A Phonological Typology of Modern Arabic Varieties, Lecture notes of Arabic

Download A Phonological Typology of Modern Arabic Varieties and more Lecture notes Arabic in PDF only on Docsity! A Phonological Typology of Modern Arabic Varieties Emily Lindsay-Smith St. Catherine’s College University of Oxford A thesis submitted for the degree of Doctor of Philosophy Hilary 2021 Acknowledgements This thesis would not have been possible without the support and encouragement of many people. I am profoundly grateful to all those who have helped me to this point. First and foremost, I wish to express my heartfelt thanks to my supervisor, Professor Aditi Lahiri. Over the past four and a half years, it has been a privilege and honour to receive such steadfast support and guidance. There are no words sincere and deep enough to express my gratitude for all you have done for me. I must also thank those who have played a role in developing the ideas of this thesis. Particular thanks go to Wolfgang de Melo and Holly Kennard for their useful feedback at my Confirmation of Status; subsequent thanks also to audiences at the Manchester Phonology Meeting and Phonetik und Phonologie Tagung who heard earlier versions of this work; anonymous reviewers at the Texas Linguistics Society; not to mention Ricardo Bermudez-Otero for his useful insights. I am also very grateful to my examiners, Elan Dresher and Wolfgang de Melo, for an enjoyable viva experience and valuable comments on my thesis that have improved this final version. I am deeply thankful to my linguistics colleagues and friends at the Language and Brain Laboratory at Oxford, who have provided me with so much support, cups of tea, and fascinating chats. I have missed our lunches so much during the interruptions of lockdown and COVID, and I am so grateful for the opportunity to know you all. Thank you to Yoolim, Kim, Nina, Swetlana, Sandra, Hilary, Emily D., Colin, Henning, Yaxuan, Holly, Josh, Zoe, Anna, and to my non-LABLAB friends: Maren, Hannah, Amanda, Edwin, Raquel, Fin. Particular thanks must go to Colin, Amanda, Emily D. and Anna for proofreading, and to Leah for her help compiling the data appendix. Your support has been invaluable, and all remaining errors are my own. Without the generous support of the Economic and Social Research Council, this work would not have been possible, and I would not have been able to spend the past few years immersing myself in a topic that has brought me such joy, and so bored my (non-linguist) friends and family. I never would have made it this far without the ongoing support of my family. Thank you to my parents, Yvonne and Andrew Lindsay, who have always believed that I could do anything and so completely supported me. Thank you to my sisters, Joey and Georgie, for picking me up when things are hard, and putting up with nonsensical linguistics chat. Thank you to my grandmothers, Pat Jones and Elisabeth Lindsay, your support and love is deeply felt, and I wish my grandfathers were here to see this. Thank you to my in-laws, Diana and Kevin Smith, as well as Ellie, Joe and Sam Williams. Your support has been invaluable and I feel so welcomed into your family. I cannot list the myriad ways in which each of you have been there, but you have made all the difference. Finally, and most importantly, I give my deepest thanks to my wife, Tris. Thank you for being there for me throughout, for loving me when I am stressed and when I am excited and all the states in between. You are the best woman I have ever met. You may not believe in syllables but I love you anyway. Abstract Modern Arabic colloquial varieties exhibit significant phonological variation that has only partially been incorporated into previous typological efforts. This thesis motivates a phonological typology of sixteen modern varieties, exploring not just the crucial ingredients of the phonological grammar including syllable and foot structure, but also the lexical and postlexical nature of prosodic structure. I argue that there are two axes for phonological variation in Arabic: Tolerance and Repair, that is, varieties differ in terms of which types of syllables they tolerate and in how they repair violations of syllable structure, and these axes do not covary. In terms of Tolerance, varieties differ in acceptance of quantity. Varieties that permit long segments have restrictions on the segmental level, whereas varieties that do not have restrictions on the moraic level, and varieties that allow long consonants or clusters but not long vowels have restrictions on the X-slot tier. In addition to the canonical 2 or 3 elements, individual varieties may also permit an extra element in their medial syllable rime according to variety-specific restrictions including coronality, fast speech, and sonority. As such, this typology provides mechanisms to account for the data in each and every variety, rather than merely providing a canonical set of features that not all varieties fit. Repair is concerned with how violations of syllable structure are repaired in terms of epenthesis direction and the licensing of word edge extrasyllabic segments. I propose that catalectic moras are required to incorporate extrasyllabic material across the word. The Onset/Coda distinction can therefore be expressed as whether the extrasyllabic material is incorporated as the onset or coda of a syllable containing a catalectic mora. Epenthesis occurs where the catalectic mora is not permitted to surface without overt segmental material. The difference between domain edge behaviour therefore follows from this — where the catalectic mora is the peripheral element, the surface cluster is permitted; but where it is non- peripheral, epenthesis occurs as the catalectic mora is filled by an epenthetic vowel. ‘Extreme’ Onset/Coda varieties are those that permit catalectic moras word-internally in domain edge syllables. Furthermore, I explore the prosodic word structure of Arabic, finding that there is a preference for affixes to attach after a stressed syllable that accounts for lengthening phenomena and apparent stress exceptions. x Contents 2 The Role of Quantity in Syllable Structure Restrictions 39 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.2 Previous Work on Arabic Syllable Structure Typology . . . . . . . . 41 2.3 Previous Work on Quantity . . . . . . . . . . . . . . . . . . . . . . 53 2.4 Analysis of Sixteen Varieties . . . . . . . . . . . . . . . . . . . . . . 57 2.4.1 Cairene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.4.2 SQanQa:ni: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.4.3 Mak:an . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 2.4.4 Iraqi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 2.4.5 Qatari . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 2.4.6 Muscat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 2.4.7 Palestinian . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 2.4.8 Lebanese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 2.4.9 Wadi Ram: . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 2.4.10 Tunisian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 2.4.11 Algerian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 2.4.12 Moroccan Casablanca . . . . . . . . . . . . . . . . . . . . . 107 2.4.13 Libyan Tripoli . . . . . . . . . . . . . . . . . . . . . . . . . . 110 2.4.14 Rufaidah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 2.4.15 Rwaili . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 2.4.16 Ha:yil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 2.4.17 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 2.5 Tolerance of Medial Syllables . . . . . . . . . . . . . . . . . . . . . 129 2.5.1 Tunisian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 2.5.2 Wadi Ram: . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 2.5.3 Rufaidah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 2.5.4 Rwaili . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 2.5.5 Ha:yil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 2.5.6 Algerian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 2.5.7 Palestinian . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 2.5.8 Lebanese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 2.5.9 Muscat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 2.5.10 SQanQa:ni: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 2.5.11 Iraqi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 2.5.12 Qatari . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 2.5.13 Mak:an . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 2.5.14 Cairene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 2.5.15 Moroccan Casablanca . . . . . . . . . . . . . . . . . . . . . 154 2.5.16 Libyan Tripoli . . . . . . . . . . . . . . . . . . . . . . . . . . 156 Contents xi 2.5.17 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 2.6 New Proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 2.6.1 Tolerance . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 2.6.2 Analysis of Sixteen Varities . . . . . . . . . . . . . . . . . . 164 2.6.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 2.6.4 Repair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 2.7 Conclusion and Next Steps . . . . . . . . . . . . . . . . . . . . . . . 181 3 The Relationship between Stress and Syllable Structure 183 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 3.1.1 Previous Approaches to Superheavy Syllables . . . . . . . . 184 3.2 Approaches to Stress and Extrametricality . . . . . . . . . . . . . . 187 3.2.1 What is Stress? . . . . . . . . . . . . . . . . . . . . . . . . . 187 3.2.2 Stress Analyses . . . . . . . . . . . . . . . . . . . . . . . . . 187 3.2.3 Degenerate Feet and Extrametricality . . . . . . . . . . . . . 194 3.2.4 Catalexis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 3.3 Analysis of Stress in Arabic Varieties . . . . . . . . . . . . . . . . . 199 3.3.1 Pattern 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 3.3.2 Pattern 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 3.3.3 Pattern 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 3.3.4 Pattern 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 3.3.5 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251 3.4 Superheavy Syllables: The Shape of Final Syllables . . . . . . . . . 252 3.4.1 Catalexis Revisited . . . . . . . . . . . . . . . . . . . . . . . 255 3.4.2 Catalectic X-slot . . . . . . . . . . . . . . . . . . . . . . . . 264 3.5 Correlation with Syllable Typology . . . . . . . . . . . . . . . . . . 265 3.6 Furthering the Syllable Structure Typology . . . . . . . . . . . . . . 267 3.7 Next Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 4 The Classification of Clitics and Affixes 271 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 4.1.1 Structure of the Chapter . . . . . . . . . . . . . . . . . . . . 276 4.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 4.2.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 4.3 Diagnosing the Clitic versus Affix Distinction . . . . . . . . . . . . 281 4.3.1 Host Selectivity and Ordering . . . . . . . . . . . . . . . . . 281 4.3.2 Arbitrary Gaps . . . . . . . . . . . . . . . . . . . . . . . . . 284 4.3.3 Morphophonological Idiosyncracies . . . . . . . . . . . . . . 286 4.3.4 Conclusion: Clitic versus Affix . . . . . . . . . . . . . . . . . 295 4.4 Pre-‘clitic’ lengthening . . . . . . . . . . . . . . . . . . . . . . . . . 297 xii Contents 4.4.1 Pre-subject lengthening . . . . . . . . . . . . . . . . . . . . 315 4.4.2 Conclusion: Pre-affix Lengthening . . . . . . . . . . . . . . . 321 4.5 Exceptions to Stress Parameters . . . . . . . . . . . . . . . . . . . . 321 4.5.1 Pre-Affix Stress Assignment . . . . . . . . . . . . . . . . . . 321 4.5.2 Feminine Past Tense Verbs with Pronominal Objects . . . . 325 4.5.3 Final Stressed Long Vowels . . . . . . . . . . . . . . . . . . 330 4.5.4 Unexpectedly Stressed Epenthetic Vowels . . . . . . . . . . . 334 4.5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 4.6 Degree of Incorporation into the Prosodic Word . . . . . . . . . . . 336 4.6.1 Arabic Prosodic Word Formation . . . . . . . . . . . . . . . 341 4.6.2 Impact on Stress . . . . . . . . . . . . . . . . . . . . . . . . 343 4.6.3 Impact of Syllable Structure . . . . . . . . . . . . . . . . . . 346 4.6.4 Qatari /h/ Assimilation . . . . . . . . . . . . . . . . . . . . 348 4.6.5 Conclusion: Incorporation into the Prosodic Word at Level 2 350 4.7 Do We Need a Clitic versus Affix Distinction in Arabic? . . . . . . . 351 4.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354 5 Conclusion 359 5.1 Outcomes and Consequences of this Thesis . . . . . . . . . . . . . . 361 5.1.1 Advantages and Benefits . . . . . . . . . . . . . . . . . . . . 367 5.2 Geographical Distribution of Patterns . . . . . . . . . . . . . . . . . 369 5.3 Broader Implications of these Results . . . . . . . . . . . . . . . . . 376 5.3.1 Derivational versus Constraint-based Models of Phonology . 378 5.4 Directions for Future Research . . . . . . . . . . . . . . . . . . . . . 380 Appendices A Data Appendix 385 A.1 Moroccan Casablanca . . . . . . . . . . . . . . . . . . . . . . . . . . 385 A.2 Algerian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386 A.3 Tunisian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 A.4 Libyan Tripoli . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389 A.5 Cairene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390 A.6 Lebanese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 A.7 Palestinian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392 A.8 Wadi Ram: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 A.9 Iraqi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394 A.10 Muscat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395 A.11 Qatari . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 A.12 Rufaidah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 List of Figures 1.1 Map of the Middle East and North Africa . . . . . . . . . . . . . . 2 1.2 Traditional Classification of Semitic Languages . . . . . . . . . . . . 3 1.3 Afroasiatic Language Family focusing on Semitic . . . . . . . . . . . 4 1.4 Geographical Classification of Modern Arabic Varieties . . . . . . . 8 1.5 Map of Variety Location . . . . . . . . . . . . . . . . . . . . . . . . 14 1.6 Geographical Classification of Modern Arabic Varieties in the Thesis 20 1.7 Phoneme Inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 1.8 Syllable Constituency . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.9 Saporta and Contreras (1962) Syllable Model . . . . . . . . . . . . 24 1.10 Kubozono (1989) Syllable Model . . . . . . . . . . . . . . . . . . . . 24 1.11 Languages by Mora Distribution . . . . . . . . . . . . . . . . . . . . 26 2.1 Arabic Syllable Types with Traditional Arabic Names . . . . . . . . 41 2.2 Modern Standard Arabic Verb Forms . . . . . . . . . . . . . . . . . 43 2.3 Broselow Adjunction to Mora . . . . . . . . . . . . . . . . . . . . . 48 2.4 Kiparsky (2003) Adjunction to Word . . . . . . . . . . . . . . . . . 49 2.5 CV Theory Approach to [ta], [ta:] and [tat] . . . . . . . . . . . . . . 54 2.6 X Theory Approach to [ta], [ta:] and [tat] . . . . . . . . . . . . . . 55 2.7 Moraic Theory Approach to [ta], [ta:] and [tat] with Weight By Position 56 2.8 Syllable Structure for [kita:b-na] with Moraic and Segmental Tiers . 162 2.9 Syllable Structure for [kitab-na] with X-slot and Segmental Tiers . . 163 2.10 Moroccan Casablanca Syllable Structure with X-Slots . . . . . . . . 175 2.11 Moroccan Casablanca Moraic Structure . . . . . . . . . . . . . . . . 175 3.1 Arabic Syllable Types . . . . . . . . . . . . . . . . . . . . . . . . . 184 3.2 McCarthy (1979) Approach to Superheavy Syllables . . . . . . . . . 185 3.3 Halle and Vergnaud (1979) Approach to Superheavy Syllables . . . 185 3.4 Kiparsky (2003) Approach to Superheavy Syllables . . . . . . . . . 185 3.5 Selkirk (1981) Approach to Superheavy Syllables . . . . . . . . . . . 186 3.6 Broselow Approach to Superheavy Syllables . . . . . . . . . . . . . 186 3.7 SPE Stress Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 3.8 Metrical Tree: ‘metricality’ . . . . . . . . . . . . . . . . . . . . . . . 189 xv xvi List of Figures 3.9 Metrical Grid: ‘metricality’ . . . . . . . . . . . . . . . . . . . . . . 189 3.10 Prince (1983) Perfect Grid . . . . . . . . . . . . . . . . . . . . . . . 190 3.11 Halle and Vergnaud (1987) Metrical Grid: ‘metricality’ . . . . . . . 191 3.12 Hayes (1995) Metrical Stress Theory Parameters . . . . . . . . . . . 191 3.13 Kager 2007 OT Stress Constraints . . . . . . . . . . . . . . . . . . . 193 3.14 OT: Non-Finality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 3.15 Palestinian [darasna] with Moraic Trochee . . . . . . . . . . . . . . 201 3.16 Palestinian [darasna] with Iamb . . . . . . . . . . . . . . . . . . . . 201 3.17 Palestinian [katabu] with Correct Moraic Trochee . . . . . . . . . . 201 3.18 Palestinian [katabu] with Incorrect Iamb L−R . . . . . . . . . . . . 201 3.19 Palestinian [duk:a:n] with Correct End Rule Right . . . . . . . . . . 202 3.20 Palestinian [duk:a:n] with Incorrect End Rule Left . . . . . . . . . . 202 3.21 Foot Extrametricality Analysis of Palestinian I . . . . . . . . . . . . 202 3.22 Foot Extrametricality Analysis of Palestinian II . . . . . . . . . . . 203 3.23 Foot Extrametricality Analysis of Palestinian III . . . . . . . . . . . 203 3.24 Syllable Extrametricality Analysis of Palestinian I . . . . . . . . . . 203 3.25 Syllable Extrametricality Analysis of Palestinian II . . . . . . . . . 203 3.26 Mora Extrametricality Correct: [Qal"lamato] . . . . . . . . . . . . . 204 3.27 Mora Extrametricality Incorrect: [*Q]al"lamat . . . . . . . . . . . . 204 3.28 Syllable Extrametricality Correct: [Qal"lamato] . . . . . . . . . . . . 205 3.29 Syllable Extrametricality Correct: ["Q]allamat . . . . . . . . . . . . . 205 3.30 Left-to-Right Parsing Correct: ["SaÃaratun] . . . . . . . . . . . . . . 205 3.31 Right-to-Left Parsing Inorrect: ["SaÃaratun] . . . . . . . . . . . . . 205 3.32 Algerian [sabagu] with Correct Moraic Trochee . . . . . . . . . . . 206 3.33 Algerian [sabagu] with Incorrect Iamb . . . . . . . . . . . . . . . . . 206 3.34 Algerian [masa"bu:ga] with Correct ERR . . . . . . . . . . . . . . . 207 3.35 Algerian [masabu:ga] with Incorrect ERL . . . . . . . . . . . . . . . 207 3.36 Algerian ["tasabagu] with Correct Syllable Extrametricality . . . . . 207 3.37 Algerian [*tasa"bagu] without Extrametricality . . . . . . . . . . . . 207 3.38 Algerian ["tasabagu] with Syllable Extrametricality and Left-to-Right Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 3.39 Algerian [*tasa"bagu] with Syllable Extrametricality and Right-to- Left Parsing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 3.40 Rufaidah ["katabaw] with Correct Moraic Trochee . . . . . . . . . . 208 3.41 Rufaidah ["katabaw] with Incorrect Iamb . . . . . . . . . . . . . . . 208 3.42 Rufaidah [Pas"talam-at-ih] with Correct L−R Parsing . . . . . . . . 208 3.43 Rufaidah [Pas"talam-at-ih] with Incorrect R−L Parsing . . . . . . . 208 3.44 Rufaidah [Pas"talamat] with Correct ERR . . . . . . . . . . . . . . . 209 3.45 Rufaidah [Pas"talamat] with Incorrect ERL . . . . . . . . . . . . . . 209 List of Figures xvii 3.46 Rufaidah ["Paètaram] with Correct Syllable Extrametricality . . . . 209 3.47 Rufaidah ["Paètaram] with Incorrect Consonant Extrametricality . . 209 3.48 Rufaidah["Paètaram] with Incorrect No Extrametricality . . . . . . . 209 3.49 Lebanese [dQarabu] with Correct Moraic Trochee L−R . . . . . . . 211 3.50 Lebanese [dQarabu] with Incorrect Iamb L−R . . . . . . . . . . . . 211 3.51 Lebanese [maktabti] with Correct ERR . . . . . . . . . . . . . . . . 211 3.52 Lebanese [maktabti] with Incorrect ERL . . . . . . . . . . . . . . . 211 3.53 Lebanese ["naz:al] with Syllable Extrametricality . . . . . . . . . . . 212 3.54 Lebanese ["naz:al] with Consonant Extrametricality . . . . . . . . . 212 3.55 Lebanese [maktabe] with Syllable Extrametricality . . . . . . . . . . 212 3.56 Lebanese [maktabe] with Incorrect Consonant Extrametricality . . . 212 3.57 Lebanese [Sa"caratun] with Correct Right-to-Left Parsing . . . . . . 212 3.58 Lebanese [Sa"caratun] with Incorrect Left-to-Right Parsing . . . . . 212 3.59 Lebanese [da"rabana] with Correct Right-to-Left Parsing and Syllable Extrametricality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 3.60 Lebanese [da"rabana] with Incorrect Foot Extrametricality . . . . . 212 3.61 Iraqi ["fahamak] with Correct Moraic Trochees . . . . . . . . . . . . 214 3.62 Iraqi ["fahamak] with Incorrect Iamb . . . . . . . . . . . . . . . . . 214 3.63 Iraqi [mus"taSfa] with Correct ERR . . . . . . . . . . . . . . . . . . 214 3.64 Iraqi [mus"taSfa] with Incorrect ERL . . . . . . . . . . . . . . . . . . 214 3.65 Iraqi ["ka:tiba] with Syllable Extrametricality . . . . . . . . . . . . . 215 3.66 Iraqi [ka:tiba] with Incorrect Consonant Extrametricality . . . . . . 215 3.67 Mak:an [katabu] with Correct Moraic Trochee . . . . . . . . . . . . 215 3.68 Mak:an [katabu] with Incorrect Iamb . . . . . . . . . . . . . . . . . 215 3.69 Mak:an [na:"de:t] with Correct ERR . . . . . . . . . . . . . . . . . . 216 3.70 Mak:an [na:"de:t ] with Incorrect ERL . . . . . . . . . . . . . . . . . 216 3.71 Mak:an ["sa:faru] with Correct Syllable Extrametricality . . . . . . . 216 3.72 Mak:an ["sa:faru] with Incorrect Consonant Extrametricality . . . . 216 3.73 Mak:an ["sa:faru] with Incorrect No Extrametricality . . . . . . . . . 216 3.74 Muscat ["QaSa] with Correct Moraic Trochees . . . . . . . . . . . . . 218 3.75 Muscat ["QaSa] with Incorrect Iambs . . . . . . . . . . . . . . . . . . 218 3.76 Muscat /"qalami/ with Correct Moraic Trochee . . . . . . . . . . . 219 3.77 Muscat /"qalami/ with Incorrect Iamb . . . . . . . . . . . . . . . . 219 3.78 Muscat [daSa"diShum] with End Rule Right and Syllable Extrametri- cality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 3.79 Muscat [daSadiShum] with End Rule Right and Incorrect No Extra- metricality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 3.80 Muscat [daSa"diShum] with Incorrect End Rule Left . . . . . . . . . 220 3.81 Muscat ["kansalu] with Syllable Extrametricality . . . . . . . . . . . 220 xx List of Figures 3.156Libyan Tripoli [y@Ù"k@s:@r] with Correct R−L Parsing . . . . . . . . 250 3.157Libyan Tripoli [y@Ù"k@s:@r] with Incorrect L−R Parsing . . . . . . . 250 3.158Libyan Tripoli [saknin] with Correct ERR . . . . . . . . . . . . . . 250 3.159Libyan Tripoli [saknin] with Incorrect ERL . . . . . . . . . . . . . . 250 3.160Structure of Consonant Clusters . . . . . . . . . . . . . . . . . . . . 254 3.161Structure of Geminates . . . . . . . . . . . . . . . . . . . . . . . . . 254 3.162Structure of Singleton Consonants . . . . . . . . . . . . . . . . . . . 254 3.163Catalectic Mora Structure . . . . . . . . . . . . . . . . . . . . . . . 255 3.164Initial Cluster Onset Variety . . . . . . . . . . . . . . . . . . . . . . 257 3.165Initial Cluster Coda Variety . . . . . . . . . . . . . . . . . . . . . . 257 3.166Final Cluster Onset Variety . . . . . . . . . . . . . . . . . . . . . . 258 3.167Final Cluster Coda Variety . . . . . . . . . . . . . . . . . . . . . . . 258 3.168Catalectic Mora with [Pal:amat] Incorrectly Predicts Stress . . . . . 259 3.169Correct Stress Analaysis of [Pal:amat] Without Catalectic Mora . . 259 3.170Initial Complex Cluster Representation of Subminimal Words . . . 262 3.171Catalexis Analysis of Apparently Subminimal Words . . . . . . . . 262 3.172Catalectic Mora Repair of [faxr-na], ‘our coal’ . . . . . . . . . . . . 265 4.1 Cairene Expected but Unattested Stress on [ramitu] . . . . . . . . . 326 4.2 Palestinian Opaque Stress Derivation . . . . . . . . . . . . . . . . . 335 4.3 Prosodic Hierarchy of Selkirk (1978) . . . . . . . . . . . . . . . . . . 337 4.4 Selkirk (1995): Clitic Typology . . . . . . . . . . . . . . . . . . . . 339 4.5 PWord Clitic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 4.6 Free Clitic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 4.7 Affixal Clitic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 4.8 Internal Clitic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 4.9 Rule Ordering Solves Opacity . . . . . . . . . . . . . . . . . . . . . 342 5.1 Geographical Classification of Modern Arabic Varieties in the Thesis 360 5.2 Syllable Structure for [kita:b-na] with Moraic and Segmental Tiers . 362 5.3 Syllable Structure for [kitab-na] with X-slot and Segmental Tiers . . 363 5.4 Initial Cluster Onset Variety . . . . . . . . . . . . . . . . . . . . . . 364 5.5 Initial Cluster Coda Variety . . . . . . . . . . . . . . . . . . . . . . 364 5.6 Final Cluster Onset Variety . . . . . . . . . . . . . . . . . . . . . . 365 5.7 Final Cluster Coda Variety . . . . . . . . . . . . . . . . . . . . . . . 365 5.8 Map of Variety Location . . . . . . . . . . . . . . . . . . . . . . . . 371 List of Tables 1.1 Data Sources Used in the Thesis . . . . . . . . . . . . . . . . . . . . 13 1.2 Realisation of j̄im across Modern Arabic Dialects . . . . . . . . . . 22 1.3 Cross-linguistic Syllable Inventory . . . . . . . . . . . . . . . . . . . 28 2.1 Broselow (1992) Syllable Behaviours across Varieties . . . . . . . . . 47 2.2 Kiparsky (2003) Dialect Group Criteria . . . . . . . . . . . . . . . . 50 2.3 Kiparsky (2003) Analysis . . . . . . . . . . . . . . . . . . . . . . . . 50 2.4 Watson (2007) Analysis . . . . . . . . . . . . . . . . . . . . . . . . 51 2.5 Farwaneh (2009) Analysis: Level of Final-C Application . . . . . . 51 2.6 Epenthesis Position and Evidence . . . . . . . . . . . . . . . . . . . 58 2.7 Arabic CSS-Morph . . . . . . . . . . . . . . . . . . . . . . . . . . 60 2.8 CV Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 2.9 VC Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 2.10 C Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 2.11 Tolerance of Medial CVXC Syllables . . . . . . . . . . . . . . . . . 159 2.12 Moroccan Casablanca Length and Weight Disparity . . . . . . . . . 174 2.13 Summary of Syllable Tolerance Restrictions . . . . . . . . . . . . . 177 2.14 Tolerance of medial CVXC Syllables, Type of Tolerance Restriction and Epenthesis Type . . . . . . . . . . . . . . . . . . . . . . . . . . 178 2.15 Initial Summary of Repair . . . . . . . . . . . . . . . . . . . . . . 180 2.16 Second Summary of Repair . . . . . . . . . . . . . . . . . . . . . . 180 2.17 Summary of Tolerance . . . . . . . . . . . . . . . . . . . . . . . 181 2.18 Summary of Repair . . . . . . . . . . . . . . . . . . . . . . . . . . 181 3.1 Stress across Arabic Varieties . . . . . . . . . . . . . . . . . . . . . 251 3.2 Covariance between Stress and Tolerance across Arabic Varieties 266 3.3 Overview of Tolerance . . . . . . . . . . . . . . . . . . . . . . . 268 3.4 Overview of Repair . . . . . . . . . . . . . . . . . . . . . . . . . . 268 3.5 Tolerance and Repair of Arabic Varieties . . . . . . . . . . . . . 269 4.1 Non-Nominative Formative Paradigm . . . . . . . . . . . . . . . . . 282 4.2 Arabic Hollow Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . 287 4.3 Closed Syllable Shortening Across Arabic Varieties and Endings . . 288 xxi xxii List of Tables 4.4 Palestinian Dative Behaviour with Epenthesis . . . . . . . . . . . . 291 4.5 Lebanese Pre-Suffix Gemination (Abdul-Karim 1980) . . . . . . . . 301 4.6 Pre-Clitic Lengthening . . . . . . . . . . . . . . . . . . . . . . . . . 302 4.7 Weak Verb Paradigms (Watson 2002, p. 144) . . . . . . . . . . . . . 315 4.8 Weak Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316 4.9 Classical Arabic versus SQanQa:ni: and Cairene Biliteral Verb Paradigm (Watson 2002, p. 146) . . . . . . . . . . . . . . . . . . . . . . . . . . 317 4.10 Biliteral Verbs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 4.11 Classical Arabic versus SQanQa:ni: and Cairene Sound Verb Paradigm320 4.12 Strategies to Create Stressed Syllables Before Affixes . . . . . . . . 329 4.13 Final /h/ in 3rd Person Masculine Singular Object . . . . . . . . . 333 4.14 Qatari Arabic Objects Trigger Syncope in the Stem (Al-Sulaiti 1993, p. 30) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 4.15 Affixation After Stressable Foot Mechanisms . . . . . . . . . . . . . 356 4.16 Affixation After Stressable Foot Mechanisms by Repair Position . . 357 4.17 Affixation After Stressable Foot Mechanisms by Syllable Restriction 358 5.1 Overview of Tolerance . . . . . . . . . . . . . . . . . . . . . . . 362 5.2 Overview of Repair . . . . . . . . . . . . . . . . . . . . . . . . . . 366 5.3 Tolerance and Repair of Arabic Varieties . . . . . . . . . . . . . 366 5.4 Distribution of Phonological Typology Parameters Across Varieties 370 5.5 Affixation After Stressable Foot Mechanisms by Syllable Restriction 375 A.1 Data from Moroccan Casablanca . . . . . . . . . . . . . . . . . . . 386 A.2 Data from Algerian . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 A.3 Data from Tunisian . . . . . . . . . . . . . . . . . . . . . . . . . . . 388 A.4 Data from Libyan Tripoli . . . . . . . . . . . . . . . . . . . . . . . . 389 A.5 Data from Cairene . . . . . . . . . . . . . . . . . . . . . . . . . . . 390 A.6 Data from Lebanese . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 A.7 Data from Palestinian . . . . . . . . . . . . . . . . . . . . . . . . . 392 A.8 Data from Wadi Ram: . . . . . . . . . . . . . . . . . . . . . . . . . 393 A.9 Data from Palestinian . . . . . . . . . . . . . . . . . . . . . . . . . 394 A.10 Data from Muscat . . . . . . . . . . . . . . . . . . . . . . . . . . . 395 A.11 Data from Qatari . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 A.12 Data from Rufaidah . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 A.13 Data from Rwaili . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 A.14 Data from Ha:yil . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 A.15 Data from Mak:an . . . . . . . . . . . . . . . . . . . . . . . . . . . 400 A.16 Data from SQanQa:ni: . . . . . . . . . . . . . . . . . . . . . . . . . . 401 List of Abbreviations C . . . . . . . . Coda, elsewhere Consonant CSS-Morph . Morphologically-triggered vowel shortening and raising CSS-Phon . . Phonologically-triggered vowel shortening without raising ERR . . . . . . End Rule Right H . . . . . . . . Heavy (syllable) L . . . . . . . . Light (syllable L–R . . . . . . Left-to-Right Parsing Directionality MSA . . . . . . Modern Standard Arabic MST . . . . . . Metrical Stress Theory MT . . . . . . . Moraic Trochee N . . . . . . . . Nucleus Non-Fin . . . . Non-Finality O . . . . . . . . Onset OT . . . . . . . Optimality Theory R . . . . . . . . Rime R-L . . . . . . Right-to-Left Parsing Directionality S . . . . . . . . Strong (branch of a foot) SSC . . . . . . Stress Subordination Convention V . . . . . . . . Vowel W . . . . . . . Weak (branch of a foot) X . . . . . . . . X-slot σ . . . . . . . . Syllable µ . . . . . . . . Mora xxv xxvi 1 Introduction Contents 1.1 Arabic Context . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Investigating Sixteen Modern Arabic Varieties . . . . . 12 1.3 Beyond the Phoneme Structure: Why Syllables Matter 22 1.4 Beyond the Word Structure: Prosodic Word Formation 31 1.5 Overview and Structure of the Thesis . . . . . . . . . . 34 Modern Arabic vernacular varieties display drastic differences in their phonology, syntax and lexicon despite their close historical origins. This thesis deals with the phonological typology of sixteen modern varieties, exploring not just the crucial ingredients of the phonological grammar, including syllable and foot structure, but also the lexical and postlexical nature of prosodic structure. 1.1 Arabic Context Arabic is a non-concatenative Central Semitic language with around 335,176,770 native speakers (Eberhard et al. 2020) across North Africa, the Middle East, and expatriate communities around the world from Australia to Venezuela. ‘Arabic’ is an umbrella term that covers Modern Standard Arabic, the formal standard variety, a wide range of vernacular varieties, and Classical Arabic as a liturgical language. 1 4 1.1. Arabic Context Figure 1.3: Afroasiatic Language Family focusing on Semitic Afro-Asiatic Cushitic Egyptian Coptic Chadic Berber Semitic West Semitic Central Semitic Arabic Maltese Old South Arabian Northwest Semitic Canaanite Hebrew Phoenician Moabite Aramaic Ugaritic SamPalian Deir QAllā Modern South Arabian Ethiopian East Semitic Akkadian Eblaite Omotic approach has been disputed (Nebes 1994; Porkhomovsky 1997; Stempel 1999; Lipiński 2001; Haelewyck 2007), it remains the foundation for subsequent work (Rubin 2011, p. 261). The family tree in Figure 1.3 below is a modification of his model (Rubin 2008, p. 62) with the main subgroupings of the Afroasiatic family added (van der Hulst and Hellmuth 2010, p. 617). Of course, the genealogical model of the family tree cannot account for the effects of language contact explained by a wave model, and indeed as Labov (2007, p. 345) argues ‘any general view of language descent must be prepared to integrate the two models’. Furthermore, Ullendorff (1970) , Garbini (1984) and Al Sharkawi (2017) question the legitimacy of attempting to propose a genetic relationship between these languages at all, given quite the degree of contact between them. However, this thesis is not concerned with the fine details of Semitic language family development, so the tree in Figure 1.3 is given to contextualise Arabic amongst related languages1. 1For more information, see Faber (1997), Bennett (1998), Appleyard (2003), Rubin (2008), Weninger et al. (2011), Versteegh (2014) and references within. 1. Introduction 5 1.1.2 Types of Arabic ‘Arabic’ covers a multitude of varieties distributed across time and space. For the varieties distributed over time, the following terms are used:2 Old Arabic refers to pre-diasporic Arabic though the exact chronological bounds vary between scholars. Classical Arabic is the language of the Qur’ān, used from pre-Islamic Arabia3 to the Abbasid Caliphate4. This is the variety discussed by classical grammarians such as Si:bawayhi (d.793) and lexicographers such as Ibn Khali:l (d.791) . It is still used today across the world as a liturgical language. Middle Arabic refers to texts deviating from Classical Arabic Grammar. This is not a chronological stage per se, but a style.5 More important for this thesis are the varieties used today. Colloquial varieties are learned as mother tongues, and used in all non-formal domains (Jarrar et al. 2017). While primarily used orally, they are found in literature, lyrics, films, soap operas, social media, cartoons and adverts. There is no standardised spelling in either the Arabic or Latin script, but conventions particularly on social media are emerging. For vernacular variety use in literature, see Aguadé (2006) for Morocco, and see Rosenbaum (2004) and Woidich (2010) for Egypt. A standard form known as Modern Standard Arabic (MSA) is used across the Arabic speaking world, but is not learnt as a mother tongue — rather it is acquired at school. It is used in literature, official documents, news bulletins, and sermons. 1.1.3 Ancestor of the Modern Varieties Whether the modern colloquial varieties shared the same common ancestor, or a range of closely related ancestors, remains unresolved and controversial. This is not an unusual problem for Historical Linguistics — later varieties tend to continue a sister variety of a better attested earlier variety. For example, Anglo-Saxon is best 2These terms do not correspond to coherent historical stages, or even necessarily linguistic commonalities. See Owens (2006) for a criticism. 3According the traditional narrative, Muhammad was born c.570, began preaching from c.609, and died 632. By his death, most of the Arabian peninsula had been converted to Islam. 4Abbasid Caliphate ruled 750-1258, and then 1261-1417 under the Mamluk Sultanate of Cairo. 5See Larcher (2001). 6 1.1. Arabic Context attested in the West Saxon form, but Middle English is descended from Anglian, Pali continues a sister variety of classical Sanskrit, and Middle Chinese is best attested in the Wu group rather than Mandarin. Different approaches have been taken to this question for Arabic. Where monogenesis is assumed, there are disagreements with regard to this origin — are the modern varieties descendants of the Classical Arabic described by the Arab grammarians, or are they descendants of a colloquial language spoken in Mekka and Medina before the advent of Islam (Vollers 1906; Versteegh 1984; Holes 2004), or are they descendants of a post-Islam military koiné (Fück 1950; Ferguson 1959a)? However, others argue that the modern varieties are descended from many different ancient varieties (Edzard 1998; Cohen 1970). While many features distinguishing modern varieties from Classical Arabic have been identified by Ferguson (1959a), Cohen (1970) and Versteegh (1984), Behnstedt and Woidich (2005) demonstrated that some dialects fail to exhibit many of these features, whereas some of these supposedly modern innovations may have existed in ancient varieties of Arabic. Furthermore, it is clear that the Arabic spoken in the 6th to 10th centuries was not a single variety but rather a multiple distinct varieties. These varieties were not neatly contained, but rather there is evidence of extensive intertribal contact (Al Sharkawi 2017). Arab grammarians noted and named the allophonic variation present in dialectal phenomena, such as QanQanah,6 kaSkaSah,7 taltalah,8 and QajQajah,9 although these colloquial varieties were not their primary focus (Sibawayhi and Hārūn 1982/AH 1402; Rabin 1951; Cadora 1992; Al Sharkawi 2017). Investigating the historical origins of the modern varieties is complicated by the limited material, but remains a fruitful and important avenue of research. Regardless of the exact origin of these varieties, it is clear that they exhibit both close similarity and drastic differences. How to classify these axes of variation has 6QanQanah: the substitution of the pharyngeal fricative Q for a glottal stop, as in Pan:aka > Qan:aka, that you. 7kaSkaSah: substitution of post-alveolar fricative S for velar plosive k in feminine singular suffix, as in biki > biSi, in you (f.s.). 8taltala: use of /i/ as the theme vowel in the imperfective subject marker, rather than /a/. 9QajQajah: substitution of /ig/ for /i:/, as in tamimi: > tamimig from Tamimi. 1. Introduction 9 In terms of social divisions, classifications occur on grounds of Sedentary versus Nomadic, Urban versus Rural, and Sectarian differences. Marçais (1938) first introduced the Sedentary versus Nomadic or Bedouin distinction to account for the patterns of arabization of North Africa. This distinguishes between the varieties related to those of the nomadic Bedouin and those from settlements. Bedouin varieties have distinct features, such as the voiced [g] reflex of the Classical Arabic *q, feminine plural forms, and interdental consonants. Bedouin varieties retain [T] and [D] for the Classical Arabic *T and *D, whereas sedentary varieties have [t] and [d] respectively. The Urban versus Rural distinction is similar, with the glottal stop or voiceless reflex of the Classical Arabic *q noted as a particular urban feature. Sectarian divides are also noted, such as in Baghdad where traditionally a distinction has been made between Muslim, Jewish and Christian varieties. However, these sectarian lines really reflect population movement, where the Baghdad Muslims came from Bedouin communities in southern Iraq but Christians came from towns in Northern Iraq (Blanc 1964; Abu-Haidar 1991; Mansour 1991). However, these classifications are increasingly unreliable, due to dialect levelling, spread of urban speech around urban centres and along trade routes, as well as population movement. 1.1.5 Diglossia Western Arabicists have explored the complex relationship between MSA and colloquial varieties since the early 20th century, and often describe it as diglossia — though exactly what this term means has changed over time. Marçais (1930) claimed ‘la diglossie arabe’ in North Africa involved two com- pletely separate entities: ‘l’arabe écrit’ and ‘l’arabe parlé’. Lecerf (1932) makes similar claims about the Levant, but suggested there were signs of a shift with more mixing between the two following the advent of the printing press. Diglossia in the sense of a hierarchical contrast between two varieties used by the same language community was formalised by Ferguson (1959b, pp. 328-335) who made the following distinction between the standardised or prestigious form, H, and the vernacular form, L. He argues there are specialised functions for each 10 1.1. Arabic Context variety with limited overlap. Ferguson’s claims as to the specialised functions and distributions are as follows. The H variety is considered superior to L by native speakers, and there is usually a large body of esteemed literature written in H. L is acquired as a mother tongue while H is acquired through formal education. As a result, speakers have a better command of L. There is a long tradition of grammatical study of H while limited or descriptive studies of L either do not exist or were written very recently. This diglossic situation continues for centuries, though uncodified intermediate forms can arise with vocabulary borrowings from H but use of L morphology and syntax. There is often much difference between H and L grammatical structures and those of L are often (viewed as) simpler. The majority of vocabulary is shared between H and L with variations in form and semantics, but often there are pairs of lexemes where usage is in solely H or L contexts. Finally, the phonological system can be similar or can diverge between H and L. Where it diverges, L is the basic system, and H diverges as a subsystem. Subsequent approaches have described diglossia as a series of 4 or 5 levels (Meiseles 1980; Badawi 1973)11 or a gradual continuum (Blanc 1960) between colloquial and literary forms with potentially infinite numbers of levels (Bassiouney 2009). Albirini and Powesland (2011) argue that the distinction between MSA and colloquial dialects is determined by function not by context. Eid (1982; 1988) proposed applying the code-switching framework to analyse ‘diglossic’ switching between H and L. This was taken up by Walters (1996a; 1996b), Boussofara-Omar (2003) and Bassiouney (2006) among others. Earlier works imply a stricter division between H and L than this, perhaps reflecting a change over time following increased education and literacy, and therefore wider fluency in MSA. This trend is likely to continue following the advent of social media, with more writing in colloquial Arabic and more mixing of varieties (Mejdell 2018). 11Badawi and Jankowsky (1985, p. 16) distinguishes between fusQèa al-tura:tQ, ‘language of heritage’, that is, Classical Arabic; fusQèa al-QasQr, ‘contemporary language’, that is MSA; Qa:m:iyat al-mutQaq:afi:n ‘colloquial of the educated’; Qa:m:iyat al-mutanaw:ari:n ‘colloquial of the semi-educated’; Qa:m:iyat al-Pum:iy:i:n ‘colloquial of the uneducated’. 1. Introduction 11 1.1.6 Research on Arabic Dialectology from the 19th Cen- tury Onwards Arabic dialectology began at the end of 19th century, with initial focus on grammars and dictionaries such as Spitta (1880). Dialect atlases began to be produced in the early 20th century, such as Bergsträßer (1915) for Palestinian, Cantineau (1940) for Hōrān, Syria and Cantineau (1946) for Palmyra, Syria. Abul-Fadl (1961) pioneered research on the geographical distribution of phonological and morphological features in the Sarqiy:a province in Nile Delta. Subsequent dialect atlases have been produced for Egyptian varieties (Behnstedt and Woidich 1985; Behnstedt 1987; Behnstedt 1988), North Yemeni varieties (Behnstedt 1987; Behnstedt 1992) and Syrian varieties (Behnstedt 1997). Today, many works can be found describing individual varieties, ranging in detail and length. However, coverage of the whole Arabic speaking world is still patchy (Versteegh 2014). Studies of Arabic phonology have primarily focused on changes to phoneme inventories and segmental processes of syncope and epenthesis (Garbell 1958; Birkeland 1952; Jakobson 1957; Mustawafi 2018; Holes 2018). There are a multitude of studies on contemporary stress (see Watson (2011a) for a summary), and some work on the correlation of syllable structure and other phonological processes across groupings of varieties (Broselow 1992; Kiparsky 2003; Watson 2007; Farwaneh 2009). However, the existing work on the phonological typology of modern Arabic varieties is incomplete, unable to account for non-canonical behaviours, and uses a disparate range of strategies to solve apparent exceptions. In this thesis, I contribute a new bipartite typology of Arabic stress and syllable structure, with better empirical coverage of quantity tolerance, superheavy syllables, stress, and the morphological interface with prosodic structures. This research has broader consequences for linguistic theory, in particular with regards to theories of syllable structure restrictions, the reanalysis of catalexis as a tool to repair degenerate syllables as well as degenerate feet, and the clitic versus affix distinction. 14 1.2. Investigating Sixteen M odern A rabic Varieties Figure 1.5: Map of the Locations of the Varieties from the thesis 1. Introduction 15 Moroccan Casablanca is a Maghrebi variety spoken in Casablanca, a coastal city in Western Morocco. As the biggest economic and industrial city in Morocco, Casablanca has a heterogenous population due to movement particularly from the local rural areas into the city, and there has been historical and modern contact with Berber12, French, Spanish and English among other languages. Despite the heterogeneity of the population, a recognisable homogenous variety of ‘Casablanca Moroccan Arabic’ exists, and exhibits Bedouin features due to the historical Bedouin settlements in the area. The data source for this variety is Boudlal’s 2001 doctoral thesis completed at the Université Mohammed V in Rabat, Morocco. Boudlal, a native speaker of Moroccan Casablanca, elicited data from family members and friends born in Casablanca whose parents do not speak Berber and have lived in Casablnca for a long time. In addition to the elicited data, his work includes data from published work (especially Harrell (1962) and Abdelmassih (1973)) that is confirmed by native speakers. Beyond this, Boudlal is not explicit about the nature of data collection or origins of individual items, except for the stress data. For stress placement, he carried out a native speaker judgement test on a corpus of 39 items with 60 subjects as well as measuring the fundamental frequency, intensity and duration of syllables in recordings from 5 subjects. Algerian is a Maghrebi variety spoken in Oran, a city on the North-West coast of Algeria. There has been extensive contact with other languages in Oran, notably Berber, Spanish, Turkish, French and other local varieties. The data source for this variety is Bouhadiba’s 1988 doctoral thesis completed at the University of Reading. Bouhadiba is a native speaker, and as such the data he cites does in part come from his own introspection; however, he states some data is collected from 20 hours of tape recordings of apparently representative speakers of Oran Arabic collected in Oran and surrounding suburbs. Informants are chosen based on phonetic/semantic considerations of their speech. Which data comes from his own introspection and which from elicitation is not specified. 12This is disputed; see Diem (1979, pp.52–55), El Aissati (2006), Taine-Cheikh (2007), Chtatou (2009), Bensoukas and Boudlal (2012) and Aguadé (2018, p. 35) inter alia. 16 1.2. Investigating Sixteen Modern Arabic Varieties Tunisian is a Maghrebi variety spoken in Nabeul, a seaside resort on the North-East coast of Tunisia. As with the other Maghrebi varieties, there has been extensive dialect contact with Spanish, Turkish, Italian and French. The data source for this variety is Maamouri’s 1967 doctoral thesis completed at Cornell. As a native speaker born in Nabeul, part of the data comes from his own introspection. However, some is also specifically elicited from 9 male Tunisian students aged 20-30 studying at Cornell. These informants come from across Tunisia, all speak fluent French and at least some English. Libyan Tripoli is a Maghrebi variety that was spoken in by the Jewish population of Tripoli, on the Mediterranean coast of Libya about 100km from the Tunisian border. In 1948, there were around 20,000 Jews in Tripoli, presumably speaking this variety, but most have left through subsequent migrations. The data source for this variety is Yoda’s 2005 research monograph describing the phonology and morphology of this variety. The data comes from 35 hours of recordings with 9 informants (5 men, 4 women) who previously lived in Tripoli city but have subsequently migrated to Israel (7 in 1948-1951, 2 in 1967). Cairene is an Egyptic variety spoken in Cairo, the capital of Egypt. Cairo is a large city, and has faced extensive language contact including French, English, Turkish and other Arabic varieties. Cairene is a well-studied Arabic variety. The main data source for this variety is Watson’s 2002 research monograph discussing the phonology and morphology of Cairene and SQanQa:ni:. Watson does not specify the source for her data, but the data she includes does not contradict the expected patterns from other work on Cairene. All Cairene data comes from Watson unless otherwise specified. The second source for Cairene data is Broselow’s 1976 doctoral thesis completed at the University of Massachusetts, Amherst. Broselow uses predominately already published sources of data as well as a native informant. Palestinian is a Levantine variety spoken in Palestine, Jordan, and diaspora communities around the world. The main data source for this variety is Abu Salim’s 1982 doctoral thesis completed at the University of Illinois at Urbana-Champaign. Abu Salim is disappointingly vague as to the origin of his data and indeed specificity 1. Introduction 19 (1980, p.24)). The main data source for this variety is Watson’s 2002 research monograph discussing the phonology and morphology of Cairene and SQanQa:ni:. Watson is not explicit as to the sources of her data, but she did conduct fieldwork in SanQa so this is the likely source. 1.2.1 Justification of Choice of Varieties These varieties were chosen as they represent a wide range of types of Arabic varieties, there was sufficient data to explore all the questions raised, and the data available was internally coherent in terms of the wider literature, expected phonological patterns, and my own knowledge of Arabic. The four Maghrebi varieties cover the different countries in North Africa. Cairene is perhaps the most studied Arabic variety, as well as the variety best understood across the region as it is spoken on TV. The three Levantine varieties provide a range of the phenomena found in this region, and include a Bedouin variety. The six Peninsula varieties include varieties from Oman and the Arabian Gulf coast, as well as four varieties including two Bedouin varieties from Saudi Arabia that represent the main dialect groups in this region. SQanQa:ni: is included as a Yemeni variety, that also sometimes is included as its own group distinct from the Peninsular varieties. This distribution across the geographical classifications is illustrated in Figure 1.6. I did not include Arabic spoken in central Africa, such as in the Sudan, as the high degree of contact with neighbouring non-Semitic languages has influenced its phonology as noted in Watson (2007). Nor did I include the Arabic varieties spoken in parts of Asia, Mauritania or Malta, for the same reason. Wadi HadQrami would have been a good addition to the Yemeni group, but the data was contradictory as to the behaviour of medial geminates (Al-Saqqaf 1999). Other studies of Bedouin varieties lacked the comprehensive array of data needed for this particular study, but the analysis developed here would be enriched by inclusion of Bedouin data in future work. This array of data is sufficient for the work completed here. Future investigations on this topic would benefit from an even larger pool to confirm the results, and in 20 1.2. Investigating Sixteen Modern Arabic Varieties Figure 1.6: Geographical Classification of Modern Arabic Varieties in the Thesis Arabic Eastern Yemeni SQanQa:ni: Peninsula Gulf Qatari Omani Muscat Najdi Rwaili Ha:yil Hejazi Rufaidah Mak:an Egyptic Cairene Mesopotamian Iraqi Levantine Northern Lebanese Southern Palestinian Wadi Ram: Western Maghrebi Moroccan Casablanca Tunisian Algerian Libyan Tripoli 1. Introduction 21 particular more Bedouin and rural varieties would give a more thorough overview of the variation found in Arabic. 1.2.2 A Note on Transcription The International Phonetic Alphabet (IPA) is used to transcribe the data. This is not standard among the data sources used, nor indeed on much work written by Arabists rather than linguists, who use the America Library Association – Library of Congress system, or variants thereof, thus I have converted the disparate systems of the sources into one consistent transcription for the whole thesis. Note that both geminate consonants and long vowels are marked with : here following IPA conventions. Transcriptions throughout the thesis are broad rather than narrow, and much phonetic detail is therefore obscured. The consonant inventory of the modern varieties is broadly as in Figure 1.7. Figure 1.7: IPA Chart of the Arabic Consonantal Phoneme Inventory At least three vowel qualities are distinguished for modern Arabic varieties [a,i,u], however some varieties also display mid vowels [e,o], and some transcriptions include a much broader array of vowels. Vowel and consonant length are contrastive in all modern Arabic varieties. However, there are a few points on which even broad transcriptions will show major differences between varieties. Geographical variation in the realisation of the phoneme known in Classical Arabic as j̄im is particularly notable. Classical Arabic realised j̄im as a voiced palatal stop (Gairdner 1925; Fischer and Jastrow 1980; 24 1.3. Beyond the Phoneme Structure: Why Syllables Matter Other constituency proposals include that in Figure 1.10 proposed for Japanese by Kubozono (1989), on the grounds that speech errors split Japanese syllables between the nucleus and rest of the syllable, and that in Figure 1.9 proposed by Saporta and Contreras (1962) and supported by Clements and Keyser (1983) due to co-occurence restrictions. Co-occurrence restrictions between the onset and nucleus occur for example in English, where anterior fricatives /f,v,s,z,T,D/ cannot occur before /u:r/, and stop plus /w/ clusters cannot occur before back vowels including /u:, U, aw/ (Clements and Keyser 1983, p. 20). Similarly, co-occurence restrictions occur between the nucleus and the coda, as short vowel + CCC sequences can be found with a final coronal member, but with a long vowel only V:+CC sequences can be found (Selkirk 1978), and as such they argue there are no grounds to favour one grouping of constituents over another. Figure 1.9: Saporta and Contreras (1962) Syllable Model Key: σ = Syllable, O = Onset, N = Nucleus, C = Coda σ O b N a C t Figure 1.10: Kubozono (1989) Syllable Model Key: σ = Syllable, O = Onset, N = Nucleus, C = Coda σ O b N a C t How the syllable functions in Arabic is not a solved matter. There are ongoing questions about the permitted size of syllables in different varieties and how violations of syllable structure are repaired which I engage with in the course of this thesis. To understand the broader context of this work, one must be familiar with the issues of sonority discussed in Section 1.3.1, syllable weight discussed in Section 1.3.2, and superheavy syllables discussed in Section 1.3.3. 1.3.1 Sonority The syllable can be analysed in terms of organising segments in waves of sonority (Sievers 1881; Jespersen 1904; Fischer-Jorgensen 1952; Blevins 1995). Sonority 1. Introduction 25 can account for restrictions on segment ordering within syllables, with sonority increasing to the nucleus and decreasing to the coda, ending often more sonorously than the onset, as encapsulated by the Sonority Sequencing Principle (Clements 1990; Selkirk 1984; Morelli 2003). There is no single phonetic correlate to sonority, but the hierarchy in (2) is widely accepted (Bell and Hooper 1978; van der Hulst 1984; Clements 1987; Clements 1990; Kenstowicz 1994; Smolensky 1995; Holt 1997). (2) The Sonority Hierarchy Plosives < Fricatives < Nasals < Liquids < Glides < Vowels The vowel is the optimal nucleus, but sonorant13 nuclei exist, as in the US English [jr " n] ‘yearn’. Some scholars identify obstruent nuclei in Nuxalk Salishan14, such as in [sx " s] ‘animal fat’. However, this is controversial, with Newman (1947) arguing for no syllabification, Bagemihi (1991) viewing obstruents are non-syllabic, and Shaw (1993) proposing non-nuclear syllables instead. Not only have patterns of sonority been identified within syllables, but also importantly sonority-based patterns also occur across syllable boundaries. There are cross-linguistic tendencies for sonority patterns between adjacent syllables (Hooper 1976; Murray and Vennemann 1983; Vennemann 1988). This is termed the Sonority Contact Law which states that: ‘A syllable contact A.B is the more preferred, the greater the sonority of the offset A and the less the sonority of the onset B. From Davis and Shin (1999, p. 286) That is, al.ka with falling sonority across the syllable boundary is preferred to ak.la with rising sonority across the syllable boundary. This accounts for a range of synchronic and diachronic processes. For a review, see Seo (2011). This has been adapted by Alhammad (2018) to account for a preference for falling sonority between a syllable boundary and a semisyllable in Arabic. 13Consonants can be divided into sonorants and obstruents. Sonorants are produced with continuous non-turbulent airflow, and include nasals, liquids, and glides; obstruents are produced with obstructions to the airflow, so include plosives and fricatives. 14Spoken in the town Bella Coola in British Columbia, Canada. 26 1.3. Beyond the Phoneme Structure: Why Syllables Matter 1.3.2 Weight Some languages treat syllables differently on account of their internal structure, as if some syllables are heavier than others (Allen 1973). Pre-generative approaches discuss this in terms of syllable structure types. However, the Prague school (Jakobson 1971) proposed the mora as a unit of weight mediating between the CV tier and the syllable (Hayes 1984; Hyman 1985). Language specific mora distribution explains why the same syllable structure has different weights in different languages (Gordon 2006). The onset never carries a mora as it makes no contribution to stress in any known language. Vocalic nuclei always carry moras, but whether long vowels carry two moras is language dependent. The greatest variation is in whether the coda carries moras through Weight by Position (Hayes 1989). The different options for mora distribution allows for three types of languages (Hayes 1995), as shown in Figure 1.11.15 Figure 1.11: Languages by Mora Distribution Type Example Heavy(H) Light(L) Rime Weight English CV: CVC CV Nucleus Weight Selkup CV: CVC CV Coda Weight Dutch CVC CV: CV Subsequent research has suggested that stress assignment in quantity sensi- tive languages can be more complex than a simple heavy versus light syllable distinction. Proposals include weight preference scales for Hindi (Kelkar 1968), Mehri (Watson and Al-Mahri 2018) and SQanQa:ni: (Watson 2002); vowel height for Nanti (Crowhurst and Michael 2005); and a sonority threshold for moraicity in Lithuanian (Zec 1995). Context-dependent weight has also been found where CVC syllables are heavy unless word final (Kager 1989; Alber et al. 1997; Rosenthall 15While ‘H’ and ‘L’ were used in the discussion of diglossia above, their meaning is different in phonology. As such, through the rest of this thesis ‘H’ will refer to heavy syllables, and ‘L’ to light syllables. 1. Introduction 29 in (6) (Kaye et al. 1990). Note that the CV:C is acceptable at the end of the word as the final consonant is extrametrical. (6) Closed Syllable Shortening in Wolof a. rO:f ‘to put in’ (imperfective) b. rOp:i ‘to take out’ (inversive) c. yE:w ‘to tie’ (imperfective) d. yEw:i ‘to untie’ (inversive ) In Koryak17, the most complex syllable permitted is CVC, so if a larger syllable is created this needs to be repaired through epenthesis, as shown in (7) (Spencer 1996, p. 63–64). (7) Epenthesis in Koryak [t@p.N@.lon] t sbj.1sg -pNlo -ask -n -obj.3sg ‘I asked him’ In Bengali18, no coda clusters are permitted, so this is repaired in (8) through degemination (Lahiri 2001, p. 1349). (8) Deletion in Bengali /boS-Ù:i/ boS-Ùi sit -prs.ind ‘he sits’ 17A Paleosiberian language spoken in Kamchatka. 18An Indo-Aryan language spoken in Bengal 30 1.3. Beyond the Phoneme Structure: Why Syllables Matter Similarly in Spanish, medial CVCC syllables are not permitted so when created through affixation the /p/ is deleted in (9-b). (9) Spanish Stray Erasure a. eskulp-ir ‘to sculpt’ b. eskul-tura ‘sculpture’ c. *eskulptura However, some consonants do seem to persist outside of canonical CVC syllable templates, in particular at word edges, and it has been questioned whether they are truly part of the syllable, or instead part of an extrasyllabic unit of some sort. Representations of these tend to be on an ‘ad hoc, language–specific basis’ (Vaux and Wolfe 2009, p. 102). Reflecting the heterogeneity of representations, a wide range of terms are also used to describe them, including semisyllables, reduced syllables, the first third of sesquisyllables, presyllables, half syllables, headless syllables, degenerate syllables, consonantal syllables and minor syllables (Herr 2011, p. 35). Vaux and Wolfe (2009, p. 104–5) classify these different types of representations as follows. Complex Margins involve segments associated with the syllable directly or via an extra onset node to form more complex onset or coda clusters. Degenerate Syllables involve a syllable with an empty nucleus and overt coda or just a coda. Appendices attach a segment to a prosodic node higher than the syllable whereas stray segments do not attach to higher level structures at all. Different languages have different restrictions on the number of semisyllables — or equivalent — in a row. Cho and King (2003, p. 194) argue there is ‘at most one semisyllable per morpheme’ but this is called into question by cases such as Dutch her-f-s-t, ‘autumn’, (Trommelen 1983; van der Hulst 1984). They can occur word initially, word medially, or word finally. Word initially, they are found in Acoma (Miller 1965), Ladakhi (Koshal 1979; Bell and Saka 1983), Greek, Sanskrit and Bella 1. Introduction 31 Coola (Steriade 1982). Word medially, they are found in English, with stop-fricative- stop clusters (Vaux and Wolfe 2009, p. 128). Word finally, they are found in French (Rialland 1994; Dell 1995), English and Wolof (Charette 1985; Kenstowicz 1994). Whether or not semisyllables contain moras seems to vary cross-linguistically. Moraic semisyllables are argued for Estonian and Saami (Bye 1997), Burmese (Green 2005), and Arabic (Kiparsky 2003). Mora-less syllables work for French (Féry and van de Vijver 2003), Mon-Khmer languages (Donegan and Stampe 1983), Lemi Chin (Herr 2011), and Proto-Indo-European (Keydana 2012). A few scholars suggest both are needed, such as Shaw (1992) to account the moraicity of liquid semisyllables but amoraicity of obstruent semisyllables in Kammu. 1.4 Beyond the Word Structure: Prosodic Word Formation The prosodic word mediates between phonology and morphology, and phonology and syntax, and roughly corresponds to the grammatical word. Evidence for its existence can be found in phenomena that use the prosodic word as their domain, including glide formation, syllabification, and stress. Hannahs (1995, p. 1131) argues that prosodic words are the domain for glide formation in French. Underlying high vowels /i, y/ can become glides [j, 4], but only in stem+suffix strings such as (10-b), not at the end of a prefix as in (10-c) or start of a compound as in (10-d). (10) French Glide Formation Hannahs (1995, p. 1131) a. colonie [kOlOni] ‘colony’ b. colonial [kOlOnjal] ‘colonial’ 34 1.5. Overview and Structure of the Thesis 1.5 Overview and Structure of the Thesis Modern Arabic varieties differ significantly in their phonology, grammar and lexicon. In the phonology alone, the variation includes epenthesis position, syncope application, vowel lengthening processes, vowel shortening processes, permissibility of medial and final consonant clusters, permissibility of larger than CVC medial syllables, foot type, parsing directionality, extrametricality, which elements are incorporated into the prosodic word, and exceptions to stress. How to reconcile this wealth of variation into a typology with predictive power has been attempted as will be discussed in Chapter 2, Section 2.2. These existing attempts, however, are incomplete, focusing on smaller sets of correlations, and face empirical limitations. This thesis explores not just the crucial ingredients of the phonological grammar, including syllable and foot structure, but also the lexical and postlexical nature of prosodic structure across sixteen modern Arabic varieties, ranging from Morocco to Iraq to Yemen. I explore these issues through the following more specific research questions: • How can we account for the distribution of syllable types across Arabic? • What is the structure of final CVXC syllables? • How are stress parameters distributed across Arabic varieties and how can we account for exceptions to them? • How are linearly attached formatives incorporated into the prosodic word? These questions allow me to develop a more comprehensive phonological typology than my predecessors that not only better accounts for the variation found across Arabic, but also has consequences for phonological theory more broadly. In Chapter 2, I review previous work on Arabic syllable structure typology, which suggests that varieties can be classified in terms of where they epenthesise into a morphologically derived tri-consonantal cluster. I test this approach against the sixteen varieties considered here. I find that there are more medial CV:C and 1. Introduction 35 CVC: syllables than predicted by Kiparsky (2003) and Watson (2007), and propose a new analysis. I argue that there are two axes for phonological variation across Arabic: Tolerance and Repair, that is, what type of syllables they tolerate and how they repair violation to syllable structure. These axes do not covary. I find that Arabic varieties differ in how they tolerate quantity. Some dialects permit medial syllables with long vowels or geminate consonants but not consonant clusters. This can be accounted for with restrictions at different levels of syllable structure. Varieties that permit long segments have restrictions on the segmental level, whereas varieties that do not have restrictions on the moraic level, and varieties that allow long consonants or clusters but not long vowels have restrictions on the X-slot tier. In addition to the canonical two or three segments, moras or X-slots, individual varieties can also permit an extra element in their medial syllable rime according to variety-specific restrictions, thus allowing for some restricted consonantal clusters. These restrictions can include coronality, fast speech, or sonority. As such, this typology provides mechanisms to account for the data in each and every dialect, rather than merely providing a canonical set of features that not all dialects fit. In Chapter 3, I delve into the issue of the structure of final CVXC syllables using evidence from stress. I revisit the stress analysis of the sixteen varieties in question. All final CVXC syllables are unaffected by extrametricality, which means that the final C must be in a separate syllable that contains a catalectic mora to ensure that the C itself is non-peripheral. I demonstrate that not only does catalexis account for the behaviour of extrametricality, it also accounts for the correlation between medial epenthesis position and whether domain edge clusters are permitted. In this way, I construe catalexis not merely as a device to resolve degenerate feet, but also degenerate syllables. Previous work on Arabic syllable typology found no link with stress position. However, considering the new typology of syllable behaviours I propose in Chapter 2, I revisit this. I find that while the syllable that stress falls on does not covary with the syllable parameters (as found by Kiparsky (2003)), there is a relationship between the extrametrical unit parameter and whether varieties permit long segments (the new axis of variation 36 1.5. Overview and Structure of the Thesis I motivate in Chapter 2). Specifically, varieties with syllable restrictions on the segmental level have syllable extrametricality, varieties with syllable restrictions on the moraic level have consonant extrametricality, and varieties with X-slot syllable restrictions do not exhibit any extrametricality. In Chapter 4, I explore cliticization and affixation in the sixteen varieties, considering their impact on syllable and foot structure. I review previous literature on clitics, including the assumptions in the literature that particular endings are clitics (specifically, attached direct and indirect object pronouns, dative markers, and negation markers). These assumptions include low selectivity; phonological differences; optionality; and stressability. I dispute the claim that these are clitics and show that they are in fact suffixes. These endings do not have low selectivity in the sense used elsewhere in the literature. The apparent differences in phonological behaviours compared to subject markers come from paradigm behaviours in the subject forms combined with a requirement for other endings to attach after a heavy monosyllabic foot. However, I show this preference is also found in the subject markers of biliteral21 and weak22 verbs. This preference for attaching to a stressed foot also accounts for well–recognised but ill–understood exceptions to stress rules across Arabic varieties. In many varieties, the third person feminine singular past tense verb plus a vowel initial bound pronominal object receives stress and is never syncopated. Rather than needing exceptional reversals in parsing directionality or other solutions, this can be accounted for as the same requirement to attach after a stressed foot as other affixes. It also accounts for why final long vowels can escape syllable extrametricality: they are heteromorphemic, contain an affix, and so require stress. I show these endings attach at Level 2 in the lexical phonology. This has to still be part of phonological word formation as words with only subject markers also undergo the phonological processes on Level 2. Therefore, it is not that these processes are occurring within a clitic group level instead. Finally I 21The canonical Arabic root contains three consonantal segments. However some verb roots only have two, and as such are named biliteral. 22The canonical Arabic root contains three consonantal segments. Weak verbs are those where the final root segment is a glide. 2 The Role of Quantity in Syllable Structure Restrictions Contents 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.2 Previous Work on Arabic Syllable Structure Typology 41 2.3 Previous Work on Quantity . . . . . . . . . . . . . . . . 53 2.4 Analysis of Sixteen Varieties . . . . . . . . . . . . . . . . 57 2.5 Tolerance of Medial Syllables . . . . . . . . . . . . . . . 129 2.6 New Proposal . . . . . . . . . . . . . . . . . . . . . . . . . 161 2.7 Conclusion and Next Steps . . . . . . . . . . . . . . . . 181 2.1 Introduction Arabic varieties have previously been divided into groups based on the position of an epenthetic vowel in a morphologically derived triconsonantal cluster. In (13-a), Cairene epenthesises afterft r the central consonant, and in (13-b) Iraqi epenthesises beforef r the central consonant, whereas in (13-c) Moroccan does not epenthesise at all. Note that here and throughout epenthetic vowels are underlined. (13) Epenthesis Position in a Triconsonantal Cluster (Kiparsky 2003, p. 3) a. Cairene CCvC 39 40 2.1. Introduction Pul say.pfv -tiii -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ b. Iraqi CvCC gil say.pst -iiit -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ c. Moroccan CCC q@l say.pst -t -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ How to account for these syllabification patterns, and the extent to which the epenthesis position is correlated with other syllable behaviours, has been a key focus in studies of Arabic phonology to date. In this chapter, I argue that previous typological accounts of Arabic syllable behaviour are incomplete and that a bifurcated typology, with one axis for medial syllable tolerance and another axis for syllable repair and domain edge behaviour, is required to account for the patterns discussed here. I demonstrate that varieties can be grouped by the level of structure they impose syllable restrictions on — whether segmental, moraic, or X-slot — which accounts for the variation in tolerance of medial CVXC syllables, and that this axis does not covary with the relationship between epenthesis position and whether word edge clusters are permitted. By CVXC I mean syllables where the X can be a vowel or consonant, thus subsuming CV:C, CVC: and CVCC. In Section 2.2 I discuss previous work on Arabic syllable typology. In Section 2.3 I discuss previous work on quantity which has not been adequately considered in the Arabic phonology literature, before evaluating the extent to which the sixteen varieties I explore here fit the existing typological expectations in Section 2.4. I demonstrate in Section 2.5 that more medial CVXC syllables surface than expected, and that CV:C and CVC: syllables pattern together. Informed by previous work on quantity, I lay out a new proposal for Arabic syllable typology in Section 2.6. 2. The Role of Quantity in Syllable Structure Restrictions 41 2.2 Previous Work on Arabic Syllable Structure Typology A range of syllable structure types are found in Arabic. These are summarised in Table 2.1, which includes the traditional Arab grammarian names for these syllable types. Figure 2.1: Arabic Syllable Types with Traditional Arabic Names Key: C = consonant, V= vowel Light al kasi:r al maftu:è CV kataba ‘he wrote’ Heavy al tawi:l al maftu:è CV: sa:lim ‘intact’ al muGlaq CVC man ‘who’ Superheavy al muqfal CV:C qa:l ‘he said’ al madi:d al muqfal bisamilain CVCC nasr ‘prose’ CVC: mad: ‘he extended’ Azzabi (2001, p. 112–113) The distribution of CV and CVC syllables is unrestricted.1 Usually, researchers claim that CVXC is only found word or utterance finally — however, as this chapter will demonstrate, there is a much higher prevalence of CVCC, CVC: and CV:C syllables word medially than has previously been recognised. Templatic morphological gemination is a very common process across Arabic, and there are no restrictions as to which segments can undergo gemination. Gemination 1Most works agree there are no onset-less syllables in Arabic. However, Azzabi (2001) claims VC syllables do exist. Such apparently onsetless syllables are often described as beginning with a glottal stop such as those in (i). (i) a. Pana ‘I’ b. Palbanja ‘Albania’ Not all scholars agree this should be considered an onset consonant (Azzajaj 1979; Ibn Jinnii 1970). The Arab grammarians distinguished between two types of glottal stops: hamzat al-qatQ that is always pronounced and present in the underlying morphophonological structure; and hamzat al-wasl that is only pronounced before vowel-initial morphemes without a preceding consonant, with inclusion controlled by the phonological environment. Contrary to Azzabi’s (2001) claim not to produce any hamzat al-wasl glottal stops, Kaye (1991) argues based on anecdotal evidence and orthography that the contextually-produced hamzat al-wasl is shifting to become a hamzat al-qatQ — that is, prothetic glottal stops are becoming part of the underlying representation. 44 2.2. Previous Work on Arabic Syllable Structure Typology Note that vowel and consonant length is important for maintaining morphological distinctions, as shown in (14) and (15). (14) Vowel Length in Morphological Distinctions a. sQadQara stem.pfv -∅ -sbj.3sg.m ‘he stemmed, originated, appeared’ Form I b. sQa:dQara seize.pfv -∅ -sbj.3sg.m ‘he seized, impounded, confiscated’ Form III (15) Consonant Length in Morphological Distinctions in Algerian (Bouhadiba 1988, p. 323) a. kad:b accuse.of.lying.pfv -u -sbj.3pl ‘they accused someone of lying’ b. kadb lie.pfv -u -sbj.3pl ‘they lied’ In addition to the the verb template system, geminate is used (less productively) in the formation of nouns of place and profession as in (16-a–b) as well as the plurals of lexicalised active participles in (16-c–d) (Rosenthall 2018) and in intense/habitual adjective formation in (16-e–f) (El Zarka 2018). (16) Sources of Gemination in Arabic (Rosenthall 2018; El Zarka 2018) a. kal:a:f ‘stablehand’ b. xab:a:z ‘baker’ c. ba:hil (sg), buh:al (pl.) ‘free’ d. Pa:biq (sg.), Pab:a:q ‘fugitive’ 2. The Role of Quantity in Syllable Structure Restrictions 45 e. sik:i:r ‘drunken, drunkard’ f. kaD:a:b ‘liar’ Geminates can also occur through assimilation of the definite article /l/ to a following apical consonant (traditionally called ‘sun letters’, compared to non-assimilatory ‘moon letters’) as shown in (17-a-b). In additional to the apical consonants, in some modern colloquial varieties there is also assimilation to [k,g,b,f,m,n,j,Ã] as in (17-c), and in West Yemenite varieties such as Jiblah the article assimilates to all consonants (Zaborski 2018) as in (17-d). Some gemination can also occur through sandhi phenomena, haplology and both regressive and progressive assimilation (Danecki 2018). (17) Definite Article Assimilation/Gemination (Zaborski 2018) a. /al-dars/ → [ad-dars] ‘the lesson’ b. /al-rajul/ → [ar-rajul] ‘the man’ c. /il-kitab/ → [ik-kitab] ‘the book’ d. /al-be:t/ → [ab-be:t] ‘the house’ How syllabification works in Arabic, and indeed how to account for differing patterns of epenthesis, has been an ongoing controversy. Selkirk (1981) argued that syllables are built around vowels to form CVC syllables, with unassigned consonants triggering vowel epenthesis. The position of the epenthetic vowel is determined by whether the unsyllabified consonant is assigned directly to a degenerate syllable, or to a nuclear mora of a degenerate syllable — and as such, forms the onset or rime respectively of the syllable. This can be seen in (18), where Cairene and 46 2.2. Previous Work on Arabic Syllable Structure Typology Iraqi adapt the loanword name ‘Fred’ according to whether the initial consonant /f/ forms an onset or coda. (18) Onset vs Coda Dialects a. Cairene: An Onset Dialect fiii.red ‘Fred’ b. Iraqi: A Coda Dialect Piiif.red ‘Fred’ Itô (1986), Itô (1989) and Farwaneh (1995) argued instead that the distinction is due to the direction of syllabification. Itô (1989) assumes the onset and nucleus share a mora, as in the model of syllable constituency proposed by Kubozono (1989). Ito assigns moras before building syllables from either edge, adding vowels where needed to form legal syllables. This is shown in Example (19), where in the second line of (19-a) and (19-b) the syllabification without epenthetic vowels is shown, before the epenthetic vowels are inserted in the third line. (19) a. Cairene Left to Right /ka.tab.t.lu/ katab write.pfv -tiii -sbj.1sg -l -dat -u -obj.3sg.m ‘I wrote to him’ b. Iraqi Right to Left /ki.ta.bt.lu/ kitab write.pfv -iiit -sbj.1sg -l -dat -u -obj.3sg.m ‘I wrote to him’ An alignment constraints account of this analysis was proposed by Mester and Padgett (1994). However, Broselow (1992, p. 25) showed that this cannot account for cases in right-to-left syllabification where an unsyllabified consonant has no consonant 2. The Role of Quantity in Syllable Structure Restrictions 49 Onset varieties are named ‘CV’ (see (20-a)), rime varieties ‘VC’ (see (20-b)), and those that do not epenthesise are ‘C’ (see (20-c)): (20) Three Epenthesis Types (Kiparsky 2003, p. 3) a. Cairene: CV Pul say.pfv -tiii -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ b. Iraqi: VC gil say.pfv -iiit -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ c. Moroccan: C q@l say.pfv -t -sbj.1sg -l -dat -u -obj.3sg.m ‘I said to him’ The features he correlates with these groups are shown in Table 2.2.3 Kiparsky does consider whether stress might pattern with the groups as well, but finds that there is no correlation between epenthetic position and stressed syllable. He argues that mora licensing can account for this pattern, where unsyllabified consonants (those outside the CV, CV:, or CVC syllable template) are dominated by a mora that is not affiliated with a syllable but instead attached to the word node directly, as shown in Figure 2.4 below. Figure 2.4: Kiparsky (2003) Adjunction to Word ω σ C µ V µ C µ C 3I have not included pausal glottalisation or desonorisation here, viewing them as areal features as argued by Watson (2007). 50 2.2. Previous Work on Arabic Syllable Structure Typology Table 2.2: Kiparsky (2003) Dialect Group Criteria CV VC C -CC# clusters permitted not permitted /sonority restrictions permitted #CC- clusters no high vowel dele- tion in initial posi- tion initial vowel pros- thesis permitted Initial geminate clusters vowel prosthesis permitted permitted Stem metathesis no yes drop vowel Postgeminate syncope no yes yes Closed Syllable Shortening yes no no Opaque stress/epenthesis interaction no yes no The difference between these dialect groups is in whether they permit semi- syllables lexically or postlexically. He argues that CV dialects do not permit semisyllables; VC dialects only permit semisyllables at the lexical level; and C dialects permit semisyllables at both the lexical and postlexical levels. This is summarised in Table 2.3 below. Table 2.3: Kiparsky (2003) Analysis Dialect Group Lexical Postlexical CV No No VC Yes No C Yes Yes Watson (2007) noted that a set of varieties from Saudi Arabia and Yemen that appear to be CV varieties based on epenthesis after the second element in a triconsonantal cluster also display features associated with VC varieties. In 2. The Role of Quantity in Syllable Structure Restrictions 51 particular, they permit medial CV:C syllables, permit final CVCCC and CV:CC, and delete high vowels after geminates. In fast speech, they share some features of C varieties, such as permitting initial consonant clusters and medial -CCC- clusters. Watson argues that these varieties, which include Mak:an and SQanQa:ni:, form a fourth dialect group that she labels ‘Cv’. This analysis is summarised below in Table 2.4. Table 2.4: Watson (2007) Analysis Dialect Group Lexical Postlexical Mora Sharing CV No No Word-final CV:C syllables VC Yes No CV:C syllables Cv No No CV:C or CVCC syllable C Yes Yes none Farwaneh (2009) links the VC/CV/C analysis with the earlier Onset/Coda dis- tinction, arguing that VC and C varieties are Coda dialects that differ postlexically, with CV as Onset dialects. She argues that these distinctions can be accounted for within Optimality Theory as the constraint Final-C applying at different levels of structure. I summarise this in Table 2.5, where CODA (extreme) refers to the North African C dialects. Table 2.5: Farwaneh (2009) Analysis: Level of Final-C Application Dialect Type Stems Words Syllables Moras ONSET Yes Yes CODA Yes Yes Yes CODA(extreme) Yes Yes Yes Yes There are several unresolved issues with these proposals. Mostly crucially, as I will demonstrate in the subsequent parts of this chapter, more medial CVXC syllables surface than predicted. Whilst there is some recognition in the literature of medial CV:C syllables surfacing, more CVCC syllables occur than expected, and different types of CVCC syllables behave differently. 54 2.3. Previous Work on Quantity ‘All other things being equal, a rule which splits up a geminate cluster is less highly valued than a rule which must be constrained from doing so’ From Kenstowicz and Pyle (1973, p. 27) They recognised a broad typological split, with segmental rules (such as vowel harmony and other vowel quality alternations) treating long vowels as a single segment, but prosodic rules (such as accentual alternation, metathesis, epenthesis and syncope) treating them as a sequence of segments. However, the best way to represent these different phenomena and the relationship between the different levels of representation needed was still elusive. The development of autosegmental representations for tone (Leben 1980) and non-concatenative morphology (McCarthy 1981) formalised the relationship between these different representations. However, disagreement persists as to the nature of these representations, in whether a skeletal tier where two timing units dominate a single segment, or moraic representations where a mora dominates the segment best accounts for the data. In both approaches, there is a separation between a melodic or segmental tier containing all feature infor- mation and a more abstract tier that encodes length through timing or weight units. Skeletal Models began with a CV tier, where timing units also encode syllable structure positions (McCarthy 1979; Leben 1980; Clements and Keyser 1983; Hayes 1986), as shown in Figure 2.5. Figure 2.5: CV Theory Approach to [ta], [ta:] and [tat] C t V a C t V a V C t V a C t Given that the consonantal or vocalic status of segments is usually predictable from syllable structure, Levin (1985) developed a model of X-Slots, as shown in Fig- ure 2.6. 2. The Role of Quantity in Syllable Structure Restrictions 55 Figure 2.6: X Theory Approach to [ta], [ta:] and [tat] Key: σ = Syllable, O = Onset, R = Rime, N = Nucleus, C = Coda, X = X-slot σ O X t R N X a σ O X t R N X a C X σ O X t R N X a C X t This model was better than the CV tier representation at accounting for compensatory lengthening, as the loss of a C in the coda led to lengthening of the preceding vowel, but could not account for the lack of lengthening in the case of onset loss. In (21-a), we see that when the Ingvaeonic parent language of Friesian and English lost nasals before fricatives, the vowel lengthened to compensate for the loss of the segment by associating to the mora of the nasal. However, in (21-b), when Middle English lost velar stops before nasals, the change to the onset did not trigger Compensatory Lengthening as there is no mora to associate to. (21) Compensatory Lengthening a. munT → mu:T ‘Mouth’ b. knOt → nOt ‘knot’ This asymmetry is better accounted for under a moraic model, where segmental root nodes are associated to syllable nodes via moras, as shown in 2.7. 56 2.3. Previous Work on Quantity Figure 2.7: Moraic Theory Approach to [ta], [ta:] and [tat] with Weight By Position Key: σ = Syllable, µ = Mora σ t µ a σ t µ a µ σ t µ a µ t Here, long vowels are bimoraic, short vowels monomoraic, geminates inherently moraic, and singletons amoraic though the latter can be moraic in closed syllables via weight by position (Hayes 1989, p. 258). With moras, the lack of onset compensatory lengthening follows from the fact that the onset does not carry a mora, so there is no floating mora to associate to another segment. Thus, this approach predicts that prosodic processes affect moras, not segments, so segments that are amoraic will not be affected. However, whilst a moraic model does work better for some phenomena, it is not sufficient for all languages. It would predict that initial geminates should be weightless as they are in the syllable onset — but in some languages they do contribute to weight, including Trukese where [tto] ‘clam (sp.)’ is an acceptable minimal word, but *[t@] ‘islet’ is not so surfaces as [t@@] (Davis 1999, p.95). Rubach (1993) found that compensatory rules in Slovak need skeletal representation rather than a weight representation. Furthermore, in some languages there is no isomorphy between rhyme structure and weight (Lahiri and Koreman 1988) — for example, in Seneca V: is light whilst VC is heavy, and in Malayalam medial geminates do not affect syllable weight while long vowels do (Mohanan and Mohanan 1984). Thus, for some researchers, a mixed model is necessary. Kraehenmann (2003) opts for a model where geminates are a single root node linked to two timing slots on an X-slot based skeletal tier, and weight is independently represented through projection of rhyme structure rather than as a function of length. This is necessary where length 2. The Role of Quantity in Syllable Structure Restrictions 59 Where CV varieties are more tolerant of medial CV:C than expected, they can be classified as Cv under Watson (2007). The canonical expectation of a VC variety is: 1. to not permit final CC clusters or to do so only with sonority restrictions; 2. to permit initial CC clusters or prothesise a vowel before an initial CC cluster; 3. to permit initial geminate clusters; 4. to undergo stem metathesis from CCVC to CVCC; 5. to undergo postgeminate high vowel deletion; 6. to permit internal CV:C syllables; 7. to exhibit opaque interactions between stress and epenthesis. The canonical expectation of a C variety is: 1. to permit final CC clusters; 2. to permit initial CC clusters; 3. to permit initial geminate clusters; 4. to syncopate the stem vowel from CCVC to CCC; 5. to undergo postgeminate high vowel deletion; 6. to permit medial CV:C syllables; 7. to not undergo opaque interactions between stress and epenthesis. Note with the licensing of medial CV:C syllables there is a complexity sometimes overlooked in the literature. Arabic contains two separate processes both often referred to a Closed Syllable Shortening (CSS). One type, referred to hereafter as CSS-Morph, is a morphophonological alternation where in one type of Arabic verbs the central long vowel is shortened and often raised in the non-3rd person forms, as shown in Table 2.7 below. 60 2.4. Analysis of Sixteen Varieties Table 2.7: Arabic CSS-Morph Verb Translation Verb Translation Sa:f ‘he saw’ Suf-t ‘I saw’ qa:l ‘he said’ qul-ti ‘you (sg) said’ na:m-at ‘she slept’ nim-tum ‘you (pl) slept’ sa:r-u ‘they walked’ sir-na ‘we walked’ CSS-Morph occurs in all Arabic varieties. The second type, referred to hereafter as CSS-Phon, is the vowel shortening but not raising that occurs when a CV:C syllable is too large for syllable structure restrictions. This is commonly seen in CV varieties. The exact behaviours and distributions of CSS-Morph and CSS-Phon will be discussed in Section 4.3.3 on page 286. 2.4.1 Cairene From the epenthesis position shown in (22), we would expect Cairene to behave as a CV variety. (22) CV Epenthesis (Kiparsky 2003, p. 3) Pul say.pfv -tiii -sbj.1sg -l -dat -ak -obj.2sg.m ‘I told you (m.)’ Final Clusters Cairene permits phrase final CC clusters regardless of the sonority profile of the cluster. In the following (23-a) has a sonority plateau, (23-b) has a sonority fall, and (23-c) has a sonority rise, but none undergo epenthesis. (23) Cairene Final CC Clusters Permitted (Watson 2002, p.70, 71, 190) a. katab write.pfv -t -sbj.1sg ‘I wrote’ b. bint ‘girl’ 2. The Role of Quantity in Syllable Structure Restrictions 61 c. masQr ‘Egypt’ This fits what we would expect from a CV variety. Initial Clusters Cairene does not permit phrase initial CC clusters. These are never created by syncope, and loanwords undergo epenthesis to avoid initial clusters, as shown in (24). Similarly there are no phrase initial geminate clusters. (24) Cairene Epenthesis in Initial CC Clusters (Watson 2002, p.135) a. biii.las.tik ‘plastic’ b. ku.lub: ‘globe’ This fits what we would expect from a CV variety. Epenthesis in Triconsonantal Clusters All medial triconsonantal clusters are broken up by epenthesis to the right of the central consonant, whether these clusters are word internal as in (25-a)–(25-b), or across words as in (25-c)–(25-d). Note that epenthesis occurs across words as the domain of syllabification is the phrase in Cairene. (25) Cairene Triconsonantal Cluster Epenthesis (Watson 2002, p.64) a. Pul say.pfv -tiii -sbj.1sg -l -dat -ak -obj.2sg.m ‘I told you’ b. kul:u all -hum -obj.pl.m ‘all of them’ c. kuntiii be.sbj.1sg hina here ‘I were here’ 64 2.4. Analysis of Sixteen Varieties (30) CV Epenthesis (Watson 2002, p.65, 106) gul say.pfv -ta -sbj.1sg -l -dat -ih -obj.3sg.m ‘I said to him’ Final Clusters SQanQa:ni: does permit phrase final CC clusters regardless of sonority. In (31-a) there is a sonority plateau and in (31-b)–(31-c) there is a sonority fall. (31) Final CC Clusters (Watson 2002, p.73) a. ktab write.pfv -t -sbj.1sg ‘I wrote’ b. élis stay.pfv -t -sbj.1sg ‘I stayed’ c. gambar sit.pfv -t -sbj.1sg ‘I sat’ This fits what we would expect from a CV variety. Initial Clusters SQanQa:ni: permits phrase initial CC clusters regardless of sonority. In (32-a) there is a sonority plateau; in (32-b) there is a sonority fall; and in (32-c) there is a sonority rise. (32) Initial CC Clusters (Watson 2002, p.73) a. fhim understand.pfv -ti: -sbj.3sg.f ‘you (f.s.) understood’ b. hrib flee.pfv -at -sbj.3sg.f ‘she fled’ c. ètQab ‘wood’ 2. The Role of Quantity in Syllable Structure Restrictions 65 This does not fit what we would expect from a CV variety, but is more consistent with a Cv variety (as in Watson’s (2007) analysis). Epenthesis in Triconsonantal Clusters In normal speech all medial triconsonantal clusters are broken up by epenthesis to the right of the central consonant, as shown in (33). (33) Triconsonantal Cluster Epenthesis (Watson 2002, p.64) a. baQdama: ‘after’ b. gablama: ‘before’ c. gul say.pfv -ta -sbj.1sg -l -dat -ih -obj.3sg.m ‘I told him’ We know that these are epenthetic vowels because they only appear in cases of triconsonantal clusters. Elsewhere, we find for example (34-a) without a following vowel, and (34-b) without such a vowel before the dative as well. (34) Evidence of Epenthetic Vowels (Watson 2002, p. 67) a. gul say. -t -sbj.1sg ‘I told’ b. yi sbj.3 -bi:Q -sell. -l -dat -ak -obj.2sg.m ‘he sells to you (m.s.)’ However, in fast speech CCC can occur, as shown in (35), where the optionally syncopated vowel is marked with brackets. (35) Fast Speech Triconsonantal Cluster Variation (Watson 2002, p.115) a. ji sbj.3 -kt(a)sib -earn.impf ‘he earns’ 66 2.4. Analysis of Sixteen Varieties b. ji sbj.3 -ft(a)hin -rest.impf ‘he rests, feels better’ c. ift(a)han rest.pfv -∅ -sbj.3sg.m ‘he rested, felt better’ This does not fit what we would expect from a CV variety, but does fit a Cv variety. Stem Metathesis SQanQa:ni: does not undergo metathesis in verb stems. This fits what we would expect from a CV or Cv variety. Postgeminate High Vowel Deletion SQanQa:ni: undergoes postgeminate high vowel deletion and subsequent degemina- tion, as shown in (36) where the first line of each example gives the underlying form of the verb. Note in (36-b) and (36-c), this deletion is postgeminate vowel deletion, rather than specifically high vowel deletion. (36) Postgeminate High Vowel Deletion and Degemination (Watson 2002, p.73) a. /jilab:isu:/ ji sbj.3 -labs -dress.impf -u: -sbj.pl.m ‘they (m.) dress (someone)’ b. /rag:aQu:/ ragQ-u: sew.pfv -sbj.3pl.m ‘they (m.) sewed’ c. /jiéam:aQayn/ ji sbj.3 -éamQ -collect.impf -ayn -sbj.pl.f ‘they (f.) collect’ This is not what we would expect from a CV variety, but does fit a Cv variety. 2. The Role of Quantity in Syllable Structure Restrictions 69 Epenthesis in Triconsonantal Clusters Mak:an epenthesises in medial triconsonantal clusters to the right, as shown in (42). (42) Epenthesis in Triconsonantal Clusters (Kabrah 2004, p. 69) a. kalba dog -kum -poss.2pl ‘your dog’ b. ruè go.pfv -t -sbj.1sg -al: -dat -u -obj.3sg.m ‘I went to him’ We know that these are epenthetic vowels because they only appear in cases of triconsonantal clusters. Elsewhere, we find for example (43-a) without a following vowel, and (43-b) without such a vowel before the dative as well. (43) Evidence of Epenthetic Vowels (Kabrah 2004, p. 23, 113) a. ruè go.pfv -t -sbj.1sg ‘ I went b. katab write. -∅ -sbj.3sg.m -l -dat -u -obj.3sg.m ‘he wrote to him’ This is what we would expect from a CV variety. Stem Metathesis Mak:an doesn’t undergo stem metathesis, as shown in (44) where the CCVC stem sur- faces. (44) Stem metathesis (Kabrah 2004, p. 41) ni sbj.1 -ktub -write.impf -sbj.pl ‘we write’ This is what we would expect from a CV variety. 70 2.4. Analysis of Sixteen Varieties Postgeminate High Vowel Deletion Mak:an doesn’t undergo postgeminate high vowel deletion, as shown in (45). (45) Postgeminate High Vowel Deletion (Kabrah 2004, p. 141) a. mudar:is teacher -i -poss.1sg ‘my teacher’ b. *mudar:s-i This is what we would expect from a CV variety. Closed Syllable Shortening Mak:an does shorten internal CV:C syllables through CSS-Phon, though these are often broken by epenthesis instead. In (46-a) we see a closed CVC syllable surface [tub], although the corresponding form in (46-b) has a long vowel [tu:]. However, in (46-c), there is no shortening, but epenthesis has occurred instead. (46) Closed Syllable Shortening (Kabrah 2004, p. 80, 61) a. maktub be.written.ptcp.m -l -dat -u -obj.3sg.m ‘it was destined/written for him’ b. maktu:b be.written.ptcp. -at -sbj.f -l -dat -u -obj.3sg.m ‘she was destined for him’ c. ba:ba door -ha -poss.3sg.f ‘her door’ However, Mak:an does create some internal CV:C syllables through syncope, as shown in (47). (47) CV:C Syllables through Syncope (Kabrah 2004, p. 136) a. ji sbj.3 -sa:fr -travel.impf -u -sbj.pl ‘they travel’ 2. The Role of Quantity in Syllable Structure Restrictions 71 b. sQa:èb friend.m -i -poss.1sg ‘my male friend’ We would expect CV varieties to shorten all medial CV:C syllables, so in this regard Mak:an is closer to C varieties. However, the variation in whether it is shortened is unusual. Opaque Stress and Epenthesis Mak:an does not undergo opaque interactions between stress and epenthesis. Evaluation Just like SQanQa:ni:, Mak:an is not a canonical CV variety, in that there are more CV:C syllables than expected — thus is closer to a Cv analysis as in the work of Watson (2007). 2.4.4 Iraqi Based on the epenthesis in the example shown in (48), we would expect Iraqi to behave as a VC variety. (48) Epenthesis in a Triconsonantal Cluster (Majdi 1988, p. 201) ra:s head iiikbi:r big ‘a big head’ Final Clusters Iraqi rarely permits phrase final CC clusters. In (49-a) there is a sonority fall, in (49-b)–(49-c) a sonority rise; in (50-a–b) they are glide plus consonant clusters; and in (50-c–d) they are a nasal or liquid plus consonant cluster with optional epenthesis, marked by brackets. (49) Phrase Final CC Clusters (Majdi 1988, p. 193) a. galub ‘heart’