«7:... ‘1 a», L _. N - ‘ W1C Ill lllli'llllllllllllllllllll s v Lmnnmca , ( ) 31293 00910 87 ll || ‘ zgaé’fi' 58 LIBRARY ' Michigan State . AUniverslty ' This is to certify that the dissertation entitled Aspects of Comparative Jordanian and Modern Standard Arabic Phonology presented by Khalil AlSughayer has been accepted towards fulfillment of the requirements for #J—degree in Linguistics MSU is an Amrmalive Action/Equal Opportunity Institution fiajor professor 0-12771 . UBRARY Michigan State 5 University W“ PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE MSU Is An Affirmative Action/Equal Opportunity Institution c:\clrc\deledue.pma—p.1 Mic in partial Departmi Slavic, ASPECTS OF COMPARATIVE JORDANIAN AND MODERN STANDARD ARABIC PHONOLOGY BY Khalil Ibrahim Al—Sughayer A Dissertation Submitted to Michigan State University in partial fulfilment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Linguistics and Germanic, Slavic, Asian and African Languages 1990 Kha The studg analgzes th liodernStenderd Arabic ( phonemes, stress assign! gaminetes. These struct linear and non~lineer phi The two varieties ht tiiierin respect to /q/ let/k1, and /g/ while ABSTRACT ASPECTS OF COMPARATIVE JORDANIAN AND MODERN STANDARD ARABIC PHONOLOGY Khalil Ibrahim Al-Sughayer The study analyzes the relation between Jordanian (JA) and Modern Standard Arabic (MBA) in refererence to inventory of phonemes, stress assignment, high vowel alternation, and geminates. These structures are analyzed in reference to both linear and non—linear phonology. The two varieties have the same inventory of phonemes but they differ in respect to /q/, /g/, /k/, /c/, /g/, /ay/ and /aw/. MSA has /q/, /l/, respectively. MSA /ay/ and /aw/ are realized in JA as [eel and [00], respectively. The analyses of stress indicate that the two varieties have the same stress rules and that they differ only in respect to the rules which change the segmental string to which the stress rules apply. The analysis of stress also indicates that the metrical appr'roach sometimes fails to yield the correct prediction. ttilllttll review of the up to these vowels. As ior geminates, the segments underlyingly or behaviour of geminates i nominate glides. In JA.! hop to not. in the application 01 Proper inclusion Preced their structural descrip mophologicallg conditir Rules oi epenthesis art lhonotactic constraint Principle (P8P). Epent PSP is not violated. Alternating high vowels appear only in JA in the context of two consonants word—finally or that of three consonants word—medially. In nouns, alternating high vowels are predictable. Accordingly, they are epenthetic. ln verbs, the alternating high vowels which appear between the first and second radicals in the context of a vowel—initial suffix are unpredictable. Accordingly, we argue that they are underlying. The analyses of alternating high vowels involve a critical review of the approach of non—linear phonology approach to these vowels. As for geminates, the analyses indicate that they behave as two segments underlyingly and as one long segment phonetically. The behaviour of geminates is similar in both varieties except for geminate glides. In JA, glides degeminate word-finally. But in MSA, they do not. In the application of rules, the study invokes the principle of Proper Inclusion Precedence (PIP). Otherwise, rules apply whenever their structural description is met. Another principle is that mophologically conditioned rules take precedence over other rules. Rules of epenthesis are late rules which apply to satisfy a phonotactic constraint in compliance with the Peak of Sonority * Principle (P8P). Epenthesis does not apply to geminates because the PSP is not violated. I thank Professor Gil committee, and the com Prolessor Carolyn Hario lortheirhelpiul comme’ lhanks are also due to I PhD committee until sl ACKNOWLDGEMENT I thank Professor Grover Hudson, the Chairman of my Ph.D committee, and the committee members: Professor Barbara Abbott, Professor Carolyn Harford, and Professor Seok Song. I thank them for their helpful comments on earlier versions of this dissertation. Thanks are also due to Professor Carol Myers Scotton who was in my Ph.D committee until she left Michigan State University. iv LisloITables l Ihhreviations Chapteri: lntrodo 1.1. Object 12. Overvi 1.3. Gener 14- Sociol 15- Litera Chapter2 Segn 21 1 Gone 21 2 Distir 2‘ 3 Vowo 2‘ 4 Disth 221 Con: 222- Dist 223- Vov 2'2-4‘ DIS List of Tables Abbreviations Chapter 1: 1.1. 1.2. 1.3. 1.4. 1.5. Chapter 2: 2.1.1. 2.1.2. 2.1.3. 2.1.4. 2.2.1. 2.2.2. 2.2.3. 2.2.4. TABLE OF CONTENTS Introduction Objectives Overview General background Sociolinguistic background Literature review Segmental phonology Consonants in MSA Distinctive features of consonants in MSA Vowels in MSA Distinctive features of vowels in MSA Consonants in JA Distinctive features of consonants in JA Vowels in JA Distinctive features of vowels in JA V page xi xiii 11 13 1 6 1 6 20 22 24 25 27 29 30 it. m onus: Stress ix _ r: or: 3.1.1. STGPap 3.1.9. Tinned 3.121. Rhyme 3.12.2. Foota: 3.12.3. Buildin 2.3. 2.3.1. 2.3.2. Chapter 3: 3.1.1. 3.1.2. 3.1.2.1. 3.1.2.2. 3.1.2.3. 3.1.2.4. 3.1.2.5. 3.2.1. 3.2.2. 3.2.3. 3.2.4. 3.2.5. 3.2.6. The relation between consonants & vowels in MSA & JA Consonants in MSA and JA Vowels in MSA and JA Stress assignment in Jordanian and Modern Standard Arabic STGP approach to stress assignment The metrical approach to stress assignment Rhyme projection Foot assignment Building the metrical tree superstructure LabeHng The principle of extrametricality Stress Assignment in MSA (linear approach) Some problematic data to rule (15) Stress assignment in MSA (non-linear approach) Hayes' metrical model Stress assignment in JA (linear approach) Stress assignment in JA (non-linear approach) Hayes' model and stress assignment in JA (non-linear approach) Superheavy syllables and stress assignment vi 33 33 34 35 35 36 37 4o 42 45 49 49 53 57 69 73 76 81 82 ..v.e.s ",1 W sag -. .. 4: ,. 41.1 commons retreat-tot.rei::-.-e-.-1~.-;iii;u~.:.:__;-.e-:‘-, r _. W“. m £3 :1" - .3; 4,1,1. annfi 4.12 Stressin‘ 4.1.3. Alternate Iollowin 4.1.4. Concludi 4.2.1 Rules of 432 Censor 4-4~ Consor 45- The pa 4-3 Epenti vowe 41 Epenti non~ 4.7.1 . 3.3.1. 3.3.2. Chapter4: 4.1. 4.1.1. 4.1.2. 4.1.3. 4.1.4. 4.2.1 4.2.2. 4.3.1. 4.3.1.1. 4.3.1.2. 4.3.1.3. 4.3.2. 4.4. 4.5. 4.6. 4.7. 4.7.1. Evaluation of the metrical approach to stress assignment Evaluation of the STGP approach to stress assignment Alternating high vowels Evidence for epenthesis Vowel-initial suffixes Stress in nouns and verbs Alternaiion high vowels preceding or following a sonorant Concluding remarks Rules of epenthesis High vowel deletion Consonant clusters in JA Consonant clusters word-finally Consonant clusters word-medially Consonant clusters word-initially Consonant clusters in MSA Consonant clusters in reference to sonority The peak of sonority principle (PSP) Epenthesis in the context of a preceding high vowel Epenthesis in the context of a preceding non-high vowel Epenthesis in the context of ioiowing iabiais vii 86 90 91 93 93 94 95 100 101 102 108 108 108 110 111 113 116 121 123 126 Chapters: Protein 51 Probie 52 Problt (34). 5 3 Impei (W 531 The. 5 a 2 The 533 The 54 Con Chitters A El Th 4.7.2. 4.7.3. 4.7.4. 4.7.5. 4.7.6. 4.7.7. 4.8. 4.9. Chapter 5: 5.1. 5.2. 5.3. 5.3.1. 5.3.2. 5.3.3. 5.4. Chapter 6: 6.1. Epenthesis in the context of following velars Epenthesis in the context of following emphatics Epenthesis in the context of following non-emphatic-coron-ais Epenthesis in the context of following laterals Epenthesis in the context of a following flap Epenthesis in the context of following pharyngeais Rule application Summary Problematic data Problematic data to rule (26) Problematic data to rules (7'), (32), (34), and (40) Imperfect verbs and rules (7‘), (32), (34), and (40) The underlying vowel analysis The metathesis analysis The epenthesis analysis Conclusion A Non-linear Approach The non-linear approach viii 127 130 133 135 136 138 140 143 146 147 150 155 158 166 171 176 177 177 elmmou-aileznqnzs-ncn " l 5-. .. _ l.1.a‘ '- . -"‘1f.; "Huh". _- gnivooiio'. .- - ..ao.. c a 71 Gemirn 71 1 Renter 712 Gemin 7.12.1. Glider degei 7.12.1.1. Glide 7.12.1.2. Glide 7.1.2.1 .3. Glide 71 3 Gemi 7 l 4 Gemi 72 Gem 721 Repr 722 Gerr 6.1.1. 6.1.2. 6.1.3. 6.1.4. 6.1.5. 6.1.6. 6.1.7. 6.2. 6.3. Chapter 7: 7.1. 7.1.1. 7.1.2. 7.1.2.1. Syncope Epenthesis Association Default Emphasis spreading, rounding and backing Tier conilation Redundancy rules Application of non-iineari rules to JA data Evaluation Geminates Geminates in JA Representation Geminate glides Glide deletion, vocalization, and degemination 7.1.2.1.1. Glide deletion 7.1.2.1.2. Glide vocalization 7.1.2.1.3. Glide degemination 7.1.3. 7.1.4. 7.2. 7.2.1. 7.2.2. 7.2.2.1. 7.2.2.2. Geminates and stress Geminates and epenthesis Geminates in MSA Representation Geminate glides Glide deletion Glide vocalization ix 177 179 180 183 184 189 190 191 197 203 204 204 207 21 1 21 1 212 218 219 220 225 225 226 227 228 81 invenbr 82 Stressa 83 Aitemai 34. Gemini 85 Conoiw Reierences 7.2.2.3. 7.2.3. 7.2.4. 7.3. 7.3.1. 7.3.2. 7.3.3. 7.3.4. 7.4. Chapter 8: 8.1. 8.2. 8.3. 8.4. 8.5. References Glide degemination Geminates and stress Geminates and epenthesis The analysis of geminates in CV phonology Geminates and stress assignment Geminates and epenthesis Geminates and glide vocalization Geminates and glide degemination Summary Conclusion inventory of phonemes Stress assignment Alternating high vowels Geminates Conclusion 229 230 230 231 231 231 233 237 239 241 241 243 249 252 253 257 Table (1): consonants in' Table (2): Distinctive Teal Tahiei3):Disiinctiveieal Table (4): Consonants in Table (5): Distinctive Tea Table (6): Distinctive Tea Table (7): Epenthesis in LIST OF TABLES xi 17 21 24 26 28 30 125 Figure (1): Historical reIa (first assump Figure (2): Histories] rele' (second as ngre (3): Vowels in M1 Figure (4): Vowels in J! LIST OF FIGURES Figure (1): Historical relation between JA and MSA (first assumption) 10 Figure (2): Historical relation between JA and MSA (second assumption) 10 Figure (3): Vowels in MSA 23 Figure (4): Vowels in JA 29 xii u 1‘ l Wagering Consl Constraint on Assimilafio Classical Arabic (CA) Emphasis (E) Emphasis Rounding (EF High Vowel Deletion (Hl High Stem Vowel Deleti High Vowel Harmony (l Jordanian Arabic (JA) Lexical Category Prom‘ ABBREVIATIONS Adjacency-Identity Constraint (AlC) Constraint on Assimilation Rules (CAR) Classical Arabic (CA) Emphasis (E) Emphasis Rounding (ER) High Vowel Deletion (HVD) High Stem Vowel Deletion (HSVD) High Vowel Harmony (HVH) Jordanian Arabic (JA) 'Lexical Category Prominence Rule (LCPR) Linking Constraint (LC) Modern Standard Arabic (MSA) Obligatory Contour Principle (OCP) Proper Inclusion Precedence (PIP) Phonetic Realization (PR) Peak of Sonority Principle (PSP) Progressive High Vowel Harmony (PHVH) Regressive High Vowel Harmony (RHVH) Sound Patterns of English (SPE) Strong (S) xiii Well Formedness Condition (WFC) xiv The main objectivI Jordanian Arabic (JAH comparison is based d phonemes, rules of st oi geminates in both may gieid some evide between the two var Another objectiv theoretical phonoiog examination of the t Precedence (PIP), ti phonology vs. non-ii nature of geminatei JA is the Arabic identified as that c or regional aifiiiat Arabic oi the acco holg book of lsiam academic subject. MSA, one is not su mag not be alwagi Chapter I Introduction 1.1. Objectives: The main objective of this study is to compare the phonology of Jordanian Arabic (JA) to that of Modern Standard Arabic (MSA). The comparison is based on the synchronic analysis of the inventory of phonemes, rules of stress assignment, epenthesis, and the behavior of geminates in both varieties. The comparison of the two varieties may yield some evidence which reflect on the historical relation between the two varieties. Another objective is to examine a number of issues of theoretical phonology through this comparison. This includes an examination of the application of the principle of Proper Inclusion Precedence (PIP), the approach to stress assignment of linear phonology vs. non-linear phonology, rules of epenthesis, and the nature of geminates. JA is the Arabic dialect spoken in northern Jordan. MSA may be identified as that dialect of Arabic which does not reflect the home or regional affiliation of its speakers. MSA is exemplified in the Arabic of the accomplished readers of the news, of the qumaan "the holy book of Islam”, and of educated people formally discussing an academic subject. Accordingly, listening to a native speaker of MSA, one is not supposed to know where he comes from. But this may not be always the case. Historically, MSA is a descendant of I ' tip 793 AD. QSemaa the advent of Islam. It has to be pointed out t sometimes inexplicitly used. refer to MSA data. This has differences between CA and orunidentifiable. MSA still which CA has. But It may dl realization of these phonem Idl, which is a voiced empha what the phontic realizatio corresponding to CA [9/ fit In morphology and synt if any. Differences may ap endings. Speakers of MSA endings utterance—finally. result of lack of competer MSA is the learned vers Iirst concerns the availab establishes an adequate n The second question conc On these grounds, it is go identity perfectly unchar 2 Classical Arabic (CA). CA may be be identified as the language of the religious Islamic texts and the literature written during early Islam. It is the standard variety of early Islam as described by the ancient Arab grammarians such as Sibawaihi, who is believed to have died in 793 AD. (Semaan,l968: 38), the second century after the advent of Islam. It has to be pointed out that the terms MSA and CA are sometimes inexplicitly used. The term CA is sometimes used to refer to MSA data. This has to be attributed to the fact that differences between CA and MSA are either considered insignificant or unidentifiable. MSA still has the same inventory of phonemes which CA has. But it may differ in respect to the phonetic realization of these phonemes. For example, MSA /g/ is realized as IdI, which is a voiced emphatic stop. In CA, we do not know for sure what the phontic realization of /d/ is. The decription of the phone IcorreSponding to CA /9/ fits a lateral (in modern terms). In morphology and syntax, there are no significant differences, if any. Differences may appear in the the use of inflectional endings. Speakers of MSA tend to delete the Optional inflectional endings utterance-finally. Sometimes differences may arise as a result of lack of competence by the speakers of MSA. MSA is the learned version of CA. This raises two problems. The first concerns the availability of an adequate record of CA, which establishes an adequate model for the language teacher to present. The second question concerns the plausibility of perfect learning. On these grounds, it is questionable whether CA has maintained its identity perfectly unchanged over along period of time. 7 I.2.llverview: l The study is presented in a the plan, objectives, and back literature is reviewed. Chapter 2 presents the in MSA, which is necessary as a chapters. Chapter 3 deals with the Two approaches are consider I979), the assignment of st metrical structure of a war grounds that it does not alto that it resorts to making ar against this approach is be ofJA and MSA data. In the stress is based on the dist word. Here stress rules ap theirapplication, the print can be accounted for with‘ Chapter 4 deals with t‘ appears in the context of and that of three consona distinctive behaviour, I a roles of epenthesis are a ihonoiactic constraints. some phonological rules l.2. Overview: The study is presented in seven chapters. In the present chapter, the plan, objectives, and background are introduced and the literature is reviewed. Chapter 2 presents the inventory of phonemes in both JA and MSA, which is necessary as a background for the subsequent chapters. Chapter 3 deals with the assignment of stress in JA and MSA. Two approaches are considered. In the first approach (McCarthy, l979), the assignment of stress is considered in relation to the metrical structure of a word. I argue against this approach on the grounds that it does not always make the correct prediction, and that it resorts to making arbitrary assumptions. The argument against this approach is based on detailed and systematic analysis \of JA and MSA data. In the second approach, the assignment of stress is based on the distribution of the vowels and consonants in a word. Here stress rules apply without reference to the syllable. In their application, the principle of PIP is invoked. I argue that stress can be accounted for within the framework of this approach. Chapter 4 deals with the analysis of the high vowel which appears in the context of a cluster of two consonants word—finally and that of three consonants word medially. On the basis of its distinctive behaviour, I argue that this vowel is epenthetic and that rules of epenthesis are applied to satisfy language-specific phonotactic constraints. The analysis involves the postulation of some phonological rules to account for the alternation of the vowel .,.._ We ei'yeam stir ‘ >. '. we ,heer-Jefdo .rieig on! ..I - .il .1 an. . '6? ._o .'.. ' . m .somo referehc: i iof' Its domain. I tirapter 5 deals with semi winter evidence to the rule: chapter. The problematic do some nouns, the problem is I unpredictable by the rules 0 the context of some occasio predictions of the rules of I In] in the context of coronal epenthesis predict (ii. To account for these pri hypotheses. The first hypc not adjust to the phonolog Such data enter the lexico specific marking. The pet formally accounted for by hypothesis is that the big vowel. Within the frame' accounting for the qualit to account for the f ailur stress. The analysis at: between metathesis ant Apparently epenthetic v 4 between Iil, Iul, and zero. The postulated rules apply in compliance with the principle of PIP. The analysis in this chapter includes an identification of the contexts where the vowel alternates with zero. This involves some reference to the scale of sonority and an identification of its domain. Chapter 5 deals with some problematic data which are potential counter’evidence to the rules of epenthesis postulated in the third chapter. The problematic data include some nouns and verbs. In some nouns, the problem is that the quality of the high vowel is unpredictable by the rules of epenthesis. The vowel appears as [i] in the context of some occasions of preceding [u], contrary to the predictions of the rules of epenthesis. In some verbs, the vowel is [u] in the context of coronals and preceding [a], where the rules of epenthesis predict [i]. To account for these problematic data, we consider four hypotheses. The first hypothesis is that some borrowed words do not adjust to the phonological system of the borrowing language. Such data enter the lexicon of the borrowing language with a specific marking. The peculiar behavior of such data can be formally accounted for by invoking the principle of PIP. The second hypothesis is that the high vowel in question is a metathesized vowel. Within the framework of this hypothesis the question of accounting for the quality of the vowel does not arise. But we have to account for the failure of the metathesized vowel to be assigned stress. The analysis also involves a consideration of the interaction between metathesis and syncope. The third hypothesis is that the apparently epenthetic vowel is underlying and that it is not assigned lameness presents a new nation. This Involves firs . of this approach, and then a JA data. I argue against thi‘ based on some assumptions and that it does not always In chapter 7, we consid argue that geminates are t segment phonetically. Ge to rules of epenthesis and geminates, we consider so account for their integrity Constraint (AIC) (Guerssel lace) (McCarthy, 1979), a I986). Chapter 8 sums up and compared in respect to ti assignment, hIgh VOWEI 5 l3. General background: Arabic is Spoken all there is usually more tl 5 stress as the stress rules predict because in these contexts stress is morphologized. The fourth hypothesis is that the quality of the vowel in question is determined by that of the high vowel of the verb prefix. Chapter 6 presents a non-linear approach to the vowel in question. This involves first a presentation of the basic principles of this approach, and then an application of these principles to some JA data. I argue against this approach on the grounds that it is based on some assumptions which are arbitrary and too abstract, and that it does not always yield the correct prediction. In chapter 7, we consider the representation of geminatesl argue that geminates are two segments underlyingly and one long segment phonetically. Geminates are also considered in reference to rules of epenthesis and stress assignment. In the analysis of geminates, we consider some principles or constraints which account for their integrity. This includes the Adjacency~ldentity Constraint (AIC) (Guerssel, I977), the Obligatory Contour Principle (OCP) (McCarthy, I979), and the Linking Constraint (LC) (Hayes, 1986). Chapter 8 sums up and analyzes findings. JA and MSA are compared in respect to their inventory of phonemes, stress assignment, high vowel alternation, and geminates. 1.3. General background: Arabic is spoken allover the Arab World. In every Arab country, there is usually more than one local variety of Arabic spoken ._’ J . ~ mm is - _ intimidate in marinate ni Tswov Hits-1a d‘iev n1" WWW MTG! daemons ia:i.""'-n.:-i'r o .2.-fr? .‘ 5: ':.'-....- - vsometlmes the so— eslafanr‘iq aiz’arfer-‘J ‘--?;1L‘I}:3'.'-'i!.~'-9', .' -' :" :5: '- ' a ‘. ' ofconpetenceofitsspeake -. colloquial accents. When dlf speakers of the standard va . . .: -_~-“: (“.5.-ff ' "ff-'1‘" 51:331.} (13:..-'- ' '7 ‘t variety is actually the stan. To answer this question, Arabic term which is used t Arabic term for the standa 'the clearly articulated Ar the way town dwellers user tribes. The town dwellers the nomads spoke f ast (Ani term 7al-fusliaa "the Clet belief that there is only or distinguished from the ott This term has commonly to since early Islam, probabl 0n the basis of the an we do not know whether r Weare‘left with a gramrr assignment. The lack of Open for having different unlikely that ancient Tu; 6 besides the standard variety. The local varieties, usually referred to as colloquial Arabic, are varied. In most cases they are mutually intelligible. Unlike the local varieties, the standard variety is supposed to be identical among all of its speakers. But this is not the case. Sometimes the so-called standard variety reflects a lack of competence of its speakers, and at other times it reflects their colloquial accents. When differences arise among different speakers of the standard variety, the question to be raised is: which variety is actually the standard one? To answer this question, we consider the implications of the Arabic term which is used to refer to the standard variety. The Arabic term for the standard variety is aal-sarabiyyah ml-fushaa “the clearly articulated Arabic". Clarity of articulation may concern the way town dwellers used to speak compared to the nomadic tribes. The town dwellers were known for speaking slowly while the nomads Spoke fast (Anis,l952; Al—ounaim,l985). The Arabic term 7aI-fusl‘iaa "the clearly articulated" reflects the common belief that there is only one standard variety which has to be distinguished from the other varieties for its clarity and elequence. This term has commonly been used to refer to the standard variety since early Islam, probably earlier. On the basis of the analysis of the ancient Arab grammarians, we do not know whether ancient fusfraa had stress rules or not. We are‘left with a grammar which tells us nothing about stress assignment. The lack of reference to the stress leaves the door open for having different readings of contemporary fusfiaa. It is unlikely that ancient fusfraa did not have stress; most modern ashet phonemic. Two other problems with fvpllaa in relation to ancie articulation of some of the i Interpretation of the lingui~ grammarians to describe th discussed in the section w inventory of phonemes in t1 Now, to avoid the probl contemporary fushaa in r eXplained above, we will to contemporary fushaa as ‘ Jordan. It is beyond the s HSAinits broad sense. T standard varieties as the In this study, the MSA represent the intuition o fushaa as it is spoken i IISA in Jordan is the saw this would have to be d data in all of those cou MSA is usually influ common to find that it 7 world languages have stress and so ancient fushaa is unlikely to be an exception. It is more likely that stress is missing in the analyses just because the grammarians were not aware of the role of stress in grammar. A plausible explanation for this is that stress was not phonemic. Two other problems with the identification of contemporary fusfraa in relation to ancient fusfraa concern the correct articulation of some of the consonants, on the one hand, and the interpretation of the linguistic terms used by the ancient grammarians to describe them. These related problems will be discussed in the section which deals with the identification of the inventory of phonemes in MSA. Now, to avoid the problems involved in the identification of contemporary fusfraa in relation to ancient fusfraa as we have explained above, we will use the term "MSA" to refer to “contemporary fusfraa as it is spoken and taught nowadays in Jordan. It is beyond the scope of this study to identify the term MSA in its broad sense. This would require some analysis of the standard varieties as they are spoken all over the Arab countries. In this study, the MSA data to be analyzed is supposed to represent the intuition of a competent speaker of contemporary fusfraa as it is spoken in Jordan. This does not of course mean that MSA in Jordan is the same as MSA in the other Arabic countries, but this would have to be determined on the basis of the analysis of MSA data in all of those countries. MSA is usually influenced by the local varieties. It is quite common to find that MSA is spoken with a local accent. This should backers of the local variet formal education, are quite Theyare exposed to It on t discussions, etc. moreover when they say their prayer Influence of the standard v provided in the section de JA. As for the local varietr developed. Some people a: by Blau (l98l) as "Middle imperfect learning of the According to this assump interlanguage where tear and generalization. The process of learning the a some of the inflectional lnflectional endings wh gender, and mark the ve regularly missing in in not have developed onlr The role of the nati f L -- 8 explain why MSA may differ from one Arabic country to another. But the influence of the local varieties an MSA is usually constrained by formal instruction which strictly emphasizes adherence to the standard variety especially when reading the qur7aan. On the other hand, the local varieties are also influenced by the standard variety. Speakers of the local varieties, including those who have had no formal education, are quite often exposed to the standard variety. They are exposed to it on the radio, TV, in the mosque, formal discussions, etc. Moreover, they are expected to use it especially when they say their prayers. Some evidence illustrating the influence of the standard variety on the local variety will be provided in the section dealing with the inventory of phonemes in JA. As for the local varieties, we do not know for sure how they developed. Some people assume that the local varieties, referred to by Blau (1981) as "Middle Arabic", have developed as a result of imperfect learning of'the standard variety after the advent of Islam. According to this assumption, the local varieties developed as an interlanguage where learners apply the principles of simplification and generalization. The application of these principles in the process of learning the standard variety resulted in the deletion of some of the inflectional endings. It is true that many of the inflectional endings which mark the nouns for case, number, and gender, and mark the verb for person, mood, number, and gender are regularly missing in the local varieties. The local varieties would not have developed only in this way. The role of the native language or languages spoken by the people ' dealing w I: one a which were 590k ‘- In figure (I). we illustra betweenJA and HSA under a Irom CA and accordingly it i (2). we illustrate another p and MSA under an assumpti . unidentified Arabic variety Evidence for the relation b from the synchronic analys support of figure (I) would Identical with the underlyr developed. Evidence in sup underlying forms of JA an CA. In either case, eviden underlying forms of JA or exclude the possibility it shares these features wi 9 living in Jordan before entering Islam must also be taken into consideration. This can account for the diversity of local varieties. Any realistic study dealing with the development of the local varieties must take into consideration the role of the native languages which were spoken in the area before the spread of Islam. In figure (1), we illustrate a possible historical relation between JA and MSA under an assumption that JA has developed from CA and accordingly it is a sister languageto MSA. In figure (2), we illustrate another possible historical relation between JA and MSA under an assumption that JA has developed from an unidentified Arabic variety which is a sister language to CA. Evidence for the relation between the two varieties may be obtained from the synchronic analysis of the two varieties. Evidence in support of figure (I) would be that the underlying forms of JA are identical with the underlying forms of CA from which MSA has developed. Evidence in support of figure (2) would be that the underlying forms of JA are different from the underlying forms of CA. In either case, evidence cannot be conclusive. That the underlying forms of .JA are identical with those of CA does not exclude the possibility that there might be another variety which shares these features with CA. -- . . W guitaob you}: adrenalin ynA II ' - ' 9': ':~.~ .-..£-' eel-rm "apnoea-farm: amt no. .-u a :«m neat-3977 -'..---' " I’ISA Figure (2): Historical relat assumption) CA MSA 10 Figure (1): Historical relation between JA and MSA (first assumption) Proto-Arabic I Figure (2): Historical relation between JA and MSA (second assumption) Proto-Arabic I \ CA Pro-to-JA l I I | l I l 1 MSA JA 3fé§3}3”£fflfm’ {astigmatism I.4.Sociolinguistic Backgrou In Jordan, there are three urban, and'the rural. These t . iiiil’oqual‘hrabic. The three bedouin dialect is spoken by nomadic life in the desert southern parts of Jordan. T dialects spoken in the hurt considered an extension of developed as a result of no Syrian desert. The urban dialect is so similarities with the urb Socially, it is the most pr prestigeous status is refl dialects adjust their sper it exerts some influence The rural dialect is sr countryside and lead an r dialect shares a lot of s spoken in southern Syria The rural dialect is the "JA" is used in this stur The local varieties r In formal contexts, spa standard variety. Wher 11 I4. Sociolinguistic Background: In Jordan, there are three main local varieties: the bedouin, the urban, and the rural. These three are varieties or dialects of colloqual Arabic. The three varieties are mutually intelligible. The bedouin dialect is spoken by the bedouins who used to lead a nomadic life in the desert which constitutes the eastern and southern parts of Jordan. This dialect bears great similarity to the dialects spoken in the northern parts of Arabia, and can be considered an extension of them. This dialect is said to have developed as a result of nomadic migrations from Arabia into the Syrian desert. The urban dialect is spoken by the city-dwellers. It shares some similarities with the urban dialects spoken in Syria and Lebanon. Socially, it is the most prestigious dialect (Ibrahim, 1984). Its prestigeous status is reflected in the way the Speakers of the other dialects adjust their speech to fit it. Accordingly, we can say that it exerts some influence on the other dialects. The rural dialect is spoken by farmers who live in the countryside and lead an agricultural life in northern Jordan. This dialect shares a lot of similarity with the hawran dialect which is spoken in southern Syria, and can be considered an extension of it. The rural dialect is the object of analysis in this study. The term "JA" is used in this study to refer only to this variety. The local varieties are spoken at home and in informal contexts. In formal contexts, Speakers of the local varieties Speak the standard variety. When speakers of different dialects meet in an Informal situation, they usui where an urban and a rural S| makes some adjustment in h speaker while the speaker o hardly makes any adjustmer The rural speaker adjust phones of his dialect with t standard variety. For exam urban speaker would say: k speaker greeting another n are you?". But a rural spea keer haalak. Here the rur with the corresponding ur oiItSA. In this example in Iins borrowed from the I. both speakers would say: The following exampll urban speaker would say: rural speaker would say: context, they both use ti val-qureaan. Both the L phone IqI in the word qu corresponding to MSA q [7] and [av] with lee}. ( [q] with [g] and [a7] wit Where the rural va rural speaker does not f 12 informal situation, they usually adjust their speech. In a situation where an urban and a rural speaker meet, the rural speaker usually makes some adjustment in his Speech to fit that of the urban speaker while the speaker of the urban dialect, the prestigious one, hardly makes any adjustment in his speech. The rural speaker adjusts his Speech by replacing some of the phones of his dialect with the corresponding ones in the urban or standard variety. For example, an urban speaker greeting another urban speaker would say: kiif fiaalak "How are you?". A rural speaker greeting another rural Speaker would say: c'eef haalak “How are you?". But a rural Speaker talking to an urban speaker would say: keef haalak. Here the rural Speaker replaced the rural phone [6] with the corresponding urban one ik] which is also one of the phones of MSA. In this example we can not tell for sure whether the phone [k] is borrowed from the urban dialect or MSA. In a formal situation, both speakers would say: kayfa fraaluka, which is MSA. The following example illustrates the influence of MSA. An urban speaker would say: 7areet 7iI-qur7aan "I read the Qurvaan". A rural speaker would say: gareet 7il—qur7aan. But in a formal context, they both use the corresponding MSA form: qaraflu pal-dumaan. Both the urban and the rural speakers used the MSA phone [q] in the word qur7aan. But in urban ?areet "I read“, corresponding to MSA qara7tu, the urban speaker replaces {q} with I9] and Iae] with [ee]. 0n the other hand, the rural Speaker replaces [q] with [g] and [a7] with [ee]. Where the rural variety shares the same phone with MSA, the rural Speaker does not adjust his Speech to fit that of the urban (enemas, It is unlikely t urban phones is] and [z] for on these grounds, we ca Influenced by both MSA and reflected in the way rural . oi the other varieties. MS . influence one another. But that of the other dialects constrained by formal ins I adherence to the grammar 1.5. Literature review: There have been a nur aspects of MSA and rural related to the main objei addressed. 0n the other other Arabic varieties it includes studies of Pale Studies of MSA and distinction between th varieties are either ig phonological analyses. 13 speaker. For example, MSA and the rural variety share the phones [8] and [’o] in hadiie 7ibaasiy "a radio interview". But the urban phones corresponding to the MSA and the rural variety are is] and [2] as we can see in the correSponding urban words hadiis 7izaasiy. In this phrase, it is unlikely that the rural Speaker will substitute the urban phones [s] and [z] for his [9] and lb] which are also MSA phones. On these grounds, we can say that the rural dialect is likely to be influenced by both MSA and the urban variety. This influence is reflected in the way rural speakers adjust their speech to fit that of the other varieties. MSA and the local varieties mutually influence one another. But the influence of the rural dialect and that of the other dialects on the standard variety is constantly constrained by formal instruction which emphasizes strict adherence to the grammar of ancient fushaa. IS. Literature review: There have been a number of studies which deal with different aspects of MSA and rural JA. But none of these studies is directly related to the main objective to which this Study is directly addressed. On the other hand, there are some relevant studies of other Arabic varieties which are closely related to JA. This includes studies of Palestinian, Lebanese, and Bedouin JA. Studies of MSA and CA generally fail to make a clear—cut distinction between the two varieties. Differences between the two varieties are either ignored or considered insignificant in respect to phonological analyses. In most cases, what happens is that some In this study, to avoid t terms MSA and CA in a veg anidentlflcation of MSA. anative speaker of Arabic Arabic was aimed at achie informant should represen speakers of rural JA. Thi . MSA as it is spoken in dif Differences have to be de of representative sample countries. Studies of CA of Abdo (I969), Brame (i there are not that many 1 related to some aSpects (I987). Abdo (1969) analyse that the term CA is use analyses are done withi his analyses, the rules extrinsically ordered. intrinsically ordered. argued for the collapsr postulated. The press 14 data which are labelled MSA are identical to what might be labelled CA. In either case, the data is based on the current pronunciation of CA as documented in the ancient literature, and modern Arabic literature, as well. In this study, to avoid the problems related to the use of the terms MSA and CA in a vague sense, we have provided in 1.1 and 1.3 an identification of MSA. The informant of the MSA data analysed is a native speaker of Arabic, the author, whose formal education in Arabic was aimed at achieving fluency in CA. The MSA of the informant should represent the version of the MSA spoken by speakers of rural JA. This MSA does not necessarily differ from the MSA as it is spoken in different areas in the Arabic countries. Differences have to be determined on the basis of explicit analysis of representative samples of MSA data as Spoken in different Arabic countries. Studies of CA or MSA related to this study include those [of Abdo (1969), Brame (1971), and McCarthy (1979). As for JA, there are not that many studies related to the present one. A study related to some aSpectS of the present one is that of Alghazo (I987). Abdo (1969) analysed the assignment of stress in CA, claiming that the term CA is used to refer to both CA and MSA (1969: 1). His analyses are done within the framework of generative phonology. In his analyses, the rules postulated to account for stress are extrinsically ordered. The present study argues that rules are intrinsically ordered. Brame (1971) reviewed the work of Abdo and argued for the collapse into one rule of the rules which Abdo postulated. The present study extends to the analysis of stress in II‘I analyses of st anus in some Arabic data, called CA Is not different f This study argues against t complicated, resorts to so fails to achieve observatio Alghazo (1987) analys Arabic on the basis of rur. is the object of analysis i done within the framewor phonology. Thie present 3 resorts to very arbitrary observational adequacy. The present study is of the relation between t problems involved in an ' and geminates, on the at an understanding of the draws attention to the I relation between MSA 2 Arab world. 15 JA the application of the principles on the basis of which this rule is postulated. Unlike the previous analyses, McCarthy postulate a metrical model for the analyses of stress and applied it to the assignment of stress in some Arabic data, including CA. The data which McCarthy called CA is not different from the MSA data analyzed in this study. This study argues against this model on the grounds that it is complicated, resorts to some'arbitrary assumptions, and above all fails to achieve observational adequacy. Alghazo (1987) analysed syncope and epenthesis in Levantine Arabic on the basis of rural JA data, the same Arabic variety which is the object of analysis in the present study. Alghazo's analysis is done within the framework of autosegmental and metrical phonology. Thie present study argues that his analysis of epenthesis resorts to very arbitrary assumptions and above all fails to achieve observational adequacy. The present study is motivated by the lack of a systematic study of the relation between JA and MSA, on the one hand, and the problems involved in an integrated analysis of stress, epenthesis, and geminates, on the other. The present study should contribute to an understanding of the the relation between MSA and JA. The study draws attention to the need for comprehensive research on the relation between MSA and the local varieties Spoken all over the Arab world. Seg As necessary backgroun presents the inventory of p divided into three sections. inventory of phonemes in t1 identified by place and ma by distinctive features, on some reference to the cor section, the vowels and c manner of articulation, 0 features, an the other. In between the inventory of 2.I.I.Consonants in MSA: Table (1) presents th represented in the articn consonants are describe articulation. Chapter 2 Segmental Phonology AS necessary background to subsequent chapters, this chapter presents the inventory of phonemes in MSA and JA. The chapter is divided into three sections. In the first section, we present the inventory of phonemes in MSA. Here, the vowels and consonants are identified by place and manner of articulation, on the one hand, and by distinctive features, on the other. The analyses also involve some reference to the corresponding phonemes in CA. In the second section, the vowels and consonants in JA are identified by place and manner of articulation, on the one hand, and by their distinctive features, on the other. In the third section, we draw a comparison between the inventory of phonemes in both varieties. 2.1.1. Consonants in MSA: Table (1) presents the inventory of consonant phonemes in MSA represented in the articulation of the informant of this study. The consonants are described in respect to place and manner of articulation. 16 stop It emphatic stop fricative f emphatic f ricative aifricate nasal m lateral flap glide W 1 = labial 2 S : alveolar 6 9: pharyngeal 10 Table (1) shows that six stops (b,t,d,k,q,?). 1 eleven fricatives tr,e,'a (he), one affricate (I), semivowels (y,w). The consonant phon similar to those of CA 17 Table (I): Consonants in MSA stop b t d k q 7 emphatic stop tg fricative f as sz 75 x25 he h emphatic fricative ‘b s affricate j nasal m n lateral l flap . r glide w y 1 = labial 2 = labio-dental 3 : interdental 4 : dental 5 = alveolar 6 : palatal 7: velar 8 : uvular 9 = pharyngeal 10 = glottal Table (1) shows that MSA has twenty—eight consonant phonemes: six stops (b,t,d,k,q,v), two emphatic stops (Lg) two nasals (m,n), eleven fricatives (f,e,’o,s,z,§,x,b,h,q,h), two emphatic fricatives (Ens), one affricate (I), one lateral (I), one flap (r), and two semivowels (y,w). The consonant phonemes in MSA (contemporary fushaa ) are quite similar to those of CA (ancient fushaa ). There is hardly any ‘itt‘ih'CA they I ' problem lies in the interpr (1863-93) as quoted by Se pronounced with the vote -. to this Interpretation, the this case, the I!) which he Thisleads us to the concl the description of the pho have changed from being Sibawaihi was probab of the role of the vocal c “the unawareness of the least the role of the voc accounts for some min0l works (Samoan, 1968: SE inaccurate in his descri support of this can be d unemphatic phone corre Dhones are: ((1, [d], ('91. ones are: [t]. (d), to], an the CA emphatic phone corresponding unemph think, that Id) is the c (1316 A. H. Islamic Lu 18 significant difference between the CA consonants as described by Sibawaihi and those of MSA as illustrated in Table (1), except for the phones: It], [g], [g], and probably [r]. In MSA, It) and Iql are voiceless. But in CA they are described as being majhuurah. The problem lies in the interpretation of the term majhuur. For Lane (1863-93) as quoted by Semaan (1968: 43) the term means "pronounced with the voice, and not with the breath only". According to this interpretation, the phones it] and [q] in CA must be voiced. In this case, the [t] which he described is equivalent to current MSA [d]. This leads us to the conclusion that either Sibawihi-is inaccurate in the description of the phones it] and [d], or that since his time they have changed from being voiced into being voiceless. Sibawaihi was probably inaccurate. He was probably not aware of the role of the vocal cords in the process of sound production. "the unawareness of the Arab phoneticians of the existence, or at 'least the role of the vocal cords in the process of sound production accounts for some minor inaccuracies in early Arabic phonetic works (Semaan,1968: 55)". Thus it is likely that Sibawaihi is inaccurate in his description of these phones. Some evidence in support of this can be derived from his assumption that the unemphatic phone corresponding to It] is [d]. In MSA, the emphatic phones are: [t], [g], [’b], and [s], and the corresponding unemphatic ones are: [t]. [d], [’o], and [s], respectively. According to Sibawaihi, the CA emphatic phone which we refer to as [g] does not have a corresponding unemphatic one. This is contrary to what we tend to think, that [d] is the corresponding unemphatic CA phone. Sibawaihi (1316 A. H. Islamic Lunar Calendar: 406) says: "wa Iawlaa become It] and there would (language)? In (I), we illu phones and the correspondt the claim of Sibawaihi. (I): It/ : /d/ /§/ : /s/ It)! : /b/ /d/ : --- All the emphatic phol phonemes except the pho corresponding to the amp expected. If this is the something different fro 118A, respectively. The ancient Arab gr description of MSA [d], wsn Ipl (Anis, 1971: 64 and Id]. In the articula issald (Ants, 1971: 48 side of the mouth or bl 19 7a]—7i,fbaaq Iasaaraf pal-pas? deal-an wa 2aI-Saad Sim—an wa pal-£5607 baal-an wa-Ia—xarafat pal-,daadmin 7aI-kalaam.” The English translation reads: ”Had there been no feature of emphasis, It] would have become Id] and is] would have become is], It») would have become lb] and there would have been no [pl in the Speech (language)." In (I), we illustrate this by providing the emphatic phones and the corresponding unemphatic phones in accordance with the claim of Sibawaihi. (I) /t/ : /d/ /s/ : /8/ /©/ : /b/ /d/ : -—- All the emphatic phonemes have non—emphatic corresponding phonemes except the phoneme, /d/. The non—emphatic phoneme corresponding to the emphatic phoneme /t/ is /d/, not /t/ as may be expected. If this is the case, then emphatic [g] and [ti in CA must be something different from what we know to be emphatic [gland It] in MSA, respectively. The ancient Arab grammarians description of "[t]" fits the description of MSA [d], while their description of "[g]" does not fit MSA [d] (Anis, 1971: 64). CA [d1 seems to be something between [’b] and Id]. In the articulation of CA "(d)" by the time of early Islam, it is said (Anis, 1971: 48—49) that the air goes out from either one Side of the mouth or both Sides. If this description is accurate, CA ~ .1: We“ the ancient . he mmufllmm‘h _, .1 aim nah-mam: unite»! ,. 11,”,an W b", [y] amazed avert aluow .- ., ,_ ' def (ad on need sever: blurrw '— ' ‘. "" 1': ""3!” can" Irl 3"1'79-Jd . - '. ._ _,_.,.-. maqaqah 'deltcate;lig . . ...rl -I:---I-.'-: ‘- Jiladamg art? gflib’i'o'm I: ‘ linguistics are: emphatic a ’ 1'“. 553,75; -l,-.'.'.= .--' 1’5“?” ' N ' . , Iheyusedonearthographi- ' the differences were not . considered one phoneme (I emphatic consonant. It is example, in MSA, la/ is b ofapreceding lrl, it is r with the emphatic phone 2.1.2. DistinctIVe feature Table (2) presents th 20 "ng" must share some similarity with the lateral sounds such as [l], and accordingly what the ancient Arab grammarians called "[gl" must be a different phone which is neither [g] nor [a]. As for [r], the ancient Arab grammarians recognized two types of [r]: one is mufaxxamah "dignified,- dark" and the other is muraqqaqah "delicate,- light". The equivalent terms in modern linguistics are: emphatic and non-emphatic, respectively. However, they used one orthographic symbol to indicate both types of /r/ as if the differences were not contrastive. Accordingly, they are considered one phoneme (Anis, 197-1: 67). in MSA, [r] behaves as an emphatic consonant. It is [+back] like the emphatic coronals. For example, in MSA, /a/ is basically realized as Ice}, but in the context of a preceding /r/, it is realized as [a]. in this way, /r/ patterns with the emphatic phonemes: /s/, /’o/, /t/, and /g/. 2.1.2. Distinctive features of consonants in MSA: Table (2) presents the distinctive features of consonants in MSA. . .' was!“ man . mum/s Ia-siene 9:1: “110? a . If; . er 9'10 11! lo :i‘ngm” risz.-a.-'..- -. 7 I slimline w one ‘Nh . n" to: h-f- BM tide] ”mien.” .-~.- urb-5:! 3571"?!" "' -' -' ' . h f I " ""-‘"‘-""-' “W- .. -.... -_--- I ' - u. - l- . .. ' - - ' . . .. ., -_ eons.“- +++ ++++ / '75:.5M’E.” --:9r:-'.-' - ' - . . . . i.‘ is! . I ' . ' - . - ‘ cont. -+ ++ + nus. -- --- ~—-— ant, ++ +++ ++++ cor. -- +++ ++++ back-- ~~+ -+-+ ‘vod.+- -++ —-++ 21 Table (2): Distinctive features of consonants in MSA bf 86?; ttdg] sszs kqxo he ?h mnlr wy syll.-— ——— ————— ———— ———— —- -— ——-— -— con$,++ +++ +++++ ++++ ++++ ++ ++ ++++ -— son_ —— —-— ————— ---- -—--- —— —— ++++ ++ cont.—+ +++ ————— ++++ ——++ ++ ~+ -——+ ++ nas.—- --- ----- ---- ---- -- -- ++-- -- antI++ +++ ++++- +++— ~—-— —— -— ++++ ~— cor.-- +++ +++++ ++++ —-—- -—- —— —+++ -- back—— ——+ —+—+— -+—- ++++ —— —- ———-+ +— high—— --— ———-—+ -——+ ++++ -- —— ——-— ++ low -— ——— ————— -——— ——-— ++ —— ——-—- —— uw Home: in? 5x1? 2‘3-7. i lI-contl Thi w «men that the Mini oiolr flow (not total bloc constriction (Chomsky and ballistic movement of the the air stream. In [t], thee in Table (2), Arabic in identified in SPE; it alwa be noted that the pharyng emphatics ls, a, t, g] are consonants are referred identified as l+cor, -higl 306-307). It has also to they are [-highl. 2.l.3. Vowels in MSA: MSA has three shor short vowels are: /i/. and /aa/. Long vowels in respect to length. l quality but not in qual 22 In Table (2), [r] and [1] are distinguished from each other by the feature [back]; [rl is [+backl while [1] is t-backl. But [+backl is an allophonic feature of [1]. To distinguish [r] from [l], we consider [r] [+cont] and [l] l-cont]. This categorization is based on the assumption that the defining characteristic of st0ps is the blockage of air flow (not total blockage of air flow) past the primary constriction (Chomsky and Halle, 1968: 318). Because flap [r] is a ballistic movement of the tongue, there is no voluntary blockage of the air stream. ln [1], there is airflow laterally. In Table (2), Arabic [h] is classified [+consl, not [—cons] as it is identified in SPE; it always functions like a fricative. It has also to be noted that the pharyngeals in, s] are [+low, +backl while the emphatics is, ’9, t, g] are [+cor, +backl. In SPE, the Arabic emphatic consonants are referred to as pharyngealized consonants which are identified as (+cor, -high, +low, +backl (Chomsky and Halle, 1968: 306—307). it has also to be noted that [x, 25} are uvular. Accordingly, they are [~high]. 2.1.3. Vowels in MSA: MSA has three short vowel phonemes and three long ones. The short vowels are: /i/, /u/, and /a/. The long vowels are: /ii/, /uu/, and /aa/. Long vowels contrast with the corresponding short vowels in respect to length. Long vowels are identical with short vowels in quality but not in quantity. This is Figure (3). .‘~-' ,- , f .. .. a? (the nuts" on, :_ L.- ' . ' mainframe” ain‘r [jeep-3-] ”than [Mow] '.r‘;:'l'il_-P. P5 'efi - .; = _ _ 5.50512“ .'i_,l ;..- f '. . ., :.I-. _ ‘ ._ I i " '- - - .' . ~-v - _- n. l ' ' - I"! .- G fi- a. . 'l l . We refer to the low v and /aa/.- These vowel p 1 They are basically l-higl Butin the context of a p to], [cm], respectively. T Rule (2): {+syll ---—> be l-high] The vowels liI, lii [uz], respectively. in t allophonic variation. In CA, Sibawaihi V 23 Figure (3): Vowels in MSA We refer to the low vowel phonemes by using the symbols /a/ and /aa/. These vowel phonemes have some allophonic variation. They are basically [-high, —back], that is [ae] and [83:], respectively. But in the context of a preceding emphatic or velar, they are [+back], [a], [(11, respectively. This is accounted for by rule (2). Rule (2): [+3le ———> [+backl / l—syl] __ l-high] [+backl The vowels /i/, /ii/, /u/, and /uu/ are realized as [i], [i:l, [u], and lu:], reSpectively. in their phonetic realization, they hardly have any allOphonic variation. in CA, Sibawaihi was unable to recognize the real nature of a [1 The long vow tithe Arabic letters yea short vowels lit, [u], and l kosrah, dommah, and fat The analysis provided significant differerence l1 represented in the inform described by Sibawaihi ( Differerences in the deso can be attributed to the descriptive accuracy or organs. 2.1.4. Distinctive featul Table (3) presents ‘ Table (3): Distinctive my bat 24 vowel. He regards the vocalization of a consonant as an inherent part of the consonant (Semaan,1968, pp. 56). Long vowels were referred to as fluruuf "letters", and short vowels as harakaat "diacritics". The long vowels /ii/, /uu/, and /aa/ were referred to by the Arabic letters yaap, waaw, and ?alif, respectively. The short vowels [i], [u], and [a] were referred to by the diacritics kasrah, gammah, and fatfiah, respectively. The analysis provided above shows that there is hardly any significant differerence between the inventory of phonemes in MSA represented in the informant of this study and those of CA as described by Sibawaihi (1316 A.H.), Semaan (1968) and Anis (1971). Differerences in the description of the articulation of some phones can be attributed to the ancient Arab grammarians“ lack of descriptive accuracy or lack of awareness of the role of the vocal organs. 2.1.4. Distinctive features of vowels in MSA: Table (3) presents the distinctive features of the vowels in MSA. Table (3): Distinctive features of vowels in MSA , . ,. .. . by . ..‘.:.‘,.‘:‘.=. '11 in the Want“ Accordingly, the vowel I to! value in respect to the dis. [+backl at the phonetic lew, orvelar consonant. Other‘ vowels are considered to vowels. 2.2.1. Consonants in JA: Table (4) identifies ‘ in respect to their place 25 in Table (3), the distinctive feature of the vowel [i] is [+high] and that of [u] is [+backl The vowel /a/ is [-high, -back]. it is distinguished by default by the feature [-high]. The feature {-back] is allophonic in the identification of the non-high vowel /a/. Accordingly, the vowel /a/ could be assigned an underlying zero value in respect to the distinctive feature [back]. The vowel /a/ is [+backl at the phonetic level in the context of a preceding emphatic or velar consonant. Otherwise, it is {-back]. The corresponding long vowels are considered to be sequences of two identical short VOW618. 2.2.1. Consonants in JA: Table (4) identifies the inventory of consonant phonemes in JA in respect to their place and manner of articulation. stop '3 emphatic stop fricative i emphatic fricative oiirlcate nasal in lateral ilap glide w l : labial 2 5: alveolar 6 g=i1tttiryngeal l0 26 Table (4): Consonants in JA stop b t d k g '2 emphatic stop ‘5 fricative f ea sz ma X 0( D" a!) 3" emphatic fricative ’9 s affricate 6 \_¢ nasal m n lateral l flap r glide w y l : labial 2 : labio—dental 3 : interdental 4 = dental 5 : alveolar 6 : palatal 7: velar 8 : uvular 9 : pharyngeal 10 : glottal Table (4) shows that JA has twenty—eight consonant phonemes: six Stops /b,t,d,k,g,?/, one emphatic stop /t/ two nasals /m,n/, eleven fricatives /f,8,’b,s,z,s,x,25,h,s,h/, two emphatic fricatives ”DAV, two affricates /c,j/, one lateral /l/, one flap /r/, and two Semivowels /y,w/. In comparison with l’lSA, .JA differs in respect to the consonants '-i: . '- c lg. unemphat is Ipl. on It! corresponds to l developed in the COMOXI 0 other contexts. Thus, we 'ieiling lies“ and in that 1 we can have lklin the con something". Without gett contexts of Id/ and /k/, l M and /é/. The appeara attributed to borrowing 1 2.2.2. Distinctive feature lnTable (5). we pro: anA. 27 /g/, /(:/, and /’o/. JA /g/ corresponds to the MSA phoneme /q/. JA /'9/ corresponds to MSA /g/ and /’o/. JA has no emphatic /<_j/ corresponding to unemphatic /d/. MSA /g/ and /tp/ have merged into one, that is /’t_>/. JA /6/ corresponds to MSA /k/. JA /6/ seems to have first developed in the context of a preceding or following /i/ and then to other contexts. Thus, we can have [6] in the context of [i] as in c'ibib "telling lies" and in that of /a/ as in c'aoab "he said a lie". Still, we can have /k/in the context of /i/ as in kisrih "a fragment of something". Without getting further into the analysis of the contexts of /c/ and /k/, we can say that MSA /k/ corresponds to .JA /k/ and /c':/. The appearance of [kl before the vowel [i] can be attributed to borrowing including dialect borrowing. 2.2.2. Distinctive features of consonants in JA: In Table (5), we present the distinctive features of consonants in JA. ’ atimm o: \d\ zliadi.ane l ! .tz'irl evori a: erresa " :3“ At, ‘-..-"'x .A'.'“'- c1 .u 151 \3 .u. : - - . .q. . -.. .- . .. -' .-' . ‘. a. 'q':‘:""7:|_. “I h. u.-- ----- i 0, l‘9r{:- -.- I'Id ‘ : I' TTII '1'. 1" : i'.’ ||'I '1 {. ___ _ _ L.‘ I I _ I _ g” t I I I wm++ +++ ++++l C;"‘:'._ .-: E." I.‘ - --I. . . .. _ I ‘ mt-++++ ---- M&-- --- ---- m.+++++ +++- W.--+++ ++++ back -. ..-+ _+__ high ----- ___‘ lW.-- ___ ‘_ VN,h _++ __+ 28 Table (5): Distinctive features of consonants in JA bf 83?} ttdé] sszs kgxd he 7h mnlrwy syll. -- —-— ----- —--- —--— —- —— ——-- —— COhS.++ +++ +++++ ++++ ++++ ++ ++ ++++ —— SOI’l. " ‘-— ----- ---- —--— -— —- ++++ ++ CODt.—+ +++ ————— ++++ -—++ ++ —+ ———+ ++ i183. -- -—— ————— -——— ——_.. .._ __ +4...- __ ant, ++ +++ +++—— +++- —-—- —— -— ++++ .— cor, —— +++ +++++ ++++ -—--— —— —— -+++ .— back -— ——+ —+——- -+-— ++++ —— —— ——-+ +— hlgh -- —-— ——-—++ ——-+ ++++ —— ——- —-—— ++ low.—— ——— _____ ____ ____ ++ _- __-_ __ VCd.+- —++ ——+—+ -—+-— —+—+ —+ —- +++++ + 2.2.3. Vowels in JA: Like t‘lSA, JA has three ones: liil, luul, and /aa/ in reference to the positi Figure (4): JA vowels [,11 29 The non—syllabic segments in JA are similar to the non—syllabic segments in MSA except for the segments [q] and [g] which JA lacks, and the segments [6], and [g] which MSA lacks. Accordingly, the distinctive feature analyses in MSA largely apply to the analyses of JA. 2.2.3. Vowels in JA: Like MSA, JA has three short vowels: /i/, /u/, /a/, and three long ones: /ii/, /uu/, and /aa/. in figure (4), these vowels are illustrated in reference to the positions of the cardinal vowels. Figure (4): JA vowels JA also has the phones [e:] and [0:]. These are not phonemes underlyingly. They are the phonetic realization of the underlying glides /ay/ and /aw/, respectively. Thus the phonetic realization 0f _ , him I!) chem”; .- yie‘e'isi were at aseylene sauiae‘ O'JH'Ji‘HSalh At. 22.4. Distinctive ieatureg lnTable (a), we prose anA, which are identica? Table (6): Distinctive f 81 high back Aproblem arises i1 vowels laa/ from the ‘ realization of Iay/ ant and [o] from lay/ and [low] so as to distingl is also (back, ~high] The JA vowel /a/ consonant other than i 30 /bayt/ "home" and/gawm/ "people" is [be:t] and [gozm], respectively. The underlying glides show up in the plural forms: [byuut] "homes" and [gwaam] "peoples", patterning with [sort] "condition" and [sru:t] "conditions". 2.2.4. Distinctive features of vowels in JA: In Table (6), we present the distinctive features of the vowels in JA, which are identical to those in MSA. Table (6): Distinctive features of JA vowels high + + — + + - back —+--+ - A problem arises in the process of distinguishing the long vowels /aa/ from the long mid—vowels [e:], and [0:], the phonetic realization of /ay/ and /aw/, respectively. in the derivation of [e:] and [0:] from /ay/ and /aw/, we need the non—distinctive feature [low] so as to distinguish [e:] which is {—back, —high] from [aez] which is also {-back, -high]. The JA vowel /a/ is raised when followed by a glide and a consonant other than a glide. This is rule (3) 01’ VOWGI raising. Rule (3) applies to Id [olbeiorea w glide. Rd Igawml, yielding beyi rule (4), which vocalizes consonant other than a y This is rule (4). Rule (4): icons] «4 l-syl] ln ('5), we illustra“ the derivation of [siid [900ml "people“. (5): 31 Rule (3): [+syll ——-—> [-low] / ..._ [+sonl [—syl] [aback] [~consl [+cons] [obeCk] Rule (3) applies to /a/, raising it to [e] before a y glide, and to to] before a w glide. Rule (3) applies to words like /bayt/ and /gawm/, yielding beyt and gowm, respectively. Rule (3) feeds rule (4), which vocalizes glides between a [—low] vowel and a consonant other than a glide. This includes the vowels [i], and [u]. This is rule (4). Rule (4): [~cons] ———> [+syl] / [+syl] __ [+cons] {-syl] lo (tb l-highl Rule (2) applies to 32 [siid] Ikuus] [beet] [goom] PR In the derivation of [beet] and [goom] in (5), the rules do not have to be extrinsically ordered. The rules apply whenever their SD is met. The SD of rule (4) of glide vocalization is not satisfied by either /bayt/ or /gawm/,- rule (4) does not apply to [+low] vowels. The SD of rule (4) is satisfied when rule (3) of vowel raising has applied. Rules (3) and (4) are in a feeding relation,- rule (3) of vowel raising feeds into rule (4) of glide vocalization. The vowels /a/ and /aa/ are basically, I assume, {-backl, [ael and [aez]. But they are [+back], [a] and [e:] after velars and emphatics. This is accounted for by rule (2) of low vowel backing postulated above and repeated here for convenience. Rule (2): [+syl] —-—> [+backl / [-syl] __ {-highl [+backl Rule (2) applies to /_salab/, /talab/, /tSaIab/, /xalas/, /‘oalam/, yielding an [a] vowel after the emphatic or velar consonant. In the context of a preceding [a] vowel, light [I] becomes emphatic [1]. This is accounted for by rule (6). Rule (6): [+son] ———> [+backl/ [+syll __ {-cont] [+backl I—nasl 1 Italobl, 1 2.3.The relation betwee 1 In this section, we di respect to the vowels an 2.3.1. Consonants in MSA Consonant phomemi ltl, lql, /g/, /g/, and I present the MSA phoner (7): 118A [kl correspor which is identical to phonemic split in JA corresponds to JA It Ia] and [t3] merged in 33 Rule (6) applies to the output of rule (2), yielding an emphatic [l] in the context of a preceding [+backl vowel, including [a]. The output of rule (6) feeds into rule (2) which applies iteratively, yielding [salabL [talabL [oalab], [xalas], and [balam], respectively. 2.3. The relation between consonants & vowels in MSA & JA: In this section, we draw a comparison between JA and MSA in reSpect to the vowels and consonants they have. 2.3.1. Consonants in MSA and JA: Consonant phomemes in MSA and JA are similar except for /k/, /5/, /q/, /g/, /d/, and /'o/. This is illustrated in (7) where we present the MSA phonemes and the corresponding JA ones. (7): .fl J43 k k, 5 CI 9 a MSA [k] corresponds to JA [6]. But in some cognates JA has [kl which is identical to the corresponding MSA phone. This indicates a Dhonemic Split in JA. MSA [q] corresponds to JA [9]. MSA [d] corresponds to JA [’29]. This indicates a phonemic merger in JA; MSA [d] and [’o] merged into JA [’9]. Other differences between MSA and _ may; “area 'I . . ' gs. .12»!an .iaaiaal .iuniail denies! WNW f. In (a), we present th: ,. . -- -. a- «:..‘: non. (at Vowels (13A 1 it u uu a 68 all aw Considering the M1 vowels in both variet' vowels lee] and [on] w Accordingly, we can a Without exception at #4 34 JA, if any, are insignificant and without phonological significance. 2.3.2. Vowels in MSA and JA: In (8), we present the vowels in MSA and the corresponding ones in JA. (8): Vowels M_SA QA i : 1' Il' : it u : u uu : uu a : a aa : aa ay : ee (/ay/) aw : oo (/aw/) Considering the MSA and JA vowels in (8), we can see that the vowels in both varieties are similar except for the JA mid—long vowels [ee] and [00] which derive from /ay/ and /aw/, respectively. Accordingly, we can say that vowels in MSA and JA are similar without exception at the underlying level. 39110 git ' In this chapter, we a . Jordanian (JA) and Moder] be done in reference to t (STGP) (Chomsky and Hal (thorthy,1979a; 1979b The chapter is divide we present the principle Thisincludes a present assignment, on the one The presentation of the or McCarthy (1979) am we analyse the assignr third section, the two based on the capacity prediction and the app 3.1.]. The STGP apprt According to the 1 distribution of vowe Chapter 3 Stress assignment in Jordanian and Modern Standard Arabic In this chapter, we analyze the assignment of stress in both Jordanian (JA) and Modern Standard Arabic (MSA). The analysis will be done in reference to the standard theory of generative phonology (STGP) (Chomsky and Halle, 1968), and metrical phonology (McCarthy,1979a; 1979b; Hayes,1981;1982). The chapter is divided into three sections. In the first section, we present the principles according to which stress is assigned. This includes a presentation of the STGP approach to stress assignment, on the one hand, and the metrical approach, on the other. The presentation of the metrical approach is based on the analysis of McCarthy (1979) and Hayes (1981,1982). In the second section, we analyse the assignment of stress to JA and MSA data. In the third section, the two approaches are evaluated. The evaluation is based on the capacity of each approach to make the correct prediction and the apparent validity of the principles they claim. 3.1.1. The STGP approach to stress assignment According to the STGP, stress assignment rules are based on the distribution of vowels and consonants. Stress rules apply to a 35 Clusters are identified a Isdeiined as a string co followed by no more the identified as a string co followed by two or more followed by any number An example of a compt - type in the English wort identified as VCC or v 3.1.2. The metrical app In the metrical app McCarthy (1979a; 1971 Hayes (1980; 1982;19 structure of the metr? building of metrical t each level, there are account for the varia levels are: rhyme prc andlabelling. 36 string of syllabic and non-syllabic segments. In STGP, neither the syllable structure nor the syllable boundary are involved in the formulation of the stress rules. In Arabic, the stress rules seem to be sensitive to the kind of clusters vowels and consonants make. Clusters are identified as either strong or weak. A "weak cluster" is defined as a string consisting of a single vocalic nucleus followed by no more than one consonant. A "strong cluster" is identified as a string consisting of either a vocalic nucleus followed by two or more consonants or a complex vocalic nucleus followed by any number of consonants (Chomsky and Halle,1968: 29). An example of a complex vocalic nucleus is the part written in bold type in the English words "paint". In Arabic, a strong cluster is identified as VCC or W (Abdo, I969; Brame, 1971). 3.1.2. The metrical approach to stress assignment In the metrical approach as evidenced in the analysis of McCarthy (1979a,- 1979b) and the modifications introduced into it by Hayes (1980,- 1982;1985), stress assignment is based on the structure of the metrical tree in a word. According to McCarthy, the building of metrical tree of a word is realized in four levels. In each level, there are parametric variations which are supposed to account for the variations between different languages. The four levels are: rhyme projection, foot assignment, word—level structure, and labelling. .f fir‘ «my. ,.,,:» .. g - 1 from": in exile] an! a: avliranfu u ‘Iectatt rooted-or Identified. “my «,3 prim; - we're fir- berth-“- '-:: e'io anneal: _ , r . _ _. .-.-;. ,, hasisoftheirhavinga ...1” 1 ob! «1:75 / " ' "'1 , 5'1 formally defined the rhy ' n. e - ‘ ' root (McCarthy, 1979a: the syllable nucleus, wh Iotheright of it, if any syllable nucleus is lobe syllable onset and coda less than or equal to th syllable nucleus, which reference to the syllab rhyme of a heavy sylla rhyme of alight syllab a dominates the sylla (w). In both MSA and . is schematically illu 37 3.1.2.1. Rhyme projection: In the first step, rhyme projection, the rhymes in the syllables of a word are projected. "Projected" is used in the sense of selected, or identified. The selection of the rhymes is done on the basis of their having a well defined characteristic. McCarthy formally defined the rhyme as the right branch of co, the syllable root (McCarthy, 1979b: 7, 8, 42). The right branch of ex. dominates the syllable nucleus, which must be a vowel, and any other segments to the right of it, if any, which’could be a vowel or a consonant. The syllable nucleus is labeled strong (5), and the other segments, the syllable onset and coda, are labeled weak (w) since their sonority is less than or equal to that of the nucleus (McCarthy, 1979b: 7). The syllable nucleus, which is always a vowel, is labeled as strong in reference to the syllable onset or coda. 0n the other hand, the rhyme of a heavy syllable is labeled strong in reference to the rhyme of a light syllable which is labelled weak. The left branch of co dominates the syllable onset which must be l—syll] and is labeled (w). In both MSA and JA, syllables do not begin with a vowel. This is schematically illustrated in (1). The rhymes of a syll rhyme consists of a she W, or VC, or, word fin structures are the rhy The types of rhyme are (2. A): Week Rhyme C V (2.8): Strong Rhymes A. I I I 38 S A C V {V} c The rhymes of a syllable may be either weak or strong. A weak rhyme consists of a short vowel. 'A strong rhyme consists of either W, or VC, or, word finally either VVC or VCC. The latter structures are the rhymes of what are termed superheavy syllables. The types of-rhyme are illustrated in (2. A—C). (2. A): Weak Rhyme (2.8): Strong Rhymes --_ “"“~—-$ . : h' ‘ e '- new v c, Double rhymes appe 'superheavy syllables h second nonbranching. T with heavy penults and The types of rhym the rhymes in words 5 contrasts with jaarhur (3). (3): The third conson rhyme structures (2 problems for the co 39 (2C): Double Rhymes S w . C V V C, C Double rhymes appear in what is termed "superheavy syllables. "superheavy syllables have two rhymes, the first branching and the second nonbranching. This makes them formally equivalent to words with heavy penults and light ultimas...” (McCarthy, 1979b: 59) The types of rhyme postulated in (2.A—C) do not accommodate the rhymes in words such as jaarrhum “(he) pulling them" which contrasts with faarhum “their neighbour". This is illustrated in (3). .(3): sw. SW 1111 II illllcll jaarr um The third consonant of jaarrhum is not considered in the rhyme structures (2.A-C). We will see later that this creates problems for the correct asignment of stress in such words. fo it ‘ in the allows a great deal of pa respect to size, directioi assignment, and the kind In respect to size, f ei These types are schemal (McCarthy,1979b: 45). E Sillfibles (McCarthy, 19 (4A): binary 40 3.1.2.2. Foot assignment: In the second level, foot assignment, feet are assigned to the rhgmes identified in the first step. The foot assignment level allows a great deal of parametric variation. Here feet could differ in respect to size, direction of branching, direction of foot assignment, and the kind of head and tail the foot could have. In respect to size, feet are either binarg, ternary, or n—arg. These tgpes are schematicallg illustrated in (ALA-C) , respectivelg (McCarthg,1979b: 45). Binarg feet are assigned to pairs of light sgllables (McCarthg, 1979b: 55, 57, 64). (4A): binarg (B): ternarg (C): n-arg /\ In (4), n stands for a vowel or a consonant in the syllable rhgme. In reSpect to direction of branching, feet branch either to the left or to the right. (4. A—C) illustrate left branching feet and (4‘.A'—C') illustrate right—branching ones. In! 1 Jews emcee: «I: . It wit" .qua met an: at aeiiunsai zsmgm . /\ / 1'5,“ Owl-i .nollat1aV'3i'-3-:u - :"r' ':c if. :1:- i2'--2r_- -' :"-VE‘HL‘ --|1|. M “I iD-‘Ii‘3-1:.-'t3.I-'.'.‘:."‘,g:‘:f-""-.."-" Mm 1"" it...» Feetcan differin rI couId be l-head, -tall], | Most Is [+headl if Its l branching. This means rhgme of a heavy sgllal the difierent tgpes of I (5A): l-head, ~taill (5.8): l+head, ~taiIl /< 4i (4'.A‘): binarg (B'): ternarg (C'): n~arg n1 n2 n1/n2>\n3 n1 -“ m Feet can differ in respect to the features of [head, tail]. A foot could be [-head, -tail], [+head, —tail], [~head, +tail], or [+head, +tail], A foot is [+head] if its head is branching, and [+tail] if its tail is branching. This means that a branching head or tail must be the rhyme of a heavg sgllable. In (5.A~D) we schematicallg illustrate the different tgpes of branching a binarg foot mag have. (5A): l-head, -’tall] (5.8): [+head, —tail] Binarg feet which or toternarg feet. This s1 Accordinglg, feet have However, in his analgs and assumed that tern represent a marked tg -taill are considered l 3.1.2.3. Building the n In the third level gather the feet assic strag rhgmes which rhome Is the light 0| 42 (5C): [-head, +tail] (5D): [+head, +tail] Binarg feet which are [+head, —tail] or I-head, +tail] are identical to ternarg feet. This subsumes the need for a ternarg foot categorg. Accordinglg, feet have to be classified as either binarg or n—arg. However, in his analysis, McCarthy maintained three tgpes of feet and assumed that ternarg feet are derivative in the sense that theg represent a marked tgpe of binary feet. Binarg feet that are I—head, V-tail] are considered more basic and represent the unmarked form. 3.l.2.3. Building the metrical tree superstructure: In the third level, the building of the tree superstructure, we gather the feet assigned to the rhgmes in the earlier level and the strag rhgmes which are not assigned to feet. An example of a strag rhgme is the light one which is not assigned to feet in (6). I I cv.cv In (6), the penultima. binarg feet which are at left to right. It cannot because the ultimate is this ultimate branching that word-final heave; 1 Accordinglg, the rhyme In building the tree know whether the tree Illustrates a Ieit-brar Which are assigned f r alight-branching one (7A): left-branching 43 In (6), the penultimate is a strag rhgme in a language that has binarg feet which are assigned over pairs of light sgllables from left to right. It cannot pair with the ultimate to form a binary foot because the ultimate is a branching rhgme of a heavg sgllable. But this ultimate branching rhgme is labeled weak. McCarthg assumes that word—final heavg sgllables function as light sgllables. Accordinglg, the rhgme of a heavy sgllable is labeled weak. In building the tree superstructure at the word level, we have to know whether the tree branches to the right or to the left. (7A) illustrates a left-branching tree of a word which has binarg feet which are assigned from the left to the right, while (7.8) illustrates a right—branching one. (7A): left—branching (Till: right-branching C l The choice between This means that the chc choice between tgpes 0 has n-arg feel if it allt number of sgllables as (littering, 1979a; I97 assignment of stress I number of sgllables a stress to the initial 3 Arabic. Knowing whe 44 (7.8): right—branching . _ s w C Vl CJ CJ C VI A The choice between right and left branching is experimental. This means that the choice is based on trial and error. In respect to choice between tgpes of feet, it is possible to tell that a language has n—arg feet if it allows the retraction of stress to an unlimited number of sgllables as it has been assumed for Classical Arabic (McCarthg, 1979a; 1979b). This means that n—arg feet allow the assignment of stress to the initial sgllable, regardless of the number of sgllables a word has. This accounts for the assignment of stress to the initial sgllable in multi-sgllabic words in Classical Arabic. Knowing whether a language has binarg or ternarg feet as,for example, in JA and MSA where stress does not retract begond the antepenultimate is also a matter of experimental testing. In our analgsis of JA and MSA, we assume that their feet are binarg. This assumption is based on the fact that stress in these varieties does not retract begond the antepenultimate, and the assumption that ternarg feet represent a form which is derivative from binarg feet. SIM. Labeling: At the fourth level, I wehave built at the thir branching node and all tI (strong). All the otherr heterthg applies the Le which sags of a pair strong (3) if and onlg If he adds, "If the tree is terminal, the terminal will mark the stress-t "the designated termir sgllable dominated bg the trees in (TA-B). (M): 45 3.I.2.4. Labeling: At the fourth level, labeling, we label the tree superstructure we have built at the third level. In labelling, the right most branching node and all the nodes dominating it are labelled "s" (strong). All the other nodes are labelled "w" (weak). In labeling, McCarthg applies the Lexical Categorg Prominence Rule (LCPR) which sags of a pair of sister nodes, the right one is labeled strong (3) if and onlg if it branches". (McCarthg, 1979b: 48). Then, he adds, "If the tree is assembled in this wag, the designated terminal, the terminal node of the tree that is dominated onlg bg s's, will mark the stress—bearing sgllable." (McCarthg, l979b: 56). By "the designated terminal element" he means the nucleus of the sgllable dominated bg s‘s. McCarthg, 1979b: 7). In (BA—B), we label the trees in (7.A-B). (8A): IicCarthg's analgsi explicit. He assumes tI light one. Accordinglg, labeled weak. This ra weak rhgmes at the To Given a word which he assuming that it has t right, we would assig (9.8). (9A): 46 (8.5): /w\ s/ \W I i | i i C V C V C V C V C McCarthg's analgsis of heavg rhgmes word—finallg is not explicit. He assumes that a word-final heavg sgllable functions as a light one. Accordinglg, the rhgme of a word—final heavg sgllable is labeled weak. This raises the question of whether theg function as weak rhgmes at the foot assignment level or at the labeling level. Given a word which has a CVC. CV. CV. CV. CVC structure, and assuming that it has binarg feet that are assigned from left to right, we would assign it either the foot structure in (9A) or that in (9.8). (9A): In (9.B), the penulti In (8A) they do not has light rhgme at the lab -_ Accordinglg, this inex different metrical tre two different positio branching metrical tre (9.A‘-B'). 47 (9.8): /\W _/\ | CV CVVC <____...U) In (9.8), the penultimate and the ultimate form a binarg foot, but in (BA) theg do not because the ultimate is assumed to function as a light rhgme at the labelling level, not the foot assignment one. Accordinglg, this inexplicitness will allow the building of two different metrical trees, and finallg the assignment of stress to two different positions. This problem arises in building left branching metrical tree to the structures in (9.A-8) as illustrated in A(9.A'-8'). /SiI|V 48 (9.A'): (9.8'): immored‘bg the s there'tliages l982: 22 constituent such as rhg domain of the stress Following, we conside metrical model of Bru Is similar to McCarthy extrametricalitg. In 0 MSA, we will see that llctarthg‘s assumptio sgllable functions as rhgmes. In reference 4229) sags: word-ii the hierarchg of sglla heavg, while heavg s‘ will be seen to have for the JA and MSA c metrical parameters 3. 2. 1. Stress assig In this section, a sets of data in (I0. structures and vary 49 3.I.2.S. The principle of extrametricality: "In the metrical theory of stress , a syllable is extrametrical if it is ignored by the stress rules,- that is, treated as if it were not there." (Hayes I982: 227). For Hayes, a single phonological constituent such as rhyme, segment, consonant, or suffix in the domain of the stress rules can be extrametrical (Hayes, 1982:228). Following, we consider the assignment of stress in reference to the metrical model of Bruce Hayes (I980; I982,- I985). Hayes's model is similar to McCarthy's except for the principle of extrametricality. In our analysis of stress assignment in JA and MSA, we will see that this principle comes as an alternative to McCarthy's assumption according to which the rhyme of the ultimate syllable functions as light and that superheavy syllables have double rhymes. In reference to word—final syllables in CA, Hayes (I982: 229) says: word—final syllables are demoted one position down the hierarchy of syllable weight: superheavy syllables are treated as heavy, while heavy syllables are treated as light." These principles will be seen to have a significant role in building the metrical trees for the JA and MSA data, and will be considered one of the different metrical parameters. 3.2. 1. Stress assignment in MSA (Linear Approach): In this section, we consider the assignment of stress to the MSA sets of data in (IO.A—D) which include MSA words of varying structures and varying number of syllabic segments. '1- I "arm to am} Nahum-nau- ~ v ' . . ‘ ~ T.» s - air—i ' I _ D0 ”5;.” i541 :ggrm 329112 9:11 yd banana! a! H N 9 '--mee’*’abaut' ”(HWY)”, fl ” ' . _ n: I” If? _ 9'1;- 3,-3.- .'s.'h.-5H.'- "2'1er galti 'heart' unfrifll'cfio’i': 31.7.7 3 1'“.- _ . . _ -- .. '.:-:-:2;.=:J;".=.'3 laa 'no' I” L ..l'-."1":‘ -' i - . laar 'neighbo .U '-" i H . . I Iaarr 'prepos , . I "J. (l0.B):BI-sgllablc wor llma 'wh balad "co k'aatib “W harsbaat 's taasdun 'dl jawaad "h maqarr "I rihlah dar'ast “ minsaar (l0.C): Tri-sgllabic kataba salaraat madrasah 50 (lO.A):Mono—syllabic vrords we 'and" can “about" qalb “heart" laa "no” jaar "neighbor" jaarr "preposiUonal,puHing" (IO.8): Bi—syllabic words lima "why" balad "country" kéatib 'eriter“ harsbaat "spears" taasdun “disease" jawaad "horse" rnaqarr "headquarters" rihlah "journey" darast ”Istudied" nfinsaar "saw” (l0.C): Tri-syllabic words kataba "heevrote" sajaraat “trees“ inadrasah "school" musaawaah e Jaaréatun 'n mumtaazun 'e mustaqbal (l0.D): Words of four a katabaka kat'abakum tajaratuka maml'akatun mustasmaratiy mustasmaratul mazrasatukum faaeidatun saasadatka mustawdasun mustawdacukt mustawdasaa‘ mustawdacaa mustawdasaa makaatibuhur kitaabaahume 51 maktabun "office" tealibun "student" t'alabun "application" mak'aatib "desks" musaawaah "equality" jaaraatun "neighbors, f." mumt'aazun "excellent" must’aqbal "future" (10.0): Words of four or more syllabic segments katabaka "he wrote you" katabakum "he wrote you (plural)" sajaratuka “your tree" mamlakatun "kingdom" mustacmaratiy “my colony" mustasmarétuka "your colony" mazrasatukum “your farm" faael’datun "benefit" saasadétka "she helped you" mustéwdasun “a store" mustawdasuka "your store" mustawdasaat "stores" mustawdasaatun "stores" mustawdasaatuka "your stores" makaatl'buhum "their offices" kitaabaahuma "their (dual) book" Some of the and some of ti example, CVCC, CWC,‘ These tgpes are identic In his studg of the split CWCI: tgpe, they are a? (197%) in his study off In (10.8), the word: ultimate vowel if it is followed bg a consonai penultimate. In (1012] to the ultimate if it i‘ not It is assigned to t consonant cluster; at In (MD). the WOFI 52 The mono—syllabic words of (10A) illustrate the types of mono—syllables that are allowed in SA: CV, CVC, CVCC, CVV, CVVC, and CVVCC. Some of these syllables are restricted to specific contexts, and some of them are more frequent than the others. For example, CVCC, CVVC, and CVVCC syllables occur only word finally. These types are identical to those postulated by Al—An’i (1978: 117) in his study of the syllable structure in Arabic. Except for the CVVCC type, they are also identical to those analysed by McCarthy (I979b) in his study of Clasical Arabic. In (10.8), the words are bi—syllabic. Stress is assigned to the ultimate vowel if it is long and followed by a consonant or short but followed by a consonant cluster,- otherwise stress is assigned to the penultimate. In (l0.C), the words are trisyllabic. Stress is‘assigned to the ultimate if it is long or followed by a consonant cluster,- if not it is assigned to the penultimate if it is long or followed by a consonant cluster; otherwise it is assigned to the antepenultimate. In (10.0), the words have more than three syllablic segments. Considering the data in (10.0), we see that stress is never assigned to the preantepenultimate or any other syllabic segment preceding it. In other words, stress does not go beyond the antepenultimate no matter what the quality of the vowels and the following segment is. Stress assignment for the monosyllabic words in (10A) can be accounted for by rule (I I). (H): V ___, [ + stress ]/ ___ C0 ii V___,[+stressl/. The words in (10.8) 9i role(l2) makes the we precludes that of rule l the ultimate in *kaati‘a structures as such, we . Such a constraint is Halle,1968: 29-30). T Inclusion Precedence applies if the SD of a IUIe takes precedenci between rules (I I) a collapsed as (13). (13): V --~> i i‘ Sit" 53 Stress assignment for the bisyllabic words in (I 1.8) can be accounted for by rule (12). (l2): I V ___, t + stress I / ___ COV E, ii The words in (10.8) satisfy the SD of both rules(l I) and (12). But rule (12) makes the wrong prediction unless its application precludes that of rule (I 1). Rule (1 I) incorrectly assigns stress to the ultimate in *kaati’b "writer." To avoid getting ill—formed structures as such, we have to constrain the application of rule (I I) . Such a constraint is the principle of rule collapsing (Chomsky and Halle,1968: 29-30). This is consistent with the principle of Proper Inclusion Precedence (Koutsoudas et al, 1974: 8). This principle applies if the SD of a rule is included in another rule; the including rule takes precedence. This is the kind of relation that exists between rules (I l) and (12) which makes it possible for them to be collapsed as (I3). (13): I V ___,[+stress]/ ---CG(VCd) it It has to be noted that long vowels are considered sequences of two short vowels. In (10C), the words consist of trisyllabic words which show that stress falls on the antepenultimate if neither the ultimate nor the penultimate is along vowel or followed by a consonant cluster. V ___, [+ stress But rule (13) rather penultimate In words I the principles of Propl and rules (13) and (14 which apply disjuncti (15): V ___,[ + stre Rule (15) assigns long vowel. In actua long vowel make one and W' are equivale equivalent to laz'l. I vowel of a phonolog Rule (IS) is ider 54 Rule (14) accounts for this. (I4): I I V ___, [+stress] /___CVCVC it 0 5’ But rule (13) rather than (14) incorrectly assigns stress to the penultimate in words such as *madrésah "school." Again we invoke the principles of Proper Inclusion Precedence, and rule collapsing, and rules (13) and (14) can be collapsed into one rule the parts of which apply disjunctively: rule (15). (IS): i ' I V ___,[ + stressl/ ___C0((VC) VCg) it Rule (15) assigns stress to the second of two short vowels of a long vowel. In actuality, the two short vowels of a phonological long vowel make one unanalyzable phonetic unit. That is both V’V and W‘ are equivalent to [Vz’]. This means that 633. and _a_a are equivalent to [of]. In writing data, I always mark stress on the first vowel of a phonological long vowel. Rule (15) is identical with the rule which Brame (I971: 560) formulated for what he calls Classical Arabic in his review of the stress rules postulated by Abdo(1969) to be the general rule of stress assignment in MSA. Here, it has to be noted that vowels which are not assigned stress by the stress rules and are not marked [+stress} are considered I—stress]. _7, \_ d3“ Some of thes These data will be disq data In reference to ep Problematic data in re, Chapter 7. In this chal along vowel and those In MSA, long vowe dual morpheme. Exam (dual)“ and mazrasal resleet to mazraeatl sleakers of MSA tent words. Itis only in: long, and In this C&SI 55 Some problematic data to rule (l5): There are some data which are potential counter evidence to rule (IS). Some of these data involve epenthesis and gemination. These data will be discussed in the following chapters. Problematic data in reference to epenthesis will be discussed in Chapter 5. Problematic data in reference to geminates will be discussed in Chapter 7. In this chapter, we will deal with words which end with a long vowel and those which end with CVCVC. In MSA, long vowels show up word—finally as the marker of the dual morpheme. Examples of this are rkatab+aa "(they) wrote—they (dual)“ and mazrasat+u+humaa "farm-nom.-their (dual)". In respect to mazrasatuhuma(a), it has to be pointed out that speakers of MSA tend to shorten long vowels word—finally in such words. It is only in a careful speech style that they pronounce them long, and in this case the final long vowel receives stress: mazrasatuhuméa. More typically the vowel is shortened and stress follows rule (15): mazrasatdhuma. On these grounds, we can say that such words are not counter evidence to the general rule of stress assignment. Final vowel shortening is accounted for by rule (to). Rule(16): VV ———-> V / _# Rule (16) applies to words such as mazrasatuhumaa "their (dual) farm", yielding mazrasatuhuma and the application of rule 'I' '5': ' . -" " .- P a“ ~ '91: fl-GMW am emu m fluinndfiiag has ziasdmsqs s-riavm me 9.;- ad! emo‘e (Bi) . Wannatrg in the ii E alfanwidm‘l 21-39;;an 31-w- ' .-._: . a '_ u fr'.'- e151: “MT .T' -._- z. a ' .: -. .- - - :‘ .- ' shah antheonehand.ond m unnamemmoifi ultimate in katabéa, I constrained. This is a Rule (17): W] dual verl Rule (17) properli mmmmn ' ‘f rule (IS) oi str alJlIlication of rule ( 56 (15) correctly assigns stress to the antepenultimate. But a problem arises in verbs such as katabaa "they (dual) wrote". The final vowel of katabaa "they (dual) wrote", is not sensitive to stylistic variations,- it is pronounced long and is assigned stress. The asymmetry in the behaviour of the ultimate vowels in katabaa on the one hand, and mazrasatdhumaaa) , on the other, is due to the fact that the rule of final vowel shortening does not apply to the ultimate in katabaa, apparently because it is morphologically constrained. This is accounted for by rule (17). Rule (17): W] dual verb —“’> W / —-— # Rule (17) properly includes rule (16) and applies disjunctively in relation to it. If rule (l5) of stress assignment applied to katabaa after the application of rule (16) of vowel shortening, the result would be kétaba, which is homophonous with kétaba “(he) wrote". But rule (16) of ultimate vowel shortening does not apply to katabaa because rule (17) properly includes it and takes precedence. Words like katabéa are consistent with rule (15). The shortened dual of verbs such as katabaa would typically be homophonous with non-duals so rule (17) preserves a morphological distinction. Such homophonies do not arise in the case of the shortened dual of nouns. reference to two metn uncaring (1979a; 1979' I985). In metrical phonoiu' relation among syllabl within the framework the metrical structur variations which may order to build the me first to identify thus that language. "mariners mode basis M the Darame loa. Feet are binar I. left to righ tau. Metricai ”-95 ii. fight bra] 57 3.2.2. Stress assignment in MSA (Non—linear Approach): Now we consider the assignment of stress in MSA within the framework of metrical phonology. Here the analysis will be done in reference to two metrical models. The first model is developed by McCarthy (1979a,- 1979b), and the second by Hayes (1980; 1982,- 1985). In metrical phonology, stress is identified as a prominence relation among syllables (Hayes, 1985: 4). This relation is defined within the framework of a metrical tree structure. The building of the metrical structure is determined by different parametric variations which may differ from one language to another. Thus in order to build the metrical trees of a particular language, we have first to identify those parameters which fit into the analysis of that language. McCarthy's model allows the building of metrical trees on the basis of the parameters in (18.a—d). 18a. Feet are binary and are assigned from i. left to right. ii. right to left. 18b. Metrical trees are i. left branching. ii. right branching. 57 level The parameters in ( different metri cal tre the parameters in (18} in (19). (19): a. katabaka Ii. katabakum c. mazracatuka d. mazras‘atukur e. 'sajar'atuka f. 'salaratukum The possible cho ”Mr“. H 18. Wit-11mg! M 18 Each of the eight c the incorrecg as“ (19). Under the assu Iollowlng mama assignment 0f st. 58 18c. The ultimate syllable functions as alight one at the i.footlevel. ii.labelling level. The parameters in (18) allow the possibility of building 8 different metrical trees for each word (2x2x2:8). Now, we apply the parameters in (18) to building metrical trees for the MSA words in(19). (19): a. katabaka "he wrote you" b. katabakum "he wrote you (plural)" c. mazrasatuka "your farm" d. mazrasatukum "your farm (plural)" e. éajaratuka "your tree" f. salaratukum "your tree (plural)" The possible choices among the parameters in (18) are: 18.8-1, b-l', C-l' 18.8-1", b-l, C-ll' 18.8-1, b-ll, C-l 18.8-l, b—ll, C-l'l' 18.8-ll, b-l, C-l' 18.8-11, b-l, C-il 18.8-l'l', b-ll, C-l 188-11, b-l'l', C-li Each of the eight choices among the parameters in (18) results in the incorrect assignment of stress to one or more of the words in (19). Under the assumptions of (18.a-i, b—i, c—i), construction of the following metrical trees in (20.a—d) results in the wrong assignment of stress. (20o): 59 (20.8): (20D): w /\ >¥éktl *kaltalb (20c): (20d): w w s W/\ s W/\W \W w .1.) /\ /\/\51\ llllililllil *ma zrasatukaa’ema z um rasa k The correct stress assignment to these words is: katabaka, katabakum, mazrasétuka, and mazrasétukum. Under the assumptions of (18.a—i, b-i, c—ii), construction of the following metrical trees in (Zia—d) results in the wrong assignment of stress. (21o): 60 (21.81): (21h): (21.C): (2l.d): s / /W\ /\ w w s w s w \ \ /\ \ L\ I\ s w . s, w s w . . s W , l l I I | l , l l l l l I *‘ma zraqatuka *ma 2 qatukum The correct stress assignment to these words is: katébaka, katabakum, mazrasétuka, and mazrasétukum. Under the assumptions of (18.a—i, b—ii, C—i), construction of the following metrical trees in (22.a—d) results in the wrong assignment of stress. (22.c): 61 (22.8): (22b): w\ s W \W . /\W ./\. /\w *k 8| t 8| k 8| *k 6| t 8| 1) 8| nl’l (22C): (22d): /\ \ /\ \ W w s W w W 3/\W \ é\. s/\w /\. S s\W a‘émtla lralqal kal *mIa iralsal kulrln The correct stress assignment to these words is: katébaka, katébakum, mazrasétuka, and mazrasétukum. Under the assumptions of (18.a—i, b—ii, c—ii), construction of the following metrical trees in (23a—d) results in the wrong assignment of stress. (23o): 62 (23.8): (23D): (23C): (23.d): V s s \ s w w s w w \w /\ \ 4t \ /\ SI\ 3 w . . s w . . . s W *mL lrlqal kal 7kml lszJtulktlurL The correct stress assignment to these words is: katabaka, katabakum, mazrasétuka, and mazraqatukum. Under the assumptions of (18.a—ii, b-i, c-i), construction of the following metrical trees in (24.a—f) results in the wrong assignment of stress. 63 Ill \ Wlll m Will: 3'11 U (l .\ 0 S\ / / SIU .16 w\ 0 1r .. .1 \fl W Z \% W/ .1 A44 / 1.8 2 .la /\ s e e .la \ \S/IS k I. s 0 S\ . (24a): (24C); W\W ./\. /. *‘s l 1 gr; The correct stress a kalabakum, mazraeétu iajarétukum. Under the assumptic following metrical trel assignment of stress. (24.8): (241‘): an 01—. (3)—— '1 C9—- ,4. C ma 0...- \-—-( _‘ 7? ~16- The correct stress assignment to these words is: katabaka, katabakum, mazrasétuka, mazracétukum, s'ajaratuka, and safe/”atukum. Under the assumptions of (18.a—ii, b-i, c—ii), construction of the following metrical trees in (25.a,c,e,f) results in the wrong assignment of stress. following metrical t 65 (25a): (25c): \ I 1 51 TT I I' I jri{kata kajémazrasa ka (25.e): (25f) /\ . W /\ \ 3 W W W S /\ /\ /\ /\ /\ . S W . . ll *saljalral kaI 328MB tquLErll The correct stress assignment to these words is: katébaka, mazrasatuka, safarétuka, and s'ajarétukum. Under the assumptions of (18.a—ii, b-ii, c—i), construction of the following metrical trees in (26.a-f) results in the wrong assignment (26c): 66 of stress. (26.13): W118 \ S/ k 818 .D \ ..18 (26.8): M II (26.C): ..llli 'sajaratu The correct stress a katébakum, mazresétu. iajarétukum. Under the assumptio following metrical trel assignment of stress. 67 (26.9): (26f): g1 sa k—( m— 1 m 7? CD me O)— L—( CD— .1 CD 7? The correct stress assignment to these words is: katébaka, katabakum, mazrasatuka, mazrasétukum, safarétuka, and safaratukum. Under the assumptions of (18.a—ii, b—i, c-ii), construction of the following metrical trees in (27.a,c,e) results in the wrong assignment of stress. 68 (278): (27c): w 3 w 8 /\ g \ (Wt / \ /S \ I ' I *katell klak'mllzrlselatukal (27.9): w/ 3 / \ é \ The correct stress assignment to these words is: katébaka, mazracétuka, and s'ajarétuka. It is clear that all the possible choices among the parametric variation provided in McCarthy‘s model fail to make the correct prediction. On these grounds, we conclude that McCarthy's model in its 1979 form does not fit the analysis of stress in MSA. Accordingly, it has to be modified. i.2.3.Hages' metrical mode In this section, we recl themetrical model of Hay principle of extrametrica‘ of which the metrical tre parametric variations in 28a. Feet are binary and i.left to right. ii. right to left. 2u.b.tletrical trees are i. left branching. ii. right branching 28.c.The final syllable The parameters in different metrical tre the parameters in (28 1“Wieneeted here (19). a. katébakfi be kat'abakum c. mozraqaium d“ mazresatuktn 69 3.2.3. Hayes' metrical model: In this section, we reconsider McCarthy's model in the light of the metrical model of Hayes (1980; 1982-1985) who proposed the principle of extrametricality as one of the parameters on the basis of which the metrical trees are built. Accordingly, we modify the parametric variations in (18) into (28). 28a. Feet are binary and are assigned from i. left to right. ii. right to left. 28b. Metrical trees are i. left branching. ii. right branching. 28c. The final syllable is extametrical The parameters in (28) allow the possibility of building 4 different metrical trees for each word (2x2xlz4). Now, we apply the parameters in (28) to building metrical trees for the MSA words in (19), repeated here for convenience. (19): a. katabaka "he wrote you" b. kat’abakum "he wrote you (plural)" c. mazrasatuka "your farm" d. mazrasatukum "your farm (plural)" etaiar'atuka U0” f'sajaratukum “you The possible choices a 28.a-i,b-i,c 28.a—i,b- Two of the four choices incorrect assignment of The other two of the fat of stress. Under the assumptio following metrical tree assignment of stress. (29.8): A /$\ )(1 tal 70 e. sajaratuka "your tree" f. sajar'atukum "your tree (plural)” The possible choices among the parameters in (18) are: 28.a—i, b-i, c 28.a—i, b—ii, c 28.a-ii, b—i, c 28.a—ii, b—ii, c Two of the four choices among the parameters in (28) result in the incorrect assignment of stress to one or more of the words in (19). The other two of the four choices result in the correct assignment of stress. Under the assumptions of (28.a—i, b—i, c), construction of the following metrical trees in (20.a—d) results in the wrong assignment of stress. (29.6): (29p); S /S\ s . . Q . . I _ l l JKGtaleka) X.taba(kum) iii *ma 2 C0 The correct stress mozrasétuka, and Ea jar Under the assumptio following metrical tree assignment of stress. (30a): 71 (29C): (29d): /8\ /W\ S\ T 1 4 i ' *ma 2 98 tt)(ka) The correct stress assignment to these words is: katabaka, mazrasétuka, and safaratuka. Under the assumptions of (28.a—i, b—ii, c), construction of the following metrical trees in (30.a—d) results in the wrong assignment of stress. (30.8): (30.13): s/\ s e. . A I I %.ta ba(ka) ta ba(kum) (The): S s/\ S/\ U l I )(I so ma The correct stress a mazraeaiuka, and sajarl Construction of the r (26.a~ii, b~i, c), and (28 assignment of stress. 1 overgenerates. More th We conclude that till MSA yields the correct are binary and that the analysis dealing with t SUtterlieavy syllables v furnished in the follov analysis of Uh 72 (30C): (3011): / S w /\ /\ T): T 1 ma so tu(I \cn mazrasétuka, and safarétuka. Construction of the metrical trees under the assumptions of both (28.a-ii, b—i, c), and (28.a—ii, b-ii, c) results in the correct assignment of stress. This means that the system is faulty; it still overgenerates. More than one choice works. 1 We conclude that the metrical approach to stress assignment in MSA yields the correct prediction under the assumptions that feet are binary and that the ultimate syllable is extrametrical. Further analysis dealing with the principle of extrametricality in reSpect to superheavy syllables word—finally and word—medially will be furnished in the following sections of the chapter, mainly in the analysis of JA. 3.2.4. Stress assignment Inthis section, we Cf sets of data in (31A-D). varying structures and v assignment to these dat STGP. Consider the JA c (31.1)): Mono-syllabic w We "and" laa "no“ can "abou‘ darb "road jaar “neig jaarr "prep TUUT "beca nhaar "day nharr "gya, nsadd “Na daWa m héIar s maéérr a (smear d 73 3.2. 4. Stress assignment in JA (Linear Approach): In this section, we consider the assignment of stress to the JA sets ofdata hi(3l.A—D). The datairi(3l.A-D)includes‘words of varying structures and varying nurnber of syllabic segrnents. Stress assignnaentto these dataowiH be consideredin reference tothe STGP. Consider the JA data in (31.A—D): (3I.A):Mono—synabicivords we "and" Tea "no" can "about" darb "road“ jaar ”neighbor" jaarr "prepositional,dragging" stigt “becarne eager" nhaar "day tune" nharr 'ans sleed" nsadd “was Ughtened" (31.8):8i—syllabicxvords davva 'hnedecine" hajar "stone" rnaéarr ‘traces ofpulHng" danaaar "destruction" mal'aan "NH“ mismaar “nail" Iamlam "gathe xt'alaf "diffel d’aawa "he ()5 x'aatim "W“ taaedun "(1135 starhfi “res ntaha "en tsaawa “be x'aalha "h' xaalhum "t1 xaayfaat “tI (31.C):Tri-syllabic w salarath) salaraat 74 malaan "full" mismaar "nail" lamlam "gathered“ xtalaf "differed" daawa "he gave medical treetment" xaa‘tim “wring" taasfuun "disease" starjas "restored" ntaha "ended" tsaawa "became equal" xaalha "her uncle" x'aalhum "their uncle' xaayfaat "they (fem) are scared" (31C): Tri—syllabic words sajara(h) "tree" sajaréat "trees" mistasjil "in a hurry" xaatimhum "their ring" msaawaraat "consultations" msaawamah "bargining" sawaalif "stories" mistasjil “in a hurry" mtlhtaram "respectable" mittpaayig "he is under pressure" (31.0):Words of more thal mistaemaratiy mizracatu(h) mizracathum mi'starayaat mazaariehum mistasmarahl) aistigbaalaathun The data in (31A) I allowed in JA: CV, CVC CCWC. Also notice th (seawa, provided in( restricted in their dis monosyllabic words or consonant clusters or with final clusters or CCVCC syllables occu are restricted to hav Considering the t assignment of Stress 75 (31.0): Words of more than three syllabic segments mistasmaratiy "my colony" mizrasatu(h) "his farm" mizrasathum "their farm" mistarayaat "purchases" mazaarishum "their farms" mistasmara(h) "colony" 71stigbaalaathum "theirreceptions" The data in (31.A) show the types of mono—syllable that are allowed in JA: CV, CVC, CVCC, CCVCC, CW, CVVC, CVVCC, and CCVVC. Also notice the CCV and CCVV syllable types of xtalaf and tsaa.wa, provided in (31.8). Some of these syllables are so restricted in their distribution that they appear only in monosyllabic words or initial position. Syllables with initial consonant clusters are restricted to word initial position and those with final clusters occur only in word final position. Therefore, CCVCC syllables occur only in monosyllabic words. CVVCC syllables are restricted to having a geminate cluster in the coda. Considering the JA data in (31. A-D ), we see that the assignment of stress follows the same pattern of MSA stress assignment in (lO.A-D ). Accordingly, we assume that stress assignment in both MSA and JA can be accounted for by rule (15). Here, it has to be pointed out that some of the problems which arise in the analysis of MSA data in reference to rule (15) do not arise in the analysis of JA. JA does not have long vowels word—finally. Long vowel: inilSA do not show UD in ~ conjugationlhe JA form (dua1)wrote" is kétabuw wrote". Nor do long vowels wl if); in JA the plural form it word corresponding t (dual) farm" is mizrasa plural) farm". In the JA the antepenultimate vo‘ Iost;JA has last case n As for epenthetic ' counter evidence to ru‘ interaction with stres following chapters. 3.2.5. Stress assignm In this section, in in (32.a~d) in referer illustrative and liml assignment in metri which take a considl 132.8.mizracah "fa b. mizras’athum 76 word-finally. Long vowels which mark verbs conjugated for the dual in MSA do not show up in JA, since JA has lost this type of conjugation. The JA form corresponding to MSA katabaa "they (dual) wrote" is kat‘abuw "they (referring to both dual and plural) wrote". Nor do long vowels which mark MSA nouns for the dual show up in JA; in JA the plural form refers to both the dual and the plural. The JA word corresponding to the MSA word mazraqawhumam) "their (dual) farm" is mizraséthum "their (referring to both dual and plural) farm". In the JA word, the ultimate long vowel is short and the antepenultimate vowel ~u- which marks the noun for case is lost,- JA has lost case markers. As for epenthetic vowels and geminates, they raise potential counter evidence to rule (15). Their phonlogical status and interaction with stress assignment will be discussed in the following chapters. 3.2.5. Stress assignment in JA (Non-linear Approach): In this section, we consider stress assignment to the JA words in (32a-d) in reference to metrical phonology. Here we consider an illustrative and limited set of data because rules of stress assignment in metrical phonology are represented in metrical trees which take a considerable amount of space. (328. mizrasah "farm" b. mizrasathum "their farm" c. mistbodasuh "his 5 d. mistasmaratak “yl The parameters in (1 different metrical trees 4different trees for ea Under the assumptit metrical trees in (33a, stress. (33a): /\ / S w 5 gig I ml 2 77 c. mistoodasuh "his sore" d. mistaqmaratak "your colony" The parameters in (18) allow the possibility of building 8 different metrical trees for each word (2x2x2:8), and those in (28) 4 different trees for each word (2x2xl:4). Under the assumptions of (188—1, b—i, c—i), construction of the metrical trees in (33.8, c) results in the wrong assignment of stress. (33a): (33c): The correct stress assignment to these words is: mr’zrasah, and mistoodasuh. Construction of the metrical trees under the assumptions of (18.a~i, b-i, c-ii), results in the correct assignment of stress. Under the assumptions of (18.a—i, b—ii, c—i), construction of the metrical trees in (34.a,c) results in the wrong assignment of stress. The correct stress mistoodaeuh. Under the assumpl metrical trees in (35. (35a): 78 (34.8): (34C): The correct stress assignment to these words is: mr’zrasah, and misto’odacuh. Under the assumptions of (18.a-i, b-ii, c-ii), construction of the metrical trees in (35.a,c) results in the wrong assignment of stress. (35.8): (35C): \ /\ S s\ \\ w w w \IN /\ /\ /\ /\ .I S w ‘ W s T T T l T T 7¥'mli irel Ilifi'mli sto OQSU h . assumpti~ metrical trees in (36a, (36a): '5 w\ T/l S *m 1 2 q The correct stres mistdodaeuh. (18.a-ii,b~i, c-ii), r 79 The correct stress assignment to these words is: mr’zracah, and misto’odasuh. Under the assumptions of (18.a—ii, b—i, c—i), construction of the metrical trees in (36.a,c) results in the wrong assignment of stress. (36.8): (36b): w / s s\ /\ w w w w w /\ é\ /\ \ $/\ T i T T T l T T T T *mi 2 ea h 95ml sto 0 cu h The correct stress assignment to these words is: ml'zraqah, and misto’odacuh. Construction of the metrical trees under the assumptions of (18.a—ii, b—i, c—ii), results in the correct assignment of stress. Thus, construction of the metrical trees under the assumptions of both (18.a—i, b—i, c—ii) and (18.a—ii, b—i, c—ii) result in the correct assignment of stress. This means the system is faulty,- it still overgenerates. Under the assumptions of (18.a—ii, b-ii, c~i), construction of the metrical trees in (37.a,c) results in the wrong assignment of stress. The correct stress mistdodasuh. Under the assumpi the metrical trees in stress. (38a): w /\ l T l l *m zra 80 (37.8): (37C) \ s s\ \ w w w w w /\ /\ /\ /\ $/\ T T T T T T T T l T T 7\‘Tmi z‘sa h'TTmi sto odasu h The correct stress assignment to these words is: mizrasah, and misto’odasuh. Under the assumptions of (18.a—ii, b-ii, c—ii), construction of the metrical trees in (38a, c) results in the wrong assignment of ~stress. (38.8): (38C): The correct stress a mistoodaeun. Construction of the I both (Ida-i, b-i, c-ii) a assignment of stress. T overgenerates. Accordi Hayes‘ model and stres Now, we consider t the parameters in (28'. variations in (28) yiel (2x2x1:4). Construct assumptions of all of assignment. Again,t overgenerates. All 0 possible choices oft correct prediction. ' foot assignment nor the correct predicti to direction of foot equivalently. That insignificant for Jr The assumptio ((1098 not count in assigned to the De antepenultimate i‘ 81 The correct stress assignment to these words is: mr’zrasah, and misto’odasuh. Construction of the metrical trees under the assumptions of both (18.a-i, b-i, c—ii) and (18.a—ii, b—i, c—ii) result in the correct assignment of stress. This means the system is faulty; it still overgenerates. Accordingly, it has to be modified. Hayes' model and stress assignment in JA (non-linear approach): Now, we consider the assignment of stress in accordance with the parameters in (28). Here the possible choices of the parametric variations in (28) yield four different metrical trees for each word (2x2xlz4). Construction of the metrical trees under the assumptions of all of these choices results in the correct stress assignment. Again, this shows that the system is faulty; it still .overgenerates. All of the possible choices work. We see that all the possible choices of the parametric variations in (28) yield the correct prediction. This gives evidence that neither the direction of foot assignment nor that of tree branching play any role in getting the correct prediction. That is, the parametric variations in respect to direction of foot assignment and tree branching apply equivalently. That is, distinctions within these parameters is insignificant for JA. The assumption that the ultimate syllable is extrametrical (does not count in stress assignment) means that stress has to be assigned to the penultimate if it is a heavy syllable or to the antepenultimate if it is heavy. If both the penultimate and the antepenultimate are ligh left branch of which is t will arise if both the an arelight; the rhyme of e the right branch of whit stress will be incorrect JAdoes not have lights problem will not arise. have a sequence of mor as katabaka "(he) in that binary feet are as right to the left. 3.2.6. Superheavy Sylli MCCarthy (1979b. . m (39), 82 antepenultimate are light, they will be assigned a binary foot the left branch of which is to be labelled "strong". However, a problem will arise if both the antepenultimate and the preantepenultimate are light,- the rhyme of each syllable will be assigned a binary foot the right branch of which has to be labeled "strong'. Accordingly, stress will be incorrectly assigned to the preantepenultimate. But JA does not have light syllables in these positions. Thus this problem will not arise. This problem arises in MSA where you can have a sequence of more than two light syllables. MSA words such as ka.té.ba.ka "(he) wrote you" raise a problem unless we stipulate that binary feet are assigned over pairs of light syllables from the right to the left. 3.2.6. Superheavy Syllables and Stress Assignment: McCarthy (1979b: ll) assigns superheavy syllables the structure in (39). In (39), the 111181 preceding syllable W‘ heavy syllable struci syllable, presentsits phenomena l superordinar outtohavei accentualpl IU'(39)“yie 01 en‘with non-branch 83 (39): ./S w ./\. A. l I ('7 In (39), the final consonant is Chomsky—adjoined to the preceding syllable with the indicated structure which represents a heavy syllable structure. McCarthy claims that a superheavy syllable, presents itself as to surface representation and phenomena like meter as a single syllable, the superordinant es in "(39)“ and this property turns out to have no little significance in the operation of accentual processes... Specifically, the two m-nodes in "(39)" yield two rhymes, if defined as the right branch of co, with the first rhyme branching and the second non—branching(McCarthy, 1979b211). The first branching rhyi second rhyme dominates Chomsky-adjoined to th are assumed to be form and light ultimas (McCt superheavy syllable is the penultimate "stron This assumption mt superheavy syllables a into the analysis of a syllables appear both hot analyze the struc Nor did he analyze CV With) Other Arabil suDer-Stiperheavy so In line with McCarth Word-ilnally, we m1 have T ”We rhyme. application of Chom 84 The first branching rhyme in (39) dominates W or VC and the second rhyme dominates C, the final consonant which is Chomsky-adjoined to the preceding syllable. Superheavy syllables are assumed to be formally equivalent to words with heavy penults and light ultimas (McCarthy, 1979b: 59). The ultimate rhyme of the superheavy syllable is labeled "weak" since it is non—branching and the penultimate "strong" since it is branching. This assumption may fit into the analysis of MSA where superheavy syllables appear only word—finally. But it does not fit into the analysis of a language such as JA where superheavy syllables appear both word—finally and word—medially. McCarthy did not analyze the structure of superheavy syllables word—medially. Nor did he analyze CVVCC syllable structures which appear in JA and probably other Arabic dialects. We might call this type of syllable super—superheavy so as to distinguish it from superheavy syllables. .In line with McCarthy's analysis of superheavy syllables word-finally, we might assume that super-superheavy syllables have a triple rhyme. Under this assumption we have to extend the application of Chomsky-adjunction to word—medial positions. We also have to assume that the final segment in super-superheavy syllables is Chomsky—adjoined to a preceding superheavy syllable in which the preceding consonant is also Chomsky-adjoined to the preceding heavy syllable. This is illustrated in (40). (e10): In (40), under the superheavy syllables segments in a super- establish a binary fo then we will have to syllable structure it binary fact. In this syllable which is a phonology where stl syllables, not parts assumption that th Dresents itself as ‘ the STGP model. (40): S/OV\W m / S/ \W Ov/ w/ >s\w I I I c v {2?} c c In (40), under the assumption that the ultimate segment in superheavy syllables functions as a light syllable, the ultimate two segments in a super—superheavy syllable could be paired toyather to establish a binary foot. If this type of analysis is to be accepted, then we will have to assign the last two segments of the CVVCC syllable structure in the JA word jaarr.hum "(he) dragging them" a binary foot. In this case, stress will be assigned to part of the syllable which is a violation of the basic principles of metrical phonology where stress is viewed as a relation of prominence among syllables, not parts of syllables. This also contradicts McCarthy's assumption that the janus-like character of superheavy syllables presents itself as a single syllable. None of this is a problem with the STGP model. 3,3.I.Evaluation of the I According to the me on the identification of evaluation of the metr' according to which the the capacity of the me on the other. The principles on t include the projectim direction of branchin and the Principle of l The rhyme projec rhymes of the syllat structures and boun “0i Drovide us with US With some Syllal Suitabification. Th Of a ”“9“598 Whic onset, and Wthh n SlIIIable coda. The constitute the Syl 86 3.3.1. Evaluation of the metrical approach to stress: According to the metrical approach, stress assignment is based on the identification of the most prominent syllable. In our evaluation of the metrical approach, we will examine the principles according to which the metrical trees are built, on the one hand, and the capacity of the metrical model to make the correct predictions, on the other. The principles on the basis of which the metrical trees are built include the projection principle, the assignment of feet, the direction of branching at both foot and tree superstructure levels, and the principle of extrametricality. The rhyme projection principle involves the recognition of the rhymes of the syllables in a word. This implies that the syllable structures and boundaries are already identified. But McCarthy did .not provide us with syllable boundary rules. However, he provided us with some syllable tree structures which he considers rules of syllabification. These syllable tree structures fit into the analysis of a language which has one consonantal segment as a syllable onset, and which might have one consonantal or vocalic segment as a syllable coda. The coda and the preceding vocalic segment constitute the syllable rhyme. These analyses account for the syllable structures CV, CW, and CVC. McCarthy also provided us with tree structures for CVVC and CVCC syllables word—finally in which the ultimate segment is Chomsky—adjoined to the node dominating the preceding syllable. CVVC, CVCC, and CVVCC, which show up word—medially in JA, are not accounted for. Thus we cannot tell whether the clusters as in the JA WI preceding or following structures are illustra structures are illustra not know whether the I word-medially or not. where the syllable str explicitly identified, rhymes. Now we examine t The assignment of fe the size of the foot, direction of foot ass variations involve sf Should not find in a as519M from the it Well. 0n the othe ‘tIpe of foot which -taill. There is 51$ “he“! +tail] binar to building the assumed that the 1 Metering did not; ”Emmet relate tt foot assignment c determine the Cor 87 cannot tell whether the ultimate consonantal segment in CVVCC clusters as in the JA word jaarrhum, syllabifies with the preceding or following vocalic segment. These types of syllable structures are illustrated in (39) and (40). CWC and CVCC syllable structures are illustrated in (39), and CVVCC in (40). We also do not know whether the principle of Chomsky-adjunction applies word-medially or not. Following his procedures, in the contexts where the syllable structures or the syllable boundaries are not explicitly identified, we will be unable to predict the syllable rhymes. Now we examine the assumptions related to foot assignment. The assignment of feet may involve several parametric variations: the size of the foot, the kind of head or tail it may have, the direction of foot assignment, and the direction of branching. These variations involve some redundancies and arbitrariness which we .should not find in a good model. It is arbitrary that feet are assigned from the left or the right, or that they branch to the right or left. On the other hand, it is redundant to postulate a ternary type of foot which is identical with a binary foot which is [+head, ~taiI]. There is also a kind of redundancy between identifying a [+head, +tail] binary foot and a pair of consequtive binary feet. In building the metrical tree superstructure, it is arbitrarily assumed that the trees branch either to the left or to the right. McCarthy did not provide us with a principle which might, for example, relate the direction of tree branching to the direction of foot assignment or foot branching. Such a principle would determine the correct parametric choices on the basis of which the L_'— metrical trees are built. Now we examine the McCarthy‘s model, the u' assumed to function as branching rhyme. McCai syllables function as li functions as alight syl will be capable of mak syllable. But if it funt then it will not pair w binary footThese twr building of two differ of stress to two diffs assumed to have dout penultimate is heavy consideration that St The fact that tic variations means th Drocess. As I Show: DrWillie have to be parametric choices model lacks 6 Nina aDittopriate perm the metrical trees Added to the pi fails to make the I illustrated b9 our 88 metrical trees are built. Now we examine the analysis of the ultimate syllables. In McCarthy's model, the ultimate heavy syllable is arbitrarily assumed to function as a light syllable even though it has a branching rhyme. McCarthy did not specify the level at which heavy syllables function as light ones. If the ultimate heavy syllable functions as a light syllable at the foot assignment level, then it will be capable of making a binary foot with the preceding light syllable. But if it functions as a light syllable at the labeling level, then it will not pair with the preceding light syllable to make a binary foot. These two possible interpretations will result in the building of two different metrical trees, and finally the assignment of stress to two different positions. Superheavy syllables are assumed to have double rhymes: the ultimate is light and the penultimate is heavy. Furthermore these analyses fail to take into consideration that superheavy syllables show up word—medially. The fact that McCarthy's model allows too many parametric variations means that the building of metrical trees is not a simple process. As I showed above, many attempts which lack a guiding principle have to be made in order to identify the appropriate parametric choices which yield the correct prediction. McCarthy’s model lacks a principle on the basis if which we can choose the appropriate parameters when approaching the process of building the metrical trees. Added to the problems we have pointed out, McCarthy's model fails to make the correct prediction in the analysis of MSA. This is illustrated by our analysis of the MSA data in (I7) above. However, L— . -_-» itis possible to modify principle of extrametri assumption, McCarthy's illustrated in (40-43). against the principle 0 underlying motivation to the application of i‘ The variation int illustrated by applyin Arabic. In Cairene Ar oppliction of the prin make it possible to a which, according tor the rhyme of the ulti problem arises in th dialect spoken in Jo: penultimate of Wm tinksérat "she got the DrmClDie of ext commit to the fac‘ sea" is assigned 81 other Arabic vartep extrametrical; 0th We“ Drediction To sum “D, the 89 it is possible to modify McCarthy's model by introducing into it the principle of extrametricality (Hayes,1982,~ 1985). Under this assumption, McCarthy's model yields the correct prediction as illustrated in (40—43). However, some arguments can be raised against the principle of extrametricality. The first relates to the underlying motivation for this principle, and the second one relates to the application of it to more than one variety of Arabic. The variation in the application of this principle can be illustrated by applying it to a number of closely related varieties of Arabic. In Cairene Arabic, the principle does not apply. The appliction of the principle to words such as padwiyatu’hu will not make it possible to assign stress correctly to the penultimate which, according to McCarthy's analysis, forms a binary foot with the rhyme of the ultimate syllable (McCarthy, 1979b: 58). The same problem arises in the analysis of Bani Hasan Arabic, a Bedouin dialect spoken in Jordan. In this dialect, stress is assigned to the penultimate of words such as msikétak “she grabbed you", and cinksérat "she got broken" (Irshid, 1984: 41, 80). In Libyan Arabic, the principle of extrametricality cannot apply,- its application is contrary to the fact that the ultimate in paréb "he hit" and bahar "sea" is assigned stress (Abumdas, 1985: 99). In MSA, unlike the other Arabic varieties, the ultimate heavy syllable has to be extrametrical; otherwise McCarthy's model will fail to yield the correct prediction. To sum up, the metrical model allows many parametric variations which are selected for the purpose of achieving observational adeqacy. The model lacks a general principle on the 3.3.2. Evaluation of th In the linear apprd assignment is based i the type of clusters i data has shown that : framework of the ST the identification of 90 basis of which we can make the appropriate parametric choices for building the metrical trees. Accordingly, the metrical model has to be modified. 3.3.2. Evaluation of the linear approach to stress assignment: In the linear approach which is based on the STGP, stress assignment is based on one guiding principle: the identification of the type of clusters in the word. The analysis of the JA and MSA data has shown that stress can be accounted for within the framework of the STGP on the basis of explicitly defined principles; the identification of weak and strong clusters. In this chapter, the phonological status of of two consonants wor consonants word-med first hypothesis is th second hypothesis is first hypothesis is ex involves the postulai involves the analysis with that of the othi htIItothesis is based that it is contrastii The chapter is d we Dresent and ana vowel in question i alternation of the . Chapter 4 Alternating High Vowels In this chapter, the main objective is the identification of the phonological status of the high vowel which shows Up in the context of two consonants word—finally or word—initially, and three consonants word—medially in JA. We consider two hypotheses. The first hypothesis is that the vowel in question is epenthetic. The second hypothesis is that it is underlying. The argument for the first hypothesis is evidence that the vowel is predictable. This involves the postulation of rules which predict the vowel. It also involves the analysis of the behavior of the vowel in comparison with that of the other vowels. The argument for the second hypothesis is based on evidence that the vowel is unpredictable, and that it is contrastive. The chapter is divided into seven sections. In the first section, we present and analyze some evidence which argue that the high vowel in question is epenthetic. This includes an analysis of the alternation of the vowel with zero, and an analysis of the behaviour of the vowel in respect to stress assignment in comparison with the low vowel [a]. In the second section, we postulate some rules to account for the presence and absence of the alternating high vowels. This includes a rule of epenthesis and a rule of high vowel deletion. In the third section, we analyze some aspects of the phonetic structures of JA and MSA. The analysis involves the identification 9i of the permissible cons permissible consonant I scale of sonority. Here phonotactic constraint peaks of sonority. In t Sonority Principle (PSI unpermissible consona applies to break conso fifth section consider preceding high vowel. account for the showi vowel is a copy of tht analyzes the vowel ii argue that the qualit context. Some rules the vowel as either findings of the chap of the relation beta aIJIJIication of these Otdata. lnthe appl Inclusion Precedeh Specific constrain‘ the failure of the i 92 of the permissible consonant clusters in both varieties. The permissible consonant clusters are discussed in reference to the scale of sonority. Here, we argue that epenthesis applies to satisfy phonotactic constraints which relate to the types of permissible peaks of sonority. In the fourth section, we introduce the Peak of Sonority Principle (PSP), which is postulated to account for the unpermissible consonant—clusters. Here, we argue that epenthesis applies to break consonant—clusters which violate the PSP. The fifth section considers the vowel in question in the context of a preceding high vowel. A rule of epenthesis will be postulated to account for the showing up of the vowel in the form of [u]. Here, the vowel is a copy of the preceding high vowel. The sixth section analyzes the vowel in the context of a preceding low vowel. Here, I argue that the quality of the vowel is determined by the consonantal context. Some rules are postulated to account for the appearance of the vowel as either [i] or [u]. In the seventh section, the main findings of the chapter are summarized. This involves consideration of the relation between the different rules of epenthesis and the application of these rules in the derivation of an illustrative sample of data. In the application of the rules, the principle of Proper Inclusion Precedence is invoked. We also consider some language specific constraints. This includes a constraint which accounts for the failure of the vowel in question to be assigned stress. 41. Evidence for epenth Evidence for epenth vowel in question in di‘ the other vowels. This failure to be assigned 4.1.1.Vowel-initial St In this section, w question in the contei that of low vowels ir (IA-B). (1A): n'afis “benefit“ n'abis "spring" k'asim "shape" Sable "lion" 93 4.1. Evidence for epenthesis: Evidence for epenthesis is based on the behaviour of the high vowel in question in different contexts in comparison with that of the other vowels. This includes its alternation with zero and failure to be assigned stress. 4.1.1. Vowel-initial suffixes: In this section, we consider the behaviour of the high vowel in question in the context of a vowel-initial suffix in comparison with that of low vowels in similar contexts. Consider the JA data in (1.A-B). (1.A)z nafis “benefit" nafs-ak "your benefit" nabis "spring" nabs—ak "your spring" kasim “shape" kasm—ak "your shape" sabis “lion" sabs-ak "yourlion" (1B): satas "thirst" satas—ak "your thirst" galam "pencil" galam-ak "your pencil" balad "country" balad-ak "your country" warag "paper" warag-ak "your paper" In (1.A), the high vowel alternates with zero in the context of a vowel—initial suffix. But in (LB), the low vowel does not. In (2.A-B), we consider the behavior of the high vowel in nouns andverbs. Consider the (MI: nafis ”benefit" n'abis "spring“ kasim "shape" sabis ”lion" (2B): sirif “he knew" simis "he heart ‘sirib “he drank kirih "he hated In (2.A-B) the high vc zero in the context 0 The analyses of t high vowels in nouns alternate with zero analyses of the data vowel in nouns is s vowel in both verbs avowela'nitial sufi 94 and verbs. Consider the JA data in (2.A—B). (2A): nafis ”benefit" nafs—ak "your benefit" nabic "Spring" nabs—ak “toyr Spring" kasim "shape" kasm—ak "your shape" sabic "lion" sabs—ak "yourlion" (2B): sirif "he knew" sirf—ak "he knew you" simis "he heard" sims—ak "he heard you" sirib “he drank" sirb-ak "he drank you" kirih "he hated" ki’rh—ak "he hated you" In (2.A-B) the high vowel in both nouns and verbs alternates with zero in the context of a vowel-initial suffix. The analyses of the data in (1.A—B) argue that the behavior of .high vowels in nouns contrasts with that of low vowels,- high vowels alternate with zero while low vowels do not. On the other hand, the analyses of the data in (2.A—B) argue that the behavior of the high vowel in nouns is similar to that of high vowels in verbs,- the high vowel in both verbs and nouns alternates with zero in the context of a vowel—tnitial suffix. 4.1.2. Stress in nouns and verbs: In this section, we consider with regard to stress the behaviour 0f the high vowel in nouns in the context of a consonant-initial suffix in comparison with the high vowel in verbs in similar ITAlzn'afis-hum “theii n'abis-hum "theii kasim-hum "the s’abis-hum "thei I3.B):sirif—hum "hek simis-hum “he I tirib-hum "hec kirih-hum "he I The data in (3A) contrasts with that (3.8). In (3A), the v contrary to the prec to the right-most v (3.8), the correspor rules predict. 4.1.3. Alternating I In this sectior the context of ap data in (4.A,B)_ (4A): caiiaxaib 95 contexts. Consider the .JA data in (3.A—B). (3A): nafis—hum “their benefit" nabis-hum "their spring" kasim-hum "their shape" sabis-hum "their lion" (3B): sirif—hum "he knew them" simis-hum "he heard them" sirib—hum "he drank them" kirih—hum "he hated them" The data in (3A) show that the high vowel in question in nouns contrasts with that of the corresponding high vowel in verbs as in (3.8). In (3A), the vowel fails to be assigned stress, which is contrary to the prediction of the stress rules, which assign stress to the right-most vowel followed by a consonant cluster. But in (3B), the corresponding high vowel is assigned stress as the stress rules predict. 4.1.3. Alternating high vowels preceding or following a sonorant: In this section, we consider the behaviour of the high vowel in the context of a preceding or following sonorant. Consider the JA data in (4.A,B). (4A): calib~calb "dog" sari'sx'ar's "crown x’alif-x'alf "behinc silim.cilm “know (4.8):h'abilfhabl "rop jahii,*]ahi "ignt sahinfsahn "pl fikirfitkr "tho The alternation of alternation in (4.8) c: sonority. 0n the basi Vennemann 1972;Ho proposed the scale a (5)1 M low vowels mid vowels 96 caris~sars "crown" Xalif~xalf "behind" silim~silm "knowldge" (4.8): habithabl "rope" jahilJ‘jahl "ignorance" s'ahinfiséhn "plate" fl'klhf‘fl'kr "thought" The alternation of the high vowel in (4A) and the absence of the alternation in (4.8) can be explained in reference to the scale of sonority. 0n the basis of earlier studies (Saussure 1966: 44-62; Vennemann 1972; Hooper 1976: 203), Hogg and McCully (1987: 33) proposed the scale of sonority provided in (S). (5): M Sonority values Examples low vowels 10 /a/ ; mid vowels 9 /e/, /o/ t high vowels 8 /i/, /u/ flaps 7 /r/ laterals 6 /l/ nasals 5 /n/ voiced fricatives 4 /b/ voiceless fricatives 3 /8/ voiced stop 2 /d/ voiceless stop 1 /t/ in (5), vowels and between 10 for the mo vowels are assigned tl lowest value. In this either p, ”s, L ormisl following consonant v preceding one which i The scale of sono languages. There may languages (Vennemar voiceless fricatives are stronger than f ri hiQher on the scale t scale of sonority in We provided by H. The JA data in ( sonority in (5). Int sonorant is higher Which is an ODStrut alternates with 2e an obstruehi is m consonant Whicht does not fitternais Consider now (6A): basie~bas8 haxoilclia'bf 97 In (5), vowels and consonants are given values which range between 10 for the most sonorous and 1 for the least sonorous. Low vowels are assigned the highest value and voiceless stops the lowest value. In this scale, the final consonant in (4A) which is either p, i, L or r_n_ is less sonorous than 1, or E. But in (4.8), the following consonant which is eitherL n, org is higher than the preceding one which is either 1;, ll. 11, or lg. The scale of sonority in (5) does not necessarily apply to all languages. There may exist considerable differences among languages (Vennemann, 1988: 9). Hooper (1976: 302) argues that voiceless fricatives are stronger than voiced fricatives and stops are stronger than fricatives. But in (5), voiced fricatives rank higher on the scale of sonority than voiced stops. Accordingly, the - scale of sonority in JA may be different from the scale in (5) or the scale provided by Hooper. The JA data in (4.A—B) SUpport the predictions of the scale of sonority in (5). In (4A), the preceding consonant which is a sonorant is higher on the scale of sonority than the following one which is an obstruent. Accordingly, the high front vowel freely alternates with zero. But in (4.8), the preceding consonant which is an obstruent is lower on the scale of sonority than the following consonant which is a sonorant. Accordingly, the high front vowel does not alternate with zero. Consider now data of (6.A-F). (6A): baci8~bace "resurrection" habif~habf "deletion” nasi'scnaQE "CO” daciécdasc "Will lop): lasisfilass “sti nafisfnafs "se waxiszaxz "I nasijcnas] “we (6.C):'sasib~'sasb “pe xazigcxazg "pt naiid-na'sd "nt wasid~wasd " (6D): nabisfinabs ‘ nagizfinagz Iadibfilado ' nabi‘bfinab’b (6E): 'sabithabt panichahi - Sobukcsabk éabuk~§abk (6.F)'.natig~xmS qa“§~*sat Sfikubd‘sal SGtUbJ'sat 98 nasiscnacs "coffin" dasic~dasc "wrinkling" (6.8): lasicfilass "stinging" nafisjnafs "self" waxiz~*waxz "poking" nasij~nasj "weaving" (6C): sasibNEasb "people" xazig~xazg ”puncturing" najid~nasd "name of an are in Arabia" wasid~wasd "promise" (6D): nabis~*nabq “spring" nagiz~*nagz "jumping" ladiofilado "biting" nabi’b..*nab’b "abandoning" (6E): sabit~§abt "falling in a trance" nabit~nabt "fully loaded" sabuk~sabk "a method of treating iron" sabuk~§abk "entangling" (8F): natigJ‘natg "vomitting" satigfsatg "setting someone free" sakubfsakb "raining heavily" satubfisatb "crossing out" alternates with zero. avoiceless fricative 1 following consonant v high front vowel does preceding consonant 1 scale of sonority tha Accordingly, the higl in (6D), the precedir the scale of sonoritt fricative. Accordin zero. In (6E), the p higher on the scale voiceless stop, Ac With zero. But in( Stop is lower on ti which is a voiced alternate With zei 99 In (6A), the preceding consonant which is a voiced fricative is higher on the scale of sonority than the following one which is a voiceless fricative. Accordingly, the high front vowel freely alternates with zero. But in (6.8), the preceding consonant which is a voiceless fricative is lower on the scale of sonority than the following consonant which is a voiced fricative. Accordingly, the high front vowel does not alternate with zero. In (6C), the preceding consonant which is a voiced fricative is higher on the scale of sonority than the following one which is a voiced stop. Accordingly, the high front vowel freely alternates with zero. But in (6D), the preceding consonant which is a voiced stop is lower on the scale of sonority than the following consonant which is a voiced fricative. Accordingly, the high front vowel does not alternate with zero. In (6E), the preceding consonant which is a voiced stop is higher on the scale of sonority than the following one which is a voiceless stop. Accordingly, the high front vowel freely alternates with zero. But in (6.F), the preceding consonant which is a voiceless stop is lower on the scale of sonority than the following consonant which is a voiced stop. Accordingly, the high front vowel does not alternate with zero. To sum up, the JA data in (4.A—B) and (6.A—F) are explained by sonority in (5) in the sense that in the data the alternation takes place when the preceding consonant is higher than the following consonant on the scale of sonority. The data also show that high vowel alternation does not take place when the following consonant is higher on the scale of sonority than the preceding one. 4.1.4. Concluding rema The analyses of ti vowels contrast with zero while some high vowel-insertion suff verbs (see (2)). Witt contrast appears: in does not receive str Postponing further f consider phonologic to the scale of sonc preceded by a cons: consonant. We will below eIlenthetic while t lion front vowel o epenthesis, and th (Utes of deletion. mainly concerned (ll‘st ttlpe. 4.2.I.Rules of 9o DOStUtate rm 9 (_ 100 4.1.4. Concluding remarks: The analyses of the data in (1—4) and (6) indicate that high vowels contrast with low vowels,- low vowels do not alternate with zero while some high vowels in some contexts do (see (1)). With vowel-insertion suffixes this vowel alternates in both nouns and verbs (see (2)). With consonant-initial suffixes, however, a contrast appears: in nouns the vowel in the alternating contexts does not receive stress, whereas in verbs it does (see (3)). Postponing further discussion of the contrast of nouns and verbs, we consider phonological environments of alternation with reference to the scale of sonority of (5). Zero appears for the high vowel when preceded by a consonant lower in sonority than the following consonant. We will below argue that the alternating high vowel of nouns is epenthetic while that of verbs is underlying. The appearance of the high front vowel of the first type has to be accounted for by rules of epenthesis, and the absence of the high vowel of the second type by rules of deletion. However, in our subsequent analyses, we will be mainly concerned with the appearance of the high vowels of the first type. 4.2.1. Rules of epenthesis: To account for the alternating high vowels of the first type, we postulate rule (7) of epenthesis. Rule (7)1 g _—-> V I+high l-bacl In compliance wi' hphesobhgatornyi isy,and that of the apmiesinthe conte> onthescaleof sono Rwe(7)predicts th oftwoconsonantSi word-rnedially. Ifitwaspossu consonantal segrne rvle (7'). ‘IO‘i Rule (7): 8———> V / C__C{C} # [+high] {—back] In compliance with the scale of sonority, rule (7) of epenthesis applies obligatorily if the sonority value of the preceding consonant is n, and that of the following consonant m. But it optionally applies in the contexts where the value of the preceding consonant on the scale of sonority is n and that of the following one is n_—_j_. Rule (7) predicts the presence of a high front vowel in the context of two consonants word—finally and three consonants word—medially. If it was possible that the word-final position functions as a consonantal segment, we could simplify rule (7) of epenthesis into rule (7'). Rule (7'): Z —--—> V / C __ C C [+high] {—back] However, some questions can be raised against rule (7‘) of epenthesis. The rule does not account for the appearance of the vowel in the form of [u] in some contexts. Further evidence related to this will be analyzed in sections 6 and 7 in this chapter. l-stressl Rule (8) of high vowels in the conte is intrinsically ord the stress rules. I‘ stress rules. However, some vowel deletion. RI some contexts. TI evidence to rule ( 102 4.22. High vowel deletion: To account for the absence of the high vowels of the second type, we postulate rule (8) of high vowel deletion. Rule (8): V ——-—> B / ___CV [+high] [-stress] Rule (8) of high vowel deletion accounts for the absence of high vowels in the context of a vowel-initial suffix as in (1.C). Rule (7) is intrinsically ordered after the stress rule. That is, it is fed by the stress rules. It has the feature [-stressl which is fed by the stress rules. However, some questions can be raised against rule (8) of high vowel deletion. Rule (8) of high vowel deletion fails to apply in some contexts. There are some data which are potential counter evidence to rule (8) of high vowel deletion. Consider the data in (9.A,B). (9A): zihig "he got bored" zhfg—na "we got bored" fihim "he understood" fhi’m—na "we understood" xi’sir ”he lost" xsir—na "we lost" sfrif "he knew" <:ri’f—na “we knew" ribih "he won" rbih—na "we won" lob): zihig "he got bor fihim "he unders xisir "he lost" sirif "heknew" ribih “he won“ In (9A), the worc and the subject is no marked for the first applies to the words unstressed high vow words are transitiv suffix -na. The sub 0t rule (8) to the w would result in for corresponding intr To account for (95), we consider the assumption th Analtlsis within t milking some 0thE the primary stres and that later in 103 (9.8): zihig "he got bored” zihi’g—na "he got bored with us" fihim "he understood" fihim-na "he understood us" xisir "he lost" xisir-ha "he lost us" sirif "he knew" sirif—na "he knew us" ribih "he won" ribih—na "he won us over" In (9A), the words are intransitive verbs which take no object, and the subject is not marked for the third person singular. But it is marked for the first person plural by the suffix :3. Rule (8) applies to the words in the second column of (9A), deleting the unstressed high vowel to yield the correct results. But in (9.8), the words are transitive verbs the subject of which is marked by the suffix -na. The subject, as in (9A), is not marked. The application of rule (8) to the words in (9.8) would yield the wrong prediction,- it would result in forms which are homophonous with the corresponding intransitive verbs. To account for the failure of rule (8) to apply to the data in (9.8), we consider two approaches. The first approach is based on the assumption that phonological rules apply in a cyclic fashion. V Analysis within the framework of this approach also involves making some other assumptions. This includes the assumption that the primary stress assigned on the first cycle becomes secondary and that later in the derivation it deletes (Alghazo: 22—3). In the cyclic approach, the rules first apply at the first cycle which consists of the verb and its subject prefix. Then rules apply at the second cycle which consists of the verb and the object suffixes attached to it. The second approach is based on the assumption that some phonological rule Inhis analysis of a (1987: 22—3) argues tl first cycle, he argues, intransitive verb, ass in fihim. 0n the sec stem and the object s penultimate vowel a: stresses are assigne assumes, the stress Then a rule of the ft (Motion, 1987: 22) 104 some phonological rules are morphologically constrained. In his analysis of JA and following Brame (1974), Alghazo (1987: 22-3) argues that rules apply in a cyclic fashion. 0n the first cycle, he argues, stress rules apply to the verb stem of the intransitive verb, assigning a primary stress to the first vowel as in fl’him. On the second cycle, stress rules re—apply to the verb stem and the object suffixes attached to it, assigning stress to the penultimate vowel as in f/‘hl’mna. In this way, two primary stresses are assigned to the stem of the verb. By convention, he assumes, the stress assigned on the first cycle becomes secondary. Then a rule of the form in (10) deletes the secondary stress (Alghazo, 1987: 22). Rule (10): ‘v —-—> v /___ CO V’ In (10), the symbol [‘1 over the vowel stands for a secondary stress. Rule (10) deletes the secondary stress which has been derived from a primary stress by convention. In (11), we illustrate the derivation of the intransitive verb fhfm-na ”we understand" and the transitive one fth'm-na "he understood us" within the framework of the cyclic approach. First Cycle flhim—na fhim-na Second Cycle 105 (ll): [fihim—nal "we understood" [[fihiml—nal "he understood us" UR First Cycle fihfm—na ffhim Rule (15) in ch. 3: Stress fhi'm-na ——- Rule (8): High vowel deletion Second Cycle -—— fihimna Rule (15) in ch. 3: Stress ___ —-- Rule (8): High vowel deletion -—— f‘ihimna By convention -—— fihfmna Rule (10): Secondary stress del. [fhl’m-na] = [fihim-na] PR In (1 1), on first cycle, rule (15) of stress applies, assigning stress to the penultimate high vowel of the intransitive form in which the suffix :pa is a subject marker. Rule (15) also assigns stress to the penultimate high vowel of the transitive verb in which the subject is not marked. Then rule (8) of high vowel deletion applies to the intransitive form, deleting the unstressed high vowel, and yielding the correct prediction, fhl'm-na "we understood". On the second cycle, rule (15) of stress assignment re-applies to the transitive form which has the suffix :n_a as an object marker, assigning stress to the penultimate high vowel which is followed by a consonant cluster. At this level, the transitive form is assigned two primary stresse penultimate on the f the second cycle. In the cyclic ap deletion cannot appl satisfied. Stress is the the first cycle 1 way, the cyclic app asssumptions are p primary stress bec secondary Stress ( are ordered and ap stress assigned u secondary, Thenr vieiaihg rihl’m-n (Cannotoooly. Rul after rule (a) of 1 “WI deletion in *(hIm-na "he UI intransitive fort 106 two prhnary stresses. The first stressis assigned tothe penulthnate on thefirstcycle andthe second tothe penultnnate on the second cycle. In the cyclic approach, rule (8) of unstressed high vowel deletion cannot apply to the transitive form because its SD is not satisfied. Stress is assigned to first high vowel of the verb stem at the the first cycle and to the the second on the second cycle. In this vvay,the cyclic approach prevents honoophony. Three arbitrary asssurnptions are postulated. 'fhefirst assurnptionis that a prnnary stress becornes secondary,the second oneis that a secondary stress deletes by convention,and the thirdis that rules are ordered and applycnfly once percycle. Accordingly,the prnnary stress assigned to the penulthnate vovvelon thefirst cycle becornes secondary. Then rule(1())of secondary stress deletion applies, yielding fihfm-na "he understood us",-to which high vowel deletion Acannot apply. Rule (10)of secondary stress deletionrnust apply after rule (8) of high vovvel deletion. Othersise,rule (8) of high vovveldeletion wall delete the unstressed high vovvel,yielding *fhl'm-na "he understood us" which is homophonous with the intransitiveforni flhfln-na ”we undersoodfl The problem with this type of analysis as provided in (11) is that it resorts to three strictly analyticjarbitrary assumptions the rnotivation ofvvhichis achieving observationaladequacy.lVe have noindependent evidenceto supportthe assunnptHNithata prwnary stress becornes secondary and then deletes laterin the derivation, orthat high vovveldeletion necessarily precedes secondary stress deletion. An alternative ap that rule (a) of high form because it is m transitive verbs. Th to maintain contrasI Accordingly, we pos applying to transiti Rule (12): V -——> V I+highl l-stressl In (12), “Vt" 31 effect of preventi applies to intrans yielding flil'm-na "he understood us conditioned takes (12) maintains tI and flifm~na "i without resortir this approach ll 107 An alternative approach to the vowel is based on the hypothesis that rule (8) of high vowel deletion does not apply to the transitive form because it is morphologically constrained from applying to transitive verbs. The motivation for the morphological constraint is to maintain contrast between the transitive and intransitive forms. Accordingly, we postulate rule (12), which constrains rule (8) from applying to transitive verb stems which have high vowels. Rule(12): v ———> v / __c v cth [+high] [+high] [-stress] In (12), "Vt" stands for transitive verbs. Rule (12) has the effect of preventing homophony. Rule (8) of high vowel deletion .applies to intransitive verbs such as fihl’m-na "we understand", yielding th’m-na, but not to transitive verbs such as rth'm-na "he understood us" because rule (12) which is morphologically conditioned takes precednce. The morphological constraint on rule (12) maintains the contrast between fihl’m-na "he understood us" and fhim-na "we understand". Rule (12) yields the correct result without resorting to the three arbitrary assumptions. In support of this approach, it has to be pointed out that the morphological constraint applies to a very limited set of data represented only in transitive verbs with a high stem vowel, and without this morphological constraint, these transitive verbs and the corresponding intransitive verbs will be homophonous. 43.1. Consonant clus In this section,‘ clusters in JA. This clusters word-final consider the constr reference to the sc 4.3.1.1. Consonant l Word-finally, J consonant of the c is non-sonorant. ' 4.3.1.2. Consonant Consider the (I3.A):sLimur "a §uoul "wt sisir "pr habit "rc s'alam “f (‘3-3)1sirif "hj I‘ibih “h tarah "j 108 4.3.1. Consonant clusters in JA: In this section, we identify the types of permissible consonant clusters in JA. This includes the identification of the permissible clusters word-finally, word—medially, and word-initially. We also consider the constraints on the permissible consonant clusters in reference to the scale of sonority. 4.3.1.1. Consonant clusters word-finally: Word-finally, JA freely allows consonant clusters when the first consonant of the consonant cluster is a sonorant and the second one is non—sonorant. This is illustrated in (3A). 4.3.1.2. Consonant—clusters word—medially: Consider the JA data in (13.A,8,C,D). (13A): camur "age" sumur-ha "her age" stipul "work" suaul-ha "her work" sfsir "price" sisir-ha "its price" habil "r0pe" habil—ha "her r0pe" salam "flag" <:t’alam-ha "her flag" (13.8): sirif "he knew" sirl’f-ha "he knew it" ribih "he won" ribih—ha "he won it" sarah "he explained" sarah-ha "he explained it" katab "hE‘NrOI' oakal “he ate" uaclaflmuf‘hge" ‘sdbul "work" fisil "action" sisir "price" habH "r009" (13.8):tirib “he dran slrif “heknev ribih "he wor flhim "he unc xisir “he lost In (13A), the att noun stem apparent‘ (13.8), the attachmi creates a cluster of avowel-initial suf 109 katab "he wrote" katab—ha "he wrote it" 7akal “he ate" 7akal—ha "he ate it" (13C): scimur "age" somr—ak "your age" saaui "work" suZSI-ak "your work" ffsil "action" ffsl-ak "your action" sisir "price" sisr-ak "your price" habil "rope" habl-ak "your rope" (13D): sirib "he drank" sirb-ak "he drank you (metaphoric)" sirif "he knew" sfrf—ak "he knew you" ribih “he won" ribh—ak "he won you" ffhim "he understood" ffhm-ak "he understood you" xisir "he lost" xisr—ak "he lost you" In (13A), the attachment of a consonant—initial suffix to the noun stem apparently creates a cluster of two consonants. Also, in (13.8), the attachment of a consonant-initial suffix to the verb stem creates a cluster of two consonants. In (13.C,D), the attachment of a vowel-initial suffix to the noun or verb with a high vowel stem results in the absence or deletion of the second high vowel, and a two-consonant cluster is formed. Here, it has to be pointed out that under the assumption that the second vowel of the stem in (13A) is epenthetic, we can say that the attachment of a consonant-initial suffix to the noun stems in (13A) creates a three—consonant cluster. These types of clusters are not allowed in JA. Hence, rules of epenthesis apply to break them up. Under the sat attachment of a vowel eliminates the need fc Word-medially, JA unless the first consc non-sonorant as in (4 Underthe assumpi the first column in (' consonants are creat attached. In verbs a stem vowel is under when a consonant-ii However, when column of the nouns cluster is created. because the rules c consonants word-r Created as a result deletion. 110 them up. Under the same assumption, we can also say that the attachment of a vowel-initial suffix to the nouns in (13C) eliminates the need for epenthesis. Word—medially, JA does not allow clusters of three consonants unless the first consonant is sonorant and the second one is non-sonorant as in (4A). Under the assumption that the second high vowel in the nouns in the first column in (13A) is epenthetic, clusters of three consonants are created in JA when a consonant-initial suffix is attached. ln verbs as in the first column in (13.8), where the second stem vowel is underlying, clusters of two consonants are created when a consonant—initial suffix is attached to the verb stem. However, when a vowel-initial suffix is attached to the first column of the nouns in (13C) and the verbs in (13D), a two consonajt cluster is created. In nouns, the two-consonant cluster is created because the rules of epenthesis do not apply in the context of two consonants word—medially. In verbs, the two consonant—cluster is created as a result of the application of rule (8) of high vowel deletion. 4.3.1.3. Consonant clusters word-initially: Word-initially, JA allows clusters of two consonants. Consider the JA data in (14). (14): ktaabfliktaab "book" lséannils’aan "tounge" slaahnislaetl "V plaadniblaad "C raaaomrcaab "5 “Mahatma 'st'aracoiEtara "h nkasar,7inkasaf lt'ahagceilt'ahag htaramcvihtara The data in (14) S word-initially. Thes 21-. This prefix ha purelyphonetic. Its in reference to eper 4.3.2. Consonant clL Unlike JA, MSA Consider the MSA c JA data in (13). (ISA): stimrt~u-n) §u251(~u~n) fisl(~u~n) slsr(~u—n) habl(~u-n) ill slaahnislaah "weapon" blaadmiblaad "countries" rcaabuircaab "stirrup" stamas~7istamas "he listened" starauistara "he bought" nkasar~9inkasar "he got broken" ltahagwiltahag "he joined" htaram~7ihtaram “he respected" The data in (14) show that JA allows two-consonant clusters word-initially. These clusters are optionally preceded by the prefix zi-. This prefix has no semantic function. Its function seems to be purely phonetic. Its function and derivation will be further analysed in reference to epenthesis in section 5 in this chapter. 4.3.2. Consonant clusters in MSA: Unlike JA, MSA does have two-consonant clusters word-finally. Consider the MSA data in (15.A,B) which correspond to some of the JA data in (13). (15A): sumr(-u-n) “age-nom-indef." suau-u-n) "worknom-indef." ficl(—u-n) "action—nom—infef.“ sicr(—u—n) "price—nom-indef." habl(-u-n) "rope—nom-indef." (15.8): 's'arib(—a),*'sa sarift-a)~*sa rabih(-a)-*rf i'ahlm(-a),,*f The data in (15. word-finally. Thes suffix -u-n "norr (15.8) show that v: whether they have (in these grour alternation word analysis of guest impermissible col nouns in JA wher . Word-mediaih arise in MSA. Th "their" to the no cluster because Optional word-r The data in (16‘, (16): gL"(Tlr(~u-i 112 (15.8): ”sarib(-a)~*'sarb(~a) "he drank“ sarif(-a)~*sarf)—a) "he knew" rabih(-a)~*rabh(-a) "he won" fahim(-a)..*fahm(-a) "he understood" The data in (15.A) show that MSA allows two—consonant clusters word—finally. These types of clusters show up in nouns where the suffix -u-n "nominative—indefinite" is optional. The MSA data in (15.8) show that verbs do not have consonant clusters word-finally whether they have the optional subject marker suffix -a or not. On these grounds, we can say that the question of high vowel alternation word—finally does not arise in MSA. Accordingly, the analysis of questions related to phonotactic constraints on unpermissible consonant clusters word—finally will be restricted to nouns in JA where consonantvclusters are not allowed. Word—medially, the question of three-consonant clusters does not arise in MSA. The attachment of a consonant—initial suffix -hum “their" to the noun stem as in (16) does not create a three consonant cluster because the case marker -u-n "nom.~indef." which is optional word-finally is obligatory with -hum and other suffixes. The data in (16) illustrate this. (l6):sumr(-u—n) "age-nom.-indef." cymr-u-hum "theirage" scam-um) ”work—nom.-indef." suol—u—hum "their work" fisl(-u*n) "action—nom.-infef." ffsl-u—hum "theiraction" sisr(—u-n) "price-nom-indef." sisr—u—hum "their price" habl(—u—n) ”rope-nom.—indef." habl-u-hum "their rope" 1‘,“ "f The initial conso! three consonant ctul questions related to} word-medially do n epenthesis word-m 4.4. Consonant clus In this sectior clusters of two—c clusters WOl'd-ms (17A): 9(ll'S-qt'jru (it‘llbutst'llu mllh~mn gir(Min caioash Itrblar Qm‘hgc Zia—(”zl sin-(“9 113 qabr(—u-n) "grave—nom—indef." qabr—u~hum "theirgrave" samal(—u—n) "work—nom-indef." samal—u—hum "theirwork" salam(—u-n) "flag—nom-indef." salam—u—hum "theirflag" The initial consonant of the attached suffix will fail to form a three consonant cluster with the preceding noun stem. Thus questions related to constraints on three-consonant clusters word-medially do not arise in MSA. Accordingly, the analysis of epenthesis word—medially will be limited to JA. 4.4. Consonant clusters in reference to sonority: In this section, we identify the contexts where JA allows clusters of two-consonants word—finally and three—consonant clusters word-medially. Consider the JA data in (17.A,B). (17.A):st1rs,.curus "wedding" ouleSulub "defeat,loss" mflh~mflih "salt" sirg~s1rig "branch" caib~caiia "dog" Jarh~jaruh "wound" gt'Jl-t~gul-it "I said" ZOF-t~zur-lt "I Visited" cin—t~sfn—it "I helped" (17.8): SUrS-kUWNCUI’US-kuw "your wedding" sirg-kuWfim eats-macaw para-koala" got-t-kuwal't zur-t-kuw~2( J sin-t-I V / it __ C C [+high] t-back] The analysis of that epenthesis api geminates called fl epenthesis applies Epenthesis applies stopped talking", t non-application of with the predictit does not. To acco epenthesis is prii .ond that epenthes geminates, but is In (24), we 11 Involve the high (24) "[109" “mp (Salt) héb 120 Rule (23): a———> {—syl] / it __ [+syll I—cont] [+low] The analysis of the .JA data we have considered so far indicate that epenthesis applies when the PSP is violated. But in a type of geminates called fake geminates as we will see in Chapter 8, epenthesis applies optionally even though the PSP is not violated. Epenthesis applies Optionally to the fake geminate in sakétt "I stopped talking", yielding sakétit "I stopped talking". The non—application of epenthesis to geminates as in sakétt complies with the predictions of the PSP while its application as in sakétit does not. To account for this, I assume that the motivation for epenthesis is primarily to break unfavourable consonant clusters _and that epenthesis is generalized to other clusters including fake geminates, but is obligatory where the cluster violates the PSP. In (24), we illustrate the derivation of some words which involve the high vowel alternation. (24): /calb/ /habl/ /habl—na/ /Isaan/ /sirif—na/ UR "dog" "rope" "our rope" "tongue" "we knew" calb habl habl—na Isaan siri’f—na Rule (15) in Ch.3: Stress -_.- ___ ___ --— crif—na Rule (8): HVD (canal IhaoiII mQQAMrm marSDismeL F of high _vowel delf feedsinto rule (7 VOWBl insertion l ,48.Epenthesis n The main objf aHernatiny high section lin this examinethe con WMchitshows Q5A)hMum “ katub " rujum futur ‘ 121 cam: habil h'abil-na -—- ———- Rule (7'): Epenthesis ——~ ——— —-— ilsaan ——— Rule (22): Vowel inser. —-— —-- ——— ailsaan ——- Rule (23): Glottal stop insertion [calib] [habil] [habil-na] [eilsaanl [srff—na] PR In (24), the rules do not have to be ordered. They apply whenever their SD is met. Rule (15) of stress assignment feeds into rule (8) of highvowel deletion (HVD). On the other hand, rule (8) of HVD feeds into rule (7) of vowel insertion. Rule (22) of word-initial vowel insertion feeds into rule (23) of glottal stop insertion. 4.6. Epenthesis in the context of a preceding high vowel: The main objective of this section is to account for the alternating high vowel when it follows a high vowel. We argued in section 1 in this chapter that this vowel is epenthetic. Here, we ‘ examine the contexts in which the vowel shows up as [i] and those in which it shows up as [u]. Consider (25.A—8). (25A): hulum "dream" kutub ”books" rujum "heap of stones" futur "mushroom" hfikum “jut §tlt1ul 'wor 'st'tsul 'kint judul 'twi‘ jt’tgum "th‘ kuhul 'dar; kufué 'crt‘ xubue "ma gubus "kit buguc “sp zulSut "sp (25.3):Iisim "bot lisir "bri fikir “thf sitig “top Stilt) “pri hImll "lo §lmlt5 ( ”till i: I'lzly "m 122 susub "grass" 20rug "blue ones" hukum "judgment“ sudul "work" stisul "kind of sheep" j't’idul "twisted ears of sheep" jagum "things which are not straight" kahul "dark black cosmetic powder " kt'tfus "crippled hands" xubue "malice, ill feeling" glibus "kind if hat" buguc "spotted animals" zuZSut "spur" (25.8):j'l’sim "body" j'I'sir "bridge" fikir "thought" sisig "love" sfj'in "prison" himil "load" silim "science" jidis "branch" tfrti "baby" simits "glue" fitin "smart" ri'zig "property" sihim "share" i zlgit 'DI'OIJe In (2511-13), the vowel. The data sh the vocalic context have similar conso‘ Iilin (25.8). In eit vowel. Rule (7') of account for the vo epenthesis. Rute (26) of E alireceding [+hit Includes and pre IIOWel in the car 4-7- Epenthesis t“ this secti context of a or 123 tibin "hay" sifin "purse" zigit "proper name, a religious person" In (25.A-B), the alternating vowel is a copy of the preceding high vowel. The data show that the quality of the vowel is determined by the vocalic context, not the consonantal one. In (25A) and (258), we have similar consonantal contexts. But the vowel is [u] in (25A) and [i] in (25.8). In either case, the vowel is a copy of the preceding high vowel. Rule (7') of epenthesis predicts the [i] vowel in (25.8). To account for the vowel [u] in (25A), we postulate rule (26) of epenthesis. Rule (26): B -——-> V / V C __ C C [+high] [+high] [+backl [+backl Rule (26) of epenthesis accounts for the vowel in the context of a preceding [+high, +backl vowel. Rule (26) of epenthesis properly includes and precludes rule (7‘) of epenthesis which predicts an [i] vowel in the context of three consonants. 4.7. Epenthesis in the context of a preceding non—high vowel: In this section, we consider the quality of the vowel in the context of a preceding low vowel. Consider (27). (272nafis ”self" nasif "eXDIOG babix "too mu baxit "luck" nafux "blowir taguf “small Easib “heel“ nabis "spring nabub “pulse nagus "decre gatuf ”pickir lahim "meat batin “belly“ xanig "suffo farit "furnit farux "Chic na§ur "sawi oaoul "mule basil "wasl 124 (27: nafis "self" nasif "exploding" babix "too much spending" baxit “luck" nafux "blowing up" Eaguf “small stones" casib "heel“ nabis “spring of water" nabutp "pulse" nagus "decrease" gatuf "picking fruit" lahim "meat" batin "belly" xanig "suffocating" faris "furniture" farux "chick" nasur "sawing wood, publishing" batSul "mule" basil "washing" In (27), stress is not indicated because it is irrelevant. The vowel is either ii] or [u], even though the preceding vowel is consistently [a]. Accordingly, we can not say that the quality of the vowel is determined by the vocalic context when the preceding vowel is low. Below, we analyze all the consonantal contexts with a preceding [a] vowel. This includes an analysis of the vowel in the contexts of pmcemng andfollo‘ th,andpharyngea altheconsonantal TaMe(7):Epenthes HrstConsonant Dhittingealtphary flap emphatic (emph) Valor (vel) literal (lat) C0f0nal(c0r) labia((lab) The VETth Thehoriz Coasoha The VOWE Vowel 125 preceding and following labial, coronal, lateral, velar, emphatic, trill, and phargngeal. This is illustrated in Table (7) which covers all the consonantal contexts in which the vowel shows up. Table (7): Epenthesis in consonantal contexts First Consonant Second (following) Consonant ___—mm_——_.___ lab cor lat vel emph flap phargn phargngeal (phargn) [i] [i] [i] [u] [u] [U] “- flap [u] [i] -—— [u] [u] -—— [u] emphatic (emph) [u] [i] [U] [U] --— [U] [U] .Velar (vel) [U] [i] [U] --— [U] [U] [U] lateral (lat) [i] [i] --- [i] [U] --- [i] coronal (cor) [i] [i] [i] [i] [U] [U] [i] labial (lab) --- [i] [i] [U] [U] [U] [1'] The vertical row indicates the first (preceding) consonant. The horizontal column indicates the second (following) consonant. The vowel in brackets indicate the qualitg of the epenthetic vowel. The dotted line indicates the absence of these contexts. ; empr of a following velai it is also in] in the unless the followioi [i]. In the followin: contexts in which ‘ be postulated to at 4.7.]. Epenthesis i In this sectio following labial o [NA-C), resoect [23-A):farum "at haruf "a harub "w karUm "9 km] "er (285): “Mm «‘ 98§uf “s satub “c [Sasub «1 gem] “I 126 Table (7) indicates the contexts in which the vowel appears as [i] and those in which it appears as [u]. The vowel is [U] in the context of a following emphatic or flap. The vowel is also [U] in the context of a following velar and unless the preceding consonant is a coronal. It is also [u] in the context of a preceding emphatic, flap, or velar unless the following consonant is coronal. Otherwise, the vowel is [i]. [n the following sections, we provide some data illustrating the contexts in which the vowel appears as [u], and then some rules will be postulated to account for the vowel in these contexts. 4.7.1. Epenthesis in the context of following labials: In this section, we analgze the vowel in the context of a following labial and a preceding flap, emphatic, and velar as in (28.A-C), respectively. (28A): farum "cutting" haruf "aletter" harub "war“ karum "garden" jaruf "eroision" (28.8):qa’bum "bones“ gasuf "shelling" satub "cancelling" oasub "forcing" gatuf "picking fruit" . 4.- :‘I: 1" . , '47» 3' ("F d - '. dagum '10! Eaguf 'sma naxub 'ng sakub 'kinq in the context (28.A,B,C), respec 1h behaves as an em; [+back], we postu1 4'71 Epenthes [n [MS SEC‘ [Ollomng Vela 127 (28.C): faoum "mouth bite" dagum "to say something face to face" saguf "small stones" naxub "by selection" sakub "kind of iron" In the context of a preceding flap, emphatic, or velar as in (28.A,B,C), respectively, the vowel is [u]. Here, the JA Arabic [r] behaves as an emphatic. Assuming that emphatics and velars are [+back], we postulate rule (29) of epenthesis. Rule(29): B——-> V / C ._ C C [+high] [+backl [+labial] [+backl Rule (29) of epenthesis accounts for the vowel in the context of a preceding consonant which is [+back], predicting an epenthetic [u]. This includes emphatics and velars. Rule (29) properly includes rule (7') which predicts [i], and applies disjunctively in relation to it. Here, it has to be noted that labials are [+ant, —cor]. The feature [labial] is used for clarity of presentation. 4.7.2. Epenthesis in the context of following velars: In this section, we analyze the vowel in the context of a following velar and a preceding pharyngeal, trill, emphatic, lateral, bacug 'bab zacug 'scr mahug "ca nasug “the [305):farux 'a y harug “but faruk "rut maruo 'ct barug 'li [30.C): latux "Sl tnabub " fasux “u masuk “ natug "a BQDbmaHk‘t xalig "c ialix "t Ealix " dalik "1 128 coronal, and labial as in (30.A-F). (30A): lacug "tasting" basug "baby cry" zaeug "screaming" mahug "causing to perish" naeug "the sound of a crow" (30.8): farux "a young bird" harug "burning" faruk "rubbing togather" maruo "covering or soiling with dust" barug "lightening" (30.C): latux "splashing" ma’ouo "grinding food in the mouth" fasux "untying' masuk "sound of slapping' natug "articulation" (30D): malik "king" xalig “creating" falix "to split apart" calix "a kind of wild plant" dalik "smoothening" (30E): ba’oix "extravagance" [30.Flzsabuk "enta: nafux 'pum' dafug 'dasl; kamux ‘kin dabuo “tar In the context coronal as in (30.1 contexts the vow. the Arabic latera preceding phargn respectively, the We Dostulate rut Rule (31) u - B —-.. Rule (31) 01 V9]flrs_ HOWew [29 xanig "suffocating" hatik "violating social norms" ceajig "messing" nasix "copying" (30.F): sabuk "entangling" nafux "pumping up" dafug "dashing of water" kamux "kind of soft clay between layers of rock" dabuo "tanning" In the context of a following velar and a preceding lateral, or coronal as in (3o.D,E), respectively, the vowel is [i]. In these contexts the vowel is predictable by rule (7') of epenthesis. Here, the Arabic lateral [l] behaves as a coronal. But in the context of a preceding pharyngeal, flap, emphatic, or labial as in (30.A,B,C,F), respectively, the vowel is [u]. To account for the vowel in (30.B,C), we postulate rule (31). Rule(3l): Z———> V / C __ C C [+high] [+back] [+back] [+back] Rule (31) of epenthesis predicts the vowel in the context of a preceding and following consonant which is [+backl This includes velars. However, rule (31) does not account for the vowel in the Rule (32) of 91 context of a f one one which is l-cm Dreceding coronal (7'). Rules (31) a disjunctively in l ,4-7-3- Epenthesis 130 context of a preceding pharyngeal or labial and a following velar as in (30,A,F). Accordingly, we postulate rule (32) of epenthesis. Rule(32): B———> V / C ___. C C [+high] l-cor] [+backl [+backl Rule (32) of epenthesis predicts that the vowel is [u] in the context of a following consonant which is [+backl and a preceding one which is [-cor]. In the context of a following velar and a preceding coronal as in (30.D,E), the vowel is predictable by rule (7'). Rules (31) and (32) properly include rule (7') and apply disjunctively in relation to it. 4.7.3. Epenthesis in the context of following emphatics: In this section, we analyze the vowel in the context of a following emphatic and a preceding pharyngeal, flap, velar, lateral, coronal, and labial as in (33.A-F). (33A): gahut "draught" masut "pulling hair out" fahus "examination" waqu’b "preaching" (33.8): faru’o "obligatory" 5e; “lam guru! ' 30!] (33c): sagut 'kic] maliu’g 'stl malius 'st §axut 'scr raku’g 'rui (33b): xalut "mi malut "to bolus "to .[33£):na§ut "ti genus “r gasut "; hasut 'u mazut " 131 sarut "condition” garus "stinging" earu’o "land" garut " sound of eating something crunchy" (33C): sagut "kicking" maout "stretching" maous "stomach ache" saxut "scratch" raku’b "running" (33D): xalut "mixing" malut "to tear apart without using a knife“ balus "to get money by cheating" . (33E): nasut "tie with a loop" ganus “hunting" gasut "part of a payment" hasut "thrown in disorder" mazut "slipping off" (33.F):nabu’2_> "pulse" gamut "packing together" nafut "petrol,oil" nafub "sprinkling" xabus "mixing things together" Inthecontextl pharyngeal,flap, (33.A,B,C,D,E,F), r B2)donotpredic mdapmcemngCt contextof preced (32)predicts an [ consonant(includ [ultoaccountfc consonantis [+co rme(34)ofepen 132 [n the context of a following emphatic and a preceding pharyngeal, flap, velar, lateral, coronal, or labial as in (33.A,B,C,D,E,F), respectively, the vowel is [u]. Here, rules (31) and (32) do not predict the vowel in the context of a following emphatic and a preceding coronal. Rule (31) predicts an [u] vowel in the context of preceding back consonants (including emphatics), and rule (32) predicts an [u] vowel in the context of a following back consonant (including emphatics) and a preceding [~cor]. Thus they fail to account for the vowel in (33.D,E) where the preceding consonant is [+cor] and the vowel is [u]. Accordingly, we postulate rule (34) of epenthesis. Ru]e(34): Z—-—> V / C __ C C [+high] [+back] [+back] [+cor] Rule (34) of epenthesis predicts an [u] vowel in the context of a following emphatic. In this respect, emphatics have to be distinguished from velars; both velars and emphatics share the feature [+back], but they do not share the feature [+cor],- emphatics are [+cor] while velars are {—cor]. In the context of a following emphatic, the quality of the preceding consonant does not have to be specified. This explains the need for rule (34). Rule (34) properly includes rule (7') and applies disjunctively in relation to it. In thIs section following non-em emphatic, velar, l‘ I (351A): bahis 'dig racid 'thu‘ nacie 'sc nasis 'coi bahie 're (35.5): kari‘s “be harie “pl farid "pi [BITE "fu 36“] "s: (3511]: mm “c hasid "I ha‘gin ~ 98§Id " batin -- 133 4.7.4. Epenthesis in the context of following non—emphatic coronals: In this section, we analyze the vowel in the context of a following non—emphatic coronal and a preceding pharyngeal, flap, emphatic, velar, lateral, coronal and labial as in (35.A-G). (3S.A): bahi§ "digging" rasid "thunder“ nasie "scattering" naei§ "coffin" banie "research" (35.8): karis "belly" harie "ploughing" farid "pistol" faris "furniture" sari] "saddle" (33C): batis "cruelty" hasid "harvesting" habin "hugging, embracing" gasid "intention" batin "belly" (35D): faxib "thigh" hagin “injection" wakit "time" (35.9: jalid “whim colic "chew malis 'sma halif 'swei falit "to ie, (35.F):ha§id 'Cf'Ol‘ xadi§ 'scrt najid '1: pr natis "a co be[If] 'tor .[3S.G):gamiz ‘ju nafis “se] kabi§ “rm samin "pu 55bit] 'lii 134 sagid "contract" zaoit "punching with a nail" (35E): jalid "whipping" salic "chewing gum" malis “smoothening” halif "swearing" falit "to let out, to set free" (3S.F): hasid "crowd" xadi§ “scratch" najid "a province in Arabia" natis "a cut in cloth resulting from hooking" bajid “torn because of high pressure" (35.6): gamiz "jumping" nafis “self" kabié “ram" samin "purified butter" cabid "liver" In the context of a following non—emphatic coronal and a preceding pharyngeal, lateral, flap, emphatic, velar, lateral, coronal or labial as in (35.A,B,C,D,E,F,G), respectively, the vowel is [i]. In all of these contexts the vowel is predictable by rule (7') of epenthesis. [In (I‘ll; ' tion, following lateral a (3en):satui "buck fasul “chap; easul “orig‘ fabul “gene wasul "rec (36.8): nagul “trai [36511] “mul cagul "mlr naxul "pal haout “fie In the contex 135 4.7.5. Epenthesis in the context of following laterals: In this section, we analyze the vowel in the context of a following lateral and a preceding emphatic, and velar as in (36.A-B). (36A): _satul "bucket" fasul "chapter“ easul "origin" fabul "generosity" wasul ”receipt" (36.8): nagul "transfer" baoul “mule" cagul "mind" naxul "palm trees pl., lever" nagul "field" In the context of a preceding emphatic, or velar and a following [I] as in (36.A,B), respectively, the vowel is [u]. In these contexts, the vowel is not predictable by other rules of epenthesis. [l] is [+cor] and rule (7') predicts that the vowel is [i] in the context of a preceding velar or emphatic and a following coronal. To account for this problem, we assume that [l] is an emphatic coronal in the context of a preceding velar or emphatic. Elsewhere [I] is non-emphatic. Rule (37) of emphasis assimilation or emphasis Spread accounts for the appearance of emphatic [I]. Rule (37) replac‘ non-emphatic [l] inj consonant, yielding Accordingly, rule (1 (3t). Rule (37) of feeding relation tt that f or convenien chart of distincti' features: [+son, +| 41.6.Epenthesis In this seen following flap a and labial as in 136 Rule (37): [+lateral] --—-—> [+back] / C (V) __ [+back] Rule (37) replaces rule (6) of Chapter 2. It applies to non—emphatic [I] in the context of a preceding velar or emphatic consonant, yielding an emphatic [l] which is [+back, +cor]. Accordingly, rule (37) applies to the data in (36.A,B), feeding rule (31). Rule (37) of emphasis spread or emphasis assimilation is in a feeding relation to rule (34) of epenthesis. Here, it has to be noted that for convenience I identify [1] by the feature [+lateral]. In the chart of distinctive features in Chapter 2, [ll has the distinctive features: [+son, +cont, ~nas]. 4.7.6. Epenthesis in the context of a following flap: In this section, we analyze the vowel in the context of a following flap and a preceding pharyngeal, emphatic, velar, coronal, and labial as in (38.A-E). (38A): sasur "hair" bahur "sea' nahur "killing, stabbing" jaqur "the sound of a cow" jahur "glaring" (38.8): satur "splitting" 7'er :,]' gggur 'pala masur 'Eg (381:):fagur 'pove saxur 'snou makur 'det saour “soa bakur “pro (38blznaeur “pr: cajur "ra nasur “s: jadur “U natur 'S (38.0: xamur “i hafur 'd jamur “ habur “ labor ' 137 satur "writing line" habur "cufew" gasur "palace" masur "Egypt" (38C): fagur "poverty" saxur "snoring: makur "deceiving, tricking" saour "soaring“ bakur "proper name" (38.0): naeur "prose, scattering" sajur "raw, not ripe" nasur "sawing wood" jadur "the area within the city limit" natur "speaking forcefully" (38E): xamur "wine" hafur "digging” jamur "glowing charcoal" habur "lean meat" jabur "preper name" In the context of a following flap and a preceding pharyngeal, emphatic, velar, coronal, or labial as in (38.A,B,C,D,E), respectively, the vowel is [U]. Here, Arabic [r] behaves as an emphatic. Accordingly, rule (34) of epenthesis which accounts for the vowel in 4.7.7. Epenthesis in In this section] Ionowingphargngi (39.A-C). (39.A):'saruh "exp sarue "lat Iaruh “we garuc "pt jaruc “Si ]395)39°]U9 “at tabuh “a §8[Uh “r batuh "k maguq n (3913): 'saguh . segue " naxuq - faguq . ragug 138 the context of a following emphatic accounts for the vowel in the context of a following [r]. 4.7.7. Epenthesis in the context of following pharyngeals: In this section, we analyze the vowel in the context of a following pharyngeal and a preceding, flap, emphatic, and velar as in (39.A-C). (39A): §aruh "explanation" saruq "law" jaruh "wound" garuq "pumpkin" jaruq "swallowing" (39.8): gatuq "cutting" fabuh "disclosure, making secrets public" satuh "roof batuh "knocking someone" masue "slipping" (39C): saguh "crack, slice" saguq "sound of pouring water" naxuq "to spit noisily to clear the throat" faguq "bubbling" ragus "patching" Rule (40) pro: relation (0 It. To sum up, th as in rule (26). ‘ velar or emphatl Vowel is [U] in t‘ non-coronal. In (32). The vowe this context, tr 139 In the context of a preceding flap, emphatic, or velar as in (39.A,8,C), respectively, the vowel is [u]. In these contexts, the vowel is predictable by rule (40) of epenthesis which predicts an [u] vowel in the context of a preceding consonant which is [+back] and a following non-coronal. Rule(40): B—-——> V / C __ C C [+high] [+back] [-corl [+back] Rule (40) properly includes rule (7'), and applies disjunctively in relation to it. To sum up, the vowel is [u] in the context of a preceding [u] vowel as in rule (26). The vowel is also [u] in the context of a preceding velar or emphatic and a following non-coronal as in rule (40). The vowel is [u] in the context of a following velar and a preceding non—coronal. In these contexts, the vowel is accounted for by rule (32). The vowel is [u] in the context of a following emphatic. In this context, the vowel is accounted for by rule (34). Elsewhere, the vowel is [i] as rule (7') predicts. Table (7) illustrates the contexts where the vowel is [i] and those where it is [u]. The analysis of the JA data in (25), (27), (28), (30), (33), (35), (36), (37), and (39) support the assumption that the vowel in question is predictable. The vowel is a copy of the preceding high vowel as rule (26) predicts. But in the context of a preceding low In the context of a | I-corl as rule (34) I velar and a precedil context of a f allow the vowel is [I] as I accounted for by rt (32), (30.3) by (40 4.8. Rule applicat In this sectic Involved in the d have argued to b rules of epenthe rules of epenthe Rules (26), epenthetic [u] [3 these four Me In relation to I “We Dostulate ] and (40). 140 vowel, the quality of the vowel is determined by that of the preceding or following consonant. If the preceding and following consonants are [+back], [u] is predicted by rule (31). The vowel is [u] in the context of a preceding emphatic or velar and a following [—cor] as rule (34) predicts. It is [u] in the context of a following velar and a preceding [~cor] as rule (32) predicts. It is [u] in the context of a following emphatic as rule (34) predicts. Elsewhere the vowel is [i] as rule (7') predicts. But in fact all the data accounted for by rule (3I) are accounted for by other rules: (30.A) by (32), (30.8) by (40), (30.c> by (40), and (30.F) by (32). 4.8. Rule application: In this section, we consider the relation between the rules involved in the derivation of the alternating high vowel which we _have argued to be epenthetic. This involves the relation between the rules of epenthesis, on the one hand, and the relation between the rules of epenthesis, deletion, and stress assignment, on the other. Rules (26), (32), (34), and (40) of epenthesis which predict an epenthetic [u] properly include rule (7') which predicts [i]. Each of these four rules properly includes rule (7') and applies disjunctively in relation to it. Following, we repeat the rules of epenthesis we have postulated in this chapter, That is rules (7’), (26), (32), (34), and (40). Rules (7‘) (7') aDulles In 141 Rule (7'): B ——-—> V / C __ C C [+high] {-back] Rule(26x B -———-> V / V C __C C [+high] [+high] [+back] [+back] Rule (32): B———> V / C ____ C C [+high] {—cor] [+back] [+back] ,Rule(34k B ——-> V / C C C [+high] [+back] [+back] [+cofl Rule(40k B———> V / C ___. C C [+high] [+back] [—cofl [+back] Rules(7) and(26)ofepenthesisaretvubdflferentrules Rule (7) appliesin the context of three consonants. Here,the vocalic high vowel. The Ill that of the epenth In (42), we III. (34), and (36) in t (4I.A):nafis "sel Blzhulum 'dr (I2):§abuk “en (D):nabut3 "pt (E): batin "be (F): ca‘oum "bl (G): fasul "te The words i context of a pre Include emphat‘ Include the app III and lot [4]): [MW lhulm. n'afs [IUIm 142 context is not included in the SD of the rule. The quality of the vowel is determined by the preceding or following consonant. But rule (26) applies in the context of three consonants preceded by a high vowel. The quality of this preceding high vowel determines that of the epenthetic vowel. In (42), we illustrate the application of the rules (7'), (26), (32), (34), and (36) in the derivation of the JA words in (4I). (4I.A): nafis "self" 8): hulum "dream" (C): sabuk "entangling" (D): nabu’t} "pulse" (E): batin "belly" (F): sabum "bones" (G): fasul "term" The words in (41) include the appearance of the vowel in the context of a preceding high back or a low vowel. The words also include emphatic and non-emphatic contexts. These contexts include the appearance of the alternating high vowel in the forms of [i] and [u]. (41): /nafs/ /hulm/ /§abk/ /nabub/ /batn/ /qa’{)m/ /fasl/ UR n’afs ht’ulm sabk nab?) batn cabm fast Rule(l5)in Ch.3: Stress [nafis] [huluml I! In the deriv whenever their applies to eitht I+backt This it Of a following need for rule 0 DIODerIy inclui In this ch: contrast wItI II[)Wels. High ___ holum ___ ——— ~—— -—- —~— Rule (26): Epenthesis ___ ___ _-- ___ -—- ——— fas] Rule (37): Emphasis ass. ___ ___ sabuk nébub —-- ——— ——- RU]e (32): Epenthesis ___ ___ ___ ___ ___ --— fasul Rule (34): Epenthesis ___ ___ _—_ ——— ——— qa’bum —-— RUIe (40): Epenthesis nafis -—- ——— ——— batin ——~ --— Rule (7'): Epenthesis [nafis] [hillum] [sabuk [nabulp] [batin] [sebum] [fasuI] PR In the derivation of the PR of the data in (41) the rules apply whenever their SD is met. Rule (37) of emphasis assimilation applies to either fasl or fasil, yielding an emphatic [,I] which is [+back]. This feeds into rule (34) which predicts [u] in the context of a following emphatic. In the derivation of all the PR forms the need for rule ordering does not arise,- rules (32), (34), and (40) properly include rule (7') and apply disjunctively in relation to it. 4.9. Summary: In this chapter, the analyses of JA vowels show that high vowels contrast with low vowels and that there are two types of high vowels. High vowels contrast with low vowels [a] in reSpect to deletion; high vowe butlow vowels do alternate with zer vowels which shot corresponding can those of the secor the second type a) In Chapter 3 pred stress at all. On these gm eItenthetic and U ”I“ e[tenthesis phonotactic con: Word-(inalig an constraints Con I“ MSA, the clusters of Inn 01 three When) a consommHp consonantS (101 these mutant 144 deletion,- high vowels delete in the context of a vowel-initial suffix but low vowels do not. Those which show up in nouns Optionally alternate with zero in the context of a preceding sonorant but high vowels which show up in verbs do not alternate with zero in the corresponding contexts. Vowels of the first type contrast with those of the second type in respect to stress assignment; vowels of the second type are assigned stress as the stress rules postulated in Chapter 3 predict but vowels of the first type are not assigned stress at all. On these grounds, we assume that vowels of the first type are epenthetic and those of the second type underlying. We also assume that epenthesis applies in JA to satisfy some language Specific phonotactic constraints which do not allow two consonant clusters word-finally and three consonant clusters word-medially. These constraints comply with the peak of sonority principle. In MSA, the question of epenthesis does not arise. MSA allows clusters of two consonants word—finally. Word-medially, clusters of three consonants do not arise in MSA. In MSA, the attachment of a consonant—initial suffix to nouns which have final clusters of two consonants does not create a three consonant cluster because in these contexts a vowel which marks case shows up obligatorily. To account for the appearance of the alternating high vowel in question in the form of [i], we postulated rules (7') of epenthesis. To account for its appearance in the form of [u], we postulated rules (26), (32), (34), and (40). In the derivation of the vowel in question, the rules apply whenever their so is satisfied. The need for a rule ordering constraint does not arise because the principle of Proper I'Diftw ~ .'-'.'.-.:=-i'.e In this chapte counter evidence The chapter is diI consider some da (26) which epent second section, i evidence to rule (32) of epenthes [+back] consonat rule (34) of epe coronal emphatv Epenthesis whit coronal and a fl section, we C0! second radical: VOWeI~initiai 1 Re[lordIng this underlying; th: I epenthetic am . VOWe) which E [0”th Seetim Chapter 5 Problematic Data In this chapter, we consider some data which are potential counter evidence to the rules of epenthesis postulated in Chapter 4. The chapter is divided into four sections. In the first section, we consider some data which are potential counter evidence to rule (26) which epenthesizes [u] in the context of a preceding M. In the second section, we consider some data which are potential counter evidence to rule (7') of epenthesis which predicts an [i] vowel, rule (32) of epenthesis which applies in the context of a following [+back] consonant and a preceding [-cor] and predicts an [u] vowel, rule (34) of epenthesis which applies in the context of a following coronal emphatic and predicts an M vowel, and rule (40) of epenthesis which applies in the context of a preceding [+back] coronal and a following [~cor] and predicts an [u] vowel. In the third section, we consider the vowel which appears between the first and second radicals in imperfect verb forms in the context of a vowel-initial suffix and the first person singular prefix [ea—l. Regarding this vowel, we consider three analyses: that the vowel is underlying,- that it is a metathesized vowel; and that the vowel is epenthetic and that its quality is determined by the underlying vowel which appears between the second and third radicals. The fourth section is a conclusion. I46 In this sectld counter evidence predicts the vow Consider the dat' (IA): ruin (*ro; been (*bl (LB): xubuz~xu burujchI xuruj~xu (It): gttlitNgu sumit~§I busit~bi soritssu zurit~21 I" (1.A), u 147 5.1. Data problematic to rule (26): In this section, we consider some data which are potential counter evidence to rule (26) of epenthesis in Chapter 4, which predicts the vowel to be a copy of the preceding [+high,+back] vowel. Consider the data in (I.A,B,C): (LA): fujl](*ft1]u]) "raddish" busl'] (*busuD "bushel" (1.8): xubuz~xubiz "bread" buruj~burij "tower" xt’uruj~xurij "kind of bag" (1C): gUIit~gulut "I said" I sumit~sumut "I fasted" bysit~busut "I kissed" sIint~§Ufut "Isaw" zurit~20rut "I visited" In (1.A), the vowel is [i] which is contrary to the predictions of rule (26). In (1.8), the vowel optionally alternates between Ii] and [u]. The forms with [i] are potential counter evidence to rule (26) which predicts an [u] vowel. In (112) also, the vowel optionally alternates between [i] and [u], but it is more commonly [i]. The [u] alternate is more common among the older generation and less common among the younger generation. This alternation is not underlying becau! Chapter 4 are epq claim, consider ( (an): fujil-hum bfisll-hum' (2.8): xubiz-hur I ‘ burij—hun I xurij-hur (21:): gull-t-hu sumi-t-h btlsi-t-ht 'squ-t-ht zuri-t—h 148 predictable by rule (26). The high vowel in question in (I.AwC) cannot be considered underlying because it patterns with the vowels which we argued in Chapter 4 are epenthetic. For further evidence in support of this claim, consider (2.A,8,C), with regard to stress. (2A): fUJiI-hum ~*fu]iI-hum "their raddish" busil—hum ~*bu§il-hum "their bushel“ (2.8): xubiz-hum~*xubiz-hum "their bread" bt’urij—humfiburij—hum "their tower" xurij—hum~*'xur1]—hum "their bag" (2C): guIi—t-hum~*guli-t-hum "I said to them" sumi—t-hum~*_sumI-t—hum "I fasted on them" bt’usi—t-humfibusi—t—hum "I kissed them" stifi-t—humNi‘sufi-t—hum " I saw them" ZUrI—t—hum~*zuri-t—hum "I visited them" In (2.A,8,C), the attachment of the consonant-initial suffix creates a two-consonant cluster. Accordingly, the stress assignment rules predict that the vowel before the cluster be assigned stress. But contrary to this prediction, the vowel is not assigned stress. In this behavior, the vowel patterns with epenthetic vowels. To account for the appearance of the vowel in the form of [i] instead of [u] as in fLin/ "raddish", We assume that the vowel Rule (3) apt context is rela‘ minimized, inc lfujll, and lbu that /r/ is [+bf contexts whic native speake. which satisfy (IA) are acco 149 appears as [i] in the context of unemphatic coronals. This is rule (3) of epenthesis, which properly includes and so takes precedence on (26). Rule (3): B —--> V / V C __ C C [+high] [+high] [+cor] [+cor] {-back] [+back] {-back] {-back] Rule (3) applies in an environment in which the consonantal context is relatively complex and therefore vowel distinctions are minimized, including the contrast of [i] and [u]. Rule (3) applies to /fujl/, and /bu§l/, yielding [fujil] and [busil], respectively. Note that /r/ is [+back] like an emphatic consonant. Note also that the contexts which satisfy the SD of rule (3) are very limited. As a native speaker informant of JA, I cannot think of any other words which satisfy the SD of the rule. On these grounds, the words in (IA) are accounted for. In (1.8), to account for the alternation of the vowel, we assume that the vowel appears as [u] in compliance with rule (26) and that it appears as [i] under the influence of borrowing from other Arabic dialects in which the corresponding vowel is [i]. In the urban dialects, the vowel is [i] as in xUDiz. This also reflects a language change which goes in the direction of using [1] instead of [u]. This change is probably initiated under the influence of the other dialects with which JA is in contact. I assume the prnounciations with [1] represent code-switching and as such are outside the rules pIII (its... In (I .C), the VI of the vowel in t appearance in thi with [i], as I hav such are outside I To sum up, tIi indicate that tht yonger generatii progress in whi contexts, repIaI explantion for t grounds, we eat exceptions to r [5.2. Data prob] In this sec Counter 8V1del cOneider the : 150 of JA. In (1C), the vowel alternates between [u] and [i]. The appearance of the vowel in the form of [u] is predictable by rule (26), but its appearance in the form of [i] is not. I assume the prinvunciations with [i], as I have indicated above, represent code—switching and as such are outside the rules of JA. To sum up, the analyses of the data in (I.A,B,C) and (2.A,8,C) indicate that the vowel tends to show up as [i], especially among the yonger generation. We take this as evidence of a language change in progress in which the use of the [1] form is generalized to more contexts, replacing [u] which tends to lose ground to [i]. A possible explantion for this is that [i] is less marked than [u]. On these grounds, we can say that the data in (I.A,B,C) represent principled exceptions to rule (26) of epenthesis. 5.2. Data problematic to rules (7', 32, 34, and 40): In this section, we consider some data which are potential counter evidence to rules (7‘), (32), (34), and (40) in Chapter 4. Consider the data in (4.A,8,C,D). Stress is regular and not indicated. (4.A):gabuq "peeling" galub "heart" salub "crucifiction" safuh "forgiveness" tasum "taste" rafus "lifting” salib 'robl safih 'fooi daeim "co nafis 'ben (4.C):saaris “st Eaabik "e' faaxif "I saadir “i maakir ‘ raafi’b (4D)§arus "I l sabuk “e faxur "p éabur “s makur " rafu’g The data It We] But co 151 (4.8): nabiq "spring" calib "dog" salib "robbing" safih "foot of a hill" dasim "collision" nafis "benefit" (4C): saaric "street" saabik "entangled" faaxif "valuable" saaoir "vacancy" maakir "treacherous" raafi’b "rejecting" (4.D):§arue "law" sabuk "entangling" faxur "pride, bragging" sazsur "soaring" makur ”treachery" rafu’o "rejection" The data in (4A) satisfy the SD of rule (7') which predicts an [i] vowel. 8ut contrary to the predictions of rule (7'), the vowel is [u]. In (4.8), the directly preceding and following consonants are identical to those of the corresponding vowel in (4A), but the vowel is [i] as rule (7') predicts. The data in (4C) satisfy the SD of rules (32), (34), and (40-) which predict an [u] vowel. But contrary to the corresponding CU To account f e have to take IntI In (4A), the init (4B), it is neitli affects the f olIi context of a prt vowel is I-back 152 predictions of these rules, the vowel is [i]. In (4D), the directly preceding and following consonants are identical to those of the corresponding consonants but the vowel is [u] as the rules predict. To account for the appearance of the vowel as [u] in (4A), we have to take into consideration the quality of the preceding vowel. In (4A), the initial consonant is either emphatic or velar, while in (4.8), it is neither emphatic nor velar. In (4A), the initial consonant affects the following low vowel which becomes [+back] in the context of a preceding velar or emphatic. In other contexts, the low vowel is {-back], [ae]. This is rule (5) (repeated from (2) in Chapter 2). Rule (5): [+syll] ———> [+back] / I-syll] __ [-high] [+back] Rule (5) predicts the low preceding vowel to be [+back] after velars and emphatics. To account for epenthesis of [u] in this context, we postulate rule (6). Rule (6): 8 ———> V / V C __ C C [+high] {-high] [+back] [+back] RUIe (6) predicts the vowel to be [u] in the context of a preceding [~high, +backl vowel as in (4A). Rule (6) does not apply to the data Rules (6) abo simplified into a Rule (7): 13 -——> Role (7), wt general than at context of a pr contexts of bot [+high, +backl : In (M) are ac 153 in (4.8) because its SD is not satisfied,- the preceding vowel is {—back] Rules (6) above and (26) of epenthesis in Chapter 4 can be simplified into one rule. This is rule (7). Rule (7) B ———> V / V C ___. C C [+high] [+back] [+back] Rule (7), which is a simplification of rules (6) and (26), is more general than either rule. Rule (7) predicts an [u] vowel in the context of a preceding [+back] vowel. This captures the preceding contexts of both rule (6) and (26) in which the preceding vowel is [+high, +backl and [—high, +backl, respectivly. On this basis, the data in (4A) are accounted for. To account for the [i] vowel in (4C), we consider the data in (BA-8), which include the same roots as (4C) and (4D), respectively. (8A): saaris+hum "their street" saabik+hum "entangling them" faaxir+hum "their valuable thing" saaoir+hum "their vacancy" maakir+hum "their treacherous one" raafitg+hum "rejecting them" faxur—hum tenor-hum makur-nun refuti—hum In (8A), the rules predict. 1 consonant-initi (88), the corre that the status that of the cor assume that tI the correspont '0] I‘vlienthesis IIDDIy to a clu: there is no so CIuster. Thu. 154 (8.8): sarus~hum (*saruq~hum) "law" sabuk—hum (*sabuk—hum) "entangling" faxur—hum (*faxur—hum) "pride, bragging" saour-hum (*satSur-hum) “soaring" maI [+stress] / ___l X [imperfect ~V Rule (13) of stress morphologization assigns stress to the prefix of the verb in (l2.C,F). In their application, morphologically conditioned rules take precedence over other rules. This includes the general stress rules postulated in Chapter 3,- those rules are not morphologically conditioned. On these grounds, rule (13) which is morphologically conditioned applies before the general stress rules. To account for the absence of the first vowel, we postulate rule (14) of high stem vowel deletion (HSVD). Rule (14): v———> e / __ c v CIstemC [+high] Rule (14) accounts for the absence of the first vowel in the context of a consonant—initial suffix as in (10.8,E), or no suffix as in (10,A,D) if we think of a word boundary as equivalent to a consonant. To account for the absence of the second vowel, we postulate rule (15) of high stem vowel deletion (HSVD). Rule (15): V "—‘—'> g / ___...— C] stem V [+high] By using the conapsedinto on Rue(16x V [+high]) Rule (16) a contexts. In(lT),vn representation ofthe rules vvi (ll): lea-nus ?a~nust ia-nus 162 Rule (15) accounts for the absence of the second vowel in the context of a vowel—initial suffix as in (10.C,F). By using the angled brackets notation, rules (14) and (15) can be collapsed into one rule. This is rule (16). Rule (16): V —-——> e / ___... Clstem[<‘> +syll [+high] Rule (16) applies to both vowel-initial and consonant-initial contexts. In (17), we present the derivation of the phonetic representation (PR) of /7a-nusuf—ak/, /7a—nusuf—hum/ on the basis of the rules we postulated: under the underlying vowel analyses. (17): /?a—nusuf—ak/ /?a—nusuf-hum/ UR, rule (I I) ea—nusuf—ak --- Rule (13): Stress ea—nusf-ak --- Rule (16):HSVD —-— 7a—nusc’uf—hum Rule (15) in ch. 3: Stress ——— 7a—nsUf-hum Rule (16): HSVD [ea—nusf—ak] [ea—nsuf—hum] PR In (17), under the assumption that the first vowel is underlying, the application of the stress and high vowel deletion rules to /?a—nusuf—ak/ and /?a-nusuf—ak/ yield the correct PR. deletion, respec before the genei requirement cor rules is the grit before other ru .However, so that the first i represents ste‘ (18): o'a-smae- . ea-emal oa-Erab- ea-rfas In (18), ii The stem of I the context t evidence for 163 Morphologically conditioned rules, that is rules (13) and (15), pre—empt the general rules of stress and rule (16) of high vowel deletion, respectively. The analyses require that rule (13) apply before the general stress rule. One interpretation of this requirement consistent with the exclusion of extrinsic ordering of rules is the principle that morphologically conditioned rules apply before other rules. However, some questions can be raised against the assumption that the first vowel is underlying. Consider the data in (18) which represents stems with low vowels. (18): 7a-smas-ak~*7a—sams—ak "I hear you" 7a-smal-ak~*?a-qamI-ak "I do yoy" ea-srab—ak~*?a—sarb—ak "I drink you" e'a—rfas—akfivé—rafs—ak “I lift you" In (18), the low stern vowel gives no evidence to alternation. The stem of the verb has only one vowel which does not alternate in the context of vowel or consonant—initial suffixes. The data give no evidence for the existence of a vowel between the first and second radicals. This argues that verbs with an underlying [~highl vowel have an underlying stem of the form /—CCVC—/. Thus under rule (11),we miss a generalization that the underlying form of the imperfect verb stern in Arabic is /—CCVC—/. To maintain this generalization under the underlying vowel analyses, we assume that the underlying form of the imperfect stem in Arabic is as in (19). In (19), the (19) has some it related to the S and second radi third radicals, postulate rules [In the basis of position, but it Under the I the vowel bet‘ I second and th‘ Rule (20); lmg 164 (19): Imperfect verb stem: /—V, CCC—/. In (19), the position of the stem vowel is not specified. Rule (19) has some advantages over rule (I I). In (19), the problems related to the simultaneous presence of the vowel between the first and second radicals, on the one hand, and between the second and third radicals, on the other, do not arise. We do not have to postulate rules which delete either the first or the second vowel. On the basis of rule (19), the vowel can only appear in either position, but not simutaneously in both. Under the assumption in (19), to account for the appearance of the vowel between the first and second radicals and between the second and third radicals, we postulate rules (20) and (21). Rule (20): Imperfect verb stem: - C V C C-/ -V [+high] Rule (21): Imperfect verb stem: —CCVC— Rule (20) accounts for the imperfect stem of verbs with a high stem vowel in the context of a vowel—initial suffix as in (10C). Rule (21) accounts for the imperfect verb stem of verbs with a high stem vowel and in the context of a consonant—initial suffix as in (10D). It also accounts for the imperfect stem of verbs with a [-high] stem vowel as in (18). Rule (20) properly includes rule (2I) and applies disjunctively in relation to it. Rules (20) and (21) over rule (21). l and a vowel-iniI consonant initia as in (IO.A,D). . The function IInearize a item does not arise rue(19) The Accordingly, n In (22), we zénusfak '1 e assumptions t (22): ?o-nusf-a té-nusf-t 165 comply with the analysis of Hudson (1985) of Arabic root and pattern morphology. In their application, rule (20) takes precedence over rule (21). Rule (20) applies in the context of a high stem vowel and a vowel-initial suffix, and rule (21) else where. This includes consonant initial contexts and contexts where no suffix is present as in (IO.A,D). The function of rules (20) and (21) in relation to rule (19) is to linearize a non-linearized input. The need for this Iinearizarion does not arise if we postulate rules (20) and (21) independently of rule (19). The postulation of rules (20) and (21) pre—empt rule (19). Accordingly, we dispense with it. In (22), we illustrate the derivation of the imperfect verbs 7anusfak "I explode you" and 7ansdfhum "explode them" under the assumptions of rules (20), and (21). (22): 7a-nusf—ak (Rule 20) m-nsuf-hum (Rule 21) ea—nusf—ak --- Rule (13): Stress ——— ea-nsUf-hum Rule (15) in ch. 3: Stress [ea-nusf-akl [ea-nSt'Jf-hum] PR The analysis in (22) yields the correct results. However, a question related to stress assignment might be raised against the analyses based on rules (20) and (21). In (22), stress is assigned to the prefix vowel vowel by a morphologically conditioned rule. That prediction. [In I reconsidered in [S the metathei 5.3.2.The meta To account (23) of metatl Rule (23): C 166 is rule (13). In our analysis, we resort to this morphologically conditioned rule when other analyses fail to yield the correct prediction. On these grounds, the underlying analysis will be reconsidered in the light of the findings of the other analysis, that is the metathesis and epenthesis analyses. 5.3.2. The metathesis analyses: To account for the derivation of the vowel, we postulate rule (23) of metathesis. Rule (23): C C V C[imperfect ‘V Under the usual assumption about the stem shape of the imperfect, that the form of the underlying representation of the imperfect stem is /-CCVC-/, the verb forms of the data in (10.C,F) satisfy the SD of rule (23) of metathesis. Rule (23) of metathesis is similar to the rule of metathesis posited by Alghazo (1987: 92) in his analysis of JA except for the stem boundary and the features which specify the identity of the metathesized vowel. Alghazo's rule does not include the stem boundary in its SD, and the metathesized vowel is [+high,—long,-stressl. In his analysis, Alghazo (1987: 92~94) argues against a metathesis rule. His main argument against metathesis is I l rule to superfig This argum‘ show that the i Mammmw always accoun some contexts by the rules a deletion f eeds consonant clu But the rule t appearance a The rules of pé-iutm-ak 016 preceoii environment the environ) VOWSIS by n In (24), be applying 167 that the function of a rule of metathesis is exactly the same as that of high vowel deletion (sync0pe in his terms) followed by epenthesis, which he argues are independently motivated (see Alghazo,1987). Accordingly, he views metathesis as an additional rule to superficially account for certain facts in the language. This argument against metathesis can be successful if we can show that the vowel in question can be accounted for by applying the rules of high vowel deletion and epenthesis. But these rules do not always account for the appearance of the vowel in question. In some contexts, the appearance of the vowel cannot be accounted for by the rules of sync0pe and epenthesis. The rule of high vowel deletion feeds into the rule of epenthesis,- it creates a three consonant cluster which satisfies the SD of the rule of epenthesis. But the rule or rules of epenthesis fail to account for the appearance of the vowel in the form of [u] in contexts such as (10F). The rules of epenthesis predict the vowel to be [i] in (10F), but in ?a-§utm—ak "I curse you" as in (10F) the vowel is [u] in the context of a preceding and following coronal. That is, [u] appears in the environment (10F) in exactly and only the stems which have [u] in the environment (10E). The [u] of (10E) and (10F) are the same vowels by metathesis. In (24), we present the derivation of ?a-s'utm~ak "I curse you" by applying the rules of high vowel deletion and epenthesis. (24): /7a—§tum-ak/ UR (assumption 20) ea—stum—ak Rule (I3): Stress 7a—stm—ak Rule (15): HSVD In [24), rule (TI predicted an [I] further discuss Under the rt harmony (see b: ‘ilob of three ru the epenthesis deletion, epen' metathesis be vowel harmon . In (25), v nDDIying rule 168 7a—sitm—ak Rule (7') in ch.4: Epenthesis *[e’a—sitm—ak] PR In (24), rule (7') of epenthesis yields the wrong prediction: it predicted an [i] vowel instead of [u]. The epenthesis analysis is further discussed in (5.3.3). Under the metathesis analysis, the need for a rule of vowel harmony (see below) does not arise. The rule of metathesis does the job of three rules-~deletion, epenthesis, and harmony as it arises in the epenthesis analysis. On the other hand, the rules of high vowel deletion, epenthesis, and vowel harmony can not do the job of metathesis because the rule of deletion bleeds the context of the vowel harmony rule. In (25), we present the derivation of the PR of /?a-stum—ak/, aIDIDIying rule (23) of metathesis. (25): /ea-stum-ak/ UR ea—stum—ak Rule (15, Ch. 3): Stress 7a-sutm—ak Rule (23): Metathesis [7a—sutm~akl PR The analysis in (25) yields the correct prediction if we impose a condition on the application of the stress rule and assume that it applies only once. Without this condition stress will have to shift to the penultimate vowel, yielding *[ea—EUtm-ak]. In our analysis, however we assume that rules apply whenever their SD is satisfied, but constraints on the application of rules are the principle w take precedence metathesis anat assignment of I principle of rul (25) yields the ordering const postulated in I stress assign) rule of stress SD of both rul aDuly once, a the prefix vo Imet. Accord: metathesize 169 allowed in the form of some general principles which are linguistic universals such as the principle of Proper Inclusion Precedence and the principle which requires that morphologically conditioned rules take precedence. On these grounds, we have either to reject the metathesis analysis or think of another way to account for the assignment of stress. An alternative analysis follows the latter principle of rule application. Under this assumption, the analysis in (25) yields the correct result without the imposition of a rule ordering constraint. In (25), both rule (15) of stress assignment postulated in Chapter 3 and the morphologically conditioned rule of stress assignment in (13) assign stress to the prefix vowel. The rule of stress and the rule of metathesis apply simultanously; the SD of both rules is met. Without the constraint that stress rules apply once, a problem arises if we assume that stress is assigned to the prefix vowel by rule (15). In 76§utm~al<, the SD of rule (15) is met. Accordingly, rule (15) predicts the shifting of stress to the metathesized vowel. To account for the problem related to the failure of the metathesized vowel to be assigned stress, we consider two analyses. The first is that rules of stress assignment are extrinsically ordered and that they apply before the rule of metathesis and only once. The second is that metathesized vowels are not assigned stress because stress in imperfect verbs is morphologically conditioned as in rule (13). We reject extrinsic ordering of rules because it is an arbitrary constraint the main motivation of which is achieving observational adeqacy. Extrinsic ordering of rules allows the application of rules apply whenever Linguistic thee) power. The all] general princip on the nine mophologically phonologically assuming rule in which it op (15). Still a on conditioned s (26): ea-st’a- sa-sta- 7a~sta 170 in the order we want, just in order to get the desired results. In our analysis, we assume that rules must be intrinsically ordered; rules apply whenever their SD is met (Koutsoudas, et al, 1974: 8). Linguistic theory as such is simpler, and has more explanatory power. The application of rules has to be constrained only by general principles, which have to be universally valid. On the other hand, if rules are ordered intrinsically and mophologically conditioned rules take precedence over phonologically conditioned rules, the correct forms are generated assuming rule (13), which pre—empts (15) in all lines of derivations in which it applies, including (25), except with (13) rather than (15). ' Still a question related to rule (13) of morphologically conditioned stress assignment can be raised. Consider (26). [(26): aa-sta-gbil "I welcome" 7a-sta—gbil—hum "I welcome them" 7a—sta-gibl-ak “I welcome you" In (26), stress assignment adjusts to the changes in the structure of the word. A problem arises in the assignment of stress to the word, 7a-sté-gbe-ak. The general stress rules, that is rule (15) in Chapter 3, incorrectly assign stress to the high penultimate vowel, yielding *7a-sta-gl’bI-ak. The stress morphologically conditioned rule, that is rule (13), assigns stress to the prefix vowel, yielding *7a-sta-gbiI-ak. To account for this problem, we assume that the variable "[ X l" in rule (13) stands for a single Igl in 7a-sté-gl evidence to NH 0)) these gr account for the radicals in the? stress assignn assignment wI stem. Now, tb metathesized. notational, no advantage of imperfect ve Imperfect ve attached and Imperfect Vi metathesis more genera underlying, framewcrk evaluated. l7i morpheme. Under this interpretation of "i X 1", rule (l3) will assign stress to the vowel preceding the first radical of the stem, which is [gl in 7a-sta-gbr‘I—ak. Thus the data in (26) are not counter evidence to rule (13) of stress assignment. On these grounds, we conclude that rule (23) of metathesis can account for the presence of the vowel between the first and second radicals in the context of a vowel-initial suffix. The question of stress assignment is accounted for bg rule (13) of stress assignment which assigns stress to the vowel preceding the verb stem. Now, the question arises whether the vowel is underlging or metathesized. The differences between the two analgses mag be notational, but the metathesis analgsis mag be said to have the advantage of capturing more generalitg about the structure of the imperfect verb stem. In the underlging analgsis, we argued that the imperfect verb stem is -CVCC- when a vowel—initial suffix is attached and -CCVC— elsewhere. But in the metathesis analgsis, the imperfect verb stem is -CCVC—.in all contexts. This argues that the metathesis analysis has to be preferred on the grounds of capturing more generality about the imperfect verb stem. In Chapter 6, the underlying, metathesis, and epenthesis analgses (within the framework of both linear and non—linear phonologg) will be further evaluated. 5.3.3. The epenthesis analysis: Under this assumption, we have to give evidence which argues that the vowel is predictable. But as we have argued in 5.3, the fwmsofUth (27A): gi-kisr-i ti-kisr-a ni-kisr- 9i-kisr~ (275):7a-kisr (27.C):gi-ksir n-ksir ni-ksh 9i-ksh (27D):?a-ks (27.'E) till-nu td-nu nu—ni ?t'l~ni (27-G)1UU—r tu~i 172 rules of epenthesis postulated in Chapter 4 fail to predict the appearance of the vowel in the form of [u] in the imperfect verb forms of (iO.F). Consider (27.A—H). (27A): yi—kisr-ak "he breaks you" ti-kisr-ak "she breaks you" ni—kisr-ak "we break you" 7f-kisr~uh "you sg. break (imperative) him" (27.8): ea—kisr-ak "1 break you" (27C): yi-ksir-hum "he breaks them“ ti-ksir-hum "she breaks them" ni—ksir~hum "we break them" ei—ksir—hum "you 39. break (imperative) them" (27D): ea-ksirehum "I break them" (27E): yu-nusf—ak "he blows you up" tt’r-nusf—ak “she blows you up“ nu-nusf—ak "we blow you up" eu—nusf—uh "you $9. blow (imperative) him up" (27F): ea-nusf—ak "l blow you up" (27.6): yu—nsuf—hum "he blows them up" tu-nsuf—hum "she blows them up” (27H): ea-nsfif- In (27A), 1. vowel; both vo, shows up betw consonant-inii (27.B,D,F,H) is identical to ti to the vowel ‘ in the contex the prefix vo Likewise, in second vowe quality of a Mel, and t M the bran because the To acco redressive Rule (28): 173 nu—nsuf—hum "we blow them up" eu-nsdf—hum "blow them up (sg. imperative)" (27.H): ea—nsc’rf—hum "I blow them up" In (27A), the vowel in question is identical to the high prefix vowel,- both vowels are [i] and are identical to the vowel which shows up between the second and third radicals in the context of a consonant-initial suffix as in (27C). The low prefix vowel of (27.B,D,F,H) is invariably [a]. In (27E), the vowel in question is also identical to the prefix vowel; both vowels are [u] and are identical to the vowel which shows up between the second and third radicals in the context of a consonant—initial suffix as in (27.6). In (27.A,C), the prefix vowel, the first vowel and the second vowel are all [1]. Likewise, in (27.E,G), the prefix vowel, the first vowel and the second vowel are all [u]. On these grounds, we assume that the quality of a prefix high vowel is determined by that of the second vowel, and that the quality of the first vowel is determined by that of the prefix vowel,- it cannot be determined by the second vowel because the two 'stem vowels do not show up simultaneously. To account for the prefix vowel, we postulate rule (28) of regressive high vowel harmony (RHVH). Rule (28): V -——> [aback] / it C _- C C v [+high] [+high1 {aback} prefix vowels, high stem vow radicals, yield time, rule (28 harmony whic epenthesis be: Rule (29): I+si [+h' 174 Rule (28) of regressive high vowel harmony applies to the imperfect prefix vowels, harmonizing the prefix high vowel to the underlying high stem vowel which appears between the first and second radicals, yielding the correct form of the prefix vowel. At the same time, rule (28) of (RHVH) feeds into a rule of progressive vowel harmony which applies to the vowels inserted by the rules of epenthesis between the first and second radicals. This is rule (29). Rule (29): {+syll]———> [aback] / iiC V —C _. C C [+high] l+highl [aback] Rule (29) of PHVH harmonizes the epenthetic vowel to the prefix vowel. THis accounts for the correspondence between the forms of the first and second vowels which appear in the verb stem in the context of a preceding high prefix vowel. But the problem is to account for the quality of the high vowel which appears between the first and second radicals in the context of a preceding i-highi prefix vowel where the SD of rule (29) of PHVH is not satisfied as in (27.B,F). In (27.B,F), the first vowel can not be determined by the prefix vowel which is [a]. This is illustrated in the derivation of yi—kisr-ak, 76-kr'sr-ak, yd-nusf—ak, and 7a-nusf-ak in (30). gi-ksr-ald gi-kisr-a (30):yi-ksir—ak yi-ksir—ak yi-ksr-ak yi-kisr-ak [yi—kisr-ak] [ea-kisr-akllyu—nusf—ak]*[9a—nisf-akl ea—ksir—ak ea—ksir—ak 7a-ksr-ak ea-kisr—ak 175 yi—nsuf—ak yi-nsuf—ak yu-nsuf-ak yu-nsf—ak yu—nisf-ak yu—nusf-ak ea-nsuf-ak ea—nsuf-ak 7a—nsf-ak 7a—nisf—ak In(30),rules(28)of RHVtIandrule(l6)ofIISVD apph; UR Rule(lS) Stress Rule (28): RHVH Rule (16): HSVD Rule(7vz Epen. Rule (29): PHVH PR shnultaneously. After the application of (lS)cn stress assignrnent, ‘the SEJOf both rules is satisfied. Rule (lS)is hie feeding relation to rule (l6).iAccordingly,they areintrinsicaly ordered. 0n the other hand, rule (16) of HSVD feeds into rule (7') of epenthesis whichalsofeedsintorule(29)ofPH\di Rule(29)ofPHVtI harmonizes the high vowel inserted by rule (7‘) of epenthesis in the context of a preceding high suffix vowel but it does not harmonize the epenthetic vovvelin the context oftaprecedingl-highlprefix vowelasin zé-nisf—ak becauseits SDis notsatnfiied Accordingly, the vowel ii] does not harmonize with the preceding l-highlprefix vovvel. This yields *lza—nisf-akl,notlea-nusf—akl,the correct forni On these grounds, we tentatively conclude that the high vowel unpred the framework analgses accor metathesized. the vowel. 5.4. Conclusio- The analg counter evide indicate that first type ap Alternating I accordingly appear betw stem are un We consider verb, in the the vowel ; 176 which appears between the first and second radicals in imperfect verbs is unpredictable by the rules of epenthesis postulated within the framework of linear phonology. This argues in support of the analyses according to which the vowel is either underlying or metathesized. In Chapter 6, we consider the non—linear approach to the vowel. 5.4. Conclusion: The analysis of the problematic data which pose potential counter evidence to the rules of epenthesis postulated in Chapter 4 indicate that there are two types of alternating high vowel. The first type appears in nouns and the second type appears in verbs. Alternating high vowels which appear in nouns are predictable and accordingly they are epenthetic. But alternating high vowels which appear between the first and second radicals in the imperfect verb stem are unpredictable by the rules of epenthesis and accordingly we consider them either underlying or metathesized. In this type of verb, in the context of a vowel-initial suffix, stress is assigned to the vowel preceding the stem by a morphologically conditioned rule. In this cha vowel in ones. (I987) of JA . first section) This includes postulated to the second St to the derive the autosegr ' linear appro Chapter 6 A Non—linear Approach In this chapter, we review the autosegmental approach to the vowel in question. Our review is based on the analysis of Alghazo (1987) of JA data. The chapter is divided into' three sections. In the first section, we present the autosegmental approach to the vowel. This includes a presentation of the autosegmental rules which are postulated to account for the derivation of the vowel in question. In the second section, we apply the rules presented in the first section to the derivation of the vowel in some JA data. In the third section, the autosegmental approach is evaluated in comparison with the linear approach presented in Chapters 4 and 5. 6.1.The non-linear approach: In this section, we present the rules involved in the derivation of the vowel. This includes rules of syncope, epenthesis, association, default, tier conflation, emphasis spreading, emphasis rounding, and vowel harmony. These rules are either as postulated by Alghazo or on the basis of his analysis when the rule is informally presented. 6.l.l. Syncope: Syncope is viewed by Alghazo (1987:100) as the deletion of a vocalic slot from the skeletal tier which is associated with a 177 syllable (Algha formal identifi slot is not affa not provide us syncope. Butt; autosegmental In (I), the vc the skeletal floating. In derivation c It feeds rul 178 matrix containing a l+highl feature in the appropriate environment. The appropriate environment is that of an unstressed, nonfinal, open syllable (Alghazo referred to the syllable without providing a formal identification of it). The matrix associated with the deleted slot is not affected, but remains floating on its tier. Alghazo did not provide us with an explicitly formalized autosegmental rule of syncope. But on the basis of his analysis, we provide the autosegmental syncope rule in (1). Rule (1): V—-—> B / __ C V I {—strxess] [+high] In (I), the vocalic slot of an unstressed high vowel is deleted from the skeletal tier and the matrix "X" associated with it remains floating. In (3), we see the application of rule (1) of syncope in the derivation of the words in (2). Rule (15) precedes rule (1) because it feeds rule (I). (2): fihim “he understood" fihm—ak "he understood you" fariig-ak “your team" katab—ak "he wrote you" lfihiml l, Rule(l)oi katabak beca vowel is l+hig fariigak, the is [+stress], I the SD of the l-highl. IntE fI'hmak whe CV. EDenthe empty voca environmei his analys Rule (4): I79 (3): /fihim/ /fihim~ak/ /fariig—aI V / C___. C t } #. < grind-l fihhg llluulh the CV tier are epenthesis is v the CV tier. TI rule (4); the all rules which w association, q' obligatory co 6.1.3. Associi The rule a melodic tier floating mat ,SUnCODe Wll' associate 180 In rule (4), only the CV tier is provided; the matrices on the melodic tier associated with the consonantal and vocalic slots at the CV tier are not provided because they are not involved since epenthesis is viewed as the insertion of an empty vocalic slot into the CV tier. The matrices in the melodic tier are not determined by rule (4); the quality of the inserted vowel is determined by other rules which will be discussed later. This includes the rules of association, default, and Spreading in accordance with the Obligatory Contour Principle (OCP). 6.1.3. Association: The rule of association associates an unassociated matrix on the melodic tier to an unassociated slot on the CV tier. Accordingly, the floating matrix which results from the application of rule (I) of syncope will be associated with the unassociated vocalic slot inserted by rule (4) of epenthesis (Alghazo, l987: 102). This | formalize as rule (5) (Alghazo, 1987: 102). Rule (5): X l vctC {C}———> c 2% 457! CVC In (5), the "X" on the left stands for a floating matrix of a [+high] vowel the vocalic slot of which on the CV tier is deleted. Rule (5) associates the unassociated floating matrix to the empty vocalic is one. The ass feature matrix (It? according. represented se This is illustr? ré-kutb-ak “ _...——-< Iisr 181 slot inserted by rule (4) of epenthesis on the CV tier. In the context where there is not an unassociated matrix, the vocalic slot is associated with the matrix of a high vowel on the same tier if there is one. The association of two vocalic slots on the CV tier with one feature matrix on the melodic tier takes place in compliance with OCP according to which two adjacent identical segments can not be represented separately on the same tier (Alghazo, 1987: 108—9). This is illustrated in the derivation of the words jl’sim "body" and 76—kutb-ak "I write you" in (6). (6): /j s m/ /7a- k t b —ak/ UR I I I I I I C V C C C C V C i u jism ea-ktub—ak Rule (15) in Ch.3: Stress -—~ ea- k t b -ak Rule(l):Syncope III C C C 182 j s m ea— k t b -ak Rule(4):Epenthesis CVCVC CVCC l I j s m ea- k t b -ak Rule(S):Association IC V (I: V l IC V (II lC l/ \. [jisiml [ea—kutb—akl PR In the derivation of jisim, rule (I) of syncope does not apply. Rule (4) of epenthesis applies, inserting an empty vocalic slot into the CV tier. The empty slot is then associated with the matrix of the preceding high vowel on the same tier as rule (5) of association predicts. This complies with the OCP. In the derivation of za-kutb—ak, rule (I) of syncope applies to the CV tier of the unstressed high vowel which appears between the second and third radicals, deleting its vocalic slot on the CV tier and leaving behind it its feature matrix floating on the melodic tier. This creates a cluster of three consonants which satisfies the SD of rule (4) of epenthesis. Rule (4) applies, inserting an empty vocalic slot between the first and second radicals. The inserted empty vocalic slot is then associated with the floating matrix, which is generated by rule (1) of syncope, as rule (5) of association predicts. matrix, def out This is based I specified vow is rule (7) (Al If there is Rule (7): T0 illus derivation o 183 6.1.4. Default: if there is neither a high vowel on the same tier nor a floating matrix, default rules will introduce the matrix of the vowel [i]. This is based on the assumption that the vowel [ii is the least specified vowel underlyingly and has no features in its matrix. This is rule (7) (Alghazo, 1987: lOl). Rule (7): V I l———> [i] To illustrate the application of rule (7), we present the derivation of the word c'élib “dog“ in (8). (8): b/ UR calb Rule (15) in Ch. 3: Stress In em datault n emphasi s 184 a l b Rule (4): Epenthesis C V C V C I a c l b Rule (7): Default C V C V C I | a ll] [Calib] PR In (8), the empty vocalic slot inserted-into the CV tier by rule (4) of epenthesis does not have a feature matrix on the melodic tier. On the other hand, there is not a floating matrix with which it can associate according to rule (5). In this context, rule (7) of default comes to the rescue. Rule (7) of default inserts the feature matrix of the high vowel [i]. In this way, the correct phonetic realization of /calb/ is realized. 6.1.5. Emphasis spreading, rounding and backing: In emphatic contexts, the vowel [i] which is introduced by default rules gets rounded,- the vowel first undergoes rules of emphasis spreading (Alghazo, 19872130) and then emphasis rounding ‘ Iam ovei reference to emphasis sp and E stands to segment: split“. Thai 185 provided in (9) and (l 1), respectively (Alghazo, 1987: 130, I35). Rule (9): [X X X X Xlstem*——>[X X X X Xlstem l \ I I I I I \ r I I I I ‘ ' I [I I, : “ : I’ll/l [ llf’ I ' I VII”, I v E E I am overlooking the problem that could be raised by Alghazo's reference to the category "stem" in the SD of the rule. in rule (9) of emphasis spreading, the X slots stand for any segment in the stem, and g stands for emphasis. That rule (9) of emphasis spread applies to segments in the stem is illustrated by the JA word [ti-stahl "to split". That rule (9) applies to epenthetic vowels in the stem is illustrated by the JA word [tabul] "drum". In (IO.A,B), we illustrate the application of rule (9) to ti-stah, and tabul, respectively. (10A): ti—s t a h————>ti—s t a h- : I x I \ I I/’ a \"I I \IIII : \\:I/’ ll : ‘I E E In (IO.A),. the stem as I nuspmadu But it spreat ofemphasis consonantlg labil and e word. Alghi 0T the 1 pop that an em; rounding, 1 135). Rule(ii); m-—-——— 186 (IOB): fl 0) U" I l I V {-0- CD U" m -_-—___...— mv———-——- . \‘ In (IO.A), emphasis spreads only to to the segments which are in the stem as rule (9) of emphasis spreading predicts. Emphasis does not spread to to the prefix 1.1; because it is not part of the stem. But it spreads to the vowels and consonants in the stem. The spread of emphasis to the stem consonant /s/ yields the emphaticized consonant [s]. In (10.8), rule (9) of emphasis spreading applies to tabil and emphasis spreads to the vowels and consonants in the word. Alghazo provides us with no evidence about the phonetic form ‘of the j_ vowel which is in the domain of emphasis. But he argues that an emphaticized 1 vowel gets rounded by a rule of emphasis rounding. This is rule (11) of emphasis rounding (Alghazo, I987: 135). Rule(il): i———>u/[XaC__Y] m-_—_—_..—— In rule (I I), “I rounds the emf emphasis to tI environment f rule (I 1). However, rounded in an leiyxarbitl "tI get rounded t problem, Alg to he market vowel in we rounded if i‘ rounding“. Alghazo it is ad hoc facts in thr which accc examples. It has to b 187 In rule (I I), "X" and "Y" are variables within the stem. Rule (1 1) rounds the emphaticized 1 vowel, yielding [u]. The spread of emphasis to the vowel 1 by rule (9) creates the apprOpriate environment for rule (1 I) of emphasis rounding,- rule (9) feeds into rule (1 1). However, there are contexts in which the vowel 1 does not get rounded in an emphatic context. For example, the Ii] vowel in [eiyxarbit] "to confuse" is in an emphatic context. But it does not get rounded by rule (1 i) of emphasis rounding. To account for this problem, Alghazo (1987:134) says that underlying /i/ vowels have to be marked with [-ER], "-emphasis rounding". Accordingly, the i vowel in words such as [eiyxaribtu] "to confuse him" does not get rounded if it is underlying and so marked with [-ER], "—emphasis rounding". Alghazo (1987: 137) rejects this assumption on the grounds that [it is ad hoc in the sense that it is adopted to account for certain facts in the language and argues that a better solution would be one which accounts for the facts and explains the apparent counter examples. He claims that a non—linear analysis will do this. (Here, it has to be noted that rule (9) of emphasis spreading is a non-linear rule. Alghazo seems to have forgotten this.) In his non—linear analysis, Alghazo (1987: 145) argues that emphasis be represented by the feature [+back]. Accordingly, rule (9) of emphasis Spreading is revised into rule (12) of emphasis backing (Alghazo, 1987: 159). 188 Rule(12): X X X X X———> X X X X X a b c d e a b c d e r \ r I I I I ‘\ I ’I ’I I” : \‘ : “”’,” I ‘I’l" I \III,’ r V“ [+back] [+back] The rule of emphasis backing as postulated by Alghazo does not include the input provided in rule (12),- only the output of the rule is provided by Alghazo. In (12), the X's represent the vocalic and consonantal slots on the CV tier, and the small letters represent the feature matrices of the consonantal and vocalic segments on the melodic tier. "b“ is the feature matrix of an emphatic consonant (The emphatic feature could be on "a", "c", 'd" or "e“.). In rule (12), the feature [+backl Spreads from an emphatic segment to the other segments within its domain. The function of rule (12) of emphasis backing will be illustrated in the derivation of the epenthetic vowel in (17). Rule (12) of emphasis backing feeds into the redundancy rules provided in (15). The features inserted as a result of the application of the redunduncy rules takes into consideration the feature [+backl provided by rule (12) of emphasis backing. Before seeing this, we present some related principles which are invoked in the non—linear approach. Since vowels on for rule (9) of empha tiers conflate into a} emphasis to spread 1 they are no longer or In (13). Rule (13): o——-———n n———————(‘) 189 6.1.6. Tier conflation: Since vowels and consonants are represented on separate tiers, for rule (9) of emphasis spreadingto apply, it is necessary that all tiers conflate into one tier. Tier conflation allows the feature of emphasis to spread from an emphatic consonant to a vowel since they are no longer on separate tiers. Tier conflation is illustrated in(i3). Rule (13): CVCVC—-——>CVCVC Rule (13) of tier conflation puts the vowels and consonants of a word which has a CVCVC structure on the same tier. In (13), the capital letters stand for the consonantal and vocalic slots on the CV tier and the small letters for the feature matrices of the vowels and consonants on the melodic tier. Tier conflation allows emphasis to spread from an emphatic consonant or vowel to the other non—emphatic consonants and vowels of a stem. ‘ In his analysis 01 principles of the thi framework of this i; unspecified underly underspecified vowI respectively. This We (14): IMMH Ista I+round]_ within t he “‘90 postulatec in the missing {at 19871156) Rites (15); (A): I 1 190 6.1.7. Redundancy rules: In his analysis of epenthesis, Alghazo invokes some of the principles of the theory of underspecification. Within the framework of this theory, he argues that the vowel [i] in JA is unSpecified underlyingly and that the distinctive features of the underspecified vowels, [u] and [a] are [+round] and I-high], respectively. This is illustrated in (l4.A-C). Rule (14): (A): I I = l (B): [+rI = u (C): I-high] : a I | | v v v In (14), I ] stands for an empty matrice, and [+rI for the feature [+round]. Within the framework of the theory of underSpecification, he also postulated the redundancy rules in (l5.A—F). These rules fill in the missing features of the vowels in (14.A—C) (Alghazo, 1987:156). Rules (15): (A): I I ——-—> [+high] (B): [ ] ———> I-round] (C): I I -——> Io I+D8CI {—back] By using the reli principles of the thI claims that the und This is illustrated 6.2. Application of In this section first section abov Consider the JA d: (ISA): tagis "We (BIZ butln "till it): 's'abuk "en The JA data ' ahaItIZed. The J; °I Alghazo (198‘ 191 (F): I I —-——> {around} / [___—_l «back, -low By using the redundancy rules (15.A—F) in compliance with the principles of the theory of underspecification, Alghazo (1987: 158) claims that the underlying vowel does not have to be marked l—ERI. This is illustrated in the derivation of some JA words in 6.2. 6.2. Application of non—linear rules to JA data: In this section, we apply the non-linear rules postulated in the first section above in the derivation of the JA vowel in question. Consider the JA data in (l6.A—B). (16A): tagis "weather" (B): batin "tummy,- belly" (C): sabuk "entangling" The JA data in (16) are from the same JA dialect which Alghazo analyzed. The JA word in (16C) is cited in the appendix of the study of Alghazo (1987: 176). (I6.A,B) are counter evidence to the non-linear approach presented above. The application of the non—linear rules provided above in the derivation of the epenthetic vowel in (16.A,B) yield the wrong results; the rules incorrectly predict an [u] vowel. But in (16C), the rules yield the correct results if we extend the application of rule (12) of emphasis spreading to include velars. of the renal is ill in the co I+back,+corl consona the context of a pre consonant. (17A): liags/ "weather“ 192 In (17.A,B,C), we apply the non—linear rules presented above in the derivation of the data in (16.A,B,C). Note that in (16.A,B), the vowel is [i] in the context of a preceding {-hi] vowel and a [+back,+cor] consonant. Note also that the vowel in (16C) is [u] in the context of a preceding I—hi] vowel and a following [+back,—cor] consonant. (17A): (17B): (17C): /tags/ /batn/ /§abk/ UR "weather" "tummy" "entangling" I+I|3I B / [~cons] __ if {-Sgl] {-sgl] Rule (4) of glide degemination applies to /sarabigg/ and /sadaww/, gielding [sarabig] and [sadaw], respectivelg. However, there are some data which are potential counter evidence to rule (4) of glide degemination. Consider the data in (5.A,B). (5A): aaww "light" h’agg "a residential area" (5.8): lag—g "for—me" salag-g "on—me" In (SA), the words are monosgllabic. In (5.8), there is a morpheme boundarg between the two segments of the geminate cluster. Rule (4) of final glide degemination does not applg to the data in (SA) and (5.8). To account for the failure of rule (4) of final glide degemination to applg to the data in (5.A,8), we postulate constraint (6). Glide dege words or q Constraint (6)! degemination iron to a geminate of w independent morpl bisgllabic words The data acco accounted for bg 209 Constraint (6): Glide degemination rules do not applg to monosgllabic words or across 8 morpheme boundarg. Constraint (6) blocks the application of rule (4) of glide degemination from applging to geminates of monosgllabic words or to a geminate of which the second segment represents an independent morpheme as it is the case in the monosgllabic and bisgllabic words gaww, and salay-y, respectivelg. The data accounted for informallg bg Constraint (6) can also be accounted for bg postulating rules (7), and (8). (7): [—cons] ———> [-con] / it C V [~cons] __ ii l-sgl] {-891} {-8911 (8): [-cons] -——> [—con] / [~cons] +___ it l—sgi] I—sgll {-891} 8g postulating rules (7) and (8) we can dispense with Constraint (6). Rule (7) accounts for monosgllabic words as in (SA) and rule (8) for words in which the geminate glide is Split bg a morpheme boundarg as in (5.8). Rules (7) and (8) properlg include rule (4) of glide degemination and applg disjunctivelg in relation to it. In this wag, we formallg account for exceptions bg properlg including rules. This complies with the assumption that all exceptions are lsarabigg/ / "an Arab“ sarablgg I earabig sarabig ls’arabigl I In (9), rule because it is a monosgllabic v includes rule ( does not W19 these words is 210 accounted for bg properlg including rules. In (9), we illustrate the derivation of some words which include geminate glides, (9): /sarabigg/ /‘baww/ /lagg/ /salagg/ UR "an Arab" "light" "for—me" "on—me“ sarablgg baww lagg salagg Stress Assignment sarabig -—— —-— —-- Rule (4): Glide Degemination sarabig —-- —-- -—— Rule (15) in Ch. 3: 1 Stress [earabig] [bawwl [l’agg] [salaggl PR In (9), rule (4) of glide degemination does not applg to [baww] because it is a monosgllabic word,- final geminate glides in monosgllabic words are accounted for bg rule (7) which properlg includes rule (4) and applies disjunctivelg in relation to it. Rule (4) does not applg to llagg] and [salagg] because the geminate cluster in these words is split bg a morpheme boundarg; the final segment of the geminate is the pronominal morpheme —y "me”. This satisfies the SD of rule (8) which also properlg includes (4) and applies disjunctivelg in relation to it. Rule (4') applies to carabl’yy, gielding sarabiy. Then stress is adjusted to the change and is assigned to the first sgllabic segment as the stress rules predict. Some evidence two segments can the rules of interi 7.1.2.1.1. Glide de The rule of in intervocalicallg, the JA data in (1 (10A): katab "he jalas "he (10.8): xaaf “he daar "he (WC): tear "it maal "hr The data it (‘O-AA'). But in (WA) it do geminate 911d inter-poem”. tire; hiding”, 211 7.1.2.1. Glide deletion, vocalization, and degemination: Some evidence which argues for the behavior of geminates as two segments can also be obtained from their behavior in reSpect to the rules of intervocalic glide deletion and glide vocalization. 7.1.2.1.1. Glide deletion: The rule of intervocalic glide deletion deletes a single glide intervocalicallg, but it does not delete a geminate glide. Consider the JA data in (lO.A—C). (10.A):katab"he wrote" (IO.A'): kattab "he made someone write" jalas "he sat down" jallas "he made someone sit" (10.8): xaaf “he got scared" (10.8'): xawwaf "he scared someone" daar "he turned" dawwar "he made someone turn" (10C): taar "it flew" (10.C'): tagger "he made it flg" maal "he bent" maggal "he bent something" The data in (10.8,C) and (10.8',C') pattern with the data in (IO.A,A'). But in (10.8,C) the second radical does not show up, while in (10.A) it does. In (10.8”,C'), the second radical shows up as a geminate glide. We take this as evidence that a single glide deletes intervocalicallg. That is the underlging forms of xaaf, and taar are: /xawaf/, and /tagar/, respectivelg. This also gives further I-consl 4 {-5911 Rule (1 I) spoil but in (10.530). t behave as two set and (10.13). 7.12.1.2. Glide VI Further evide segments can al Consider the JA (12.A):9axuw " ramuw ' ligig "hl nisig "h 212 evidence that geminates do not behave as a single segment. Rule (1 1) accounts for intervocalic glide deletion. Rule (11): l—cons] ———> z / [+sglI] --— l+sglll l-sgl] Rule (1 1) applies to (10.8,C), deleting an intervocalic single glide but in (10.8',C'), the rule does not applg because geminate glides behave as two segments, not one segment as it is the case in (10.8) and (10.C). 7.1.2.1.2. Glide vocalization: Further evidence about the behavior of geminates as two segments can also be obtained from a rule of glide voclization. Consider the JA data in (12.A-C). (12A): 7axuw "brother" ramuw "theg threw" ligig "he found" nisig "he forgot" (12.8): 7axuu—ha "her brother“ ramuu—ha "theg threw it" llgll-ha "he found it" nisii-ha ”he forgot it" ll hagg “resi Eagg 'burn (12.0): baww-na laww-na hagg-na “ éagg—na “ In (12A), we bgavowel. In (I tothe data in (1 preceding vowel m(I2D),where notvocalize as 1“112-C,D)cont (12.A,B); the 9]. (1213.13) behave glides, we post Rule (13); l-sgll 1‘C0n: to rule (13 213 (12c): aaww "light" jaww "skg; weather" hagg "residential area" cagg "burn with a hot iron rod" (12D): baww-na "our light" jaww—na "our skg; weather" hagg—na "our residential area" cagg—na "our burn with iron“ In (12A), we have glides of one segment word-finallg preceded bg a vowel. In (I28), where a consonant-initial suffix is attached to the data in (12A), the glide vocalizes, merging with the preceding vowel. In (12C), we have geminate glides word-finallg. In (12D), where a consonant—initial suffix is attached, the glides do not vocalize as it is the case in (12.8). This argues that the glides in (l2.C,D) contrast with those in the correSponding positions in (12.A,8); the glides in (12.A,B) behave as one segment while those in (l2.C,D) behave as two segments. To account for the vocalization of glides, we postulate rule (13) of glide vocalization. Rule (13): [-sgl] ———> [+sgll] / [+sgll] __ [+cons] [-cons] [th] l—low] [oguuf "guests" (14C): filim "film" jisim "bodg" gisim "part" (14D): (oi)—flaam (91)-]saam (oil-gsaarrI (9i)-elaam1 The data in (1} the words are vet root is a vocalize data in (10.5-5) I c‘tm‘vsllonding $1: With, the second (145), the secor aIlpear in the COI 'tht‘ the Preced' Dlvral forms, 111 °1 We vocaiiz ”andthe Short the long VOWels glide preceded postulate in (1‘ (MB)- 0n thes from the Com; 215 silim “science,- news" (14D): (9i)-flaam "films" (7i)-j'saam "bodies" (oi)-gsaam "parts" (7i)-<;laam "sciences; news" The data in (14.A,B) pattern with the data in (I4.C,D). In (14A), the words are verbal nouns in which the second consonant of the root is a vocalized glide as we have argued in the analgsis of the data in (10.8-C) above. The data in (14.8,D) are the plural of the corresponding singular forms in (1-4.A,C), respectivelg. In the plural form, the second stem consonant is followed bg a long vowel. In (14.8), the second stem consonant is a glide. This glide does not appear in the corresponding singular form,- the glide has vocalized (with the preceding vowel, gielding the long mid vowel [ee,oo]. In the plural forms, the glide does not vocalize because the SD of rule (13) of glide vocalization is not met; glides vocalize in the context of a preceding short vowel and a following consonant. This shows that the long vowels lee] and [00] in (14A) derive from an underlging glide preceded bg a low vowel. The underlging glide which we postulate in (14A) shows up in the corresponding plural forms in (14.8). On these grounds, we assume that the data in (14A) derive from the corresponding underlging forms in (MA). (14.A'): /gawl/ /mawt/ leagfl Rule (4) of 91 and repeat here f of vowel raising I gielding 19001], int I Rule (4): I-sull -: I-consl 216 /bagt/ /§agf/ Rule (4) of glide vocalization which we postulated in Chapter 2 and repeat here for convenience applies in conjunction with rule (3) of vowel raising in Chapter 2 to the underlging forms in (14.A'), gielding [gool], [moot], [beet], and [beef], respectivelg. Rule (4): I-sgl] —--> [+sgl] / [+sgl] __ [+cons] I-cons] [oaww—na/ /7axuw—na/ /lat$w-na/ UR baww—na 7axdw-na laow-na Rule (15): Stress assignment [‘daww-nal | 7.12.1.3. Glide dI In JA, a rule geminate glides again the JA dat (18A): sarabig badawig s'ajnabig sadaw (. (18.8): sarabfgg badawlg ?o]nabil sadaww In (18.A),I 218 ——— oaxuu—na —-—- Rule (13): Glide vocalization ——— ——- labuw-na Rule (32) in ch. 4: Epenthesis [oaww—na] [oaxdu—na] [latSuw-na] PR 7.1.2.1.3. Glide degemination: In JA, a rule of glide degemination applies to word-final geminate glides but not to word—medial geminate glides. Consider again the JA data in (18.A,B). (18A): sarabig "Arabic" badawig "a bedouin" ?ajnabig "a foreigner" <:adaw (Neaduww = MSA) "an enemg" (18.8): sarabl’gg-ih "Arabic fem." badawigg-ih "A bedouin fem." oajnabl’gg-ih "A foreigner fem." sadaww—ah "an enemg fem." In (18A), the word-final single glides derive from the corresponding geminate glides in (18.8). Note that w—glides preceded bg a low vowel are verg rare. The native JA form sédaw is verg commonlg replaced bg the corresponding MSA word, saddww. The data in (18.A,8) argue that geminate glides word-finallg degemination whi convenience. In rule (4), at glide degeminati lsadawwl, gield application of H of the segmetal effects a stress 71.3. Geminate The represg 219 undergo a rule of glide degemination. This is rule (4) of glide degemination which we postulated above and repeat here for convenience. (4): l—cons] —--—> B / {-consl __ ii 1-Sgl] {-sgll In rule (4), again, [it] does not equal a consonant. Rule (4) of glide degemination applies to geminate glides as in /sarabigg/ and /sadaww/, gielding sarabiy and cadaw, respectivelg. The application of rule (4) of glide degemination changes the structure of the segmetal string to which the stress rules applg. This change effects a stress shift which we discuss in 7.1.3. 7.1.3. Geminates and stress: The representation of geminates as two segments underlginglg and as one long segment phoneticallg raises no problems to the rules of stress assignment. Consider the JA data in (19.A,8,C,D). (19A): sakat-t "I stopped talking" nahat-t "I carved" (19.8): mahall "place,- room" mamarr "passage,- wag" sédavv ”e (19.8): sarabigg-t cadaww-t InUQAfiD) represented as t thepredictionSI arMeongdedl segmentand str mnvahonofso (20) Isakatt/ sakatt 220 (19C): sarabig "Arabic" sadaw "enemg" (19.D): sarabl’gg—ih "Arabic fem." sadaww-ah "enemg fem." In (19.A,8,D), under the assumption that geminates be represented as two segments, stress is assigned in compliance with the predictions of the stress assignment rules. But in (19C) where a rule of glide degemination applies, the geminate becomes one segment and stress is adjusted accordinglg. In (20), we present the derivation of some JA words which involve geminates. (20): /sal V / 2, if the focus X takes on some or all of the features of the determinant 2, then ang boundarg which intervenes between V and Z in the output of that rule is deleted bg convention." According to this condition, geminates which are derived by assimilation behave as true geminates,- boundaries including morpheme boundarg delete at the site of assimilation. Further evidence in support of Guerssel's CAR can be obtained from the metrical phonologg analgsis of Palestinian Arabic bg Abu-Salim (1988: 60). In his analgsis, Abu—Salim argues that rules of both partial and complete assimilation block the application of rules of epenthesis. This explains the failure of the rules of epenthesis to applg to the geminates in (21.8) in which the rule of assimilation which created the geminate cluster has alreadg applied. Epenthesis does not applg before assimilation because rules of epenthesis are late rules. In (21C), assimilation mag be considered to have applied vacuouslg. The optional application of the rules of epenthesis to fake geminates can also be explained in reference to the Peak of Sonoritg Principle presented in Chapter 4. The two segments which form the fake geminate are identical. Hence, no peaks or vallegs of sonoritg will be formed between ang two consecuetive sgllabic segments or a sgllabic segment and a word-boundarg. Thus the scale of sonority principle is not violated and the grammaticality of geminates which are not split by epenthesis follows. In 7.3.2, we present the framework of CV 7.2. Geminates in In l’ISA, gemlr geminates mag ai lnflectional endil the end of a sent analgsis of gemi geminates. 7.2.1. Represent The represe Potential countl postulated in cI representation because in MSA Geminates be r 1°99 Segment I 225 analgsis of geminates in reference to epenthesis within the framework of CV phonologg. 7.2. Geminates in MSA: In MSA, geminates appear mainlg word-mediallg. Word-finally, geminates mag appear in the context where the word-final inflectional ending in optional. lnflectional endings are optional at the end of a sentence, mainlg before pause. Accordinglg, our analgsis of geminates in MSA will be limited mainlg to word-medial geminates. 7.2.1. Representation: The representation of geminates as one segment would be potential counter evidence to the rules of stress assignment postulated in chapter 3. In respect to epentheses, the representation of geminates does not raise any problems in MSA because in MSA rules of epenthesis do not applg. We argue that geminates be represented as two segments underlginglg and as one long segment phoneticallg. Consider the MSA data in (22.A~C). (22A): (22.A'): katab—a "he wrote" kattab-a "he made someone write" oakal-a "he ate" ?akkaI-a "he made someone eat" deras-a "he studied' derras-a "he taught" rakatp-a "he ran" rakkab-a "he made someone run" katab-a 'he WI’ déras-a “he stl talab-a "he rei salab-a ”he re (2212): 'sadd-a "he tig sadd-a "he clt h'alI-a “he sol taqq-a "he sp The data in data In (22.A'). non-geminates ‘(22.8,B',C,C'), t from the perfe cttmllared with WhICh derive f (225). The ge as their par-III 722- Gemlnal 226 (22.6): (22.8'): katab—a "he wrote" maktuub-un "something written" dares-a "he studied" madruus-un "something studied” talab-a "he requested" matluub—un "something requested" salab—a "he robbed" masluub—un "something robbed" (22C): (22.C'): s’add-a “he tightened" masduud—un "something tightened" s'add—a "he closed" masduud-un "something closed" hall-a "he solved" mahluul—un "something solved" saqq-a "he split" masquuq-un "something split" The data in (22A) minimallg contrast with the corresponding data in (22.A'). This argues that geminates contrast with non—geminates phoneticallg and phonologicallg word—medially. In -(22.B,8',C,C'), the derivation of the passive participles in (22.8) from the perfect verbs with a final geminate as in (22C) is compared with the derivation of the passive participles in (22.8) which derive from the perfect verbs which have no geminates as in (22.8). The geminates of (22C) derive from roots of CIC2C2 tgpes as their participles show. 7.2.2. Geminate glides: The behavior of geminate glides in MSA is similar to that of geminate glides in JA in reSpect to intervocalic glide deletion; single glides in both .JA and MSA delete intervocalicallg while In respect to glit IISA, one segmerl vowel and a folll Geminate glides, differ from glid word-finally, ge‘ in JA. 7.2.2.1. Glide (It In MSA, sin glides do not. (23A): - katab-a "he jalas-a “he I (23.3): 227 geminate glides do not. But glides in MSA differ from glides in .JA in reSpect to glide vocalization and final glide degemination. ln MSA, one segment glides vocalize in the context of a preceding high vowel and a following consonant, and so do single glides in JA. Geminate glides do not vocalize in both varieties. But glides in MSA differ from glides in JA in respect to glide degemination: word-finallg, geminate glides do not degeminate in MSA but theg do in JA. 7.2.2.1. Glide deletion: in MSA, single glides delete intervocalicallg, but geminate glides do not. Consider the MSA data in (23). (23A): (23.A') katab-a "he wrote" kattab—a "he made someone write" jalas—a "he sat down" jallas-a "he made someone sit" (23.8): (23.8'): xaaf-a "he got scared" xawwaf-a "he scared someone" daar-a "he turned" dawwar-a “he made someone turn“ (23.C): (23.0): taar—a "it flew" tagger—a "he made it fly" maal—a "he bent" maggal-a "he bent something" The MSA data in (23) are similar to the JA data in (lo). deletion in MSA. 7.2.2.2. Glide vo ln MSA, unli; preceding high i vocalize in the glides vocalize (12.A,B) and pri glides in both t (MA-C). (24A):€Hd “n diin "rl ruuh “: SUur "i (24.8): qawm lavvh ' bagl " tegr “ (2413): film ' qlsm Sil‘b ' 228 Accordinglg, the analgsis of intervocalic glide deletion in JA presented in 7.1.2.l.l. applies to the analgsis of intervocalic glide deletion in MSA. 7.2.2.2. Glide vocalization: In MSA, unlike JA, single glides vocalize in the context of a preceding high vowel and a following consonant. But theg do not vocalize in the context of a preceding low vowel. in JA, single glides vocalize in the context of preceding high vowels as in (12.A,B) and preceding low vowels as in (MA), as well. Geminate glides in both JA and MSA do not vocalize. Consider the MSA data in (24.A-C). (24.A):siid "festival" (24.A'):?aqgaad "festivals" diin “religion" 7adgaan "religions" ruu‘fi "spirit" earwaah "spirits" suur "wall" 7aswaar "walls" (24.8): qawm "people" (24.8”): eaqwaam "peoples" lawh "board" ealwaah "boards" bagt "line of poetrg" pabgaat "lines of poetrg" tagr "bird" 7atgaar "birds" (24C): film "film" (24C'): 7aflaam "films" qism "part" vaqsaam "parts" sirb "a flock" easraab "flocks" (24.0): siggaarah xaggael ' sawwaah suwwaah; The data in t in (24A), the un plural forms in high vowel and ; preceding low v context, the bel that of one segi vocalize Wheth alurther evidenc with that of no behfiViOl’ 0f ger Similar to that above. 229 jism "bodg“ 7ajsaam "bodies" (24.8): siggaarah "a car" xaggaal "a horseman" sawwaah "a tourist" suwwaah "tourists" The data in (24.A,A') and (2488') are like the data in (24.C,C'). ln (24A), the underlging glide which appears in the corresponding plural forms in (24.A') has vocalized in the context of a preceding high vowel and a following consonant. But in the context of a preceding low vowel as in (24.8) the glide does not vocalize. In this context, the behavior of one segment glides in MSA is different from that of one segment glides in JA. In (24.0), geminate glides do not vocalize whether the preceding vowel is high or low. This gives further evidence that the behavior of geminate glides contrasts with that of non—geminate glides. It also gives evidence that the behavior of geminate glides in MSA in respect to vocalization is similar to that of geminate glides in JA as we have shown in 7.1.2.2, above. 7.2.2.3. Glide degemination: in MSA, geminate glide do not degeminate in either word-final or word-medial contexts. Consider the MSA data in (25). (25): sarabigg~sarabigg-un "an Arab(-nom.)" nablggunab daddwwsa The data in l geminate glides which marks caI glides do not un‘ the behavior of geminate glide: degemination, i the stress rule position. Stres cluster as the (7.2.3. Geminati 230 badawigg~badawigg—un "a bedouin(-nom.)" nabiggmabigg—un "a prophet(—nom.)" saduww~saduww—un "an enemg(—nom.)“ The data in (25) show that it is possible to have word-final geminate glides in MSA in the context where the inflectional ending which marks case is optional. The data also show that word—final glides do not undergo a rule of glide degemination. Thisshows that the behavior of geminate glides in JA is different from that of geminate glides in MSA. Due to the absence of a rule of glide degemination, there is no change in the segmental string to which the stress rules applg. Accordinglg, stress does not shift to a new position. Stress is assigned to the vowel preceding the geminate cluster as the stress rules predict. 7.2.3. Geminates and stress: The analgsis of geminates as one long segment phoneticallg and as two segments underlginglg raises no problems to the rules of stress assignment. In the contexts which involve geminates, stress is assigned as the stress rules postulated in Chapter 3 predict. 7.2.4. Geminates and epenthesis: In I’ISA, the question of epenthesis does not arise. MSA allows two consonant clusters word—finally. Word—medially, MSA does not have three consonant clusters. Geminates are not directly followed segments does word—medially. 7.3. The analgse In CV phonoI the CV tier and within the from assignment, epi 7.3.1. Geminate In CV phone the CV tier on: With the gong: representation tier, In the an reference to tl °°“50hantdl si feature matic on the melodl. trees; gemina 7‘3-2. Gemina Within th 231 by consonants. Thus, the representation of geminates as two segments does satisfg the SD of the rules of epenthesis word-medially. 7.3. The analgses of geminates in CV phonologg: In CV phonologg, geminates are represented as two segments on the CV tier and as one segment on the melodic tier. Following, within the framework of CV phonologg, we analyze stress assignment, epenthesis, glide vocalization, and glide degemination. 7.3.1. Geminates and stress assignment: In CV phonologg, geminates are represented as two segments on the CV tier and as one segment on the melodic tier,- in compliance .with the Obligatorg Contour Principle (OCP) which prohibits the representation of two adjacent identical segments on the melodic tier. In the analgsis of stress in CV phonologg, there is no explicit reference to the melodic tier. It refers only to vocalic and consonantal segments on the CV tier; there is no reference to the feature matices associated with the vocalic and consonantal slots on the melodic tier. Accordinglg, in the building of the metrical trees, geminates are represented as two segments on the CV tier. 7.3.2. Geminates and epenthesis: Within the framework of CV phonologg, the failure of epenthetic predIctIons of th lines. The Inser violates the wel between the CV crosses the asst epenthesis doeSI "replg“. In fake I there is a morph geminate cluste geminate cluste represented on : sakétit. In (26 the CV tier and Words rédd, a (26A): radd, Sdkét-t’ I 232 vowels to split a true geminate cluster follows from the predictions of the Well Formedness Condition (WFC) on association lines. The insertion of an epenthetic vowel in a true geminate violates the well formedness condition. The association line between the CV tier and the melodic tier of the epenthetic vowel crosses the association lines of the geminate cluster. Accordinglg, epenthesis does not applg to true geminate clusters as in *radid "replg". In fake geminates as in sakat+t "I stopped talking". where there is a morpheme boundarg between the two segments of the geminate cluster, epenthesis is allowed since the segments of the geminate cluster belong to different morphemes which are represented on separate tiers. This explains the grammaticalitg of sakatit. In (26.A,8), we illustrate the association lines between the CV tier and the melodic tier of the geminate clusters in the JA words rédd, and sakatt. (26A): /[1\ X\ rédd, C V C C#-—-> *C V C V C# (26.8): [If] IIIIH II II/ sakét-t,CVCVC-C#-——>C V C V C V C# In (26A), the epenthesis. ACCI (26.8), the assoc CV phonologg acl lake geminates. absence of open in rédd ;CV ph before epenthes line crossing (H In mg linear geminates is at Geminate clust geminate clust e(lual sonoritg 7.3.3. Geminats The JA rul Context of a f< Within the fra glide vocalize involves a p10 "Associauoni EXhHVotive“ (I gilde VOCoIth tier to dlStin “We M the 233 In (26A), the association lines cross each other as a result of epenthesis. Accordinglg, *rédi'd is ruled ungrammatical. But in (26.8), the association lines do not cross each other. In this wag, CV phonologg accounts for the optional application of epenthesis in fake geminates. This explains the grammaticalitg of sakétit. The absence of epenthesis in words like sakatt, (=sakat—t):explains as in rédd ; CV phonologg here assumes that assimilation has applied before epenthesis to yield the branching structure which disallows line crossing (Hages, 1986: 327). In my linear analgsis, the optionalitg of epenthesis in fake geminates is accounted for bg reference to the scale of sonority. Geminate clusters do not violate the PSP; the segments of the geminate cluster do not create a sonoritg peak because theg have equal sonority values, so epenthesis is optional. 7.3.3. Geminates and glide vocalization: The JA rule of glide vocalization which vocalizes glides in the context of a following consonant does not applg to geminate glides. Within the framework of CV phonologg according to Hages (1986), glide vocalization is blocked from applging to geminates because it involves a violation of the Linking Constraint (LC) which sags, "Association lines in structural description are interpreted as exhaustive" (Hages, 1986: 331). According to Hages (I986: 334) glide vocalization rules necessarilg refer to two tiers: the melodic tier to distinguish glides from other consonants, and the CV tier to carrg out the structural change C ——> V Hages adds that glide -' .I' In: ~'vl.~' - ..., i. one «an mm , ',.,‘ __ “v.5" firm"?! '2'}; 3f jg‘i‘oliqg; \ ‘ reg: vocal‘izationin Constraint pred morpheme boun alterable as th‘ 332-334). Thi vocalization ru "he gets arrest In MSA, gli and a followin non-linear gli Rule (27): C- Before r which has Ih merging the glide (Hages' 234 vocalization respects the distinction between true and fake geminates, supporting his assumption with evidence from glide vocalization in Berber. A true geminate is inalterable as the Linking Constraint predicts. But a fake geminate which is split by a morpheme boundarg does permit glide vocalization and so it is alterable as the Linking Constraint would predict (Hages, 1986: 3323534). This is illustrated by the application of non-linear glide vocalization rules in the derivation of the MSA words lgu—wqaf-u/ "he gets arrested" and /xuwwifa/ "he got scared". In MSA, glides vocalize in the context of a preceding high vowel and a following unidentical consonant. This is rule (27) of non—linear glide vocalization. Rule (27): Before rule (27) applies, Hages (I986: 335) assumes that a rule which has the effect of the Obligatorg Contour Principle must applg, merging the melodic representations of the vowel and the following glide (Hages' did not formallg provide such a rule). Rule (27) applies to MSA /gu—wqaf-u/ "he gets arrested", gielding Iguuqafu] but it fails to applg to MSA xawwifa "he got scared". Rule (27) applies to /gu—wqaf-u/, gielding guuqafu without violating the Linking Constraint,- one feature matrix on the melodic tier has two association lines to link it with the preceding vowel flyytaé‘lfl ”I“! IIDV'J ‘lfi; ('1 l"(?."". I '3. 1 .. ~.-.'-v,r:Tf¢cr ollfiaP ”z” I I;ri‘Ji’I'."-"¥"~”_ - '- - _ . r - .713 " ”"1 I andthe onto-con in(28)(Hagesrt -» at some? w'" . In (28), the vocalization W association lir of (27) are not Within th questions cou geminate in x analgzes gem analgsis does argues that I lerIsIeIong IIIIIII 235 and the vocalized glide. But the rule cannot applg to xuwwz’fa because here one feature matrix will have three links, to the vowel and the two consonants of the geminate cluster. This is illustrated in (28) (Hayes, 1986: 335). (28): In (28), the application of rule (27) of non—linear glide vocalization would be in violation of the LC, according to which association lines are interpreted exhaustively. Thus the two lines of (27) are not found in the representation (28). Within the framework of this non—linear analgsis, some questions could be raised in reference to the identification of the geminate in xuwwifa. Hages, in line with McCarthg (1981), analgzes geminates in Form II verbs as true geminates. This tgpe of analgsis does not take into consideration some evidence which argues that the two segments of the geminate cluster in Form II verbs belong to two different morphemes. It has to be noted that McCarthg (1979: 388) analgzed both segments of the geminate of form II verbs as part of one morpheme, namelg, the stem morpheme. In doing so, McCarthg denies that the root consists of the three consonants (usuallg referred to as the I .1... 2." .g'v’t'croxtr , in m we -‘-.7- .,. ,. - e mm ”ain't- ;;"13.1r.‘.:flfi 3, lverb, the root which is a For kattab, is an a the same wag t analgzed as an kattab, and k McCarthg's one his analgsis of in 7a-srac "h- "he talked" wh received (visit morphemes (it that the two 3 which is a for Accordinglg, geminate. In vocalization It because Iu This means I ungrammatic However, iIIlnIIIe 236 three radicals) and that other consonants cannot be part of the tri—radical root. For example, in katab "he wrote“, which is a Form I verb, the root is //ktb// and in kattab "he made someone write", which is a Form II verb, the root is still //ktb//. The second /t/ in kattab, is an affix which represents an independent morpheme in the same wag the initial /n/ in nkéttab which is a Farm VII verb, is analgzed as an independent morpheme. katab contrasts with kattab, and kattab contrasts with nkétab. This shows that McCarthg's analgsis of the affix in Form II verbs is inconsitent with his analgsis of the affixes in the other verb forms. The affixes /?/ in za-sras "he hurried" which is a Form IV verb, /t/ in ta-kallam "he talked" which is a Form V verb, and /s/ in 7is-taqbal "he received (visitors)"JForm X,are all analgsed as independent morphemes (McCarthg, 1981:388—389). On these grounds, I argue that the two segments of the geminate glide in words like xawwaf which is a form It verb belong to two different morphemes. Accordinglg, the geminate glide in this form would be a fake geminate. In this case, the application of rule (27) of glide vocalization to xuwwifa will not be considered a violation of the LC because [u] would be associated with just two association lines. This means that the LC is incapable of accounting for the ungrammaticality of *Ixuuifa]. However, we have to take into consideration that the derivation of Form II verbs is not exactlg the same as the derivation of the other verb forms. In Form II verbs, the affix morpheme is marked by a consonant which is alwags identical to the second stem consonant. The affix consonant alwags takes the identitg of the stem 9,- . - ' ' -‘ mfwmutmww'h this case. It '92 assimilation it the CV tier. Another Dr vocalization 7" effect of the I and the follow non-linear SI”I identical feat application of glide vocalize vocalized 911d 7.3.4. Geminet In JA, 3 r’ word—finally framework 0 princilfl'as E”: this is the 8‘ in Attic in r /ss/ degem‘ a i.W'-"“u‘3rr He argues tI which slmp 237 consonant. In other words, it assimilates to it. On these grounds, the geminate cluster in Form II verbs is created bg assimilation. In this case, it can be argued that rule (27) is blocked after assimilation forms a true geminate which is not a VC sequence on the CV tier. Another problem with the CV phonologg approach to glide vocalization relates to the postulation of a rule which has the effect of the OCP to merge the melodic representation of the vowel and the following glide. Such a rule which supplements rule (27) of non-linear glide vocalization assumes that vowels and glides have identical feature matrices on the melodic tier before the application of the rule of glide vocalization itself. It is rule (27) of glide vocalization itself which has the effect of merging the vocalized glide with the preceding vowel. 73.4. Geminates and glide degemination: In JA, a rule of degemination applies onlg to geminate glides word—finallg. Our analgsis of glide degemination-in JA within the framework of CV phonologg will be based on the CV phonologg principles applied in the analgsis of other geminates. Related to this is the analgsis of Hages (1986:346-347) of /ss/ degemination in Attic in reference to the Linking Constraint. In his analgsis of /ss/ degemination in Attic, Hages argues that /ss/ degemination is a two-tier rule which involves both the CV tier and the melodic tier. He argues that /ss/ degemination is different from /s/ deletion which simplg deletes /s/ before another consonant. In (29) and (30) deletion, respecti ’ (29): lssl—Degemi. c ———> o (30): Isl—Deletlo Rule (29) is 1 segment. In this of the rule and n Accordinglg, he Constraint. ln( following consc consonant are n The probler happens to the the CV tier wit have to assumt Inline wit account for 91 238 we present the rules he postulated for /ss/ degemination and /s/ deletion, respectivelg (Hages,1986: 346). (29): /ss/—Degemination (30): /s/—Deletion C———>B/__C Rule (29) is a two-tier rule which alters onlg half of a long segment. In this rule, the multiple linkings are mentioned in the SD of the rule and match one—to—one with those of a geminate /ss/. Accordinglg, he concludes, the rule does not run a foul of the Linking Constraint. In (30), /s/ is not linked to the representation of the following consonant on the CV tier because /s/ and the following consonant are not a geminate as it is the case in (29). The problem with this analgsis is that it is not clear about what happens to the association line which links the deleted segment on the CV tier with its feature matrix on the melodic tier. Here, we have to assume that it automaticallg deletes by convention. In line with the analgsis of Hages, we postulate rule (31) to account for glide degemination in JA. I (31): Glide Degemf C ——-) 3 Rule (31), app ltarablgg/ "Arabi 7.4. Summarg: To sum uI3. ge- Stress assignme as two segment V°Wel preceding Itis two segments from MSA and J) non‘Qeminates. This is support; vocalization an epenthesis, We and fake gemin identical segm geminates do r and CAR (Guer Constraint (He 239 (31): Glide Degemination Rule (31), applies to the .JA glides in /sadaww/ "enemg" and /sarabigg/"Arabic",gielding sadaw, and sarabiy, reSpectivelg. 7.4. Summarg: To sum up, geminates do not present counter evidence to the stress assignment rules postulated in Chapter 3. Geminates behave as two segments underlginglg. Hence stress is assigned to the vowel preceding geminates word—finallg. The analgsis of geminates [as two segments is supported bg phonetic and phonological evidence from MSA and JA data. Phoneticallg, geminates contrast with non-geminates. Phonologicallg, geminates behave as two segments. This is supported bg their behavior in reSpect to the rules of glide vocalization and intervocalic glide deletion. In respect to epenthesis, we recognize two tgpes of geminates: true geminates and fake geminates. True geminates are adjacent; theg consist of identical segments and are not split bg a morpheme boundarg. True geminates do not allow epenthesis as both the AIC (Gurresel, 1977) and CAR (Guerssel, 1978) in segmental phonologg, and the Linking Constraint (Hages,l986) in CV phonologg predict. True geminates Constraint postult geminates allow I predict. I The behavior I respect to glides. of final glide deg. words and across Which applg (11st to account for mt geminate is Split the context of a includes the co“. ghde degeminau itDDIies onlg int 240 also include geminates created bg assimilation rules. Epenthesis does not applg to assimilated geminates as the Assimilation Constraint postulated bg Abu-Salim (I988) predicts. Fake geminates allow epenthesis as the AIC and the Linking Constraint predict. The behavior of geminates in JA differs from that of MSA in reSpect to glides. Word-finallg, glides degeminate in JA. The rule of final glide degemination applies in JA except for monosgllabic words and across morpheme boundarg. Properlg including rules which applg disjunctivelg in relation to other rules are postulated to account for monosgllabic words and words in which the final geminate is Split bg a morpheme boundarg. In JA, glides vocalize in the context of a preceding vowel and a following consonant. This includes the context of a preceding low vowel. In MSA, the rule of glide degemination does not applg and the rule of glide vocalization .applies onlg in the context of a preceding high vowel. Inthischapte according to the t phonemes of each vowels, and gemi two varieties wi In the first secti MSA.Intheseco andMSA.Inthe inJAandMSA,a JAandMSA. Th 8.1. Inventorg oi Chapter 8 Conclusion In this chapter, we review the comparison of JA and MSA according to the findings of our analgsis of the inventorg of phonemes of each varietg, stress assignment, alternating high vowels, and geminates. Similarities and differences between the two varieties will be illustrated bg the analgsis of some cognates. In the first section, we review the inventorg of phonemes of .JA and MSA. In the second section, we compare stress assignment in .JA and MSA. In the third section, we compare alternating high vowels in JA and MSA, and in the fourth section, we compare geminates in JA and MSA. The fifth section is a conclusion. 8.1. Inventorg of phonemes: The analgses of the inventorg of phonemes in JA and MSA in Chapter 2 indicate that the inventorg of phonemes in both varieties are similar except for /q/, /g/, /g/, /é/, /ag/, and /aw/. MSA has /q/, /g/, /aw/, and /ag/ while JA has /g/, /’o/, lee] ( on the I38: 244 (F): bitnaagasuw "theg are discussing" (F'): gatanaaqasuuna (G): carabig "Arabic" ((3'): carabiggun The JA words in (2.A-G) and the corresponding MSA words in (2.A‘-G') show that stress is assigned to both varieties as the stress rules postulated in Chapter 3 predict. In both varieties stress constantlg adjusts to the changes in the segmental string to which the stress rules applg. The different positions to which stress is assigned are due to the rules which change the structure of the segmental string itself. The rules which involve a change in the segmental string are either historical or sgnchronic. Historical rules account for the loss of the infectional endings such as I-u-nl "nominative—indefinite" which marks nouns for case and indefiniteness, and the word-final glottal stop in the JA words in .(2.A-D, G) correSponding to the MSA words in (2.A'-D', G). Sgnchronic rules account for vowel shortening as in (2.A-8), unstressed high vowel deletion as in (2.0), and word-final glide degemination as in (2.6). Unstressed high vowel deletion is accounted for bg rule (8) (in Ch. 4). Final—glide degemination is accounted for bg rule (4) (in Ch. 7). Vowel shortening is accounted for bg rule (16) in (Ch. 3) and repeated in (3) for convenience. Rule (3): W ———> V / __ it On the basis of the analgsis provided above, putting aside differences in vet words: some, flag from: samtiaeun, respectivelg. Thi corresponding ne- these JA words. . related to vowel feminine marker appearance of so In (2.C,D). It is t issues which are (4): lsamaa/ I some some some 245 differences in vowel and consonant qualitg, we assume that the 3A words: sama, héya, ml'drasih, and mistamsih historicallg derive from: samaa7un, hayéwun, madrésatun, and mustamicatun, respectivelg. These postulated JA forms are similar to the corresponding MSA cognates. In (4), we provide the derivation of these JA words. In these derivations, we put aside questions related to vowel or consonant qualitg such as the change of the feminine marker [t] into In] word—finallg as in (2C) and the appearance of some vowels in the form of [it instead of [a] or [u] as in (2.C,D). It is begond the scope of this studg to deal with these issues which are not directlg related to the issue under analgsis. (4):/samaa/ /hagaa/ /midrasih/ /mistamisih/ UR sam’aa hagaa midrasih mistamisih Rule (15): Stress sama haga -—- -—— Rule (3): V—shorten. sama haga --- --- Rule (15): Stress -_- _-- —-— mistamsih Rule (8) (in Ch.4): HVD [samal lh'agal [midrasihl Imistamsih] PR In (4), the rules are not ordered. Theg applg whenever their SD is met. Rule (15) of stress assignment re—applies when rule (3) of vowel shortening has applied. This accounts for the shift of stress from the second st some, and flags. when rule (15) 0f In JA eigraayi: sgllabic segment stress rules pred is assigned to thi consonant cluste In JA bitnaag sgllabic segmenI gatanaagaétiuna I segment as the : conjugational ei plural indicativ environment for appearance of t In on sarab sgllabic segme 888tones to the “We I‘unl. T Of the on rule In (5), we eigraagithum, 246 from the second sgllabic segment to the first sgllabic segment in same, and hag/a. The SD of rule (8) of high vowel deletion is met when rule (15) of stress assignment has applied. In JA 7igraayl'thum as in (2E), stress is assigned to the third sgllabic segment which is followed bg a consonant cluster as the stress rules predict. But in MSA qiraa7étuhum as in (2.E‘), stress is assigned to the third sgllabic segment which is not followed bg a consonant cluster. In JA bitnaagasuw as in (2F), stress is assigned to the second sgllabic segment as the stress rules predict. But in MSA yatanaaqa§duna as in (2.F'), stress is assigned to the fifth sgllabic segment as the stress rules also predict. In the JA form the conjugational ending {—na] which marks the imperfect verb for the plural indicative is lost. The loss of this suffix provides an environment for rule (4) of glide vocalization, explaining the appearance of word—final [uwl in JA. In JA cérabiy as in (2.13), stress is assigned to the first sgllabic segment. But in MSA carabl’gyun as in (26‘) stress is assigned to the third sgllabic segment. JA has lost the inflectional ending I—un]. This loss created a new context which satisfies the SD of the JA rule of final-glide degemination. In (5), we illustrate the derivation of the JA words: zigraayl’thum, bitnaagasuw, and carabiy. (5): /giraagltuhum/ giraagituhum graaglthum igraagithum eigraagithum 247 (5): /giraagituhum/ /bitnaaga§uw/ /sarabigg/ UR giraagituhum bitn'aagasuw sarabigg Rule (15): Stress graagithum ——- ——— Rule (8): HVD igraagithum ——- --- Rule (22) (in Ch.4): Word-Initial Epen. eigraagithum ~—- ——— Rule (23) (in Ch.4): Glottal Stop Inser. ——- sarabig Rule (4) in (ch.7): Glide Degemination ——— sarabig Rule (1'5) (in Ch.3): Stress [eigraagithuml [bitnaagasuwl [sarabigl PR In (5), the sgnchronic rules do not have to be ordered. Theg applg whenever their SD is met. In the derivation of carabiy, stress is adjusted to the change in the segmental string which resulted from the application of rule (4) of glide degemination. This eXplains the assignment of stress to the second sgllabic segment in MSA carabl’ggun and to the third sgllabic segment in JA Isarabig]. In the derivation of JA eigraayithum, stress is assigned to the second sgllabic segment. But in the corresponding MSA cognate qiraamfituhum, stress is assigned to the third sgllabic segment. The assignment of stress to two different positions is due to the loss of the vowel which follows [t] which marks the verbal noun for C838. The 1038 Of the C858 marker created 8 consonant cluster and so stress is assig as the stress rule stress is assignel rules predict. Bu yatanaaqaéduna, the stress assigr to different posi suffix l—nal in JA verbs for the ind yatanaaqasuw-n vocalization wh word, the glide not. The applic: yatanaaqasuw-l glide gields a it 248 so stress is assigned to the sgllabic segment preceding this cluster as the stress rules predict. in the derivation of JA biynaagasuw, stress is assigned to the second sgllabic segment as the stress rules predict. But in the corresponding MSA cognate yatanaaqas‘duna, stress is assigned to the fifth sgllabic segment as the stress assignment rules also predict. The assignment of stress to different positions in the two varieties is due to the loss of the suffix {—na] in JA. The MSA suffix [me] which marks imperfect verbs for the indicative has been lost in JA. The MSA word yatanaaqas’uw-na satisfies the SD of rule (13) (in Ch?) of glide vocalization while the JA word bitnaagas'uw does not. In the MSA word, the glide is followed by a consonant but in the JA word it is not. The application of rule (13) of glide vocalization to MSA yatanaaqasuw-na gields yatanaaqasuuna. The vocalization of the glide gields a long vowel. This long vowel is assigned stress as the stress rules predict. in the corresponding JA cognate, bitnaagas’uw, the glide does not vocalize because the SD of rule (l3) is not satisfied,- the loss of the suffix {-na] bleeds the context of rule (13). Accordinglg, no long vowel is created and so stress is assigned to the third sgllabic segment. The above analgses of JA and MSA cognates makes it clear that JA and MSA may be considered to have the same stress assignment rules. Differences are onlg in the structure of the segmental string to which the stress rules applg. Stress rules alwags adjust to the changes in the segmental string. To sum up, the analgses of stress in JA and MSA give inconclusive evi :1 Figure (i). (In th possibilitg that is a sister langu illustrated in Fig 8.3. Alternating . The analgses VOWei which op; ihd three conso: that MSA does it (6-A'E) and the 1A (6A); n'abn 'ro 3): daluw "b C): Uifiirfak D): Ufikutbak in (6A5), déiw in (6.A' to the high V0 high voweiSd Desmon in re For example, 249 inconclusive evidence that JA is a descendant of CA as illustrated in Figure (1). On the other hand, the analgses do not exclude the possibilitg that JA is a descendant of another Arabic varietg which is a sister language to CA from which MSA has developed as illustrated in Figure (2). 8.3. Alternating high vowels: The analyses in Chapter 4 indicate that JA has an epenthetic vowel which appears in the context of two consonants word-finalig and three consonants word-medially. The analgses also indicate that MSA does not have this vowel. Consider the JA cognates in (6.A-E) and the corresponding MSA cognates in (6.A'-E'). . JA M_S_A (6A): habil "rope" (6.A'): habl B): daluw "bucket" B): dalw C): gisirfak "he knows gou" C): gasri’fuka D): gukutbak “he writes gou" D): gaktdbuka in (6.A,B), JA hébil and déluw correspond to MSA hébl and dalw in (6.A',B'), reSpectively. The two cognates differ in respect to the high vowels, [i] and [u], which do not appear in MSA. These high vowels differ from other high vowels in the corresponding position in reference to stress assignment and glide vocalization. For example, in the context of a consonant initial suffix as in déluw-hum, the glide does not v0 (15) of stress as péxuw "brother gields 7axdu-hur (15) of stress a: vocalizes as rul that the status of the correspor suffix, as in bi presence and at calling them “a from other von in (6C), JA (6.0). Here, th and second cor Similar to it i call the JA his in the context knows them“. bBtWeen the s We arsued th: A"altisis indi metathesizel metathesizel the StruciUr. 250 da’luw-hum, the vowel is not assigned stress and the following glide does not vocalize, which is contrary to the predictions of rule (15) of stress assignment and rule (i3) of glide vocalization. But in eéxuw "brother", the attachment of a consonant—initial suffix gields 7aqu-hum in which the vowel is assigned stress as rule (15) of stress assignment predicts, and the following glide vocalizes as rule (13) of glide vocalization predicts. This argues that the status of the high vowel in déluw is different from that of the corresponding one in 7éxuw. In the context of a vowel-initial suffix, as in habl-ak, and délw-ak the vowel is absent. The presence and absence of the these tgpes of high vowels motivate calling them "alternating". This term is used to distinguish them from other vowels which do not alternate. In (6C), JA yi’sirfak corresponds to MSA yasrifuka as in _(6.C'). Here, the alternating high vowel appears between the first and second consonants of the stem. In MSA, an unalternating vowel similar to it appears between the second and third radicals. We call the JA high vowel "alternating" because it alternates with zero in the context of a consonant-initial suffix as in yisrif—hum "he knows them". In MSA, the corresponding vowel alwags appears between the second and third consonants of the stem. In Chapter 5, we argued that this alternating high vowel cannot be epenthetic. Analgsis indicates that the vowel is either underlging or metathesized. The analgsis which indicates that the vowel is metathesized has the advantage of capturing more generalitg about the structure of the imperfect stem vowel. Accordinglg, it is preferred to the 1: Differences bl vowels indicate a vowel insertion i deletion and the I created unfavour came to the resc loss of the inflel the MSA word a cluster. This ch (4.5). This tgpe the MSA forms r on does not nec developed. To sum up, ' inconclusive ev Figure (1). Un‘ possibilitg tha is a sister lang illustrated in I 251 preferred to the analgsis which argues that the vowel is: underlging. Differences between JA and MSA in reference to alternating high vowels indicate a language in which JA inserted these vowels. High vowel insertion is motivated by the rules such as high vowel deletion and the loss of some inflectional endings. These rules created unfavourable consonant clusters, and the rules of insertion came to the rescue by breaking these clusters. For example, the loss of the inflectional ending {-u-n] "nominative-indefinite" from the MSA word habl-un "rope“ resulted in flab] which has a i-bliil cluster. This cluster is not permitted in JA; it violates the PSP in (4.5). This tgpe of cluster does not arise in MSA. This argues that the MSA forms represents an earlier form of JA. This earlier form of JA does not necessarilg have to be CA from which MSA has developed. To sum up, the analgses of epenthesis in JA and MSA give inconclusive evidence that JA is a descendant of CA as illustrated in Figure (1). 0n the other hand, the analgses do not exclude the possibilitg that JA is a descendant of another Arabic varietg which is a sister language to CA from which MSA has developed as illustrated in Figure (2). However, underthe assumption that that the vowel which appears in the context of two consonants word—finallg or three consonants word-mediallg is underlging, the analgses give conclusive evidence that .JA is not a descendant of CA as illustrated in Figure (2). In this case, the absence of the vowel in CA has to be accounted for by a rule of syncope. 8.4. Geminatesz The analgses that theg behav segment phonetil word-finallg. JI while MSA doesl (YA-D) and the . JA (7A): sarabig " B): sadaw “e C): mistasid D): mistagir in (7.A,B), and saddwwu mistagirr C0 respectivelg. (in on?) app} degeminatior MSA in the d' To sum o inconclusivs Figure“). I Possibilitu 252 8.4. Geminates: The analgses of geminates in JA and MSA in Chapter 7 indicate that they behave as two segments underlginglg and as one long segment phoneticallg. But theg differ in respect to geminate glides word-finallg. JA has a rule of glide degemination word-finally while MSA does not have this rule. Consider the JA cognates in (7.A-D) and the corresponding MSA ones in (7.A'—D'). AA ELSA (7A): sarabig "Arabic" (7.A'): sarabi’gg-un B): sadaw "enemg" B'): saduww-un C): mistasl’dd “readg” C'): mustasidd-un D): mistagirr "has got settled" D'): mustaqirr-un In (7.A,B), JA sarabiy and sadaw correspond to MSA sarabiyyun and saddwwun, respectivelg. In (7.C,D),JA mistasfdd and mistagl’rr correspond to MSA mustasfdd and mustaql’rr, respectivelg. In JA, the rule of glide degemination, that is rule (4) (in Ch.7) applies word—finallg. In MSA, the rule of glide degemination does not applg. This indicates a language change from MSA in the direction of JA. To sum up, the analgses of geminates in JA and MSA give inconclusive evidence that JA is a descendant of CA as illustrated in Figure (t). On the other hand, the analgses do not exclude the possibilitg that JA is a descendant of another Arabic varietg which illustrated in F 8.5. Conclusionzi The analgse phonemes, stre geminates indi phonemes, the [00]. MSA [g] of the other hand, [8W], respectii varieties have Which change Alternating hi tWD-consonan Word-median 253 is a sister language to CA from which MSA has developed as illustrated in Figure (2). 8.5. Conclusion: The analgses of JA and MSA in reference to the inventorg of phonemes, stress assignment, alternating high vowels, and geminates indicate these findings. in reference to the inventorg of phonemes, the two varieties differ in respect to [g], [g], [61, [ee], and [00]. MSA [g] and [q] correspond to JA [25], and [g], respectivelg. 0n the other hand, JA [6], [ee], and [00] correspond to MSA [k], [ag] and [aw], respectivelg. In reference to stress assignment, the two varieties have the same stress rules. Differences are in the rules which change the segmental string to which the stress rules apply. Alternating high vowels appear onlg in JA. MSA allows two—consonant clusters word—finallg while JA does not. Word—mediallg, three consonant clusters do not appear in MSA. Accordingly the analgses of alternating high vowels are limited to those which appear in JA in the context of two consonants word—finally and three consonants word—medially. Thus questions related to high vowel alternation do not arise. The studg argues that geminates in both JA and MSA behave as two segments underlginglg and as one long segment phoneticallg. But theg differ in reSpect to the behavior of geminate glides. .JA has a rule of glide degemination which applies to word-final geminate glides except in monosgllabic words, while MSA does not have such a rule. The analgs the segmental (15) of stress I framework of ‘ prediction. StI model (McCArt incorporation 1982) into licl prediction. The analgs indicate that ‘ the other apps between the f accordinglg in related to the that in this Cl the first radi In nouns, it is epenthei and in the col VOWel is date Consonants. 254 The analyses of stress assignment in both JA and MSA argue for the segmental linear approach against the metrical approach. Rule (15) of stress assignment which is formulated within the framework of linear'phonology in Chapter 3 yields the correct prediction. Stress assignment on the basis of McCarthy's metrical model (McCArthy, 1979) fails to yield the correct prediction. The incorporation of Hayes' principle of extrametricality (Hayes, 1981, 1982) into McCarthy's model fails also to yield the correct prediction. The analyses of alternating high vowels which appear in JA indicate that they are of two types. One type appears in nouns and the other appears in verbs. In verbs, the vowel which appears between the first and second radical is unpredictable and accordingly we argue that it is underlying. Here, there is a question related to the failure of the vowel to be assigned stress. We argue that in this context stress is assigned to the prefix vowel preceding the first radical by a morphologically conditioned rule. In nouns, the vowel is predictable and accordingly we argue that it is epenthetic. The vowel is a copy of the preceding high vowel, and in the context of a {-highl preceding vowel, the quality of the vowel is determined by that of the preceding and following consonants. The vowel is [u] in the context of a preceding emphatic or velar and a following emphatic or velar except when the preceding consonant is coronal. Elsewhere, the vowel is Iii. Here, the vowel poses potential counter evidence to the stress assignment rules. The vowel is not assigned stress contrary to the predictions of the stress rt rules which ap| evidence in we obtained from 7axuw when t gielding eaxuul because in 96x epenthetic. Ti stress applies The study phonotactic Ct does not allov sgllabic segm _ sgllabic segm Regarding phonologg, tht analysis of A‘ prediction. T vowel in the . COnsonant as Wars as in [ ”Oil-linear m arrangement Will be funda Epenthes DU 8 morpher 255 of the stress rules, and i assume that rules of epenthesis are late rules which apply after and do not feed stress or other rules. Some evidence in support of epenthesis following other rules can be obtained from rule (13) of glide vocalization. The rule applies to 7éxuw when the consonant-initial suffix I-hum] is attached, yielding 2ax0uhum. But the rule fails to apply to daluw-hum because in 7éxuw the vowel is underlying while in déluw it is epenthetic. This indicates that the rule of glide vocalization like stress applies at a level where there are no epenthetic vowels. The study argues that rules of epenthesis apply to satisfy a phonotactic constraint in compliance with the PSP. This principle does not allow the occurence of peaks of sonority between two syllabic segments, and when they do, epenthesis applies, inserting a syllabic segment which represents a peak. Regarding epenthesis within the framework of non—linear phonology, the study argues that the non-linear phonological analysis of Alghazo (1987) does not always yield the correct prediction. The analysis fails to account for the appearance of the vowel in the form of [i] in the context of a preceding emphatic consonant as in [batin] " tummy", and the appearance of Ii] before velars as in [nagid] "criticism". If Alghazo chooses to patch up the non-linear model he will have to take into consideration the linear arrangement of vowels and consonants. In this case, his patching up will be fundamentally linear in nature. Epenthesis optionally applies to fake geminates which are split by a morpheme boundary. The optional application of epenthesis to fake geminates geminate clustl epenthesis doe: indivisible unit applies to fake Well-Formedne between the Cl other, fake ger geminates, ept to the same ti the crossing o Well-Formedn In the app‘ and other rela their SD is me rules are allo the principle morphologies other rules, a Which do not R996rding JA and MSA g 256 fake geminates does not violate the scale—of-sonority principle,- the geminate cluster does not have a sonority peak. In true geminates, epenthesis does not apply because true geminates make one indivisible unit. Within the framework of CV phonology, epenthesis applies to fake geminates because it does not violate the Well-Formedness Condition according to which association lines between the CV tier and the melodic tier should not cross each other; fake geminates belong to two separate tiers. In true geminates, epenthesis does not apply because true geminates belong to the same tier. Hence, the application of epenthesis will result in the crossing of the association lines which is a violation of the Well-Formedness Condition. In the application of the rules of epenthesis, stress assignment, and other related rules, the study argues that rules apply whenever their SD is met. However, constraints on the application of the rules are allowed in the form of general principles. This includes the principle of Proper Inclusion Precedence, the principle that morphologically conditioned rules take disjunctive precedence over other rules, and the principle that rules of epenthesis are late rules, which do not provide input for other rules. Regarding the relation between JA and MSA, the analyses of the JA and MSA data give inconclusive evidence that JA is a descendant of CA as Figure (1) predicts. At the same time, the analyses do not exclude the possibility that JA is a descendant of another Arabic variety which is a sister language to CA the variety from which MSA has developed. Abdo,Daud A. approach. Abumdas,Abd dissertati Abu-Salhn,ls Lingua 74 lAI-Anl, Salm Alghazo,m. non-line Urbana- Al-liunaim, l ?Bi~kita ?GI-mak References Abdo, Daud A. 1969. On stress and Arabic phonology, a generative approach. Beirut: Khayyats. Abumdas, Abdul Hamid Ali. 1985. Libyan Arabic Phonology. Ph.D dissertation, University of Michigan, Ann Arbor. Abu-Salim, Isam M. 1988. Consonant assimilation in Arabic. Lingua 74: 45—66. lAl—Ani, Salman. 1978. The phonological structure of the syllable in Arabic. In Salman Al-Ani (ed), Readings in Arabic linguistics. Bloomington, Indiana: Indiana University Linguistics Club. Alghazo, M. 1987. Epenthesis and syncope in Levantine Arabic: A non~linear approach. Ph.D dissertation, University of Illinois at Urbana—Champaign. Al—aunaim, _saalihah Rashid bunaim. 1985. eal—lahajaat fii eal—kitaal li-Sibawayh vaswaatan wa bunyatan. jeddah: daar 7al—makna. 257 . _ _ '..| 1:: .!l l- L1 _ Anis, Ibrahim. Matbasat I _. 1971. 7 val-enjlo Brame, Michat‘ of Langua, _ 1974. T and Spani Chomsky, Noe New York Guerssel, Moi Linguisti ~. 1978, , 4: 223- Hages, Bruce disserta 258 Anis, Ibrahim. 1952. fii 7al—lahajaat 9al~sarabiyyan Cairo: Matbasat Lajnat val—bayaan eal—sarabiyy. _. 1971. eel—eswaat eal—lubawiyyah. Cairo:Maktabat oal-enjlo eal—misriyyah. Brame, Michael K. 1971. Stress and Arabic phonology. Foundations of Language 7: 556—91. __ 1974. The cycle in phonology: Stress in Palistinian, Maltese, and Spanish. Lingua 5.1: 39-60. Chomsky, Noam and Moris Halle. 1968. The sound pattern of English. New York: Harper and Row. Guerssel, Mohammed. 1977. Constraints on phonological rules. Linguistic Analysis 3:267-305. __ 1978. A constraint on assimilation rules. Linguistic Analysis 4: 223-54. Hayes, Bruce. 1980. A metrical theory of stress rules. Ph.D dissertation, MIT. _. 1986. Inalterability in CV phonolOgy. Language 62: 331-51. Hogg, Richard and C. B. McCully. 1987. Metrical phonology: a Hooper,Joan 48:525-1 _.19_76. A Academi Hudson,Grov tiers. Jo lbrahhn,Moh Prestigh Jonathan Linguist Universi lrshied, 0mg Bedouin Illinois, Jernudd,5j( Arabh Koutsoudas Dhonol 259 coursebook. Cambridge: Cambridge University Press. Hooper, Joan B. 1972. The syllable in phonological theory. Language 48: 525—40. _. 1976. An introduction to natural generative phonology. Academic press. Hudson, Grover. 1985. Arabic root and pattern morphology without tiers. Journal of Linguistics 22:85-122. Ibrahim, Mohammad H. 1984. On the notions 'Standard' and Prestigious' in Arabic sociolinguistics. In Issam Abu-Salim and Jonathan Owens (eds), Proceedings of the Third Annual Linguistics Conference. Department of English, Varmouk University, Irbid, Jordan. Irshied, Omar. 1984. The phonology of Arabic: Bani Hassan - A Bedouin Jordanian dialect. Ph.D dissertation, University of Illinois, Urbana—Champaign. Jernudd, Bjorn H. and Mohammad Ibrahim (eds). 1986. Aspects of Arabic sociolinguistics. New York: Mouton de Gruyter. Koutsoudas, A., Sanders, G. and N011, C. 1974. The application of phonological rules. Language 50: 1—28. ’ff ii IL Inquiry 1 __ 1979b. Ph.D dissI Linguisti __1981.A Linguist Semaan, Kha points 0 ._ 1968. 'Sibawagh (g SibaWaI Reimpre Vennemann, Linguis 260 McCarthy, John. 1979a. On stress and syllablfication. Linguistic Inquiry 10:443-465. __ 1979b. Formal problems in Semitic phonology and morphology. Ph.D dissertation, MIT. Bloomington: Indiana University Linguistics Club, l982. __ 1981. A prosodic theory of non-concatenative morphology. Linguistic Inquiry 12: 373-418. Semaan, Khalil. 1963. Arabic phonetics: lbs Sina‘s Risaalah on the points of articulation of the speech sounds. Lahore. __ 1968. Linguistics in the middle ages. Leiden: E. J. Brill. ‘Sibawayh (sali Bisr bin 9amr, died by 180 AH. or 797 AD). Kitaab Sibawayh. vol. 2 distributed by Aj-Muthanna Library, Baghdad. Reimpression: Bulaq Press 1316 A.H. (Islamic Lunar Calender). Vennemann, Theo. 1972. On the theory of syllabic phonology. Linguistic Berichte 18:1-18. __ 1988. Preference laws for syllable structure. Berlin: Mouton de Gruyter. Zwicky, Arnold M. 1972. Note on a phonological hierarchy in English. Linguistic change and generative phonology, Robert P. Stockwell ..h. LIBRQRIES INICHIGQN STQTE UNIV. III 758 8 1119 I 1011 312L300 I 1.2.: ..