frx, i [+rd]/:

-45-

/frx, INITIAL-CC *Min-LH LH LL DEP-IO

{i, [+rd]}Af/
a. (fi.rəx) *! * *
b. (f.rix) *!
)c. f.(riy.yəx) * ***

281
Note that in 45 details about labialization have not been included. The first candidate is
eliminated on the ground that it violates higher-ranked INITIAL-CC. It has made recourse to
schwa epenthesis to satisfy the LL iamb. Candidate 45b is also eliminated because it violates
*Min-LH. The third candidate is the optimal one in spite of the fact that it has proceeded to
augmentation to satisfy LL, thus incurring three violations of DEP-IO. Note further that in all the
diminutive forms, the initial syllabic consonant, which forms a minor syllable, is allowed outside
the iamb. Incorporating it into foot structure would constitute a fatal violation of FT-BIN. Stating
that it is not syllabic at all and that it should be included in the iamb would satisfy FT-BIN and
solve the problem in words such as (f.riyµ.yəxµ) but would raise it in all non-derived trisegmental
items such as (k.təbµ) and (kəlµ.b). In this work, we will continue to assume that the initial
consonant in the diminutive is moraic and is associated with a minor syllable which is adjoined
directly to the PWd.
Two other potential optimal candidates which are not included in 45 would be *[friyyux]
and *[frixəx]. Take first the candidate *[friyyux]. Given that LH dominates LL, *[friyyux] should
be chosen as optimal since it conforms to the most harmonic iamb, i.e. LH. In order to eliminate
this candidate, we make recourse to the constraint DEP-u (see chapter 5) which has the effect of
penalizing any cases of u-epenthesis. This constraint has to out-rank LH to avoid epenthesis in
cases such as [friyyəx].

-46-

/frx, INITIAL-CC DEP-u LH LL DEP-IO

{i, [+rd]}Af/
)a. f.(riy.yəx) * ***
b. f.(riy.yux) *! * ***

Notice that the ranking in 46 makes clear that INITIAL-CC must dominate DEP-u, a fact which
will become clear as we proceed in the analysis.
The second candidate that competes with the actual output is *[frixəx]. It satisfies both
INITIAL-CC and LL and would therefore be preferred to [friyyəx] on the ground that it incurs
only one violation of DEP-IO. How is it possible to block a form such as [frixəx] from being the
optimal output?

282
 To answer this question, it should be noted that *[frixəx] has made recourse to both
epenthesis and consonant spreading to satisfy the constraint LL. One way to block spreading in
cases like [frixəx] is to invoke the NO-SPREAD constraint proposed by McCarthy (1997:9) and
stated as follows:

-47-

NO-SPREAD S1-S2 (τ,ς)τςτℜ

Let τi and ςj stand for elements on distinct autosegmental tiers in two related
phonological representations S1 and S2, where
τ1 andς1 ∈ S1,
τ2and ς2 ∈ S2,
τ1 ℜ τ2, and
ς1 ℜ ς2
if τ2 is associated with ς2,
then τ1 is associated withς1.

The version of the NO-SPREAD constraint we will be using for CMA diminutives is one that
relates a mora and a consonant. Thus NO-SPREAD (µ, C) prohibits the spreading of a consonant,
resulting in the gain of new association lines. This constraint was originally proposed by
McCarthy (1997) to account for the non-spreading of the consonant [t] in the singular/plural
mapping [xaatam]/[xawaatim] instead of the unattested form *[xataatim], which violates NO-
SPREAD (µ, C). What the classical Arabic example shows is that the epenthesis of [w] is better
than the spreading of [t]. Likewise in the CMA diminutive cases, adding a whole syllable is more
highly valued than spreading the final consonant and then epenthesizing a schwa to satisfy LL. In
the tableau below we show how [friyyəx] wins over *[frixəx]. Since spreading a mora leads to
the satisfaction of LL and LH (cf. *[frixux], for example), it follows that NO-SPREAD (µ, C) has
to dominate both LH and LL in order to derive the correct output as shown in 48:

283
-48-

/frx, DEP-u NO-SPREAD LH LL DEP-

{i, [+rd]}Af/ (µ, C) IO
)a. f.(riy.yəx) * ***
b. f.(ri.xəx) *! * *
c. f.(riy.yux) *! * ***

Notice that the [+rd] feature has not been realized in the above form because the first consonant is
not dorsal or geminate labial. We have shown that labial consonants are labialized under special
circumstances, i.e. when they are geminated. Note also that forms such as *[friyyix] and *[frixix]
are also excluded although they conform to an iambic foot of the type LH. The reason is that they
epenthesize another root vowel node which is filled out with a copy of the diminutive vowel. This
epenthesis incurs a violation of DEP-V, which bans epenthesis of full vowels. (On the need to
separate DEP-V and DEP-ə, see chapter five, section 4.2)
Now, consider another example whose first consonant is dorsal. The tableau below
exposes the different candidates for the input
/klb, {i, [+rd]}Af.

-49-

/klb, NO-SPREAD INITIAL-CC DEP-u LH LL DEP-

{i, [+rd]}Af/ (µ, C) IO
)a. kw.(liy.yəb) * ***
b. kw.(li.bəb) *! * *
c. (kwi.ləb) *! * *
d. kw.(liy.yub) *! * ***

Candidate 49b has achieved the LL iamb by geminating the final consonant of the base, thus
causing a fatal violation of NO-SPREAD (µ, C). Candidate 49c does not conform to an LH iamb
and is still an iamb of the type LL exactly like the optimal candidate. But it is excluded for
violating the constraint INITIAL-CC. Finally, 49d is eliminated for having made recourse to u-
epenthesis to achieve an LH foot type.

284
 Next, consider inputs whose second segment is a vowel. These include all the forms in 1d
and 1e above. They all show augmentation by the addition of a whole syllable whose nucleus is
the schwa. Two remarks need to be made about these diminutives. First, if the base first segment
is a labial consonant, the diminutive form surfaces with a labialized geminate. The labialization is
the result of the juxtaposition of [w] and the labial consonant of the base and this to avoid
gratuitous violation of the constraint OCP (lab). As to the gemination it is dictated by the
requirement that the diminutive start with two consonants in respecting the constraint INITIAL-
CC. But where does the [w] that causes labialization come from? Is it the realization of the [+rd]
feature-morpheme or is it some other segment that is epenthesized there whenever the language
needs one? Second, if the base second segment is a vowel, it is consistently replaced by the
diminutive [-i-], a case of melodic overwriting (McCarthy and Prince 1990b).
On the basis of items such as [bit], [mus] and [RaS] (cf. the examples in 1d and 1e in
section 2 above) the glide [w] could be argued to be the language default segment that is
epenthesized to satisfy the constraint INITIAL-CC. Given that the base second segment is a
vowel, no form would obey this constraint were the base vowel to be realized. It has been shown
in chapter 5 that the passive participle makes recourse to u-epenthesis to satisfy iambicity. As to
the diminutives, we believe that the /u/ that shows up in the passive participle is the same as the
one that consistently appears in the second position of diminutive forms whose second segment
of the base is a vowel. The only difference is that the default segment /u/ in the PP retains its
moraic status, whereas in the diminutive it loses this moraic status and gets realized as a glide to
serve as an onset to the diminutive vowel if needed. The diminutive forms whose second segment
of the base is a vowel show that ranking INITIAL-CC above DEP-u is justified in that it is this
epenthesis that leads to the satisfaction of INITIAL-CC.
Consider the possible candidates obtained from an input such as /RaS, {i, [+rd]}Af/:

-50-
/RaS, NO-SPREAD INITIAL-CC DEP-u LH LL DEP-IO
{i, [+rd]}Af/ (µ, C)
)a. R.(wiy.yəS) * * ****
b. R.(wi.SəS) *! * * **
c. (Ri.wəS) *! * **
d. R.(wi.yuS) **! ***

285
The base vowel in all the candidates is filled with the diminutive morpheme [-i-]. Candidate 50b
is ruled out for reasons that have become clear now. The candidate in 50c has resorted to u-
epenthesis. However the epenthesized /u/ is placed after the diminutive vowel to serve as on
onset to the schwa syllable, thus causing a fatal violation of INITIAL-CC. Finally the candidate
in 50d is ruled out because it incurs two violations of DEP-u: the first /u/ serves as the onset of
the first syllable of the foot; the second as a nucleus to the heavy syllable of the same foot.
It should be noted that cases like [RwiyyəS], [kwiyyəs] and [γwiyyər], which are derived
from the nouns [RaS] “head”, [kas] “a glass” and [γar] “cave”, respectively, should be treated
differently from the rest of the items. As it has been shown above, the second segment of the base
does not show up; it is systematically replaced by the diminutive vowel. Were the input vowel to
surface, we would end up with forms such as *[Ra.iS] or *[RawiS] which violate INITIAL-CC,
or *[RwawiS] which incurs two violations of DEP-u.
Other diminutives forms that deserve special treatment are output forms such as
[kwiyyəs] and [γwiyyər] which are special in the sense that they present an apparent challenge to
the constraint requiring the labialization of simple dorsal consonants as the tableau in 51 shows:

-51-

MAX-IO [rd] *Kw IDENT-IO [rd]

w
0a. k .wiy.yəs * *
b. k.wiy.yəs *!

Our constraints wrongly predict that the output is 51a rather than 51b. Clearly another constraint
is at play, and it is a constraint that must outrank MAX-IO [rd].
Recall that in chapter 5, we assume that there is an OCP constraint on labials and more
particularly on rounded labials dubbed *RdRd that militates against having output forms such as
*[məʃwuf] instead of [məʃyuf]. It seems that it is this constraint that forces the non-labialization
of [k] in the diminutive [kwiyyəs]. Further support for this assumption comes from items that
have two diminutive forms: one with the labialization of the initial dorsal consonant and a
syllable whose onset is the glide [y]; the other with a non-labialized dorsal consonant and a
second syllable whose onset is the glide [w].

286
-52-

Base Diminutive Unattested Dim. Gloss

Sγir Sγwiyyər / Sγiwwər *Sγwiwwər small

rqiq rqwiyyəq / rqiwwəq *rqwiwwəq thin
qSir qwSiyyər / qSiwwər *qwSiwwər short
γwzal γwziyyəl / γziwwəl *γwziwwəl nice
glil gwliyyəl / gliwwəl *gwliwwəl short

When the initial consonant is labialized, the epenthesized segment is the glide [y]. The facts in 52
are reminiscent of Tashlhit Berber where two labials of the same rank are not allowed in the same
word (Selkirk 1993). This is exemplified by the data below taken from Bensoukas (1999: 15):

-53-

a. Imperative Perfective Gloss

knu kwni bend

gnu gwni sew
xlu xwli become crazy
qlu qwli fry

b. Singular Plural Gloss

taglut tigwla oar

taγrrust tiγwrras a piece (of fish)
agru igwra frog
aγyyul iγwyyal donkey

These examples show that whenever the glide [w] occurs with the vowel [u], the labialized
consonant loses its secondary articulation as a result of the OCP. The same phenomenon takes
place in the CMA data in 52 above.
What is relevant to us here is that the non-labialization of the dorsal consonant in
[kwiyyəs] is the result of a dissimilatory process that bans the occurrence of two round segments
of the same rank, an instance of the OCP. The constraint *RdRd should be allowed to dominate
MAX-IO [rd] to produce the correct output:

287
-54-

*RdRd MAX-IO [rd] *Kw IDENT-IO [rd]

)a. k.wiy.yəs *
b. kw.wiy.yəs *! * *

This example shows that satisfying the OCP constraint *RdRd is more important than satisfying
MAX-IO [rd] by labializing the dorsal consonant.
Before we close this subsection, it is reasonable to ask why in the case of trisegmental
CCC roots, augmentation applies but u-epenthesis does not whereas in CVC roots both
augmentation and u-epenthesis apply? In other words, why isn’t the diminutive of [sdər] realized
as *[swidər] instead of [sdiyyər]? The answer comes from ranking DEP-u above the general
DEP-IO. Since [sdiyyər] does not violate DEP-u while [swidər] does, it follows that the optimal
candidate is the one that satisfies INITIAL-CC with its own consonantal material without ever
having to resort to epenthesis to fill in the second position.
To recapitulate, it has been shown that the diminutive forms in 1a-e proceed to
augmentation by the addition of a whole syllable to avoid the minor iamb. It has also been shown
that augmentation never leads to achieving an ideal iambic foot of the type LH. Instead it tries to
avoid feet that would otherwise surface as minor iambs. Because of higher-ranking DEP-u and
NO-SPREAD (µ, C), the only iambic foot that the diminutive form could achieve is of the type
LL. Further support for our analysis comes from the diminutive forms that achieve the LL iamb
through the suffixation of the feminine morpheme [-a].

6.2 Augmentation as the Feminine Morpheme Suffixation

This subsection will consider items that are totally different from the ones considered
above. Up to this point we have shown that the diminutive form of nouns consisting of a single
major syllable struggles to avoid a minor iambic foot and this by the addition of a whole syllable
to satisfy an LL iamb. There is yet another category of nouns which proceeds to a different kind
of augmentation to satisfy the iamb. These include nouns that are inherently marked for the
feature [feminine] (cf. The items in 1f above):

288
-55-

Base Diminutive Gloss

wdn wdina ear

ʃəmʃ ʃəmʃ sun
zit zwita oil
DaR dwira2 house
bənt bnita / bniyya girl

A look at these items shows that the diminutive has an additional final vowel, marking the
feminine, which is not part of the input. Where did that vowel come from? Given the constraint
system available to us so far, a noun such as [wdən] may have the following candidates in the
diminutive:

-56-

NO-SPREAD INITIAL- *Min-LH LH LL DEP-IO

(µ, C) CC
a. (w.din) *! *
0b. w.(diy.yən) * ***
c. w.(di.nən) *! *! *
d. (wi.dən) *! *! * *

The constraints we have so far seem to make the wrong prediction as to the optimal form. So,
how is it possible to derive the form [wdina] instead of *[wdiyyən]?
Boudlal (1993) has shown that the lexical entries of the items in 1f (i.e. items like those in
55) are different from those in 1a-e in the sense that the former have an inherent specification for
the feminine feature while the latter do not. In other words, the lexical entry of words such as
[wdən] includes among other things, the phonological representation of the word, its syntactic
category, diacritic features such as [feminine] and any idiosyncratic behavior. Words that are not
specified for the feature feminine are simply unmarked (i.e. [ufeminine]) and acquire the [-
feminine] specification by default.

2
Notice here the deemphatization of the the consonants [D] and [R] which is triggered by the diminutive morpheme
[-i-].

289
 Therefore and in order to account for the diminutive of forms such as [wdən], we propose
that the input should also include the feature [+feminine]. It is this feature that calls for the
affixation of the morpheme [-a] to the base to satisfy foot requirements. In order for the feature
[+feminine] to be materialized, we need a constraint that forces the suffix
[-a] to surface in words such as those in 55 (and 1f) above. This constraint could be stated as
follows:

-57-

FEMININE = [-a] (henceforth FEM = [-a])

The feminine morpheme must be realized as [-a] in the output.

Satisfaction of this constraint simply means the realization of the [+feminine] feature as the
vowel [-a]. Note here that this constraint is blind to all the masculine forms since the feature
[+feminine] is not included in their input and that explains why they augment their bases by the
addition of the default syllable. The constraint FEM = [-a] is undominated and as such should be
ranked high in the hierarchy.
In the tableau below we reexamine the two competing candidates [wdina] and [wdiyyən]:

-58-

/wdn, {i, FEM = [-a] *Min-LH LH LL DEP-IO

[+rd], [+fem]} Af/
a. w.(diy.yən) *! * ***
)b. w.(di.na) * *

Candidate 58a loses to candidate 58b because it fails to realize the [+feminine] feature which is
normally associated with words that are inherently feminine.
It should be noted here that the word [bənt] listed among the items in 1f above presents a
special case in the sense that it has two diminutive forms: a regular form obtained by the
affixation of the feminine suffix [-a], giving the output [bnita], and another form which deletes
the final [t] of the base and proceeds to the suffixation of [-a] thus giving the output [bniyya]. The
glide is epenthesized to serve as an onset to a syllable whose head is the feminine suffix. Within a
derivational framework, [bniyya] could be obtained as follows:

290
-59-

Input: /bnt/
Dim morpheme: bnit
Final C deletion bni
Fem. a-suffix bnia
Glide epenthesis bniyya
Output [bniyya]

Leaving aside the deletion of [t] which is really unpredictable, the example above shows that
nouns which are inherently feminine take a final [-a] in order to avoid surfacing with a minor
iamb. This example also shows that the constraint ONSET is observed in the language and thus
never violated.
Closely related to ONSET satisfaction is a class of diminutives whose base ends up in a vowel
(cf. 1g above). In 50 we repeat these examples:

-60-

a. ʕʃa ʕʃiwa
γda γdiwa

b. mRa mRiyya / mRiwa

bRa bRiyya / bRiwa
ʃta ʃtiwa
ʕSa ʕSiyya/ʕSiwa

The common characteristic among all these bases is that they all end up with the vowel [a], which
is part of the base and therefore shouldn’t be confounded with the feminine suffix [-a]. However,
the respective diminutive of these bases is feminine (Compare this with diminutive forms in
Tashlhit which are all feminine). The difference between these items is that the items in 60a are
masculine (i.e. [-feminine]) whereas those in 60b are inherently specified as [+feminine]. The
final [a] in 60b does not add any feature specification for gender.
The items in 60 are different from the ones considered so far in that the diminutive
surfaces as disyllabic in spite of the fact that the base consists of one major syllable, and this
without having to make recourse to augmentation (be it schwa syllable addition or a-suffixation).

291
Given a base form such as /ʕSa/, one should expect the diminutive morpheme [-i-] to be placed
after the second segment of the base in conformity with the constraint INITIAL-CC. This
placement causes the morpheme to be contiguous to the base final vowel, thus creating a hiatus
that the language resolves by epenthesizing a glide or turning a high vowel into a glide to serve as
the onset of the base final vowel.
In the tableau below, we show how [ʕSiyya] wins over two other candidates:

-61-

/ʕSa, INITIAL-CC *Min-LH LH LL DEP-IO

{i, [+rd]}Af/
)a. ʕ.(Siy.ya) * **
b. (ʕəS.ya) *! * *
c. (ʕi.Sa) *! *

The items in 60 above also present a special case in as far as ONSET satisfaction is concerned. It
is known that in the case where a glide needs to be epenthesized, this glide needs to share the
features of rounding with the immediately preceding vowel. If the preceding vowel is [i], the
epenthesized glide is [y]; otherwise it’s [w]. Such is not always the case as the items in 41 show
in spite of the fact that the diminutive form of some items such as [ʕSa] shows an option between
[y] and [w].
To sum up, we have shown that CMA diminutive proceeds to augmentation to avoid a
minor iambic foot by suffixing the feminine morpheme [-a] to bases that are inherently specified
as [+feminine]. We have also shown that this augmentation is not necessary in bases whose final
vowel is [a] which should be distinguished from the feminine suffix [-a]. For these cases we have
argued that iambicity follows from the interaction of constraints, namely the need to satisfy
ONSET, an undominated constraint in the language.

7. UNAUGMENTED DIMINUTIVE FORMS

These include the diminutive forms of disyllabic bases (1h-j) and trisyllabic ones (1k,l).
Consider representative examples given in 62 below:

292
-62-

Base Diminutive Gloss

a. bəlγa bliγa oriental (women’s) slippers

TəbSil TbiSil plate
sarut swirit key
SəBBaT SBiBiT a pair of shoes

b. limuna lwimina an orange

hiDuRa hwiDiRa sheepskin
nəwwaRa nwiwiRa flower

The diminutive forms in 62 do not undergo any augmentation because they already satisfy the
constraint on iambicity and as it has been shown, augmentation is achieved for the sole purpose
of avoiding a minor iamb.
In the constraint tableau 63, we present some of the diminutive forms obtained from the
base [bəlγa]:

-63-

/blγ-a, INITIAL-CC *Min-LH LH LL DEP-IO

{i, [+rd]}Af/
)a. b.(li.γa) *
b. (bil.γa) *! * *
c. b.(liy.γa) * *!

Although candidate 63a cannot achieve the true LH iamb, it is optimal because it conforms to a
LL iamb, something the other candidates could not achieve. In placing the diminutive vowel after
the first segment of the base, candidate 63b has violated the constraint INITIAL-CC as well as
iambicity. Finally candidate 63c has proceeded to glide epenthesis causing a fatal violation of
DEP-IO. However, it should be noted that a form such as [bliyγa] is a possible alternative output
and it does conform to an LL iamb given the fact that the glide here is not moraic as shown in 64:

293
-64-

PWd

Ft

σ σ σ
│ │
µ µ µ
│ │
b l i y γ a

The structure above reflects a phenomenon that relates to regional variation. Generally, rural
varieties epenthesize a glide whereas in the rest of MA varieties, the diminutive form obtained is
identical to the output form in 63 above.
The next class of disyllabic bases we will consider is that of cases whose prefinal segment
is the vowel [i] or [u], i.e. cases that contain what Al Ghadi (1990) calls Class I derivational
affixes. These affixes are systematically replaced by the diminutive vowel [-i-] as a result of
melodic overwriting.
The base/diminutive mapping is shown below for [TəbSi]/[TbiSil]:

-65-

PWd PWd
|
Ft Ft

σ σ σ σ σ
│
µ µµ µ µ µ µ
│ │ │
T əb S l T b S l

i i

The output of the diminutive here is a true LH iamb; the initial minor syllable is adjoined to the
prosodic word since it cannot form a foot on its own as this would constitute a violation of the
constraint FT-BIN. The tableau in 66 presents some candidates obtained from the input /TbSil/:

294
-66-

/TbSil, INITIAL-CC *Min-LH LH LL DEP-IO

{i, [+rd]}Af/
)a. T.(bi.Sil) *
b. (Tib)(Sil) *! * *
c. T.(bi.Səl) *! *

The candidate in 66b is ruled out for violating INITIAL-CC. The placement of the diminutive
morpheme after the first segment of the base has created a situation where we have two
contiguous heavy syllables, thus incurring a violation of both LH and LL. The candidate in 66c is
ruled out because it does not conform to an LH iamb as the optimal candidate does. 66c can also
be ruled out because the diminutive morpheme fails to spread to the final V-position that is
occupied by the [i] in [TəbSil], thus violating MAX-V. Notice further that a potential candidate
such as [TwibSil] is ruled out on the ground that it has resorted to u-epenthesis to satisfy
INITIAL-CC. In the following tableau, we show how [TbiSəl] and [TwibSil] are ruled out in
favor of [TbiSil]. We Assume that MAX-V and DEP-u are not ranked with respect to each other.

-67-

/TbSil, MAX-V DEP-u LH LL DEP-IO

{i, [+rd]}Af/
)a. T.(bi.Sil) *
b. T.(wib)(Sil) *! * *
c. T.(bi.Səl) *! * *

Both candidates 67b and 67c are ruled out for different reasons: 67b is ruled out because it has
resorted to the epenthesis of /u/ (realized as a glide in onset position), thus violating DEP-u; 67b
is ruled out because it has deleted a vowel of the input, causing a fatal violation of MAX-V. In
addition to these two constraints, both candidates can be eliminated on the ground that their
prosodic shape does not conform to an LH iamb.
Similarly disyllabic bases whose prefinal segment is the vowel [u] behave in the same
way as words like [TəbSil]. For illustration, consider the structures of the pair [sarut] / [swirit]
given in 68 below:

295
-68-

PWd PWd
|
Ft Ft

σ σ σ σ σ
│
µ µµ µ µ µ µ
│ │ │
s a r t s w r t

u i

In the diminutive, the base vowel [a] is not realized because of the constraint INITIAL-CC. It has
already been shown that whenever the base second segment is a vowel, it fails to show up in the
output form but its position is filled in with the diminutive vowel. Given that these bases do not
start with a cluster of consonants, recourse is made to the default segment /u/ (realized as the
glide [w]) to satisfy INITIAL-CC.
The final cases of disyllabic diminutives we will consider are those in 1j above where the
diminutive morpheme [-i-] splits up the geminate. Adopting the Two-Root Theory of geminates
(Selkirk 1990, 1991), a word such as [SəBBaT] has the representation in 69 below:

-69-
σ σ

µ µ µ

RC RC

S ə B a T

In the following tableau, we consider some possible candidates from the input /SBBaT/. To rule
out any candidate deleting a root consonant, we need the constraint MAX-RC which ensures that
all root consonants of the input appear in the output. This constraint along with MAX-V need to
be undominated:

296
-70-

/SBBaT, MAX-V MAX-RC LH LL DEP-IO

{i, [+rd]}Af/
)a. S.(Bi.BiT) *
b. S.(Biy.yiT) *! * **
c. S.(Bi.BəT) *! * *

The candidate in 70b has resorted to glide epenthesis at the expense of a root segment, causing
fatal violation of MAX-RC. The candidate in 70c is also ruled out because the base prefinal
vowel-position, which is supposed to be filled out by the diminutive [-i-], is deleted, thus causing
violation of MAX-V.
Trisyllabic bases do not need any other constraints; their diminutive forms can be
obtained much in the same way as those of disyllabic ones. The only difference is that the output
of the diminutive of trisyllabic bases consists of a minor LL iamb followed by an LL iamb as
could be seen in 71 below:

-71-
PWd

Ft Ft

σ σ σ σ
│
µ µ µ µ
│ │ │ │
l w i m i n a

The output in 71 differs from any other output considered so far. We have shown that all the
augmented forms correspond to an iamb of the type LL preceded by a minor syllable which
cannot form a foot on its own given the constraint FT-BIN. In disyllabic bases, the diminutive
may consist of an LL iamb preceded by a minor syllable or an LH iamb preceded by a minor
syllable depending on whether or not the input contains one of the affixes [-i-] and [-u-] found in
[TəbSil] and [sarut], respectively. The diminutive of trisyllabic bases is consistently an LL iamb
preceded by a minor LL iamb where the first light syllable is minor. Trying to decrease the
number of syllables in a word such as [limuna] would result in the diminutive form [lmina] which

297
incurs a fatal violation of MAX-V. This prohibition against deletion constitutes strong evidence
for the undominated nature of MAX-V.
The tableau below gives some candidates obtained from the base [limuna]:

-72-

/limun-a, MAX-V INITIAL- DEP-u LH LL DEP-IO

{i, [+rd]}Af/ CC
)a. (l.wi)(mi.na) * *
b. li.(mi.na) *! *
c. l.(wim.na) *! * * *
d. l.(mi.na) *!

Candidate 72b is ruled out because it violates INITIAL-CC. The one in 72c is also ruled out
because the spreading of the diminutive affix, which is supposed to fill the position that used to
be occupied by the base vowel [u], fails to apply, thus leading to the violation of MAX-V and LH
and LL iambs. Finally, even if candidate 72d conforms to an iamb of the type LL, it is eliminated
because it incurs a fatal violation of MAX-V.
Trisyllabic bases consisting of a geminate behave like disyllabic ones with regard to
diminutive formation. Given the Two-Root Theory of geminates mentioned above, the
diminutive vowel is expected to be placed between the first and the second part of the geminate
and spread to any other position to its right.
To recapitulate, the constraints needed to account for the diminutive in CMA are ranked
in 73 below:

298
-73-

MAX-V MAX-RC INITIAL-CC FEM = [-a] NO-SPREAD

(µ, C)

DEP-u

*Min-LH

LH

LL

DEP-IO

Note that these constraints need to be complemented by the markedness and faithfulness
constraints presented in 38 above and which are responsible for the labialization or non-
labialization of initial consonants in the diminutive form.
To sum up, disyllabic and trisyllabic bases present further evidence for the interaction of
prosodic and faithfulness constraints. Disyllabic cases have been found to conform to a foot of
the type LH or LL with the initial syllabic consonant lying outside the iamb. As to trisyllabic
cases, they all consist of a sequence of two LL iamb with the first syllable being minor. In both
disyllabic and trisyllabic cases, it has been shown that the diminutive forms do not resort to
augmentation because they already meet the required shape.

8. CONCLUSION
In this chapter, we have provided an OT account of diminutives in CMA. We have shown
that in order to provide an adequate account of this morphological class, reference has to be made

299
to prosody and its interaction with other constraints in the grammar. In particular, we have shown
that the diminutive abides by a prosodic constraint which requires that the output conform to an
iambic foot of the type LL or in the ideal cases LH. In order to achieve this prosodic requirement,
certain forms augment their bases. The augmented cases considered are of two types: those that
proceed by the addition of the language default syllable for masculine nouns, and those that
suffix the feminine morpheme [-a] to nouns which are inherently specified as [+feminine]. In
both cases, we have argued that augmentation never leads to the most harmonic LH iamb;
instead, it is undertaken in order to avoid a minor LH.
We have also considered the process of labialization in its relation with the diminutive.
We have shown that only dorsal consonants and labial geminates occurring in initial position are
labialized and that their labialization follows from the constraint requiring the alignment of a
labialized consonant with the left edge of the prosodic word and the need to satisfy MAX-IO [rd].
We have argued that the underapplication of labialization to the rest of the consonants (including
simple labial consonants) results from the interaction of markedness constraints on simple
labialized consonants and faithfulness constraints such as MAX-IO [rd] and IDENT-IO [rd]. We
have shown that the failure of these consonants to labialize follows from ranking the constraints
prohibiting labialized coronals, pharyngeals and simple labials above MAX-IO [rd] and below
the constraint prohibiting labialized dorsals and geminate labials. Also, we have shown that the
gemination of the labial consonants follows from the combination of INITIAL-CC, the OCP (lab)
and markedness constraints on geminated labials.

300
 Conclusion

Throughout this dissertation, we have tried to show that the framework of OT and CT is the most
appropriate for the treatment of syllable structure and stress as cases related to CMA prosodic
phonology, and the treatment of the nisba adjective, the causative, the passive participle and the
diminutive as cases related to CMA prosodic morphology. It has been argued that this framework
allows for a better understanding of these cases in terms of the interaction of constraints
pertaining to Universal Grammar and ranked on a language-specific basis. Constraint interaction
takes the form of conflict. It has been shown that lower-ranked constraints are allowed to be
violated for the sole purpose of securing higher order constraints which determine the optimal
shape of the output.
In dealing with CMA syllable structure, we have established a distinction between a
minor syllable, which consists of a moraic consonant, and a major syllable, whose nucleus is the
schwa or one of the full vowels of the language. Minor syllables arise to avoid fatal violations of
the constraint *COMPLEX, prohibiting branching onsets and codas. Granting a syllabic status to
minor syllables allows for the satisfaction of FT-BIN, particularly in nonderived trisegmental
lexical words having the pattern CCV, CCəC and CəCC, and where the first consonant of the
initial cluster in CCV, CCəC and the second consonant of the final cluster in CəCC are associated
to a minor syllable. Furthermore, the analysis offered in this work has enabled us to come up with
an explanation to the problematic cases of schwa epenthesis while still maintaining our
predecessors’ claim that the schwa in nouns, but not verbs or adjective, is dependent on the
sonority of the surrounding consonants. In particular, it has been shown that the placement of the
schwa before the final consonant in nonderived trisegmental verbs and adjectives derives from
the constraint ALIGN-R (Vb/Adj, σ′), requiring stem-prominent syllable right-alignment, thus
giving rise to an iambic foot type. This verb- and adjective-specific alignment constraint must
rank higher than the general stem-prominent syllable right-alignment constraint ALIGN-R-σ′,
needed to account for iambicity in nouns, verbs and adjectives. In both cases, a minor syllable
can never appear in a prominent position because of the undominated constraint *Min-σ′. In
order to account for nominal schwa syllabification, we have had recourse to a set of markedness

301
constraints favoring schwa syllables with a higher sonority coda. Because these markedness
constraints are noun-specific, they must rank higher than the general stem-prominent syllable
right-alignment constraint in order to account for nominal cases on the pattern CəCC, where the
schwa is epenthesized before the second consonant, thus leading to a noun whose right syllable is
minor.
Another issue related to syllable structure concerns cyclic syllabification, which
necessitates recourse to a type of faithfulness involving two output forms. It has been shown that
cyclic syllabification is not warranted on both theoretical and empirical grounds. Theoretically, a
cyclic account is operational in that an input form has to pass through different intermediate
stages before reaching the final stage of phonetic realization. Empirically, a cyclic account of
syllabification makes the wrong predictions in certain cases involving affixation of an object
clitic to the verb stem. The analysis that has been proposed in this work is more powerful in that
it derives the effect of cyclic syllabification without having to refer to intermediate stages, and
this by a set of O-O constraints much in the spirit proposed in Basri et al. (1998) and Selkirk
(1999). In order to account for the asymmetry between cases of affixation to the stem and
affixation to the word exhibited in forms such as [DRəbt] and [DəRbək], we have proposed two
different types of O-O faithfulness to prosodic edges: O-Ostem ANCHOR (σ, σ, Initial) and O-
Oword ANCHOR (σ, σ, Initial). We have shown that O-Ostem ANCHOR (σ, σ, Initial) must
dominate O-Oword ANCHOR (σ, σ, Initial) in order to account for faithfulness to the base stem
[D.Rəb] in the derived output [D.Rəb.t], which does not epenthesize a schwa before the subject
suffix, and lack of faithfulness to the stem exhibited by the output form [DəR.bək], which
epenthesizes a schwa before the object suffix. Finally, we have further shown that the constraint
O-Ostem ANCHOR (σ, σ, Initial) is never violated except when the higher order markedness
constraints ONSET and *COMPLEX are at stake.
The model of O-O correspondence adopted in this work has been tested to see if it can
account for the formation of the causative and a class of nisba adjectives whose bases undergo
truncation. We have argued that the truncation witnessed in nisba adjectives derived from
compound nouns and nouns with the affixes [(ʔ)a-] and [ta-...-t] is not prosodically motivated and
therefore does not require recourse to an O-O constraint relating a base and the truncated form or
any other constraint limiting the prosodic size of its output. We have instead argued that in order
to explain the truncation seen with these nisba adjectives, we need both markedness constraints

302
and affixational ones. In particular, we have shown that truncation follows from ranking *[N +
Adj]PWd along with the affixation constraint AFFIX-TO-Nstem and the alignment constraint
ALIGN-Nisba-R above the constraint MORPH-REAL, thus forcing deletion of the leftmost stem
in compounds and the nominal affixes [ta-...-t] and [ʔa-] in nouns designating localities. As to the
causative form, we have shown that the analysis appealing to prosodic faithfulness to some
designated syllable edge is inadequate in that it fails to account for bases on the pattern /CVC/
and also to block long distance consonantal spreading, whereby it is the third and not the second
segment of trisegmental bases which is geminated. We have instead proposed, following Imouzaz
(forthcoming), to analyze the causative in terms of O-O correspondence constraints relating a
base and its reduplicant. We have also argued for the prosodic nature of the causative by showing
that the output always conforms to an iambic foot consisting of a sequence of two light syllables,
a fact which blocks total reduplication.
In order to provide a thorough understanding of the stress system of CMA, we have
judged it necessary to undertake two experiments: one quantitative; the other instrumental. The
quantitative experiment has allowed us to quantify the results of native speakers’ intuitions about
the placement of stress. The instrumental experiment, in which we have considered both words
in isolation and words in context, has allowed us to see to what extent the results obtained from
the quantitative experiment are reliable. The results obtained from the instrumental test about
words in isolation confirm to a large degree those obtained from the quantitative test, namely that
the language is quantity-sensitive with stress on the final syllable if it is heavy or else on the
penultimate. The instrumental test has also revealed that in context words, stress is consistently
on the final syllable. Having discovered that the stress system of CMA is both iambic and
trochaic, we have been faced with the intricate question of finding an appropriate way to solve
this puzzle. The OT framework provides the appropriate tools for solving this puzzle. We have
offered a unitary OT account based on the idea that iambic feet take priority over trochaic ones,
and this by ranking the constraint IAMB above TROCHEE. We have shown that the location of
stress and consequently the foot types depend on the organization of prosodic words into
phonological phrases. In a phrase with a single prosodic word (i.e. a word in isolation), the foot
type that surfaces as optimal is trochaic with stress on the final or penultimate syllable.
Penultimate stress is derived by positing the constraint NON-FINALTY

303
(σ′, PPh), requiring that the prominent syllable be non-final within a phonological phrase. In a
phrase with more than one member, one word will be final and the others non-final surfacing
with an iambic foot. In both isolation and context cases, it has been shown that restricting stress
to the last two syllables of a word results from the undominated constraint ALIGN-R (Ft′, PWd),
demanding right-alignment of the PWd and the prominent foot. Because this constraint is
undominated, we have been led to recognize a trochaic foot of the type HL which violates RH-
HARM and an iambic foot of the type L which violates FT-BIN. Both types occur word-finally.
The fact that the stress system of CMA is basically iambic is justified in the prosodic
morphology of the PP and the diminutive. The idea defended in chapters five and six is that these
two morphological categories are governed by prosodic constraints requiring that the output
conform to an iambic foot of the type LH or LL. Assuming that the PP marker is the prefix [m-]
and not the discontinuous morpheme [m-...-u-...], we have shown that the PP forms fall into two
classes, based on whether or not they undergo the epenthesis of the lexical default segment /u/. In
particular, we have argued that augmentation by u-epenthesis is a side effect of the requirement
that the output conform to an LH iamb. The forms that resort to augmentation include the class of
non-derived trisegmental verbs with the exception of verbs whose final segment is a vocoid. The
foot structure of these forms still conforms to an iambic foot but of the type LL. We have argued
that augmentation to achieve an LH iamb in these exceptional forms results in fatal violations of
the higher-ranked constraints DEP-C or NO-LONG-V or else IDENT-IO [cons]. Other classes
that do not show augmentation include the PP of quadrisegmental and derived trisegmental verbs.
We have shown that if augmentation were to apply in these forms, the resulting output would
violate the O-O constraints O-Ostem ANCHOR (Ft, Ft, Initial), requiring positional faithfulness of
the initial segment of the base foot in the derived output, and O-Ostem IDENT-σ, demanding
conservation of weight identity between two output stems.
The diminutive provides further support for the iamb-based analysis proposed for the PP.
In dealing with this morphological category, we have provided arguments calling for the need to
incorporate in the grammar of CMA the constraint INITIAL-CC which requires a PWd to start
with a cluster of two consonants. This constraint, which interacts with the constraints calling for
an iambic output, has allowed us to explain not only the necessity for placing the diminutive
morpheme after the second segment of the base, but also to explain the gemination of the initial
labial consonant of some bases. It has been shown that bases which consist of one major syllable

304
proceed to augmentation in the diminutive to achieve an iambic foot consisting of a sequence of
two syllables, preceded by a minor syllable which is directly adjoined to the PWd. The
augmented cases we have considered are of two types: those that proceed to the addition of a
schwa syllable if the base is masculine, and those that suffix the feminine morpheme [-a] to bases
that are inherently specified as feminine. In both cases, we have shown that augmentation takes
place in order to avoid diminutive cases that surface with a minor LH iamb, where the light
syllable is a minor syllable. The analysis has also considered diminutive cases derived from
disyllabic and trisyllabic bases. These forms do not resort to augmentation because they already
meet the required prosodic shape. The diminutive forms derived from trisyllabic bases surface
with a PWd consisting of a sequence of two iambs. For these we have argued that decreasing the
number of syllables by deleting a vocalic element from the base would lead to a fatal violation of
undominated MAX-V.
Closely related to the formation of the diminutive is the process of labialization which
affects any dorsal or geminate labial consonants occurring word-initially. It has been shown the
round-feature morpheme, responsible for labialization, is part of the diminutive morpheme and
that its realization depends on the interaction of markedness constraints on labialized consonants
and the faithfulness constraints MAX-IO [rd], calling for the realization of this morpheme along
with IDENT-IO [rd] demanding the preservation of featural identity of the input segments in the
input-output mapping. It has also been shown that the failure of consonants other than dorsals and
geminate labials to labialize is the result of ranking the markedness constraints prohibiting
labialized coronal, pharyngeal and simple labial consonants above MAX-IO [rd] and below the
constraints on labialized dorsal and geminate labial consonants. The gemination of a word-initial
labial followed by [w] has been shown to be derived from the combination of markedness
constraints on labialized geminates, the OCP (lab) and INITIAL-CC.

305
 Appendix A

The Constraints Listed in the Dissertation

The constraints needed to account for the prosodic phonological and morphological
aspects of Casablanca Moroccan Arabic dealt with in this dissertation are listed below:

1. AFFIX-TO-Nstem
The nisba morpheme is suffixed to a noun stem.

2. ALIGN-E (Root, PrWd)

The left edge of the root must correspond to the left edge of the PWd and the right
edge of the root must correspond to the right edge of the PrWd.

3. ALIGN-FT-R
The right edge of every foot must be aligned with the right edge of the prosodic
word.

4. ALIGN-Nisba-R
The right edge of the nisba affix must be aligned with the right edge of the PWd.

5. ALIGN-R
The right edge of the root must be aligned with the right edge of the the syllable.

6. ALIGN-R (Ft′, PWd)

The right edge of the prominent foot must be aligned with the right edge of the
PWd.

7. ALIGN-R-σ′
The right edge of the stem must be aligned with the right edge of the prominent
syllable.

8. ALIGN-R (verb/adjective, σ′) (ALIGN-R (Vb/Adj, σ′)

The right edge of the verb/adjective stem must be aligned with the right edge of
the prominent syllable.

9. ANCHOR (Base, L, RED, L)

The left edge of the base must correspond to the left edge of the reduplicant.

10. *COMPLEX
Codas and onsets must not branch.
11. CONSTRAINTS ON LABIALIZED CONSONANTS
a. *Tw, *Hw, *Bw >>*Kw
b. *TTw, *HHw, *KKw >> *BBw

12. DEP-IO
Every segment of the output has a correspondent in the input.

13. FEM = [-a]

The feminine morpheme must be realized in the output as [-a].

14. FT-BIN
Feet are binary under syllabic or moraic analysis.

15. *GEM-σ1
Geminates are prohibited in the first syllable.

16. *GEM
Geminates are prohibited.

17. H-NUC
A higher sonority nucleus is more harmonic than one with lower sonority.

18. IAMB
Feet are right-headed.

19. IDENT-BR [cons]

The base featural specification for [cons] must be preserved in the reduplicant.

20. IDENT-IO [cons]

Featural specification for [cons] must be preserved in the input/ output mapping.

21. IDENT-IO [rd]

Identity of the feature [round] must be preserved in the input/output.

22. INITIAL-CC
Words must begin with two consonants.

23. LABIALIZE (C, L, PWd, L)

The left edge of the prosodic word must be aligned with a labialized consonant.

24. MAX-IO:
Every segment of the input has a correspondent in the output.

25. MAX-IO [rd]

The feature [round] must be preserved in the input/output mapping.

307
26. MAX-Rt-BR
Every root segment of the base has a correspondent in the reduplicant.

27. *Min-LH
A LH foot whose light syllable is minor is prohibited.

28. *Min-σ
Minor syllables are prohibited.

29. *Min-σ′
Prominent minor syllables are prohibited.
30. MORPH-REAL
An input morpheme must be realized in the output.

31. NO-CODA:
syllables must not have a coda.

32. NON-FINALTY (σ′, PPh)

The main-stressed syllable of a word is not in phrase final position.

33. NO-LONG-V
Long vowels are prohibited.

34. NO-SPLITTING
Epenthesis cannot apply to geminates.

35. NO-SPREAD (µ, C)

The spreading of a consonant is prohibited.

36. *[Noun + Adjective]PWd

A sequence of a noun and an adjective is prohibited.

37. ONSET
Syllables must have an onset.

38. O-OStem ANCHOR (Ft, Ft, Initial)

A foot initial segment in the affiliate output form must correspond to a foot initial
segment in the base output form.

39. O-Ostem ANCHOR (σ, σ, Initial)

Where two strings S1 and S2 are in an O-Ostem correspondence relation and S1 is
the base and S2 the affiliate of that correspondence relation, a syllable-initial
segment belonging to S2 must correspond to a syllable-initial segment belonging
to S1.

308
40. O-Oword ANCHOR (σ, σ, Initial)
Where two strings S1 and S2 are in an O-Oword correspondence relation and S1 is
the base and S2 the affiliate of that correspondence relation, a syllable-initial
segment belonging to S2 must correspond to a syllable-initial segment belonging
to S1.

41. O-OStem IDENT-σ

Light/heavy syllables in the affiliate output form must correspond to light/heavy
syllables in the base output.

42. PARSE-σ
Syllables must belong to a foot.

43. PARSE-seg
Every segment must belong to a syllable.

44. *RdRd
Sequences of round segments are prohibited.

45. SONORITY (in nouns)

*µ >> *µ >> *µ >> *µ >> *µ

ə S ə F ə N ə L ə G

46. TROCHEE
Feet are left-headed.

47. VERB ROOT = [µ µ]

A verb root must correspond to two moras.

48. WSP
A heavy syllable is stressed in foot structure.

309
 Appendix B

The Quantitative Test: The Questionnaire

This questionnaire is undertaken in the framework of a research study whose objective is

to determine some phonological aspects of Casablanca Moroccan Arabic.

Please respond to the questions below. Thank you for your contribution.

A. FIRST NAME:

B. LAST NAME:

C. SEX: † Male † Female

D. DATE OF BIRTH:

E. PLACE OF BIRTH:

F. OCCUPATION:

G. HAVE YOU TAKEN ANY LINGUISTICS COURSES?

† Yes † No

H. IF YES, FOR HOW LONG?

I. DO YOU SPEAK BERBER?

†. Yes † No

J. PLACE STRESS (′) ON THE FOLLOWING ITEMS.

The symbol [y] stands for the palatal glide (IPA [j]). All capitalized symbols stand
for emphatic sounds. The rest of the symbols adopted are those of the IPA.Syllable
boundaries are marked by a period.

i. DISYLLABIC WORDS

1. law.yin wilted (pl.)

2. kal.kum he ate you (pl.)
3. məl.yun one million
4. Drəb.hum he hit them

310
5. wal.mək it (mas.) suits you
6. naw.ya intending (fem.)
7. bab.ha her/its (fem.) door
8. li.mun oranges
9. da.ruh they did it (mas.)
10. li.kum for/to you
11. məl.məl he shook (sth.)
12. wəl.dək your (sg.) son
13. ma.yəl bent (mas.)
14. ʕu.dək your (sg.) stick
15. Rəm.la sand
16. Səm.na we fasted
17. wa.lu nothing
18. Da.Ru his house
19. li.na for/to us

ii. TRISYLLABIC WORDS

20. law.yin.hum they are twisting them

21. man.Da.Rin clementine
22. wal.mu.kum they fit you (pl.)
23. bəZ.Tam.kum your (pl.) wallet
24. mər.məd.nak we trailed you (in dust)
25. ru.bəl.kum he disturbed you (pl.)
26. ʕəR.ga.nin they are sweaty
27. mqul.bi.nək they are deceiving you
28. wəl.dat.ni she gave birth to me
29. min.ʒa.ra sharpener
30. ban.ya.ha she built it (fem.)
31. rub.lu.ni they disturbed me

311
32. gən.Du.Ra a Moroccan gown
33. yəD.Rəb.na he hits us
34. məl.mlu.na they shook us
35. li.mu.na an orange
36. ʕi.ni.na our eyes

iii. POLYSYLLABIC WORDS

37. ban.ya.ha.lih she is building it (fem.) for him

38. di.ri.ha.li.ha do it (fem.) for her
39. gəl.bu.ha.li.kum they reversed it (fem.) for you (pl.)

312
 Appendix C

The Instrumental Test: The Data

1. lawyin wilted (pl.)

2. kalkum he ate you (pl.)
3. məlyun one million
4. Drəbhum he hit them
5. walmək it (mas.) suits you
6. babha her/its (fem.) door
7. limun oranges
8. daruh they did it (mas.)
9. məlməl he shook (sth.)
10. wəldək your (sg.) son
11. mayəl bent (mas.)
12. Rəmla sand
13. walu nothing
14. lina for/to us
15. lawyinhum they are twisting them
16. manDaRin clementine
17. mərmədnak we trailed you (in dust)
18. ʕəRganin they are sweaty
19. mqulbinək they are deceiving you
20. wəldatni she gave birth to me
21. banyaha she built it (fem.)
22. gənDuRa a Moroccan gown
23. yəDRəbna he hits us
24. məlmluna they shook us
25. limuna an orange

313
26. banyahalih she is building it (fem.) for him
27. dirihaliha do it (fem.) for her
28. gəlbuhalikum they reversed it (fem.) for you (pl.)

314
 Appendix D

Scores of Subjects in Test Items in Isolation

The charts below give the values for fundamental frequency (F0), intensity (IT) and
duration (D) of the 28 test items in Appendix C as pronounced in isolation by four
subjects. The first three subjects are males; the lat is a female.

1. lawyin

F0 law yin IT law yin D law yin

S1 124 140 S1 32 30 S1 213 289
S2 120 128 S2 37 37 S2 253 354
S3 130 147 S3 29 23 S3 240 380
S4 206 225 S4 32 31 S4 270 316

2. kalkum

F0 kal kum IT kal kum D kal kum

S1 128 93 S1 37 26 S1 304 187
S2 141 121 S2 33 27 S2 213 236
S3 167 117 S3 28 24 S3 262 255
S4 240 242 S4 31 29 S4 289 278

3. məlyun

F0 məl yun IT məl yun D məl yun

S1 104 128 S1 28 38 S1 156 347
S2 114 134 S2 31 25 S2 194 335
S3 126 138 S3 25 28 S3 213 337
S4 208 229 S4 23 30 S4 156 327

315
4. DRəbhum

F0 DRəb hum IT DRəb hum D DRəb hum

S1 142 108 S1 35 30 S1 261 195
S2 147 117 S2 31 25 S2 240 255
S3 166 124 S3 30 22 S3 274 147
S4 261 202 S4 30 27 S4 251 209

5. walmək

F0 wal mək IT wal mək D wal mək

S1 139 112 S1 36 18 S1 261 234
S2 142 105 S2 30 18 S2 262 274
S3 155 130 S3 27 19 S3 228 175
S4 244 178 S4 29 24 S4 262 289

6. babha

F0 bab ha IT bab ha D bab ha

S1 137 105 S1 36 21 S1 316 163
S2 134 92 S2 36 22 S2 346 140
S3 146 124 S3 28 17 S3 304 133
S4 236 183 S4 31 18 S4 281 175

7. limun

F0 li mun IT li mun D li mun

S1 114 142 S1 32 34 S1 152 335
S2 112 143 S2 32 32 S2 160 350
S3 129 145 S3 24 25 S3 171 388
S4 202 234 S4 28 27 S4 216 330

8. daruh

F0 da ruh IT da ruh D da ruh

S1 118 148 S1 34 32 S1 220 316
S2 125 144 S2 36 31 S2 213 262
S3 128 148 S3 27 28 S3 171 323
S4 203 234 S4 30 33 S4 228 323

316
9. məlməl

F0 məl məl IT məl məl D məl məl

S1 139 104 S1 34 32 S1 179 243
S2 156 96 S2 33 29 S2 194 217
S3 152 117 S3 28 22 S3 232 207
S4 286 190 S4 33 35 S4 228 278

10. wəldək

F0 wəl dək IT wəl dək D wəl dək

S1 147 116 S1 36 34 S1 209 228
S2 156 115 S2 34 28 S2 240 236
S3 166 127 S3 26 23 S3 252 162
S4 256 202 S4 30 32 S4 259 240

11. mayəl

F0 ma yəl IT ma yəl D ma yəl

S1 137 106 S1 35 22 S1 240 201
S2 142 93 S2 35 23 S2 327 186
S3 144 115 S3 26 23 S3 203 227
S4 212 192 S4 27 29 S4 194 144

12. Rəmla

F0 Rəm la IT Rəm la D Rəm la

S1 144 113 S1 34 32 S1 232 198
S2 144 85 S2 33 28 S2 190 144
S3 174 123 S3 27 21 S3 196 154
S4 265 179 S4 27 21 S4 220 144

13. walu

F0 wa lu IT wa lu D wa lu
S1 145 113 S1 35 20 S1 266 129
S2 138 98 S2 31 21 S2 251 156
S3 134 114 S3 29 24 S3 195 143
S4 212 182 S4 28 22 S4 248 124

317
14. lina

F0 li na IT li na D li na
S1 142 106 S1 35 26 S1 209 190
S2 147 112 S2 34 27 S2 228 160
S3 159 115 S3 26 19 S3 186 158
S4 267 170 S4 34 29 S4 217 152

15. lawyinhum

F0 law yin hum IT law yin hum D law yin hum

S1 112 152 115 S1 31 36 29 S1 187 237 218
S2 121 147 101 S2 36 30 24 S2 247 205 213
S3 136 164 121 S3 28 25 19 S3 209 255 224
S4 189 248 192 S4 26 26 20 S4 236 240 209

16. manDaRin

F0 man Da Rin IT man Da Rin D man Da Rin

S1 109 119 147 S1 28 34 33 S1 175 191 288
S2 114 118 140 S2 34 32 21 S2 241 144 245
S3 129 129 131 S3 23 28 26 S3 158 165 297
S4 212 204 214 S4 28 30 30 S4 217 148 274

17. mərmədnak

F0 mər məd nak IT mər məd nak D mər məd nak

S1 116 122 148 S1 31 31 34 S1 109 171 409
S2 111 113 130 S2 31 29 31 S2 175 132 386
S3 139 134 155 S3 26 26 26 S3 162 190 256
S4 204 203 211 S4 25 27 28 S4 163 213 323

18. ʕəRganin

F0 ʕəR ga nin IT ʕəR ga nin D ʕəR ga nin

S1 111 121 140 S1 29 34 30 S1 171 249 288
S2 112 124 133 S2 29 34 25 S2 209 179 259
S3 135 146 149 S3 24 25 24 S3 217 171 300
S4 209 195 275 S4 28 31 26 S4 186 186 300

318
19. mqulbinək

F0 mqul bi nək IT mqul bi nək D mqul bi nək

S1 123 147 107 S1 35 32 24 S1 253 218 253
S2 119 138 102 S2 36 26 22 S2 354 140 226
S3 142 174 161 S3 29 25 21 S3 396 179 160
S4 222 227 196 S4 31 28 23 S4 373 190 236

20. wəldatni

F0 wəl dat ni IT wəl dat ni D wəl dat ni

S1 118 149 109 S1 30 36 21 S1 202 276 159
S2 111 133 105 S2 33 34 19 S2 190 259 190
S3 139 153 121 S3 25 28 22 S3 234 285 127
S4 214 250 201 S4 28 35 25 S4 190 285 167

21. banyaha

F0 ban ya ha IT ban ya ha D ban ya ha

S1 116 145 113 S1 32 33 23 S1 226 245 159
S2 114 133 96 S2 35 27 19 S2 210 237 163
S3 135 154 138 S3 28 27 19 S3 233 175 167
S4 206 231 215 S4 27 26 17 S4 255 194 133

22. gənDuRa

F0 gən Du Ra IT gən Du Ra D gən Du Ra

S1 116 145 117 S1 30 35 26 S1 186 236 144
S2 118 139 134 S2 35 29 27 S2 266 217 127
S3 139 152 148 S3 24 31 25 S3 228 194 110
S4 211 222 190 S4 28 29 21 S4 266 217 127

23. yəDRəbna

F0 yəD Rəb na IT yəD Rəb na D yəD Rəb na

S1 119 148 122 S1 28 35 27 S1 182 228 186
S2 117 144 98 S2 30 32 25 S2 191 152 163
S3 132 164 111 S3 22 26 20 S3 191 196 122
S4 213 238 194 S4 22 25 19 S4 190 137 133

319
24. məlmluna

F0 məl mlu na IT məl mlu na D məl mlu na

S1 120 154 109 S1 30 35 27 S1 175 315 152
S2 115 146 98 S2 32 30 21 S2 202 269 136
S3 139 159 130 S3 23 25 18 S3 152 278 137
S4 211 239 188 S4 26 30 23 S4 144 251 114

25. limuna

F0 li mu na IT li mu na D li mu na
S1 123 165 112 S1 23 35 26 S1 156 228 194
S2 117 140 93 S2 33 33 24 S2 183 167 152
S3 141 158 127 S3 23 26 23 S3 171 214 122
S4 210 224 176 S4 28 26 21 S4 152 179 110

26. banyahalih

a. Fundamental Frequency
F0 ban ya ha lih
S1 118 142 129 148
S2 123 124 122 144
S3 138 146 141 150
S4 209 237 195 202

b. Intensity
IT ban ya ha lih
S1 30 33 30 29
S2 34 25 25 25
S3 28 24 22 21
S4 25 28 23 25

c. Duration
D ban ya ha lih
S1 167 175 167 296
S2 288 124 144 296
S3 285 137 171 323
S4 237 171 163 276

320
27. dirihaliha

a. Fundamental Frequency
F0 di ri ha li ha
S1 137 134 132 148 113
S2 133 135 133 148 120
S3 140 141 140 161 139
S4 223 236 223 229 184

b. Intensity
IT di ri ha li ha
S1 33 34 32 31 23
S2 38 36 34 29 25
S3 26 22 22 25 18
S4 34 37 30 34 20

c. Duration
D di ri ha li ha
S1 163 113 159 218 171
S2 230 120 120 136 171
S3 243 118 140 182 133
S4 144 125 148 156 167

28. gəlbuhalikum

a. Fundamental Frequency
F0 gəl bu ha li kum
S1 119 131 125 131 90
S2 125 132 131 139 96
S3 142 160 157 158 65
S4 205 204 190 213 199

b. Intensity
IT gəl bu ha li kum
S1 29 35 30 27 20
S2 31 34 26 23 21
S3 26 25 24 25 21
S4 28 28 26 23 16

321
 c. Duration
D gəl bu ha li kum
S1 167 144 167 163 167
S2 144 132 132 159 210
S3 125 148 133 144 217
S4 167 156 140 175 190

322
 Appendix E

Pitch Tracks of Selected Test Items in Isolation

5. [walmək] “it (masc.) suits you”

323
7. [limun] “Oranges”

324
9. [məlməl] “He shook (sth.)

325
11. [mayəl] “Bent (masc.)

326
15. [lawyinhum] “They are twisting them”

327
16. [manDaRin] “Clementine”

328
25. [limuna] “An orange”

329
39. [gəlbuhalikum] “They reversed it (fem.) for you (pl.)”

330
 Appendix F

Scores of Subjects in Test Items in Context

The charts below give the values for F0, I and duration D of the 28 items in Appendix C
as pronounced by 5 subjects in the middle of the sentence: [gal [target word] ʒuʒ
məRRat] “he said ...... twice.” The first three subjects are males; the lat two are females.

1. lawyin

F0 law yin IT law yin D law yin

S1 120 164 S1 28 24 S1 247 283
S2 131 176 S2 35 27 S2 317 347
S3 122 170 S3 30 28 S3 225 312
S4 201 268 S4 27 25 S4 256 320
S5 204 260 S5 36 32 S5 286 340

2. kalkum

F0 kal kum IT kal kum D kal kum

S1 119 140 S1 31 26 S1 244 255
S2 131 165 S2 31 28 S2 263 259
S3 124 136 S3 29 27 S3 250 246
S4 208 262 S4 34 29 S4 237 312
S5 207 270 S5 34 31 S5 258 261

3. məlyun

F0 məl yun IT məl yun D məl yun

S1 117 155 S1 25 25 S1 177 375
S2 126 165 S2 26 26 S2 190 333
S3 122 159 S3 27 25 S3 189 313
S4 202 270 S4 31 31 S4 196 367
S5 200 264 S5 25 34 S5 196 349

331
4. DRəbhum

F0 DRəb hum IT DRəb hum D DRəb hum

S1 131 162 S1 28 27 S1 180 299
S2 138 157 S2 32 28 S2 243 339
S3 132 159 S3 30 29 S3 213 320
S4 220 257 S4 31 30 S4 292 205
S5 193 231 S5 33 34 S5 225 334

5. walmək

F0 wal mək IT wal mək D wal mək

S1 127 170 S1 29 23 S1 258 233
S2 139 179 S2 32 25 S2 336 250
S3 132 178 S3 30 24 S3 279 240
S4 207 264 S4 38 33 S4 285 269
S5 205 233 S5 35 31 S5 293 270

6. babha

F0 bab ha IT bab ha D bab ha

S1 118 161 S1 30 30 S1 249 193
S2 139 165 S2 32 30 S2 366 230
S3 129 164 S3 31 30 S3 302 280
S4 208 256 S4 37 34 S4 234 209
S5 204 226 S5 35 33 S5 322 249

7. limun

F0 li mun IT li mun D li mun

S1 128 154 S1 22 26 S1 187 322
S2 137 173 S2 25 29 S2 226 315
S3 131 164 S3 24 27 S3 210 318
S4 216 262 S4 31 33 S4 176 321
S5 198 242 S5 23 34 S5 256 319

332
8. daruh

F0 da ruh IT da ruh D da ruh

S1 127 177 S1 30 26 S1 169 320
S2 132 177 S2 32 26 S2 222 319
S3 130 170 S3 31 28 S3 212 309
S4 202 254 S4 35 32 S4 184 225
S5 204 229 S5 34 31 S5 243 351

9. məlməl

F0 məl məl IT məl məl D məl məl

S1 147 141 S1 25 26 S1 185 247
S2 133 145 S2 28 26 S2 253 253
S3 129 153 S3 27 26 S3 198 250
S4 211 261 S4 30 33 S4 177 201
S5 198 235 S5 29 31 S5 213 260

10. wəldək

F0 wəl dək IT wəl dək D wəl dək

S1 142 183 S1 26 29 S1 211 237
S2 141 159 S2 29 31 S2 291 217
S3 138 170 S3 28 30 S3 241 218
S4 225 284 S4 33 30 S4 233 258
S5 197 226 S5 27 33 S5 228 264

11. mayəl

F0 ma yəl IT ma yəl D ma yəl

S1 120 154 S1 27 23 S1 192 199
S2 134 156 S2 28 24 S2 272 211
S3 128 152 S3 28 25 S3 230 206
S4 203 268 S4 35 32 S4 163 247
S5 211 227 S5 32 31 S5 234 280

333
12. Rəmla

F0 Rəm la IT Rəm la D Rəm la

S1 129 140 S1 25 30 S1 183 214
S2 130 170 S2 24 31 S2 260 245
S3 129 157 S3 24 30 S3 232 230
S4 204 274 S4 32 32 S4 217 250
S5 206 233 S5 33 33 S5 264 202

13. walu

F0 wa lu IT wa lu D wa lu
S1 130 186 S1 30 20 S1 258 191
S2 137 178 S2 31 21 S2 299 234
S3 134 184 S3 30 22 S3 278 212
S4 197 274 S4 36 35 S4 220 208
S5 206 240 S5 33 31 S5 257 225

14. lina

F0 li na IT li na D li na
S1 126 174 S1 22 29 S1 207 211
S2 135 164 S2 25 28 S2 253 236
S3 132 172 S3 24 28 S3 230 245
S4 219 276 S4 27 32 S4 208 267
S5 209 233 S5 30 33 S5 243 219

15. lawyinhum

F0 law yin hum IT law yin hum D law yin hum

S1 125 138 180 S1 27 20 25 S1 245 175 234
S2 137 136 165 S2 31 24 28 S2 246 221 268
S3 133 135 170 S3 29 23 27 S3 239 202 250
S4 205 217 306 S4 36 32 34 S4 239 225 272
S5 206 207 242 S5 34 25 31 S5 230 167 254

334
16. manDaRin

F0 man Da Rin IT man Da Rin D man Da Rin

S1 127 129 170 S1 26 30 22 S1 215 155 236
S2 144 140 163 S2 26 31 19 S2 244 191 242
S3 132 136 169 S3 26 32 20 S3 230 180 241
S4 213 198 269 S4 33 32 31 S4 230 165 253
S5 210 212 238 S5 30 35 29 S5 215 183 240

17. mərmədnak

F0 mər məd nak IT mər məd nak D mər məd nak

S1 127 125 177 S1 22 20 27 S1 149 177 344
S2 145 137 168 S2 23 22 27 S2 175 189 315
S3 136 130 165 S3 24 23 25 S3 180 198 323
S4 209 207 268 S4 34 30 34 S4 197 184 256
S5 210 211 233 S5 31 29 34 S5 150 224 285

18. ʕəRganin

F0 ʕəR ga nin IT ʕəR ga nin D ʕəR ga nin

S1 121 124 172 S1 24 24 22 S1 162 214 219
S2 134 137 173 S2 28 30 22 S2 244 232 269
S3 125 131 165 S3 26 28 23 S3 187 223 255
S4 215 209 290 S4 33 34 30 S4 151 193 265
S5 182 211 238 S5 31 34 25 S5 168 205 245

19. mqulbinək

F0 mqul bi nək IT mqul bi nək D mqul bi nək

S1 149 143 193 S1 27 25 23 S1 343 175 215
S2 153 136 161 S2 31 27 24 S2 389 197 212
S3 151 140 178 S3 29 25 26 S3 360 187 215
S4 227 217 264 S4 32 31 30 S4 348 176 158
S5 222 215 235 S5 34 25 30 S5 343 167 218

335
20. wəldatni

F0 wəl dat ni IT wəl dat ni D wəl dat ni

S1 125 132 187 S1 20 30 21 S1 175 282 242
S2 140 139 173 S2 23 32 23 S2 191 260 221
S3 128 140 182 S3 22 33 24 S3 182 278 232
S4 213 213 275 S4 31 36 30 S4 208 255 286
S5 200 212 247 S5 23 35 24 S5 160 245 203

21. banyaha

F0 ban ya ha IT ban ya ha D ban ya ha

S1 124 127 177 S1 26 26 26 S1 263 203 227
S2 136 128 172 S2 32 24 27 S2 230 200 281
S3 129 125 174 S3 28 25 28 S3 254 202 233
S4 210 205 249 S4 33 34 30 S4 216 200 281
S5 212 208 246 S5 33 34 29 S5 228 215 219

22. gənDuRa

F0 gən Du Ra IT gən Du Ra D gən Du Ra

S1 127 127 164 S1 30 24 28 S1 218 160 190
S2 139 134 170 S2 28 31 27 S2 275 204 208
S3 129 132 166 S3 28 26 25 S3 222 180 198
S4 239 211 276 S4 31 29 29 S4 243 204 208
S5 211 121 233 S5 33 33 31 S5 210 159 159

23. yəDRəbna

F0 yəD Rəb na IT yəD Rəb na D yəD Rəb na

S1 194 125 168 S1 15 27 27 S1 196 154 179
S2 145 135 155 S2 21 31 26 S2 209 269 211
S3 165 126 164 S3 19 25 26 S3 198 174 203
S4 227 204 279 S4 23 26 30 S4 152 163 241
S5 197 214 247 S5 26 34 32 S5 155 199 179

336
24. məlmluna

F0 məl mlu na IT məl mlu na D məl mlu na

S1 127 136 171 S1 19 22 27 S1 223 219 161
S2 144 140 174 S2 25 26 26 S2 292 228 230
S3 129 138 179 S3 23 25 28 S3 285 234 196
S4 220 201 291 S4 24 22 26 S4 150 252 208
S5 208 215 235 S5 26 30 32 S5 185 261 174

25. limuna

F0 li mu na IT li mu na D li mu na
S1 126 130 168 S1 20 24 29 S1 134 190 200
S2 144 140 171 S2 25 28 28 S2 186 249 214
S3 134 139 161 S3 23 26 29 S3 141 220 211
S4 210 203 288 S4 27 24 29 S4 149 158 206
S5 210 214 225 S5 25 32 33 S5 174 189 186

26. banyahalih

a. Fundamental Frequency
F0 ban ya ha lih
S1 122 124 124 179
S2 141 135 135 174
S3 123 125 127 169
S4 217 207 197 288
S5 204 206 215 245

b. Intensity
IT ban ya ha lih
S1 26 23 27 23
S2 30 28 27 25
S3 27 25 27 24
S4 28 27 28 23
S5 32 34 30 31

337
 c. Duration
D ban ya ha lih
S1 195 134 152 256
S2 242 174 196 311
S3 224 179 198 279
S4 206 159 191 253
S5 200 184 162 201

27. dirihaliha

a. Fundamental Frequency
F0 di ri ha li ha
S1 129 128 128 127 163
S2 147 146 139 128 168
S3 131 126 126 129 161
S4 234 225 204 200 352!!
S5 207 204 209 207 249

b. Intensity
IT di ri ha li ha
S1 22 25 27 22 28
S2 28 29 29 26 26
S3 26 27 29 24 29
S4 30 29 26 26 27
S5 28 29 33 34 31

c. Duration
D di ri ha li ha
S1 146 115 124 152 203
S2 170 127 204 170 251
S3 187 134 142 165 209
S4 144 163 163 167 265
S5 115 146 163 114 193

338
28. gəlbuhalikum

a. Fundamental Frequency
F0 gəl bu ha li kum
S1 126 129 127 129 161
S2 139 148 144 137 159
S3 137 141 138 134 164
S4 233 216 210 202 265
S5 203 211 211 203 242

b. Intensity
IT gəl bu ha li kum
S1 24 27 29 18 27
S2 24 30 31 21 24
S3 24 28 30 21 27
S4 30 31 31 27 30
S5 32 31 31 31 31

c. Duration
D gəl bu ha li kum
S1 200 136 156 131 247
S2 150 150 196 196 233
S3 198 145 179 144 239
S4 149 154 160 143 259
S5 195 107 128 157 225

339
 Appendix G

Pitch Tracks of Selected Test Items in Context

5. [walmək] “it (masc.) suits you”

340
7. [limun] “Oranges”

341
9. [məlməl] “He shook (sth.)

342
11. [mayəl] “Bent (masc.)

343
15. [lawyinhum] “They are twisting them”

344
16. [manDaRin] “Clementine”

345
25. [limuna] “An orange”

346
39. [gəlbuhalikum] “They reversed it (fem.) for you (pl.)”

347
 Bibliography

Abdelmassih, Ernest (1973) An Introduction to Moroccan Arabic. The University of

Michigan: Ann Arbor, MI.

Abu-Mansour, Mahasen (1995) Optimality and Conspiracy in the Syllable Structure of

Arabic. In University of Massachusetts Occasional Papers in Linguistics
18: Papers in Optimality Theory, ed. Jill Beckman, Laura Walsh Dickey, and
Suzanne Urbanczyk. Amherst, MA: GLSA publications.

ADDaR AlbayDaaʔ (Casablanca):1, March 1999. A magazine edited (in Arabic) by the
“Centre de Documentation et d’Information de la Grande Commune Urbaine de
Casablanca”. Casablanca: Dar An-Nashr Al Maghrebiyya.

Aguadé, Jordi (1994) The Arabic Dialect of Skura (Southern Morocco). In J. Aguadé, F.
Corriente and M. Maruhan (eds.) Actas del Congreso Internacioal Sobre
Intereferencias Linguisticas Arabo-Romances Y Paralelos Extra-Iberos. Zaragoza:
Navarro and Navarro.

Ait Hammou, Youssef (1988) Syllabe et processus phonologiques en arabe marocain.

Mémoire de D.E.S. Faculté des Lettres, Rabat.

Alber, Birgit (1997) Quantity Sensitivity as the Result of Constraint Interaction. To

appear in G. Booij and J. van de Weijer (eds.) Phonology in Progress. HIL Phonology
Papers III, 1-45. The Hague: Holland Academic Press.

Al Ghadi, Abdellatif (1990) Moroccan Arabic Plurals and the organization of the
Lexicon. D.E.S. Thesis. Faculté des Lettres, Rabat.

Al Ghadi, Abdellatif (1994) An OT Account of Moroccan Arabic Prosody. Ms. Delaware

University, and Université Mohamed V, Faculté des Lettres, Rabat.

Allen, Margaret (1978) Morphological Investigations. Ph.D. dissertation. Cambridge,

MA: MIT.

Aronoff, Mark (1976) Word formation in Generative Grammar. Cambridge,

Massachusetts: MIT Press.

Bakovic, Eric (1996) Foot Harmony and Quantitative Adjustments. Ms. New Brunswick,
NJ: Rutgers University. Rutgers Optimality Archive 168.

348
Basri, Hassan, Ellen Broselow, Daniel Finer, and Elisabeth Selkirk. (1998)
Morphologically Grounded Affix Classes: the Case of Makassarese. Ms. State
University of New York at Stony Brook, NY and University of Massachusetts at
Amherst, MA.

Beckman, Mary and Janet Pierrehumbert (1986) Intonational Structure in Japanese and
English. In Phonology Yearbook 3, 255-309.

Beckman, Jill (1995) Shona Height Harmony: Markedness and Positional Identity. In
University of Massachusetts Occasional Papers in Linguistics 18: Papers in
Optimality Theory, ed. Jill Beckman, Laura Walsh Dickey, and Suzanne
Urbanczyk, pp. 53-75. Amherst, MA: GLSA publications.

Beckman, Jill (1998) Positional Faithfulness. Ph.D. dissertation. Amherst, MA:

Uinversity of Massachusetts at Amherst.

Bellout, Zoubida (1987) Moroccan Arabic Syllable Structure. D.E.S. thesis. Faculté des
Lettres I, Casablanca.

Benhallam, Abderrafi. (1980) Syllable Structure and Rule Types in Arabic. Ph.D.
Dissertation. Gainesville, FL: University of Florida.

Benhallam, Abderrafi (1988) Vers une description non-linéaire de la syllabe en arabe

marocain. In Le Maroc et la Hollande, Publications de la Faculté des
Lettres, Rabat, pp. 244-56.

Benhallam, Abderrafi (1989) Aspects de la recherche en phonologie de l’arabe marocain.

In Langue et Socité au Maghreb: Bilan et Perspectives 13, pp. 13-23. Faculté des
Lettres, Rabat.

Benhallam, Abderrafi (1990a) Moroccan Arabic Syllable Structure. In Langues et

littératures VIII, pp. 177-191. Faculté des Lettres, Rabat.

Benhallam, Abderrafi (1990b) Native Speaker Intuitions about Moroccan Arabic Stress.
In Maghreb Linguistics, ed. Jochen Pleines, pp. 91-109. Rabat: Editions Okad.

Benhallam, Abderrafi (1991) On Geminates in Moroccan Arabic. In Linguistica

Communicatio. Publications de la Faculté des Lettres de Fès.

Benhallam, Abderrafi and Mohamed Dahbi (1990) Accents of Moroccan Arabic: A

Preliminary Study. In Maghreb Linguistics, ed. Jochen Pleines, pp. 111-125.
Rabat: Editions Okad.

Benhallam, Abderrafi (1995) Etude des accents de l’arabe marocain. In Dialectologie et

Sciences Humaines au Maroc 38. Faculté des Lettres, Rabat.

349
Benhallam, Abderrafi (1998) Contact and Historical Evolution of Languages in Morocco.
In Langues et Littératures XVI. Faculté des Lettres, Rabat.

Benkaddour, Adelfettah (1982) A Nonlinear Analysis of Some aspects of the Phonology

and Nonconcatenative Morphology Of Arabic. Ph.D. Dissertation. School of
Oriental and African Studies, London.

Benkirane, Thami (1982) Etude phonétique et fonction de la syllabe en arabe marocain.

Thèse de 3ème cycle. Université de Provence.

Bennis, Said (1992) La formation du causatif en arabe marocain. Mémoire de D.E.S.

Faculté des Lettres, Rabat.

Bensoukas, Karim (1994) Tashlhit Agentive Nouns: An Optimality-Theoretic Approach.

D.E.S. Thesis. Faculté des Lettres, Rabat.

Bensoukas, Karim (1999) Labial Alternations in Tashlhit Revisited: A Constraint-based

Analysis. Ms. Faculté des Lettres, Rabat.

Benua, Laura (1995) Identity Effect in Morphological Truncation. In University of

Massachusetts Occasional Papers in Linguistics 18 Papers in Optimality Theory,
ed. By Jill Beckman, Laura Walsh Dicky and Suzanne Urbanczyk, pp. 77-136.
Amherst: MA, GLSA publications.

Benua, Laura (1997) Transderivational Identity: Phonological Relations Between Words.

Ph.D. dissertation. Amherst, MA: University of Massachusetts at Amherst.
Rutgers Optimality Archive 259.

Bobaljik, Jonathan (1997) Mostly Predictable: Cyclicity and the Distribution of Schwa in
Itelmen. To appear in WECOL, University of California at Santa Cruz. Rutgers
Optimality Archive 208.

Bohas, Georges, Abdeljebbar Hammoumi and N. Chami (1989) Accentuation et

effacement dans le parler de Tanger. In Langues Orientales Anciennes:
Philosophie et Linguistique 2.

Borowsky, Toni (1999) Some Consequences of Word Faithfulness. To appear in

University of Massachusetts Occasional Papers in Linguistics 23, ed. Paul
de Lacy and Anita Nowak, pp. 1-24.

Boudlal, Abdelaziz (1993) Moroccan Arabic Glides: A Lexical Approach. D.E.S. Thesis.
Faculté des Lettres, Rabat.

Boudlal, Abdelaziz (1995) Some aspects of the Lexical Phonology of Moroccan Arabic.
In Langue et Littératures XIII, pp. 23-43. Faculté des Lettres, Rabat.

350
Boudlal, Abdelaziz (1996) Constraint Interaction in the Phonology and Morphology of
Moroccan Arabic. Doctoral prospectus. Faculté des Lettres, Rabat.

Boudlal, Abdelaziz (1998) A Diachronic Analysis of Labialization in Moroccan Arabic.

In Langue et Littératures XVI, pp. 45-60. Faculté des Lettres, Rabat.

Boudlal, Abdelaziz (to appear a). An Optimality-Theoretic Account of Moroccan Arabic

Syllable Structure. In Proceeding of the Study Day on Moroccan Arabic,
organized by the “Unité de Recherche Linguistique et Didactique” (5 April 2000).
Faculté des Lettres, Ben M’Sik, Casablanca.

Boudlal, Abdelaziz (to appear b). A Note on Four Morphological Classes in Moroccan
Arabic. In Revue de la Faculté des Lettres de Beni-Mellal 4. Faculté des Lettres,
Beni-Mellal.

Boukous, Ahmed (1979) Le profil sociolinguistique du Maroc. In Bulletin Economique et

Social du Maroc 40, pp. 5-31.

Boukous, Ahmed (1998) La situation sociolinguistique au Maroc. In Plurilinguismes (Le

Maroc) 16, pp. 5-30. Centre d’Etudes et de Recherches en Planification
Linguistique, Paris.

Bouziri, Raja, Hassan Nejmi and Mohamed Taki (1991) L’accent de l’arabe marocain
parlé à Casablanca et à Tunis: étude phonétique et phonologique. In Actes du
XIIème Congrès International des Sciences Phonétiques (Aix-en-Provence),
Vol. 5, pp. 134-137.

Brame, Michael (1974) The Cycle in Phonology: Stress in Palestinian, Maltese, and
Spanish. In Linguistic Inquiry 5, pp. 39-60. Cambridge, MA: MIT.

Brunot, Louis (1950) Introduction à l’arabe marocain. Paris: Maisonneuve.

Buckley, Eugene (1994) Alignment in Manam Stress. Ms. University of Pennsylvania.

Rutgers Optimality Archive 183.

Burzio, Luigi (1996) Surface Constraints Versus Underlying Representation. In Current

Trends in Phonology: Models and Methods, ed. Jacques Durand and
Bernard Laks, pp. 123-42. Manchester, England: European Studies Research
Institute, University of Salford.

Cabré, Tereza and Michael Kenstowicz (1995) Prosodic Trapping in Catalan. In

Linguistic Inquiry 16: 4, pp. 694-705.

Cantineau, Jean (1950) Reflexions sur la phonologie de l’arabe marocain. In Hespéris 37,
pp. 193-207.

351
Chen, Matthew (1970) Vowel length Variation as a Function of the Voicing of the
Consonant environment. In Phonetica 22, pp. 129-59.

Chomsky, Noam (1970) Remarks on Nominalization. In Studies on Semantics in

Generative Grammar, ed. Noam Chomsky. The Hague: Mouton.

Chomsky, Noam and Morris Halle (1968) The Sound Pattern of English. New York:
Harper and Row.

Clements, George (1988) The Sonority Cycle and Syllable Organization. In Phonologica,
ed. Dreshler et al., pp. 63-76. Cambridge, MA: MIT Press.

Clements, George (1993) Lieu d’articulation des consonnes et des voyelles: une théorie
unifiée. In L’architécture de la représentation phonologique, ed. B. Laks and A.
Rialland. Editions CNRS, Paris.

Clements, George and Samuel Keyser (1983) CV Phonology: A Generative Theory of the
Syllable. MIT Press. Cambridge, MA.

Clements, George and Elisabeth Hume. (1995) The Internal Organization of Speech
Sounds. In The Handbook of Phonological Theory, ed. John Goldsmith, pp. 245-
306. Oxford, UK: Blackwell.

de Lacy, Paul (1998) The Effect of Consonant Clusters on Vowel Duration in English.
Ms. Amherst, MA: University of Massachusetts at Amherst.

de Lacy, Paul (1999) Circumscriptive Morphemes. To appear in Proceedings of the Sixth

Meeting of the Austronesian Formal Linguistics Association, ed. C. Kitto and C.
Smallwood. Holland Academic Graphics. Rutgers Optimality Archive 339.

Dell, François and Mohamed El Medlaoui (1985) Syllabic Consonants and Syllabification
in Imdlawn Tachelhiyt Berber. Journal of African Languages and Linguistics 7,
pp. 105-130.

El Hadri, Mustapha (1993) A Metrical Approach to Stress in Moroccan Arabic Verbs.

D.E.S. Thesis. Faculté des Lettres, Rabat.

El Himer, Mohamed (1991) La morphologie verbale de l’arabe marocain: verbes

simples. Mémoire de D.E.S. Faculté des Lettres, Rabat.

El Himer, Mohamed (1993) La gémination en tant que procésus prosodique en arabe

marocain. Ms. Faculté des Lettres, Beni-Mellal.

El Medlaoui, Mohamed (1992) Aspects de representations phonologiques dans certaines

langues chamito-sémitiques. Thèse d’état. Faculté des Lettres, Rabat.

352
Fares, Najiba (1993) Stress in Moroccan Arabic Nouns and Adjectives: A Metrical
Approach. D.E.S. thesis. Faculté des Lettres, Rabat.

Farwaneh, Samira (1995) Directionality Effects in Arabic Dialect Syllable Structure.

Ph.D. dissertation. University of Utah.

Farwaneh, Samira (1996) The Interaction of Stress and Epenthesis in Arabic. In

Proceedings of the 12th Eastern States Conference on Linguistics (1995), ed. M.
Przezdziecki and L. Whaley. Cornell University, Ithaca.

Ferguson, Charles (1959) Diaglossia. In Word 15, pp. 325-340.

Gafos, Adamantios (1996) The Articulatory Basis of Locality in Phonology. Ph.D.

dissertation. Johns Hopkins University.

Gafos, Adamantios (1998) Eliminating Long-distance Consonantal Spreading. In Natural

Language and Linguistic Theory 16, pp. 223-278.

Giangola, James (1997) Constraint Interaction and Brazilian Portuguese Glide

Distribution. To appear in Proceedings of theNorth East Linguistic Society 27.
Rutgers Optimality Archive 182.

Gnanadesikan, Amalia (1997) Phonology with Ternary Scales. Ph.D. dissertation.

Amherst, MA: University of Massachusetts at Amherst.

Goldsmith, John (1976) Autosegmental Phonology. Ph.D. dissertation. Cambridge, MA:

MIT.

Goldsmith, John (1979) The Aims of Autosegmental Phonology. In Current Phonological

Theories, ed. D. Dinnsen. Bloomington, Indiana: Indiana University Press.

Goldsmith, John (1990) Autosegmental and Metrical Phonology. Cambridge and Oxford:
Blackwell.

Goldsmith, John (ed.) (1995) The Handbook of Phonological Theory. Oxford: Blackwell.

Green, Antony (1997) The Prosodic Structure of Irish, Scots Gaelic and Manx. Ph.D.
dissertation, Cornell University. Rutgers Optimality Archive 196.

Guerssel, Mohamed (1978) A Condition on Assimilation Rules. In Linguistic Analysis 3,

pp. 267-305.

Guerssel, Mohamed (1986) Glides in Berber and Syllabicity. In Linguistic Inquiry 17, pp.
1-12. Cambridge, MA: MIT.

353
Hale, Mark, Madelyn Kissock, and Charles Reiss. (1997) Output-Output Correspondence
in Optimality Theory. To appear in proceedings of theWest Coast Conference on
Formal Linguistics XVI. Rutgers Optimality Archive 202.

Halle, Morris (1973) Prolegomena to a Theory of Word Formation. In Linguistic Inquiry

4, pp. 3-16. Cambridge, MA: MIT.

Hammari, Adelouahed (1996) Place Features in Rhafsai Moroccan Arabic. D.E.S.

Thesis. Faculté des Lettres, Rabat.

Hammari, Abdelouahed (2000) An Optimality-Theoretic Analysis of Rhafsai Moroccan

Arabic. Doctoral dissertation. Faculté des Lettres, Rabat.

Hammoumi, Abdeljebbar (1988) Syllabation, accentuation, éffacement et abrègement

dans le parler de Fès et d’autres parlers marocains. Thèse de doctorat. Université
de la Sorbonne Nouvelle, Paris III.

Hargus, Sharon and Siri Tuttle. (1999) Augmentation as Affixation in Athabaskan

Languages. Ms. University of Washington.

Harris, Zellig (1942) The Phonemes of Moroccan Arabic. In Journal of the American
Oriental Society 62, pp. 309-18.

Harrell, Richard (1962) A Short Reference Grammar of Moroccan Arabic. Georgetown

University Press, Washington, DC.

Hayes, Bruce (1986) Inalterability in CV Phonology. In Language 62: 2, pp. 321-351.

Hayes, Bruce. (1989) Compensatory Lengthening in Moraic Phonology. In Linguistic

Inquiry 20: 2. Cambridge, MA: MIT.

Hayes, Bruce (1995) Metrical Stress Theory: Principles and Case Studies. Chicago:
University of Chicago Press.

Hume, Elizabeth (1992) Front Vowels, Coronal Consonants, and Their Interaction in
Nonlinear Phonology. Ph.D. dissertation. Cornell University, Ithaca.

Hong, Soonhyn (1997) Prosodic Domains and Ambisyllabicity in Optimality Theory.

Ph.D. dissertation, University of Pennsylvania. Rutgers Optimality Archive 209.

Hung, Henrietta (1995) The Rhythmic and Prosodic Organization of Edge Constituents:
An Optimality-Theoretic Account.. Indiana University Linguistics Club.
Bloomington, Indiana.

Hyman, Larry (1985) A Theory of Phonological Weight. Foris, Dordrecht.

354
Imouzaz, Said (1991) La dérivation du nom d’action en arabe marocain. Mémoire de D.E.S.
Faculté des Lettres, Rabat.

Imouzaz, Said (forthcoming) Dérivation nominale à base verbale et optimalité. Thèse d’état.
Faculté des Lettres de Ben M’Sik, Casablanca.

Ito, Junko (1989) A Prosodic Theory of Epenthesis. In Natural Language and Linguistic
Theory 7, pp. 217-260.

Jakobson, Roman (1962) Selected Writings. The Hague: Mouton.

Jebbour, Adelkrim (1996) Morphologie et contraintes prosodiques en berbère:

analyselinguistique et traitement automatique. Thèse d’état. Faculté des Lettres,
Rabat.

Kager, René (1996) Surface Opacity of Metrical Structure in Optimality Theory. In The
Derivational Residue in Phonology, ed. B. Hermans and M. van Oostendorp. Tilburg,
Holland. Rutgers Optimality Archive 207.

Kager, René (1999) Optimality Theory. Cambridge, MA: Cambridge University Press.

Kager, René, Harry Van der Hulst, and Wim Zonneveld, (eds.) (1999) The Prosody-
Morphology Interface. Cambridge University Press, Cambridge, Massachusetts.

Kaye, Jonathan, Jean Lowenstamm, and Jean-Roger Vergnaud (1985) The Internal Structure of
Phonological Elements: A Theory of Charm and Government. In Phonology 2, pp. 305-28.

Kaye, Jonathan, Jean Lowenstamm, and Jean-Roger Vergnaud (1990) Constituent Structure
and Government in Phonology. In Phonology 7: 2.

Keer, Edward (1998) Spirantization and Geminate Inalterabity. In Rutgers Occasional Papers, ed.
R. Artstein and M. Holler, pp. 147-167.

Keer, Edward (1999) Geminates, the OCP and the Nature of Con. Ph.D. dissertation. New
Brunswick, NJ: Rutgers University.

Kenstowicz, Michael (1994) Syllabification in Chukchee: A Constraints-based Analysis. To

appear in proceedings of the Formal Linguistics Society of Mid-America 4. Rutgers
Optimality Archive 30.

Kenstowicz, Michael (1996) Base-Identity and Uniform Exponence: Alternatives to Cyclicity. In

Current Trends in Phonology: Models and Methods, ed. Jacques Durand and Bernard
Laks, pp. 363-94. Manchester, England: European Studies Research Institute, University
of Salford. Rutgers Optimality Archive 103.

Kenstowicz, Michael (1997) Uniform Exponence: Exemplification and Extension. Ms.

Cambridge, MIT. Rutgers Optimality Archive 218.

Khomsi, A. (1979) Etude phonétique et phonologique de l’arabe marocain de Casablanca.

Thèse de 3ème cycle. Paris VI.

355
 Kingston, John (1998). Class Handouts from Linguistics 820: Developments in Laboratory
Phonology. Amherst, MA: University of Massachusetts at Amherst.

Kiparsky, Paul (1982) Lexical Morphology and Phonology. In Linguistics in the Morning
Calm, Linguistic Society of Korea, ed. I.S. Yang. Hanshin, Seoul.
Kiparsky, Paul (1998) Paradigm Effects and Opacity. Ms. Stanford University.

Kluender, Keith, Randy Diehl and Beverly Wright (1988) Vowel-Length Differences
before Voiced and Voiceless Consonants: an Auditory Explanation. In Journal of
Phonetics 16, pp. 153-169.

Laefur, Christiane (1992) Patterns of Voicing-conditioned Vowel Duration in French and

English. In Journal of Phonetics 20, pp. 411-40.

Lasri, A. (1989) Le diminutif en arabe marocain. Ms. ILPGA, Paris III.

Leben, William (1973) Suprasegmental Phonology. Ph.D. dissertation. Cambridge, MA:

MIT.

Lombardi, Linda and John McCarthy (1991) Prosodic Circumscription in Choctaw

Morphology. In Phonology 8, pp. 37-71.

Mack, Molly (1982) Voicing-dependent Vowel Duration in English and French:

Monolingual and Bilingual Production. In Journal of the Acoustic Society of
America 71:1, pp. 173-78.

Marsil, Ouafa (1988) La structure syllabique en arabe marocain. Mémoire de D.E.S.

Faculté des Lettres, Rabat.

Mawhoub, Mourad (1992) Intonation et organisation de l’énnoncé en arabe marocain

parlé à Casablanca: aspects prosodiques et structures énonciatives. Thèse de
doctorat. Paris III, Université de la Sorbonne Nouvelle.

McCarthy, John (1979) Formal Problems in Semitic Phonology and Morphology. Ph.D.
dissertation, MIT, Cambridge, MA.

McCarthy, John (1981) A Prosodic Theory of Nonconcatenative Morphology. In

Linguistic Inquiry 12, pp. 373-418. Cambridge, MA: MIT.

McCarthy, John (1982) Prosodic Templates, Morphemic Templates and Morphemic

Tiers. In The Structure of Phonological Representations 1, ed. Harry van der Hulst
and Norval Smith, pp. 191-225.

McCarthy, John (1983) Consonantal Morphology in the Chaha Verb. In Proceedings of the
second West Coast Conference on Formal Linguistics. Stanford University.

356
McCarthy, John (1986) OCP Effects: Gemination and Antigemination. In Linguistic
Inquiry 17, pp. 207-263.

McCarthy, John (1993) A Case of Surface Constraint Violation. In C. Paradis and D. La

Charité, eds., Constraint-Based Theories in Multilinear Phonology, special issue
of Canadian Journal of Linguistics 38, 169-195

McCarthy, John (1995) Extensions of Faithfulness: Rotuman Revisited. Ms. Amherst,

MA: University of Massachusetts at Amherst. Rutgers Optimality Archive 110.

McCarthy, John (1997) Faithfulness and Prosodic Circumscription. To appear in The

Pointing Finger: Conceptual Studies in Optimality Theory, ed. Joost Dekkers,
Frank van der Leeuw and Jeroen van de Weijer. Amesterdam: HIL.

McCarthy, John and Alan Prince (1986) Prosodic Morphology. Ms. University of
Massachusetts, Amherst and Brandeis University.

McCarthy, John and Alan Prince (1990a) Foot and Word in Prosodic Morphology: The
Arabic Broken Plurals. In Natural Language and Linguistic Theory 8, pp. 209-83.

McCarthy, John and Alan Prince (1990b) Prosodic Morphology and Templatic
Morphology. In Perspectives on Arabic Linguistics: Papers from the Second
Symposium, ed. M. Eid and J. McCarthy, pp. 1-54. Amesterdam, Benjamins.

McCarthy, John and Alan Prince (1993a) Prosodic Morphology I: Constraint Interaction
and Satisfaction. To appear, MIT Press. Technical Report no. 3, Rutgers
University, Center for Cognitive Science.

McCarthy, John and Alan Prince (1993b) Generalized Alignment. In Yearbook of

Morphology pp, ed. Geert Booij and Jaap van Marle, pp. 79-153. Dordrecht:
Kluwer.

McCarthy, John and Alan Prince (1994a) The Emergence of the Unmarked: Optimality in
Prosodic Morphology. In Proceedings of the North East Linguistic Society 24, ed.
Mercè Gonzàlez, pp. 333-79. Amherst, MA: GLSA Publications.

McCarthy, John and Alan Prince (1994b) Two Lectures on Prosodic Morphology
(Utrecht, 1994). Part I: Template Form in Prosodic Morphology. Part II:
Faithfulness and Reduplicative Identity. Ms. Amherst, MA: University of
Massachusetts at Amherst and New Brunswick, NJ: Rutgers University.

McCarthy, John and Alan Prince. (1995) Faithfulness and Reduplicative Identity. In
University of Massachusetts Occasional Papers in Linguistics 18: Papers in
Optimality Theory, ed. Jill Beckman, Laura Walsh Dickey, and Suzanne
Urbanczyk, pp. 249-384. Amherst, MA: GLSA publications. Rutgers Optimality
Archive 60.

357
McCarthy, John and Alan Prince. (1999) Faithfulness and Identity in Prosodic
Morphology. In The Prosody-Morphology Interface, ed. René Kager, Harry van
der Hulst and Wim Zonneveld. Cambridge, MA: MIT Press.

Meliani, Mohamed (1995) Formation de l’adjectif en arabe marocain: approche

prosodique. Mémoire de D.E.S. Faculté des Lettres, Rabat.

Mester, Armin and Jaye Padgett (1994) Directional Syllabification in Generalized

Alignment. In Phonology at Santa Cruz 3,pp. 79-85.

Mohanan, Karuvannour P. (1986) The Theory of Lexical Phonology. Foris, Dordrecht.

Moren, Bruce (1997) Markedness and Faithfulness Constraints on the Association of

Moras: the Dependency Between Vowel Length and Consonant Weight. M.A.
thesis. University of Maryland at College Park. Rutgers Optimality Archive 248.

Moumine, El Amine (1990) Sociolinguistic Variation in Casablanca Moroccan Arabic.

D.E.S. thesis. Faculté des Lettres, Rabat.

Moumine, El Amine (1995) Variation in Casablanca Moroccan Arabic. In Dialectologie

et Sciences Humaines au Maroc 38. Faculté des Lettres, Rabat.

Nejmi, Hassan (1993) Contribution à une étude des processus accentuels de l’arabe
marocain de Casablanca: approache phonétique et phonologique (accent de mot).
Thèse de doctorat. Paris III, Université de la Sorbonne Nouvelle.

Nejmi, Hassan (1995) Les variations prosodiques de l’accent de l’arabe marocain parlé à
Casablanca: approche phonétique et phonologique (accent de mot). In Recherches
Linguistiques. Faculté des Lettres, Ben M’Sik, Casablanca.

Nelson, Nicole (1998) Right Anchor, Aweigh. Ms. New Brunswick: NJ: Rutgers
University. Rutgers Optimality Archive 284.

Nespor, Marina and Irene Vogel (1982) Prosodic Domains of External Sandhi Rules. In
The Structure of Phonological Representations 2, ed. Harry van der and Norval
Smith, pp. 225-56. Dordrecht: Foris.

Nespor, Marina and Irene Vogel (1986) Prosodic Phonology. Dordrecht: Foris.

Paradis, Carole (1988a) On Constraints and Repair Strategies. In The Linguistic Review 6,
pp. 71-97.

Paradis, Carole (1988b) Towards a Theory of Constraint Violations. In McGill Working

Papers in Linguistics 5, pp. 1-44.

358
Paradis, Carole (1996) The Inadequacy of Filters and Faithfulness in Loanword
Adaptation. In Current Trends in Phonology: Models and Methods, ed. Jacques
Durand and Bernard Laks. Manchester, England: European Studies Research
Institute, University of Salford, Salford.

Paradis, Carole and Darlene La Charité (eds.) (1993) Constraint-Based Theories in

Multilinear Phonology. Special issue, Canadian Journal of Linguistics 39: 2.

Pater, Joe (1995) On the Nonuniformity of Weight-to-Stress and Stress Preservation

Effects in English. Ms. McGill University. Rutgers Optimality Archive 107.

Peterson, Gordon and Ilse Lehiste (1960) Duration of Syllable Nuclei in English. In
Journal of the Acoustic Society of America 32: 6, pp. 693-703.

Pierrehumbert, Janet (1980) The Phonology and Phonetics of English Intonation. Ph.D.
dissertation. Cambridge, MA: MIT.

Prince, Alan and Paul Smolensky (1991) Connectionism and Harmony Theory in
Linguistics. Report no. CU-CS-533-91. University of Colorado, Boulder.

Prince, Alan and Paul Smolensky (1992) Optimality: Constraint Interaction in Generative
Grammar. Paper presented at the 12th West Coast Conference on Formal
Linguistics. Los Angeles, CA.

Prince, Alan and Paul Smolensky (1993) Optimality Theory: Constraint Interaction in
Generative Grammar. Technical Report no. RuCCS-TR-2. New Brunswick, NJ:
Rutgers University Center for Cognitive Science.

Rguibi, Samira (1990) Moroccan Arabic and Lexical Phonology. D.E.S. thesis. Faculté
des Lettres, Rabat.

Rguibi, Samira (forthcoming) Phonological Variation and Change in Moroccan Arabic:

The Case of the Tazi Dialect. Doctoral dissertation. Faculté des Lettres, Rabat.

Rose, Sharon. (1997) Theoretical Issues in Comparative Ethio-Semitic Phonology and

Morphology. Ph.D. dissertation. McGill University.

Rosenthall, Samuel (1994) Vowel-Glide Alternation in a Theory of Constraint Interaction.

Ph.D. dissertation. Amherst, MA: University of Massachusetts at Amherst.

Sagey, Elizabeth (1986) The Representation of Features and relations in Nonlinear

Phonology. Ph.D. dissertation. Cambridge, MA: MIT.

Samek-Lodovici, Vieri (1993) A Unified Analysis of Crosslinguistic Morphological

Gemination. New Brunswick, NJ: Rutgers University.

359
Schein, Barry and Donca Steriade. On Geminates. In Linguistic Inquiry 17, pp. 691-744.
Cambridge, MA: MIT.

Selkirk, Elisabeth (1978) On Prosodic Structure and its Relation to Syntactic Structure. In
Nordic Prosody, ed. T. Fretheim. Trondheim: TAPIR.

Selkirk, Elisabeth O. (1980) The Role of Prosodic Categories in English Word Stress. In
Linguistic Inquiry 11, pp. 563-605. Cambridge, MA: MIT.

Selkirk, Elisabeth (1981) Epenthesis and Degenerate Syllable in Cairene Arabic. In H.

Borer and J. Aoun, eds., Theoretical Issues in the Grammar of the Semitic
Languages, pp. 111-40. Cambridge, MA: MIT.

Selkirk, Elisabeth (1982)The Syntax of Words. Linguistic Inquiry Monograph 7.

Cambridge, MA: MIT Press.

Selkirk, Elisabeth (1984) Phonology and Syntax: The Relation Between Sound and
Structure. Cambridge, MA: MIT Press.

Selkirk, Elisabeth (1986) On Derived Domains in Sentence Phonology. In Phonology

Yearbook 3, pp. 371-405.

Selkirk, E. (1990) A Two-Theory of Length. In University of Massachusetts Occasional

Papers 14, ed. J. Padgett and E. Dunlap. Amherst, MA: GLSA publications.

Selkirk, E. (1991) On the Inalterability of Geminates. In Papers from the (1990 Cortona
Phonology Meeting, ed. P. M. Bertinetto, M. Kenstowicz and M. Loporcaro.

Selkirk, Elisabeth (1993) [Labial] Relations. Ms. Amherst, MA: University of

Massachusetts at Amherst.

Selkirk, Elisabeth (1995a) The Prosodic Structure of Function Words. In University of

Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory,
ed. Jill Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk, pp. 439-69.
Amherst, MA: GLSA publications.

Selkirk, Elisabeth (1995b) Language-Particular Violations of a Universal Constraint: the

OCP from the Perspective of Optimality Theory. Paper presented at the Marrakech
Round Table on Contact and Historical Evolution of Languages in Morocco
11-13 Janauary, 1995. Faculté des Lettres, Rabat.

Selkirk, Elisabeth (1999) Morphologically Governed Output-Output Constraints in a

Noncyclic Optimality Theoretic Grammar: Evidence from the Makassar
Languages. Paper presented at Sophia University, Tokyo.

360
Shaw, Patricia (1993) The Prosodic Constituency of Minor Syllables. In E. Duncan, D.
Farkas and P. Spaelti (eds.) The Proceedings of the 12th West Coast Conference
on Formal Linguistics. University of California at Santa Cruz.

Siegel, Dorothy (1974) Topics in English Morphology. Ph.D. dissertation. Cambridge,

MA: MIT.

Smith, Jennifer (1997) Noun Faithfulness: On the Privileged Behavior of Nouns in

Phonology. Ms. Amherst, MA: University of Massachusetts at Amherst. Rutgers
Optimality Archive 242.

Smolensky, Paul (1996) The Initial State and Richness of the Base. Ms. University of
Colorado at Boulder. Rutgers Optimality Archive 154.

Spaelti, Philip (1997) Dimensions of Variation in Multi-Pattern Reduplication. Ph.D.

dissertation. University of California at Santa Cruz.

Urbanczyk, S. C. (1996) Patterns of Reduplication in Lushootseed. Ph.D. dissertation.

Amherst, MA: University of Massachusetts at Amherst, Reproduced by GLSA.

Walker, Rachelle (1994) Buriat Syllable Weight and Head Prominence. In Phonology in
Santa Cruz 3, pp. 99-110. Rutgers Optimality Archive 142.

Walker, Rachelle (1996) Prominence-Driven Stress. Ms. University of California at Santa

Cruz. Rutgers Optimality Archive 172.

Waksler, Rachelle (1990) A Formal Account of Glide/Vowel Alternation in Prosodic

Theory. Ph.D. dissertation. Harvard University.

Youssi, Abderrahim (1977) L’arabe marocain conceptuel. Thèse de 3ème cycle. Paris V.

Youssi, Abderrahim (1983) La triglossie dans la typologie linguistique. La linguistique

19, 2, pp. 71-83.

Youssi, Abderrahim (1986) L’arabe marocain médian. Analyse fonctionnaliste des

rapports syntaxiques (de la synchronie dynamique dans les corrélations des
normes sociolinguistiques et des formes phonologiques, morphosyntaxiques, et
lexicales). Thèse d’état. Université de Sorbonne Nouvelle, Paris III.

Youssi, Abderrahim (1989) Changements socioculturels et dynamique linguistique. In

Langue et Socité au Maghreb: Bilan et Perspectives 13, pp. 101-16. Faculté des
Lettres, Rabat.

Youssi, Abderrahim (1990) Lexical Processes in the Berber of the Media in Morocco. In
linguistics in the Maghreb, ed. Jochen Pleines, pp. 264-81. Rabat: Editions Okad.

361
Youssi, Abderrahim (1992) Grammaire et lexique de l’arabe marocain moderne.
Casablanca: Wallada.

Youssi, Abderrahim (1998) Fieldwork “Fallouts” and the Linguistic Bi-and Multilateral
Relations in the Phonematic Structure. In Langues et Littératures XVI. Faculté des
Lettres, Rabat.

Zec, Draga (1988) Sonority Constraints on Syllable Structure. Ph.D. dissertation.

Stanford, CA: Stanford University.

362