The Vowels of Urban Qatari Arabic

University of North Dakota
UND Scholarly Commons
Theses and Dissertations Theses, Dissertations, and Senior Projects
12-1-2020
The vowels of Urban Qatari Arabic

Mark Daniel Shockley
Follow this and additional works at: https://commons.und.edu/theses
Recommended Citation
Shockley, Mark Daniel, "The vowels of Urban Qatari Arabic" (2020). Theses and Dissertations. 3390.
https://commons.und.edu/theses/3390
This Thesis is brought to you for free and open access by the Theses, Dissertations, and Senior Projects at UND
Scholarly Commons. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator
of UND Scholarly Commons. For more information, please contact [email protected].
THE VOWELS OF URBAN QATARI ARABIC
by Mark Daniel Shockley
Bachelor of Arts, Sam Houston State University, 2009
A Thesis
Submitted to the Graduate Faculty
of the University of North Dakota
in partial fulfillment of the requirements
for the degree of
Master of Arts
Linguistics
Grand Forks, North Dakota
December
2020
This thesis, submitted by Mark Daniel Shockley in partial fulfillment of the require
ments for the Degree of Master of Arts from the University of North Dakota, has been
read by the Faculty Advisory Committee under whom the work has been done and is
hereby approved.
_______________________________________
Dr. Adam Baker
_______________________________________
Dr. John M. Clifton
_______________________________________
Dr. Mark E. Karan
This thesis is being submitted by the appointed advisory committee as having met all
of the requirements of the School of Graduate Studies at the University of North Dakota
and is hereby approved.
____________________________________
Chris Nelson
Dean of the School of Graduate Studies
____________________________________
Date
ii
PERMISSION
Title The Vowels of Urban Qatari Arabic
Department Linguistics
Degree Master of Arts
In presenting this thesis in partial fulfillment of the requirements for a graduate de
gree from the University of North Dakota, I agree that the library of this University
shall make it freely available for inspection. I further agree that permission for exten
sive copying for scholarly purposes may be granted by the professor who supervised my
thesis work or, in his absence, by the Chairperson of the department or the dean of the
School of Graduate Studies. It is understood that any copying or publication or other use
of this thesis or part thereof for financial gain shall not be allowed without my written
permission. It is also understood that due recognition shall be given to me and to the
University of North Dakota in any scholarly use which may be made of any material in
my thesis.
Mark Daniel Shockley
December 14, 2020
iii
ACKNOWLEDGMENTS
Many thanks to Dr. Adam Baker, my committee chair, whose patience and scientific
acumen made this project possible. Dr. Mark E. Karan and Dr. John M. Clifton provided
timely and detailed feedback that greatly improved the quality of this thesis.
Dr. Darwish AlEmadi, Acting Vice President for Research at Qatar University, pro
vided expert advice in preparing this study. Dr. Eiman Mustafawi, Associate Professor
in the Department of English Language and Linguistics at Qatar University, was very
generous in suggesting research questions and assisted me with recruiting participants.
I am grateful to my Arabic teachers in Qatar, the United Arab Emirates, and Oman, who
are too many to name here.
Participants in this study included Ali Albader, Muhammad alMurri, Buthaina Alkhu
laifi, Abdulla AlHajri, and Noora AlAnsari. I remain humbled by their generosity in
offering me their time, as well as their insights and intuitions on their language, which
certainly exceed the bounds of this study.
I owe a special thanks to my wife, Dr. Bethany Shockley, for her unwavering support
throughout this project.
iv
ABSTRACT
Urban Qatari Arabic is a variety of Gulf Arabic [afb] spoken by Qataris with tradi
tionally sedentary tribal backgrounds. This study examines phonetic and phonological
aspects of Urban Qatari Arabic vowels using acoustic phonetic data gathered in Qatar.
A new phonemic vowel inventory is proposed, including five long vowels and two short
vowels. This finding contradicts published studies on Gulf Arabic, which include three
or more short vowels; however, it is not unexpected when studies are compared from
nearby Arabic varieties, such as Najdi Arabic and Mesopotamian Arabic. The vowel in
ventory is also investigated using four linear mixedeffects regression models. In Gulf
Arabic, variation in short vowel backness is largely accounted for by the point of artic
ulation of adjacent consonants, as indicated by the statistical results. Wordfinal vowel
raising is found to be conditioned by the preceding vowel in Urban Qatari Arabic. Typ
ically, in other varieties of Arabic, wordfinal vowel raising is consonantconditioned.
A nuanced investigation of word stress, including interactions between stress and
epenthesis, demonstrates significant microlinguistic variation. Epenthetic vowels gen
erally can be stressed in Gulf Arabic, but for some speakers, stress assignment also
depends on which type of epenthesis occurs. Diachronic variation in Gulf Arabic word
stress is tending toward the elimination of stigmatized stress patterns and the introduc
tion of stress patterns from nonArabian dialects.
Acoustic correlates of word stress, including pitch, duration, intensity, and changes
in formants, are examined using the regression models. Pitch, duration and intensity
are found to be statistically significant markers of stress. An examination of formants
shows that contrasts in vowel quality are maximized in stressed vowels.
A phonological analysis of epenthesis shows that it is generally predicted by sonor
ity. Historical vowel deletion and synchronic epenthesis are demonstrated to be in a
counterfeeding relation, which has previously led to conflicting claims about word
final epenthesis. In threeconsonant clusters, epenthesis is expected after the first con
sonant, but sometimes epenthesis occurs after the second consonant or does not take
place. Although stress may be assigned to epenthetic vowels, epenthetic vowels dif
fer phonetically from other short vowels. Furthermore, postlexical epenthesis differs
phonetically from lexical epenthesis.
v
vi
Contents
1 Introduction 1
1.1 Problem and findings . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The state of Arabic dialectology . . . . . . . . . . . . . . . . . . . . . 2
1.3 Urban Qatari Arabic in context . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Classifying Arabic dialects . . . . . . . . . . . . . . . . . . . . 4
1.3.2 North Arabian dialects . . . . . . . . . . . . . . . . . . . . . . 5
1.3.3 Language contact in the Gulf . . . . . . . . . . . . . . . . . . . 7
1.3.4 Linguistic situation . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Existing research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.1 Researcher qualifications . . . . . . . . . . . . . . . . . . . . . 11
1.5.2 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.3 Interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.4 Stimuli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.5 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.6 Statistical methods . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.7 Supplemental data . . . . . . . . . . . . . . . . . . . . . . . . 16
2 Vowel inventory 19
2.1 Phonemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.1 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.2 Long vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.1.3 Short vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.4 Diphthongs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 The phonemic status of [u] in Gulf Arabic . . . . . . . . . . . . . . . . 27
2.2.1 Primary data . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.2 Minimal pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.3 Putative minimal pairs . . . . . . . . . . . . . . . . . . . . . . 29
2.2.4 Further evidence from functional load . . . . . . . . . . . . . . 30
2.2.5 Secondary data . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3 Allophones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.1 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.2 Consonantconditioned vowel backing . . . . . . . . . . . . . . 35
2.3.3 Mutual reinforcing of backing consonants . . . . . . . . . . . . 38
2.4 Wordfinal vowel raising (ʔimāla) . . . . . . . . . . . . . . . . . . . . 39
2.4.1 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
vii
3 Stress 47
3.1 Stress position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Epenthesisstress interaction . . . . . . . . . . . . . . . . . . . . . . . 50
3.3 Acoustic correlates of stress . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.1 Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.2 Duration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.3 Intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.4 Fundamental frequency . . . . . . . . . . . . . . . . . . . . . . 55
3.3.5 Formants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Epenthesis 59
4.1 Position of epenthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1.1 Epenthesis in wordfinal clusters . . . . . . . . . . . . . . . . . 59
4.1.2 Lexical variants . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1.3 Epenthesis at morpheme boundaries . . . . . . . . . . . . . . . 62
4.1.4 Consonant reduction and epenthesis . . . . . . . . . . . . . . . 63
4.2 Quality of epenthetic vowels . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Acoustics of epenthetic vowels . . . . . . . . . . . . . . . . . . . . . . 66
4.3.1 Underlying and epenthetic vowels . . . . . . . . . . . . . . . . 66
4.3.2 Lexical and postlexical epenthesis . . . . . . . . . . . . . . . 68
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
A Variation in stress assignment 85
B Wordlist 87
C Statistical results 93
C.1 Long vowel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
C.2 Short vowel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
C.3 Vowel raising model . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
C.4 Epenthetic vowel model . . . . . . . . . . . . . . . . . . . . . . . . . 96
D Supplemental figures 99
viii
List of Figures
1.1 Gulf Arabic in geographical context . . . . . . . . . . . . . . . . . . . 5

1.2 Qatari dialects within North Arabian dialects . . . . . . . . . . . . . . 6
1.3 Spectrogram of /ʔidāra/ [ḭdāra] ‘administration’ (Speaker 6) . . . . . . . 14
1.4 Spectrogram of /yiʃribūn/ [īʃrbōn] ‘they drink’ (Speaker 1) . . . . . . . 15
1.5 Spectrogram of /χirza/ [χrza] ‘a bead’ (Speaker 1) . . . . . . . . . . . . 15
2.1 Effect of list position on fundamental frequency, by speaker . . . . . . . 22

2.2 Effect of list position on intensity and duration . . . . . . . . . . . . . . 22
2.3 Effect of adjacent backing consonants on vowel backness (F2), by length 23
2.4 Formant scatterplots of long vowels . . . . . . . . . . . . . . . . . . . 24
2.5 Effect of adjacent backing consonants on short vowel backness (F2) . . 34
2.6 Formant scatterplots of short vowels . . . . . . . . . . . . . . . . . . . 35
2.7 Effect of preceding place of articulation on short vowel backness (F2) . 36
2.8 Wordfinal vowel raising . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.9 Wordfinal vowel raising, by consonant environment . . . . . . . . . . 41
3.1 Effect of stress and word position on vowel duration, by length . . . . . 54

3.2 Effect of stress on vowel duration, by vowel . . . . . . . . . . . . . . . 55
3.3 Effect of stress on vowel duration, by length . . . . . . . . . . . . . . . 55
3.4 Effect of stress on intensity . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5 Effect of stress on fundamental frequency . . . . . . . . . . . . . . . . 56
3.6 Effect of stress on fundamental frequency, by vowel . . . . . . . . . . . 57
3.7 Effect of stress on formant values, by allophone . . . . . . . . . . . . . 58
4.1 Statistically significant differences between lexical and epenthetic vowels 67

4.2 Intensity of epenthetic vowels, by vowel . . . . . . . . . . . . . . . . . 67
4.3 Statistically significant differences between lexical and postlexical epenthe
sis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
D.1 Effect of adjacent backing consonants on vowel height (F1), by length . 100
D.2 Effect of adjacent backing consonants on fundamental frequency, by
length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
D.3 Effect of adjacent backing consonants on vowel intensity, by length . . 100
D.4 Effect of adjacent backing consonants on vowel duration, by length . . 101
D.5 Effect of adjacent backing consonants on duration of short vowels, by
vowel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
D.6 Effect of adjacent backing consonants on fundamental frequency of short
vowels, by vowel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
D.7 Effect of adjacent backing consonants on intensity of short vowels, by
vowel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
ix
x
List of Tables
2.1 Vowel formants in Urban Qatari Arabic (speaker 1, male) . . . . . . . . 24

2.2 Vowel formants in Urban Qatari Arabic (speaker 6, female) . . . . . . . 24
2.3 Items including the allophone [u] in Urban Qatari Arabic . . . . . . . . 28
2.4 Variant transcriptions for *ħibb and *ħubb in North Arabian dialects . . 29
2.5 Complementary distribution of [i] and [u] in Gulf Arabic pronouns and
affixes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.1 Epenthesis by sonority sequence in Gulf Arabic . . . . . . . . . . . . . 60
A.1 Accounts of stress in North Arabian dialects . . . . . . . . . . . . . . . 86
B.1 Wordlist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
C.1 Statistical results for vowels . . . . . . . . . . . . . . . . . . . . . . . 94

C.2 Statistical results for short vowels . . . . . . . . . . . . . . . . . . . . 95
C.3 Statistical results for wordfinal vowel raising . . . . . . . . . . . . . . 96
C.4 Statistical results for epenthetic vowels . . . . . . . . . . . . . . . . . . 97
xi
xii
Chapter 1
Introduction
1.1 Problem and findings
Urban Qatari Arabic is the spoken language of Qatar’s traditionally sedentary tribes
and social groups, as well as the prestige variety of the nation. Urban Qatari Arabic
represents a subgroup of Gulf Arabic, which is a continuum of closely related dialects
that are spoken today along the eastern coast of the Arabian Peninsula.1 Gulf Arabic
is comprised of the urban dialects of the United Arab Emirates, Qatar, Bahrain (Sunni
speakers), northeastern Saudi Arabia, and Kuwait; it is also spoken in and around Bu
raimi, Oman, and in some coastal areas of Iran (Eberhard et al. 2020). These dialects
have a very high degree of mutual intelligibility, and differences in syntax, morphology,
and phonology are minor.
Most researchers of Gulf Arabic are focused on questions about morphology and
syntax. There are in addition several studies of phonetics, especially sociophonetics (al
Amadidhi 1985; Holes 1986, etc.). Phonology is given comparatively brief treatment
in the descriptive grammatical works of Qafisheh (1977), alRawi (1990), and Holes
(1990; 2016). A forthcoming volume on Emirati Arabic is likewise written by linguists
who specialize in syntax (Leung et al. 2020). This study seeks to bring acoustic pho
netic evidence to bear on phonological issues, including the vowel inventory, stress, and
epenthesis.
1
In some previous studies, the variety of Gulf Arabic spoken in Qatar has been called Qatari Arabic.
Previous research has shown that both Gulf Arabic and Najdi Arabic are spoken in Qatar. Therefore, the
term Urban is employed here for Qatari tribes and social groups that speak Gulf Arabic, as opposed to
the Bedouin groups that speak Najdi Arabic.
1
2 CHAPTER 1. INTRODUCTION
This study proposes a novel vowel inventory in which backness is not phonemically
contrastive in short vowels (§2.1, 2.2). This crucially requires a new description of the
effects of consonants on short vowels (§2.3). Vowel raising is also examined instrumen
tally in Urban Qatari Arabic for the first time, and its conditions are found to differ from
those described in any nearby dialect (§2.4).
Previous accounts of stress in Gulf Arabic have agreed in general but varied in some
details, as shown in Appendix A. This study found that diachronic changes in stress
position exhibit an ongoing pattern of dialect leveling (3.1). Stress and stressepenthesis
interaction are also found to vary from speaker to speaker (§3.2).
A detailed account of epenthesis in Gulf Arabic is provided from primary and sec
ondary data in Chapter 4. Synchronic epenthesis in Gulf Arabic may be explained with
reference to a sonority hierarchy with a few exceptions (§4.1.1). While other researchers
have written of free variation in epenthesis, this study found that lexical variants were
sometimes at play, and speakers readily employed either variant depending on register
(§4.1.2). The quality of epenthetic vowels is not necessarily predictable (§4.2). Even
though epenthesis usually precedes stress assignment, epenthetic vowels are found to
differ from lexical vowels in their acoustic properties (§4.3.1); types of epenthetic vow
els likewise differ phonetically from one another (§4.3.2). Suggestions for future re
search are given (§4.4).
1.2 The state of Arabic dialectology
This thesis addresses a felt need within Arabic dialectology for more detailed descrip
tion of the dialects of the Arabian Peninsula (Behnstedt & Woidich 2013:321). In doing
so, this thesis seeks to make a careful distinction between speech community and na
tion. One of the important shifts in contemporary Arabic dialectology is the shift away
from conflating geopolitical labels with Arabic dialects. In the early twentieth cen
tury, the community of speech considered was often either tribe or village (e.g., Feghali
1919). In the late twentieth century, the Western conception of Arabic dialects became
1.2. THE STATE OF ARABIC DIALECTOLOGY 3
increasingly preoccupied with the concepts of nation and capital. Reference grammars
appeared for Moroccan Arabic (Harrell 1962), Iraqi Arabic (Erwin 1963), Syrian Arabic
(Cowell 1964), and later, Gulf Arabic (Qafisheh 1977), Egyptian Arabic (AbdelMassih
et al. 1981), Yemeni Arabic (Qafisheh 1992), and Nigerian Arabic (Owens 1993). These
frequently suffer from the weakness of conflating nation and dialect; researchers write
(and Arabs speak) of “Egyptian Arabic” when they mean the dialect of certain social
classes in Cairo. Very few of the earlier modern descriptive grammars dealt with vari
ation beyond the national level, some notable exceptions being Shawkat’s grammar of
“educated Damascene Arabic” (1962), Blanc’s Communal dialects of Baghdad (1964),
and Khalafallah on Ṣaʕīdī Egyptian (1969).2
Increasingly, though, there are now serious variationist treatments of even smaller
areas—like de Jong’s work on Sinai’s Arabic dialects (de Jong 2000; 2011)—or full
volumes on the dialects spoken in single communities, like Grigore’s book on Mardīn
(Grigore 2007) and Lahdo’s descriptive grammar on Arabic in Tillo (Lahdo 2009). As
linguists focus more closely on smaller and smaller geographical areas, diversity contin
ues to emerge (Watson 2011a). Longdistance links have been discovered, for instance,
in a sparsely attested nominal linking suffix, found in Nigeria, the Gulf, and Central Asia
(Owens 2006). Nevertheless, the goal of compiling microlinguistic fieldwork on Arabic
into macrolinguistic descriptions that are comprehensive in geographic and social scope
remains incomplete. While detailed linguistic atlases have been completed in regions
like Egypt, North Yemen, and Syria, Behnstedt and Woidich point out the embarrassing
irony that the Arabian Peninsula itself is the most neglected area:
“On the core of the Arab world, the Arabian Peninsula, in particular
its northwestern area Ħid͡ʒāz, its southern parts ʕAsīr, Ħaḍramawt, and ʕU
mān, we have only very limited knowledge. And the same is true for the
Sudan and the subSaharan areas and for Libya, parts of Algeria, let alone
the “Sprachinseln”, such as those in Uzbekistan, Afghanistan, and Iran.
2
Arabic words are Romanized using the International Phonetic Alphabet with two alterations typical
in Arabic dialectology literature: 1) the underdot is used to signify “emphasis” in consonants with back
coarticulation; 2) long vowels are marked with a macron.
All these areas still need much more fieldwork.” (Behnstedt & Woidich
2013:321)
The same authors’ Wortatlas der Arabischen Dialekte represents a notable effort
towards a comprehensive atlas of Arabic dialects, but it is only as good as the primary
source material it is made up of, and this is particularly lacking in the Arabian Peninsula
(Behnstedt & Woidich 2010).
1.3 Urban Qatari Arabic in context
1.3.1 Classifying Arabic dialects
Although Arabic is popularly framed as a single language spoken by hundreds of mil

lions, it is in reality a vast continuum of spoken lects with varying mutual intelligibility.
These spoken varieties are linked together by widespread use of Modern Standard Ara
bic in certain media and religious domains.
Previous researchers have broadly classified Arabic dialects into Bedouintype and
sedentarytype groups (e.g., Rosenhouse 2006). This linguistic dichotomy is derived
historically from the Arabian vocational division between lbadū ‘Bedouin, nomadic
people’, who traditionally subsisted through a pastoral lifestyle, and lħað̣ar ‘seden
tary, urban dwellers’, who traditionally subsisted through trade and agriculture. In this
scheme, all Gulf Arabic varieties are Bedouintype varieties of Arabic because they are
spoken by tribes that were historically nomadic (Watson 2011a). Most of these tribes
settled in coastal towns in the eighteenth century. Within Qatar, the term badū refers to
tribes of relatively recent Najdi origin and dialect. These tribes settled in Qatar much
more recently and retain stronger connections to their pastoral heritage. For typologi
cal purposes, all Arabic dialects spoken in Qatar (whether Najdi Arabic or Gulf Arabic)
are classified as “Bedouin varieties” and may be usefully compared with other Bedouin
type varieties of Arabic. To avoid confusion, the term “Bedouintype” will be employed
in this study when referring to this dichotomy, and “Bedouin” will be used to refer to
recently nomadic tribes.
1.3. URBAN QATARI ARABIC IN CONTEXT 5
1.3.2 North Arabian dialects
Gulf Arabic (χalīd͡ʒī), of which Urban Qatari Arabic is a dialect, is spoken mainly in
coastal cities and villages from Kuwait to the United Arab Emirates. As seen in Figure
1.1, Gulf Arabic is in contact with Mesopotamian Arabic (the prestige dialect of Iraq,
“Iraqi Arabic” in cited sources), Najdi Arabic (in Kuwait, Qatar, and the United Arab
Emirates), Baħārna Arabic (in Bahrain), and Omani Arabic. Of these, Najdi Arabic,
Mesopotamian Arabic, and Gulf Arabic are all classified as Bedouintype dialects in
Arabic dialectology, and share a number of characteristics. Throughout this study, these
three dialect groups (Najdi, Mesopotamian, Gulf Arabic) will be referred to collectively
as “North Arabian dialects”, a term used by Ingham (1982:1).
Figure 1.1: Gulf Arabic in geographical context
Najdi Arabic, Mesopotamian Arabic, and Gulf Arabic have their origin in the Najd
region of the eastern Arabian Peninsula (see Figure 1.1). The ruling tribes of Kuwait
(ʔāl ṣabāħ), Bahrain (ʔāl χalīfa), Qatar (ʔāl θānī), and the United Arab Emirates (ʔāl
nahajjān) settled on the Gulf coast in the late eighteenth century (Johnstone 1967b:xxiii;
Fromherz 2017:53).3 The prestige varieties spoken in central and southern Iraq likewise
originated in waves of migration from the Najd, but in a much longer time frame (Ingham
3
A few tribes, such as alMohannadi in Qatar, claim an older origin than this.
1982:1; 1994:89). Gulf Arabic is understood, therefore, to be closely akin to Najdi

Arabic and Mesopotamian Arabic. Figure 1.2 illustrates this relationship.
North Arabian dialects
Mesopotamian Arabic Najdi Arabic Gulf Arabic
Northern Sunni Bahraini
NorthernCentral Urban Emirati
Central Urban Qatari
Southern kabasa
ʔāl murrah mahānida
banī hād͡ʒir ʔāl bū kuwāra
almanāṣīr banī kaʕb
almanāniʕa
alχulaifāt
salaṭa
Figure 1.2: Qatari dialects within North Arabian dialects

Blue = variety spoken in Qatar; green = name of tribe in Qatar
Baħārna Arabic and Omani Arabic are regarded as sedentarytype dialects, and di
verge significantly from the North Arabian dialects (Watson 2011b, Holes 2016, Shock
ley 2020). Ingham (1982) and Holes (2016) have categorized these dialects as “South
Arabian dialects”, a designation that also includes other Arabic varieties spoken in Oman
and Yemen. The world’s largest sand desert, the Empty Quarter (from its Arabic name,
arrubʕ alχālī), poses a considerable geographic barrier between North Arabian and
South Arabian dialects.
Alongside Gulf Arabic speakers, Qatar and the United Arab Emirates also contain
tens of thousands of members of Bedouin tribes with more recent Najdi origins. The
Bedouin tribes of the Arabian Peninsula settled in the twentieth century (Mandaville
1.3. URBAN QATARI ARABIC IN CONTEXT 7
2011). Thus, even though they are Bedouin in tribal affiliation, they are no longer no
madic. Nonetheless, they retain distinct dialect characteristics, neighborhoods, and so
cietal roles in Gulf communities (Ingham 1986). The dialect of the Āl Murrah tribe,
spoken in Saudi Arabia and Qatar, is categorized as a Southern Najdi dialect and is
strictly differentiated from Urban Qatari Arabic in this study (Ingham 1986). Accord
ing to Johnstone (1967b), the speech of the Banī Had͡ʒir tribe shares some characteristics
with Najdi Arabic and others with Gulf Arabic (see §1.4).
Other varieties of Arabic also coexist with Gulf Arabic. A pidgin variety of Gulf
Arabic is spoken by many South Asians (Smart 1990; Næss 2008). In addition, in and
around the Musandam Peninsula, Shiħħī Arabic is spoken by some thousands of Omanis
and Emiratis, and is widely divergent from any of the North Arabian or South Arabian
dialects described above (Lorimer 1908; Bernabela 2011). Kumzari, a West Iranian lan
guage, is also spoken in Musandam (Thomas 1930; van der Wal Anonby 2015; Al Jahd
hami 2016).
1.3.3 Language contact in the Gulf
In Gulf Arabic, there are Aramaic borrowings dating from preIslamic times, which are
not shared with Classical Arabic (Holes 2001; 2002). Lexical evidence given in Syr
iac ecclesiastical correspondence points to a form of Aramaic being spoken in the Bēt
Qaṭrāyē community of northeastern Arabia before the Islamic conquests, and an inscrip
tion found in Mleiha [Milēħa], Sharjah, U.A.E. corroborates the use of Aramaic there
(Contini 2003; Romeny 2014; Overlaet et al. 2016; cf. Kozah et al. 2021, forthcoming).
Holes admits that there remains much to be done in understanding substrate elements in
Arabian dialects (Holes 2001:xxix).
Gulf Arabic varieties have a large stock of nonSemitic loanwords resulting from
language contact with Persian, Portuguese (since 1507), and Turkish (from the mid
sixteenth century). More recently, Gulf Arabic has been in ongoing contact with Persian,
Urdu, and English (Holes 2001; 2002). In the early twentieth century when J. G. Lorimer
studied the Gulf, he described great ethnic and linguistic diversity in Dubai, Sharjah, and
Oman; however, he wrote that there were “no Indian or other British subjects anywhere
in Qatar” (Lorimer 1908:1535). This state of isolation changed very rapidly after the
discovery of oil (Johnstone 1967b:xxv). In Qatar, the proportion of residents who are
Qatari was only 11.6% according to a 2015 estimate (CIA World Factbook 2020).
1.3.4 Linguistic situation
Heavy immigration from other Arabicspeaking nations has had a profound affect on
the sociolinguistic situation in Qatari society. A key observation for framing the present
study is that Qataris and Emiratis are outnumbered in their cities by speakers of non
Arabian dialects.4 In spite of the Gulf’s prestige in other domains, its language is a
stigmatized, minority variety of Arabic, spoken by around 2% of all speakers of Arabic
(Eberhard et al. 2020). In Qatar, at least, the feeling is mutual; Johnstone (1967b) wrote
of “Qatar’s lack of respect for the “prestige” dialects” (p. xxix) of Arab immigrants
from outside the Arabian Peninsula. NonArabian dialects exert considerable influence
through media. In addition, the linguistic ideology of the region enforces diglossia,
teaching that vernacular Arabic is “not Arabic”, which does not encourage use of local
varieties (Ferguson 1959). Decades ago, Johnstone (1967b) wrote that Kuwaiti Arabic
was a disappearing dialect. This prediction was overly dire, but a pattern of dialect
leveling is observed nonetheless.
Today, Qatar ranks highest in the region in six out of seven measures of quality of
education, and aspires to become the education hub of the region (Alpen Capital 2018).
Qatari women are much more likely than Qatari men to pursue postsecondary education,
and bilingualism is likely more pervasive among women as a result (Ministry of Devel
opment Planning and Statistics 2017). Social changes introduced after the discovery
of oil and gas have reportedly disrupted the family structure enough to result in some
language attrition, and during interviews, it was a point of pride among participants to
speak Urban Qatari Arabic fluently. Among them, the Gulf Arabic variety with the
4
Confusingly, this has also resulted in several publications that putatively describe “Kuwaiti Arabic”,
“Qatari Arabic”, and “Emirati Arabic” that bear no relation to the variety of Arabic being described in
this thesis.
1.4. EXISTING RESEARCH 9
most prestige is the native dialect of a Qatari royal, known to one participant as llahd͡ʒa
lbayð̣āʔ ‘the white dialect’. In addition, another participant stated that there is a type
of cant or cryptolect used by many teenage Qataris to mislead those outside their peer
group.
1.4 Existing research
T. M. Johnstone’s Eastern Arabian dialect studies (1967b; data, 19581959) is a ground

breaking volume on Gulf Arabic, and offers notes on the dialects of Kuwait, Bahrain,
Qatar, Abu Dhabi, Dubai, and Buraimi. Throughout his work, Johnstone singles out
Qatari speech as the most conservative among Gulf Arabic dialects, describing it as hav
ing distinctive Najdi affinities. Recent research has pointed to an almost total absence
of the features that were regarded by Johnstone as distinctively Qatari among Gulf Ara
bic dialects (Bettega 2014; 2017; cf. Johnstone 1967b:41, 117). Antepenultimate stress
(e.g., ˈjinkisar ‘it is broken’) as observed by Johnstone in Qatar was likewise poorly
attested in this study, but regularly occurs in the Najdi dialects of Qatar’s Āl Murrah
tribe. Johnstone interviewed a man from the Banī Hād͡ʒir (Bedouin) tribe, stating that
his dialect “does not differ radically” from that of the capital (p. xxix). Moreover, he
describes Qataris generally as “the camelrearing aristocracy of the desert” (p. xxix),
though most Qatari tribes are made up of urbandwellers that have not been pastoral
for two centuries. As a trailblazer, Johnstone evidently held a flawed expectation of
linguistic uniformity in the Qatar Peninsula, and he did not identify the dialectal divide
between Bedouin and Urban Qataris (cf. alQenaie 2011:173). Therefore, his findings
are not taken as a useful index for the speech of today’s Urban Qataris.
Since Johnstone (1967b), Gulf Arabic has been described much more thoroughly.
Kuwaiti Arabic is very well documented by Ingham (1982; see AlBader 2015 for more
references). Working primarily in Abu Dhabi, Hamdi Qafisheh produced a reference
grammar (1977) and dictionary (1997). AlRawi’s dissertation (1990) is another very
useful study of the dialect of Abu Dhabi. AlTajir’s Language and linguistic origins
in Bahrain (1982) also addresses Bahrain’s unique linguistic milieu. Clive Holes gave
us a pedagogical work on Gulf Arabic (1984) as well as his own reference grammar
(1990), both based on extensive fieldwork in Bahrain. Feghali (2008) gives texts from
Gulf nationals with a brief grammatical treatment. Finally, Holes’ Dialect, culture, and
society in Eastern Arabia comprises, in three volumes, a glossary (2001), ethnographic
texts (2005), and a grammar (2016) describing the dialects of Bahrain with data from
the 1970s. Clearly, Qatar is the least represented in linguistic fieldwork among the Gulf
nations.
In addition to the above, many books have been published in Arabic, the most helpful
being’s Ħanð ̣al’s dictionary of Emirati Arabic (1998), Ħammūz’s dictionary (2008), and
alMālikī’s thorough study on loanwords in Qatar (2000). Numerous valuable books on
local culture have been published in Gulf countries, mostly in Arabic.
Qatar is sometimes regarded in international relations as the most isolated of the

Gulf nations. In his Gazetteer, Lorimer wrote of attempting to fill a lack of information
concerning Qatar, but even his map of Qatar’s coastline was very inaccurate (Lorimer
1908:1505). Despite its massive scope, Behnstedt & Woidich’s Wortatlas (2010) lacks
reliable primarysource data from Qatar. Owens’ A linguistic history of Arabic (2006),
which is the first modern scholarly attempt to trace the history of Arabic dialects as a
whole, likewise makes no mention of primarysource data from Qatar.
What follows is an account of primary sources on Arabic varieties spoken in Qatar.

Johnstone (1967b) did the first modern linguistic fieldwork in Qatar, but he somewhat
confounded the speech of Qatar’s Bedouin and Urban tribes. Unpublished dissertations
written by native speakers are the largest source of targeted research on Qatar’s language
(alAmadidhi 1985; Bukshaisha 1985; alMuhannadi 1991; alSulaiti 1993; Mustafawi
2006). Feghali has published texts in Qatari Arabic from fieldwork accomplished among
Qataris in the United States (Feghali 2008:315397). Mustafawi has written several
articles of theoretical phonology based on her mother tongue (Mustafawi & Mahfoudhi
2002; Mustafawi 2007; 2011). Bettega has recently written two articles on Urban Qatari
Arabic (Bettega 2014; 2017). Finally, Ingham (1986) includes two chapters about the
1.5. METHODS 11
Najdi dialect spoken by the Āl Murrah Bedouin tribe in Qatar.
1.5 Methods
1.5.1 Researcher qualifications
The principal researcher lived in Doha, Qatar from August 2014 to August 2016, in
Simaisma, Qatar from August 2016 to July 2018, and in Sharjah, United Arab Emirates
since August 2019. During this period, the principal researcher studied Modern Stan
dard Arabic parttime at the Translation and Interpreting Institute of Hamad Bin Khalifa
University (one semester, fall 2014), and fulltime at Qatar University (five semesters,
from January 2015 to May 2017). Intensive, immersive study at Qatar University was
conducted by the generous consent of the Arabic for NonNative Speakers Center. This
was followed by shorter periods of formal study of Gulf Arabic at the Gulf Arabic Pro
gramme in Buraimi, Oman (two weeks in summer 2017, intensive), and the American
University of Sharjah in the United Arab Emirates (one semester, fall 2019, parttime).
1.5.2 Subjects
Eight Qataris participated in this study. All participants were native speakers of Ara
bic and were raised in Arabicspeaking households in Qatar. Four men and four women
participated. They were recruited mainly through the author’s personal network and con
tacts at Qatar University. In terms of sociolinguistic grouping, two were from Bedouin
tribes (Najdi Arabic speakers), five were from sedentary tribes (Urban Qatari Arabic
speakers), and one was not affiliated with a tribe (an Urban Qatari Arabic speaker).5
Most were conversant with English, and a few were highly bilingual. For the purposes
of the acoustic study, only data from Qataris with sedentary tribal affiliation was used in
order to make the scope of the study more representative of Gulf Arabic. While Bedouin
5
In the Arabian Peninsula, including Qatar, not all Arabs have a tribal affiliation. According to Qataris
themselves, during modernization, many nontribal Qataris adopted new family names as a way of better
assimilating. Nontribal Arabs generally go unmentioned in previous linguistic studies. Like the majority
of Qataris, they are speakers of Gulf Arabic.
Qataris were interviewed, these interviews were not included in the acoustic study be
cause Āl Murrah Arabic has been classified as a Southern Najdi dialect (Ingham 1986;
1996).
1.5.3 Interviews
Three interviews were recorded in Doha for the purpose of preparing wordlists and
gathering information about Qatar’s linguistic situation. Two of these interviews were
recorded (a Najdi Arabic speaker, male, 30s; a Gulf Arabic speaker, male, 40s); the
third was not recorded because of time constraints (a Gulf Arabic speaker, female, 40s).
All recordings were made using a Tascam DR05 portable digital recorder with omni
directional microphones. Interviews were conducted in Gulf Arabic; the Najdi Arabic
speaker generally replied in English. The length of preliminary interviews ranged from
twenty minutes to one hour and a half. Preliminary interviews were useful in creating
stimuli that reflect Gulf Arabic as spoken in Qatar since published sources are lacking
in lexical data from Qatar. Participants also cleared up questions in relation to general
izations found in Johnstone (1967b), as discussed in §1.4.
1.5.4 Stimuli
Five elicitation sessions were recorded for the purpose of observing phonetic and phono
logical phenomena in Qatar’s dialects. Participants who completed the wordlist included
one Najdi Arabic speaker (male, 30s) and three Gulf Arabic speakers (two females, one
male, all between 20 and 40). All elicitation sessions were accompanied by followup
questions and discussion. During elicitation sessions, the author asked each participant
to pronounce five repetitions each of 223 words in a frame sentence.6 Only the target
words were included in the written stimuli. The frame sentence ͡tʃūf ___ lħīn ‘look _
__ now’ was adapted from Bukshaisha (1985:344). Most of the words elicited in this
frame were uninflected nouns, but no word class was excluded. Some of the resulting
sentences are awkward or ungrammatical; however, this frame enables comparison with
6
The wordlist is provided in Appendix B.
1.5. METHODS 13
previous studies. It also avoids elements of accommodated speech and high register.
In creating the wordlist, words were chosen to test a number of phonetic and phono
logical variables, including each vowel in the environment of back (pharyngeal or uvu
lar), emphatic, and nonemphatic consonants.7 Various syllable weight patterns were
included to thoroughly test for stress position. A number of words predicted to include
epenthesis were also included. Most of the words were chosen from Qafisheh’s Gulf
ArabicEnglish dictionary (1997), which was produced in Abu Dhabi and is the only
published dictionary in English that claims to be representative of Gulf Arabic as a
whole. After 207 words were chosen to test for each vowel in various environments,
16 random filler words were also added. During preliminary discussions, a few words
were omitted or exchanged to make them more locally representative of Urban Qatari
Arabic, as opposed to other varieties of Gulf Arabic. The wordlists were randomized for
each speaker. Because the Arabic writing system only optionally includes short vowels,
some explanations had to be furnished during elicitation sessions to avoid ambiguity
between homographs.8 Instructions were provided in Gulf Arabic whenever possible
during elicitation sessions.
1.5.5 Measurements
Two of the resulting corpora were chosen for the acoustic portion of the study. Both
speakers (Speaker 1, male; Speaker 6, female) were in their twenties and live in Doha.
These recordings were chosen because of their quality and completeness, and the speech
of both speakers was found to be representative of Urban Qatari Arabic as described in
7
In the study of spoken Arabic, “emphatic” usually refers to consonants with coarticulated dorsal or
pharyngeal constriction. In North Arabian dialects, they are usually treated as pharyngealized (Qafisheh
1977; Bukshaisha 1985; Ingham 1994) or as velarized (Holes 1990; 2006). Arabic varieties differ in
regard to the quantity and quality of their emphatic consonants, even within Qatar. In previous studies
of Gulf Arabic, “emphatic” consonants include /ṣ/, /ṭ/, /ð ̣/ and /ḷ/. Very rarely, emphatic bilabials /ḅ/ and
/ṃ/ have also been heard in Gulf Arabic, but these were not observed during this study. Emphatic con
sonants are perceptually distinguished from nonemphatic consonants by a lowering and backing effect
on surrounding vowels. Evidence presented in §2.3 indicates that uvulars /χ/, /ʁ/, and /q/ have a similar
effect. Tentatively, /ɡ/ (< *q) may trigger backing in Gulf Arabic, but this has not been well established.
8
During preliminary interviews, Bedouin Qataris (including one from the Āl Murrah tribe and another
from the Banī Hājir tribe) frequently pointed out homographs that were not known to Urban Qatari Arabic
speakers. The vocabulary of Urban Qataris had only minor discrepancies from published sources on Gulf
Arabic.
prior research, with no obvious interference from bilingualism or from other Arabic
varieties. These two interviews were segmented manually in Praat software to obtain
measurements of duration, fundamental frequency, intensity, as well as first, second, and
third formants. All segmentation was performed at positive zerocrossings of amplitude.
Following Baart (2010), segmentation was performed where transitions were detected in
amplitude, following Praat’s intensity trace as well as the waveform itself. If transitions
were not visible in amplitude, segmentation was made at a transition in the waveform.
Difficulties in segmentation were sometimes posed by breathy and creaky phonation.
The initial glottal stop /ʔ/ regularly surfaced as a creaky onset on the vowel, as shown
by the erroneously high pitch line in Figure 1.3. In general, the glottal fricative /h/ also
surfaced as [ɦ] when following a voiced segment. Both findings have been described
in Palestinian Arabic by Shahin (2011). Where no transition of amplitude was visible
in Praat, transitions in the waveform were used in segmentation. When a wordinitial
glottal stop was realized as a creaky vowel onset, silence often intervened between the
frame and the vowel onset, and vowels were measured after the silence.
Figure 1.3: Spectrogram of /ʔidāra/ [ḭdāra] ‘administration’ (Speaker 6)
Another special segmentation problem encountered in Urban Qatari Arabic was the
elision of short high vowels when adjacent to [ɾ]. Impressionistically, /yiʃribūn/ ‘they
drink’ was transcribed as [īʃiɾbūn] for some speakers. Indeed, this word is a prototypical
example of metathesis frequently encountered in the study of Arabian dialects (e.g.,
Holes 2005:215; Alkalesi 2006:143; cf. Johnstone 1967a:14); however upon closer
inspection, there is no vowel on either side of [ɾ], as seen in Figures 1.4 and 1.5.
After segmentation, Praat was used to extract measurements of duration, fundamen
1.5. METHODS 15
Figure 1.4: Spectrogram of /yiʃribūn/ [īʃrbōn] ‘they drink’ (Speaker 1)
Figure 1.5: Spectrogram of /χirza/ [χrza] ‘a bead’ (Speaker 1)
tal frequency, intensity, and formants. All vowel measurements were taken at 25%,
50%, and 75% of vowel duration. Most vowels, with the exception of creaky variants,
were found to be steadiest at the midpoint. Measurements were obtained from 1243
vowels for Speaker 1 (male) and 1780 vowels for Speaker 6 (female), yielding 3023
vowels in total. Formant measurements were extracted by Praat using standard settings
(five formants at 5500 Hz for female speakers; five formants at 5000 Hz for male speak
ers). These were supplemented by manual formant measurements where formants were
visible and distinct but Praat did not successfully extract measurements. For example,
in some cases, Praat detected one formant where two were close together but distinctly
visible. Statistical outliers were measured manually.
Pitch measurements were extracted using a range of 75 to 600 Hz. Praat was not able
to extract pitch measurements in several cases where creakyvoiced phonation was used.
In these cases, pitch measurements were obtained by manually measuring periodicity in
the waveform (Frequency (Hz) = Time (sec) / 1).
1.5.6 Statistical methods
Two batteries of repeatedmeasures (withinspeaker) analyses of variation (ANOVAs)

were performed using measurements of 2779 vowels obtained from elicitations from two
Urban Qatari Arabic speakers (Max & Onghena 1999). All ANOVAs were computed
using the linear mixedeffects regression test of the lmerTest package in R statistical
software.
The first battery of ANOVAs examined the characteristics of both long and short
vowels in Urban Qatari Arabic. Insufficient quantities of measurements were obtained
for two vowels (ō, ū), which had to be excluded from the model. This was the result
of a few lexical items that were not known to Qatari participants, as well as the unpre
dictability of these two vowels. High back rounded ū is generally avoided in unstressed
positions; it is frequently lowered to ō, especially near pharyngeal consonants or in
wordfinal position. Openmid ō was also sporadically realized as a diphthong aw or,
in one case, lowered to ā. In spite of this instability in long back vowels, minimal pairs
are numerous and phonemic contrast was maintained by all speakers. Full discussion of
findings from this model is given in §2.1.1 and 3.3.
A second battery of ANOVAs was performed using 1992 measurements of short

vowels, including both lexical and epenthetic vowels. These measurements are a sub
sample of the tokens used in the first battery. The results for the short vowel model are
discussed in §2.3.1, 3.3.2, and 4.3.1.
A third model was created to better understand the conditions of wordfinal vowel
raising, using 1144 tokens of /a/. This is discussed further in §2.4.1.
Another battery of ANOVAs was performed using 225 measurements of high epenthetic
vowels. This model is discussed in §4.3.2.
Full statistical results for all four models are given in Appendix C.
1.5.7 Supplemental data
In addition to data gathered directly from native speakers in Qatar, the author compiled
a lexical database of 13,400 Gulf Arabic stems from the best available sources for the
1.5. METHODS 17
purpose of collating variant forms and testing phonological variables. The most im
portant sources were Qafisheh (1997), Holes (2001), alRawi (1990), and Bukshaisha
(1985); thus, Urban Qatari Arabic data gathered for this thesis was compared to the
best sources on Gulf Arabic from neighboring countries. Variant forms were included
from Qafisheh (1977; 1996), Hoffiz (1995), alMālikī (2000), Mustafawi (2006), and
Feghali (2008). Some data was incorporated from articles (Mustafawi & Mahfoudhi
2002; Mustafawi 2011) and unpublished theses (alMuhannadi 1991; alSulaiti 1993).
Lexical data was gathered from important cultural publications (Lorimer 1908; Arabian
American Oil Company 1958). Finally, when of interest, comparative data was also
collated from linguistic studies in Mesopotamian Arabic (Woodhead et al. 1967) and
Najdi Arabic (Ingham 1994; Mandaville 2011).
Occasionally, data was also compared from a database of 6000 proper nouns in
the eastern Arabian Peninsula (Shockley 2020), gathered and analyzed from Lorimer
(1908). Thus, in the sections that make use of secondary data, generalizations are drawn
from around 19,000 lexical items from North Arabian dialects.
Chapter 2
Vowel inventory
2.1 Phonemes
2.1.1 Findings
Urban Qatari Arabic has seven phonemic vowels: five long vowels (/ā, ē, ī, ō, ū/) and
two short vowels (/a, i/). The number of long vowels is well established in the literature
on Gulf Arabic. The number of phonemic short vowels is more prone to controversy.
In particular, this study accounts for [u] as an allophone of /i/ (§2.2). In most modern
studies of Gulf Arabic, the phonemic vowel inventory is presented with five long vowels
(/ā, ē, ī, ō, ū/) and three short vowels (/a, i, u/) (Bukshaisha 1985:51; alRawi 1990:23;
Holes 1990:264265; alSulaiti 1993:8; Mustafawi 2006:8; Holes 2016:65). The short
allophones [e] (< /a/) and [o] (< /ō/), which are occasionally treated as phonemes, are
addressed in §2.1.3.
Vowels in North Arabian dialects are subject to a range of consonant effects which
are difficult to perceive without the use of instruments. For instance, Bukshaisha (1985:40
51) assigns three allophones to each respective vowel: one quality adjacent to pharyn
geal consonants, one quality adjacent to emphatic consonants, and one quality when
neither is present. Other researchers on Gulf Arabic have written that coronal conso
nants may have a raising effect (§2.4). Labial consonants have a backing and rounding
effect (Holes 2016:6667). There is also the issue of hundreds of points of variance in
19
20 CHAPTER 2. VOWEL INVENTORY
the transcription of short vowels in secondary data. In some studies these approach free
variation (overlapping distributions), and in others they approach allophony (comple
mentary distributions). This putative variation in short vowels has not been treated using
acoustic phonetic data. Clearly, the effects of consonants on vowels are too numerous
and finegrained for the researcher’s perceptual threshold. Therefore, a multivariate sta
tistical model appeared to be a promising way of approaching vowel allophones.
To test this account of Urban Qatari Arabic vowels, a number of statistical tests
were performed using 2779 tokens from two speakers (§1.5). Four regression models
were created. The first model, described here, investigated long and short vowels. The
second model investigated short vowels and is described in §2.3.1. The third model
tested for correlates of wordfinal raising of /a/ (§2.4). The fourth model compared
types of epenthesis and is described in §4.3.2.
In the long vowel model, the continuous dependent variables were F1, F2, funda
mental frequency (F0), intensity, and duration.
List intonation was accounted for by the factor variable ListPosition {Initial, Me
dial, Final}.
Vowel was coded as a factor variable with seven levels: five long vowels (/ā, ē, ī,
ō, ū/), and two short vowels (/a, i/). This coding is based on the foregoing analysis of
the phonemic inventory of Urban Qatari Arabic. Short and long vowels were coded
separately rather than in length pairs because, in spite of transcription practices, they
differ significantly in quality (Bukshaisha 1985). Long /ā/ is a back [ɑː], and short /a/
is usually central [ɐ] or front [æ]. Additionally, short vowels interact differently with
consonant context, as will be shown.
Two dichotomous factor variables BackingConsBefore and BackingConsAfter were

used to capture the presence or absence of immediately adjacent labial, emphatic, or
uvular consonants. The effects of labial and emphatic consonants are well documented
(Blanc 1964; Johnstone 1967b; Bukshaisha 1985; Holes 2016). The backing effect
of uvulars is attested in other Arabic dialects, as well as crosslinguistically (Yasin &
Owens 1984; Herzallah 1990:170; Bessell 1991; Choi 1991; Lahdo 2009:69; Abdure
2.1. PHONEMES 21
him 2014; see Bessell 1998:56 for further references). Because these backing effects
are mainly noted in short vowels in previous studies, the model also tested for twoway
interactions between vowel and backing consonant before, as well as vowel and backing
consonant after.1
The dichotomous variable Stress {Stressed, Unstressed} was also introduced as an

independent variable. The model also tested for interaction between vowel and stress
since previous research has found that stress affects long and short vowels differently
(Bukshaisha 1985).
Five withinspeaker ANOVAs were performed. The dependent variables are F1,
F2, fundamental frequency (F0), intensity, and duration. The independent variables
are those listed after the equals sign, separated by the plus sign. The asterisk indicates
multiplication of variables, which tests for their interactions.
lmer([ depvar ] = (ListPosition + Vowel + Stress + BackingConsBefore + Backing

ConsAfter + Vowel * Stress + Vowel * BackingConsBefore + Vowel * BackingConsAfter)
+ (1|Speaker))
The main effect of the vowel variable was statistically significant in all five tests,
which justifies this treatment of Urban Qatari Arabic vowels. List position was found to
have a statistically significant effect on F0. The effect of list position on pitch is shown
in Figure 2.1. Final repetitions had lower pitch for Speaker 1 (male), and higher pitch
for Speaker 6 (female). List position also had statistically significant effects on intensity
and duration, shown in Figure 2.2.
The dichotomous variable for the presence of backing consonants before the vowel
had a significant main effect on F2; likewise backing consonant after had a significant
effect on F2. Overall, the presence or absence of backing consonants affected vowel
backness as measured at the midpoint, for both long vowels and short vowels.
Consonant backing effects were found to be quite complex. Vowel and the presence
of a backing consonant before had a statistically significant interaction on F1, F2, inten
sity, and duration. Vowel and a backing consonant after the vowel also had a statistically
1
This model tests for the effects of adjacent consonants on vowel quality (cf. Blanc 1964); however,
emphasis spread is shown elsewhere to affect the quality of nonadjacent vowels (see §2.3.2).
Figure 2.1: Effect of list position on fundamental frequency, by speaker
Figure 2.2: Effect of list position on intensity and duration
significant twoway interaction on F2, F0, intensity, and duration. Overall, these inter
actions indicate that consonant environment has a more salient effect on short vowels
2.1. PHONEMES 23
than on long vowels. In Figure 2.3, the effect of both backing environments is additive
in long vowels, but it is multiplicative in short vowels.
Figure 2.3: Effect of adjacent backing consonants on vowel backness (F2), by length
Not all of the statistical results given above are relevant to the major findings of
this chapter; however, additional boxplots relating to consonant effects are given in
Appendix D.
Results of this model pertaining to stress are discussed in §3.3.
Mean formant values for all vowels are given in Tables 2.1 and 2.2.
Table 2.1: Vowel formants in Urban Qatari Arabic (speaker 1, male)
Vowel Mean F1 (Hz) St Dev Mean F2 (Hz) St Dev

ā 690 82 1246 127
ē 450 44 1722 141
ī 394 50 1867 203
ō 495 63 1080 125
ū 435 65 984 214
a 565 102 1418 202
i 412 50 1466 297
Table 2.2: Vowel formants in Urban Qatari Arabic (speaker 6, female)
Vowel Mean F1 (Hz) St Dev Mean F2 (Hz) St Dev

ā 813 74 1406 292
ē 570 70 2380 119
ī 434 52 2586 181
ō 594 88 1040 161
ū 444 76 977 182
a 741 78 1856 220
i 506 66 1917 402
2.1.2 Long vowels
Scatterplots of long vowel measurements for both speakers are shown in Figure 2.4. The
long vowels occupy the periphery of the vowel space and tend to be more stable than
the short vowels.
Figure 2.4: Formant scatterplots of long vowels
There is clear overlapping distribution between the front vowels /ī/ and /ē/, and be
tween the back vowels /ū/ and /ō/. Both of these overlaps appear to be caused by the
2.1. PHONEMES 25
lowering of high vowels near emphatic or pharyngeal consonants. Phonemic contrast

was maintained by all speakers, as shown in (1) and (2).
(1) a. ʕōd ‘great’
b. ʕūd ‘incense’2
(2) a. sēf ‘sword’
b. sīf ‘beach’
Lowering of /ū/ toward [ō] appeared to be more sporadic and unpredictable.
2.1.3 Short vowels
In alveolar contexts, both short vowels are front or nearfront. Backness is not phone
mically distinctive in short vowels, but is determined by phonological environment, as
described in §2.3.2. The status of [i] and [u] is dealt with in detail in §2.2.
The short closemid vowel [e] (< /a/) of wordfinal vowel raising is marginally
phonemic in Urban Qatari Arabic. Hypothetically, minimal pairs between [a] and [e]
are possible based on the data collected during this study.
(3) a. sidre ‘a lotus tree’
b. sidra ‘his lotus trees’ (unattested)
Minimal contrasts between [e] and [i] are not obtainable for all speakers since short
[i] only surfaces wordfinally for some speakers. Minimal pairs demonstrating either
opposition ([a] vs. [e], [e] vs. [i]) were not elicited in this study because the complexity
of wordfinal vowel raising was not anticipated by previous treatments. Therefore, [e]
is regarded as marginally phonemic in Gulf Arabic; it may be phonemic for certain
speakers and not others. This nuanced phenomenon is described in detail in §2.4.
Some authors include a short vowel [o], only appearing wordfinally, where it is
in complementary distribution with long /ō/ (Johnstone 1967b:22; Qafisheh 1977:15;
2
See Ingham 1996:101 on the status of these vowels among Qatar’s Najdi Arabic speakers.
Hoffiz 1995:54; cf. Blanc 1964:30). In this study, final vowels were not uniformly
shortened, which allows the possibility of minimal pairs for stress position, as also seen
in Hassan (1981).
(4) a. /ɡālō/ → ˈɡālō ‘they said’ (in phrase)
b. /ɡālōa/ → ɡāˈlō ‘they said it’
2.1.4 Diphthongs
The expected Gulf Arabic cognates of Classical Arabic *aj and *aw are the long, close
mid vowels [ē] and [ō], respectively. The diphthong [aw] and the long vowel [ō] alter
nate according to phrase position (cf. Prochazka 1988:18):
(5) a. tid͡ʒamˈɦarō ‘they gather together’ (In phrase)
b. tid͡ʒamˈɦaraw ‘they gather together’ (In isolation or phrasefinal)
Monophthongization is much less common when /w/ or /j/ is a root consonant, and
was only observed in one speaker:
(6) mōð ̣ūʕ ‘topic’ [ √wð ̣ʕ ] (Speaker 1)
In (7), three different vowel variants were observed. This variation has no clear
explanation.
(7) ṣālōn ~ ṣālawn ~ ṣālān ‘salon; broth’
As with /w/ and /j/, when the semivowel [j] is the result of lenition of /d͡ʒ/, [aj] is usu
ally maintained, as in majlis ‘a sitting room for guests’ but mēlis has been documented
elsewhere (Qafisheh 1996:79; 1997:52).
In (8), the semivowel [w] was maintained by all speakers:
(8) dawrāt ‘courses’ [ √dwr ]3
3
Compare the minimal pair dawra ‘course’ and dōra ‘rotation’ in Qafisheh 1997:234.
2.2. THE PHONEMIC STATUS OF [U] IN GULF ARABIC 27
Overall, contrasting forms like these suggest that the diphthongs and closemid vow
els are capable of contrast for most speakers, and monophthongization is an ongoing but
unstable process in Urban Qatari Arabic (on the instability of diphthongs in Arabic di
alects, see Ferguson 1957; Johnstone 1967b:25, Morano 2019:65).
One educated female Qatari never used the vowel [ē], resulting in an asymmetrical
vowel inventory: /ā, ay, ī, ō, ū, a, i/. Although some older speakers of Baħārna Ara
bic use diphthongs [ay] and [aw] over [ē] and [ō] (Holes 2016:166), frequent use of
diphthongs may also be the result of standardization.
2.2 The phonemic status of [u] in Gulf Arabic
2.2.1 Primary data
Recall that this study treats [i], [ɨ] and [u] as allophones of a single phoneme. Because
this claim differs from published studies on Gulf Arabic, further remarks are necessary to
establish the relation between the data presented here and that of previous studies. From
the wordlist used here (223 words), 11 wordlist items were consistently pronounced
with a short, high back rounded vowel [u] in Urban Qatari Arabic, and all of them were
in phonological contexts demonstrated to cause backing in short vowels, as described
in §2.3. These items are given in Table 2.3 in narrow transcription, with comparison
from Qafisheh’s dictionary (1997, broad transcription) from Abu Dhabi, as well as the
corresponding literary forms used in Modern Standard Arabic (Wehr 1979, final case
suffixes omitted).
2.2.2 Minimal pairs
Eight minimal pairs have been claimed for [i] and [u] in Gulf dialects. Five of them were
contradicted in primary data from Urban Qatari Arabic; some were also contradicted in
secondary data from other Gulf Arabic dialects. The most important pairs will be treated
first.
Table 2.3: Items including the allophone [u] in Urban Qatari Arabic
Urban Qatari Arabic Gulf Standard Gloss

buɡɑr baɡar baqar ‘cow’
bunṣɨr bunṣur binṣir ‘ring finger’
buṣɑl baṣal baṣal ‘onion’
ɡɑˈbuḷna ɡabilna qablunā ‘before us’
(ә)ħṭubo ħṭuba ħaṭaba ‘a piece of wood’
ṣubɑʕ ṣubiʕ ʔiṣbaʕ ‘finger’
ṣufɨr ṣufur ṣafr, ṣifr, ṣufr ‘zero’
ṭubb ṭibb ṭibb ‘medicine’
ʔuˈχitha ʔuχutta ʔuχtuha ‘her sister’
χuˈbizkum — χubzukum ‘your (pl.) bread’
χunṣɨr χunṣur χinṣir ‘little finger’
(9) a. ħubb ‘love’
b. ħibb ‘large jar for water’
This is a minimal pair that is frequently referenced in studies of Gulf Arabic (al
Rawi 1990:23; Qafisheh 1997:115116; Holes 2001:99100). The word ħibb ‘large jar
for water’ is somewhat archaic, though, so the following minimal pair has also been
used (Bukshaisha 1985:40; cf. de Jong 2011:29):
(10) a. ħubb ‘love’
b. ħibb ‘love!’
Urban Qatari Arabic speakers claimed that these were not homophones; but during
the acoustic portion of the study, both words in (10) were elicited and pronounced as
[ħɨbb], with similar formant values.
As Owens (2006) points out, ħibb and ħubb were not in sustained contrast in Classi
cal Arabic. There is likewise evidence presented in Table 2.4 that they are not in contrast
in contemporary North Arabian dialects. Words in bold differ from Modern Standard
Arabic, exemplified here by Wehr. It is possible that in previous studies the central
quality of the vowel causes difficulty for the researcher, who often uses Classical Ara
bic as a baseline, and therefore anticipates front [i] or back rounded [u]. Meissner alone
captures the central unrounded [ɨ] in his transcriptions, and his transcriptions here are
corroborated more than 100 years later in data from Qatar.
Table 2.4: Variant transcriptions for *ħibb and *ħubb in North Arabian dialects
Source Dialect ‘love’ ‘love!’ ‘water jar’

Ibn Manẓūr (1883/13th c.) Classical ħubb~ħibb — ħubb
Wehr (1979:179) Modern Standard ħubb — ħibb
Wetzstein (1868:86) Najdi ħibba ‘a kiss’ ħibb —
Meissner (1903:117) Mesopotamian ħubb~ħɨbb — ħɨbb
Landberg (1920:333) Najdi — ħubb —
Johnstone (1967b:74, 80) Gulf — ħibb ħibb
Qafisheh (1997:23) Gulf ħubb ħibb ħibb
alRawi (1990:23) Abu Dhabi ħubb — ħibb
Holes (2001:99100) Bahrain ħubb — ħibb
Hashemi (2015:263) Abu Dhabi ħibb — —
Shockley Qatari (Urban) ħɨbb ħɨbb —
Sulaiti (1993:137138) offers two other minimal pairs, given below.
(11) a. ṭibb ‘study of medicine’
b. ṭubb ‘jump!’
This contrast was directly contradicted by Urban Qatari Arabic data; three Urban
Qatari Arabic speakers consistently pronounced /ṭibb/ ‘medicine’ as [ṭubb]. Holes (1990:264)
gives the first word in narrow transcription as [ṭәḅḅ]; Bukshaisha (1985:44) gives the
second word in narrow transcription as [ṭɔbb].
(12) a. bidd ‘distribute’
b. budd ‘need’
The second word is transcribed badd, bidd, and budd by Holes (2001:31; cf. Holes
2016:295).
2.2.3 Putative minimal pairs
Qafisheh’s dictionary (1997) includes a few putative minimal pairs, given here in cita
tion form.
(13) a. zibda ‘butter’ (p. 289)
b. zubda ‘gist’ (p. 289)

The form in (13b) is the Modern Standard Arabic pronunciation, unattested in other
Gulf Arabic research. Qatari participants stated that (13a) and (13b) were the same
word, zubda being merely the Standard pronunciation. Data elicited in Urban Qatari
Arabic includes [zibde] and [zɨbde] but not [zubda].
(14) a. ħusun ‘excellence’ (p. 137, cf. Woodhead et al. 1967:102)
b. ħisin ‘beauty’ (p. 137, cf. Woodhead et al. 1967:102)
For meaning (14a), Qafisheh gives the phrase liħusn lħað̣ð ̣ ‘fortunately’; Hashemi
(2015:159) gives the same phrase as liħisin lħað̣ð.̣
(15) a. luwa ‘to bend’ (p. 545)
b. liwa ‘major general’ (p. 545)
For the first meaning, Holes gives lawa (Holes 2001:486); the second word liwa is
a formal title (cf. Modern Standard liwāʕ).
(16) a. luʁa ‘language’ (p. 545)
b. liʁa ‘to talk incessantly’ (p. 545)
Both [luʁa] and [liqa] were recorded in Qatar during interviews. Holes has liʁa for
the first meaning and laʁa for the second (Holes 2001:479480).
2.2.4 Further evidence from functional load
A full treatment of functional load is beyond the scope of this study, but it will suffice to
say that there is no morphophonological position in which [i] and [u] are found in sus
tained contrast in Gulf Arabic (cf. Johnstone 1967a:4; Abboud 1979:468). Conditioned
variation is exhibited in every morphological category. In pronouns and pronominal af
fixes, [u] is only observed adjacent to [m], and [i] adjacent to [n], as shown in Table 2.5,
with data from Qafisheh (1977:160). Forms given with an asterisk are Classical Arabic
cognates.
Table 2.5: Complementary distribution of [i] and [u] in Gulf Arabic pronouns and affixes
hum ‘3p masc. pl. pron.’ hin < *hunna ‘3p fem. pl. pron.’
ʔintum ‘2p masc. pl. pron.’ ʔintin < *ʔantunna ‘2p fem. pl. pron.’
kum ‘2p masc. pl. poss.’ kin < *kunna ‘2p fem. pl. poss.’
hum1 ‘3p masc. pl. poss.’ hin < *hunna ‘3p fem. pl. poss.’
tum ‘2p masc. pl. subj. ag.’ tin < *tunna ‘2p fem. pl. subj. ag.’
in < *un ‘nominal linking suffix’
1
See examples of him in Johnstone 1967b:27, Qafisheh 1997:557, Feghali
2008:60.
2.2.5 Secondary data
Generalizability
It is established in what precedes that [i] and [u] are not phonemically distinct for some
Urban Qatari Arabic speakers. On historical and typological grounds, this is not a sur
prising discovery. However, the external validity of these findings may be questioned.
Therefore, in this section, the lexical data of three previous researchers (Bukshaisha
1985; Qafisheh 1997; Mustafawi 2006) is thoroughly compared to the model of short
vowel allophony given here.
Bukshaisha (1985), Urban Qatari Arabic
Bukshaisha (1985:51) gives F2 for [i] and [u] in “nonemphatic context” as 1750 Hz
and 1000 Hz respectively. But her dataset only contains [u] in nonemphatic context in
two words, both with labial consonants: dub ‘fat’ (p. 546) and tuf ‘phew!’ (p. 499).
In both words, the “falling transition” she describes (p. 245) is undoubtedly the effect
of the labial context on the vowel. Thus, in Bukshaisha’s data (1985:478507) [u] only
occurs in backing environments, in complementary distribution with [i].
Qafisheh (1997), Urban Emirati Arabic
Qafisheh’s dictionary (1997) includes [u] in 781 stems (7.6% of 10,238). Following the
analysis presented here, 547 (70% of 781 stems containing [u]) are adjacent to backing
consonants. Of the remaining stems, a further 109 have documented variants in which
[i] and [u] alternate, many of them in Qafisheh’s data (1996; 1997). This leaves only
105 stems (1% of 10,238 stems) in which [u] appears to be unconditioned by consonant
environment. Many of them are formal or semiformal words, e.g. rusul ‘apostles,
messengers’, suʔāl ‘question’, ruʔasa ‘presidents’, taχarrud͡ʒ ‘graduation’. It is likely
that these transcriptions are influenced by Classical Arabic, as is the case with most
of the minimal pairs in §2.2.2. All of them belong to morphophonological patterns in
which [u] alternates with [i] elsewhere, as noted in §2.2.4. For comparison, in the same
dictionary, 3943 stems (38.5% of 10,238 stems) include [i].
Qafisheh also transcribes at least 65 examples of [i] where [u] is expected, e.g. ṭibb
‘medicine’, yṣibb ‘he pours’. These are also generally contradicted elsewhere. In §2.2.2
and 2.2.3 above, five minimal pairs in Qafisheh’s data are analyzed. Three were contra
dicted in primary data, and the other two are contradicted elsewhere in secondary data
on Gulf Arabic.
Mustafawi (2006), Urban Qatari Arabic
Data from Mustafawi (2006:165296) anticipates a lack of phonemic contrast between

[i], [ɨ], and [u], but her topic is consonant variation, and thus no vowel inventory is
given. Among thousands of Urban Qatari Arabic forms in narrow transcription, only
26 items have the vowel quality [u], generally near bilabials or emphatics.4 The usual
reflexes of /i/ are [i] and [ɨ], conditioned by consonant context.
2.3 Allophones
2.3.1 Findings
A second model was created to test variables that only pertain to short vowels (/a/, /i/).
The dependent variables were the same variables as in the previous model: F1, F2, pitch,
intensity, and duration.
All of the independent variables from the long vowel model (§2.1.1) were included:
4
Based on the analysis presented here, the sole exception, among 131 pages of narrowly transcribed
Urban Qatari Arabic paradigms, is hudʒūm ‘attacking’ (v.n.) (p. 289).
2.3. ALLOPHONES 33
list position, vowel, stress, backing consonant before, and backing consonant after. Pre
vious research has suggested an interaction between labials and emphatics for /i/ (Holes
2016:138). Therefore, the model includes a test for a twoway interaction between back
ing consonant before and backing consonant after, as well as a test for a threeway in
teraction between vowel, backing consonant before, and backing consonant after.
Other variables were also needed to explore phonological relationships suggested

in previous research. Holes previously claimed that the short high vowels were more
distinctive in closed syllables in Sunni Bahraini Arabic (Holes 2006; cf. Yasin & Owens
1984:205; de Jong 2011:29), but this statement was later mitigated (Holes 2016). To
assess the earlier claim, the present model tested for the main effect of SyllableType
{Open, Closed} as well as a test for an interaction between syllable type and vowel.
Vowel duration in Arabic dialects is generally sensitive to word position (McCarthy

2005). A dichotomous variable WordPosition {Final, NonFinal} was included in the
short vowel model to control for the effect of wordfinality.
The model also uses the variable Underlying {Underlying, Epenthetic} to test for
differences between lexical vowels and underlying vowels (Gouskova & Hall 2009; Hall
2013). Tests were also performed for the interaction of Underlying with vowel. The
model equation is given here, with the asterisk representing the interaction of two or
more variables.
lmer([ depvar ] = (ListPosition + Vowel + Stress + Underlying + BackingCons

Before + BackingConsAfter + SyllableType + WordPosition + Vowel * Stress + Vowel
* Underlying + Vowel * BackingConsBefore + Vowel * BackingConsAfter + Backing
ConsBefore * BackingConsAfter + Vowel * BackingConsBefore * BackingConsAfter +
Vowel * SyllableType) + (1|Speaker))
Five withinspeaker ANOVAs were performed, one for each of the continuous de
pendent variables: F1, F2, pitch, intensity, and duration.
As expected, vowel had a statistically significant effect on F1 but not F2. The pres
ence of a backing consonant before a vowel had a statistically significant main effect on
F1 and F2. The presence of a backing consonant after likewise had a statistically signifi
cant main effect on F2, showing that consonant environment is an important determiner
of both height and backness in short vowels.
For the dependent variable F2, there was also a statistically significant twoway
interaction of vowel and the presence of a backing consonant before the vowel, as well
as vowel and the presence of a backing consonant after the vowel. This indicates that
the effect of consonant environment differs between /a/ and /i/, as shown in Figure 2.5.
Furthermore, for the dependent variable F2, there was a statistically significant two
way interaction of a backing consonant before a vowel and a backing consonant after a
vowel.
Figure 2.5: Effect of adjacent backing consonants on short vowel backness (F2)
The threeway interaction of vowel, backing consonant before, and backing conso
nant after was also statistically significant for dependent variables F1, intensity, and du
ration. These effects are not relevant to the main claims of the thesis, but the graphs are
available in Appendix D for the reader who is interested in the phonetic detail. The same
interaction was not statistically significant for F0 or F2. The effect of a second back
ing consonant has a compound effect on short vowels in Urban Qatari Arabic. These
results suggest that short vowels are extremely prone to the effects of surrounding con
sonants in North Arabian dialects, and phonetic studies should examine vowel quality at
the level of the syllable minimally. Short vowel formants for both speakers are shown
below in Figure 2.6. This scatterplot shows compelling evidence that there are only two
phonemic short vowels. There is no discrete boundary between backed and nonbacked
2.3. ALLOPHONES 35
vowels since other factors are at play, such as longdistance coarticulation effects.
Figure 2.6: Formant scatterplots of short vowels

Aqua = backing environments; magenta = nonbacking environments
Results related to stress and word position in the short vowel model are discussed in
§3.3.2. Acoustic differences between lexical and epenthetic vowels are given in §4.3.1.
2.3.2 Consonantconditioned vowel backing
In short and epenthetic vowels, backness is largely determined by preceding and fol
lowing consonants.
The short low phoneme /a/ is frequently realized as [ɐ], but varies by consonant con
text. It is fronted and raised toward [ɛ] when following velar or postalveolar consonants.
It is backed and lowered toward [ɑ] when uvular or emphatic consonants are adjacent.
For simplicity, this study transcribes front or central allophones as [a] and backed allo
phones as [ɑ]. In the feminine singular suffix, it may be raised to a closemid vowel [e],
[ә], or [o].
The short high phoneme /i/ is nearfront [ɪ] or central [ɨ] when backing consonants
are not present, but usually central [ɨ] when a backing consonant is adjacent. /i/ is further
backed and rounded to [u] when adjacent to more than one backing consonant (§2.3.3).
When /i/ is adjacent to the palatal glide /j/ (as in the imperfect verbal prefix ji), there
is no distinguishable transition between the two phones; rather, they merge to [ī], as
already shown in Figure 1.4. This suggests four distinct allophones [i], [ɪ], [ɨ], and [u],
but the distributions of these allophones do overlap.
Consonant effects on short vowels are complex and gradient, posing a challenge
for phonological representation which utilizes a necessarily finite number of discrete
categories (Cohn 2006; McCollum 2018). The effect of the passive place of articulation
of the preceding consonant on short vowel backness is shown in Figure 2.7.
Figure 2.7: Effect of preceding place of articulation on short vowel backness (F2)
While pharyngeals did have a slight backing effect, it was not as salient as that of
uvulars or emphatics. There are also coarticulation effects that extend across multi
ple syllables, and this model only distinguishes backing effects in adjacent consonants.
For these reasons, there is noticeable overlapping distribution between backed and non
backed short vowels, and these allophones cannot be treated as fully discrete categories.
Short vowels are nearfront when no backing consonants are present:
(17) a. dass ‘he hid’
b. kanz ‘treasure’
(18) a. θilθ ‘onethird’
b. ḭðin ‘ear’
2.3. ALLOPHONES 37
c. rid͡ʒūʕ ‘returning (v.n.)’
Labials have a backing and rounding effect on short vowels (Johnstone 1967b; Holes
2016:66), which is most pronounced when a bilabial consonant follows /i/:
(19) /ħibbna/ → [ħɨbbna] ‘our love’
The effect of labiodental [f] on /a/ is not often noticeable:
(20) /daftar/ → [daftar] ‘notebook’
The effect of emphatic consonants on vowel quality is well documented in Arabic.

All short vowels, whether underlying or epenthetic, are backed when preceding or fol
lowing emphatic consonants.
(21) a. /baṭal/ → [bɑṭɑḷ] ‘champion’
b. /saṭl/ → [ṣɑṭɑḷ] ‘bucket’
(22) a. /ṣibaʕ/ → [ṣubɑʕ] ‘finger’
b. /ṣikk/ → [ṣɨkk] ‘close (imp.)’
Emphasis spread in Urban Qatari Arabic is thoroughly documented by Bukshaisha,

who notes that it can spread in both directions, across syllable and word boundaries,
for up to six segments (Bukshaisha 1985:220). The effect of emphatic consonants on
vowels crosses syllable boundaries in many words:
(23) a. /ṣūra/ → [ṣūrә] ‘picture’
b. /rṭūba/ → [rṭūbә] ‘humidity’
Uvulars were found to have a backing effect even more salient than that of emphatics
or labials.
(24) a. /diχūl/ → [dɨχūl] ‘entrance’
b. /θaqb/ → [θɑqub] ‘hole’
In this study, uvulars are grouped together with emphatic consonants in Urban Qatari
Arabic.5 The same effect was noted for the velar /ɡ/ (< *q) in Urban Qatari Arabic.
(25) a. /ɡallib/ → [ɡɑḷḷub] ‘turn over!’ (Bukshaisha 1985:507)
b. /ɡabl/ → [ɡɑbuḷ] ‘before’6
Holes gave (25b) as an example of a word that is “emphatised when no emphatic

consonant is present” in Bahraini dialects. It is possible that Gulf Arabic /ɡ/ (< *q) is
underlyingly uvular, as is convincingly argued in a recent study on Palestinian Arabic
(Abo Mokh & Davis 2020).
Holes also gives two examples of unexpected emphasis near uvulars: /tχammar/
[ṭχɑṃṃɑṛ] ‘to ferment’, and /χtarab/ [χṭɑṛɑḅ] ‘to go rotten’ (Holes 2016:7879). Quranic
reading tradition has classed uvulars and emphatics together as alħurūf almustaʕlija
‘heavy letters’ since the twelfth century (Zamakhshari 1859:190; cf. Herzallah 1990:170).
This justifies the position that uvulars and emphatics are closely related. In Gulf Arabic,
coarticulation is perhaps maintained in /ɡ/, though no longer uvular.
2.3.3 Mutual reinforcing of backing consonants
The presence of more than one backing consonant has a mutual reinforcing effect in Gulf
Arabic. The twoway interaction of backing consonant before and backing consonant
after was statistically significant in the short vowel model. In Figure 2.5, this effect
is seen where a short vowel is both preceded and followed by a backing consonant (a
labial, emphatic, or uvular consonant).
5
Further research on coarticulation could better establish this position, and this is being pursued by
Noora AlAnsari.
6
The extent of emphasis spread in this word is used by locals as a shibboleth distinguishing Urban
Qatari and Bedouin Qatari speech. One Bedouin Qatari juxtaposed Urban [ɡɑbuḷ] with Bedouin [ɡɑbil]
‘before’. Emphasis does spread across a syllable boundary from [ɡ] in the Āl Murrah Arabic word [ɡɑʃrɑː]
‘insipid’.
2.4. WORDFINAL VOWEL RAISING (ʔIMĀLA) 39
When a short vowel is adjacent to a labial or uvular consonant and in emphatic

context, the backing effect is multiplied as seen in (26) (cf. de Jong 2011:29; Holes
2016:138).
(26) a. /ṭibb/ → [ṭubb] ‘medicine’
b. /ħṭiba/ → [ħṭubo] ‘a piece of firewood’
As noted, emphasis spread crosses syllable boundaries in both directions. In exam

ple (27), the initial uvular has a backing effect on /i/, which is further compounded by
the emphatic /ṣ/ in the following syllable.
(27) /χinṣir/ → [χunṣɨr] ‘little finger’
The same compound backing is seen when velar [ɡ] is near a labial consonant:
(28) a. /ɡabl/ → [ɡɑbuḷ] ‘before (prep.)’
b. /baɡar/ → [buɡɑr] ‘cow’
2.4 Wordfinal vowel raising (ʔimāla)
2.4.1 Findings
In Urban Qatari Arabic, the feminine singular suffix a is raised to [e] or [i] when it
follows an underlying high vowel.
(29) a. /ʃjara/ → [iʃjara] ‘tree’
b. /ʕɡida/ → [ʕɡideɦ] ‘knot; a popular Yemeni dish’
Scant attention is paid to wordfinal vowel raising in published grammars of Gulf

Arabic. It is usually believed to be conditioned by the preceding consonant, as is the case
in several nonArabian varieties.7 However, this study is the first to explore Gulf Arabic
7
Vowel raising is claimed to be consonantconditioned in Gulf Arabic by alSulaiti (1993:158),
Mustafawi (2006:163), and Feghali (2008:14); this does not explain numerous examples in Bukshaisha
(1985), alRawi (1990), and alMālikī (2000). References on consonantconditioned vowel raising in
Levantine dialects are given in Durand (2011) and Abo Mokh & Davis (2020).
vowel raising with acoustic phonetic data. Raising of low vowels is known in Arabic
linguistics as ʔimāla (lit., “tilting” or “leaning”), and takes place in many varieties of
Arabic. In Najdi Arabic, the feminine singular suffix generally surfaces as [e] or [i], but
vowel raising does not apply equally to other morphemes (Johnstone 1967a; Abboud
1979:489). It is likewise lexically conditioned in Urban Qatari Arabic.
A third model was designed using only observations for the phoneme /a/ to con
firm these observations about vowel raising. The dependent variable was F1. Vari
ables included ListPosition{Initial, Medial, Final}, Stress, Underlying{Underlying,
Epenthetic}, BackingConsBefore, BackingConsAfter, SyllableType{Open, Closed},
WordPosition{Final, Nonfinal}. For reasons explained in §2.1.1, the model tested for
the interaction of backing consonant before and backing consonant after.
Two dichotomous variables were also introduced to capture the unique conditions
of wordfinal vowel raising: AfterHighVowel and FeminineSingular. In total, 1144
measurements were included in this test.
lmer(F1 = (ListPosition + Stress + Underlying + BackingConsBefore + Backing

ConsAfter + SyllableType + WordPosition + AfterHighVowel + FeminineSingular +
BackingConsBefore * BackingConsAfter + AfterHighVowel * FeminineSingular) +
(1|Speaker))
There was a statistically significant twoway interaction of the feminine singular

suffix and a preceding (underlying) high vowel on F1. The results suggest that word
final vowel raising is both phonologically and lexically conditioned among Urban Qatari
Arabic speakers. Figure 2.8 shows raised vowel quality (lowered F1) where both con
ditions are met. This result holds whether the suffix follows a coronal consonant or not,
as shown in Figure 2.9.
2.4.2 Discussion
The two speakers studied differed in the degree of vowel raising, as seen in their short
vowel scatterplots in Figure 2.6.
Figure 2.8: Wordfinal vowel raising
Figure 2.9: Wordfinal vowel raising, by consonant environment

(30) a. /ħmisa/ → [ħmisi] ‘a turtle’ (Speaker 1)
b. /ħmisa/ → [ħmise] ‘a turtle’ (Speaker 6)
Stemfinal long vowels are typically shortened in Arabic dialects (McCarthy 2005).
This also varied in this study, with one speaker shortening final long ī and the other
maintaining stemfinal vowel length.
(31) a. /ʕālī/ → [ʕālī] ‘high’ (Speaker 1)
b. /ʕālī/ → [ʕāli] ‘high’ (Speaker 6)
It therefore appears that Gulf Arabic speakers who shorten final /ī/ will not fully
raise the /a/ of the feminine singular suffix. Certain varieties have undergone a chain
shift in which final vowel shortening (/ī/ → [i]) has partially suppressed vowel raising
(/a/ → [e]). In other varieties, final vowel length is retained (/ī/ → [ī]), and therefore
vowel raising is unrestrained (/a/ → [i]). If the vowel of the feminine singular suffix
were considered to be phonemic, it could only be phonemic for speakers of the first
type.
Since the raised vowel of the feminine singular suffix differs from both [a] and [i]
for some speakers, it is transcribed here as [e].
(32) /wriɡat/ → [wriɡet] ‘a piece of paper; playing cards’ (construct form)
(33) /brika/ → [(u)brike] ‘blessing’
In Urban Qatari Arabic, vowel raising occurs after both open and closed syllables
(cf. Holes 2006):
(34) /zibdat/ → [zibdet] ‘gist; butter’ (construct form)
In example (35), vowelraising is the effect of a high vowel that is no longer present
in the surface form.
(35) /ʁāriɡa/ → [ʁārɡe] ‘drowned‘
In North Arabian dialects, underlying /a/ is frequently raised in nonfinal open syl
lables (Johnstone 1967a; Holes 2006). When a preceding high vowel is underlyingly an
/a/, no vowel raising is present.
(36) /maktaba/ → [maktiba] ‘library’
In words where the feminine singular suffix follows a low vowel, the wordfinal
vowel quality is usually maintained:
(37) a. /ʔidāra/ → [ḭdāra] ’administration’
b. /ʃjara/ → [iʃjara] ‘tree’
In the present study, there is also no vowel raising where a wordfinal a follows a
high vowel but is not from the feminine singular suffix (again, unlike Sunni Bahraini
Arabic):
(38) a. /ɡablna/ → [ɡabuḷna] ‘before us’
b. /ħibbna/ → [ħɨbbna] ‘our love’
c. /xitha/ → [χɨtta] ‘her sister’
Like all short vowels in Urban Qatari Arabic, the raised vowel can also be backed
when in the environment of backing consonants:
(39) a. /ħṭiba/ → [ɨħṭubo] ‘a piece of firewood’
b. /rɡiba/ → [ɨrɡubo] ‘neck’
In the eighth century (A.D.), Sībawayh recorded that wordfinal vowel raising was
not usually conditioned by the adjacency of a high vowel, as it was in wordmedial
vowel raising (Levin 1992). However, in the data presented here, the preceding vowel
is the conditioning factor (cf. Owens 1998:59; BarMoshe 2019:4950). This differs
significantly from the analyses of vowel raising in Gulf Arabic by alSulaiti (1993),
Mustafawi (2006), Feghali (2008), and Holes (2016). Future research will determine
if this finding is unique to the speech of Qataris or is also found in other Gulf Arabic
varieties (alMaṭlabī 1978:130).
2.5 Conclusion
Most linguistic studies of Gulf dialects are written in broad transcription without acous
tic phonetic research, making it difficult to make any definitive statement about the
distribution of short vowels. The research presented in this chapter from an acoustic
basis includes four findings that differ from previous research on Gulf Arabic:
1. Backness is not phonemic in short vowels, and the short high vowels [i] and [u]
were not found to be phonemically distinct.
2. This study found the short vowels /a/ and /i/ to be central or nearfront in quality.
3. Uvular consonants have a noticeable backing effect on all short vowels, which is
compounded when labial or emphatic consonants are present.
4. Wordfinal vowel raising is conditioned by a preceding high vowel.
Of these findings, the first is a topic that has been much discussed across Arabic
dialects. Many Arabic dialects have been found to lack phonemic contrast in short high
vowels in the following dialect groups, all of which are classed as Bedouintype dialects:
1. North Mesopotamian dialects in Iraq, Syria, and Turkey (Khan 1997; Grigore
2007; Lahdo 2009);
2. Mesopotamian dialects spoken in southern Iraq and Khuzistan (Ingham 1982;

1994);
3. Najdi dialects spoken in ‘Anaiza (Lehn 1967), Riyadh, Sudair, and Hā’il (Proc
hazka 1988);
2.5. CONCLUSION 45
4. Gulf dialects spoken in Kuwait (Ingham 1982), Hofuf (Prochazka 1988), and
Qatar (as presented above);
5. Some dialects spoken in the Levant, including those of the West Bank (Herzallah
1990) and Beirut (Sakr 2019).
In addition to the above references, one frequently encounters the attenuated state
ment that the distribution of [i] and [u] is “virtually complementary” (Holes 2016:67).
In his book on Najdi Arabic, Ingham (1994) writes that “there is a certain distributional
overlapping” and “a great deal of nondistinctive variation” (p. 14), but treats them as
separate phonemes (cf. de Jong 2011:30). Therefore, it would be in keeping with oth
ers’ research to say that the instability of these two vowels is a feature shared by most
North Arabian and Mesopotamian varieties of Arabic.
The North Mesopotamian dialects of Iraq, Syria, and Turkey were brought to that
region immediately following the Islamic conquests (Owens 2006:274). Ingham (1982)
writes that “the area of northeastern Arabia and the river valleys of Iraq and Khuzistan
have from the beginning of the Christian era been subject to immigration from the central
Arabian peninsula.” (p. 1) It is possible that dialects of both type, with and without
phonemic contrast in high short vowels, have existed since before the divergence of the
North Mesopotamian dialect group, as contested by Owens (2006) and corroborated by
this review.
If this phoneme merger precedes this migration pattern, then it is very old indeed, and
is possibly a question that should be more effectively brought to bear in North Arabian
dialects, and in Arabic dialects as a whole.
Chapter 3
Stress
3.1 Stress position
Although stress was not discussed by early grammarians of Arabic, all dialects of Ara
bic, including Gulf Arabic, exhibit word stress dictated by syllable weight and position
(Watson 2011c). In Gulf Arabic, like many Arabic dialects, the wordfinal mora is extra
metrical for purposes of stress assignment. Mora here refers to either a final consonant
or the length on a long vowel. In nonfinal syllables, Cv and vC are light syllables, and
all others are heavy. In final syllables, Cv̄ and CvC are also light and do not attract
stress.
Detailed data on stress assignment was gathered for three Urban Qatari Arabic speak
ers during elicitation sessions. One interview also yielded conversational data in Gulf
Arabic sufficient for a detailed analysis of stress. The following is a summary of stress
assignment as observed in Urban Qatari Arabic.
1) If the final syllable is heavy (CvCC or Cv̄C), stress is final as exemplified in (40).
(40) a. kiˈtabt ‘I/she wrote’
b. riˈd͡ʒūʕ ‘return (n.)’
c. ṣīˈbān ‘nits (coll. n.)’
d. yawʕāˈnīn ‘hungry (pl.)’
47
48 CHAPTER 3. STRESS
2) If the final syllable is light and the penultimate syllable is heavy, stress is penul
timate.
(41) a. ˈnaħla ‘bee’
b. fōˈɡānī ‘upper’
c. ɡɑˈbuḷna ‘before us’
3) If the final two syllables are both light, and the antepenultimate syllable is heavy,
stress varies between penultimate and antepenultimate (cf. Johnstone 1967b:111; Holes
1984:319).
(42) ˈmaktaba ~ makˈtiba ‘library’
During this study, penultimate stress in such words appeared to be more representa
tive of Urban Qatari Arabic as a whole (contra Johnstone 1967b). Thus, variants anal
ogous to ˈmaktaba are taken as formal variants (note the lack of vowel raising), and
makˈtiba is taken as representative of an older phase of the language. This is consistent
with the earliest available evidence on Gulf Arabic stress, which includes penultimate
stress both in heavylightlight words (barˈṭufī, ‘a reef in Bahrain’, Lorimer 1908:213)
and in words of three light syllables (kiˈbisa ‘a tribe’, Lorimer 1908:1035).
4) If the final three syllables are light, stress is antepenultimate. This primarily
occurs in formal borrowings, which were often preferred by female speakers during
elicitations.
(43) a. ˈbaraka ‘blessing’
b. raˈzaɡa=nī ‘he (God) granted me’
Penultimate stress in strings of three light syllables (here written as CvCvCv) was
recorded in earlier sources on North Arabian dialects, but is not observed in newer stud
ies (references in Appendix A). This is also consistent with the elision of the unstressed
antepenultimate syllable in light syllable strings (CvCvCv → CCvCv).
The definite article proclitic has no effect on stress assignment (cf. ilHazmy 1975;
Watson 2011c):
3.1. STRESS POSITION 49
(44) a. /l=mara/ → [il=ˈmɨra] ‘my wife’ (lit., ‘the woman’)
b. /l=hid͡ʒin/ → [il=ˈhid͡ʒin] ‘the racing camels’
In Gulf Arabic, suffixes and enclitics are always taken into account when assigning
word stress:
(45) a. raˈzaɡa=nī ‘he granted me’
b. muħīˈṭāt ‘oceans’
c. dīˈret=na ‘our homeland’
Prepositional adjuncts, such as those formed from li ‘to, for’, are frequently attached
to the verb in Gulf Arabic. These are taken into account during stress assignment:
(46) a. tintɨˈbɨh=lih ‘pay attention to it’
b. ʔɑṣɑlˈlɨħ=lik ‘I correct for you’
In (47c), the doubleobject construction leads to the possibility of a phonological

word with four heavy syllables, two of which receive stress.
(47) a. ˈʔā ktib

1S write
‘I write’
b. ʔā kˈtib= li k

1S write= for 2SG.M
‘I write for you’
c. ʔā kˌtib= li k= ˈjā h

1S write= for 2SG.M IO 3SG.M
‘I write it for you’

3.2 Epenthesisstress interaction
In this section, speakerspecific variation in the handling of stress in words with epenthe
sis is described. In all studies of Gulf Arabic, when the penultimate syllable is heavy, it
receives the stress. All Urban Qatari Arabic speakers stressed the penultimate syllable
in words of this pattern.
(48) /χibz=kim/ → [χuˈbiz=kum] ‘your (pl.) bread’
Thus, in words with a lightheavylight syllable pattern, epenthesis precedes stress

assignment.
Underlying /bint=na/
Epenthesis binit=na
Stress Assignment biˈnit=na
Surface [biˈnit=na]
In words with a heavylightlight syllable pattern, stress assignment varies from

speaker to speaker. In general, Urban Qatari Arabic speakers pronounced these with
penultimate stress (contra Johnstone 1967b):
(49) a. mak.ˈti.ba ‘library’
b. jin.ˈki.sar ‘it is broken’
c. ti.d͡ʒam.ˈɦa.raw ‘they crowded around’
In Johnstone (1967b) and Holes (1990), epenthetic vowels are ignored during stress
assignment in words with a heavylightlight syllable pattern, e.g. ˈkillә=na ‘all of us’.
This was observed for one Urban Qatari Arabic speaker; thus, for some speakers of Gulf
Arabic, there are two types of epenthesis. The first epenthesis rule is incorporated into
stress assignment, and the second (postlexical) epenthesis rule is not.
Underlying /kill=na/ /bint=na/

Epenthesis 1 — biˈnit=na
3.2. EPENTHESISSTRESS INTERACTION 51
Stress Assignment ˈkill=na biˈnit=na

Epenthesis 2 ˈkillә=na —
Surface [ˈkillә=na] [biˈnit=na]
Results on this point were far from uniform. Of four Urban Qatari Arabic speakers,
only one followed the pattern given in Johnstone (1967b) and Holes (1990). In the
speech of Speaker 6, underlying HLL words received penultimate stress as in (50a) and
(50b), but epenthesized HLL words received antepenultimate stress as in (50c).
(50) a. mak.ˈti.ba ‘library’
b. jin.ˈki.sar ‘it is broken’
c. ˈṭaɡ.ɡә.ni ‘he hit me’
Speaker 1 (male) pronounced all HLL words with penultimate stress, regardless of
epenthesis:
(51) a. jin.ˈki.sar ‘it is broken’
b. ti.d͡ʒam.ˈha.raw ‘they crowded around’
c. ṭaɡ.ˈɡә.ni ‘he hit me’
The variant in (51c), with penultimate stress on the epenthesized vowel, is charac
teristic of nonArabian dialects such as Cairene Arabic. It is mentioned in passing by
Qafisheh, who notes that it is unusual and degemination is expected in “normal speech”
(Qafisheh 1977:35).
Speaker 3 (male, 40s, nontribal) exhibited the same pattern in spontaneous speech
(impressionistic transcription):
(52) ʕinˈdu=hum ‘at their house’1
Speaker 4 (female, 30s) pronounced monomorphemic HLL words with antepenulti

mate stress as in (53a), but pronounced multimorphemic HLL words with stress on the
penultimate (epenthetic) syllable as in (53b):
1
The [u] here may be an example of hypercorrection, as elsewhere the same speaker distinctly said
kʊlma ‘word’ rather than kilma or kilme.
(53) a. ˈmak.ta.ba ‘library’
b. ħab.ˈbә.ha ‘he loved her’
The results concerning stress and stressepenthesis interaction in Urban Qatari Ara
bic contradict the geographic generalizations made by Johnstone (1967b), a pattern con
tinuing from recent research by Bettega. The degree of variation observed in stress
epenthesis interaction also has not been described in prior studies on Gulf Arabic and
may indicate ongoing stress shift. Speakerspecific variation in stress assignment ap
pears to be shifting toward forms that are used in nonArabian dialects. In the past 50
years, penultimate stress in lightsyllable strings has already disappeared from North
Arabian dialects due to the influence of standardization.
3.3 Acoustic correlates of stress
3.3.1 Findings
This section discusses five continuous variables with statistically significant correlations
with stress in Urban Qatari Arabic. Four of these are examined acoustically for the first
time in a Gulf Arabic dialect. This study uses measurements obtained from two Urban
Qatari Arabic speakers for 2779 vowels. The statistical model is described in §2.1.1,
and full results are given in Appendix C. The short vowel model described in §2.3.1 is
also used to better understand the effect of stress on duration. The dependent variables
tested include duration, fundamental frequency, and intensity, as well as first and second
formants.
Since Fry’s (1955) study on stress in American English, numerous acoustic stud
ies have found duration, fundamental frequency, intensity, changes in formants, and
spectral tilt to be meaningful acoustic correlates of phonological stress in a multitude of
languages. Studies along this vein are also available in several Arabic dialects. Buk
shaisha (1985) investigated duration as a marker of stress in Urban Qatari Arabic, but
found mixed results (discussed in §3.3.2). In his grammar of Gulf Arabic, Holes claims
3.3. ACOUSTIC CORRELATES OF STRESS 53
that stress is marked by consonant release (Holes 1990:272). The present study, how
ever, only measured the effect of stress on vowels. Hassan (1981), in a study of south
ern Mesopotamian Arabic, found duration, fundamental frequency, and intensity to be
statistically significant markers of stress.2 In a study of formal spoken Arabic, alAni
(1992) found that stress was marked by increases in intensity and duration, but not fun
damental frequency.
3.3.2 Duration
A withinsubjects ANOVA was performed on vowel measurements to test the effect of

stress and other variables on duration. In the long vowel model, there was a signifi
cant main effect of stress on duration, as well as a significant twoway interaction of
vowel and stress on duration, indicating that stress does not affect duration equally for
all vowels. Stressed long vowels are markedly longer in duration than unstressed long
vowels.
Stress was also included in the short vowel model, which controlled for word posi
tion (§2.3.1). In the short vowel model, both word position and stress had a statistically
significant main effect on duration. As presented in Figure 3.1, stress results in increased
duration when word position is accounted for.
In Figure 3.2, the lengthening effect of stress on individual vowels is shown. Very
few measurements were extracted for unstressed [ū] because this vowel is generally
restricted to stressed positions in Gulf Arabic stems.
This finding differs from that of Bukshaisha (1985). In Bukshaisha’s study, stress
only affected duration for long vowels (Bukshaisha 1985:433). However, Bukshaisha
did not control for word position. As Figure 3.3 shows, the interaction of stress and
word position results in the false impression that short vowel duration is unaffected by
stress.
Duration is highly sensitive to word position in Urban Qatari Arabic, as in other
Arabic dialects (alAni 1992; McCarthy 2005). When word position is controlled for,
2
Hassan writes that he is from Basrah, Iraq, but identifies his dialect as both “Baghdadi” and “Iraqi
Spoken Arabic” (Hassan 1981:2).
Figure 3.1: Effect of stress and word position on vowel duration, by length
stress increases duration for both short and long vowels in Urban Qatari Arabic.
In general, duration is a less reliable marker of stress in studies where word position
is accounted for (Gordon & Roettger 2017). However, in this case, accounting for word
position made duration a more reliable marker of stress. Likewise, in Hassan’s (1981)
experiment in Mesopotamian Arabic, which carefully accounted for word position by
utilizing minimal pairs, duration was a reliable marker of stress.
3.3.3 Intensity
A withinsubjects ANOVA was performed on vowels to test the effect of stress and
other variables on intensity. There was a significant main effect of stress on intensity,
with stressed vowels having a higher intensity. The interaction of vowel and stress on
intensity was not statistically significant. Intensity was found to be a reliable marker of
stress in Urban Qatari Arabic, as shown in Figure 3.4.
Figure 3.2: Effect of stress on vowel duration, by vowel
Figure 3.3: Effect of stress on vowel duration, by length
3.3.4 Fundamental frequency
A withinsubjects ANOVA was performed on vowels to test the effect of stress and other
variables on fundamental frequency. There was a significant main effect of stress on fun
damental frequency in the long vowel model. In Urban Qatari Arabic, stressed vowels
generally have a higher fundamental frequency than their unstressed counterparts. The
overall effect of stress on fundamental frequency is shown in Figure 3.5.
Figure 3.4: Effect of stress on intensity
Figure 3.5: Effect of stress on fundamental frequency
There was also a significant twoway interaction of vowel and stress on fundamental
frequency. The effect of stress on fundamental frequency is not equal across all vowels;
however, as seen in Figure 3.6, each vowel has a higher pitch in stressed position.
As noted in southern Mesopotamian Arabic by Hassan (1981), fundamental fre

quency appears to be an important marker of phonological stress in Urban Qatari Ara
bic.
Figure 3.6: Effect of stress on fundamental frequency, by vowel
3.3.5 Formants
In addition to the above stress correlates, changes in formants were observed. Cross
linguistically, it is often expected that unstressed vowels are more centralized (Gordon
& Roettger 2017). In Urban Qatari Arabic, that is generally true but not always the case.
In the long vowel model, stress had a statistically significant main effect on both F1
and F2. The twoway interaction of vowel and stress likewise yielded significance for
both F1 and F2.
In the short vowel model, stress had a statistically significant main effect on F1, but
the main effect of stress on F2 was not significant. Twoway interactions of vowel and
stress yielded significance for both F1 and F2. This indicates that the effect of stress
on formants varies across vowels. The mean first and second formant values for each
allophone are shown in the vowel plot in Figure 3.7, with the effect of stress indicated
by the arrow.
The closemid vowels /ē/ and /ō/ are both lowered in stressed positions, which serves
to maximize contrast with high vowels /ī/ and /ū/. The effect of stress on the allophones
of /i/ are similar: [i] is backed in stressed position, but backed allophones [ɨ] and [u] are
both slightly less back when in stressed position. Thus, all three allophones are pushed
approximately toward a retracted [i], maximizing contrast with long [ī] and [ū]. This
Figure 3.7: Effect of stress on formant values, by allophone
result would be very unexpected under the prevailing belief that [i] and [u] are short
phonemes in Gulf Arabic corresponding to long [ī] and [ū], in which case [i] should be
fronted in stressed position, and [u] should be backed.
Taking all this into account, unstressed vowels are not always centralized; rather,
stressed vowels exhibit increased quality contrast.
3.3.6 Conclusion
In conclusion, duration is only found to be a reliable marker of stress in Urban Qatari

Arabic when word position is controlled for. Intensity was a much more salient acoustic
marker of stress, with fundamental frequency being the most salient. Stressed vowels
also exhibit increased quality contrast. These are novel findings in Gulf Arabic, but they
do not differ from what has been found in other North Arabian dialects.
Chapter 4
Epenthesis
4.1 Position of epenthesis
4.1.1 Epenthesis in wordfinal clusters
In Arabic varieties, including Urban Qatari Arabic, wordfinal CC clusters are fre
quently broken up by epenthesis. In (54a), the final cluster is broken by epenthesis,
but in (54b), the addition of the pronominal suffix ī makes epenthesis unnecessary:
(54) a. ɡɑbuḷ ‘before’
b. ɡɑbḷī ‘before me’
Words that are underlyingly disyllabic do not reduce when followed by a vowel
initial suffix as in (55b) . This confirms that the alternation in (54) results from epenthe
sis and not deletion.
(55) a. ħað ̣ar ‘urban, sedentary people’
b. ħað ̣arī ‘urban, sedentary (adj.)’
In Urban Qatari Arabic, wordfinal epenthesis generally happens when sonority is

increasing in the final consonant cluster, as shown below (cf. Clements 1990).
Glide (y, w) > Rhotic (r) > Nasals (m, n), Liquids (l, ḷ) > Fricatives > Affricates,
Stops
59
60 CHAPTER 4. EPENTHESIS
Examples of each pairing in this sonority hierarchy are given below in Table 4.1
(data, Qafisheh 1997). Words in bold are epenthesized forms, all of which alternate
with unepenthesized forms when followed by a vowelinitial suffix in Qafisheh (1997).
Fields marked with an asterisk are unattested.
Table 4.1: Epenthesis by sonority sequence in Gulf Arabic
Rhotic Nasals Liquids Fricative Aff, Stop

Rhotic * ɡirm * ʔarð ̣ burd
Nasals tamir dimin ramil kanz band
Liquids * ʕilim * fils wild
Fricatives biðir ʕað ̣im fað ̣il ʃuχṣ, nafs nad͡ʒd
Aff, Stop sidir nad͡ʒim ʕaɡil χubiz ʕabd
There are a few exceptions to this hierarchy:
• A voiceless fricative following a voiceless stop does not induce epenthesis, e.g.
fatħ ‘opening’, ʕaks ‘opposite’. Exceptions to sonority sequencing involving cer
tain stopfricative sequences are well attested crosslinguistically.
• If two nasals or a nasal and a liquid occur together in a final cluster, epenthesis
usually occurs, e.g. ħɨlɨm ‘a dream’.
• If the final consonant is a voiced uvular fricative, epenthesis occurs, e.g. rizɨʁ
‘wrist’.
• If the final consonant is a glottal stop, epenthesis occurs, e.g. d͡ʒiziʔ ‘part’.
• If the final consonant is a voiced pharyngeal [ʕ], epenthesis may occur, e.g. faɡiʕ
‘mushrooms’. The data is inconsistent on this particular point, and it may depend
which allophone is in use (Butcher & Ahmad 1987). In general, /ʕ/ is either a stop
or an approximant.
• A voiced obstruent following a voiceless obstruent also triggers epenthesis, e.g.

fahad ‘leopard’, θɑqub ‘hole’ .
Several of the languagespecific exceptions listed here involve the class of conso
nants regarded as “gutturals” (McCarthy 1989; Hayward & Hayward 1989). The other
4.1. POSITION OF EPENTHESIS 61
exceptions are, on the whole, unsurprising crosslinguistically.
4.1.2 Lexical variants
For some words, monosyllabic and disyllabic variants were readily produced by all Ur
ban Qatari Arabic speakers. In these cases, speakers were aware that the disyllabic
variant was the “older” form and the reduced variant was formal, but had no problem
producing both:
(56) a. taħat ~ taħt ‘under’
b. ṣɑṭɑħ ~ ṣɑṭħ ‘roof, surface’
Other examples from conversational data show that epenthesis is not obligatory in
Urban Qatari Arabic for these consonant clusters:
(57) a. baħt ‘pure’
b. riħt ‘I went’
c. fatħ ‘opening’
One difficulty in accurately describing epenthesis of this type (CvCC → CvCiC) in

Arabic is that unstressed high vowels have been elided in some Bedouintype dialects
(CvCiC → CvCC), resulting in lexical variants with and without syncope:
(58) *kabid → kabd ~ kabid ‘liver’ (Cadora 1989)
Many such pairings occur in Gulf Arabic:
(59) a. kanz ~ kaniz ‘treasure’ (Qafisheh 1997:508; alRawi 1990:266)
b. ʕɑnz ‘nanny goat’ (Qafisheh 1997:447)

(60) a. ͡tʃabd ~ kabid ‘liver’1
b. ʕɑbd ‘slave’
(61) a. ʕilt͡ʃ ~ ʕilit͡ʃ ‘chewing gum’ (cf. Bukshaisha 1985:484)
b. malt͡ʃ ‘signing the marriage contract’ (Feghali 2008:379)
Epenthesis (CvCC → CvCiC) and high vowel deletion (CvCiC → CvCC) are thus
in a counterfeeding order. It is possible that words such as kaniz are borrowings from
nearby varieties of Arabic in which high vowel deletion did not take place. This means
that in some lexical items, the point of variation is not (synchronic) epenthesis, but
(historical) syncope. In light of the above data, the preceding examples are taken to be
subregional variants, not epenthesis in free variation.
4.1.3 Epenthesis at morpheme boundaries
During elicitations, a few CC clusters were broken up at word boundaries, as in (62). It is

likely that these were motivated by the juxtaposition of two superheavy (Cv̄C) syllables,
which induces epenthesis in Najdi Arabic.
(62) a. /t͡ʃūf ʁēn/ → [t͡ʃūf u ʁēn] ‘a letter in the Arabic abjad’
b. /t͡ʃūf lēt/ → [t͡ʃūf ɨ lēt] ‘light’
A short high vowel in an unstressed open syllable is normally deleted in Gulf Arabic,
so the type of epenthesis in (62) is analyzed to occur postlexically (as discussed in
§4.3.2).
In the literature on North Arabian dialects, clusters of three consonants at morpheme
boundaries are expected to be broken up by the insertion of a (high) vowel after the first
consonant (Qafisheh 1977:24, cf. Johnstone 1967b:26).
(63) a. /ktāb/ → [ktāb] ‘a book’
b. /l=ktāb/ → [li=ktāb] ‘the book’

1
kabid, kibd in Ibn Manẓūr (1883:3806).
4.1. POSITION OF EPENTHESIS 63
Some Arabic dialects do not break up CCC clusters (CCC → CCC). Those that
do epenthesize generally fall into two categories: those that usually epenthesize after
the first consonant (CCC → CvCC) and those that usually epenthesize after the sec
ond consonant (CCC → CCvC). Kiparsky (2003) comprehensively classifies Arabic
dialects using this tripartite scheme (C, vC, Cv). Gulf Arabic falls into the second cat
egory, but all three patterns are documented in Gulf Arabic. Threeconsonant clusters
were observed from several speakers of Urban Qatari Arabic (CCC → CCC):
(64) a. /bint=na/ → [bint=na] ‘our daughter’
b. /mtfayɡ=lik/ → [mtfayɡ=lik] ‘available for you’
c. /χirza/ → [(u)χrza] ‘a bead’2
In threeconsonant clusters, the middle consonant is usually a voiceless stop.
Epenthesis can also happen after the second consonant in Gulf Arabic (CCC →
CCvC). The final cluster in the common preposition ʕind ‘with’ is never broken up, but
epenthesis occurs after the second consonant, as in ʕindәna ‘with us’, ʕindәkum ‘with
you (pl.)’, and so on. It is not clear why this paradigm has developed, but alMotairi
(2015) proposes that such examples involve maintaining morphemic identity. To com
plicate matters further, Holes writes that some speakers in the Gulf always epenthesize
after the second consonant in threeconsonant clusters (Holes 1990:281).
4.1.4 Consonant reduction and epenthesis
Geminates cannot be broken up by epenthesis in Urban Qatari Arabic. Clusters involv

ing geminates surface differently from speaker to speaker:
2
Others are listed by Johnstone (1967b:26) and Hoffiz (1995:62). Typical examples include:
ɡiltluh ‘I told him’ (Qassimi (Najdi) Arabic, alMotairi 2015:48)

strīħ ‘relax!’ (Qafisheh 1977:24, Hoffiz 1995:62; cf. Shaaban 1977:64)
ð̣rabtha ‘she hit it’ (Dubai, Hoffiz 1995:62; cf. Holes 1990:274)
ʃiftkum ‘I saw you’ (Holes 1990:270)
kiʃtbān ‘thimble’ (Qafisheh 1997:502). Johnstone only lists CCC clusters in Qatari speech.
(65) a. /ħabb=ha/ → [ħabb=ha] ‘he loved her’ (Speaker 6)
b. /ħabb=ha/ → [ˈħabb=aha] ‘he loved her’ (Speaker 1)
c. /ħabb=ha/ → [ħabˈb=aha ]‘he loved her’ (Speaker 4)
When epenthesis did not occur after a geminate consonant, the length of the geminate
was maintained:
(66) /ħaɡɡ=ha/ → [ħaɡɡ=ha] ‘for her’
Holes wrote that degemination was common in Sunni Bahraini Arabic (Holes 2016:71).
Degemination is also frequently observed by previous researchers in Bahraini and Kuwaiti
Arabic (Johnstone 1967b; Feghali 2008), as well as in Abu Dhabi (Qafisheh 1997).
Qafisheh’s grammar points to degemination with the example kilhum ‘all of them’ (Qafisheh
1977:29, 64), but elsewhere he gives killahum (p. 35) with the same meaning, writing
that kilhum is the more common form. His later dictionary, however, includes at least
ten examples of killahum ‘all of them’, and never kilhum (Qafisheh 1997). It is possi
ble that the absence of degemination in this study of Urban Qatari Arabic points to the
influence of nonArabian dialects.
Reduction of threeconsonant clusters was likewise uncommon. Consonant reduc
tion was observed sporadically in conversational data, only in two common words shown
in (67).
(67) a. /ʕind=na/ → [ʕidna] ‘with us’ (one attestation)
b. /ɡilt=lik/ → [ɡitlik] ‘I told you’ (multiple speakers)
In these examples, consonant reduction appears to be lexical, not an active part of

the phonology.
One Urban Qatari Arabic speaker, with roots in northern Qatar, used both ʕindәhum
‘with them’ and ʕidna ‘with us’ during the same interview. In published sources, the
reduced form is characteristic of Bahraini and Kuwaiti dialects; the epenthesized form
is more common in Qatar and the United Arab Emirates. Therefore, this variant may be
characteristic of the north of Qatar.
4.2. QUALITY OF EPENTHETIC VOWELS 65
4.2 Quality of epenthetic vowels
Epenthetic vowels at word boundaries were high, such as the initial vowels in (68):
(68) a. uṭħīn ‘flour’
b. ɨnħaram ‘he was deprived’
c. iʃtibaɦ ‘he likened’
Like other short vowels, epenthetic vowels may also be backed and rounded in labial
and emphatic environments (Johnstone 1967b:26, alRawi 1990:30). This is exempli
fied in (69):
(69) θɑqub ‘hole’
Within words, epenthetic vowels exhibit vowel harmony (alRawi 1990:29, alSulaiti
1993:145147, Holes 2016:7475; cf. Blanc 1964:5556):
(70) a. laħam ‘red meat’
b. sidir ‘lotus tree’
c. ħɨṣɨn ‘fortress’
However, a competing tendency towards disharmony is seen in (71):
(71) a. ɡiṭɑn ‘cotton’
b. a̰mur ‘order’ (n.)
c. laħin ‘melody’
In (72) this resulted in homophony or nearhomophony between /CiCC/ nouns and

/CiCaC/ verbs of the same root:
(72) a. ħɨlam ~ ħɨlɨm ‘dream’
b. ħɨlam ‘he dreamed’3

3
In Qafisheh (1997), compare the nouns θiɡal ‘heaviness’ (p. 82), ṣiʁar ‘young age’ (p. 294), and
ɡiṣar ‘shortness’ (p. 522).
The epenthetic vowel in (72a) is not expected to be lowered in this consonant en
vironment, so the alternating height of the second vowel may involve phonotactic con
straints. Highvowel sequences are dispreferred, and ia is the most stable sequence of
short vowels (cf. Lahrouchi 2011; Faust 2019).
From the preceding examples, it is clear that epenthetic vowels are sensitive to the
effects of consonant environment, word position, preceding vowel, and the phonotactic
constraints of Urban Qatari Arabic.
4.3 Acoustics of epenthetic vowels
4.3.1 Underlying and epenthetic vowels
Two studies on Lebanese Arabic (Gouskova & Hall 2009; Hall 2013) found statistically
significant differences between underlying and epenthetic vowels. This has not been
tested in any North Arabian variety of Arabic. To test if this finding could be replicated
in Urban Qatari Arabic, each of the five ANOVAs for short vowels (described in §2.3.1)
tested for the main effect of the dichotomous variable Underlying {Epenthetic, Under
lying}. Each ANOVA also tested for the twoway interaction of vowel and underlying.
There was a significant main effect of underlying for F1, F0, mean intensity, and
duration (four of five dependent variables). These variables are shown in boxplots in
Figure 4.1.
Epenthetic vowels have lower intensity and lower fundamental frequency. The dif
ference in duration between lexical and epenthetic vowels is even more salient, epenthetic
vowels being much shorter than lexical vowels. Finally, nonfinal epenthetic vowels
have a much lower F1 (corresponding to a higher vowel) than nonfinal lexical vow
els. These statistical results indicate that epenthetic vowels differ from lexical vowels
in their acoustic properties.
In addition, there was a significant twoway interaction of vowel and underlyingness

on mean intensity. This interaction is shown in Figure 4.2, where high epenthetic vowels
are characterized by lower intensity.
4.3. ACOUSTICS OF EPENTHETIC VOWELS 67
Figure 4.1: Statistically significant differences between lexical and epenthetic vowels
Full statistical results are reported in Table C.2.
Epenthetic vowels are therefore phonetically distinct from lexical vowels, despite
being traditionally transcribed as a, i, or u.
Figure 4.2: Intensity of epenthetic vowels, by vowel

4.3.2 Lexical and postlexical epenthesis
During data analysis, variable stressepenthesis interaction demonstrated that there are at
least two epenthesis rules in some varieties of Gulf Arabic, as shown in §3.2.4 For certain
speakers, one type of epenthesis is capable of receiving stress assignment while another
is not. Acoustic phonetic research regarding variable epenthesis has been pursued in
Tripolitanian Libyan Arabic (Plug et al. 2019), but not in any of the North Arabian
dialects.
In elicited items, most epenthesized forms resulted from lexical epenthesis, which is
capable of receiving stress assignment. As discussed above, certain epenthesized forms
resulted from postlexical epenthesis. In (73), epenthesis follows stress assignment.
(73) /ˈṭaɡɡ=nī/ → [ˈṭaɡɡ=әni] ‘he loved her’ (Speaker 6)
In (74), epenthesis follows a rule that deletes short high vowels in unstressed open
syllables.
(74) a. /ʁēn/ → [uʁēn] ‘a letter in the Arabic abjad’ (both speakers)
b. /lēt/ → [ɨlēt] ‘light’ (Speaker 1)
c. /ˈʕuqda/ → [ˈʕuqɨda] ‘a knot; a complication’ (Speaker 6)
The above forms differed from other epenthesized forms in multiple ways. Post
lexical epenthesis was much less consistent than lexical epenthesis. Speakers of Urban
Qatari Arabic were readily aware of alternations produced by lexical epenthesis. Im
pressionistically, vocoids that resulted from postlexical epenthesis seemed perceptibly
less distinct than vowels that surfaced during lexical epenthesis.
To test the hypothesis that lexical and postlexical epenthesis are phonetically dis
tinct, a statistical model was created to examine measurements of epenthetic vowels.
The continuous dependent variables were F1, F2, fundamental frequency (F0), in
tensity, and duration.
4
I owe thanks to John Clifton for suggesting this acoustic investigation of variable epenthesis.
4.3. ACOUSTICS OF EPENTHETIC VOWELS 69
This model includes independent variables described in §2.1.1 and 2.3.1. Only high
epenthetic vowels were tested since low epenthetic vowels are infrequent. Therefore, the
variable for vowel was not included. In addition, a new independent variable Epenthe
sisType {Lexical, Postlexical} was introduced to account for the two types of epenthe
sis described in §3.2.
Five tests were performed on measurements of 225 vowels. Based on the analysis
presented in this thesis, 194 of these surfaced during lexical epenthesis (capable of re
ceiving stress assignment), and 31 surfaced during postlexical epenthesis (incapable of
receiving stress assignment).
The model equation is given below, with asterisks representing the interaction of
two variables. Full statistical results are given in Table C.4.
lmer([ depvar ] = (ListPosition + Stress + BackingConsonantBefore + Backing

ConsonantAfter + BackingConsonantBefore * BackingConsonantAfter + SyllableType
+ WordPosition + EpenthesisType) + (1|Speaker))
The regression model failed some of Levene’s tests for unequal variances. There
fore, these results cannot be regarded as conclusive, especially in light of the small
number of postlexical epenthetic vowels that were recorded. Similar research may test
the robustness of these findings with a larger and more balanced sample of epenthetic
vowels.
The variable for epenthesis type yielded statistically significant results on the de
pendent variables F2 (backness) and duration. Both effects are shown below in Figure
4.3. The results show that epenthetic vowels at syllable boundaries tend to be signif
icantly more fronted in quality than other epenthetic vowels. Vocoids resulting from
postlexical epenthesis are also substantially shorter in duration. This acoustic phonetic
treatment corroborates the finding that epenthesis is variable in Gulf Arabic. Paired with
the findings in §4.3.1, this allows us to posit the existence of four categories of vocoids in
Gulf Arabic, each of which may be distinguished from the others by its duration: long
vowels, short vowels, lexical epenthetic vowels, and postlexical epenthetic vocoids.
These categories are distinguished from one another by both phonetic and phonological
characteristics.
Figure 4.3: Statistically significant differences between lexical and postlexical epenthe
sis
4.4 Conclusion
This thesis makes several advances in the study of Gulf Arabic and also presents promis
ing avenues for future research. The findings on the vowel inventory point to the need
for more nuanced phonological representation. In Gulf Arabic, short vowels are subject
to complex and gradient effects from consonant environment. The seemingly erratic be
4.4. CONCLUSION 71
havior of Gulf Arabic short vowels is explained with reference to interactive consonant
effects that extend beyond the boundary of the syllable.
The findings concerning epenthesis in Gulf Arabic point to a continued pattern of
standardization and koineization in Gulf Arabic. Even while conversing in Gulf Ara
bic, speakers often produced formal variants, e.g. baraka ‘blessing’ and ṣɑṭħ ‘roof’.
On reflection, they realized that brike and ṣɑṭɑħ were the vernacular forms they would
use. This shows that patterns of syncope and epenthesis that were recently described
as obligatory in the language are now lexical variants, and speakers may choose freely
according to the register needed. Syncope in lightsyllable strings, in particular, was
a pattern that was considered as a defining characteristic of all North Arabian dialects
according to Johnstone and Ingham; however, no freely occurring examples of syncope
were observed during this study. As Gulf Arabic speakers have a wider range of regis
ters available to them, phonological description will have to adapt.
Several phonological features that are atypical of vernacular Gulf Arabic were rel
atively much more frequent among female speakers. Some of these features were char
acteristic of prestige varieties of the language, such as preference of the uvular stop [q]
over its usual reflex [ɡ] (AbdElJawad 1987; Amara 2005) and light syllable strings
(CvCvCv) rather than reduced forms (CCvCv ~ CvCCv). Other gendermarked fea
tures were neutral in prestige, such as creaky and breathy phonation, and the interaction
of stress and epenthesis. The sample size in this study was not enough to generalize
about gender; however, this could prove an interesting area in the future as many lin
guistic studies in the Gulf draw on data from only men or only women.
Bibliography
Abboud, Peter F. 1979. The verb in Northern Najdī Arabic. Bulletin of the School of
Oriental and African Studies, University of London 42.467–499.
AbdElJawad, Hassan R. 1987. Crossdialectal variation in Arabic: Competing pres

tigious forms. Language in Society 16.359–367.
AbdelMassih, Ernest T. & AbdelMalek, Zaki N. & Badawi, ElSaid M. 1981. A

reference grammar of Egyptian Arabic. Center for Near Eastern and North African
Studies, University of Michigan.
Abdurehim, Esmael. 2014. The Lopnor dialect of Uyghur: A descriptive analysis,

volume 17 of Publications of the Institute for Asian and African Studies. Helsinki:
University of Helsinki.
Abo Mokh, Noor & Davis, Stuart. 2020. What triggers ’imala. In van Gelderen, Elly
(ed.) Perspectives of Arabic Linguistics XXXII: Papers selected from the Annual
Symposium on Arabic Linguistics, Tempe, Arizona, 35–55.
alAmadidhi, Darwish G. H. 1985. Lexical and sociolinguistic variation in Qatari

Arabic. University of Edinburgh dissertation.
alAni, Salman. 1992. Stress variation of the construct phrase in Arabic: A spectro
graphic analysis. Anthropological Linguistics 34.256–276.
Al Hashemi, Ayesha & Isleem, Nasser. 2015. Ramsah: An introduction to learning

Emirati dialect and culture. Kuttab Publishing.
73
74 BIBLIOGRAPHY
Al Jahdhami, Said. 2016. Kumzari: The forgotten language. International Journal of

Linguistics 8.27–34.
alMaṭlabī, G. F. 1978. Lahjāt Tamīm wa āthāruha fī al‘Arabiyya almuwaḥḥada

[The dialect of Tamīm and its influences on Arabic as a whole]. Baghdad: Dār
alḤurriyya.
alMotairi, Sarah. 2015. An optimalitytheoretic analysis of syllable structure in Qassimi

Arabic. Master’s thesis, Eastern Michigan University.
alMuhannadi, Muneera. 1991. A sociolinguistic study of women’s speech in Qatar.

University of Essex dissertation.
alMālikī, N. 2000. Alfāẓ dakhīla wamu‘arraba fi allahja alQaṭariyya [Extraneous

and borrowed vocabulary in Qatari dialect]. Doha: Markaz alturāt̠ alsha‘bī.
alQenaie, Dawoud Shamlan. 2011. Kuwaiti Arabic: A sociophonological perspective.

Durham University dissertation.
alRawi, RosinaFawzia B. 1990. Studien zum arabischen Dialekt von Abū Ẓabī: Nom
inalbildung und lexik, Texte [Studies in the Arabic dialect of Abu Dhabi: Noun
morphology, lexicon, and texts]. Heidelberg: Groos.
alSulaiti, Latifa Mubarak. 1993. Some aspects of Qatari Arabic phonology and mor
phology. Univerity of Lancaster dissertation.
AlTajir, Mahdi Abdalla. 1982. Language and linguistic origins in Baḥrain: The Baḥār
nah dialect of Arabic, volume 5 of Library of Arabic Linguistics. Taylor & Francis.
AlBader, Yousuf B. 2015. Semantic innovation and change in Kuwaiti Arabic: A study
of the polysemy of verbs. University of Sheffield dissertation.
Alkalesi, Yasin M. 2006. Modern Iraqi Arabic with mp3 files: A textbook. Georgetown
University Press.
Alpen Capital. 2018. GCC education industry. Alpen Capital.

BIBLIOGRAPHY 75
Amara, Muhammad. 2005. Language, migration, and urbanization: The case of Beth
lehem. Linguistics 43.883–901.
Arabian American Oil Company. 1958. EnglishArabic word list. Monterey, Calif.:
Defense Language Institute.
Baart, Joan. 2010. A field manual of acoustic phonetics. Dallas: SIL International.
BarMoshe, Assaf. 2019. The Arabic dialect of the Jews of Baghdad: Phonology,
morphology, and texts. Wiesbaden: Harrassowitz Verlag.
Behnstedt, Peter & Woidich, Manfred. 2010. Wortatlas der arabischen dialekte: Band
I: Mensch, natur, fauna und flora [Word atlas of Arabic dialects: Book I: People,
nature, fauna and flora]. Brill.
—— & ——. 2013. Dialectology. In The Oxford handbook of Arabic linguistics. Oxford
University Press.
Bernabela, Roy S. 2011. A phonology and morphology sketch of the Šiḥḥi Arabic
dialect of әlĞēdih, Oman. Master’s thesis, Leiden University.
Bessell, Nicola J. 1991. Vowel effects of pharyngeal and uvular consonants in Interior
Salish. The Journal of the Acoustical Society of America 89.1918–1918.
Bessell, Nicola J. 1998. Local and nonlocal consonantvowel interaction in Interior

Salish. Phonology 15.1–40.
Bettega, Simone. 2014. The use of dialectal tanwīn in Qatari Arabic. Ricognizioni:
Rivista di lingue, letterature e culture moderne 2.15–25.
——. 2017. Diachronic variation in Qatari Arabic: A comparison between modern

data and those gathered by T. M. Johnstone between 1950 and 1960. In al Hamad,
Muntasir Fayez Faris & Ahmad, Rizwan & Alaoui, Hafid I. (ed.) Lisan alArab:
Studies in Arabic dialects. Proceedings of the 10th international conference of
AIDA, Qatar University, 2013, Zürich. Lit Verlag.
76 BIBLIOGRAPHY
Blanc, Haim. 1964. Communal dialects in Baghdad, volume 10. Cambridge, Mass.:
Harvard University Press.
Bukshaisha, Fouzia A. M. 1985. An experimental phonetics study of some aspects in

Qatari Arabic. University of Edinburgh dissertation.
Butcher, Andrew & Ahmad, Kusay. 1987. Some acoustic and aerodynamic character
istics of pharyngeal consonants in Iraqi Arabic. Phonetica 44.156–172.
Cadora, Frederic J. 1989. Linguistic change and the Bedouinsedentary dichotomy in

Old Arabic dialects. Anthropological Linguistics 31.264–284.
Choi, John D. 1991. An acoustic study of Kabardian vowels. Journal of the Interna
tional Phonetic Association 21.4–12.
CIA World Factbook. 2020. Qatar. Page last updated on July 21, 2020.
Clements, G. N. 1990. The role of the sonority cycle in core syllabification. In Papers in
Laboratory Phonology: Volume 1, Between the Grammar and Physics of Speech.
Cambridge: Cambridge University Press.
Cohn, Abigail C. 2006. Is there gradient phonology? In Fanselow, G. & Féry, C. &
Schlesewsky, M. & Vogel, R. (ed.) Gradience in grammar: Generative perspec
tives, 25–44. Oxford: Oxford University Press.
Contini, Riccardo. 2003. La lingua del Bēt Qaṭrāyē [The language of Bēt Qaṭrāyē].
In Lentin, J. & Lonnet, A. (ed.) Mélanges David Cohen, 173–181. Paris: Maison
neuve & Larose.
Cowell, Mark W. 1964. A reference grammar of Syrian Arabic. Washington, D.C.:

Georgetown University Press.
de Jong, Rudolf Erik. 2000. A grammar of the Bedouin dialects of the northern Sinai
Littoral: Bridging the linguistic gap between the eastern and western Arab world.
Leiden: Brill.
BIBLIOGRAPHY 77
——. 2011. A grammar of the Bedouin dialects of central and southern Sinai. Leiden:
Brill.
Durand, Emilie Pénélope. 2011. Wordfinal imaala in contemporary Levantine Arabic:

A case of language variation and change. Master’s thesis, University of Texas at
Austin.
Eberhard, David M. & Simons, Gary F. & Fennig, Charles D. (eds.) 2020. Ethnologue:
Languages of the world. Dallas: SIL International.
Erwin, Wallace M. 1963. A short reference grammar of Iraqi Arabic. Washington,

D.C.: Georgetown University Press.
Faust, Noam. 2019. The apophonic chain and the form of weak and strong verbs in
Palestinian Arabic. The Linguistic Review 34.83–121.
Feghali, Habaka J. 2008. Gulf Arabic: The dialects of Kuwait, Bahrain, Qatar, UAE,
and Oman: Grammar, dialogues, and lexicon. Hyattsville, Maryl.: Dunwoody
Press.
Feghali, Michel T. 1919. Le parler de Kfar’abida (LibanSyrie) [The language of

Kfar’abida (LebanonSyria)]. Imprimerie nationale.
Ferguson, Charles A. 1957. Two problems in Arabic phonology. Word 13.460–478.
—— 1959. Diglossia. Word 15.325–340.
Fromherz, Allen J. 2017. Qatar: A modern history. Updated edition. Washington, D.C.:
Georgetown University Press.
Fry, Dennis B. 1955. Duration and intensity as physical correlates of linguistic stress.
The Journal of the Acoustical Society of America 27.765–768.
Gordon, Matthew K. & Roettger, Timo B. 2017. Acoustic correlates of word stress: A
crosslinguistic survey. Linguistics Vanguard 3.
78 BIBLIOGRAPHY
Gouskova, Maria & Hall, Nancy. 2009. Acoustics of epenthetic vowels in Lebanese
Arabic. In Phonological argumentation: Essays on evidence and motivation, 203–
225. Equinox London.
Grigore, George. 2007. L’arabe parlé à Mardin: Monographie d’un parler arabe
“périphérique” [The Arabic dialect of Mardin: A monograph on a “peripheral”
Arabic dialect]. Bucharest: Editura Universităţii din Bucureşti.
Hall, Nancy. 2013. Acoustic differences between lexical and epenthetic vowels in
Lebanese Arabic. Journal of Phonetics 41.133––143.
Harrell, Richard S. 1962. A short reference grammar of Moroccan Arabic. Washington

D.C.: Georgetown University Press.
Hassan, Zeki Majeed. 1981. An experimental study of vowel duration in Iraqi Spoken
Arabic. University of Leeds dissertation.
Hayward, Katrina M. & Hayward, Richard J. 1989. ‘Guttural’: Arguments for a new
distinctive feature. Transactions of the Philological Society 87.179–193.
Herzallah, Rukayyah S. 1990. Aspects of Palestinian Arabic phonology: A nonlinear

approach. Cornell University dissertation.
Hoffiz, Benjamin. 1995. Morphology of United Arab Emirates Arabic, Dubai dialect.
University of Arizona dissertation.
Holes, Clive. 1984. Colloquial Arabic of the Gulf and Saudi Arabia. New York:
Routledge.
——. 1986. The social motivation for phonological convergence in three Arabic dialects.
International Journal of the Sociology of Language 61.33––51.
——. 1990. Gulf Arabic. New York: Routledge.
——. 2001. Dialect, culture, and society in eastern Arabia. Volume one: Glossary.
Leiden: Brill.
BIBLIOGRAPHY 79
——. 2002. NonArabic Semitic elements in the Arabic dialects of eastern Arabia. In
Sprich doch mit deinen Knechten aramäisch, wir verstehen es [Please speak to
your servants in Aramaic, for we understand it], 269–280. Harassowitz Verlag.
——. 2005. Dialect, culture, and society in eastern Arabia. Volume two: Ethnographic
texts. Leiden: Brill.
——. 2006. Bahraini Arabic. In Encyclopedia of Arabic language and linguistics,

volume 1, 241–255. Brill.
——. 2016. Dialect, culture, and society in eastern Arabia. Volume three: Phonology,
morphology, syntax, style. Leiden: Brill.
Ḥammūz, ‘Abd alFattāḥ. 2008. Mu‘jam alfāẓ lahjat alImārāt wata’ṣīliha [A dictio
nary of the vocabulary items of Emirati dialect and their etymologies]. AlAin:
Zayed Center for Heritage and History.
Ḥanẓal, Fāliḥ. 1998. Mu‘jam alalfāẓ al‘āmmiyya fi dawlat alImārāt [A dictionary

of vernacular vocabulary in the nation of the Emirates]. Abu Dhabi: Ministry of
Information and Culture.
Ibn Manẓūr, Muḥammad ibn Mukarram. 1883. Lisān al‘Arab [The tongue of the
Arabs]. Būlāq Miṣr.
ilHazmy, Alayan Mohammed. 1975. A critical and comparative study of the spoken
dialect of the Ḥarb tribe of Saudi Arabia. University of Leeds dissertation.
Ingham, Bruce. 1982. Northeastern Arabian dialects. London & Boston: Kegan Paul
International.
——. 1986. Notes on the dialect of the Āl Murra of eastern and southern Arabia. Bulletin
of the school of African and Oriental Studies 2.271–291.
——. 1994. Najdi Arabic: Central Arabian. Amsterdam & Philadelphia: John Ben
jamins.
80 BIBLIOGRAPHY
——. 1996. Arabian diversions: Studies on the dialects of Arabia. Lebanon: Ithaca
Press.
Johnstone, Thomas Muir. 1967a. Aspects of syllabication in the spoken Arabic of

‘Anaiza. Bulletin of the school of Oriental and African studies 30.1–16.
——. 1967b. Eastern Arabian dialect studies. London: Oxford University Press.
Khalafallah, Abdelghany A. 1969. A descriptive grammar of Ṣa‘īdi Egyptian colloquial

Arabic. The Hague: Mouton.
Khan, Geoffrey. 1997. The Arabic dialect of the Karaite Jews of Hit. Zeitschrift für
Arabische linguistik. No. 34, 53102 .
Kiparsky, Paul. 2003. Syllables and moras in Arabic. In Féry, Caroline & van de Vi
jver, Ruben (ed.) The syllable in optimality theory, 147–182. Cambridge University
Press.
Kozah, Mario & Kiraz, George & AbuHusayn, Abdulrahim & AlMurikhi, Saif &
Al Thani, Haya. 2021. Beth Qaṭraye: A lexical and toponymical survey. Pis
cataway, N.J.: Gorgias Press. Forthcoming.
Lahdo, Ablahad. 2009. The Arabic dialect of Tillo in the region of Siirt (southeastern
Turkey). Uppsala: Edita Västra Aros.
Lahrouchi, Mohamed. 2011. Cooccurrence restrictions in the vocalic patterns of

Afroasiatic plurals. Folia Orientalia 2.89–102.
Landberg, Carlo. 1920. Glossaire Dat̠ înois [Dathina glossary]. Leiden: Brill.
Lehn, W. 1967. Vowel contrasts of Najdi Arabic. In Stuart, D. G. (ed.) Linguistic stud
ies in memory of Richard Slade Harrell, 123–31. Washington D.C.: Georgetown
University Press.
Leung, T. TszCheung & Ntelitheos, Dimitrios & Kaabi, Meera Al. 2020. Emirati Ara
bic: a comprehensive grammar. Calcutta: Superintendent Government Printing.
BIBLIOGRAPHY 81
Levin, A. 1992. The authenticity of Sībawayhi’s description of the imāla. Jerusalem

studies in Arabic and Islam 15.74–93.
Lorimer, John Gordon. 1908. Gazetteer of the Persian Gulf, ’Omān and central Arabia.
Calcutta: Superintendent Government Printing.
Mandaville, James P. 2011. Bedouin ethnobotany: Plant concepts and uses in a desert
pastoral world. Tucson: University of Arizona Press.
Max, L. & Onghena, P. 1999. Some issues in the statistical analysis of completely
randomized and repeated measures designs for speech, language, and hearing re
search. Journal of Speech, Language, and Hearing Research 42.261–270.
McCarthy, John J. 1989. Phonetics and phonology of Semitic pharyngeal consonants.

Linguistics Department Faculty Publication Series, Paper 86.
—— 2005. The length of stemfinal vowels in colloquial Arabic. Linguistics Department

Faculty Publication Series .
McCollum, Adam. 2018. Gradient morphophonology: Evidence from Uyghur vowel

harmony. In Proceedings of the 2018 Annual Meeting on Phonology. Linguistic
Society of America.
Meissner, Bruno. 1903. Neuarabische geschichten aus dem Iraq [Modern Arabic texts
from Iraq]. Leipzig: J. C. Hinrichs.
Ministry of Development Planning and Statistics. 2017. Education in Qatar: Statistical

profile 2016. Ministry of Development Planning and Statistics.
Morano, Roberta. 2019. The Arabic dialect spoken in the alʿAwābī district, northern
Oman. University of Leeds dissertation.
Mustafawi, Eiman. 2006. An optimalitytheoretic approach to variable consonantal

alternations in Qatari Arabic. University of Ottawa dissertation.
82 BIBLIOGRAPHY
——. 2007. Affrication in North Arabic revisited. Amsterdam studies in the theory and
history of linguistic science series 4.151.
——. 2011. The OCP as a synchronic constraint in Arabic. Canadian Journal of

Linguistics/Revue canadienne de linguistique 56.229–246.
—— & Mahfoudhi, Abdessatar. 2002. The development of binding principles: New

findings. Cahiers linguistiques d’Ottawa 30.91–111.
Næss, Unn Gyda. 2008. “Gulf Pidgin Arabic”: Individual strategies or a structured
variety? Master’s thesis, University of Oslo.
Overlaet, Bruno & Macdonald, Michael C. A. & Stein, Peter. 2016. An Aramaic
Hasaitic bilingual inscription from a monumental tomb at Mleiha, Sharjah, UAE.
Arabian Archaeology and Epigraphy 27.127––142.
Owens, Jonathan. 1993. A grammar of Nigerian Arabic, volume 10. Otto Harrassowitz
Verlag.
——. 1998. Neighborhood and ancestry: Variation in the spoken Arabic of Maiduguri,
Nigeria, volume 4 of Impact: Studies in Language and Society. Amsterdam: John
Benjamins.
——. 2006. A linguistic history of Arabic. New York: Oxford University Press.
Plug, Leendert & Shitaw, Abdurraouf & Heselwood, Barry. 2019. Interconsonantal
intervals in Tripolitanian Libyan Arabic: Accounting for variable epenthesis. Lab
oratory Phonology: Journal of the Association for Laboratory Phonology 10.1––
33.
Prochazka, Jr., Theodore. 1988. Saudi Arabian dialects. London: Kegan Paul.
Qafisheh, Hamdi A. 1977. A short reference grammar of Gulf Arabic. Tucson: The
University of Arizona Press.
—— 1992. Yemeni Arabic reference grammar. Dunwoody Press.

BIBLIOGRAPHY 83
—— 1996. A glossary of Gulf Arabic: Gulf ArabicEnglish, EnglishGulf Arabic.

Beirut: Libraire du Liban.
—— 1997. NTC’s Gulf ArabicEnglish dictionary. Chicago: NTC/Contemporary.
Romeny, Bas ter Haar. 2014. Syriac biblical interpretation from Qatar: Ahob of Qatar.
In Kozah, Mario & AbuHusayn, AbdulRahim & AlMurikhi, Saif Shaheen &
Al Thani, Haya (ed.) The Syriac writers of Qatar in the seventh century. Piscataway,
N.J.: Gorgias Press.
Rosenhouse, Judith. 2006. Bedouin Arabic. In Versteegh, Kees (ed.) Encyclopedia of

Arabic language and linguistics, 259–269. Leiden: Brill.
Sakr, Georges. 2019. A laboratory phonology investigation into the vocalic inventory
of Lebanese. Master’s thesis, University of Edinburgh.
Shaaban, Kassim A. 1977. The phonology of Omani Arabic. University of Texas at

Austin dissertation.
Shahin, Kimary. 2011. A phonetic study of guttural laryngeals in Palestinian Arabic

using laryngoscopic and acoustic analysis. In Hassan, Z. M. & Heselwood, B. (ed.)
Instrumental Studies in Arabic Phonetics, volume 319, 129–140. John Benjamins.
Shawkat, Mahmoud Hamed. 1962. A descriptive grammar of educated Damascene

Arabic. Cornell University dissertation.
Shockley, Mark. 2020. Diminutive proper nouns in Peninsular Arabic. Paper presented
at the Arabic Linguistics Forum.
Smart, J. R. 1990. Pidginization in Gulf Arabic: A first report. Anthropological Lin

guistics 32.83–119.
Thomas, Bertram. 1930. The Kumzari dialect of the Shihuh tribe, Arabia, and a vocab
ulary. London: Royal Asiatic Society.
84 BIBLIOGRAPHY
van der Wal Anonby, Christina. 2015. A grammar of Kumzari: A mixed PersoArabian
language of Oman. Leiden University dissertation.
Watson, Janet C. E. 2011a. Arabic dialects. In Weninger, Stefan & Khan, Geoffrey &
Streck, Michael P. & Watson, Janet C. E. (ed.) The Semitic languages: An inter
national handbook, 851–896. Berlin and New York: De Gruyter Mouton.
—— 2011b. Dialects of the arabian peninsula. In Weninger, Stefan & Khan, Geof
frey & Streck, Michael P. & Watson, Janet C. E. (ed.) The Semitic languages: An
international handbook, 897–908. Berlin and New York: De Gruyter Mouton.
—— 2011c. Word stress in Arabic. In The Blackwell Companion to Phonology, 1–29.

Wiley & Sons.
Wehr, H., J. Milton Cowan. 1979. A dictionary of modern written Arabic, Arabic
English. Urbana, Ill.: Spoken Language Services.
Wetzstein, J. G. 1868. Sprachliches aus den Zeltlagern der syrischen Wüste [Lin
guistics of the Syrian desert camps]. Zeitschrift der Deutschen Morgenländischen
Gesellschaft 22.69–194.
Woodhead, Daniel R. & Beene, Wayne & Stowasser, Karl. 1967. A dictionary of Iraqi
Arabic: ArabicEnglish. Washington, D.C.: Georgetown University Press.
Yasin, Raslan Bani & Owens, Jonathan. 1984. The Bduul dialect of Jordan. Anthropo
logical Linguistics 26.202–232.
Zamakhshari, Maḥmūd bin ‘Umar. 1859. alMufaṣṣal [The minute description]. Chris
tianae.
Appendix A
Variation in stress assignment
In North Arabian dialects, stress is not always predictable if the final two syllables are
both light. Table A.1 summarizes previous accounts of stress in those forms in which
stress varies.
Two points of diachronic variation are noted:
1. Penultimate stress in light syllable strings (Cv.Cv.Cv), e.g., ba.ˈra.ka ‘blessing’,

was common in the early twentieth century (Meissner 1903:14; Lorimer 1908),
transitional in the middle of the twentieth century (Johnstone 1967b:2829; Erwin
1963:4043), and is not recorded in any modern sources.
2. Penultimate stress in epenthesized words with the syllable weight pattern heavy
lightlight (CvC.Cv.Cv), e.g., ħab.ˈbә.ha ‘he loved her’, is very unusual in pre
vious publications on Gulf Arabic. Only Qafisheh (1977:29, 64) mentions this,
noting that it is unusual. However, such examples were observed in the speech of
three out of four Urban Qatari Arabic speakers, particularly after verbs ending in
geminates.
85
86 APPENDIX A. VARIATION IN STRESS ASSIGNMENT
Table A.1: Accounts of stress in North Arabian dialects
CvCvCv CvCCvCv(C) Are epenthetic vowels stressed?

Mesopotamian
Meissner 1903:14 Varies 1 Antepenultimate Varies
Erwin 1963:4043 Varies 2 Varies Varies
Alkalesi 2006:15 Antepenultimate Antepenultimate
Najdi
Johnstone 1967a:5 Syncope Antepenultimate
Abboud 1979:498 Syncope Antepenultimate
Ingham 1982:xxi Syncope Antepenultimate No
Prochazka 1988:2021 Antepenultimate Antepenultimate
Qatari
Johnstone 1967b:111 Varies Antepenultimate No
Bukshaisha 1985:89 Antepenultimate Penultimate
alSulaiti 1993:8, 94 Antepenultimate Antepenultimate
Shockley 2020 Antepenultimate Varies Varies
Gulf
Lorimer 1908 Penultimate Penultimate
Johnstone 1967b:2829 Varies 3 Penultimate No
Qafisheh 1977:3035 Varies 4 Penultimate Yes
Holes 1984:318319 Antepenultimate Varies No
Holes 1990:274282 Antepenultimate 5 Penultimate Varies
Feghali 2008:17 Antepenultimate Antepenultimate
Holes 2016:76 Syncope Penultimate
1
Antepenultimate stress is rare, mainly heard among urban speakers.
2
Stress is antepenultimate in monomorphemic examples, but varies in multimorphemic examples.
3
CvCvCv surfaces only in loanwords; normally, syncope is expected.
4
CvCvCv surfaces only in loanwords; normally, syncope is expected.
5
Stress is antepenultimate among educated speakers; syncope is expected among uneducated
speakers.
Appendix B
Wordlist
Below in Table B.1 is the wordlist used in elicitation sessions in Qatar. The transcription
given here includes the most wellattested variants in the data. Formal variants are
marked with (†). Variants that are characteristic of other Arabic dialects such as kitbat
‘she wrote’ are marked with (‡).
Table B.1: Wordlist
English Gloss Transcription (IPA) Arabic Stimulus

her sister ṵˈχitha ~ χɨtta ‫ﺃﺧﺘﻬﺎ‬
ground (adj.) a̰rð ̣iː ‫ﺃﺭﺿﻲ‬
an order a̰mur ‫ﺃﻣﺮ‬
administration ḭdɑːra ‫ﺇﺩﺍﺭﺓ‬
frame ḭṭɑːr ‫ﺇﻃﺎﺭ‬
imam ḭmɑːm ‫ﺇﻣﺎﻡ‬
ear ḭðin ‫ﺍﺫﻥ‬
he likened iʃtibah ‫ﺍﺷﺘﺒﻪ‬
China ɨṣṣiːn ‫ﺍﻟﺼﻴﻦ‬
amen ɑ̰ːmiːn ‫ﺍﻣﻴﻦ‬
he was deprived ɨnħaram ‫ﺍﻧﺤﺮﻡ‬
Iran ḭːrɑːn ‫ﺍﻳﺮﺍﻥ‬
faith ḭːmɑːn ‫ﺍﻳﻤﺎﻥ‬
door bɑːb ‫ﺑﺎﺏ‬
salmon (coll.) bɑːluːl ‫ﺑﺎﻟﻮﻝ‬
sea baħar ‫ﺑﺤﺮ‬
blessing (u)brike ~ ˈbaraka(†) ‫ﺑﺮﻛﺔ‬
onion (coll.) buṣɑl ~ bɑṣɑl ‫ﺑﺼﻞ‬
hero bɑṭɑl ‫ﺑﻄﻞ‬
stomach,pregnancy bɑṭɑn ‫ﺑﻄﻦ‬
— Continued on next page —
87
88 APPENDIX B. WORDLIST
Table B.1 continued

even bɑʕd ‫ﺑﻌﺪ‬
he went away bɑʕɑd ‫ﺑﻌﺪ ﻓﻌﻞ ﻣﺎﺿﻲ‬
cow buɡar ‫ﺑﻘﺮ‬
with you bik ‫ﺑﻚ‬
reel — ‫ﺑﻜﺮﺓ ﻟﻴﺲ ﺑﺎﻛﺮ ﻭﻧﺶ‬
our daughter bintna ~ bɨˈnɨtna ‫ﺑﻨﺘﻨﺎ‬
pinky finger bunṣɨr ‫ﺑﻨﺼﺮ‬
house beːt ‫ﺑﻴﺖ‬
money beːzɑːt ‫ﺑﻴﺰﺍﺕ‬
tobacco titin ‫ﺗﺘﻦ ﻛﻠﻤﺔ ﻗﺪﻳﻤﺔ‬
they gathered together tid͡ʒamˈɦaraw ‫ﺗﺠﻤﻬﺮﻭﺍ‬
under taħat ‫ﺗﺤﺖ‬
the nineties tisʕiːnɑːt (†) ‫ﺗﺴﻌﻴﻨﺎﺕ‬
berries (coll.) tuːt ‫ﺗﻮﺕ‬
twins toːm ‫ﺗﻮﻡ‬
figs (coll.) tiːn ‫ﺗﻴﻦ‬
hole θɑqub ‫ﺛﻘﺐ‬
a weight θiɡɨl ‫ﺛﻘﻞ ﺍﺳﻢ‬
it weighed down θaɡɡal ‫ﺛﻘﻞ ﻓﻌﻞ ﻣﺎﺿﻲ‬
onethird θilθ ‫ﺛﻠﺚ‬
men's garment θoːb ‫ﺛﻮﺏ‬
customs d͡ʒimrik ~ d͡ʒɨmrɨk ‫ﺟﻤﺮﻚ‬
love! ħɨbb ‫ﺣﺐ! ﺃﻣﺮ‬
our love ħɨbbna ‫ﺣﺒﻨﺎ ﺍﻻﺳﻢ‬
he loved her ħabbɦa ~ ħabˈbaha ‫ﺣﺒﻬﺎ ﻓﻌﻞ ﻣﺎﺿﻲ‬
fortress ħɨṣɨn ‫ﺣﺼﻦ ﺍﺳﻢ‬
a piece of firewood uħṭubo ‫ﺣﻄﺒﻪ‬
for ħaɡɡ ‫ﺣﻖ‬
for her ħaɡɦa ~ ħaɡɡɦa ‫ﺣﻘﻬﺎ‬
reign (n.) ħukum ‫ﺣﻜﻢ ﻣﻤﻠﻜﺔ ﺍﺳﻢ‬
dream ħɨlam ~ ħɨlɨm ‫ﺣﻠﻢ ﺍﺳﻢ‬
he dreamed ħɨlam ‫ﺣﻠﻢ ﻓﻌﻞ ﻣﺎﺿﻲ‬
a turtle ɨħmise ‫ﺣﻤﺴﻪ‬
a fish ħuːt ‫ﺣﻮﺕ‬
courtyard ħoːʃ ‫ﺣﻮﺵ‬
whales ħiːtɑːn ‫ﺣﻴﺘﺎﻥ‬
moment ħiːn ‫ﺣﻴﻦ‬
maternal uncle χɑːl ‫ﺧﺎﻝ‬
your (pl.) bread χuˈbɨzkum ‫ﺧﺒﺰﻛﻢ‬
circumcision χtɑːn ~ χitɑːn ‫ﺧﺘﺎﻥ‬
liar χarrɑːṭa ‫ﺧﺮﺍﻃﺔ‬
a bead uχrza ~ uχrɨza ‫ﺧﺮﺯﺓ‬
let us χɑlna ~ χɑnna ‫ﺧﻠﻨﺎ‬
ring finger χunṣɨr ‫ﺧﻨﺼﺮ‬
89
Table B.1 continued

bay/name of a city χoːr ‫ﺧﻮﺭ‬
good χoːʃ ‫ﺧﻮﺵ‬
good χeːr ‫ﺧﻴﺮ ﺍﺳﻢ‬
horses (coll.) χeːl ‫ﺣﺼﺎﻥ ﻧﻔﺲ ﺧﻴﻞ‬
entry dɨχuːl ‫ﺩﺧﻮﻝ‬
dirham dirham ‫ﺩﺭﻫﻢ‬
send! dizz ‫ﺩﺯ! ﺃﻣﺮ‬
he hid (tr.) dass ‫ﺩﺱ ﻓﻌﻞ ﻣﺎﺿﻲ‬
go in! diʃʃ ‫ﺩﺵ! ﺃﻣﺮ‬
notebook daftar ‫ﺩﻓﺘﺮ‬
knock! diɡɡ ‫ﺩﻕ! ﺃﻣﺮ‬
mast (of a ship) — ‫ﺩﻗﻞ‬
grease dihɨn ‫ﺩﻫﻦ ﺍﺳﻢ‬
courses dawrɑːt ‫ﺩﻭﺭﺍﺕ‬
our homeland diːˈretna ‫ﺩﻳﺮﺗﻨﺎ‬
religion diːn ‫ﺩﻳﻦ‬
dinar diːnɑːr ‫ﺩﻳﻨﺎﺭ‬
diwan diːwɑːn ‫ﺩﻳﻮﺍﻥ‬
golden ˈðaɦabiː ‫ﺫﻫﺒﻲ‬
taste ðoːq (†) ‫ﺫﻭﻕ ﺍﺳﻢ‬
he organized rattab ‫ﺭﺗﺐ‬
return rid͡ʒuːʕ ‫ﺭﺟﻮﻉ‬
sleeve ridin ‫ﺭﺩﻥ‬
wrist, forearm risq ~ rizɨɣ ‫ﺭﺳﻎ‬
loss rɨsuːb ‫ﺭﺳﻮﺏ‬
humidity rṭuːbә ‫ﺭﻃﻮﺑﺔ‬
he lifted rifaʕ ~ rfaʕ ‫ﺭﻓﻊ ﻓﻌﻞ ﻣﺎﺿﻲ‬
neck ɨrɡubә ‫ﺭﻗﺒﺔ‬
corner rɨkɨn ‫ﺭﻛﻦ ﺍﺳﻢ‬
he increased zɑːd ‫ﺯﺍﺩ‬
butter/essence (?) zibde ~ zɨbde ‫ﺯﺑﺪﺓ ﻟﺐ ﺍﻟﻤﻮﺿﻮﻉ‬
oil zeːt ‫ﺯﻳﺖ‬
shin sɑːɡ ‫ﺳﺎﻕ‬
poisonous sɑːm ‫ﺳﺎﻡ‬
stuff sɑːmɑːn ‫ﺳﺎﻣﺎﻥ‬
stalk sɨbɨl ‫ﺳﺒﻞ‬
lotus sidir ‫ﺳﺪﺭ‬
onesixth sidis ‫ﺳﺪﺱ ﺍﻟﺮﻗﻢ‬
roof ṣɑṭɑħ ~ ṣɑṭħ(†) ‫ﺳﻄﺢ‬
bucket ṣɑṭɑḷ ‫ﺳﻄﻞ‬
silence sikuːt ‫ﺳﻜﻮﺕ‬
age/tooth sinn ‫ﺳﻦ ﻋﻤﺮ‬
Sudan suːdɑːn ‫ﺳﻮﺩﺍﻥ‬
termites suːs ‫ﺳﻮﺱ‬
Table B.1 continued

sword seːf ‫ﺳﻴﻒ‬
falcon ʃɑːɦiːn ‫ﺷﺎﻫﻴﻦ‬
lion cub ʃɨbɨl ‫ﺷﺒﻞ‬
a tree iʃjara ‫ ﺷﺠﺮﻩ ﻧﺒﺎﺕ‬/ ‫ﺷﻴﺮﺓ‬
a piece of fat ʃaħma(‡) ~ ʃħama ‫ﺷﺤﻤﺔ‬
he tightened ʃadd ‫ﺷﺪ ﻓﻌﻞ ﻣﺎﺿﻲ‬
tighten! ʃidd ‫ﺷﺪ! ﺃﻣﺮ‬
he bought ʃira ‫ﺷﺮﻯ‬
gratitude ʃɨkɨr ‫ﺷﻜﺮ ﻣﺼﺪﺮ‬
soap ṣɑːbuːn ~ ṣɑːboːn ‫ﺻﺎﺑﻮﻥ‬
soaps ṣɑːboːnɑːt (†)(= ‫ﺻﺎﺑﻮﻧﺎﺕ‬
ṣɑwɑːbiːn)
name of an Arabic letter ṣɑːd ‫ﺻﺎﺩ ﺣﺮﻑ ﺍﻟﺼﺎﺩ‬
salon; broth ṣɑːlawn ~ ṣɑːlɑːn ‫ﺻﺎﻟﻮﻥ‬
a finger ṣubɑʕ ‫ﺻﺒﻊ‬
healthy ṣɑħħɑ ‫ﺻﺤﺔ‬
a stone ṣɑχra ‫ﺻﺨﺮﻩ‬
a goat ɨṣχɑḷɑ ‫ﺻﺨﻠﻪ‬
a seashell ṣɨdfɘ ~ ˈṣɑdafɑ(‡) ‫ﺻﺩﻓﺔ ﻋﻠﻰ ﺍﻟﺴﻴﻒ‬
zero ṣufɨr ~ ṣɨfɨr ‫ﺻﻔﺮ ﺃﻗﻞ ﻣﻦ ﻭﺍﺣﺪ‬
autumn (u)ṣfiriː ~ ṣɨfriː(†) ‫ﺻﻔﺮﻱ‬
he arranged (into rows) ṣɑffɑf ‫ﺻﻔﻒ‬
he closed ṣɑkk ‫ﺻﻚ ﻓﻌﻞ ﻣﺎﺿﻲ‬
close! ṣɨkk ‫ﺻﻚ! ﺃﻣﺭ‬
direction ṣoːb ‫ﺻﻮﺏ ﻣﺜﻞ ﺍﺗﺠﺎﻩ‬
picture ṣuːrә ‫ﺻﻮﺭﺓ‬
fasting ṣoːm ‫ﺻﻮﻡ‬
nits (coll.) ṣiːbɑːn ‫ﺻﻴﺒﺎﻥ‬
summer ṣeːf ‫ﺻﻴﻒ‬
guest ð ̣eːf ‫ﺿﻴﻒ ﺷﺨﺺ ﻓﻲ ﺑﻴﺘﻚ‬
it fell ṭɑːħ ‫ﻃﺎﺡ‬
mill ṭɑːħuːn ‫ﻃﺎﺣﻮﻥ‬
it was long, may it be long ṭɑːl ‫ﻃﺎﻝ‬
medicine ṭubb ‫ﻃﺐ‬
flour uṭħiːn ‫ﻃﺤﻴﻦ‬
he hit ṭɑɡɡ ‫ﻃﻖ ﻓﻌﻞ ﻣﺎﺿﻲ‬
he hit me ˈṭɑɡɡiniː ~ ṭɑɡˈɡәniː ‫ﻃﻘﻨﻲ‬
dust, duststorm ṭoːz ‫ﻃﻮﺯ ﻋﺎﺻﻔﺔ ﺭﻣﻠﻴﺔ‬
perfume ṭiːb ‫ﻃﻴﺐ ﺍﺳﻢ ﻣﺜﻞ ﻋﻮﺩ‬
bird ṭeːr ‫ﻃﻴﺮ‬
mud ṭiːn ‫ﻃﻴﻦ‬
he returned ʕɑːd ‫ﻋﺎﺩ ﻓﻌﻞ ﻣﺎﺿﻲ‬
duststorm ʕɑːfuːr ‫ﻋﺎﻓﻮﺭ ﻋﺎﺻﻔﺔ ﺭﻣﻠﻴﺔ‬
high ʕɑːliː ‫ﻋﺎﻟﻲ‬
91
Table B.1 continued

year ʕɑːm ‫ﻋﺎﻡ‬
two years ʕɑːmeːn ‫ﻋﺎﻣﻴﻦ‬
slave ʕɑbd ‫ﻋﺒﺪ‬
a window sill ʕtɨbɘ(ɦ) ‫ﻋﺘﺒﺔ‬
Arab (sing.) ʕɑrɑbiː ‫ﻋﺮﺑﻲ‬
prone to anger ʕɑṣɑbiː ‫ﻋﺼﺒﻲ‬
knot ʕɡideɦ ~ ˈʕuqәda (‡) ‫ﻋﻘﺪﺓ ﻣﻦ ﺣﺒﻞ‬
knowledge ʕilm (†) ‫ﻋﻟﻢ ﺍﺳﻢ ﺍﻟﻤﻌﺮﻓﺔ‬
he knew ʕɨlam ‫ﻋﻠﻢ ﻓﻌﻞ ﻣﺎﺿﻲ‬
male name ʕɑliː ‫ﻋﻠﻲ ﺍﺳﻢ‬
great ʕoːd ‫ﻋﻮﺩ ﻳﻌﻨﻲ ﻛﺒﻴﺮ‬
incense ʕuːd ‫ﻋﻮﺩ‬
holiday ʕiːd ‫ﻋﻴﺪ‬
camels (coll.) ʕiːs ‫ﻋﻴﺲ‬
eye; spring ʕeːn ‫ﻋﻴﻦ‬
two eyes ʕeːneːn ‫ﻋﻴﻨﻴﻦ‬
shipwrecked ɣɑːrɡe ‫ﻏﺎﺭﻗﺔ‬
name of an Arabic letter ɣeːn ‫ﻏﻴﻦ‬
he reacted fɑːʕal ‫ﻓﺎﻋﻞ‬
opening (n.) fatħ ‫ﻓﺘﺢ‬
he divided fɨṣɑl ‫ﻓﺼﻞ ﻓﻌﻞ ﻣﺎﺿﻲ‬
division fɑṣɨl ‫ﻓﺼﻞ ﺍﺳﻢ‬
desert truffles faɡiʕ ‫ﻓﻘﻊ‬
hotel fɨndɨq ~ funduq(†) ‫ﻓﻨﺪﻖ‬
male name; leopard fahad ‫ﻓﻬﺪ‬
above foːɡ ‫ﻓﻮﻕ‬
upper foːɡɑːniː ‫ﻓﻮﻗﺎﻧﻲ‬
there is/in it fiːɦ ‫ﻓﻴﻪ‬
before us ɡɑˈbuḷna ‫ﻗﺒﻠﻨﺎ‬
already ɡid ~ ɡidd ‫ﻗﺪ‬
he slaughtered ɡiṣɑb ~ qɑṣɑb(†) ‫ﻗﺼﺐ ﻓﻌﻞ ﻣﺎﺿﻲ‬
cotton ɡiṭɑn ‫ﻗﻄﻦ‬
I said to him ɡitlә ~ ɡiltla ‫ﻗﻠﺘﻠﻪ‬
coffee ɨɡɦawa ~ ɡaɦwa (‡) ‫ﻗﻬﻮﻯ‬
closet kabat ‫ﻛﺒﺖ‬
liver ͡ tʃabd ~ kabid ‫ ﭽﺒﺪ‬/ ‫ﻛﺒﺪ‬
books kɨtɨb ‫ﻛﺘﺐ‬
she wrote ɨktibat ~ kitbat(‡) ‫ﻛﺘﺒﺖ‬
kohl kiħɨl ‫ﻛﺤﻞ‬
mule — ‫ﻛﺪﺵ‬
treasure kanz ‫ﻛﻨﺯ‬
meat laħam ‫ﻟﺤﻢ‬
melody laħin ‫ﻟﺤﻦ ﺍﺳﻢ‬
necessities lɨzuːm ‫ﻟﺯﻭﻡ‬
Table B.1 continued

playing (v.n.) laʕab ‫ﻟﻌﺐ ﻓﻌﻞ ﻣﺎﺿﻲ‬
light leːt ‫ﻟﻴﺖ‬
lights leːtɑːt ‫ﻟﻴﺘﺎﺕ‬
available for you mitˈfajjɡәlik(‡) ‫ﻣﺘﻔﻴﻖ ﻟﻚ‬
neutral mħɑːjid ‫ﻣﺤﺎﻳﺪ‬
burnt maħruːɡ ‫ﻣﺤﺮﻭﻕ‬
stuffed maħʃiː ‫ﻣﺤﺸﻲ‬
oceans mħiːṭɑːt ~ muħiːṭɑːt(†) ‫ﻣﺤﻴﻄﺎﺕ‬
library ˈmaktaba(†) ~ makˈtiba ‫ﻣﻜﺘﺒﺔ‬
models moːdeːlɑːt ‫ﻣﻮﺩﻳﻼﺕ‬
topic mawð ̣ūʕ ~ mōð ̣ūʕ ‫ﻣﻮﺿﻮﻉ‬
a bee naħla ‫ﻧﺤﻠﺔ‬
trouble niʃab ‫ﻧﺸﺐ ﺍﺳﻢ‬
it stuck niʃab ‫ﻧﺸﺐ ﻓﻌﻞ ﻣﺎﺿﻲ‬
eyelash — ‫ﻫﺪﺐ‬
garment(s)? — ‫ﻫﺪﻢ ﺟﻤﻊ‬
garment(s)? — ‫ﻫﺪﻡ ﻣﻔﺮﺩ‬
this hɑːða ‫ﻫﺫﺍ‬
one wɑːħid ‫ﻭﺍﺣﺪ‬
piece of paper; playing wriɡe ~ ˈwaraqa(†) ‫ﻭﺭﻗﺔ‬
cards
they drink iːʃrubuːn ~ iːʃrbuːn ‫ﻳﺸﺮﺑﻮﻥ‬
Yemeni ˈyamaniː ‫ﻳﻤﻨﻲ‬
it is broken iːnˈkisar ‫ﻳﻨﻜﺴﺮ‬
hungry (pl.) yawʕɑːniːn ‫ﻳﻮﻋﺎﻧﻴﻦ‬
Appendix C
Statistical results
C.1 Long vowel model
Table C.1 gives Fstatistics and pvalues for all statistical tests in the first statistical
model, described in 2.1.1.
This battery of ANOVA tests was performed on measurements of 2779 vowels.
In the code below, all variables listed after the equals sign are independent variables.
The asterisk means that the model is testing for interaction between two or more vari
ables.
lmer([ depvar ] = (ListPosition + Vowel + Stress + BackingConsBefore + Backing

ConsAfter + Vowel * Stress + Vowel * BackingConsBefore + Vowel * BackingConsAfter)
+ (1|Speaker))
93
94 APPENDIX C. STATISTICAL RESULTS
Table C.1: Statistical results for vowels

Variable(s) F1 F2 F0 dB Duration
ListPosition F(2) = 2.3 F(2) = 0.4 F(2) = 13 F(2) = 223 F(2) = 9.2
{Initial, Medial, Final} p < 0.1 . p > 0.1 p < 0.001 *** p < 0.001 *** p < 0.001 ***
Vowel aː, eː, iː, a, i F(4) = 935 F(4) = 396 F(4) = 65 F(4) = 3.2 F(4) = 600
p < 0.001 *** p < 0.001 *** p < 0.001 *** p < 0.05 * p < 0.001 ***
BackingConsBefore F(1) = 0.3 F(1) = 255 F(1) = 0.2 F(1) = 0 F(1) = 7.5
p > 0.1 p < 0.001 *** p > 0.1 p > 0.1 p < 0.01 **
BackingConsAfter F(1) = 3.9 F(1) = 74 F(1) = 0.6 F(1) = 6.1 F(1) = 70
p < 0.05 * p < 0.001 *** p > 0.1 p < 0.05 * p < 0.001 ***
Stress F(1) = 12 F(1) = 8.5 F(1) = 662 F(1) = 97 F(1) = 155
p < 0.001 *** p < 0.01 ** p < 0.001 *** p < 0.001 *** p < 0.001 ***
Vowel* F(4) = 12 F(4) = 43 F(4) = 0.6 F(4) = 2.1 F(4) = 7.4
BackingConsBefore p < 0.001 *** p < 0.001 *** p > 0.1 p < 0.1 . p < 0.001 ***
Vowel* F(4) = 1.7 F(4) = 20 F(4) = 6.2 F(4) = 10 F(4) = 11
BackingConsAfter p > 0.1 p < 0.001 *** p < 0.001 *** p < 0.001 *** p < 0.001 ***
Vowel*Stress F(4) = 19 F(4) = 12 F(4) = 16 F(4) = 0.8 F(4) = 25
p < 0.001 *** p < 0.001 *** p < 0.001 *** p > 0.1 p < 0.001 ***
C.2 Short vowel model
Table C.2 gives Fstatistics and pvalues for all statistical tests in the statistical model
for short vowels, described in 2.3.1.
This battery of ANOVA tests was performed on measurements of 1992 vowels.

The tests were as follows:
lmer([ depvar ] = (ListPosition + Vowel + Stress + Underlying + BackingConsBefore

+ BackingConsAfter + SyllableType + WordPosition + Vowel * Stress + Vowel *
Underlying + Vowel * BackingConsBefore * + Vowel * BackingConsAfter +
BackingConsBefore * BackingConsAfter + Vowel * BackingConsBefore *
BackingConsAfter + Vowel * SyllableType) + (1|Speaker))
C.3. VOWEL RAISING MODEL 95
Table C.2: Statistical results for short vowels

ListPosition F(2) = 2.4 F(2) < 0.1 F(2) = 6.7 F(2) = 150 F(2) = 3.9
{Initial, Medial, Final} p < 0.1 . p > 0.1 p < 0.01 ** p < 0.001 *** p < 0.05 *
Vowel {a, i} F(1) = 1241 F(1) = 2.6 F(1) = 34 F(1) = 4.4 F(1) = 122
p < 0.001 *** p > 0.1 p < 0.001 *** p < 0.05 * p < 0.001 ***
Stress F(1) = 85 F(1) = 0.4 F(1) = 511 F(1) = 142 F(1) = 28
p < 0.001 *** p > 0.1 p < 0.001 *** p < 0.001 *** p < 0.001 ***
Underlying F(1) = 34 F(1) = 3.7 F(1) = 17 F(1) = 13 F(1) = 23
p < 0.001 *** p < 0.1 . p < 0.001 *** p < 0.001 *** p < 0.001 ***
BackingConsBefore F(1) = 20 F(1) = 650 F(1) < 0.1 F(1) = 0.2 F(1) = 12
p < 0.001 *** p < 0.001 *** p > 0.1 p > 0.1 p < 0.001 ***
BackingConsAfter F(1) = 4.8 F(1) = 349 F(1) = 8.4 F(1) = 7.3 F(1) = 25
p < 0.05 * p < 0.001 *** p < 0.01 ** p < 0.01 ** p < 0.001 ***
SyllableType F(1) = 5.1 F(1) = 0.1 F(1) = 9.4 F(1) < 0.1 F(1) = 0.2
p < 0.05 * p > 0.1 p < 0.01 ** p > 0.1 p > 0.1
WordPosition F(1) = 11 F(1) = 3.8 F(1) = 3.6 F(1) = 0.4 F(1) = 19
p < 0.001 *** p < 0.1 . p < 0.1 . p > 0.1 p < 0.001 ***
Vowel*Stress F(1) = 16 F(1) = 0.2 F(1) = 2.5 F(1) = 3.2 F(1) = 0.2
p < 0.001 *** p > 0.1 p > 0.1 p < 0.1 . p > 0.1
Vowel* F(1) = 6 F(1) = 3.2 F(1) = 1.1 F(1) = 5.2 F(1) = 1
Underlying p < 0.05 * p < 0.1 . p > 0.1 p < 0.05 * p > 0.1
Vowel* F(1) = 0.4 F(1) = 75 F(1) = 1.7 F(1) = 0.7 F(1) = 2.1
BackingConsBefore p > 0.1 p < 0.001 *** p > 0.1 p > 0.1 p > 0.1
Vowel* F(1) = 2.8 F(1) = 44 F(1) = 1.4 F(1) = 4 F(1) = 0.2
BackingConsAfter p < 0.1 . p < 0.001 *** p > 0.1 p < 0.05 * p > 0.1
BackingConsBefore* F(1) = 1 F(1) = 26 F(1) = 0.3 F(1) = 5.6 F(1) = 0.3
BackingConsAfter p > 0.1 p < 0.001 *** p > 0.1 p < 0.05 * p > 0.1
Vowel* F(1) = 17 F(1) = 6.5 F(1) = 23.7 F(1) = 3.4 F(1) = 4.9
SyllableType p < 0.001 *** p < 0.05 * p < 0.001 *** p < 0.1 . p < 0.05 *
Vowel*BackingConsBefore* F(1) = 13 F(1) = 0.1 F(1) = 1.4 F(1) = 6 F(1) = 6.5
BackingConsAfter p < 0.001 *** p > 0.1 p > 0.1 p < 0.05 * p < 0.05 *
C.3 Vowel raising model
Table C.3 gives Fstatistics and pvalues for all statistical tests in the third statistical
model, described in 2.4. This model was designed to test for effects on vowel height
(F1) in /a/. In total, measurements of 1144 vowels were included in this test.
lmer(F1 = (ListPosition + Stress + Underlying + BackingConsBefore +

BackingConsAfter + SyllableType + WordPosition + AfterHighVowel +
FeminineSingular + BackingConsBefore * BackingConsAfter + AfterHighVowel *
FeminineSingular) + (1|Speaker))
Table C.3: Statistical results for wordfinal vowel raising

Variable(s) F1
ListPosition F(1) = 2.7
p < 0.1 .
Stress F(1) = 46
p < 0.001 ***
Underlying F(1) = 18
p < 0.001 ***
BackingConsBefore F(1) = 18
p < 0.001 ***
BackingConsAfter F(1) = 2.2
p > 0.1
SyllableType F(1) = 2.6
p > 0.1
WordPosition F(1) = 2.2
p > 0.1
AfterHighVowel F(1) = 10
p < 0.01 **
FemSingular F(1) = 26
p < 0.001 ***
BackingConsBefore* F(1) = 12
BackingConsAfter p < 0.001 ***
AfterHighVowel* F(1) = 23
FemSingular p < 0.001 ***
C.4 Epenthetic vowel model
Table C.4 gives Fstatistics and pvalues for all statistical tests in the statistical model
described §4.3.2, which was created to investigate acoustic phonetic differences
between lexical epenthesis and postlexical epenthesis.
This battery of ANOVA tests was performed on measurements of 225 epenthetic
vowels. Based on the analysis given in Chapter 4, 194 of these epenthesis tokens were
coded as lexical epenthesis, and the remaining 31 were coded as postlexical
epenthesis.
The tests were as follows:
lmer([ depvar ] = (ListPosition + Stress + BackingConsonantBefore +

BackingConsonantAfter + BackingConsonantBefore * BackingConsonantAfter +
SyllableType + WordPosition + EpenthesisType) + (1|Speaker))
C.4. EPENTHETIC VOWEL MODEL 97
Table C.4: Statistical results for epenthetic vowels

ListPosition F(2) = 0.7 F(2) = 0.1 F(2) = 1.9 F(2) = 21 F(2) = 0.2
{Initial, Medial, Final} p > 0.1 p > 0.1 p > 0.1 p < 0.001 *** p > 0.1
Stress F(1) = 5.5 F(1) < 0.1 F(1) = 1.7 F(1) = 18 F(1) = 3.8
p < 0.05 * p > 0.1 p > 0.1 p < 0.001 *** p < 0.1 .
BackingConsBefore F(1) = 6.3 F(1) = 66 F(1) = 0.5 F(1) = 13 F(1) = 7.3
p < 0.05 * p < 0.001 *** p > 0.1 p < 0.001 *** p < 0.01 **
BackingConsAfter F(1) = 6.9 F(1) = 40 F(1) < 0.1 F(1) = 2.7 F(1) = 0.8
p < 0.01 ** p < 0.001 *** p > 0.1 p < 0.1 . p > 0.1
BackingConsBefore* F(1) = 0.3 F(1) = 3.4 F(1) = 2.7 F(1) = 25 F(1) = 1.6
BackingConsAfter p > 0.1 p < 0.01 . p > 0.1 p < 0.001 *** p > 0.1
SyllableType F(1) = 3.7 F(1) = 11 F(1) < 0.1 F(1) = 0.7 F(1) < 0.1
p > 0.1 p < 0.01 ** p > 0.1 p > 0.1 p > 0.1
WordPosition F(1) = 36 F(1) = 0.6 F(1) = 3.2 F(1) = 17 F(1) = 3.6
p < 0.001 *** p > 0.1 p < 0.1 . p < 0.001 *** p < 0.1 .
EpenthesisType F(1) = 0.9 F(1) = 6.6 F(1) = 0.8 F(1) < 0.1 F(1) = 7.9
p > 0.1 p < 0.05 * p > 0.1 p > 0.1 p < 0.01 **
Appendix D
Supplemental figures
The figures presented here are statistically significant results that are not directly
relevant to the main findings of this thesis.
Figures D.1, D.2, D.3, and D.4 represent findings mentioned in §2.1.1. The presence
of backing consonants in an adjacent consonant results in higher F1 (corresponding to
vowel lowering), higher pitch, higher intensity, and shorter duration in Urban Qatari
Arabic vowels; some of these effects appear to be more prominent in short vowels,
which are subject to stronger conditioning effects from surrounding consonants.
99
100 APPENDIX D. SUPPLEMENTAL FIGURES
Figure D.1: Effect of adjacent backing consonants on vowel height (F1), by length
Figure D.2: Effect of adjacent backing consonants on fundamental frequency, by length
Figure D.3: Effect of adjacent backing consonants on vowel intensity, by length

101
Figure D.4: Effect of adjacent backing consonants on vowel duration, by length
Figures D.5, D.6, and D.7 represent statistically significant findings mentioned in
§2.3.1. The presence of backing consonants has a shortening effect on both /i/ and /a/.
Backing consonants have a lowering effect on pitch in /i/, but the opposite effect is
shown for /a/.
Figure D.5: Effect of adjacent backing consonants on duration of short vowels, by vowel
102 APPENDIX D. SUPPLEMENTAL FIGURES
Figure D.6: Effect of adjacent backing consonants on fundamental frequency of short

vowels, by vowel
Figure D.7: Effect of adjacent backing consonants on intensity of short vowels, by vowel

The Vowels of Urban Qatari Arabic

Uploaded by

Copyright:

Available Formats

The Vowels of Urban Qatari Arabic

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Vowels of Urban Qatari Arabic

Uploaded by

Copyright:

Available Formats

University of North Dakota

UND Scholarly Commons

Theses and Dissertations Theses, Dissertations, and Senior Projects

The vowels of Urban Qatari Arabic

Follow this and additional works at: https://commons.und.edu/theses

by Mark Daniel Shockley

Bachelor of Arts, Sam Houston State University, 2009

in partial fulfillment of the requirements

for the degree of

Grand Forks, North Dakota

Title The Vowels of Urban Qatari Arabic

Degree Master of Arts

Mark Daniel Shockley

December 14, 2020

A Variation in stress assignment 85

1.1 Gulf Arabic in geographical context . . . . . . . . . . . . . . . . . . . 5

2.1 Effect of list position on fundamental frequency, by speaker . . . . . . . 22

3.1 Effect of stress and word position on vowel duration, by length . . . . . 54

4.1 Statistically significant differences between lexical and epenthetic vowels 67

2.1 Vowel formants in Urban Qatari Arabic (speaker 1, male) . . . . . . . . 24

4.1 Epenthesis by sonority sequence in Gulf Arabic . . . . . . . . . . . . . 60

A.1 Accounts of stress in North Arabian dialects . . . . . . . . . . . . . . . 86

C.1 Statistical results for vowels . . . . . . . . . . . . . . . . . . . . . . . 94

1.1 Problem and findings

1.2 The state of Arabic dialectology

1.3 Urban Qatari Arabic in context

1.3.1 Classifying Arabic dialects

Although Arabic is popularly framed as a single language spoken by hundreds of mil­

1.3.2 North Arabian dialects

Figure 1.1: Gulf Arabic in geographical context

1982:1; 1994:8­9). Gulf Arabic is understood, therefore, to be closely akin to Najdi

North Arabian dialects

Mesopotamian Arabic Najdi Arabic Gulf Arabic

Northern Sunni Bahraini

Northern­Central Urban Emirati

Central Urban Qatari

ʔāl murrah mahānida

banī hād͡ʒir ʔāl bū kuwāra

al­manāṣīr banī kaʕb

Figure 1.2: Qatari dialects within North Arabian dialects

1.3.3 Language contact in the Gulf

1.3.4 Linguistic situation

1.4 Existing research

T. M. Johnstone’s Eastern Arabian dialect studies (1967b; data, 1958­1959) is a ground­

Qatar is sometimes regarded in international relations as the most isolated of the

What follows is an account of primary sources on Arabic varieties spoken in Qatar.

Najdi dialect spoken by the Āl Murrah Bedouin tribe in Qatar.

1.5.1 Researcher qualifications

Figure 1.3: Spectrogram of /ʔidāra/ [ḭdāra] ‘administration’ (Speaker 6)

Figure 1.4: Spectrogram of /yiʃribūn/ [īʃrbōn] ‘they drink’ (Speaker 1)

Figure 1.5: Spectrogram of /χirza/ [χrza] ‘a bead’ (Speaker 1)

1.5.6 Statistical methods

Two batteries of repeated­measures (within­speaker) analyses of variation (ANOVAs)

A second battery of ANOVAs was performed using 1992 measurements of short

1.5.7 Supplemental data

Two dichotomous factor variables BackingConsBefore and BackingConsAfter were

The dichotomous variable Stress {Stressed, Unstressed} was also introduced as an

lmer([ depvar ] = (ListPosition + Vowel + Stress + BackingConsBefore + Backing­

Figure 2.1: Effect of list position on fundamental frequency, by speaker

Figure 2.2: Effect of list position on intensity and duration

Although Arabic is popularly framed as a single language spoken by hundreds of mil

1982:1; 1994:89). Gulf Arabic is understood, therefore, to be closely akin to Najdi

NorthernCentral Urban Emirati

almanāṣīr banī kaʕb

T. M. Johnstone’s Eastern Arabian dialect studies (1967b; data, 19581959) is a ground

Two batteries of repeatedmeasures (withinspeaker) analyses of variation (ANOVAs)

lmer([ depvar ] = (ListPosition + Vowel + Stress + BackingConsBefore + Backing

(3) a. sidre ‘a lotus tree’

b. sidra ‘his lotus trees’ (unattested)

b. /ɡālōa/ → ɡāˈlō ‘they said it’

b. tid͡ʒamˈɦaraw ‘they gather together’ (In isolation or phrasefinal)

(6) mōð ̣ūʕ ‘topic’ [ √wð ̣ʕ ] (Speaker 1)

(8) dawrāt ‘courses’ [ √dwr ]3

Sulaiti (1993:137138) offers two other minimal pairs, given below.

Data from Mustafawi (2006:165296) anticipates a lack of phonemic contrast between

lmer([ depvar ] = (ListPosition + Vowel + Stress + Underlying + BackingCons

2.3.2 Consonantconditioned vowel backing

Short vowels are nearfront when no backing consonants are present:

(18) a. θilθ ‘onethird’