The Jewish Neo-Aramaic Dialect of Dohok: A Comparative Perspective
By
Dorota Molin
TRINITY HALL
This thesis is submitted for the degree of Doctor of Philosophy.
Faculty of Asian and Middle Eastern Studies
University of Cambridge
March, 2021 DECLARATION
This thesis is the result of my own work and includes nothing which is the outcome of work done in collaboration except as declared in the preface and
specified in the text. It is not substantially the same as any work that has already been submitted before for any degree or other qualification except as declared in the preface and specified in the text. The dissertation does not ex- ceed the word limit of 80,000 as set by the Degree Committee for the Faculty of
Asian and Middle Eastern Studies. For Sarah Adaqi and the larger ‘Dohok’ community of Maʿoz Ziyyon, Israel.
– Yarxi xayax, Sota Sarah! CONTENTS
ABSTRACT XV
ACKNOWLEDGEMENTS XVII
LIST OF SYMBOLS AND ABBREVIATIONS XIX
LIST OF TABLES AND FIGURES XXIII
INTRODUCTION 1
Aims and their justification 1
Linguistic and historical background of NENA 3
The linguistic and social background of J. Dohok Neo-Aramaic 4
Literature review 6
Organisation of the thesis 7
Sources & methodology 9
0.6.1. Data collection and sources 9
0.6.2. Transcription 12
0.6.3. Glossing and terminology 13
0.6.4. Linguistic methodology: a theory-grounded grammatical description and study of
microvariation 17
PHONOLOGY
THE SEGMENTS: CONSONANTS AND VOWELS 23
Introduction 23
Consonants 25
1.2.1. Consonantal phonemes 25
1.2.2. Historical background of selected phonemes 27
1.2.2.1. Dental stops and fricatives /t/, /θ/, /d/ and /ð/ 27
1.2.2.2. The pharyngeals /ḥ/ and /ʿ/ 28
1.2.2.2.1. Words with a historical pharyngeal in the vicinity of an emphatic or q: 29 ii CONTENTS
1.2.2.2.2. Words with a lost historical pharyngeal in J. Dohok 30
1.2.2.2.3. Innovated pharyngeals 31
1.2.2.2.4. Historical background of pharyngeals in NENA 31
1.2.2.3. Emphatic phonemes 32
1.2.2.3.1. Phonetic properties of NENA emphatics 32
1.2.2.3.2. The emphatic phonemes in J. Dohok 34
1.2.2.3.3. The historical Aramaic emphatics /ṭ/ and /ṣ/ 36
1.2.2.3.4. Secondary emphatics in native words 37
1.2.2.3.5. Loan phonemes 39
1.2.2.4. /v/ and /f/ 40
1.2.2.5. /q/ 40
1.2.2.6. /ġ/ 40
1.2.2.7. /ʾ/ 41
Vowels 45
1.3.1. Introduction 45
1.3.2. Vowel phonemes 45
1.3.3. Phonetic properties of phonemes and their allophones 48
1.3.3.1. /a/ and /ă/ 50
1.3.3.2. /e/ 52
1.3.3.3. /i/ 53
1.3.3.4. /o/ 53
1.3.3.5. /u/ 55
1.3.3.6. /ǝ/ as allophone and phoneme 56
1.3.4. Diphthongs and their reflexes 58
1.3.4.1. *ay > e 58
1.3.4.2. *aw > o 58
1.3.4.3. *iy >i, *əw > u and *uw > u 59 CONTENTS iii
PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 61
2.1. Phonetically-conditioned processes affecting consonants 61
2.1.1. Devoicing 61
2.1.2. Aspiration 61
2.1.3. Place assimilation 62
2.1.4. Complete phoneme assimilation of the l in L-suffixes 62
2.1.5. Sonorant interchange 62
2.2. Emphasis spread (affecting consonants and vowels) 65
2.2.1. Degrees of emphasis caused by the emphatic phonemes 68
2.2.2. Directionality of spread 70
2.2.3. Domain of emphasis spread 71
2.3. Phonotactics 75
2.3.1. Consonant cluster resolution 75
2.3.1.1. Word-final CC cluster resolved CǝC 76
2.3.1.2. Word-internal CCC cluster resolved CCǝC 76
2.3.1.3. Word-internal CCC cluster with liquid as C2 resolved CǝCC 77
2.3.1.4. Word-initial CCC clusters (with a morpheme boundary between C1 and C2)
resolved CəCC 77
2.3.1.5. CC cluster with C1 /ʾ/ resolved with CVC 78
2.3.2. Syllable structure, vowel length and lexical stress 79
2.3.2.1. Lexical stress 79
2.3.2.2. Loanwords 80
2.3.2.3. Complex verbal and enclitic copula forms 81
2.3.2.3.1. Forms based on šaqəl 83
2.3.2.3.2. Forms based on šqəlle 87
2.3.2.3.3. Forms with enclitic copula 87 iv CONTENTS
2.3.2.3.4. The Imperative 88
2.3.3. Annexation units 89
2.3.3.1. Historical construct state 90
2.3.3.2. Neo-construct state 90
2.3.2.3.5. Apocopation causing syllable closure (CV.CV > CV̄C and CVC) 91
2.3.2.3.6. Apocopation causing syllable opening (C.CV > C.əC) 92
2.3.4. Forms with numerals 93
2.4. Intonation units: nucleus stress and pause 94
MORPHOLOGY
PRONOUNS 101
3.1. Independent personal pronouns 101
3.2. Demonstrative pronouns 102
3.2.1. Near and far deixis 102
3.2.2. Very far deixis 102
3.2.3. Anaphoric pronouns 103
3.3. Pronominal suffixes on nouns and prepositions 103
3.4. The independent genitive particle 106
3.5. Reflexive pronoun 107
3.6. The pronoun of independence 108
3.7. The pronoun of isolation 108
3.8. Interrogative pronouns 109
3.9. Indefinite pronouns 109
OVERVIEW OF NOUNS AND ADJECTIVES 111
Nouns 111
4.1.1. Singular endings 111
4.1.2. Plural forms 112 CONTENTS v
4.1.2.1. Semantically masculine nouns with -a (SG) and -aθa (PL) 113
4.1.2.2. Semantically feminine nouns with θa/ta (SG): -aθa and -e in PL 113
4.1.2.3. Irregular and suppletive plurals 114
4.1.2.4. Nouns with no singular form (pluralia tantum) 114
Adjectives and their agreement with nouns 115
4.2.1. Inflection of adjectives and agreement 115
4.2.2. Uninflected adjectives 116
Annexation of nouns 117
4.3.1. Annexation by means of the genitive particle –ət 117
4.3.2. Annexation with the independent genitive particle did 118
4.3.3. Historical construct state 118
4.3.4. Neo-construct state 119
VERBAL MORPHOLOGY 121
Introduction: Derivational morphology in Semitic 121
The inventory of patterns, stems and inflectional affixes 124
5.2.1. Patterns and stems and their historical background 124
5.2.2. Šaqəl and the inflectional E-suffixes 129
5.2.3. Šqəlle and the inflectional L-suffixes 130
5.2.4. The Participle 131
5.2.5. Imperative 132
Weak verbs in Pattern I 134
5.3.1. Class I of verba primae /Ø/: ʾamər ‘say’ and ʾaxəl ‘eat’ 135
5.3.2. Class II of verba primae /ʾ/ < */ʿ/: ʾarəq ‘run’ 140
5.3.3. Verba primae /y/: yasəq ‘go up’ 141
5.3.4. Verba mediae /y/, e.g. kayəp ‘bend’ 142
5.3.5. Verba mediae /w/: kawəš ‘go down’ 143
5.3.6. Verba tertiae /ʾ/: šameʾ ‘hear’ 144 vi CONTENTS
5.3.7. Verba tertiae /y/: maxe ‘hit’ 146
Weak verbs in Pattern II 148
5.4.1. Verba mediae /y/: mhayər ‘dare’ 148
5.4.2. Verba mediae /w/: mjawəb ‘answer’ 149
5.4.3. Verbs with identical C2 and C3: t-l-l ‘get wet’, x-l-l ‘wash (one’s hands)’ 149
Weak verbs in Pattern III 151
5.5.1. Verba primae /Ø/: maxəl ‘feed’ 151
5.5.2. Verba primae/ʾ/ maʾărǝq ‘cause to flee’ 153
5.5.3. Verba mediae /w/, e.g. makuš ‘bring down’ 154
5.5.4. Verbs primae /y/, e.g. mayʾəl ‘bring in’ 155
Irregular verbs 156
5.6.1. ‘Want‘ ʾə́be (< *b-ʾ-y) 156
5.1.6.1.1. Overview of forms 156
5.1.6.1.2. Historical background 157
5.6.2. ‘Know’ yaʾe (< *y-ð-ʾ) 163
5.6.3. ‘Give’ yawəl (< *y-h-w) 168
5.6.4. ‘Bring’ meθe (< *ʾ-θ-y) 170
5.6.5. ‘Be able’ ʾəṃṣe (ṃ-ṣ-y) 173
5.6.6. ‘Be born’ and ‘be’ hawe (*h-w-y) 174
Towards a typology of verbal derivation in NENA 178
5.7.1. Root and/or template opacity 179
5.7.2. Stem uniformity across a paradigm and the derivation of stems from stems 181
5.7.3. Analogical change (‘reanalysis’) of the pattern 182
5.7.4. Loanwords 184
5.7.5. Summary 184
CONTENTS vii
SYNTAX AND MORPHOSYNTAX
THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS 189
Introduction 189
Objects of monotransitive 192
6.2.1. Šaqəl-based forms except qam-šaqəl-la with L-suffixes 192
6.2.2. Past perfective forms 193
6.2.2.1. Qam-šaqəl-la with L-suffixes 193
6.2.2.2. Šqəlle with E-suffixes 194
6.2.2.3. Šqəlle with the independent prepositional pronoun ʾal- 198
6.2.3. The Imperative with L-suffixes 198
6.2.4. Wele šqila (resultative) with ʾəll- 199
Ditransitive and constructions with object-like arguments 200
6.3.1. Šaqəl-based forms except qam-šaqəl-la 200
6.3.2. Šaqəl with L-suffixes 200
6.3.3. Šaqəl with the independent prepositional pronoun ṭa- (and ʾəll-?) 202
Past perfective forms 202
6.4.1. Šqəlle with L-suffixes 202
6.4.2. Šqəlle with the independent prepositional pronouns ṭa- (or ʾəll- and ʾəbb-) 204
6.4.3. Qam-šaqəl-la with L-suffixes 205
Object (P, T and R) affixes on past perfective forms in LD in general 208
Differential object marking 212
6.6.1. Word order 214
6.6.2. Definiteness in animate objects 215
6.6.3. Definiteness in inanimate objects 215
6.6.4. (Grammatically) indefinite objects 216
6.6.5. Inalienable (definite and inanimate) objects 216
6.6.6. Objects in narrow focus 217 viii CONTENTS
Summary 217
THE MORPHOLOGY AND SYNTAX OF THE COPULA 219
Introduction 219
J. Dohok copulas: morphology and basic syntax rules 221
7.2.1. Basic copula 221
7.2.2. The wele copula (contingent state and verbal) 222
7.2.3. Negative present copula 223
7.2.4. Basic past copula 224
7.2.5. Alternative past copula (wele) 225
7.2.6. Negative past copula 226
7.2.7. Future and irrealis copulas 226
The semantics and pragmatics of NENA copula clauses 227
7.3.1. Semantic features: permanent vs contingent states 227
7.3.2. Pragmatic features of copula clauses: thetic vs categorical 228
Wele/basic copula + verbal predicate 230
7.4.1. 3rd person clauses 230
7.4.2. 1st/2nd persons: ordinary copula–predicate 236
Wele/basic copula + non-verbal predicate clauses 237
7.5.1. 4.1. 3rd person clauses: presentative copula + predicate 237
7.5.2. 4.2. 1st/2nd person clauses: basic copula + predicate 239
7.5.3. Clauses with the Participle (wele + predicate in 3rd persons) 241
Non-verbal predicate=basic copula 244
7.6.1. 3rd person copula clauses: predicate=basic copula 244
7.6.2. 1st/2nd person copula clauses: predicate + basic copula 246
Basic copula + predicate (non-verbal, permanent property) 246
Summary of present tense copula clauses in J. Dohok 247
Past tense copula clauses 248 CONTENTS ix
7.9.1. Clauses with the basic past copula 248
7.9.2. Clauses with k-hawe-wa (*h-w-y, hawe) ‘used to be’ as a copula 251
7.9.3. Alternative past copula and its identity with the present wele copula 252
Negated copula clauses (all clause types) 254
Irrealis copula clauses 255
Clauses lacking a copula 255
Comparative and diachronic discussion 258
7.13.1. Introduction: Distinctive and areal features of NENA copulas and their clauses 258
7.13.2. The L-copula (most dialects third persons only) 260
7.13.3. The quasi-verbal basic copula (1st/2nd persons) 267
7.13.4. The presentative (‘deictic’) copula 269
THE TENSE-ASPECT-MOOD SYSTEM 275
8.1. Introduction 275
8.1.1. Defining ‘meaning’: theoretical assumptions and methodology 276
8.1.1.1. Grammatical meaning 276
8.1.1.2. The cooperation between lexical and grammatical meaning 278
8.1.1.3. The cooperation between grammatical meaning and pragmatic information 279
8.1.1.4. The synchronic system as a point within a diachronic process 280
8.1.2. Categories of tense, aspect and mood and their terminology 281
8.1.2.1. Tense 281
8.1.2.2. Aspect 282
8.1.2.3. Mood 284
8.2. Šaqəl 285
8.2.1. Main clause irrealis mood 285
8.2.2. Subordinate clause irrealis mood 287
8.3. K-šaqəl 288
8.3.1. Dialogue: continuous, habitual and gnomic 289 x CONTENTS
8.3.2. Past narrative and description 291
8.3.2.1. Habitual past 291
8.3.2.2. Narrative present 292
8.3.2.3. Stative past 294
8.4. Šaqəl and k-šaqəl for irrealis mood in past purpose clauses 295
8.5. B-šaqəl 296
8.5.1. Dialogue 298
8.5.1.1. Future tense of predictive (epistemic) mood 298
8.5.1.2. Conventionalised habitual present (of Ø-m-r,ʾamər ‘say’) 300
8.5.1.3. Modal future: ‘why‘ and ‘either-or’ clauses 301
8.5.1.4. Modal relative future: apodosis of real conditional clauses 302
8.5.1.5. Modal future: protasis of real conditional clauses (vividness effect?) 303
8.5.2. Past narrative 304
8.5.2.1. Relative future 304
8.6. La k-šaqəl 304
8.6.1. Negated predictive future 305
8.7. Šqəlle (and qam-šaqəl-la) 306
8.7.1. Perfective past (narrative foreground, dialogue) 306
8.7.2. Relative past in clauses with future time reference (including protasis of real
conditionals) 307
8.8. Šqəl-wa-le 308
8.9. Šaqəl-wa 309
8.9.1. Irrealis 310
8.9.2. Realis 311
8.10. K-šaqəl-wa 312
8.10.1. Habitual past (narrative background & descriptions of lifestyle narrative) 314
8.10.2. Stative past in the verbs yaʾe ‘know’ and ʾəbe ‘want, to love’ 317 CONTENTS xi
8.10.3. Past continuous 321
8.11. B-šaqəl-wa 322
8.11.1. Future in the past (perfective past) 323
8.11.2. Unreal conditional 323
8.11.3. Habitual past 324
8.12. K-šaqəl(-wa) in a diachronic and comparative perspective 328
8.13. B-šaqəl(-wa) in a diachronic and comparative perspective 335
8.14. Copula + Participle (wele šqila) 343
8.15. Copula + in-Infinitive (wele bə-šqala) and wal/hol/hole k-šaqəl 348
8.15.1. Wele bə-šqala 350
8.15.2. Wəl/hol/hole k-šaqəl 351
8.15.3. The history of the NENA continuous forms 353
8.16. Expression of other TAM categories 358
8.16.1. Proximative 358
8.16.2. Inchoative 360
8.17. Conclusion 360
WORD ORDER 363
9.1. Introduction 363
9.1.1. Aims 363
9.1.2. Data, methodology and terminology 364
9.2. Word order and information structure 370
9.3. Default word order and its flexibility in clauses with a subject and another argument 374
9.3.1. Frequency of clauses by number of arguments 374
9.3.2. Word order in two-argument clauses 374
9.4. The core arguments S, A and P 376
9.4.1. Position 376 xii CONTENTS
9.4.2. Frequency of full expression 378
9.4.3. Position of independent personal pronouns 380
9.5. Transitive clauses 381
9.5.1. Os and general comments on information structure 381
9.5.1.1. Position and differential object marking 381
9.5.1.2. Object defocalisation 382
9.5.1.3. New topics or topic reactivation through clause-initial position and prosodic
separation (all arguments) 383
9.5.1.4. Arguments in narrow focus 385
9.5.1.5. Clause repetition and word order inversion in narrative 386
9.5.2. Agents 387
9.6. Ditransitive clauses 389
9.6.1. Constructions with an indirect and a direct object 390
9.6.2. Constructions with two direct objects 396
9.6.2.1. Verbs of filling and naming 396
9.7. Intransitive clauses: Subjects (Ss) and the argument of existential clauses 403
9.7.1. Sentence focus’ 403
9.7.2. The presence of Goals 407
9.7.3. The role of animacy and discourse status 408
9.7.3.1. Discourse status 410
9.7.3.2. Discourse status in inanimate subjects 412
9.7.3.3. Discourse status in animate Subjects 414
9.7.4. Clauses with p-y-š (payəš) ‘break out, occur’: sentence focus and non-canonical
subjects 416
9.7.5. Existential clauses and their parallels with verbal intransitive clauses 420
9.8. Goals and obliques 425
9.8.1. Goals 425 CONTENTS xiii
9.8.2. Other obliques 428
9.9. Conclusion 429
CONCLUSION 433
BIBLIOGRAPHY 443
ABSTRACT
This thesis provides a linguistic analysis of key issues in the Neo-Aramaic dialect that was historically spoken by the Jewish community in Dohok, north-western
Iraq. This linguistic variety belongs to the North-Eastern sub-branch of Neo-Ara- maic – a group of dialects and languages used by Jews and Christians in parts of
Turkey, Iraq and Iran. Like many other linguistic varieties from the North-Eastern branch, the dialect of Dohok now faces imminent extinction.
This work is based on a language questionnaire and a corpus of personal narra- tives, descriptions of traditions and folktales which I collected in my own field- work among speakers now living in Jerusalem, Israel.
My analysis includes comparisons with other North-Eastern Neo-Aramaic varie- ties. These comparative data help to identify both the distinctive and the shared features of the Jewish Dohok dialect, and thus help place this dialect within the general linguistic landscape of the North-Eastern group. Such a comparison also leads on a few occasions to a general discussion of linguistically noteworthy phe- nomena in North-Eastern Neo-Aramaic, the analysis of which is the secondary aim of this work. In addition to documenting the linguistic diversity of Neo-Aramaic, this thesis also aims to make the field of Neo-Aramaic accessible to general lin- guists – inter alia, by providing data glossing.
The discussion begins with the phonology and phonetics of both segments and larger units. This is followed by morphology, where I focus especially on the verbal xvi ABSTRACT
system. At the heart of this dissertation is the investigation of syntax and morpho- syntax. I describe and analyse strategies for the expression of pronominal objects, the syntax, semantics and pragmatics of copula clauses and the tense-aspect-modal system. The final chapter is the first quantitative study of Neo-Aramaic word order and compares the dialect of the Jews of Dohok with that of Sanandaj (Iran).
ACKNOWLEDGEMENTS
This project would not have been possible without the help and support of count- less individuals. Firstly, I am immensely grateful to all the members of the Dohok community of Maʿoz Ziyyon and Jerusalem. Israel. A special word of thanks is due to Sarah Adaqi, Tzvi Avraham and Yefet, who showed me great hospitality and patience, and dedicated countless hours of their time during my fieldwork. I also owe a great deal to Batya Aloni and Naftali Mizrahi of the Zakho community for their help in organising interviews in Israel, in learning Lishana Deni and for the interviews which I held with them. I hope that this thesis will help preserve your wonderfully rich cultural and linguistic heritage. ʾilaha ʾawəð ḅāš ta kulloxun!
My deepest thanks goes to my doctoral supervisor, Prof Geoffrey Khan. It has been an absolute honour to be trained by a scholar of such experience and class, and I remain indebted to him for introducing me to the fascinating world of Neo-Ara- maic. I also benefited greatly from Prof Khan’s kindness and moral support, and deeply admired his humility and respect with which he approaches those around him.
A special thank you belongs to my dear friend Paul Noorlander, whose moral sup- port, encouragement and good humour proved invaluable in the course of thesis writing. I also benefited greatly from our linguistic discussions. Paul is also cred- ited here for the technical support with thesis formatting, and for his contribution xviii ACKNOWLEDGEMENTS
to the chapters on word order and pronominal objects. The details of his contri- bution are listed in the relevant chapters.
I am truly grateful to Dr Aaron Hornkohl for his encouragement and our stimulat- ing linguistic discussions. I also owe a great deal to Prof Hezy Mutzafi, who helped me connect with the Lishana Deni communities in Israel, and whose fascination for and expertise in Neo-Aramaic inspired me deeply. Other colleagues and friends I would like to thank are Estara Arrant, for encouragement and support with statis- tics and coding, Dr Lidia Napiorkowska for her invaluable feedback, and Dr Mi- chael Chyet, Dr Hiwa Asadpour and Dr Masoud Mohammadirad for advice on Ira- nian languages. There were many other friends and colleagues who supported me, but who I am unable to include here. To all of you, thank you!
A special thanks belongs to my parents and brothers for their love, support and for believing in me. Our multi-lingual household also became the nurturing envi- ronment in which my linguistic curiosity was piqued for the first time. Without you, this work would not have been possible. Dziękuję! LIST OF SYMBOLS AND ABBREVIATIONS
Abbreviations
A Agent = subject of transitive
Ar Arabic
BEN beneficiary
BK Bahdini Kurdish
C. Christian North-Eastern Neo-Aramaic dialect
COP copula
CP common plural
DEIX deictic
EXIST existential
FS feminine singular
FUT future tense
G Goal
H _ Modern Hebrew
IMP imperative
INF infinitive
IRR irrealis mood
J. Jewish North-Eastern Neo-Aramaic dialect
LD Lishana Deni Jewish North-Eastern Neo-Aramaic
MS masculine singular
NEG negation xx LIST OF SYMBOLS AND ABBREVIATIONS
NENA North-Eastern Neo-Aramaic
O Object
P Patient = object of monotransitive
PERF past perfective
PL common plural
PST past
PROG progressive
R Recipient of ditransitive
REAL realis mood
RES resultative participle
S Subject = subject of intransitive
SG common singular
T Theme of ditransitive
TZ Trans-Zab Jewish North-Eastern Neo-Aramaic
V verb
Symbols
< derived from, ranks lower than
> developed into, ranks higher than
* ungrammatical/unattested and reconstructed historical form
Ø zero morpheme or phoneme LIST OF SYMBOLS AND ABBREVIATIONS xxi
[] morpheme-level and phonetic representation
// phonological/underlying morphological representation
3,2,1 person
p statistical significance
Φ correlation coefficient
LIST OF TABLES AND FIGURES
Tables
Table 1: Personal details of the participating J. Dohok speakers 10
Table 2: Forms based on šaqəl 15
Table 3: Forms based on šqəlle 16
Table 4: Forms based on the Infinitive (šqala) 16
Table 5: Forms based on the Imperative (šqul) 16
Table 6: Forms based on the resultative Participle (šqila) 16
Table 7: Forms of the copula 17
Table 8: Examples of nominal morphology glosses 17
Table 9: The consonantal phonemes of J. Dohok 25
Table 10: Reflexes of historical Aramaic dental fricatives in LD 26
Table 11: Differences in verbal stems illustrating different status of [ʾ] from /ʾ/ and /ʿ/ 37
Table 12: Vowel length in ms. in J. Dohok per syllable type 43
Table 13: F1 values in Hz of [uː], [u] and [o] (short) 46
Table 14: Sonorant interchanges in LD 64
Table 15: F2 values in Hz of /a/ in plain and emphatic environment 69
Table 16: Values of /a/ in Hz before and after an emphatic 71
Table 17: F2 values of phonemes showing effects of emphatics on other phonemes within the same word 72
Table 18: independent personal pronouns in J. Dohok 101
Table 19: Deictic pronouns in J. Dohok 102
xxiv LIST OF TABLES AND FIGURES
Table 20: Possessive suffixes in J. Dohok 103
Table 21: Inflection of the independent genitive particle with possessive suffixes in J. Dohok 106
Table 22: Inflectional stems for all verbal Patterns in J. Dohok 128
Table 23: The šaqəl inflection in J. Dohok 130
Table 24: The šqəlle inflection in J. Dohok 131
Table 25: The resultative Participle (šqila) inflection in J. Dohok 131
Table 26: The Imperative (šqul) inflection in J. Dohok 132
Table 27: Šaqəl and k-šaqəl stems in historical primae /ʾ/ verbs in J. Dohok 135
Table 28: Inflection of the verb 'want' across Jewish dialects 159
Table 29: Inflections of the verb ‘know’ across Jewish dialects 163
Table 30: Inflections of the verb ‘give’ across Jewish dialects 168
Table 31: Inflections of the verb ‘bring’ across LD 171
Table 32: Tense-aspect-modal forms of ‘be’ and ‘be born’ in J. Dohok 175
Table 33: Frequencies of the two competing transitive constructions in LD and the middle voice construction (šqəlle with construction E-suffixes) 209
Table 34: Frequencies of constructions for encoding reflexive objects in LD 210
Table 35: The basic present copula in J. Dohok 222
Table 36: The wele (contingent state/eventive) copula in J. Dohok 223
Table 37: The negative present copula in J. Dohok 224
Table 38: Basic past copula in J. Dohok 225
Table 39: Alternative past copula in J. Dohok 225
LIST OF TABLES AND FIGURES xxv
Table 40: Features of verbal copula clauses in LD and C. Barwar 235
Table 41: 1st/2nd person copulas and inflection of h-w-y 'be' in J. Dohok 268
Table 42: Presentative copulas and their properties across LD 271
Table 43: Copula paradigms in C. Diyana-Zariwaw 272
Table 44: Total counts of annotated data 365
Table 45: Illustrative contingency table 369
Table 46: Overall word order typology by verb position 374
Table 47: Position of the core arguments S, A and P relatively to the verb 377
Table 48: Argument frequency per argument’s grammatical type 378
Table 49: Differential object marking through verbal agreement and object position in J. Dohok 381
Table 50: The position of R (indirect object) and T (direct object) in J. Dohok 391
Table 51: Properties of ditransitive constructions in J. Dohok and J. Sanandaj 401
Table 52: Position of nominal Subjects in J. Dohok 403
Table 53: Position of V and S in J. Dohok intransitive clauses 407
Table 54: The association between word order and animacy in J. Dohok
(contingency table) 409
Table 55: The association between word order and discourse status in J. Dohok
(contingency table) 410
Table 56: The association between word order and discourse status of inanimate
Subjects in J. Dohok 412
xxvi LIST OF TABLES AND FIGURES
Table 57: The association between word order and discourse status of animate
Subjects in J. Dohok (contingency table) 414
Table 58: Position of Goals in J. Dohok and J. Sanandaj 425
Figures
Figure 1: The distribution of vowels in plain vs emphatic environments 68
Figure 2: Phonetic manifestation of lexical stress in ḥakoma 79
= Figure 3: lack of aspiration in p̮awe [p aːwe] 177
Figure 4: aspiration in kawe [khaːwe] 177
Figure 5: Phi coefficient (Stefanowitsch 2020, 153) 369
Figure 6: Association between discourse status (‘d-new’ and ‘d-old’), animacy
(‘animate’ and ‘inanimate’) and word order (‘VS’ and ‘SV’) in intransitive clauses with a full nominal Subject in J. Dohok (MCA) 408
INTRODUCTION
Aims and their justification
The objective of this thesis is to provide a linguistic analysis of key issues in the
Jewish North-Eastern Neo-Aramaic (henceforth ‘NENA’) that was historically spoken in the city of Dohok (north-western Iraq).1 Comparative data from other
NENA varieties are also brought to bear. Such data help to identify both the distinctive and shared features of the Jewish Dohok dialect, and thus help to place this dialect within the broader linguistic landscape of NENA. Such a comparison also leads on a few occasions to a discussion of linguistically noteworthy NENA- wide phenomena, the analysis of which is the secondary aim of this work. I discuss the historical development of these phenomena and/or consider their implications for the linguistic typology of NENA.
In doing so, this thesis addresses two urgent challenges in Semitic linguistics: to document the rapidly disappearing diversity of Neo-Aramaic, and to help build a bridge between NENA studies and general linguistics. The existing Neo-Aramaic varieties are the last surviving branch of Aramaic – one of the world’s longest documented linguistic traditions with around 3,000 years of documented history.
Thanks to this, Aramaic is a true treasure trove – not only for Semitists, but also for, inter alia, typological, historical and contact linguists. Within Neo-Aramaic,
1 ‘Dohok’ is the NENA pronunciation; the name on maps is ‘Dihok’ (after Kurdish) or ‘Dahuk’ (after Arabic). 2 INTRODUCTION
the NENA subgroup represents an especially diverse group, including some 150 distinct varieties (Khan 2007). Despite the potential which NENA represents for scholarship, many NENA varieties remain undocumented and are now severely endangered (Khan 2007, 4). This situation follows a worldwide trend: by 2100, up to 90% of the world’s languages are predicted to become extinct, causing devastating losses not only for human culture, but also for empirical linguistics
(Krauss 1992). Thus, the grammatical analysis of endangered languages such as the Jewish NENA dialect from Dohok is of primary interest to both Semitic and general linguistics.
Second, despite its unquestionable potential for linguistics at large, much of the
NENA literature remains inaccessible to non-specialist linguists. This thesis hopes to build bridges between Aramaic and general linguistics, and this is accomplished in several ways. First, for the first time in a NENA grammar, glossing is provided for all primary data in cases where this is relevant and helpful, making them accessible to non-specialists. Second, linguistic frameworks are provided for each discussion to ensure methodological and terminological clarity and transparency.
And finally, it is suggested at a few occasions that a meticulous, language-specific analysis rooted in a robust linguistic framework has the potential to enhance current linguistic theory, such as grammaticalisation pathways and typological tendencies. Clearly, this area represents a promising avenue for large-scale
(quantitative and comparative) research in the future. INTRODUCTION 3
Linguistic and historical background of NENA
In the second half first millennium BCE, Aramaic was a lingua franca across vast swaths of the Near East. It continued as a vernacular of communities and a vehicle of sacred traditions of Judaism, Christianity and Mandaeism. Following the rise of
Islam, Aramaic gradually developed to be a vernacular of ethno-religious minorities in pockets of the Middle East (Khan 2007, 3).2
The surviving vernacular dialects of Aramaic can be divided into four distinct subfamilies:
(1) the western group spoken by Christians and Muslims in Maʿlula and other
villages in the region of Damascus, Syria;
(2) the Ṭuroyo group, spoken by Christians in the Ṭur ʿAbdin region of south-
eastern Turkey;
(3) Mandaic, spoken by Mandaeans in the city of Ahwaz in Iran and the
surrounding region
(4) north-eastern-Neo-Aramaic (NENA), spoken by Jews and Christians in
south-eastern Turkey, northern Iraq and western Iran (Khan 2007, 3-4).
2 For an anthropological study of the ‘Kurdish’ NENA-speaking Jews and a brief overview of their history, see
Brauer (1993). See also Mutzafi (2008a), and especially the primary data corpus therein, which is a priceless source of information about the culture and customs of the Kurdish Jews. 4 INTRODUCTION
The NENA group is by far the largest and most diverse one. This diversity is conditioned not only by geography, but also by communal identity: Jewish (J.) and Christian (C.) varieties from the same area are commonly distinct if not mutually incomprehensible (Khan 2007). The differences between some NENA varieties—especially between those spoken in areas far-removed from each other—are substantial and even impede mutual intelligibility, meaning that it is more accurate to speak not only of NENA dialects, but also languages. The present investigation uses interchangeably the terms ‘dialect’ and ‘variety’. The former designation reflects a decision to maintain the tradition in existing NENA scholarship.
The linguistic and social background of J. Dohok Neo-
Aramaic
The dialect of the Jews of Dohok (henceforth ‘J. Dohok’) is most closely affiliated with the other Jewish dialects historically spoken west of the Great Zab river, generally known as Lishana Deni ‘our language’, henceforth ‘LD’. Dialects belonging to this group were spoken also in the towns or villages of Zakho,
Amedia, Betanure, Nerwa, Aradhin, Atrush (north-western Iraq), as well as Challa and Gzira in south-eastern Turkey. In the early 1950s, nearly all speakers of Jewish
NENA emigrated to Israel, where the linguistic transmission was soon brought to a halt due to severe social and political pressures. Today, J. Dohok is on the verge INTRODUCTION 5
of extinction, having only about twenty remaining active speakers. These speakers were born in the 1930s or 1940s in Dohok, or in the 1950s in Israel. Most of the
Dohok emigrants settled in the Jerusalem area.
The town of Dohok was historically inhabited by Kurdish Muslims, Jews and
Christians, and the Jewish presence in the town is attested in the 18th century (Ben
Yaacov 1981, 56). Jews and Christians, however, constituted a small minority of the population. According to Informant 2, there were four Jews and three
Christians in his school class, and the rest were Muslims. My informants and Ben
Yaacob (1981) report that the Jews of Dohok were craftsmen (weavers, tanners, shoemakers) or pedlars, but also agriculturalists, working in the vineyards and orchards that reportedly used to surround Dohok. Others hired themselves for labour (e.g. the transport of goods). The Christian NENA speakers in Dohok came originally from nearby villages such as Mar Yaqo or Shiuz, and recently, likely from further afield. Thus, the Christian Dohok dialect, whose brief sketch may be found in Sabar (1995), is of recent and most likely heterogenous origin.
In addition to LD, Jewish NENA dialects were also historically spoken east of the
Great Zab river – in north-eastern Iraq and western Iran. These dialects are referred to in the present work as ‘TZ’ (Trans-Zab), after Mutzafi (2008b). The
Trans-Zab group stands apart not only from the LD cluster, but also from most other Christian NENA varieties (Mutzafi 2008b), while also displaying strong internal diversification. In many linguistic areas, the TZ group displays the highest 6 INTRODUCTION
degree of interference from the Iranian languages in contact. For instance, 69 percent of nouns the J. Urmi (north-western Iran) variety are loanwords (Garbell
1965). In word order, the Trans-Zab group underwent a shift to subject-object- verb from the earlier Aramaic subject-verb-object order (Noorlander and Molin, forthcoming 2022), which is also due to Iranian interference (Khan 2009, 11-2).
Literature review
The study of Jewish NENA in general was pioneered by Hans Polotsky (e.g. 1967) who focused on J. Zakho, though much of the primary material which he collected and of his research remained unpublished (cf. Cohen 2012, 2). The present thesis is the first comprehensive description of J. Dohok; a (morphological) description of its verbal system may be found in Khaykin (2011; Master’s thesis).
Comprehensive descriptions of other LD dialects have been undertaken for J.
Betanure (Mutzafi 2008a), J. Amediya (Hoberman 1989 and Greenblatt 2008) and J. Challa (Fassberg 2010). These grammars focus for the most part on morphology, lexicon and primary data (the latter two especially in Mutzafi
2008a). Cohen’s work (2008; 2012 and 2017) makes a major contribution to the study of LD syntax. Cohen studies in great detail the functions of verbal forms in narrative and different types of copula clauses, categorised by copula and predicate type.
The grammars of Khan represent the most comprehensive description of NENA dialects at large, culminating in the voluminous grammars on C. Barwar (northern INTRODUCTION 7
Iraq) and C. Urmi (north-western Iran) (respectively, Khan 2008a, vols.1-3 and
2016, vols. 1-4). Among Khan’s key contributions are his meticulous studies of pragmatic (including discourse pragmatics) functions of different clause types, and his challenge to the traditional Semitic root-and-pattern verbal morphology, based on the C. Urmi data (2016). Khan (2006) has also laid the foundations for the study of NENA copula clauses, and especially their pragmatic functions and possible development.
A great contribution to the study of morphosyntactic alignment in NENA (the expression of pronominal subjects of intransitive and transitive, and of objects) has been made by Coghill (2016, historical reconstruction) and Noorlander (2018 and 2021; historical and typological approach respectively). The work of
Napiorkowska (2015a and b) and Khan (especially 2016) represents a landmark in the study of NENA phonetics and phonology; they apply current empirical phonetic methods to understand the nature of phenomena such as suprasegmental emphasis and emphasis spread.
Organisation of the thesis
As far as possible, this study prioritises the understudied areas within NENA. It opens with a phonological study, which subsumes the phonemic inventory (§1), key phonetic processes, phonotactics and intonation units (§2). It includes an empirical study of emphasis spread, missing in the existing LD varieties, and attempts to synthetise the rules of stress placement and syllable structure into a 8 INTRODUCTION
single set of phonotactic rules which can account for all the forms, and which is likely applicable to other NENA varieties.
This is followed by morphological overviews of pronouns (§3) and nominals (§4), whose brevity is justified by the fact that the morphology of J. Dohok bears a close resemblance to the already-described morphology of other LD dialects. The subsequent chapter offers a detailed study of verbal morphology (§5) which points out some features characteristic (exclusively) of Jewish NENA dialects. It also aims to show the limitations of the traditional Semitic derivational model for NENA, arguing for a modified model based chiefly on patterns (rather than roots and patterns), combined with elements of stem-and-affix derivation.
Systematic research on verbal and clause syntax within LD is still lacking, which motivates placing syntax at the heart of this thesis. I study the expression of pronominal objects and related arguments, showing the large inventory of object- expressing strategies in NENA, typologically highly noteworthy. This is followed by the morphology and syntax of the copula (§7) and the verbal tense-aspect-mood system (§8). Finally, the study of word order (§9) comes to fill the need in the wide NENA scholarship for statistical study of the clause which takes into account not only information structure, but also argument structure. This investigation makes generalisations about the typology of NENA word order.
INTRODUCTION 9
Sources & methodology
0.6.1. Data collection and sources
This thesis relies largely on fieldwork undertaken by the author with speakers of
J. Dohok and of other LD varieties in summer 2017, spring 2018 and summer
2019. Finding a large number of informants proved highly challenging.
Limitations in data sampling opportunities are frequent among documentation linguists working on endangered varieties (Woodbury 2003). This inevitably poses a problem for corpus-driven approaches (Cox 2015, 118), which strive to represent natural variation within a language. In my data collection, I strived for an optimal balance between speakers’ competence (especially the apparent degree of interference from other languages or LD varieties) and between wide sampling. In the end, the data come from six speakers (see Table 1 below for details), which for the highly endangered LD group already represents a large sample, relatively to some other grammars. The purpose of the data collection was explained to all speakers and they gave their full oral consent (preferred over written one due to the illiteracy of one speaker).
10 INTRODUCTION
Table 1: Personal details of the participating J. Dohok speakers
Speaker Age (at start of Sex Place of birth
interviewing)
1 82 M Dohok, Iraq
2 86 M Dohok, Iraq
3 88 F Dohok, Iraq
4 late seventies (?) M Dohok, Iraq
5 early eighties (?) M Dohok, Iraq
The data for the morphological part of the study were largely obtained through elicitation, but the remaining parts draw from a natural corpus data, unless indicated otherwise. As far as possible, elicitation was avoided, as it produces data without a discourse context (relevant for pragmatic nuance) and sometimes in an unnaturally careful register. If elicitation was necessary, I tried to create a natural conversational context and elicit the data from several speakers.
The corpus used for this study comprises personal narrative, description of past customs and traditions, and folktales, and contains almost 8,000 lexical tokens. INTRODUCTION 11
Due to the nature of endangered language fieldwork, dialogue is by and large restricted to conversations within narrative, meaning that language with present tense deixis is inevitably underrepresented. The linguistic profile of a corpus is naturally determined by the genres included (Glynn 2010, 8; Cohen 2012), so I tried to include different genres in order to gain a more comprehensive picture into the use of the language.
Material representing most other NENA dialects has been taken from published grammars, unless no source is indicated, in which case the data come from my own fieldwork. This thesis also includes references to pre-NENA or early NENA linguistic varieties. The latter includes a corpus of biblical homilies referred to here as ‘Early J. Nerwa.’ The original text of these homilies likely goes back to the early 16th century (Sabar 1976, XLY) and was probably copied in Nerwa or
Amediya (north-western Iraq) in the 17th century. Its language can generally be considered a predecessor of the contemporary LD varieties. The present thesis relies on a transcription of these homilies published by Sabar (1976). Another source of early NENA is the north-western Iraqi Christian poetry, whose parts with a linguistic commentary were published by Mengozzi (2012). This corpus comes from 17th-19th century Telkepe and Alqosh (Mosul plain), but is archaising in nature (Mengozzi 2012).
12 INTRODUCTION
0.6.2. Transcription
This thesis uses for the most part the transcription symbols specific to the NENA literature; their phonetic values are presented in §1.2.1. and §1.3.2. A minimalist transcription system of vowels is favoured; this system assumes a phonetically predictable opposition of long vowels (open stressed syllables) and of short ones
(elsewhere). Lexical stress typically falls on the penultimate syllable.
Consequently, vowel length or shortness and lexical stress are only indicated when not predictable from these simple rules (the latter is denoted by )́ . One exception to this transcription policy are monosyllabic words with a in an open syllable, which is therefore stressed. The most common of these words are xa ‘one, a certain’, the negator la, ma ‘what’ and ta ‘for (+noun)’, where the stressed vowel a is always short. Being lexically predictable, the shortness in such words is not marked. In code-switched words, long vowels are explicitly indicated.
Aside from unpredictable lexical stress, only nucleus stress is generally marked
(with the sign ),̀ and the end of an intonation unit is indicated by the symbol ˈ.
Sometimes, a single intonation unit has two nucleus stresses, in which case both are indicated in the transcription.
The symbol ‘=’ is used for enclitics. In J. Dohok, the only certain, phonetically verifiable type of clitic is the present third person copula (‘L-copula’), so this sign is used only in those cases. The symbol ‘-’ is used to show the structure of INTRODUCTION 13
morphologically complex units which are one prosodic word, and for monophonemic prepositions and particles.
Foreign words and phrases which reflect spontaneous code-switching, rather than loans, are marked with a superscript ‘H’, ‘A’ and ‘K’ – respectively, Modern
Hebrew, Arabic and Bahdini Kurdish.
This transcription method is by-and-large also applied to other NENA dialects, leading to occasional modifications of transcriptions taken from published sources.
Data from the TZ group always have lexical stress indicated; in contrast to the rest of NENA, these dialects tend to have word-final stress (Khan 2009, 4).
0.6.3. Glossing and terminology
Tables 2-8 below summarise the glossing system. Each gloss is provided with its prototypical meaning. I devised this system specifically for J. Dohok in consultation with Geoffrey Khan and Paul Noorlander. For convenience, such a system—with a few exceptions—is also used for other NENA varieties. The glosses for finite verbal forms are purely morphological – the forms are referred to by the
3MS form of the relevant inflection of the verb š-q-l ‘take’. This non-functional terminology is also used in the text to avoid the use of functional labels, which are not comprehensive. For the Infinitive, Imperative and (resultative) Participle, functional labels are used. 14 INTRODUCTION
Verbal lexemes are given by their—ideally synchronic—roots if a plausible root can be identified, followed by their 3MS šaqəl form to show their derivational pattern. Nominal forms are glossed only in rare cases. When necessary, grammatical functions of arguments are indicated; for the inventory of tags, see
§9.1.2.
INTRODUCTION 15
Table 2: Forms based on šaqəl
Form Gloss Tense Aspect Mood
Ø-šaqəl IRR- non-past irrealis
3 FUT/IRR- future irrealis/predictive
la šaqəl NEG IRR- non-past irrealis (including negative imperative) k-šaqəl REAL- present realis
wəl/hole/hol k-šaqəl DEIX present progressive realis (non-stative?) la k-šaqəl NEG REAL- present realis
future (for predictive Patterns II- IV) b-šaqəl FUT- future predictive
la b-šaqəl
4 qam-šaqəl-la PERF- past perfective realis
Ø-šaqəl-wa IRR- -PST past irrealis past habitual realis k-šaqəl-wa REAL- -PST past habitual realis
b-šaqəl-wa FUT- -PST future in the past, habitual realis past
3 *This applies to Patterns II, III and IV (‘stems II, III’ and ‘the quadriliteral stem’ in traditional Semitic terminology). In these forms, which always begin with m, the future prefix b- has been lost after being assimilated to the following m, e.g.: *b-mašxən-Ø (FUT-warm_up-he) ‘he will warm up’ > *m-mašxən > mašxən. This led to the merger of b-šaqəl with šaqəl in these verbal Patterns (Ø-mašxən-Ø IRR-warm_up-he ‘he may warm up’). In order to indicate this morphological ambiguity, all Pattern II, III and IV šaqəl forms and those that may have been underlyingly b-šaqəl are glossed as IRR/FUT.
4 The alternative to šqəlle, used with object suffixes.
16 INTRODUCTION
Table 3: Forms based on šqəlle
Form Gloss Tense Aspect Mood
šqəl-le PERF. past perfective realis
šqəl-wa-le PERF. -PST anterior past perfective realis
Table 4: Forms based on the Infinitive (šqala)
Form Gloss Tense Aspect Mood Resultativity
copula bə-šqala (Infinitive) in-INF.take present continuous (stative only?) realis resultative
Table 5: Forms based on the Imperative (šqul)
Form Gloss Mood
šqul (Infinitive) IMP. irrealis: imperative
la šqul
Table 6: Forms based on the resultative Participle (šqila)
Form Gloss Resultativity
copula + šqila (Participle) RES. resultative, experiential perfect
INTRODUCTION 17
Table 7: Forms of the copula
3MS Form (unless stated Gloss Grammatical semantics
otherwise)
rd wele (3 persons only) COP.DEIX.(he) present contingent state and presentative copula
rd ile (L-copula in 3 persons) COP.PRS.(he) basic present copula (stable/permanent property)
st nd wət (1 /2 persons) COP.PRS.(youMS)
le-we NEG-PRS.COP.(he) negative present copula
we-wa COP.(he)-PST past copula
we-le PST.be(-he) alternative past copula based on the inflection of ‘be’
(*h-w-y)
Table 8: Examples of nominal morphology glosses
FORM Gloss
ḥakom-a, ḥakom-Ø-i (ruler)-MS, ruler-MS-my
ḥakom-θa (ruler)-FS
ḥakom-e, dawəl-yaθa, bat-ane (xxx)-PL
0.6.4. Linguistic methodology: a theory-grounded grammatical
description and study of microvariation
The writing of a grammar requires the constant balancing of abstract linguistic generalisations with the linguistic data available. On the one hand, pre-existing linguistic categories (from theoretical and typological literature, as well as from existing NENA scholarship) are useful for initial working assumptions about 18 INTRODUCTION
meanings and groupings of data. On the other hand, the application of such pre- existing categories can lead to various analytical errors; for instance, features which ‘do not fit’ within those pre-existing categories can be dismissed, though such features may be crucial for the linguistic system in question (Haspelmath
2012; Ozerov 2018, 77–8).
The present study therefore draws as far as possible from ‘bottom-up’, data-driven linguistic approaches. At the same, it also provides the necessary theoretical and
NENA-specific frameworks to ensure analytical rigour, and to demonstrate the implications of the present study for linguistics in general. Linguistic approaches relevant for this thesis are Function Grammar (Dik 1981 and 1997),5 the
‘historicising’ diachronic approach (see below), typology and usage-based corpus linguistics (Bybee and Hopper 2001; Nedjalkov 2001; Givón 1977; Glynn and
Fischer 2010). In line with such bottom-up approaches, the focus of the present study is on language in its functional context. It is assumed that functional units
(conveying various types of ‘meaning’) emerge through the cooperation of lexical and rule-based components of grammar, including word order patterns, semantics and even information structure and discourse pragmatics (Ozerov 2018;
Noorlander 2018, 31–3). ‘External’ factors (e.g. frequency and its change, genre)
5 For the related concept of Constuction Grammar, see Booij (2010). INTRODUCTION 19
also have a direct bearing on meaning and often correlate with its change in the course of time.
On a few occasions, a diachronic dimension is added to the discussion. From a diachronic perspective, a given synchronic system represents just a single stage within an ongoing diachronic process (cf. Bybee, Revere and Pagliuca 1994); in other words, it is part of ‘a bigger (hi)story’. Diachrony, in turn, is often a useful tool for explaining the synchronic system. Moreover, on a few occasions, comparative synchronic data are used to offer a hypotheses on grammaticalisation pathways in NENA, and such hypotheses are compared with grammaticalisation models offered in the general-linguistic literature. Further details of these models are given in the relevant chapters.
PART ONE
PHONOLOGY
CHAPTER ONE
THE SEGMENTS: CONSONANTS AND VOWELS
Introduction
This chapter is a study of the phonology and phonetics of J.Dohok with references to other varieties of NENA, especially the LD cluster. The survey includes the pho- nemic inventory of consonants and vowels, their allophones and phonetic pro- cesses, the most notable of which is emphasis spread (i.e. pharyngealisation spread). This is followed by a discussion on phonotactics and prosody: syllable and word structure, lexical stress, consonant-cluster resolution and general fea- tures of intonation units.
The task of phonology is to establish the existence of discrete, abstract categories such as the phoneme on the basis of empirical phonetic data (Ladd 2011, 348-
350). I adopt here the traditional definition of a phoneme as the basic structural
(though not indivisible) segment in the phonology of a language, employed in the creation of a unit of meaning (e.g. words), and thus with the potential to create a difference in meaning (cf. Barry 2006). The task of phonetics relevant to this chap- ter is to describe the conditioned and therefore predictable processes applying to phonemes. Phonetics therefore also strives to identify the allophones of a pho- neme, that is, its realisational variants (Ladd 2011). In this study, in line with a 24 THE SEGMENTS: CONSONANTS AND VOWELS
linear approach to phonology, features are largely taken to be the property of individual phonological segments.1
The main challenge in this inquiry stems from the dichotomy between phonetics and phonology. A group of phonetically close and even overlapping sounds (i.e.
‘phones’) may on the phonological level represent distinct phonemes (cf.
Ladefoged and Johnson 2011, 87). Phonetic closeness can also result in a situation in which two phones are both allophonic and phonemic, meaning that one of the sounds has a phonemic status, but only a marginal one (Hall 2013). In the present inquiry, this is particularly relevant to emphatic phones, which can be either allo- phones of non-emphatic phonemes or emphatic phonemes (cf. §1.5. below), and for the word-initial glottal stop, which represents either a phoneme or the phono- logical /Ø/. Such near- or complete overlapping of phonemes is also highly perti- nent to vowel phonology, since vowels lie on a realisational continuum (Ladefoged and Johnson 2011, 87).2 One of the goals of this chapter is to make explicit the different phonological status of phonetically close or identical sounds. In the ex- isting NENA scholarship, this distinction often remains unclear for the phones [ə] and [ʾ]. Both sounds can be either phonemic or phonetic, depending on the
1 Cf. the section on emphasis spread (§2.2), which mentions suprasegmental emphasis in the C. and J. Urmi
NENA dialects. For non-linear approaches to phonology, see McCarthy (1982).
2 For NENA, cf. Napiorkowska (2015a, e.g. 19). PHONOLOGY 25
context, as illustrated by the homophone ʾazǝl /azəl/ ‘he may go’ vs ʾazǝl [ʾazǝl]
‘he may weave’.
Cases of sounds with doubtful phonemic status in J. Dohok are resolved with the commonly-used phonemicity diagnostics. If a minimal pair exist, the two sounds carry semantic contrast and are phonemic. On the other hand, if two sounds are in complementary distribution, they are most likely one another’s allophones. In both cases, however, the relationship could still be intermediate.
Consonants
1.2.1. Consonantal phonemes
In Table 9 below, the inventory of the consonantal phonemes of J.Dohok is pre- sented in IPA symbols. If the IPA symbol for a phoneme is different from the sym- bol used in the transcription system, the latter is added in round brackets. Four consonants are loan-phonemes which—to my knowledge—do not occur in the na- tive lexicon. Along with č̣ and ṛ, they are considered marginally phonemic and all of them are given in curly brackets. Voiced phonemes appear under their voiceless counterparts.
26 THE SEGMENTS: CONSONANTS AND VOWELS
Table 9: The consonantal phonemes of J. Dohok
Labial Dental Alveol. Postalv. Velar Uvular Pharyn. Glottal
Plosive p t k q {ʔ} (ʾ)
b d g
ʔ ʔ {p } (p̣) {t } (ṭ)
{bʔ} (ḅ) {dʔ} (ḍ)
Nasal m n
{mʔ}
(ṃ)
Flap ɾ (r)
ɾ̣ (ṛ)
Fricative f θ s {ʃ}(š) x {ħ} (ḥ) h
v ð z {Ʒ} (ž) {ɣ}(ġ) {ʕ} (ʿ)
{ðʔ} (ð̣) {sʔ} (ṣ)
{zʔ}(ẓ)
Affricate tʃ͡ (č)
dʒ͡ (j)
tʃ͡ ʔ (č̣)
* Approximant w l y
lʔ (ḷ)
*[w] has double realisation: voiced labial-velar approximant.
PHONOLOGY 27
1.2.2. Historical background of selected phonemes
1.2.2.1. Dental stops and fricatives /t/, /θ/, /d/ and /ð/
The dental fricatives /θ/ and /ð/ inherited from earlier Aramaic have been pre- served in J. Dohok. Table 10 below compares the realisation of these phonemes with that of other LD dialects, showing the conservative nature of J. Dohok and J.
Betanure:
Table 10: Reflexes of historical Aramaic dental fricatives in LD
Early J. Dohok J. Betanure J. Zakho J. Amediya J. Challa
*θ θ θ s (merged θ s (merged
(
*ð ð ð z (merged d (merged d (merged
(
In Classical Aramaic, these fricatives were the post-vocalic allophones of, respec- tively, /t/ and /d/. Together with /b/, /g/, /k/ and /p/ they belonged within the so-called beged kefet group, which had fricative post-vocalic allophones
(Fassberg 2010, 12–13). See Kapeliuk (1996) on the reflexes of /θ/ and /ð/ in
NENA in general. In J. Dohok, the phonemic distinction between, for instance, 28 THE SEGMENTS: CONSONANTS AND VOWELS
/t/ and /θ/ is demonstrated by the minimal pair ta ‘for (followed by a nomi- nal)’ and θa ‘come (MS)!’
1.2.2.2. The pharyngeals /ḥ/ and /ʿ/
The pharyngeals ḥ and ʿ inherited from earlier Aramaic exist in J.Dohok, but in the native lexicon have been preserved only in certain words containing q or the historical emphatics (i.e. the pharyngealised consonants ṣ or ṭ). This is common in
NENA, including JD (cf. J. Betanure, Mutzafi 2008a, 27-33). The phonology of J.
Zakho has to-date not been studied in this regard; a few details are supplied below.
In C. Barwar, pharyngeal and non-pharyngeal pronunciations sometimes coexist side-by-side (Khan 2008b, 3–56).
This attraction between the pharyngeals and the emphatics and q in NENA has generated a significant amount of scholarly interest (e.g. Polotsky 1961;
Hoberman 1985; Napiorkowska 2015a; Khan 2016). As pointed out by Hoberman, however, emphatics and pharyngeals have some distinct acoustic and articulatory properties. In the J. Dohok corpus, their difference is apparent from the fact that while emphatics cause the lowering of the F2 formant in the adjacent vowels (es- pecially /a/), the pharyngeals trigger no such lowering (see below). The interac- tion between these two groups is thus, according to Hoberman, typologically un- usual and uncommon (1985). This differences notwithstanding, both groups share PHONOLOGY 29
the ‘retracted tongue root’ feature (henceforth [+RTR]).3 This feature is stronger in the case of the pharyngeals (Hoberman 1985, 233). Though it might be typo- logically unique, the emergence of pharyngeals in a pharyngealised environment is demonstrably an areal feature: in a number of west Iranian languages including
Bahdini Kurdish, pharyngeal phonemes appeared in environments which are acoustically similar to pharyngealised environments (Barry 2019).4
Pharyngeals are also preserved in loanwords from Arabic, Kurdish and from pre- modern Hebrew (referred to here with the general term ‘liturgical’),5 for instance:
ḥaxam (Rabbinic (?) Hebrew; cf. Sabar 2002, 166) or ḥ-k-y (maḥke) ‘speak’ (Ara- bic).
1.2.2.2.1. Words with a historical pharyngeal in the vicinity of an emphatic
or q:
In the following words shared by J. Dohok, J. Zakho and J. Betanure, the pharyn- geal has been preserved:
3 In the phonetic literature, this articulatory gesture has various terms, including ‘constricted pharynx’ (cf.
Hoberman 1988, 24).
4 In syllables with bilabials, which are known to cause the F2 lowering, and with the vowel [æ], which is susceptible to F2 lowering (Barry 2019).
5 In Christian NENA, this has a parallel in the conservative pronunciation of key Syriac religious words (cf.
Coghill 2003, 21). 30 THE SEGMENTS: CONSONANTS AND VOWELS
J. Dohok, J. Zakho and J. Betanure forms:
raḥuqa ‘far’
q-ḥ-r (qaḥər) ‘be sad’
ʿamuqa ‘deep’
ṣ-ʿ-r (mṣaʿar) ‘be angry’
ḥoqa ‘ladder’ (J. Zakho huqqa)
1.2.2.2.2. Words with a lost historical pharyngeal in J. Dohok (and references
to other LD dialects)
In other words, however, a historical pharyngeal has been lost in J. Dohok, despite the presence of an emphatic or q. A brief comparison of J. Dohok, J. Zakho and
J.Betanure suggests that it is J. Betanure which exhibits the greatest tendency to retain ḥ and ʿ in the environment of emphatics and q:
J. Dohok gloss J. Zakho J. Betanure6
ṣaḅota ‘finger’ ṣaboʾta ṣaboʾta
x-n-q (xanəq) ‘drown’ x-n-q ḥ-n-q
6 (Mutzafi 2008a, 21–24, 354). PHONOLOGY 31
x-p-q (xapəq) ‘to hug’ x-p-q ḥ-p-q
1.2.2.2.3. Innovated pharyngeals
One can also find words in the LD dialects with innovated pharyngeals, including words without q or an emphatic in the synchronic form.7 For instance, the word for ‘until’ (J. Dohok hil, ḥəl) has an innovated ḥ in most LD dialects. Such cases of fortition can be viewed in parallel to the shift h > x in certain high-frequency words in dialects such as C. Harire-Batase, e.g. heš > xeš ‘still’.8 Other words vary across dialects; for instance, the word for ‘summer heat’, ‘desert wind’ or ‘thirst’ is realised sehya in J. Dohok, and ṣehya in J. Betanure (Mutzafi 2008a, 22), and J.
Zakho and J. Amediya.
1.2.2.2.4. Historical background of pharyngeals in NENA
The presence of the pharyngeals in the synchronic NENA dialects suggests their continued existence in Aramaic (cf. Mutzafi 2008a, 21). The other possibility— that the pharyngeals were reintroduced into the language—appears less likely. An impressionistic observation suggests that in native-Aramaic words, (presumably) the retention of ḥ and ʿ is more common than their innovation, suggesting a history of retention and only occasional innovations. This is despite the widely-held view
7 For more details, see Mutzafi (2008a, 22).
8 Lidia Tuwalska-Napiorkowska, personal communication. 32 THE SEGMENTS: CONSONANTS AND VOWELS
that in Jewish Babylonian Aramaic, pharyngeal weakening was widespread (cf.
Morgenstern 2011, 59-76). This is likely indicative of the fact that NENA goes back to Late Aramaic varieties distinct from JBA. Had a complete reintroduction occurred, it should not have discriminated between the environment of original versus secondary emphatics. The pharyngeals are also preserved in the liturgical
Hebrew of the NENA-speaking Jews and were likely preserved therein even in the period when these sounds were undergoing widespread weakening in Jewish Bab- ylonian Aramaic (Molin 2020), suggesting their continuous presence in the (Jew- ish-) Aramaic speech community.
1.2.2.3. Emphatic phonemes
1.2.2.3.1. Phonetic properties of NENA emphatics
‘Emphatics’ are defined here as phonemes with the coarticulatory feature of velar- isation or pharyngealisation which occur across NENA, with some phonetic varia- tion in their realisation cross-dialectally.9 In a series of consonants, emphasis is a phonemic feature (see discussion below), but because this feature spreads, it re- sults in the phonetic velarisation or pharynrealisation of the neighbouring pho- nemes (cf. §1.5. below). In J. and C. Urmi (north-eastern Iran), ‘emphasis’ is a
9 For longer disucssions, see Hoberman (1988, Khan (2013a, 32-5, 2008b) and Napiorkowska (2015a, 13-
21, 2015b). PHONOLOGY 33
suprasegmental feature, anchored in the larger phonological word (Khan 2016,
50–91, 2013a).
In the Iraqi NENA dialects such as J. Dohok, emphasis is usually said to be realised phonetically through pharyngealisation (Napiorkowska 2015a; Hoberman 1985,
1988; Khan 2013a).10 The articulatory gesture responsible for both velarisation and pharyngealisation is the retraction of the tongue root, henceforth [+RTR] (cf.
Ladefoged and Johnson 2011, 235–6). In pharyngealisation, the root is pulled down towards the pharynx, while in velarisation, it is pulled down and backwards to the velum. This tongue movement also leads to the constriction of the pharynx and the increase of muscle tension (Khan 2016, 50). This tension, in turn, usually results in the unaspiration of the vowels which follow an emphatic. It may also result in their labialisation (cf. Watson 1999, 289),11 because labialisation also in- volves a tensing and a downward movement of the tongue and a constriction of
10 This is most likely due to contact with Arabic. Watson (1999) argues that in Iraqi Arabic, emphasis is re- alised through pharyngealisation. The phonetic accuracy of this study is questionable, however, because
Watson draws on a study from 1963. Since the present work is not an in-depth analysis of articulatory pho- netics, it is assumed that emphasis in J.Dohok is manifested in pharyngealisation. This notwithstanding, in
C. Barwar (North-West Iraq), emphasis is apparently manifested in velarisation (Khan 2008b, 32). Velarisa- tion is also the likely manifestation of suprasegmental emphasis; see Napiorkowska (2015b) and the refer- ences therein. It is likely that in reality, both strategies are combined or interchange in many dialects, as is reportedly the case in C. Diyana-Zariwaw (Napiorkowska 2015a, 50).
11 For emphasis and labialisation in NENA, see also Khan (2013, 4) and Mutzafi (2008, 20). 34 THE SEGMENTS: CONSONANTS AND VOWELS
parts of the vocal tract – in this case, the mouth. Conversely, labialisation can also induce [+RTR]. For these reasons, bilabials in NENA are among the consonants most susceptible to developing emphasis (see the following sections and §2.2). In vowels, [+RTR] results in the backing of the vowel (Khan 2016, 50). Phonological emphasis exists also in Qəltu Arabic (Hassan and Esling) and in Bahdini (Thack- ston 2006), but the emphatic series in those languages are not identical to the
NENA sets.
On the acoustic level, emphasis affects the formants—the resonating frequencies of the vocal tract—causing especially the lowering of the second formant (‘F2’;
Ladefoged and Maddieson 1997, 360–363). Significant F2 lowering, however, is only clearly observable in liquids, due to their high sonority which allows for the clear manifestations of formants, and in the low vowels (/a/, realised [ɑ]).
1.2.2.3.2. The emphatic phonemes in J. Dohok
J.Dohok has ten emphatic phonemes. Of these, only two – ṣ and ṭ – are original
Aramaic emphatics. The four phonemes ḷ, ṃ, ṛ and č̣ are secondary emphatics, which developed through various phonetic and semantic mechanisms. Four pho- nemes – ḅ, ð̣, ẓ, p̣ – are attested only in loanwords, some apparently only in one lexeme. One of the six emphatics which occur in native words, č̣, is only margin- ally phonemic, since a minimal pair could not be identified and its occurrence seems to be phonetically conditioned. PHONOLOGY 35
As discussed below, the most apparent causes for the appearance of emphatics in new places and for the rise of new emphatics are the presence of a historical phar- yngeal (e.g. in ṭṃāne-ʾesar ‘eighteen’) and sound symbolism (e.g. ṭora ‘the Torah’).
In some cases, emphasis also arose in the course of loaning a word (e.g. ḅāš ‘good’) and apparently also by lexical analogy (e.g. ṭṃani ‘eighty’ < ṭṃāne-ʾesar ‘eighteen’
?), showing that the presence or absence of emphasis cannot be predicated on the basis of (historical) phonetic factors. In other cases, the semantic potential of em- phatics is used to create a minimal pair of different meanings from one historical lexeme (e.g. ṃ-ḷ-y ‘fill’ vs m-l-y ‘be enough’ and ṭoṛa ‘the Torah’ vs tora ‘bull’).12 As argued, it is not a coincidence that all of the secondary emphatics except č̣, that is, ḷ, ṃ and ṛ, and the loan phonemes ḅ and p̣ are liquids or bilabials. This is the case across NENA.13 The reason for this susceptibility of the bilabials to emphasis is that the gesture of labialisation is from an articulatory viewpoint similar to the process of increasing muscle tension and a constriction of the vocal tract, which
12 Since these conditioning factors exist, it is not appropriate to call such pharyngealisation processes simply
‘unmotivated’ (as is the case, for instance, in Fassberg (2010, 15).
13 For the emphatic sets in other NENA dialects, see Khan (2013, 12; Iraqi NENA in general), Napiorkowska
(2015, 50; C. Diyana-Zariwaw) or Hoberman (1985; J. Amediya). 36 THE SEGMENTS: CONSONANTS AND VOWELS
accompanies emphatics (see above). Liquids are a favourable locus for emphasis due to their high sonority, and thus perceivable changes in the formants.14
To establish the presence of emphasis in the case of secondary emphatics and in loanwords, I measured the value of F2 for the vowel /a/.15 In words with no /a/, the presence or absence of aspiration in stops was checked. Emphatic phonemes tend to lack aspiration in contrast to their non-emphatic counterparts, though sometimes, aspiration can also occur with the emphatic series (Khan 2013a, 112).
1.2.2.3.3. The historical Aramaic emphatics /ṭ/ and /ṣ/
The historical emphatics ṭ and ṣ occur in words which historically had an emphatic and those which acquired it.
ṭ: minimal pair: tora ‘a bull’ and ṭoṛa ‘Torah, the Hebrew Bible’ (sound sym-
bolism: the prestige of the Bible)16
(ṃṭy) ṃaṭe ‘arrive’ (historical)
14 See Napiorkowska (2015a, 47) and the references therein for the susceptibility to emphasis in liquids and bilabials.
15 For further details, see §2.2 below.
16 Making a phoneme emphatic in NENA is a common device for creating sound-symbolism; for instance, it serves to symbolise greatness, intensity and significance (e.g. Khan 2016, 178–186). PHONOLOGY 37
ṭṃāné-ʾesar ‘eighteen’ (in the environment of historical ʿ: < ṭṃāné-ʿesar
but this does not occur in tmanya ‘eight’) ṭaṃaha ‘over there, far away’ (sound symbolism: physical distance) ṭḷaha ‘three (without a nominal)’ (in the environment of a historical ḥ: < tlaḥa < tlaha < tlaθa ?) ṣ: minimal pair: ṣuḷta ‘dunghill’ sulta ‘snare’ ṣiwe ‘wood’ (historical) 1.2.2.3.4. Secondary emphatics in native words č̣: This phoneme occurs mostly in native Aramaic words, where it developed from the historical Aramaic emphatic phonemes ṣ and ṭ, or from a non- emphatic sound in the vicinity of a historical ʿ (Mutzafi 2008a, 19; Khan 2013a, 112). In J. Dohok, there is only a handful of words with č̣. It is probably not a coincidence that several of these also contain ṃ; this bilabial also anchors emphasis and apparently helps preserve it in the word in ques- tion. While č̣ is restricted to a set of lexemes and can no longer be shown to be a phonetically motivated allophone of /č/, I have not found a secure minimal pair, casting doubts on its phonemic status. The minimal pair given for J. Betanure, bəč̣ʾa ‘for nine’ and bəčʾa ‘bastard’ (Mutzafi 2008a, 19) could be elicited with some speakers, however, in natural speech, speakers use 38 THE SEGMENTS: CONSONANTS AND VOWELS the form bət ʾə̀č̣ʾa ‘for nine’, rather than bəč̣ʾa. Moreover, several of the words with č̣ include a bilabial (ṃ and w), raising the possibility that the true anchor of emphasis in the word is in fact the bilabial. ʾə̀č̣ʾa ‘nine’ (> ʾəč̣ʿa > ʾətša > tēšā) (Mutzafi 2008a, 19) č̣wč̣y (mč̣oč̣e) ‘miaow’ (in J. Betanure ‘chirp’) č̣ṃy (č̣aṃe) ‘extinguish’ č̣yṃ (č̣ayəṃ) ‘shut (eyes)’ ṃ: minimal pair: ṃ-ḷ-y (ṃaḷe) ‘to fill’, m-l-y (male) ‘to be enough’ These two roots, ṃ-ḷ-y and m-l-y are etymologically related. ṃaye ‘water’ ((maʿe <) ṃaye < spontaneous emphasis in m); the J. Sul- eymaniya form maʿe (Khan 2004, 35) suggests that the (historical) emphatic (ṃ) even induced pharyngealisation in some cases) č̣ṃy (č̣aṃe) ‘extinguish’ ṃəḷxa ( ḷ: minimal pair: ṃ-ḷ-y (ṃaḷe) ‘to fill’, m-l-y (male) ‘to be enough’ ṛ: minimal pair: tora ‘a bull’ and ṭoṛa ‘Torah, the Hebrew Bible’ ṛaḅa ‘big, very’ PHONOLOGY 39 The emphasis in these words is due to sound symbolism (for physical great- ness and cultural prestige). 1.2.2.3.5. Loan phonemes The following loan phonemes from Bahdini or Arabic, and minimal pairs could not be established for them. Note that in the case of ḅ, emphasis is found only in the NENA loanword, and not in the Bahdini source. ḅ: only in Bahdini loanwords, e.g. ḅāš ‘good’ < BK baš). No native words have been found in contrast to J.Betanure, where ḅ is said to have arisen in the word ḅaḅa ‘Oh dear!’ (< baba ‘father’; Mutzafi 2008a, 19), apparently through sound symbolism. ð̣: in loanwords from Qəltu Arabic (Mutzafi 2008a, 18), e.g. ḥað̣əṛ ‘ready’ (Ar. ḥaḍər ‘present’; via BK?). Note that the sound in question has shifted from a plosive to fricative in the borrowing process. According to Khan (2013a, 112) Arabic ḍ and ẓ (transmitted via Kurdish) also occur in the Iraqi NENA dialects. In the LD dialects, however Arabic ḍ apparently shifts to ð̣ when borrowed. The same applies to other LD dialects. For instance, the J. Zakho form of ‘ready’ is ḥaẓər, with the characteristic J. Zakho shift (ð > z; here ḍ >ẓ). This apparently is due to the medium of Kurdish, where these sounds are loaned as fricatives (cf. C. Barwar hazər). 40 THE SEGMENTS: CONSONANTS AND VOWELS ẓ: ẓalem ‘to oppress’ (ẓalam; via BK) and ẓararuθa ‘damage’ (Ar. ḍarar ‘to dam- age’; via BK). As these two examples illustrate, ẓ is the reflex of the Arabic ð̣ or ḍ, typically imported via Kurdish (Mutzafi 2008a, 401). p̣: this phoneme is attested in the loanword from C. NENA p̣ap̣a ‘pope’ (in which emphasis apparently arose through sound symbolism). 1.2.2.4. /v/ and /f/ V and f – as in other LD dialects – occur in Hebrew and Kurdish loanwords and nonce-borrowings, e.g. ʾaval ‘but’ ( 1.2.2.5. /q/ This phoneme is often fricativised before high front vowels, e.g. šqila [ʃχiːla]. 1.2.2.6. /ġ/ The sound ġ [ɣ] is phonemic in loanwords from Arabic (for instance, ġēr ‘other, else, different’) and liturgical Hebrew (where it represents the intervocalic allo- phone of gimel, /g/), e.g. maġǝlla ‘scroll, the book of Esther’. It also exists in native words, but only as the voiced (intervocalic) allophone of /x/. In many dialects, this allophone has become phonemicised in some lexemes. In J. Betanure, for in- stance, ġ has been generalised across the paradigms of the verb ġ-z-y (ġaze) ‘see’, from x-z-y (Mutzafi 2008a, 19). PHONOLOGY 41 1.2.2.7. /ʾ/ The sound [ʾ] in J. Dohok—as in many other Iraqi varieties of NENA— is found in contexts which historically had pharyngeal /ʿ/.17 Word-initially, moreover, [ʾ] can also be the reflex of the historical /ʾ/. Synchronically, however, it is best to con- sider only the former [ʾ] (from */ʿ/) phonemic in J. Dohok, while [ʾ] from the historical /ʾ/ is best regarded as a phonetic vowel onset. As shown below, this distinction is based on the fact that phonetic [ʾ] always disappears if its phonetic trigger disappears, and that /ʾ/ does not follow patterns of complementary disti- bution. Moreover, phonetic and phonological [ʾ] are found in different morpho- logical patterns. Consequently, assuming the phoneme /ʾ/ in all cases with [ʾ] does not help correctly predict the morphological shape of words. Similar argumenta- tion is applicable to many other Iraqi NENA varieties, though many grammars still do not make an explicit distinction between phonetic and phonological [ʾ]. Phonetically, syllable-initial phonemic or phonetic [ʾ] in J. Dohok is rarely realised with a full glottis closure, surfacing instead as creaky voice. In intervocalic posi- tions, creaky voice is often replaced simply with a vowel-to-vowel transition. /ʾ/ can occur in all except word-final positions, where it is never realised. In word- medial positions, it tends to occur only in slow, careful speech, being elided 17 Note that ʿ itself comes from two earlier Aramaic phonemes, namely, from /ʿ/ and from intervocalic allo- phone of /g/, that is, [ɣ]. 42 THE SEGMENTS: CONSONANTS AND VOWELS especially commonly in intervocalic positions. Even when elided, however, ʾ is still present on the phonological level, as its absence does not lead to syllable restructuring. For instance, the form k-šamʾ-i REAL-hear-they ‘they hear’ is realised [kʃamʾi] or [kʃami], but never [kʃaːmi]. In other words, the vowel a is always short, indicating that the syllable is phonologically closed. /ʾ/ is present on the underlying level even word-finally. For instance, /k-šāmeʾ/ ‘he hears’ is realised [kšaːme], without the glottal stop, but when a suffix is added, /ʾ/ can be realised; thus /k-šāmeʾ-li/ ‘he hears me’ can be pronounced [kʃaːmeʾli], indicating that /ʾ/ is preserved on the phonological level. Moreover, even when the glottal stop is elided, the vowel preceding is is not lengthened; the form is always realised [kšaːmeli], but never *[kʃaːmeːli]). This indicates that the syllable remains phono- logically closed by /ʾ/. In the synchronic Iraqi dialects, [ʾ] from the historical glottal stop occurs only in the word-initial position. In J. Dohok (and likely in other NENA varieties), this sound is indistinguishable from [ʾ] which comes from the historical /ʿ/, creating homophones such as ʾazǝl ‘that he goes’ (from historical ʾ-z-l) vs ʾazǝl ‘that he weaves’ (from the historical ʿ-z-l) (data from Hoberman 199, 318–19). Indeed, many NENA grammars consider such word-initial [ʾ] phonemic, regardless of its historical origin (e.g. C. Alqosh - Coghill 2003, 21), or at least transcribe both types of ʾ, without indicating whether such transcription is phonetic or PHONOLOGY 43 phonological.18 In these grammars, therefore, the word-initial /ʾ/ (from both /ʾ/ and /ʿ/) seems to be considered the product of phoneme neutralisation: the his- torical /ʾ/ has been lost in all but word-initial positions, where it has merged with /ʾ/ from the historical pharyngeal. This situation notwithstanding, there are reasons to consider [ʾ] from the historical glottal stop simply as a phonetic vowel. In J. Dohok, for instance, though the word- initial [ʾ] has identical pronunciation regardless of its historical background, when followed by a vowel, it shows a distinct behaviour when followed by a consonant or when a prefix is added. This is illustrated by the Infinitive and the b-šaqəl forms in Table 11 below (this realisation represents J. Dohok and the other LD dialects): Table 11: Differences in verbal stems illustrating different status of [ʾ] from /ʾ/ and /ʿ/ Verbal root Gloss šaqəl Infinitive b-šaqəl b-axəl [baxəl] Ø-x-l ‘eat’ ʾaxəl [ʾaːxəl] ʾixala [ʾixaːla] (never [bʾaːxəl]) (< *ʾ-x-l) ʾ-r-q ‘flee, ʾarəq [ʾaːrəq] ʾăraqa [ʾaraːqa] b-ʾarəq [bʾaːrəq] (< *ʿ-r-q) run’ ~ [baːrəq] 18 Cf. the grammars of J. Betanure (Mutzafi 2008a, 20); J. Challa (Fassberg 2010), C. Barwar (Khan 2008b, 57) and J. Amediya (Greenblatt 2008). 44 THE SEGMENTS: CONSONANTS AND VOWELS As this table shows, [ʾ] from the historical /ʾ/ is never realised before the future preverb b-. In the Infinitive, moreover, the epenthetic vowel after [ʾ] is different, depending on the phone’s historical background. If [ʾ] in both verbs given above represented a single phoneme, we would expect this sound to behave consistently, and not idiosyncratically, depending on the lexeme. A similar discrepancy be- tween [ʾ] from the historical /ʾ/ and /ʿ/ is found in Iraqi dialect beyond LD, sug- gesting that these two historical sounds have not merged into a single phoneme. In C. Alqosh, for instance, the infinitive of ‘say’ is ʾimara or mara, but ‘run’ has the form ʾraqa (Coghill 2003, 144). The present analysis is thus partly parallel to Khan’s approach to C. Urmi, where ʾ is considered ‘a non-etymological phonetic addition’. In C. Urmi, however, ʾ is only found word-initially – it does not occur word-medially even when from the historical voiced pharyngeal fricative (Khan 2016, 173). This analysis, moreover, implies that in NENA – unlike in most pre-modern Semitic languages – the syllable need not start with a phonological consonant. In the present study, word-initial [ʾ] is always indicated in the transcription, in line with the general policy to reflect a certain degree of phonetic detail in the transcription. A unique case in J. Dohok is the šaqəl and b-šaqəl form of ‘want’, from the histor- ical NENA root bʾy. The forms are, respectively, ʾəbe and b-ʾəbe. This glottal stop does not come from the historical /ʿ/, but is demonstrably a phonological conso- nant: it can be preserved after the preverb b-. However, its Infinitive form ʾəbaya PHONOLOGY 45 has a different epenthetic vowel than ʾ from the historical /ʿ/, e.g. ʾăraqa (the latter is lowered due to the presence of the historical pharyngeal). Thus, the par- adigm of the verb ‘want’ is best regarded as based on a distinct morphology (cf. §5.6.1). Vowels 1.3.1. Introduction The phonology of vowels has its own particular challenges: belonging to a single realisational continuum, vowels are notoriously unstable (cf. Ladefoged and Johnson 2011, 87). This instability frequently leads to phonemic overlapping and near-overlapping, creating a challenge for distinguishing between phonemes and allophones. This can lead to intermediate phonological relationships between vowels. 1.3.2. Vowel phonemes J. Dohok possesses the vowel phonemes /e/, /i/, /o/ and /u/, which are phono- logically unspecified for quantity. These are long in open, accented syllables and otherwise short. In /e/, /i/ and sometimes /o/, the short realisation is accompa- nied by a qualitative change (see below), causing the vowel to overlap with the allophone of another phoneme. In the case of a, there are grounds to postulate two underlying phonemes, /a/ (unspecified for length) and /ă/. However, /ă/ is 46 THE SEGMENTS: CONSONANTS AND VOWELS only marginally phonemic, because in the vast majority of cases, a (i.e. /a/) fol- lows the rules of complementary distribution, and because there are only a few (certain) minimal pairs involving [aː]/[a]. The vowel [ə] is marginally phonemic and in some cases remains short even in open stressed syllables. In most cases, however, it functions as a short allophone of /i/ and /e/.19 The approach adopted here is therefore that vowel length in J. Dohok and in other LD dialects is by-and-large predictable and thus phonetic, rather than phonological. By contrast, the existing LD grammars adopt a more maximalist ap- proach which assumes a phonemic opposition of /ă/ and /ā/, /ŭ/ and /ū/, along- side a series of underlyingly long vowels /ē/, /ī/ and /ō/ (e.g. J. Amediya - Hoberman 1997; J. Betanure - Mutzafi 2008a, 25-7; J. Challa - Fassberg 2010, 18). Presumably, this approach is influenced by the situation in earlier Aramaic, where vowel length was phonemic.20 As will be shown, the distribution of long and short vowels in LD cannot always be predicted by phonetic rules only; in a few cases, morpho-phonetic specification is necessary. In such cases, the occurrence of a sound is conditioned by phonetic 19 For a comparable analysis of vowels of an Iraqi NENA variety, see the work on C. Barwar by Khan (2008b, 63–94). 20 For the historical vowel phonology in Eastern Aramaic (Syriac), see Hoberman (1992) and Knudsen (2015). For the background of vowels in a NENA dialect (C. Diyana-Zariwawa), see Napiorkowska (2015a, 34–40). PHONOLOGY 47 parameters such as syllable type, but the sound in question only appears in such environments in specific morphological cases. This is clear especially in the case of /a/: it can be realised long in open syllables outside the stressed position, but only in specific verbal forms. Such morphologically-sensitive complementary dis- tribution is referred to here as derived contrast (Hall 2013, 235). Furthermore, the existence of [ă] in environments where other vowels are not realised short is ex- plained as an incomplete merger of the historical /ă/ and /ā/. With other vowels as well as a, length outside the primary stressed syllable occurs in some cases due to a synchronic, phonetic process. This lengthening process, however, can be explained phonetically – with recourse to lexical stress shifts. On the whole, therefore, vowel length is largely phonetically predictable in the native Aramaic stock, while in some cases, it is predictable morpho-phonetically. This licenses the adoption of an economic phonemic system in which contrast in length (with the exception of /ă/) is phonetic and—in a few cases—derived. The synchronic system of phonetically predictable length in NENA contrasts with earlier Aramaic. In the LD group, this change has resulted in the lengthening of stressed vowels. For instance, the loss of morphological gemination in verbal and adjectival forms has led to syllable opening. In C. Barwar, by contrast, gemination is very occasionally retained (Khan 2008b, 98–100).In stressed positions, such vowels were subsequently lengthened, for instance: *mšaddər [mʃaddər] > mšadər [mʃadər] > [mʃaːdər] ‘(that) he sends’. Outside the stressed position, the vowel 48 THE SEGMENTS: CONSONANTS AND VOWELS remained short despite the syllable opening, e.g.: *bassīmā [bassiːmaː] > basima [basiːma] (cf. C. Barwar - Khan 2008b, 194–7). The quantitative values of vowels in this chapter are based on measurements in Praat. Tokens for vowel quantity measurement were taken from a single piece of narrative by a specific speaker to ensure their comparability. Unless stated other- wise, the position of the syllable relatively to the entire intonation unit was not considered, though syllables in pause (i.e. at the end of an intonation units) were excluded. Finally, the findings below should be considered preliminary, since sta- tistical data would be necessary to eliminate the effects of coarticulation, speakers’ idiolects etc. 1.3.3. Phonetic properties of phonemes and their allophones Vowels are long (96-120 ms) only in open, accented syllables, while being short or mid-long elsewhere. As Table 12 below demonstrates, short values are charac- teristic of closed syllables, while mid-long phonemes appear in open syllables, of which word-final vowels are slightly longer than their non-word-final counter- parts. PHONOLOGY 49 Table 12: Vowel length in ms. in J. Dohok per syllable type open open open closed accented unaccented unaccented word-final attested ranges 95-120 60-80 48-78 45-59 mean average 112 73 63 55 Exceptions to this general rule which apply in a particular morpho-phonetic envi- ronment and which—to use the language of Optimality Theory—outrank (over- ride) the general rule of length are discussed below. Such changes in vowel length are usually due to the shift of lexical stress in morphologically complex verbal forms (see §2.3.2.3). Here, some issues of length of /a/ are mentioned in order to address the issue of a single or two underlying a phonemes, which has arisen in the literature. 50 THE SEGMENTS: CONSONANTS AND VOWELS 1.3.3.1. /a/ and /ă/ /a/ and /ă/ have the quality [a], or [ɑ] in emphatic environments. There are a few minimal pairs (see below), suggesting the (marginally) phonemic status of /ă/,21 but some are uncertain. măre- ‘owner/lord of’ mare ‘his owner/lord’ măru ‘say (PL)!’ maru ‘their owner/lord’ ʾăm(ə)r- ‘wool of’ ʾamər ‘that he say’ The underlying form of ămər is /am.rət/, meaning that the shortness of ă is due to the closure of the underlying syllable.22 However, since the word tends to be real- ised with the epenthetic [ə], the phonetic form of the word does have a minimal pair. Măru occurs only in fast speech, in careful speech, the form is mărun. Măre— as argued below—is likely a historical construct form, meaning that it is likely to be prototypically unstressed (cf. §2.3.3.1). This means that măre and mare are dis- tinguished by two features: lack of stress and length. Words with ă in open stressed 21 Hoberman posits the same phonemic opposition (1997, 322) for J. Amediya, but—despite affirming the existence of minimal pairs—does not give any examples thereof. The minimal pair given for J. Challa, bale ‘his mind/attention’ and băle ‘yes, indeed’, contains a loanword băle; so the situation is not necessarily rep- resentative of the native phonology. 22 Cf. similar neo-construct forms: dĕʾǝθ bəṭme [deʾəθ bətˤme] ‘terebinth resin’, from deʾθa ‘sweat, resin’. PHONOLOGY 51 syllables, where shortness cannot be attributed to complementary distribution, in- clude măni ‘who’. Long a outside the primary-stressed syllables also occurs in specific morpho-pho- netic environments in verbs in NENA: in šaqəl-based forms which contain the ob- ject-marking L-suffix or the past converter wa. Both morphemes are termed here semi-clitics due to their mixed clitic- and affix-like properties, cf. §2.3.2.3). The addition of these semi-clitics often causes stress shifts,23 but the a vowel which had previously carried lexical stress remains long, as illustrated by the following J. Dohok forms: k-páθǝx [ˈkpaːθəx] ‘he opens’ but: k-pāθə́x-wa [kpaːˈθəxwa] ‘he used to open’ qam-pāθə́x-la [qampaːˈθəxla] ‘he opened it’ While this length is also attested with other vowels (e and o), it is consistent only with /a/ (cf. §2.3.2.3). Historically, the vowel was long in all such cases, which is apparent from the fact that the vowels e and o in such contexts commonly origi- nate in diphthongs, e.g. g-meθé-wa [gme(ː)θéːwa] < *g-mēθéwa < *g-mayθé-wa ‘he used to bring’. Apparently, therefore, the vowel a ‘lags behind’ other vowels in the process of their shortening outside the stressed position. This, in turn, is most 23 For the rules which regulate the placement of lexical stress in such forms, see §2.3.2.3. 52 THE SEGMENTS: CONSONANTS AND VOWELS likely due to the ‘inherent length’ of a (Lehiste 1970, 18-9). Low vowels are ap- parently ‘inherently long’ for articulatory reasons: a longer time is required for the production of low vowels than of high ones. Lehiste shows that the long allophone of a is longer than that of a higher vowel, and its short allophone longer than the short allophone of another vowel. As demonstrated by NENA, it is thus apparently simpler to shorten higher vowels than to shorten a, leading to a more consistent preservation of a outside stress than e or o in the NENA dialects. The distribution of a is thus predictable, but through phonetic and morphological specification, so length is indicated in the transcription. In other cases, a is lengthened in penultimate syllables which cannot bear stress, and the same process applies to other vowels and is likely a synchronic phonotac- tic process (cf. §2.3.2.3). In the transcription, this type of lengthening is not indi- cated, because it is phonetically predictable. 1.3.3.2. /e/ The phonemic distinctness of /e/ from /i/ is demonstrated by the minimal pair ʾeða ‘holiday, feast’ and ʾiða ‘hand, arm’. The qualitative difference between the two, however, is typically neutralised in closed unstressed syllables. /e/ has the quality [e] in open, accented syllables (where it is long) as well as in open, unaccented word-final syllables (where it is short), e.g. k-šate ‘he drinks’. In closed syllables of the native Aramaic stock, the quality shifts to [ǝ] (short), for instance, g-emǝr ‘he says’ > g-ǝmra ‘she says’. In some loanwords, however, the PHONOLOGY 53 quality [e] is preserved, even when the vowel shortens; for instance, the Arabic ġēr [ɣeːr] is often realised ġer [ɣer]. 1.3.3.3. /i/ /i/ has the quality [i] in open, accented syllables (where it is long) as well as in open, unaccented word-final syllables (where it is short). In closed syllables, as with /e/, the quality shifts to [ǝ], for instance, pθixa ‘open (ms)’ > pθǝxta ‘open (fs)’. An exception are syllables closed with h or ʾ, in which the vowel has shifted to [e] (cf. Mutzafi 2008a), as illustrated by šmeʾle ‘he heard’, from the historical *šmiʾle (which contrasts with ǝ in strong forms, e.g. šqǝlle ‘he took’). 1.3.3.4. /o/ /o/ is phonemically distinct from u, as demonstrated by the minimal pair ṭora ‘Torah, the Hebrew Bible’ and ṭura ‘mountain’. The qualitative difference between the two, however, is sometimes neutralised in closed unstressed syllables. /o/ has the quality [o] or slightly higher in open accented syllables (where it is long) and in open unaccented syllables (where it is short). In closed syllables, the vowel tends to be slightly centralised ([ɤ]), sometimes also raised to [ʊ]. Such raising brings about a neutralisation with the short allophone of /u/; for instance: smoqa [smoːqa] ‘red (MS)’ > smoqta [smʊqta] ‘red (FS)’. As the values of F1 in Table 13 demonstrate, this raising is not consistent, meaning that [o] (short) is 54 THE SEGMENTS: CONSONANTS AND VOWELS often lower than [u]. The increase in the value of F1 corresponds to vowel lower- ing. Table 13: F1 values in Hz of [uː], [u] and [o] (short) Lexical token F1 ʿamuqa [ʿamuːqa] 519 ʿamuqta [ʿamuqta] 590 b-yawən-nox [byaːwənnɤ/ox] 668 The fluid realisation of short /o/ is also reported for C. Barwar (Khan 2008b, 80- 1). According to Mutzafi, in J. Betanure short /o/ is realised [u] (Mutzafi 2008a, 26), but in reality, there is also likely fluidity; indeed, Mutzafi consistently tran- scribes the 2MS L-suffix lox, not lux (cf. Mutzafi 2008a, 49). In the transcription of J. Dohok in this study, /o/ in closed syllables is always transcribed ‘o’ for the sake of consistency, though in reality, the realisations vary between [o], [ʊ] and [ɤ]. In syllables closed with h or ʾ, the quality [o] is always preserved (though the vowel is short; thus, č̣oʾa [čˤoːʾa] ‘smooth (MS)’ > č̣oʾŏta [čˤoʾota] ‘smooth (FS)’. PHONOLOGY 55 1.3.3.5. /u/ The vowel space of /u/ extends between [u] and [ʊ]. /u/ is long in open, accented syllables. In the environment of pharyngealised consonants, the quality tends to be lowered to [o]. In other types of syllables, the vowel is short and partly cen- tralised ([ʊ]), for instance [ʔamuːqa] ‘deep (MS)’ vs [ʔamʊqta] ‘deep (FS)’. Some speakers tend to centralise and delabialize short u, realising it close to [ǝ], for instance mušxǝnne [mǝšxənne] ‘he warmed up’. No minimal pairs could be found for J. Dohok involving [u] and [uː]. In his J. Betanure grammar, Mutzafi lists the pair xūr ‘friend of’ and xur ‘like’ to argue for two different u phonemes (Mutzafi 2008a, 28). In J. Dohok, however, both forms are short and no other certain minimal pairs were identified. Indeed, as Mutzafi also admits, [u] and [uː] are in complementary distribution in the majority of cases.24 Cases of long u in post-tonic open syllables in J. Dohok are discussed in §2.3.2.3 below; this lengthening is characteristic of penultimate syllables which cannot bear stress and occurs with other vowels as well. 24 No minimal pairs are provided for J. Amediya (Hoberman 1997; Greenblatt 2008, 20), though two pho- nemes, /ū/ and /u/, are also posited by Hoberman. 56 THE SEGMENTS: CONSONANTS AND VOWELS 1.3.3.6. /ǝ/ as allophone and phoneme As shown above, [ǝ] is in the majority of cases the short allophone of /i/ and /e/ occurring in closed syllables, for instance k-eθe ‘he comes’ > k-əθya ‘she comes’. [ǝ] also functions as a prosthetic or an epenthetic vowel (phonologically Ø) and leaves lexical stress and syllable structure unaffected. This occurs, for instance, in Pattern III šaqəl-based and imperative forms, e.g. mášxənu ‘warm up (PL)!’ The same situation obtains in other LD dialects (e.g. J. Betanure - Mutzafi 2008a, 46). Here, ə does not attract stress indicating that the underlying form is /mášxnu/. In all of these cases, therefore, [ə] is not phonemic, but is transcribed for the sake of realisational clarity.25 In other cases, [ə] is phonemic. Its phonemic status and distinctiveness from e is demonstrated by the minimal pair gəba ‘she wants’ vs geba ‘towards’. ə is also pho- nemic in the CəCCa nominal pattern, in which it cannot be shown to be synchron- ically in complementary distribution. This pattern is exemplified by words such as 25 This distinction is not always explicit in the NENA scholarship. Mutzafi lists /ǝ/ as a phoneme in J. Betanure, despite showing that it occurs in complementary distribution with the vowel phones e and i and as ‘an anaptyctic vowel’ (2008a, 32). PHONOLOGY 57 šəmma ‘name’ and šənne ‘years’, which developed respectively from *šmā and *šnē (Syriac).26 A noteworthy case of phonemic [ə] is found in the šaqəl-based paradigms of the verb ‘want’ (< *ʾ-b-y), e.g. gə́be ‘he wants’ (cf. also §5.6.1). Here, [ə] is synchron- ically not predictable, since a long vowel is expected. In fact, /ə/ in gəbe violates the rule of bimoraity of stressed syllables (CVC or CVV), which also indicates that in contrast to other vowels, it is underlyingly short (/ə̆/). Moreover, unlike pho- netic [ə], ə in gəbe attracts stress.27 [ə] in an open stressed syllable also occurs in forms with the 2PL E-suffix when followed by an L-suffix or wa, e.g. garšə́tū-le [garˈʃətuːle] ‘that you (PL) pull him’. In forms without the semi-clitics, this vowel is realised [eː], for example garšétun [garˈʃeːtun] ‘that you (PL) pull’. 26 Cf. Hoberman (2006), according to whom this development was motivated by the prosodic requirement of disyllabicity. 27 This may be compared with C. Harbole, where we find forms such as mašxína ‘she warms up’ (from the historical < *mášxǝna < *mášxna, cf. J. Dohok). In this form, the historical ǝ has also been phonemicised (and can therefore bear stress). Unlike in gəbe in J. Dohok, however, this vowel has also merged with /i/, apparently following the reinterpretation of the epenthetic [ə] as the short allophone of /i/. 58 THE SEGMENTS: CONSONANTS AND VOWELS 1.3.4. Diphthongs and their reflexes 1.3.4.1. *ay > e The diphthong *ay has contracted to e, as illustrated by the words below. This contracted e has merged with the /ē/ from earlier Aramaic. beθa ‘house’ < *beta < bayta ʾena ‘eye’ < *ʾayna In J. Amediya, the ay diphthong is sometimes still preserved e.g. ḥaywan (< Ar. ḥaywān) ‘wild animal’ (Greenblatt 2008, 20; J. Dohok ḥewan). 1.3.4.2. *aw > o The diphthong *aw contracts to o, as illustrated by the words below. This contrac- tion has brought about the merger of o with the /ō/ of earlier Aramaic. hoya < * haw.ya ‘that she is’ (contrasting with Ø-ha.we-Ø IRR-be-he) moθa < *mawθa ‘death’ The dialects of J. Betanure (Mutzafi 2008a) and J. Challa (Fassberg 2010) have also contracted this diphthong, while in J. Amediya, aw is sometimes preserved (Greenblatt 2008, 20), as is the case in C. Barwar, where aw may coexist side-by- side with o (Khan 2008b, 82–83). PHONOLOGY 59 1.3.4.3. *iy >i, *əw > u and *uw > u The diphthongs *iy, *əw and *uw have been simplified in J. Dohok, respectively to i, u and u. Interestingly, synchronic forms containing the reflexes of *iy and *əw indicate that these contractions did not occur at the same time: kθuta < *kθəwta ‘written (FS)’ mliθa < *mlita < mliyta ‘filled (FS)’ These data indicate that the diphthong əw collapsed later than iy. The latter shifted to i when θ was still an allophone of /t/, while the former diphthong contracted when the shift t > θ was no longer phonetically conditioned, and consequently did not apply after the emergence of u. CHAPTER TWO PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.1. Phonetically-conditioned processes affecting consonants 2.1.1. Devoicing Word-final obstruents are often devoiced or partially devoiced, e.g. ʾawǝð [ˀaːwǝθ] ‘he may do.’ 2.1.2. Aspiration Voiceless non-emphatic stops are generally aspirated before vowels and before [ʾ] and [h].1 Notable exceptions which are not predictable phonetically include the b- = 2 šaqəl paradigm of ‘be’, i.e. p̮awe [p aːwe] (see §2.3.6.6.). 1 The existing LD grammars do not deal with aspiration, but the situation is likely comparable to J. Dohok. 2 In the subordinating conjunction t-/d- before the independent form of the basic copula, aspira- tion does seem to be present (e.g. t=ile), though it is reportedly unaspirated in C. Barwar (Khan 2008b, 112). 62 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.1.3. Place assimilation The interdentals θ and ð assimilate in place to the following l if this l is part of the verbal L-suffix (for L-suffixes, see §5.2). In that process, they also shift to stops (respectively, [t] and [d]), e.g. g-ewǝð-Ø REAL-do-he ‘he does’ vs wǝd-le PERF.do-L- SUF.he ‘he did’ (cf. Mutzafi 2008a). 2.1.4. Complete phoneme assimilation of the l in L-suffixes As is commonly the case in NENA, the l of the L-suffix which occurs in verbs (and carries various functions) assimilates completely to the preceding consonant if preceded by an alveolar sonorant, that is, n or r. It also assimilates to the preceding alveolar t, but only if t not a part of the root (Mutzafi 2008a), for instance, ʾamr- ǝt-ti IRR-say-youMS-me (Ø-m-r) ‘you may tell me’, but qam-bahǝt-li PERF-surprise-he- me (b-h-t) ‘he surprised me’. In the former case, however, the gemination of the sonorant (i.e. [rr] or [nn]) tends to be weakened and lost in the LD dialects, for instance mšodər-re [mʃodəre] PERF.send-L-SUF.he ‘he sent’. 2.1.5. Sonorant interchange In the LD dialects, one finds occasional interchanges between the sonorants l, n and r, as shown in Table 14. While some are spontaneous (e.g. are attested only a handful of times in the speech of a single informant), others occur consistently in certain words, indicating that the new sonorant should be analysed as a PHONOLOGY 63 phonological segment of the word in question. This phenomenon is not discussed in the existing LD grammars. Parallel sonorant interchanges and shifts may also occur in Bahdini, though this has not been investigated. The shift ḷ > ṛ is attested in some dialects of Sorani Kurdish. This has apparently influenced the NENA dialect of J. Sulemaniya (Sorani area), we find the shift l > r occurs in the speech of some informants (Khan 2004, 26–7). 64 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS Table 14: Sonorant interchanges in LD Shift J. Dohok Gloss Original form Forms in form other LD varieties l > r pantarṓn ‘trousers’ bantalon (Ar.) pantaron, pantalon l > r kastar ‘Mount Castel’ Kastel (MH) (near Jerusalem) r > n beheruθa, be- ‘light’ beheruθa behruθa (J. henuta Betanure) r > n băθər băθən ‘after’ baθr l > r (J. dolamant ‘a rich man’ dawlamant (NENA; dolamant (J. Amediya) ‘rich man’) < dawla Nerwa) (Ar.; ‘state, wealth’) dawramant (J. Amediya) (Hoberman 1997, 324) PHONOLOGY 65 2.2. Emphasis spread (affecting consonants and vowels) The term ‘emphasis spread’ refers to the acquisition of [+RTR] by phonemes, re- alised through pharyngealisation or velarisation. [+RTR]. 3 Khan (2013) divides the NENA dialects into three distinct typological groups with regard to emphasis, though dialects such as C. Diyana-Zariwaw show a mixed typology (Napiorkowska 2015a; cf. below). The first group is characterised by ‘emphasis spread’, that is, a diffusion of [+RTR] from an emphatic phoneme to its wider environment. This typology is found in the Iraqi and South-Eastern Turkish dialects and corresponds broadly to the situation reported for Iraqi Arabic (Hoberman 1989; Watson 1999). The second group is represented by the western Iranian dialects – south of Lake Urmia. As in the first group, emphasis spread from emphatic phonemes to the neighbouring sounds, anchoring itself in the pharyngeal phonemes and leading to their preservation in that environment.4 Unlike in the previous group, the em- phatic phonemes have now been largely lost, leaving only the pharyngeals as a relic. In the final group which subsumes north-western Iran, emphasis is not a 3 From a more general phonetic point of view, it should be viewed as a type of assimilation (in [+RTR]), and specifically ‘long distance consonant agreement’ (cf. Khan 2013a, 4). 4 This was probably due to the fact that both classes share the [+RTR] feature, though in the case of phar- yngeals, the tongue is retracted to a much lower place (see Hoberman 1985, 223). 66 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS phonetic feature of assimilation between individual segments, but rather a supra- segmental, phonological feature of entire lexemes.5 The LD group belongs to the first group of dialects where [+RTR] spreads linearly, but there is no systematic, measurement-based study thereon in the existing LD grammars (cf. Fassberg 2010; Greenblatt 2008, 38–40; Mutzafi 2008a, 20).6 The present study comes to fill this gap. It also contributes to the understanding of NENA emphasis in general, as most current studies do not include measured, quan- tifiable data of the formants.7 The J. Dohok data indicate that emphasis is stronger in sonorants than in obstruents, but a given emphatic has a comparable effect on the vowels before it and after. Two- or three-syllabic words often show strong pharyngealisation in all segments susceptible to it (especially in the sonorants l, m and r and in a), or even throughout the word if suitable phonetic conditions arise. In such words, therefore, emphasis is best considered synchronically anchored in all the phonemes susceptible to it. It is suggested that such words were pivotal in the development of suprasegmental pharyngealisation in C. and J. Urmi. 5 In NENA studies, this is called synharmonism, suprasegmental or autosegmental emphasis (or pharyngeal- isation). For C. Urmi, see Khan (2013; 2016, 50-90) and the references therein. For J. Urmi, see Khan (2008a, 13-20). 6 There are a few references to emphasis spread in J. Zakho in Khan (2013). PHONOLOGY 67 This investigation relies on the values of F2 (and in some cases F1) obtained through Praat-assisted measurements of phonological segments, excluding the very onset and offset to eliminate coarticulation effects. Lexical tokens were se- lected from a single speaker, so that the frequencies of phonemes could be com- pared relatively to each other. For vowels, I mostly measured the values of F2 for /a/ in open, ideally accented syllables and in the emphatic bilabials and liquids for reasons of high affectedness by emphasis (see §1.2.2.3. above). Emphatic pho- nemes occurring only in loanwords except ḅ were excluded due to their infre- quency.8 The findings below, while empirically grounded, remain preliminary, as they do not proceed from statistical data. As explained at the end of this section, the convention is to transcribe all consonants which can receive phonological em- phasis as emphatic in words which have emphasis, as long as those phonemes belong within the same morpheme and are not divided by a high segment. 8 There are a few references to emphasis spread in J. Zakho in Khan (2013a). 8 Khan 2013a, 114). This notwithstanding, q displays some properties of an emphatic and is therefore con- sidered a part of the emphatic class in some studies of NENA or of other Semitic languages (e.g. Khan 2013a, 2). In some Arabic dialects, by contrast, [q] is the source of emphasis spread (Watson 2007, 273). 68 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.2.1. Degrees of emphasis caused by the emphatic phonemes Table 15 and the plot in Figure 1 below show F2 values of a in non-emphatic and emphatic environments. Segments in non-emphatic environments are henceforth called ‘plain’. Figure 1: The distribution of vowels in plain vs emphatic environments PHONOLOGY 69 Table 15: F2 values in Hz of /a/ in plain and emphatic environment Lexical token F2 Plain environment huðaya 1569 ʾĭxala 1390 ʾaθe 1407 xaθa 1455 Environment of /ṭ/ ṭala 1333 ṭalu 1261 Environment of /ṣ/ ṭṛoṣa 1336 ṃṣaḷe 1264 Environment of secondary emphatics and /ḅ/ g-mač̣ṃe 1390 ṛaḅa 1295 ḅāš 1239 gṃaḷe 1191 ṭḷa 1112 ṃaṭo 1050 As this table demonstrates, the historical emphatics ṭ and ṣ as well as č̣ and ḅ cause a smaller degree of F2 lowering within their environment than the emphatic son- orants ḷ and ṃ. The F2 of a in the environment of emphatic obstruents is always 70 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS above 1200 Fz, sometimes reaching over 1300 Hz. In the vicinity of sonorants, by contrast, a is always below 1200 Hz, with the exception of ṛ. This suggests that [+RTR] is likely stronger in sonorants than obstruents, as the phonetic properties of the former make them highly susceptible to developing emphasis (see above). A parallel phenomenon has been noted in C. Diyana-Zariwaw (north-western Iraq, close to western Iran), where emphasis in the historical emphatics ṭ and ṣ some- times leaves no effect altogether in the adjacent vowels (Napiorkowska 2015a, 50–1). The data suggest that in J. Dohok, however, the difference in the degree of emphasis is likely due to the class of the consonants (obstruents vs sonorants), rather than the emphatics, diachronic origin (original vs secondary). By contrast, the measurements demonstrate no striking difference in the directionality of em- phasis spread. 2.2.2. Directionality of spread As shown in Table 16, there is no striking difference between the degree of low- ering of /a/, depending on its position before or after an emphatic: PHONOLOGY 71 Table 16: Values of /a/ in Hz before and after an emphatic Before ṭ After ṭ Token F1 F2 Token F1 F1 qaṭe 740 1230 ṭalu 577 1261 b-qaṭe 634 1267 ṭale 622 1333 2.2.3. Domain of emphasis spread Table 17 below presents the values of F2 for the liquids l and r, the bilabial m and the vowel a in words containing an emphatic, comparing them to the attested F2 values in words with no emphatics.9 Only one segment in each word is transcribed as emphatic – typically the phoneme which was emphatic in earlier Aramaic, or the one considered emphatic in existing NENA grammars. As the data show, how- ever, often several consonants are clearly emphatic in the synchronic system, 9 Segments with lowered F2 values (relatively to plain environments) are underlined. The num- ber in brackets indicates the distance in syllables from the emphatic phoneme. 72 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS Table 17: F2 values of phonemes showing effects of emphatics on other phonemes within the same word /a/ /m/ /l/ /r/ Plain (approx- 1410-1570 1390-1480 1800- 1550-1750 imated) 2000 maṭe 1163 (1) 1235 (1) maṭo 1030 (1) ṭala ṭala 1333 (0) 1448 (1) ṭala 1384 (1) troṣa 1307 (0) 970 (1) ṣloθa 1337 (1) mṣalax mṣalax 1230 (0) mṣalax 1330 (1) mṣalya mṣalya 1355 (0) mṣalya 1362 (1) PHONOLOGY 73 Table 17: Effects of emphatics on other phonemes within the same word (continued) /a/ /m/ /l/ /r/ mṣalyat mṣalyat 1433 (0) mṣalyat 1760 (1) gmaḷewalu gmaḷewalu 1261 (1) gmaḷewalu 1530 (1) ṭamaha ṭaṃaha 1310 (0) 1072 (1) ṭaṃaha 1114 (1) ṭaṃaha 1002 (2) which provides grounds for also considering them synchronically phonologically emphatic. This table demonstrates that the F2 lowering in /a/ and in the sonorants m, l and r is commonly observable in the syllables adjacent to the syllable with an emphatic phoneme. Though it is challenging to find longer words that would show a wider extent of the spread, words such as ṭamaha clearly illustrate that entirely 74 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS pharyngealised words do occur.10 This, in turn, suggests that in many words, em- phasis is likely synchronically anchored (i.e. phonologised) in all phonemes sus- ceptible to it, rather than in a single emphatic phoneme. It is often impossible to show the primary anchor of emphasis, because strong emphasis is discernible in several segments. This, in turn, leads to short words such as ṭḷa and maḷe being pronounced with [+RTR] throughout. At the same time, as is usually the case in the Iraqi NENA dialects, the high front vowels e or i in J. Dohok block the spread of emphasis (Khan 2013, 3). For instance, the F2 of a in the syllable wa of g- maḷewalu is clearly plain (1530 Hz) and the bilabial w in this word does not anchor emphasis either. On the other hand, y does not always block emphasis spread, as illustrated by mṣalya, realised with [+RTR] throughout: [mˤsˤɑˤ̠ lˤyɑ].̠ The tendency to anchor pharyngealisation in several segments susceptible to it is reflected in my transcription policy. In words containing emphasis, all of the con- sonants from the series t/ṭ, s/ṣ, č/č̣ m/ṃ, l/ḷ, b/ḅ, p/p̣ and r/ṛ are transcribed as emphatic as long as they belong to the same morpheme and are not separated by e or i. This can be illustrated with the following examples: ṭṛoṣa, ‘truth’ (from the 10 See Khan (2013, 113) and the references therein for claims that emphasis spread tends to be ‘restricted to adjacent syllables’ In Iraqi NENA and Arabic. The Iraqi NENA in Khan (ibid) are represented by C. Barwar and J. Zakho. In C. Urmi (which has suprasegmental emphasis), [+RTR] is apparently weakened in post-stress syllables (Khan 2016, 68–75). PHONOLOGY 75 historical troṣa; cf. Mutzafi 2008a, 392), but ṭa-la ‘to her’ (rather *ṭaḷa due to mor- pheme boundary). Likely, forms such as the above-mentioned mṣalya [mˤsˤɑˤ̠ lˤyɑ]̠ and ṭamaha [tˤɑm̠ ˤɑˤ̠ hˤɑ]̠ were pivotal in the development of suprasegmental emphasis in C. and J. Urmi. Since they are emphatic throughout, they would have served as a suitable trigger for the reassignment of emphasis from individual segments to the entire lexeme (i.e. /mṣalya/ [mˤsˤɑˤ̠ lˤyɑ]̠ > /+msalya/ [mˤsˤɑˤ̠ lˤyɑ])̠ or at least, initially, to a syllable. This suggestion is in line with Khan’s reconstruction of the origin of suprasegmental emphasis in C. and J. Urmi, which occurred through the ‘phonologization of the phonetic coarticulation’ (Khan 2016, 121).11 2.3. Phonotactics 2.3.1. Consonant cluster resolution Consonant clusters are typically resolved with the epenthetic (or prosthetic) vowel ə (except for certain cases in which the epenthetic has a different quality; see be- low). The epenthetic is a phonetic vowel (e.g. it is invisible to stress), but is 11 Cf. also Napiorkowska (2015a, 55–6), who views emphasis in C. Diyana-Zariwaw—a dialect with transi- tional typology of emphasis resembling to some extent C. Urmi) as a continuum containing both emphasis spread and suprasegmental emphasis. 76 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS indicated in the transcription if featuring regularly in a given context.12 Epenthet- ics in environments which historically had a pharyngeal consonant assimilate in quality to a neighbouring vowel. The sonorants r and l in clusters are resyllabified in a way that places them in syllable offset positions. 2.3.1.1. Word-final CC cluster resolved CǝC Word-final clusters occur in the so-called neo-construct forms whose non-construct form is CVCCa. In such words, the word-final cluster emerges as a result of apoc- opation of the word-final vowel. In fact, there are reasons to consider not only the epenthesis, but also the vowel apocopation a phonetic process, as shown in §2.3.3.2 below. Thus, the neo-construct forms /CVCCa/ are realised [CVCəC], for instance pəsra ‘meat’ is realised pəsər [pǝ.sər] ‘meat of.’ 2.3.1.2. Word-internal CCC cluster resolved CCǝC Word-medial clusters are resolved through epenthesis after the second segment (except for cases specified in the following section). This epenthetic is character- istic of, for instance, the inflected šaqəl-based forms of Pattern III, for instance /mašxna/ > [máš.xə.na] ‘she may warm up.’ 12 For a detailed study of cluster resolution in another NENA dialect, see the C. Barwar grammar (Khan 2008b, 106–111). PHONOLOGY 77 2.3.1.3. Word-internal CCC cluster with liquid as C2 resolved CǝCC If the second segment of the CCC cluster is a liquid (l or r), the CCC cluster is resolved differently: ǝ is inserted after the first consonant. This is common in the šaqəl-based forms of Pattern III with a liquid, e.g. /maxlṣi/ > [ma.xəl.ṣi] (instead of *[max.lə.ṣi]) ‘that they finish.’ Phonotactically, the liquid l is resyllabified from the onset position (*[max.lə.ṣi]) to offset position ([ma.xəl.ṣi]). This is apparently done to satisfy the syllable contact principle, which stipulates that the onset of a syllable should be less sonorous than the offset of the preceding syllable (Vennemann 1987, 40). In the form in question, therefore, the sequence [ma.xəl.ṣi] is preferred to *[max.lə.ṣi]. This principle, however, apparently does not apply with nasals, though these are also more sonorous than obstruents. For instance, /madmxa/ ‘that she put to sleep’ is realised [mad.mə.xa], not *[ma.dəm.xa], producing an onset ([m]) more sonorous than the preceding offset ([d]). Moreover, this resyllabification only applies only in CCC clusters in which epenthesis is licensed phonetically (to break up a cluster), and not with all CC sequences. 2.3.1.4. Word-initial CCC clusters (with a morpheme boundary between C1 and C2) resolved CəCC Word-initial CCC clusters are only attested in cases in which C1 belongs to a prefix or a preposition. In all of those cases, the cluster is resolved CəCC, contrasting with the typical word-internal resolution CCǝC (CǝCC being reserved for liquids), 78 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS for instance /b-dmaxa/ > [bədmaxa] ‘sleeping’ and /d-tre/ > [dətre] ‘of two.’ This word-initial CəCC may thus be motivated by morpheme separation. 2.3.1.5. CC cluster with C1 /ʾ/ resolved with CVC A special type of epenthesis occurs in (word-initial and word-internal) CC clusters with ʾ as the first segment. In these cases, we find an epenthetic which assimilates in quality to the preceding vowel. If ʾ is word-initial, the epenthetic has the quality [a], as shown in the following examples involving Verba primae /ʾ/ of Pattern I: /ʾraqa/ > ăraqa [aʾraːqa] ‘flee’ This vowel is likely a relic from an earlier period where /ʾ/ was still /ʿ/, and the vowel in its vicinity was lowered through an assimilation process. In Verba mediae ʾ in Pattern III, word-internal CC clusters are broken up: /muʾrǝqle/ > muʾŭrəqle [muʾurǝqle] ‘he caused to flee’ This quality is also a historical relic, likely connected to the lenition of the histor- ical /ʿ/. Lenition and mechanisms to avoid are discussed further in Bybee (2015, 30–31) and Molin (2017, 51-7). A comparable qualitative assimilation of epenthet- ics to a nearby vowel in the vicinity of the pharyngeals and glottals (ḥ, ʿ, ʾ and h) is attested in in Tiberian Hebrew (Khan 2013c). In my transcription, these epen- thetic vowels are transcribed, since they occur consistently, but are marked with a breve to make their phonetic status apparent. PHONOLOGY 79 2.3.2. Syllable structure, vowel length and lexical stress 2.3.2.1. Lexical stress As a general rule, stress is penultimate in J.Dohok. Lexical stress is characterised by an increase in intensity, and frequently also in pitch. If the accented syllable is open, moreover, the vowel is long, unless this is [ə] (typically above 95ms). In the spectrogram of the word ḥakoma ‘ruler’ below, for instance, the accented syllable [ko] clearly exhibits the highest intensity (yellow line) and pitch (blue line). The vowel o is also clearly long (115ms). Figure 2: Phonetic manifestation of lexical stress in ḥakoma There is, however, a number of exception to these rules; these are discussed below. 80 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.3.2.2. Loanwords Loanwords appear both in phonologically adapted an unadapted form. Likely, the adapted ones entered the language at an earlier stage than the unadapted ones. Many of these loanwords entered NENA through another language (commonly, Kurdish), adding another layer to their possible phonological adaptations. a) Adapted loan nouns Form Gloss Source language and pronunciation ḥàmmam [ħámmam] ‘bath’ Arabic [ħammáːm] diwan [díːwan] ‘office’ Arabic [diːwaːn] kundắra ‘shoe’ BK ( b) Unadapted (and uninflected) monosyllabic loan adjectives end- ing in a consonant Form Gloss Source language original phonology ḅāš ‘good’ BK [baːʃ] pīs ‘dirty’ BK [piːs] xōš ‘good’ BK [xoːʃ] PHONOLOGY 81 2.3.2.3. Complex verbal and enclitic copula forms This section addresses the rules of stress placement, syllable structure and vowel length in (morphologically) complex and long forms, most of which are verbs and forms with the enclitic copula. The aim is to synthesise the phenomena attested into a few basic phonetic or morpho-phonetic rules which can account for all the forms and to relate them to the general phonotactic principles outlined above. To my knowledge, such a set of comprehensive rules has so far not been proposed in the NENA literature. First, as a general rule, stress is penultimate. There are three types of reason due to which lexical stress sometimes cannot be penultimate; the necessary conditions blocking penultimate stress are most commonly manifested in verbs and forms with the enclitic copula and mostly involve morphemes which cannot take stress. These reasons will be explained separately, and are presented here as Rule 1a (operating in verbal forms excluding the Imperative), Rule 1b (operating in forms with a cliticised copula) and Rule 1c (in Imperatives). In verbal forms excluding the imperative there are conditions blocking the penul- timate placement of stress (Rule 1). In the verbal forms, the semi-clitic L-suffixes (subject- or object-indexing) and the -wa suffix cannot take stress. In such cases, accent falls as close to the penultimate syllable as possible, for instance, on the syllable before an L-suffix and/or wa if these morphemes are penultimate 82 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS (Hoberman 1997, 331) (Rule 1a).13 This is one of the reasons for considering them semi-clitics: on the one hand, like clitics, they do not take stress. On the other hand, they attract stress ‘as close as possible to itself’, unlike clitics, which are invisible to lexical stress. Another semi-clitic property is mentioned later in this section. In all of these forms, furthermore, the application of Rule 1 creates the environ- ment for another rule to apply (‘Rule 2’). Namely, if stress cannot be penultimate, the vowel in the penultimate syllable is lengthened as long as that syllable is open. This is apparently a synchronic mechanism of lending the penultimate (i.e. canon- ically stress-bearing) syllable prominence to compensate for its lack of stress. This type of prominence can be considered secondary stress.14 There are some addi- tional phonetic processes which Rule 1 triggers (or at least correlates with) in some environments; these processes are only manifested in some of the paradigms, and are therefore discussed in the relevant sections below. 13 This rule appears to have been formulated for the first time by Hoberman for J. Amediya, but has been found to be applicable also to J.Dohok, and apparently also to the other LD dialects. This rule is also given for C. Urmi (Khan 2016, 228–29) and C. Diyana-Zariwaw (Napiorkowska 2015a, 66–7). 14 This vowel lengthening is not discussed in the existing LD studies, and indeed in other NENA dialects more generally. Despite the fact that the rule is not stated, the lengthened vowel is sometimes indeed also marked as long. PHONOLOGY 83 2.3.2.3.1. Forms based on šaqəl This category includes all the šaqəl-based with an L-suffix and/or the past affixes (qam- or wa). Consider the following forms, in which the stress shift causes the lengthening of the vowel in the penultimate syllable (Rule 2): gárša [ˈgarʃa] ‘she may pull’ garšà-li [garˈʃaːli] ‘she may pull me’ (Rule 1a) qam-garšà-li [qamgarˈʃaːli] she pulled me’ (Rule 1a) garšà-wā-li [garˈʃaːwaːli] ([aː] in wa: 95ms) ‘she would pull me’ (Rule 1a > Rule 2).15 There is an additional rule which applies to an open (pre-) pre-tonic syllable, and specifically to the syllable which bears stress if the form in question occurs without the L- or a -wa suffix. The vowel in such pre-tonic, open syllable does not shorten, even though, following the suffixation of an L-suffix or a -wa, it no longer bears stress (‘Rule 3’; cf. Mutzafi 2008a, 30): k-xáze ‘he sees’ k-xāzèle [kxaːˈzeːle] ((aː] in xa: 97ms) ‘he sees him’ (Rule 1a > Rule 3) 15 The vowel whose length is under discussion is marked bold. 84 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS qáte ‘that he cut’ qātewa [qaːteːwa] ((aː] in qa: 115ms) ‘he would cut’ (Rule 1a > Rule 3) This rule, however, consistently applies only to the vowel a. With the vowels e and o, the vowel tends to be shorter (short or mid-long), though long realisations are also attested: e: g-mĕθéwa: 58ms gezə́lwa: 75ms keθéwa: 59ms g-ʾēšíwa: 97ms o: ʾoðán-ne: 60ms ʾōðáx-le: 127ms lēpát-te 98ms As was argued in the section on /a/ (ehind other vowels in the process of short- ening outside stress. In the case of a, therefore, this phenomenon should be under- stood as a case of derived contrast (i.e. morphologically-sensitive complementary distribution (Hall 2013, 235): this vowel remains long not in all pretonic syllables, but only in words with the aforementioned semi-clitics. The other vowels (o and e) exhibit an intermediate stage between derived contrast and complementary dis- tribution which is completely phonetically-predictable: outside stress, they are al- most always short, except for a few forms containing the semi-clitic suffixes. In the transcription, such (pre-)pre-tonic length is only indicated with a, since it is predictable only with this vowel. Moreover, as stated above, this is only attested in verbs, and specifically in forms with an L- or a wa suffix. In fact, this process is likely related to the semi-clitic nature of these morphemes (see above). Namely, PHONOLOGY 85 when a remains long outside the stressed syllable, it is as in a sense as if the sylla- ble was still stressed. For instance, ṃā [maː] in g-ṃāḷewā-lu ‘he used to fill them’ would be the stressed penultimate syllable if the form had no wa and L-suffix (i.e. g-ṃáḷe [gmˤaːlˤe]). When both Rule 2 and 3 apply, they produce a form with up to three long vowels. Such prosodically weighty forms, however, seem to be dispreferred. Thus, Rule 2 operates in the two immediately following forms. By contrast, in the remaining tokens which contain the environment for both Rule 2 and 3 to apply, Rule 2 is not effected: g-meθéwā-lu [gmeθeːwaːlu] ([a] in wa: 98ms) ‘he would bring them’ k-ṣarxiwā-la [ksˤarxiːwaːla] ([a] in wa: 95ms) ‘they used to call her’ But: g-ṃāḷewa-lu [gmaːleːwa(ː)lu] ([a] in wa: 88ms) ‘they used to fill them’ dāréwa-lu [daːreːwalu] ([a] in wa: 75ms) ‘he would carry them’ g-dāréwa-lu [gdaːreːwalu] ([a] in wa: 69ms) ‘he used to carry them’ g-nābǝlwa-lu [gnaːbəlwalu] ([a] in wa: 64ms) ‘he used to take them’ In short, the vowel in the penultimate syllable is longer in forms which do not have a pretonic ā. In the first group, the vowels in the penultimate syllable should be considered long (though their values are relatively low for long vowels), while 86 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS in second group, these vowels are mid-long (in reality, likely on a continuum from long to mid-long or perhaps short). These tokens suggest, therefore, that Rule 2 (pre-tonic lengthening) is blocking the application of Rule 3 (penultimate length- ening), or at least makes it optional. As previously stated, Rule 2 is apparently a historical relic, Rule 3 being a synchronic process; the former appears to be stronger. Verbal forms with the 2PL E-suffix -etun (cf. §5.2) are distinct to the verbs discussed so far, though as will be shown, it is possible to create one general rule of stress placement in verbs which applies to all forms. The -etun suffix is the only disyllabic E-suffix, and its second syllable cannot take stress, meaning that stress will always remain on the first syllable of the E-suffix. Without a semi-clitic suffix, the stress occupies the usual penultimate position: garšétun [garˈʃeːtun] ‘that you (c.pl) pull’ The discussion so far can be summarised with one general rule of stress placement in verbs, captured succinctly by Khan (2008b, 128) for C. Barwar, but widely ap- plicable across NENA: ‘stress has to remain on a syllable that contains a radical of the verbal base,’ e.g. (the last radical in bold) g-garšétun ‘you (PL) pull, k-šaqlá-wā- le ‘she used to take him.’ As will be shown below, this does not apply to the im- perative. PHONOLOGY 87 As expected, therefore, when an L- or a -wa suffix are added do a form with the - etun morpheme, the stress remains. However, in the LD dialects, the vowel in the stressed syllable also shortens to [ə]:16 garšə́tū-le [garˈʃətuːle] that you (c.pl) pull him’ This shortening of the suffix in LD may be motivated by a pressure to shorten the prosodically heavy form (with an unusually long E-suffix, followed by an L-suffix and/or wa). 2.3.2.3.2. Forms based on šqəlle In forms based on the Past, Rules 1a and 2 apply (there is no phonetic environment for Rule 3 to be manifested). Since an L-suffix does not take stress, in some forms (e.g. Pattern I verbs), stress falls on the initial syllable. Moreover, the penultimate syllable is lengthened if it does not carry primary stress. šqə́llōxun [ˈʃqəlloːxun] ‘you (PL) took’ (Rule 1a) húllū-lan [ˈhulluːlan] ‘they gave us’ (Rule 1a > Rule 2) 16 This shortening does not occur in many dialects outside the LD group; cf. in C. Barwar, where the 2pl suffix has the form -ítu in strong verbs, and does not shorten: qaṭl-ítu ‘you (PL) kill’ qaṭl- ítu-wa ‘you (PL) used to kill.’ 88 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.3.2.3.3. Forms with enclitic copula The present 3rd persons copula are clitics, meaning that they are one step further removed from being morphologically integrated into the word than the L- and wa suffixes: they are invisible to stress altogether.17 Therefore, the stress falls on the penultimate syllable of the form preceding the copula (a non-verbal predicate, e.g. noun, adjective or an adverb) (Rule 1b). Still, even clitics show a degree of inte- gration with the form to which they cliticise, because Rule 2 also operates in forms with clitics. In other words, the penultimate syllable of the form with a copula is still lengthened, creating secondary stress, as if the entire construction were a sin- gle phonological word. When the longer copulas (ila, ile, ilu) are used, this length- ening will apply to the first syllable of the copula itself: bāš́ =īle [ˈbˤaː̠ ʃiːle] (119ms) (Rule 1b > Rule 2) bríndar=īle [ˈbrindariːle] (106ms) (Rule 1b > Rule 2) When the short copula is used, the syllable preceding it is lengthened: gúrgā=le [ˈgurgaːle] (151ms) (Rule 1b > Rule 2) ʾékē=la [ˈʾeːkeːla] (128ms) (Rule 1b > Rule 2) 17 Cf. Fassberg (2010, 34): ‘enclitic particles and words (…) do not take stress’. This is somewhat misleading, however, because clitics are altogether invisible to lexical stress. PHONOLOGY 89 2.3.2.3.4. The Imperative In contrast to other verbal forms, in the imperative, lexical stress is fixed on the first syllable (Rule 1c) in J. Dohok, as case in many other NENA dialects. Napior- kowska suggests that this lends the imperative ‘prominence’ (Napiorkowska 2015a, 76), likely meaning a sense of urgency. This stress placement is also in- duced by contact with Kurdish. Consider the following form: mšádər-re [ˈmʃaːdərre] ‘send (SG) it!’ (Rule 1c) Some forms also contain a suitable environment for Rule 2 to apply; this is com- mon in the plural forms (with the morpheme -u) followed by an L-suffix: grúšū-le [ˈgruːʃuːle] (102ms) ‘pull (PL) him!’ mšádrū-le [ˈmʃadruːle] (139ms) ‘send (PL) him!’ 2.3.3. Annexation units ‘Annexation’ refers here to the adjoining of two or more nouns (or sometimes a noun in combination with a pronoun or a numeral) in order to express relation- ships such as possession, category or a characteristic feature. There are four types of annexation units in LD: historical construct, asyndetic annexation, annexation with the genitive particle -ət and neo-construct. Constructions with the genitive - 90 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS ət particle show no exceptions to the general phonotactic rules. In this section, the historical construct and neo-construct are discussed, as these display noteworthy phonotactic features. (Some) historical construct forms reflect a high degree of phonological integration, while neo-construct forms are often less integrated – of- ten, the genitive –ət—though phonetically apocopated—is still present on the un- derlying level. 2.3.3.1. Historical construct state A few annexation constructions are likely retentions of earlier-Aramaic construct state, rather than neo-construct.18 In the J. Dohok corpus, this is attested with the nouns mare- [mare] ‘owner, lord of’ (historical *māre (?) ‘lord, owner’; cf. Mutzafi 2008a, 92), be- ‘house of’ (historical *bēθ < *bayθ ‘house of’), and bər- ‘son of’ (historical *bar ‘son of’), for instance: mar-e-dawəlta owner-GEN-wealth ‘rich man’ bər-Ø-ʾamoy-ax son-GEN-paternal_uncle-yourFS ‘your (FS) paternal cousin’ be-Ø-kalo house-GEN-bride ‘[the] bride’s house’ 18 So also likely cases of historical construct in other dialects: C. Barwar (Khan 2008a, 401), C. Urmi (2016, 542-5) and J. Betanure (Mutzafi 2008a, 92). PHONOLOGY 91 These forms reflect a high degree of phonological integration, which reinforces the hypothesis of their historical retention, rather than of innovation. This is ap- parent especially in măre, which has the short ă in an open, penultimate syllable. This shortening most likely indicates that the entire annexation unit has become a single prosodic word with only one lexical stress, and that, consequently, the syllable with mă is synchronically unstressed. 2.3.3.2. Neo-construct state Neo-construct forms are created by the apocopation of the word-final genitive morpheme -ət (or -tət/θət in the feminine singular forms with the -ta/θa suffix, e.g. baxta ‘wife, woman’> baxtət > bax ‘wife of’). For phonological purposes, two types are distinguished (see below): in one, apocopation creates an open (pho- netic) syllable, in th other, it closes a syllable, but the CC cluster is subsequently resolved on the phonetic level. In both cases, the apocopation results in changes in the word’s phonetic structure. In some cases, the phonological structure (vowel length and stress) is left unaffected by the apocopation, while in others, the struc- ture changes following the apocopation. In Khan’s grammar of C. Urmi, neo-con- structs are analysed as phonologically unapocopated, that is, preserving the –ət 92 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS annexation morpheme on the phonological level (Khan 2016, 217).19 In J. Dohok, this tendency exists, but is apparently not consistent. 2.3.2.3.5. Apocopation causing syllable closure (CV.CV > CV̄C and CVC) This apocopation pattern occurs with nouns containing a long vowel in an open penultimate syllable, which is closed, following the creation of the neo-construct. The length of the vowel in the newly-closed syllable varies: in some cases, the vowel is shortened, reflecting a phonological closure of the syllable in the neo- construct, in others, length is preserved, indicating that phonologically, the form is still unapocopated. With the vowel u, both short and long variants are attested and further data would be necessary to establish the existence of any patterns. In general, this variation most likely reflects the ongoing nature of the phonological reanalysis. a) Forms preserving vowel length palgūt́ -Ø dunye (102ms) ‘middle of the world’< pal.gu.θet ‘middle of’ lēl-Ø šabθa (106ms) ‘Shabbat eve’< le.lət ‘eve of’ brōn-Ø malka (112ms) ‘king’s son’ < bro.nət ‘son of’ 19 Cf. also C. Barwar (Khan 2008b, 75); it is unclear whether the phonological preservation of the original form is consistent. PHONOLOGY 93 In pal.gūt́ , the lack of phonological reanalysis is also apparent in the placement of stress on the final syllable, that is, the penultimate syllable of the unapocopated form pal.gú.θət. b) Forms with a shortened vowel brat-Ø ʾamoyox (76ms) ‘your (MS) paternal cousin’ < bra.tət xur-Ø broni (65ms) ‘my son’s friend’< xu.rət ’friend of’ 2.3.2.3.6. Apocopation causing syllable opening (C.CV > C.əC) This group includes nouns whose penultimate syllable is closed. In this group, the apocopation of the coda creates a word-final CC cluster, which is resolved to CəC, separating the heavy tri-moraic syllable (CVCC). In all the cases attested, however, this epenthesis is phonetic; on the phonological level, this syllable remains pho- nologically closed (CVCC). This is apparent from the fact the vowel remains short (in the transcription, this shortness is indicated by the breve sign), for instance: ʾămər ʾərwe (45ms) ‘sheep wool’ < am.rət ‘wool of’ dĕʾǝθ bəṭme (59ms) ‘terebinth resin’ < deʾ.θət ‘resin, sweat’ 94 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS 2.3.4. Forms with numerals In independent numerals—except those including teens—stress is penultimate, for instance šóʾa ‘seven’. Teens are treated as a single phonological word: there is only one syllable with prominent stress and it falls on the the ultimate syllable of the first word (the unit), for instance, ṭṃāné-ʾĕsar. The stressed vowel e, however, re- mains short – apparently because the syllable is analysed as closed, i.e. /ṭṃā.neʾ.sar/ [ṭˁṃˁaː.neʾe.sar]. In numerals modifying a nominal, the stress on the numeral is either penultimate or penultimate; in the former, the stress on the numeral is acoustically more prom- inent than the stress on the nominal. The same phenomenon is recorded for C. Barwar, where stress is said to give ‘prominence’ to the numeral (Khan 2008b, 130-1), that is, often its focal function. In sentence (1), for instance, the numeral carries narrow focus (underlined): (1) 5a/22 Context: 'For instance, Passover, the Seder, people would not celebrate in- dependently, in their own houses.‘ xamka naqle g-yatw-i-wa xamší ʾəštì naše.ˈ ‘Sometimes, fifty or sixty people would sit together.’ In J. Dohok, however, this word-final stress does not occur with teens, which can apparently only be stressed on the ultimate syllable of the first unit. PHONOLOGY 95 2.4. Intonation units: nucleus stress and pause Intonation units are speech units with a prosodic contour20 and containing one (or occasionally more) prosodically prominent segment, which carries the nucleus stress. Nucleus stress is conveyed through increased length, intensity and pitch of the stressed syllable, and this prosodic prominence typically mirrors discourse prominence (see below). Segments at the beginning and end of intonation units (‘in pause’) are often also prosodically prominent, for instance, by being length- ened. Intonation units tends to overlap with syntactic units such as the clause, and therefore—on the functional level—tend to mark an information unit within the discourse (Grice 2006, 778-786). In J.Dohok, the duration of the long vowel in syllables carrying nucleus stress is above 105ms. For the tokens collected, the mean average is 126ms, which con- trasts with the average quantity of 112ms attested for stressed vowels in non-nu- cleus and non-pause positions. In sentences with the unmarked SVO word-order, nucleus stress is placed within the broad focus group, towards the end of the clause. It tends to fall on the object if the object is present.21 A clause-final nucleus stress may therefore coincide with 20 For different types of pitch contours in a NENA dialects and their information-structural func- tions, see Napiorkowska’s grammar of C. Diyana-Zariwaw (2015, 72–76). 21 Cf. the chapter on word order (§9). 96 PHONETIC PROCESSES AND LARGER PHONOLOGICAL UNITS the form in pause. This is illustrated by the sentence in (2) with broad focus (‘they dressed him in new clothes’) and the clause-final object ‘new (clothes’) in pause and carrying nucleus stress: (2) 1a/15 Context: ‘They took the Jew.’ qam-malušíle badla xàθa.ˈ ‘They dressed him in new clothes.’ The corpus also contains a few pausal forms in which the stress is retracted, which is also attested for dialects such as C. Urmi (Khan 2016, 228–9). Whether inci- dentally or not, all of the J. Dohok pausal forms with retracted stress exhibit a ‘more archaic’ stress pattern. They are verbal forms with wa or with an object- marking L-suffix in which the stress falls on the penultimate syllable of the verb proper, as if the L-suffix or wa were disregarded by lexical stress. (3) 1a/8 hulle ṭalu ṭḷà yome ʾaθe ʾamrì-le,ˈ là,ˈ rešu màfǝr-re (V-L-SUFFIX).ˈ ‘He gave them three days [so that] they would come tell him, [if] not, he would cut off their head.’ (4) 5b/18 B-az-i-wa l-knəšta u-mṣàle-wa.ˈ ‘They would go to the synagogue and pray, then come [back home].’ PART TWO MORPHOLOGY CHAPTER THREE PRONOUNS This chapter gives an overview of the pronominal inventory. 3.1. Independent personal pronouns Table 18: independent personal pronouns in J. Dohok 3MS ʾawa 3PL ʾani 3FS ʾaya 2MS ʾahǝt/ʾāt 2PL ʾaxtoxun 2FS ʾahat/ʾāt 1CS ʾana 1PL ʾaxni 102 PRONOUNS 3.2. Demonstrative pronouns 3.2.1. Near and far deixis Table 19: Deictic pronouns in J. Dohok Near deixis Far deixis Independent Attributive Independent Attributive MS ʾoha ʾo (marginal ʾoha) ʾawa ʾaw /ʾawa FS ʾeha ʾe (marginal ʾeha) ʾaya ʾay /ʾaya PL ʾanna ʾanna ʾani ʾan 3.2.2. Very far deixis These very far deixis pronouns are only attested in elicitation, suggesting that they occur extremely infrequently. MORPHOLOGY 103 MS ʾawaha FS ʾayaha PLL ʾanaha 3.2.3. Anaphoric pronouns For anaphora, the far deixis independent or attributive demonstratives are used (cf. example (1) below), though the near deixis pronouns are also attested in a few instances. (1) 1b/87 ʾawa ḥakom-a krə̀b-le,ˈ g-emər-Ø… thatM ruler-MS PERF.be_angry-he, REAL-say-he ‘(When he heard this,) the ruler got angry and said…’ 3.3. Pronominal suffixes on nouns and prepositions In J. Dohok, there is only one set of possessive suffixes, used for both singular and plural nominals. When the possessive morphemes are suffixed to plural nouns end- ing in -e, this plural suffix is apocopated, leading to the loss of a formal distinction between plural and singular nouns in such cases. For instance, reš-oxun can mean either ‘your head’ (from reša ‘head’) or ‘your heads’ (from reše ‘heads’). 104 PRONOUNS Table 20: Possessive suffixes in J. Dohok 3MS -e (-eh before a vowel) 3PL -u 3FS -a (-ah before a vowel) 2MS -ox 2PL -oxun 2FS -ax 1cs -i 1PL -an The 1PL allomorph -eni which is recorded for J. Challa and J. Betanure (respec- tively, Fassberg, 2010, 37 and Mutzafi 2008a, 47) does not exist in J. Dohok, ex- cept for in combination with the independent genitive particle (see below). The 3FS and 3MS preserve the historical h before vowels, as illustrated in (2) below: (2) 3/interview ʾoha beθ-ah =ile thisM house-her =PRS.COP.he ‘this is her house.’ In the Mosul plain dialects, the segment h is preserved in all context (thus, -eh and -ah in C. Bariṭla - Khan 2008b, 143), or is even strengthened to ḥ (-eḥ and aḥ) in C. Qaraqosh (Khan 2002, 76). Speakers have no difficulty using the 3MS and 3FS suffixes, respectively -e and -a, despite the identity of these suffixes to the nominal endings (plural and singular respectively). This is illustrated in (3) below. In C. Barwar, by contrast, this MORPHOLOGY 105 ambiguity of the possessive suffix is sometimes resolved by adding the independ- ent possessive particle diy- to the noun (Khan 2008b, 143), leading to a double marking of the possessee. Such double marking does not occur in J. Dohok. (3) 5b/49 Context: ‘The engaged couple wound not see each other.’ kudxa bēt-Ø gyàn-e.ˈ ʾawa go bèθ-e, ʾaya go bèθ-a.ˈ everyone house-gen self-his he in house-his she in house- her ‘Everyone (would stay in) their own house: he in his own house, she in house.’ Alongside possessive suffixes, another common strategy for expressing possession in NENA is with the independent genitive particle (see below). Possessive suffixes are generally preferred with inalienable nouns (e.g. C. Qaraqosh – Khan 2002, 272). In J. Dohok, possessive suffixes occur with body parts, kinship relations, physical and emotional states, nouns associated with identity such as ‘language’1 and ‘religion’, as well as those whose ‘inalienability’ is debatable, e.g. ‘house’ (cf. (3) above). 1 Including lišan-an ‘our language’ in the exclusive sense. 106 PRONOUNS 3.4. The independent genitive particle As mentioned previously, another common strategy for expressing possession is with the independent genitive particle+possessive suffix. The inflection is given below: Table 21: Inflection of the independent genitive particle with possessive suffixes in J. Dohok 3MS did-e 3PL d-eni 3FS did-a 2MS did-ox 2PL d-oxun 2FS did-ax 1CS did-i 1PL d-ohun In J. Dohok, it is not possible to use a double possessive construction such as *beθ- e did-e house-his POSS-his ‘his house’, though this apparently occurs (in the 3MS form) in J. Betanure and C. Barwar (respecitvely, Mutzafi, 2008a, 42 and Khan 2008b, 143). MORPHOLOGY 107 The base did-, used in LD with singular possessive suffixes, contrasts with the pho- netically more developed form diy-, common in Christian dialects of the ‘core’ NENA territory (northern Iraq). Outside LD, did- also occurs in most of the TZ varieties, where it is also used with plural suffixes, e.g. didan ‘ours’ (Khan 2009, 11-2). Christian varieties with did- in the singular and (some) plural forms include C. Qaraqosh in the south (Khan 2002, 84), C. Bohtan in the north-west (Fox 2002) and C. Shaqlawa east of the Zab. 3.5. Reflexive pronoun The reflexive pronoun in J. Dohok and most other NENA dialects is formed with the base gyan+possessive suffix, as illustrated by the third person forms below: 3MS gyan-e 3PL gyan-u 3FS gyan-a The base gyan- is an Iranian loanword, though its exact origin is not entirely cer- tain. Contemporary Bahdini varieties possess both gyan and jan ‘soul, self, body, 108 PRONOUNS dear’, though jan is the standard (and perhaps more common) form.2 The form gyan also exists in Sorani and Gorani,3 neither of which are synchronically in con- tact with the LD dialects. Gorani is a north-west Iranian language, but a more distant relative of Kurdish. If the NENA gyan does not come from Bahdini, it could be a Gorani loanword. Gorani was once more widespread throughout the region than it is in the present day (Leezenberg 1993), and its contact with NENA may have led to mutual NENA-Gorani influences, including lexical borrowings. 3.6. The pronoun of independence The pronoun expressing independent agency, i.e. ‘by one’s self’, is formed of b- gyan+possessive pronoun, for instance b-gyan-i ‘by myself’. B-noši (see below) is also used with this meaning. 3.7. The pronoun of isolation Isolation, i.e. ‘one’s self’, is expressed with b-noš+possessive suffix, for instance b- noš-i ‘myself’. 2 Masoud Mohammadirad, personal communication. In the Barwar area, the two forms are reported to coexist side-by-side. 3 Masoud Mohammadirad, personal communication. MORPHOLOGY 109 3.8. Interrogative pronouns ‘who?’ măni ‘what?’ ma (realised with a short vowel: [ma]) ‘which’ ʾema 3.9. Indefinite pronouns ‘someone’ xă ‘something’ xaməndi ‘no one’ čuxxa (secondary gemination to close the syllable < čŭxa) ‘nothing’ čŭḿ əndi ‘other’ xēt/xət/xeta (interchangeable) ‘everyone, anyone’ kudxa ‘everything’ kudməndi Čŭḿ əndi and čuxxa must be accompanied by a negated predicate, for instance: (4) 1c/50 čuxxa là k-xāze -Ø -le.ˈ nobody NEG REAL-see -A.he -P.him 110 PRONOUNS ‘Nobody will see him.’ The lack of inflection for this pronoun is a distinctive feature of Jewish NENA varieties (Mutzafi 2016), both LD and TZ. Many Christian dialects inflect this pro- noun for gender and number. CHAPTER FOUR OVERVIEW OF NOUNS AND ADJECTIVES This chapter gives an overview of the nominal inflection. Nouns 4.1.1. Singular endings The most common singular endings for Aramaic words and for adapted loan- words are: a) a: grammatically masculine preferred for nouns which are semantically masculine example: ḥakom-a ‘ruler’ b) -ta/-θa: grammatically feminine preferred for nouns which are semantically feminine related endings: -iθa, -uta examples: parsəm-ta ‘wound’, nun-iθa ‘fish’ Other endings occurring in Aramaic words include -o, for instance kal-o ‘bride’. The origin of this suffix is not entirely clear; Kurdish etymology (Mutzafi 2008a, 91) and derivation from the historical diminutive ending -ūn (Khan 2016, 495) have been proposed. A few word have the singular ending e-, identical with one of the plural ending e, e.g. lel-e ‘night’ (PL lel-awaθa). 112 OVERVIEW OF NOUNS AND ADJECTIVES 4.1.2. Plural forms The most common plural endings are: a) -e example: kundar-e ‘shoes’ (SG kundar-a) b) -aθa example: šab-aθa ‘Shabbats’ (SG šab-θa) There is a number of endings similar to -aθa: -waθa, -yaθa and -awaθa. These occur with nouns which have various endings in the singular (cf. Mutzafi 2008a, 96-100): ‘holiday’ ʾeð-a (MS) ʾeð-awaθa ‘friend’ xur-a (MS) xur-awaθa ‘fish’ nun-iθa (FS) nun-yaθa ‘sister’ xa-θa (FS) xaθ-waθa Another relatively common ending is -ane; this is attested most commonly with nouns that have the MS ending -a, for instance: ‘vineyard’ karm-a karm-ane MORPHOLOGY 113 4.1.2.1. Semantically masculine nouns with -a (SG) and -aθa (PL) Most semantically masculine nouns have the plural ending -e, for instance malk-a ‘king’/malk-e ‘kings’, but other plural form exist. For instance, several male kin- ship terms have the plural ending -awaθa or aθa: ‘father’ bab-a bab-awaθa ‘brother’ ʾaxon-a ʾaxa-waθa ‘grandfather’ sawoy-a sawoy-aθa/saw-aθa 4.1.2.2. Semantically feminine nouns with θa/ta (SG): -aθa and -e in PL As a general rule, semantically feminine nouns with no masculine counterpart of the same stem have the plural ending -aθa, or a related ending such as -waθa or yaθa: ‘mother’ yəmm-a yəmm-aθa ‘sister’ xa-θa xaθ-waθa However, nouns referring to female human beings inflected from a masculine stem typically have no special feminine form, merging with the semantically masculine forms in the plural: ‘singer’ zamar-a (MS) zamar-ta (FS) zamar-a 114 OVERVIEW OF NOUNS AND ADJECTIVES 4.1.2.3. Irregular and suppletive plurals The plural stems of some nouns is slightly different to their singular stem: ‘man’ gora gure/gorane ‘house’ beθa beθawaθa/batane ‘dog’ kalba kalwe ‘year’ šata šǝnne ‘son’ brona bnone In contrast to most Christian dialects and LD dialects such as J. Betanure (Mutzafi 2008a, 383), the plural form of ‘thing’ *məndyane does not exist in J. Dohok and the inflection is suppletive: ‘thing’ məndi ʾawaye 4.1.2.4. Nouns with no singular form (pluralia tantum) The following forms are examples of nouns which only have a plural form, and therefore always agree with a plural adjective: ‘water’ ṃaye ‘sky’ šəmme MORPHOLOGY 115 ‘mercy’ raḥamime (includes an incorporated Hebrew plural ending - im) ‘life’ xaye ‘money’ pare Adjectives and their agreement with nouns 4.2.1. Inflection of adjectives and agreement In Aramaic nouns and adapted loanwords, only three inflectional endings exist: a) -a: masculine singular b) -ta/-θa: feminine singular c) -e: common plural In an attributive noun phrase, a noun always precedes its adjective and the two agree in number. Gender agreement exists only in the singular. Examples of at- tributive noun phrases are shown below: xabuš-a smoq-a xabuš-e smoq-e apple-MS red-MS apple-PL red-PL tawər-ta smoq-ta tawər-yaθa smoq-e cow-FS red-FS cow-PL red-PL 116 OVERVIEW OF NOUNS AND ADJECTIVES sawoy-a zor-a saw-aθa zor-e grandfather-MS small-MS grandfather-PL small-PL yabəš-ta kom-ta yabiš-e kom-e raisin-FS black-FS raisin-PL black-PL xay-e yarix-e life-INV long-PL 4.2.2. Uninflected adjectives Many loaned adjectives remain uninflected for gender and number: yalonk-e ḅāš-Ø child-PL good-INV ‘good children’ malək-ta xōš-Ø queen-FS good-INV ‘a good queen’ MORPHOLOGY 117 Annexation of nouns The phonotactics of annexation units have been discussed in §2.3.3. For a mor- phological and functional comparative treatment of annexation (attributive) con- structions in NENA, see Gutman (2018). There are four types of annexations in J. Dohok. In this dialect, the Kurdish geni- tive particle -e is unattested. Within LD, the existence of this particle is reported only for J. Challa, where it occurs in expressions which are originally from Ara- bic, e.g. ʾawlād-e rasūl ‘descendants of the Prophet (Muhammad)’ (Fassberg 2010, p.56). 4.3.1. Annexation by means of the genitive particle –ət The most common method of annexation is with the morpheme -ət enclitic to the first word. Though the historical form of this particle is d, it is best to consider - ət the synchronic underlying form. The reason for this is that -ət is the more common variant, and occurs not only before voiceless sounds, but also before voiced ones in some cases. Occasionally, however, -ət is voiced to -əd in assimila- tion to the following voiced phoneme (cf. (a) below). -ət as the underlying form has also been prosed for C. Barwar (Khan 2008b, 396-7). In the transcription, the voiced and voiceless allomorphs are distinguished. If the first word ends in the 118 OVERVIEW OF NOUNS AND ADJECTIVES vowel -e or -a, this vowel is apocopated. Occasionally, d- appears as a proclitic of the second word (cf. c)). a) yoməd ʾəròta from yoma ‘day’ 'Friday‘ b) našǝt qamaye from naše ‘people’ ‘the first people’ c) lʾine d-pare ‘jugs (full) of money’ 4.3.2. Annexation with the independent genitive particle did The independent genitive particle did is also used in the J. Dohok corpus, though less frequently than the enclitic -ət. In J. Betanure, the form is did (Mutzafi 2008a, 92). kǝsta dǝt pàre bag GEN money ‘a bag of money’ 4.3.3. Historical construct state As suggested in §2.3.3.1, a few annexation units reflect a construct state pre- served from earlier Aramaic, for instance: mar-e-dawəlta owner-GEN-wealth ‘rich man’ MORPHOLOGY 119 4.3.4. Neo-construct state There are occasional attestations of the so-called ‘neo-construct’, created by the apocopation of the genitive morpheme -ət. Though apocopated in speech, how- ever, there are phonetic reasons to consider the genitive particle present on the underlying level (see §2.3.3.2). The two types of neo-construct identified previ- ously are illustrated below: a) brat-Ø ʾamoy-ox daughter-GEN paternal_uncle-yourMS from < brat-ət daughter-GEN ‘your paternal cousin’ b) ʾămər-Ø ʾərw-e wool-GEN sheep-PL from amr-ət ‘wool-GEN ‘sheep wool’ PART THREE SYNTAX AND MORPHOSYNTAX CHAPTER FIVE VERBAL MORPHOLOGY Introduction: Derivational morphology in Semitic In traditional approaches to Semitic verbal morphology, verbs are said to comprise of pattern and root, ‘root’ being a set of typically three segments (‘radicals’) which is discontinuous, abstract, and (historically) consonantal. Root and pattern1 to- gether form a stem to which inflectional, that is, person-number-gender affixes are added. The manifestations of a given pattern are inflection-specific2 and the in- flectional bases are referred to here as ‘stems’, defined primarily by their distinct vocalic patterns. Root-and-pattern derivation contrasts with stem-and-affix mod- els. In the latter, the specific phonetic shape of an inflectional stem is derivation- ally unmotivated, except in the sense that it can be derived from a different stem (cf. Bybee 1985, 49-77). Recent studies, however, cast serious doubts on the real- ity of a ‘root’. This applies especially to cognitive language processing (of ‘weak’ verbs), but sometimes also for the descriptive usefulness of the root-and-pattern model. A more nuanced understanding of root radicals is articulated by 1 The term ‘pattern’ is used here for a conjugational class, as in Khan (2016). Most other NENA grammars (e.g. Mutzafi 2008a and Napiorowska 2015a) use the term 'stem', following the convention in traditional Semitic philology. Coghill (2003) employs the designation ‘class’. ‘Tem- plates’ sometimes refers to specific stems in root-and-pattern derivation. 2 The terms ‘inflection’ and ‘paradigm’ are synonymous here. 122 VERBAL MORPHOLOGY Goldenberg (2013, 117–20), who introduces the concept of radical ‘slots’. In this model, verbal patterns require three radical slots to be filled, but do not necessarily require a tri-radical root. This allows for the manifestation of patterns in the case of incomplete roots. Namely, if a radical has been lost, its slot within a given pat- tern will be filled by a new segment (for instance a vowel or a new radical), or a Ø radical (see below). For the sake of categorisation convenience and for the easy access of Semitists, this chapter uses historical roots as a starting point of the analysis, including roots which are no longer manifested in the synchronic forms.3 This presentation not- withstanding, several verbs are analysable as having new synchronic roots, that is, a new set or new sets of slot-filling segments (cf. above). It is this sense of the term ‘root’ which is used in the synchronic description, in line with Goldenberg. In contrast to the existing LD studies with only list historical roots, an effort is thus made to distinguish between a synchronic and a diachronic description. Some verbs and paradigms are even more innovative, meaning that no synchronic pat- tern and root can be identified in them. Instead, their stems are synchronically underived, or sometimes derived from another stem. 3 Such demonstrably ‘historical’ roots are introduced with an asterisk. MORPHOLOGY 123 The subsequent sections introduce the J. Dohok verbal system, present classes of weak verbs (§5.3-5.5) and the most frequent irregular verbs (§5.6),4 which are supplemented by comparative data of LD and a TZ representative. The final section reviews the discussion, attempting to capture the typology of verbal derivation in NENA. I argue for a mixed derivation which involves both stem-and-affix and pat- terns, sometimes combined within a single verbal lexeme. Crucially, however, the non-stem-and-affix derivation is considered here to be based on patterns, rather than on roots and patterns. For instance, analogical change from one verbal class to another, discussed in irregular verbs, is indicative of a paradigmatic pressure to adherence to existing derivational classes (‘weak verb’ classes or ‘strong verb’ Pat- terns). It is shown, however, that does not automatically result in a pressure to create a single synchronic root. For instance, for a single lexeme, multiple syn- chronic ‘roots’ can sometimes be postulated. In other instances, the synchronic paradigms are not easily derived from any synchronic root. If the conclusions of this chapter were to be applied consistently, the presentation would have to be altered significantly, but this is not carried out to the full extent.5 4 The absence of a particular class from the discussion (e.g. verba mediae /ʾ/) indicates that such a class follows the ‘strong’ morphology. Paradigms are typically illustrated only by the 3MS form if the rest of the paradigm is predictable from this form. 5 An example of the proposed pattern-based model in this chapter is the listing of 3MS šaqəl exem- plar for each type of weak root, that is, for a particular verbal class. 124 VERBAL MORPHOLOGY Such a new model would involve grouping verbs under derivational and inflec- tional classes and abandoning the quest for roots, instead giving particular lexical exemplars of the class. For verbs that belong to various patterns due to the lack of a single synchronic ‘root’, various exemplars would be listed for various inflec- tions. Verbs with stems that do not conform to existing patterns would have to be listed separately, though it would be possible to give a paradigm of shared inflec- tional suffixes for a few of these. The motivation for keeping the concept of roots— aside from the aforementioned attempt to accommodate to Semitists—is to test its usefulness and demonstrate its challenges, which is revisited at the end. The con- cept of ‘radical slots’ is useful for describing a position within the verbal form. The inventory of patterns, stems and inflectional affixes 5.2.1. Patterns and stems and their historical background There are four conjugational patterns (I-IV) in J. Dohok, descended from older Aramaic patterns, respectively: pǝʿal (‘Ground stem’), paʿʿel (‘Doubled/Geminate stem’) and ʾap̄ʿel (‘Causative stem’). Pattern IV has four radical slots (four radicals). For each of these patterns, the following stems (that is, inflectional bases) exist. There are tense-aspect-modal inflections not presented here, but those are based on the stems below and thus do not require to be listed separately. MORPHOLOGY 125 šaqəl (prototypically, non-past irrealis), b-šaqəl (prototypically, predictive future) and k-šaqəl (prototypically, realis present) šqəlle (prototypically, perfective past) Imperative Participle Infinitive Šaqəl is derived from the historical active participle in the absolute state. In most cases, šaqəl shares its stem with b-šaqəl and k-šaqəl, though k-šaqəl paradigms of verbs with the historical /ʾ/ as the first radical should be considered as having a different synchronic stem (see below). The preverb k- is primarily a marker of realis6 is devoiced in verbs beginning with a voiceless consonant. Moreover, this morpheme is inaudible in verbs starting with k, g or q. In those cases, the morpheme has assimilated to the following consonant and subsequently elided. In NENA dialects with the habitual realis preverb ʾi-, such as C. Marga, this preverb is in fact preserved only in initial /ʾ/ verbs, where it has undergone degrammaticalisation, becoming integrated within the verbal stem.7 6 For a semantic-syntactic discussion, see §8.3. For an overview of the function and history of the habitual-realis preverbs across NENA (typically k-/g- or ʾi-/y), see Khan (2007). 7 I owe this insight to a discussion with Paul Noorlander, with whom I conducted fieldwork on this dialect. 126 VERBAL MORPHOLOGY Overall, given that the absence of k- is ascribed to a simple phonetic cause and considering the relatively small number of cases of the preverb’s absence, it is best assumed that the preverb in verbs beginning with k, g or q is present on the un- derlying level, that is, that the forms in question are underlyingly k-šaqəl rather than šaqəl. This is therefore reflected in the transcription (e.g. q-qaṭe-Ø [qaṭe] HAB- cut-he ‘he cuts’). The b-šaqəl paradigm is formed by adding the predictive-future preverb b- to šaqəl base, but this prefix is only clearly audible in Pattern I verbs. In the other forms, the preverb is never audible, except for nonce borrowings (see below). It assimi- lated to the following m, and this gemination of m has subsequently been lost, e.g. *b-mbašəl-Ø FUT-cook-he > m-mbašəl > mbašəl ‘he will cook’. Such realisation contrasts with that of spontaneous derivations from MH verbs, where the preverb is realised and even unassimilated. Compare, for instance, the spontaneously de- rived bə-mšaləm-Ø FUT-pay-he ‘he will pay’ from MH mešalem ‘he pays’ with Ø- mšadər-Ø IRR/FUT-pay-he ‘he will/might pay’. This disparity despite the same pho- netic environments (m as the first segment of the šaqəl stem) in fact suggests that with the established lexicon, the preverb is synchronically absent even on the un- derlying level. Since it is impossible to confirm the presence of the preverb in individual cases in patterns II-IV verbs, the preverb is not transcribed in those cases and such forms should be considered ambiguous. This ambiguity is indicated in glossing, for instance Ø-maxḷəṣ-Ø IRR/FUT-save-he ‘he may/will save’. MORPHOLOGY 127 Šqəlle is derived from the historical resultative participle in the absolute state.8 The Imperative in Pattern I is based directly on the older Aramaic imperative while in the other patterns, the prefix m- has been added to the original imperative forms. The Participle the Participle is a reflex of the older Aramaic resultative par- ticiple in determined state. The infinitive Infinitive of Patterns II-IV comes from earlier Aramaic infinitive forms and in the case of Pattern I, likely from a verbal noun of the pǝʿal Aramaic pattern (‘basic stem’; Mutzafi 2008a, 45). The following forms exemplify the inflectional stems of strong verbs for all verbal patterns: 8 As Noorlander shows, the basic function of this participle is resultative, rather than passive (forthcoming 2021), but the term ‘passive participle’ for the Participle is still encountered in the literature. Designations such as ‘Perfect’ also occur, e.g. Mutzafi (2008a). 128 VERBAL MORPHOLOGY Table 22: Inflectional stems for all verbal Patterns in J. Dohok Pattern I: Pattern II: Pattern III: Pattern IV: p-θ-x (paθəx) š-d-r (mšadər) š-x-n (mašxən) g-n-d-r ‘open’ ‘send’ ‘warm up’ ‘roll down (a (transitive) stone)’ šaqəl paθǝx mšadǝr mašxǝn mgandǝr šqəlle pθǝxle mšodǝrre mušxǝnne mgundǝrre Imperative pθox mšadǝr mašxǝn mgúndǝra Participle pθixa mšudra múšxǝna9 mgandǝr Infinitive pθaxa mšadore mašxone mgandore The following sections present the specific inflections, exemplified by the Pattern I verb p-θ-x (paθəx) ‘open’. 9 In this form — as well as in the Participle in Pattern IV, in certain forms in šaqəl of Patterns III and IV and in the plural Imperatives of Patterns III and IV — the accent is pre-penultimate. This is due to the fact that ǝ is (originally) an epenthetic vowel. In other words, it is a purely phonetic vowel, invisible to accent. MORPHOLOGY 129 5.2.2. Šaqəl and the inflectional E-suffixes Historically, the 3rd person forms developed from forms of the active participle in the absolute state, while the 1st and 2nd originate in absolute state active participle forms with enclitic personal pronouns. Synchronically, all forms except 3MS have an inflectional suffix, referred to here as ‘E- suffixes.’10 A unique feature of Patterns III and IV is that the presence of an epenthetic throughout the inflections, thus in Pattern III: 3MS mašxən-Ø and 3FS mášxǝn-a, Pattern IV: 3MS mgandər-Ø and 3FS mgándǝr-a. 10 As in, for instance, Mutzafi (2008a) and Noorlander (2018), where ‘E’ refers to ‘enclitic’ (histor- ically). The terminology of these suffixes remains a matter of controversy in NENA; they are also known as ‘D- (‘direct’) suffixes, S- (‘subject’) suffixes (in Khan, e.g. 2008b) or ‘A- (‘agent’) suffixes’ (in Coghill 2003). In the view of this thesis’ author, none of these terms represent an accurate definition of these suffixes. Therefore, the employment of the term ‘E-suffixes’ in the present study is purely conventional, and does not represent a commitment to the terminology as a definition. 130 VERBAL MORPHOLOGY Table 23: The šaqəl inflection in J. Dohok 3MS paθəx-Ø 3PL paθx-i 3FS paθx-a 2MS paθx-ət 2PL paθx-etun 2MS paθx-at 1MS paθx-ən 1PL paθx-ax 1FS paθx-an 5.2.3. Šqəlle and the inflectional L-suffixes The šqəlle stem is inflected with the so-called L-suffixes, historically derived from inflected forms of the preposition l- ‘to, for, by’ (Khan 2008b, 165).11 If preceded by an alveolar sonorant, that is, r or n, l- of the suffix assimilates thereto, thus *mǝr-la > mǝr-ra PERF.say-she ‘she said’. 11 See Noorlander (2018, throughout) for a detailed study of the function of the L-sufixes across NENA. MORPHOLOGY 131 Table 24: The šqəlle inflection in J. Dohok 3MS pθǝx-le 3PL pθəx-lu 3FS pθǝx-la 2MS pθǝxl-ox 2PL pθə́x-loxun 2MS pθǝx-lax 1MS pθǝx-li 1PL pθəx-lu 5.2.4. The Participle The FS forms of Patterns II-IV have the -a- vowel in the syllable before the inflec- tional ending, e.g. Pattern II mšudar-ta. The participle is inflected with the nominal (adjectival) endings (cf. §4.2). Table 25: The resultative Participle (šqila) inflection in J. Dohok MS pθix-a PL pθix-e FS pθǝx-ta 132 VERBAL MORPHOLOGY 5.2.5. Imperative Table 26: The Imperative (šqul) inflection in J. Dohok SG pθox-Ø [pθʊx]/pθōx-Ø [pθoːx] with L-suffix pθóx-Ø- PL pθux-un/pθux-u with L-suffix pθúx-u- The singular imperative form in J. Dohok is in careful speech and in pause posi- tions realised pθōx [pθoːx], while pθox [pθʊx] occurs elsewhere. Since the careful form clearly has the vowel [o], the fast form is also analysed as having the vowel o on the underlying level. As shown in the phonology chapter (§1.3.3.4), the short allophone of /o/ has realisations which vary in quality and include alongside [o] also [ʊ] attested in the imperative. The form with long [oː] has a super-heavy, tri-moraic phonetic syllable /CVVC/. By contrast, the NENA stressed syllable is typically bimoraic: /CVC/ or /CVV/. This heavy syllable may have been created under a pressure to satisfy the prosodic minimality constraint of disyllabicity, that is, a demand for words—at least the MORPHOLOGY 133 prosodically independent words—to have minimally two syllables.12 Assuming the validity of this constraint for NENA, imperative forms such as pθōx are likely lengthened to make them disyllabic, if only on the phonetic level. Since tri-moraic syllables as a rule do not occur in NENA, the form pθōx should be considered disyllabic: [pθo.ox].13 A similar pressure of disyllabicity may be responsible for the imperative forms in the Trans-Zab dialects such as J. Sanandaj. There, along- side the historical form gruš, gə́ruš also occurs for Pattern I transitive and intransi- tive verbs (Khan 2009, 79). The vowel u in the imperative stem may be due to analogy with the u in the in- flectional ending. A similar form is attested elsewhere in LD, though for J. Challa, two allomorphs are given: ptox-un and ptux-u-le, suggesting o>u|suffix (Fassberg 2010, 112.) In the plural forms without a suffix in all verbal patterns, two allomorphs occur in the corpus, pθuxu and pθuxun. In the existing LD grammars, only the latter, more conservative one is given, though both likely occur in natural speech. 12 This constraint has been proposed for NENA by Hoberman (2006). 13 This analysis is parallel to Khan’s treatment of super-heavy syllables in Tiberian Hebrew, in which /CVVC/ is phonetically [CV.VC] (Khan 2013). Cf. also Kiparsky's notion of semisyllables applied to Arabic dialects (2003, 152–54). 134 VERBAL MORPHOLOGY Weak verbs in Pattern I In the existing grammars of LD dialects, šaqəl and the k-šaqəl are listed under the same stem, and are said to differ only in the presence or absence of a TAM preverb (e.g. J. Challa - Fassberg 2010, 106).14 However, in Pattern I of historical primae /ʾ/, and in the verb y-ʾ-y (yaʾe) ‘know’, k-šaqəl has synchronically based on a dis- tinct stem. This separation is due to various morphophonological processes which applied to k-šaqəl and is apparent in the weak and irregular forms presented in Table 27 below. By contrast, other TAM paradigms (with various preverbs) still share their stems and inflectional suffixes with the corresponding šaqəl paradigms. 14 Note that a similar process of separation of the Present base from šaqəl has taken place in the Jewish Trans-Zab dialects — see, for example, (G Khan 1999, 96) MORPHOLOGY 135 Table 27: Šaqəl and k-šaqəl stems in historical primae /ʾ/ verbs in J. Dohok šaqəl k-šaqəl Gloss Verb class l-w-š lawəš glawəš ‘wear’ strong Ø-x-l ʾaxəl kexəl ‘eat’ weak: I-ʾ Ø-m-r ʾamər gemər ‘say’ weak: I-ʾ y-ʾ-y yaʾe kiʾe ‘know’ irregular 5.3.1. Class I of verba primae /Ø/: ʾamər ‘say’ and ʾaxəl ‘eat’ This class includes all roots with /Ø/ as the the C1 and corresponding to /ʾ/ in earlier Aramaic. Such verbs conform to a single derivational pattern, except for the k-šaqəl inflection, which subdivides into two further sub-classes, and for the Imperative of ‘say’, whose stem is unique. As argued in the phonological section, historical /ʾ/ has been lost in many Iraqi NENA varieties, and the word-initial [ʾ] in places which historically had the segment /ʾ/ is simply a phonetic syllable onset 136 VERBAL MORPHOLOGY in cases with word-initial vowels.15 This also holds for this class (thus, ʾamər vs mərre), except for the Infinitive (ʾĭmara), which cannot be predicted from a root such as Ø-m-r-. However, this Class also includes some high-frequency verbs with a historical ʿ (‘do’ and ‘pass’), whose morphology arose in analogy with verbs with a historical ʾ. /Ø/ as a possible radical segment is also postulated in Khan’s analysis of C. Urmi (2016, 287). This is done to describe verbs whose (vocalic) pattern is canonical or almost canonical and could thus in principle be based on a root and pattern, but one of the radical slots is empty. šaqəl ʾamǝr k-šaqəl 3MS g-emǝr k-exǝl 3FS g-ǝmra k-ǝxla The k-šaqəl of this class has the form keC2ǝC3/kəC2C3- which contrasts with kC1aC2əC3 in other Pattern I verbs. This pattern is especially characteristic of Jew- ish dialects, occurring in all the documented LD varieties and in Early J. Nerwa 15 For consistency’s sake, verbs from other dialects with a historical glottal stop as C1 are also considered here verba primae /Ø/, though in the source grammars, /ʾ/ is typically given as the initial radical. MORPHOLOGY 137 (e.g. k-emər-Ø REAL-say-he ‘he says’; data from Sabar 1976, XXXIX). It is also found in the transitional J. Barzani cluster, e.g. Ø-ʾazəl-Ø IRR-go-he ‘he may go’ vs g-ezəl- Ø REAL-go-he ‘he goes’ (Mutzafi 2002a, 61). In the TZ dialects, the 3MS form lacks a vowel after the preverb, but forms with inflectional suffixes have an i vowel in those environments. This morphology can be illustrated for J. Sanandaj by the k- šaqəl forms for ‘eat’: 3MS k-xəl, 3FS k-əxlá [kəx.la] (Khan, 2009, p. 99). That [ə] represents in the inflected forms the historical short allophone of /e/ rather than simply being an epenthetic is suggested by the irregular verb ‘go’ *ʾ-z-l, in which hezəl (šaqəl) has apparently been formed in analogy to g-ezəl. The 3MS form of both paradigms have the vowel [e]: g-ezəl-Ø REAL-go-he ‘he goes’ > Ø-hezəl-Ø IRR-go- he. A high front vowel i or e in (historically) primae /ʾ/ verbs is also found in Christian dialects; to the author’s knowledge, it occurs in C. Alqosh, e.g. k-ixəl-Ø REAL-eat- he ‘he eats’ and k-əxl-a REAL-eat-she ‘she eats’ (Coghill 2003, 144), though not in the other Mosul Plain dialects such as C. Qaraqosh (cf. Khan 2002, 100). Christian dialects with e or i in the 3MS form from the core NENA territory (NW Iraq) include C. Təlla: k-iθ-e REAL-come-he ‘he comes’ and the ʿAqra region, e.g. k-emər-Ø (data from Al-Zebari 2018, 150). 138 VERBAL MORPHOLOGY It is difficult to find a convincing explanation for the origin of this pattern based on the synchronic data; The shift a >e is certainly not characteristic of NENA,16 though it occurred in earlier Aramaic after a glottal stop. Coupled with the reality that a similar form occurs in dialects which are neither geographically nor com- munally close to LD such as C. Alqosh, this situation suggests that this pattern represents a conservatism. It is likely that the change of the stem vowel i/e to a took place in analogy with the vowel a in strong verbs. It is furthermore possible that this change began in the Christian dialects of the core NENA territory, but did not spread to dialects which were communally and geographically more ‘remote’, such as Jewish and Mosul plain varieties. Mutzafi (2016 and 2018) presents a series of other features which are shared across Jewish dialects, often to the ex- clusion of Christian varieties, showing that communal affiliation can both catalyse and block a dialectological spread. Cases such as the verbal class in question could thus show that geographical and communal ‘distance’ can have similar effects. One could reconstruct this verbal pattern in the following way: kʾaxǝl > kyaxǝl > kexǝl, though ya does not shift to e in other environments, including other stressed positions (e.g. in g-yasəq). It is also noteworthy that a few dialects which have the present realis preverb y- (a variant of ʾi; cf. §8.12) preserve this preverb only in (historical) primae /ʾ/ verbs. This is illustrated for C. Marga (SE Turkey): 16 This is the explanation given by Mutzafi, but I am not aware of such shifts elsewhere in LD. MORPHOLOGY 139 yaxəl (ʾ/Ø-x-l) ‘he eats’ garəš (g-r-š) ‘he pulls’. The distribution of the allomorphs g vs k in J. Dohok points to anticipatory conso- nant harmony in voicing: g- occurs where the second consonant of the verb is voiced, while k- is found when that consonant is devoiced:17 g-emǝr (<ʾ-m-r) ‘he says’ k-exǝl (<ʾ-x-l) ‘he eats’ g-ewǝð (<ʾ-w-ð< ʿ-w-ð) ‘he does’ k-eθe (<ʾ-θ-y) ‘he comes’ g-ewǝr (<ʾ-w-r < ʿ-w-r) ’he passes’ This is also attested in J. Amediya (cf. data in Greenblatt 2008), while J. Betanure has developed voicing in all forms (e.g. g-emər and g-exəl). In J. Challa, both allo- morphs of k- occur, but cannot be attributed to consonant harmony (e.g. k-emər, but g-ezəl). Assuming /Ø/ as the first root radical in these LD verbs, the situation in J. Dohok is in fact more consistent with the synchronic system. By contrast, the forms with k- in dialects such as J. Dohok and J. Amediya constitute a relic of the stage when the preverb k- was still followed by the voiceless [ʾ]. šqəlle 17 For other cases of long-distance consonant harmony in NENA, see (Khan 2016, 116–7). 140 VERBAL MORPHOLOGY mǝr-re xǝl-le Imperative mar xōl măr-u(n) xol-u(n) I am not aware of many other NENA dialects which preserve the historical the- matic a vowel in the imperative of Ø-m-r < ʾ-m-r ‘say’, as is the case in the Jewish LD and TZ dialects (cf. J. Sanandaj - Khan 2009, 100). In many Christian dialects, this vowel has shifted to o in analogy to other classes, thus mur/mor say (SG)!’ Infinitive ʾǐxala Participle xila 5.3.2. Class II of verba primae /ʾ/ < */ʿ/: ʾarəq ‘run’ This class comprises some verbs with an original */ʿ/ as C1, now /ʾ/. It is quasi- strong; the only deviation from strong verbs is the insertion of an epenthesis [a] after the radical /ʾ/ when /ʾ/ occurs before a consonant, for instance ʾăraqa MORPHOLOGY 141 ‘running’ (Infinitive).18 The quality of this epenthetic is also unusual – [a], rather than [ə]. This is a relic of the stage when /ʾ/ was /ʿ/ and caused the lowering of the adjacent vowel. The phonetic character of the epenthetic [a] vowel is apparent from accent place- ment; for instance, the SG Imperative form ʾăróq has stress on the second, rather than the first, epenthetic vowel. Since stress in the Imperatives falls on the first syllable, the stress placement in ʾăróq indicates that the underlying form of this word is [ʾroq]. The phonetic character of such [a] is indicated in the transcription using the breve sign, thus, ă, not a. k-šaqəl g-ʾarǝq Note that the voiced allomorph g- is used, despite the fact that synchronically, the realis preverb is followed by the voiceless [ʾ]. Like the epenthetic ă, this voiced allomorph is a relic of an earlier stage – one in which the prefix was still followed by the voiced [ʿ]. 18 In the same class in Pattern III, this epenthetic assimilates in quality to the main vowel of the syllable which ʾ closes. In Pattern I, however, ʾ is always word-initial, so this rule does not apply. 142 VERBAL MORPHOLOGY 5.3.3. Verba primae /y/: yasəq ‘go up’ The conjugation of primae /y/ verbs is quasi-strong, the only deviation from the strong paradigm being an epenthetic i between y and the immediately following consonant (i.e. in the Past, Imperative and Infinitive), e.g. yǐsǝqle he went up’. Both J. Challa and J. Betanure apparently possess both a strong and a weak class (respectively, Fassberg 2010, 130, Mutzafi 2008a). In J. Zakho, the ‘weak’ radical seems to be preserved. 5.3.4. Verba mediae /y/, e.g. kayəp ‘bend’ šaqəl 3MS kayǝp 3FS kepa šqəlle kǝple Imperative SG kop PL kopu(n) MORPHOLOGY 143 Participle kipa Infinitive kyapa 5.3.5. Verba mediae /w/: kawəš ‘go down’ šaqəl 3MS kawǝš 3FS koša 3PL koši As in verba mediae /y/, when there is a diphthong, it contracts, in this case, aw > o. Consequently, the radical /w/ is only preserved in the 3MS form. šqəlle kušle In J. Dohok, the paradigm is weak – the original diphthong *wǝ has contracted to u. This is parallel to J. Challa, where the diphthong has contracted to ü (Fassberg 2010, 132), while J. Amediya and J. Betanure preserve the strong paradigm (re- spectively, Greenblatt 2008, 192 and Mutzafi 2008a, 72). 144 VERBAL MORPHOLOGY Imperative J. Dohok & J. Challa19 J. Betanure and J. Amediya20 SG koš kwoš PL košu(n) kwušun Participle kwiša Infinitive kwaša 5.3.6. Verba tertiae /ʾ/: šameʾ ‘hear’ Some verbs with /ʾ/ as the (historical) final radical are conjugated as verba tertiae /ʾ/, while other ones are shifting by analogy to the morphology of tertiae /y/.21 An example of the former case is š-m-ʾ ‘hear’. The root q-ṭ-ʾ ‘cut’, by contrast, is 19 Fassberg 2010, 132. 20 Respectively, Mutzafi (2008a, 72) and Greenblatt (2008, 189). 21 ‘Analogical change’ is used here largely in parallel to ‘reanalysis’, but the latter term is sometimes used specifically for root reanalysis in the Semitic literature. The fact that a verbal paradigm is synchronically derivable from a root, however, can simply be an outcome of analogical change, rather than its mechanism, so that reanalysis does not necessarily imply the existence of roots in cognitive processing. In the present work, ‘reanalysis’ thus refers primarily to classes and patterns. MORPHOLOGY 145 conjugated in the corpus both as a tertiae /y/ and tertiae /ʾ/, while also occurring in hybrid forms. For instance, the attested 3PL šaqəl forms include qaṭe [qaːtˤe] (tertiae /y/), qaṭʾi [qatˤʾi] (tertiae /ʾ/) and qaṭi [qaːtˤi].22 Verba tertiae /ʾ/ are con- jugated as follows. šaqəl 3MS šameʾ realised [ʃaːme], but šāmeʾ-li [ʃaːme(ʾ)li], not [ʃaːmeːli] 3FS šamʾa 3PL šamʾi In the inflected forms, /ʾ/ tends to be preserved in a careful register such as the grammar questionnaire, but is usually elided in free speech. In the 3MS form, it is never realised, except before a suffix. In all forms, however, /ʾ/ is part of the un- derlying form, and is therefore indicated in transcription. This is apparent from the fact that the first vowel of the inflected forms always remains short, e.g. 3PL šamʾi or šămi, but never *šami (i.e. [ʃaːmi]). šqəlle šmeʾle 22 Such processes of reanalysis are found across dialects, but dialects differ in the verbs and inflec- tions that are targeted; cf. the Participle form of š-m-ʾ ‘hear’ in J. Challa, reanalysed as verbum tertiae /y/: šǝmya ‘heard (MS)’ (Fassberg 2010, 135) 146 VERBAL MORPHOLOGY As in the inflection of šameʾ, ʾ is sometimes elided, but this does not result in the lengthening of the vowel which precedes it. For the background of the stem vowel e in this paradigm, see §1.3.3.3. Imperative SG šmoʾ PL šmoʾun Infinitive šmaʾa 5.3.7. Verba tertiae /y/: maxe ‘hit’ šaqəl In this paradigm, the third radical /y/ is preserved in the feminine singular forms where it is followed by the low vowel a. In the remaining forms, it was followed by e (or a similar, high vowel), and therefore lost because of the vowel’s articula- tory proximity to [y]. 3MS maxe 3FS maxya 3PL maxe 2MS maxǝt MORPHOLOGY 147 2FS maxyat 2PL maxetun 1MS maxǝn 1FS maxan 1PL maxax šqəlle mxele Imperative MS mxi FS mxe PL mxuwu(n)23 J. Betanure and J. Amediya agree with J. Dohok in the plural form (cf. respectively Mutzafi 2008a, 73 and Greenblatt 2008, 202) , while J. Challa appears to have the contracted form mxūn in most verbs (Fassberg 2010, 137). J. Zakho seems to pre- serve the plural imperative tertiae weak verbs C1C2aC3w attested in Late Aramaic 23 The form given for J. Dohok in Mutzafi (2002b, 481) is C1C2awun, e.g. štaw-un IMP.drink-PL ‘drink (PL)!’, but such a form was most likely produced under the interference of J. Zakho – the dominant LD dialect in the Jerusalem area today. 148 VERBAL MORPHOLOGY (e.g. Syriac; cf. Muraoka 2005, 107), but added to it another plural marker; e.g. mxawun. Participle MS məxya FS mxiθa PL məxye Infinitive mxaya Weak verbs in Pattern II 5.4.1. Verba mediae /y/: mhayər ‘dare’ This inflection is almost strong with a predictable monophthongisation of ay to e, e.g. Ø-mhayər-Ø IRR-dare-he ‘he may dare’ vs mhera IRR-dare-she ‘she may dare.’ An unusual feature is the first stem vowel in šqəlle – u, rather than o, e.g. 3MS mhuyərre. MORPHOLOGY 149 5.4.2. Verba mediae /w/: mjawəb ‘answer’ Most of the inflections are strong with a predictable monophthongisation of aw to o, for instance mjawǝb-Ø IMP.answer-SG ‘answer (SG)!’ vs mjob-un IMP.answer-PL ‘an- swer (PL)!’ In the šqəlle paradigm, assimilation has taken place, thus: mjuwuble <*mjowəble. 5.4.3. Verbs with identical C2 and C3: t-l-l ‘get wet’, x-l-l ‘wash (one’s hands)’ In šaqəl and the Imperative, two different conjugational classes exist, but a single verb may in many cases be conjugated in either of the classes. šaqəl Class I Class II 3MS mxălǝl mtălǝl 3FS mxalla mtắlǝla 3PL mxalli mtắlǝli Class I is synchronically almost strong, except for the 3MS form. In the inflected forms where the two identical radicals are adjacent, one of them is sometimes elided, but this elision is purely phonetic, as it does not affect the length of the previous vowel. 150 VERBAL MORPHOLOGY The base form mxălǝl has a short first stem vowel is short, even though this vowel is long in the strong paradigm. In Class II, an epenthetic ǝ is inserted where the two identical radicals would have been adjacent. This epenthesis serves to prevent the elision of one of the two identical radicals and is apparently still phonetic, since it does not affect the length of the preceding vowel and lexical stress. The underlying form of both classes is thus arguably identical, for instance, 3PL /mC1áC2.C3i/. In Class II, the stem vowel a is also short in the 3MS form. In both classes, this shortness is apparently in analogy to the inflected forms, but such analogical shortening does not take place in strong verbs. In Class II, it leads to the creation of a phonetically uniform stem across the inflection. Such a uniform stem is not created through the vowel shortening in Class I, however, which cre- ates a suspicion that ă the 3MS form of Class I (mxălǝl) is in analogy with the shortness in Class II (mtălǝl). šqəlle 3MS mtolǝlle Imperative Class I Class II SG mxălǝl mtălǝl PL mxallun mtắlǝlun Participle MORPHOLOGY 151 MS mxulla mtúləla FS mxulalta mtulalta PL mxulle mtúləle Infinitive mtalole24 Weak verbs in Pattern III 5.5.1. Verba primae /Ø/: maxəl ‘feed’ The historical radical /ʾ/, which used to be syllable-final, has been elided through- out the paradigms and the vowel preceding it has been lengthened. This morphol- ogy is thus as expected for verbs with /Ø/ as the first radical and is widespread across NENA (cf. C. Barwar – Khan 2008b, 230-1). It exhibits a different shift, however, than is the case with another historically prima /ʾ/ Pattern III verb ʾ-θ- y ‘bring’ (cf. below), but the roots of this difference reach back into Middle 24 For J. Betanure, Mutzafi (2008a, 78) gives the form msammone in which, as he puts it, ‘gemina- tion applies’. It is not entirely clear what the origin of this gemination might be, however, because the morphological gemination of C2 in Patterns II-IV has been lost in LD) Moreover, the same term ‘gemination’ is also to explain forms such as mxulla, but in ths case, ‘gemination’ simply means the adjunction of C2 and C3, which is clearly not the case in the Infinitive. The gemination of C2 in Infinitive could be secondary. 152 VERBAL MORPHOLOGY Aramaic (cf. Syriac - Muraoka 2005, 49). There are several verbs in this class whose Pattern I counterpart is inflected as a prima /y/ verb, e.g. Pattern III Ø-s-q ‘bring up’ vs Pattern I y-s-q ‘go up’.25 šaqəl maxǝl šqəlle muxǝlle Imperative maxǝl Participle muxla Infinitive maxole 25 So also in J. Betanure (Mutzafi 2008a, 72). MORPHOLOGY 153 5.5.2. Verba primae/ʾ/ maʾărǝq ‘cause to flee’ This is a class of verbs with an original */ʿ/ and it is quasi-strong. Its only special feature is epenthesis after the glottal stop when this phoneme precedes another consonant. This epenthetic assimilates in quality to the vowel before the glottal stop.26 šaqəl maʾărǝq šqəlle muʾŭrǝqle Imperative SG maʾărǝq PL maʾărqun Participle múʾŭrəqa Infinitive maʾăroqe 26An epenthetic after a pharyngeal, in contrast, always has the quality [a], cf. verba primae /ʾ/ in Pattern I. 154 VERBAL MORPHOLOGY 5.5.3. Verba mediae /w/, e.g. makuš ‘bring down’ šaqəl 3MS makuš 3FS mắkuša 3PL mắkuši The diphthong *wǝ in this paradigm has contracted to u, but in the inflected forms, the stem vowel a still remains short. This shortness is a relic of a stage when the syllable was still closed, e.g. 3PL mak.wǝ.ši. *In the 3MS form, by contrast, the syl- lable has been reanalysed as open, i.e. *mak.wəš > ma.kuš, leading to the length- ening of a. šqəlle mokušle The first stem vowel of this class is o, rather than u. Imperative măkuš mắkušun The Imperative also preserves a short a despite the opening of the syllable, but in contrast to the šaqəl paradigm, this shortness even occurs in the base form; MORPHOLOGY 155 contrast the SG Imperative măkuš with the 3MS šaqəl makuš. In other words, the Imperative is apparently phonetically more conservative than the šaqəl paradigm. Participle MS mokuša FS mukwašta PL mokuše Infinitive makoše 5.5.4. Verbs primae /y/, e.g. mayʾəl ‘bring in’ This class is strong, but it is precisely the ‘strength’ of this class that makes it peculiar.27 In the šaqəl the Imperative paradigms, y is part of the diphthong ay, but this diphthong does not collapse e, as is the case elsewhere. Such diphthong preser- vation is likely due to the pressure to conform to the morphology of the pattern, for instance mayʾəl = maC1C2əC3 ≠meʾəl. However, such morphological pressure does not prevent diphthong contraction elsewhere in the verbal system. 27 In Mutzafi’s grammar of J. Betanure, this class is labelled ‘strong’. 156 VERBAL MORPHOLOGY Irregular verbs The group of irregular verbs includes lexemes which do not fit into the previously- discussed weak verb classes. In these verbs, the loss of several radicals, morpho- phonological processes and morphological reanalysis and suppletion have ob- scured the original morphological pattern. Comparative tables are given to show different possible resolutions of opaque morphology. Forms are introduced here by their meaning and their historical root. 5.6.1. ‘Want‘ ʾə́be (< *b-ʾ-y) 5.1.6.1.1. Overview of forms šaqəl 3MS ʾǝbe 3FS ʾǝba 3PL ʾǝbi As expected, b-šaqəl is based directly on šaqəl: b-ʾǝbe, realised [bǝbe] or [bʾəbe]. The voiced allomorph b- is used, even when this preverb is followed by the voice- less [ʾ]. k-šaqəl 3MS g-ǝbǝ negated form: la g-be MORPHOLOGY 157 3FS g-ǝba 3PL g-ǝbe 2MS g-ǝbǝt 2FS g-ǝbat 2PL g-ǝbetun 1MS g-ǝbǝn 1FS g-ǝban 1PL g-ǝbax šqəlle ʾǝbele Infinitive ʾǝbaya Imperative MS ʾǝbi FS ʾǝbe PL bʾuwun 5.1.6.1.2. Historical background Historically derived from the NENA root b-ʾ-y (< b-ʿ-y), the verb ‘want’ in J. Dohok preserves this historical root only in the plural Imperative form. This dialect ex- hibits the greatest degree of innovation within the LD group, and even surpasses 158 VERBAL MORPHOLOGY in its development towards stem-and-affix morphology the TZ varieties. The forms are compared for Jewish dialects of NENA in Table 28 below. As shown below, ʾəbe (šaqəl) in both J. Dohok and J. Zakho has been derived from g-əbe (k-šaqəl) by back-formation, so g-əbe is discussed first. Across LD, the g-əbe paradigm does not follow any existing pattern and must thus synchronically be considered underived stem (cf. Bybee 1985, 60–64), since it does not follow the morphology of strong or weak (primae /Ø/) Pattern I verbs. In Aramaic and be- yond, the realis present is indeed the most likely locus of morphological innova- tion in the verbal system (cf. Khan 2016, 382-9). Instead of the expected a (strong) or e (primae /Ø/) first stem vowel, this inflection—excluding the 2PL—has the short vowel ǝ, despite being open and accented.28 While ə is likely phonetic (ep- enthetic) in origin, it is now clearly part of the stem, since it receives stress. 28 In many cases, ǝ is the short allophone of /e/; cf. §1.3.3.2. The negated form is la gbe, which suggests that ǝ in gəbe is not obligatory, and therefore most probably phonetic in origin. MORPHOLOGY 159 Table 28: Inflection of the verb 'want' across Jewish dialects Dohok Zakho Betanure/29 Challa31 Barzani32 Sanandaj33 Amediya30 šaqəl ʾǝbe ʾǝbe baʾe baʾe ʾəbe ʾăbé (ʾăxə́l) k-šaqəl g-ǝbe g-ǝbe g-ǝbe g-ǝbe gbe g-be (k-xəl) šqəlle ʾǝbele ʾǝbele bʾele bʾele ʾbele ʾəbele (xille) Imperative ʾǝ́bi/ bʾi bʾi – ʾbi ʾə́bu/ buʾwun ʾə́bumu(n) (xul) Infinitive ʾǝbaya ʾǝbaya bʾaya – ʾba ʾaboé (ʾax- ole) 29 Mutzafi 2008a, 80. 30 Greenblatt 2008, 260–7. 31 Fassberg 2010, 142. 32 Mutzafi 2002a, 62. 33 Khan 2009, 133-4. In brackets, the morphology of primae /Ø/(primae /ʾ/ in the source grammar) is given. 160 VERBAL MORPHOLOGY In fact, the k-šaqəl inflection of ‘want’ has even its own unique set of inflectional suffixes, shared with the šaqəl and k-šaqəl paradigms of ‘know’ (yaʾe) and ‘be able’ (ʾəṃṣe). The 3MS form g-əbe has ending -e, which suggests that the verb follows the morphology of verba tertiae /y/ (C1aC2e), but the feminine singular suffixes do not support this theory, since *y has been elided there (e.g. 3FS *gǝbya > gǝba). The inflectional suffixes of ‘want’ overlap with those of the strong verb, except for the 3MS form, which has the suffix -e contrasting with -Ø of strong verbs. The origin of g-əbe is not entirely strong. For J. Betanure, Mutzafi (2008a, 80) proposes that g-əbe is historically derived from the resultative (‘passive’) rather than on the active participle, that is, on *bʾe rather than *baʾe, and seem to recon- struct as follows: bʾe > gbʾe > gǝbʾe > gǝbe. The Early J. Nerwa corpus has forms such as gibax ‘we want’ (data form Sabar 1976, 87), which would be compatible with this explanation. On the other hand, k-šaqəl is typically derived from šaqəl, and the šaqəl paradigms of ‘want’ in J. Betanure, J. Amediya and J. Challa—baʾe— are still based on the active, rather than on the resultative participle. This suggests that both paradigms were likely originally based on the active participle.34 In ad- dition, some dialects still preserve the morphology of the active participle even in the k-šaqəl paradigm; cf. the C. Qaraqoš form g-baʾe (Khan 2002, 120–21). 34 There is indeed one verb in LD whose k-šaqəl paradigm is based on the resultative, rather than active participle, but in that case, both k-šaqəl and šaqəl are easily shown to be derived from the resultative participle; J. Betanure has the form ʾəmṣe ( MORPHOLOGY 161 An alternative explanation offered here thus assumes that the k-šaqəl forms in LD are also based on the historical active participle, but underwent drastic phonetic processes uncommon in other environments. This is plausible, considering the high frequency of the verb ‘want’, especially the present realis paradigm. We can propose for instance: gbaʾe > vowel harmonisation: gbeʾe > loss of ʾ: gbe > pres- sure to lengthen a monosyllabic word: gǝbe. In this scenario, ǝ is inserted to lengthen a monosyllabic word- a tendency which is also attested in the Pattern I SG Imperative in LD, and perhaps also TZ (see above). Like parts of the J. Dohok and J. Zakho paradigms, the TZ varieties and the tran- sitional J. Barzani cluster have developed a stem-initial ʾ which behaves like a strong radical. In TZ and J. Barzani, however, this has taken place in all the in- flections, so one could almost propose a full reanalysis of the root to ʾ-b-y, in which ʾ is a strong radical, except that in J. Barzani, like in LD, the inflectional suffixes on Ø-/k-/b-/šaqəl do not match the class of tertiae /y/ verbs. a) šaqəl The noteworthy feature of the šaqəl paradigm in J. Dohok and in J. Zakho is that it is a derived stem, formed from k-šaqəl by back-formation: gǝbe has been reana- lysed as g+ʾǝbe, that is, g+šaqəl, which gave rise to the šaqəl form ʾə́be. The ʾə́be paradigm cannot be convincingly shown to be diachronically root-based, neither is its synchronic form made to conform to existing patterns and weak verb classes. Rather, it is directly based on a different stem, which is a feature of stem-and- affix, rather than root-and-pattern morphology (cf. Bybee 1985, 60–64; Shimron 162 VERBAL MORPHOLOGY 2003b). The situation in J. Dohok and J. Zakho contrast with the other LD as well as many other Christian dialects, which preserve šaqəl forms based on the histori- cal NENA root *b-ʾ-y, e.g. baʾe.35 b) b-šaqəl The paradigm b-ʾəbe in J. Dohok and J. Zakho is realised [bəbe] or [bʾəbe] and is also stem-based, being derived from the new šaqəl paradigm ʾəbe. This also pro- vides indirect evidence that ʾəbe is a stem derived from gəbe, since [ʾ] has been added as part of the stem, which is not the case in gəbe. c) šqəlle The ʾəbele paradigm in J. Dohok and in J. Zakho is once again innovated, while the other dialects preserve the historical NENA root *b-ʾ-y. segment ʾǝbe. The J. Zakho and J. Dohok paradigms are synchronically analysable as strong and com- ply with the canonical morphology, the radical slots being filled with the sequence ʾ-b-y. This is a new, synchronic root, however, and does not follow the inflection of either primae /Ø/ (<*/ʾ/) verbs, or of primae /ʿ/ (<*/ʿ/). In those cases, we would expect respectively bele and ʾăbele. A similar innovated root which begins with /ʾ/ but behaves like a strong root is found in TZ (e.g. J. Sanandaj and J. Arbel). 35 For C. Barwar and C. Urmi, see Khan (2008b and 2016, 313-4) respectively. MORPHOLOGY 163 d) Imperative and Infinitive The singular forms of Imperative in J. Dohok—in contrast to the other dialects and in parallel to TZ—also has a new, synchronic root analysable as ʾ-b-y, which nev- ertheless follows the strong pattern. A similar reanalysis has also taken place in the Infinitive of J. Dohok and J. Zakho. 5.6.2. ‘Know’ yaʾe (< *y-ð-ʾ) As shown in Table 29 below, the historical NENA root for the verb ‘to know’ y-ð-ʾ (< *y-d-ʿ) is in LD preserved in šqəlle, the Imperative and Infinitive, while in k- šaqəl and šaqəl, this historical root and pattern has been obscured by phonetic processes. Neither šaqəl nor k-šaqəl can be demonstrated to be synchronically root- derivable, nor do they comply with the morphology of existing weak classes. Dif- ferences across LD are minimal (but cf. the šqəlle form yʾele in J. Challa) and largely phonologically predictable. The particular šaqəl and k-šaqəl stems are—to my knowledge—unique to the Western Jewish dialects, including the transitional Bar- zani cluster, while most other NENA dialects follow the strong or particular weak inflections. The situation in TZ suggests an attempt to maintain a new, cross-par- adigmatic root ʾ-l-y, but one that follows the strong class, rather than primae /ʾ/ verbs. Only the k-šaqəl form is not derivable from the root ʾ-l-y. Table 29: Inflections of the verb ‘know’ across Jewish dialects 164 VERBAL MORPHOLOGY LD dialects in the Barzani Sanandaj respective order šaqəl all LD yaʾe yae (b-šaqəl b- ʾălé (ʾăxə́l) yae) k-šaqəl all LD kiʾe kiye k-ăé (k-xəl) šqəlle Dohok, Zakho, Be- yĭðeʾle, zeʾle, ðeʾle, ydile ʾlile (xille) tanure, Amediya, ydeʾle, yʾele Challa Imperative Dohok, Zakho, Be- yĭðōʾ, zō, ðōʾ, ydoʾ ydi ʾə́lu (xul) tanure, Amediya Infinitive Dohok, Zakho, Be- yĭðaʾa, yizaʾ, yðaʾa, yda ʾăloé (ʾaxole) tanure, Amediya ydaʾa MORPHOLOGY 165 šaqəl 3MS yaʾe 3FS yaʾa 3PL yaʾe This inflection has its own unique set of inflectional suffixes, shared with the verb ‘want’ and ‘be able’. It is synchronically best analysed as based on a concatenative stem yaʾ+inflectional suffix. The 3PL form with the suffix -e prevents us from sug- gesting a reanalysis into a new synchronic root *y-ʾ-Ø, in which case the 3PL would have been *yaʾi. Possible explanations for the origin of this paradigm involve the shift of postvo- calic ð to ʾ, or the complete elisionof the former. Such a weakening of ð (or its dialectal variants d and z) is uncommon in LD and NENA in general, though a different type of lenition of ð—to l—occurs in the TZ group. Interestingly, a par- allel elision of postvocalic ð in another high-frequency paradigm takes place in many Christian dialects in NW Iraq and NE Iran; there, the independent possessive (‘genitive’) particle *did-/dið- has shifted to diy- (e.g. C. Barwar – Khan 2008b, 147; C. Urmi – Khan 2015, 245). This shift, however, does not take place in LD (did- is preserved); conversely, the weakening of ð in the paradigm for ‘know’ does not take place in the Christian NENA dialects such as C. Barwar or C. Urmi. For the k-šaqəl paradigm of ‘know’ in LD, we can propose the following process in the 3MS: yaðeʾ > yaðe > yaʾe > yaʾe; and elsewhere (e.g. 3FS) yaðʾa > yaʾʾa > 166 VERBAL MORPHOLOGY yaʾa. The inflectional endings of this paradigm do not correspond to a single reg- ular weak class, though they are identical to those in the gəbe (k-šaqəl) paradigm of ‘want’. Synchronically, the inflectional endings of both are a fusion of mor- phemes from different classes. In the yaʾe paradigm, the 3PL is problematic, as its suffix -e is characteristic of verba tertiae /y/, while the other forms have the suf- fixes of verba tertiae /ʾ/. In Early Jewish Nerwa, both yaʾe and yaʾi still coexisted side-by-side (cf. Sabar 1976, 43, footnote 60). Both of these high-frequency verbs, ʾəbe and yaʾe, exhibit the same tendency to inflect by using a stem lacking an abstract pattern and a new set of person-number-gender affixes. As shown below, the Trans-Zab dialects are different from LD, but also Christian dialects where the inflection of ‘know’ are more predictable. In the TZ group, this verb is characterised by a metathesis of ð and ʾ, the historical C2 and C3 respec- tively, and the shift of ð to l. These processes produced the new root y-ʾ-l, e.g. J. Koy Sanjaq yaʾǝl (3MS) and yaʾ(l)ex (1PL) (Mutzafi 2004, 90). k-šaqəl 3MS k-iʾe negated form: la k-ye 3FS k-iʾa 3PL k-iʾe The unusual features of this paradigm are the voiceless allomorph k- before i, and the vowel i in the first syllable, unattested in any other class of Pattern I. The former is difficult to explain, especially since historically, the realis preverb was MORPHOLOGY 167 always followed by a voiced segment, the first radical of the historical root being y. Perhaps it reflects a retention from before the voiced allomorph of the preverb, g-, even existed. The non-original vowel i in kiʾe corresponds to a of strong Pattern I verbs. It likely developed from a form such as *kiyaʾe in which iya > i, perhaps in a process similar to that which produced ke-/ge- in verba primae /Ø/. šqəlle 3ms yĭðeʾle Table 29 above has shown that while most LD dialects including J. Dohok preserve the historical root, J. Challa has the innovated past stem yʾe+L-suffix. This inflec- tion resembles the šaqəl and k-šaqəl inflection in having [ʾ] and lacking [d] (i.e. J. Dohok [ð]), but whether this is a case of weakening of d or an attempt to create a new, cross-paradigmatic root is not certain. Imperative SG yĭðōʾ PL yĭðoʾun Infinitive yĭðaʾa 168 VERBAL MORPHOLOGY 5.6.3. ‘Give’ yawəl (< *y-h-w) The historical NENA root for the verb ‘to give’ is y-h-w (< *y-h-b). The segment /l/ which has been used to re-fill the C3 slot in some forms comes from the prepo- sition l- ‘for, to’ (Mutzafi 2008a, 81). The deeper history of this verb in Aramaic can be viewed in Fox and Fassberg (2020). A similar process has occurred in the verb ‘go’, e.g. Ø-ʾazǝl-Ø IRR-go-he ‘he may go’. The l in both lexemes has a similar distribution and arguably helps maintain morphological transparency where it would otherwise be at risk. As shown in Table 30 below, the differences between the LD varieties in the inflections of ‘give’ are found only in the Infinitive and in šqəlle. All paradigms in both LD and TZ comply with the morphology of existing classes, but different paradigms are not derivable from a single synchronic root. Any proposed new roots would have to include various combinations of the seg- ments y, w, l, h and Ø. Table 30: Inflections of the verb ‘give’ across Jewish dialects Dohok Zakho Betanure36 Amediya37 Challa38 Sanandaj 36 (Mutzafi 2008a). 37 (Greenblatt 2008, 280–7). 38 (Fassberg, 2010, p. 146) MORPHOLOGY 169 šaqəl yawǝl yawǝl yawǝl yawǝl yawǝl hăwə́l (doq < *dăwə́q) k- šaqəl gyawǝl gyawǝl gyawǝl gyawǝl gyawǝl k-wəl (Ø- dwəq) šqəlle hulle hulle hūle hule hūle hiwle (dwəqle) Impera- hal hal hal hal hal hol (duq) tive Infini- ʾihawa ʾihawa yhāwa hyawa - háwoé tive (dwaqá) šaqəl 3MS yawǝl 3FS yawa 3PL yawi The historical radical h has been lost in this paradigm. The synchronic forms can- not be shown to be derived either from the same root or the same stem due to the distinctiveness of the 3MS form. For the remaining forms, the synchronic root y-Ø- w could be proposed. šqəlle 170 VERBAL MORPHOLOGY 3MS hulle 3FS hulla 3PL hullu In J. Dohok and J. Zakho, hulle is analysable as derived from h-w-l, but in the other LD dialects, this verb ‘display(s) the radicals h-w only’,39 the gemination of l having been lost (i.e., hulle > hūle). Imperative SG hal PL halun Infinitive ʾihawa The Infinitive shows some variation across LD, including a metathesis in J. Amediya (hyawa). J. Betanure (yhāwa) is the only dialect to preserve the historical root y-h-w. 5.6.4. ‘Bring’ meθe (< *ʾ-θ-y) The verb ‘bring’ comes historically from the root ʾ-θ-y, inflected in Pattern III. In the synchronic forms in LD and across NENA, the m- of Pattern III is part of the 39 Ibid. MORPHOLOGY 171 stem in all inflections, as shown in Table 31 below. It does not follow any verbal classes and thus cannot be considered as derived from a single synchronic root, though the latter part of the stems nd the inflectional suffixes conform to the pat- terns of verba tertiae /y/. In J. Dohok, the šaqəl, k-šaqəl and šqəlle inflections are arguably based on a single synchronic stem followed by the expected realisations of the third radical /y/. That is, they are formed on məθ/y/ in combination with an inflectional suffix. Assuming such a hybrid stem-and-affix and root-and-pattern derivation helps explain the vowel ə after m in all of these paradigms. The Imper- ative in J. Dohok cannot be derived from such a stem; like in J. Betanure and Amediya, its unique morphology is due to the assimilation of the first vowel to the second (cf. Mutzafi 2008a, 82). Table 31: Inflections of the verb ‘bring’ across LD Dohok Zakho Betanure Amediya Challa šaqəl meθe mese meθe meθe mese məθya masya meθya mǝθya mesyat(-tu) šqəlle mǝθele mǝsele muθele moθele musele/mǝsele Imperative miθi mesi miθi meθi misi meθe mese meθe - mese 172 VERBAL MORPHOLOGY muθun, mèθū-L- mesun muθun meθun musu suffix šaqəl The reconstruction of this inflection is as follows: maʾθe > mayθe > meθe (cf. Mutzafi 2008a, 82, footnote 76). The J. Zakho feminine singular forms preserve the original quality a, perhaps reinforced through vowel harmony with the second a (cf. the Imperative below). MORPHOLOGY 173 šqəlle The more original vowel o/u (from *muʾθele) is preserved in J. Betanure and Amediya, and partly in J. Challa. 5.6.5. ‘Be able’ ʾəṃṣe (ṃ-ṣ-y) Ṃ-ṣ-y is the only certain case of a verb whose šaqəl inflection and its derivatives k-šaqəl and b-šaqəl are based historically on the resultative, rather than active par- ticiple (though see above for the discussion of the verb ‘want’). It is derived from the resultative participle in the absolute state, that is, šqil.40 The historical root of this verb is m-ṣ-ʾ. Though the radical ʾ is not preserved in any of the forms, the inflectional endings for šaqəl are as expected for tertiae /ʾ/ verbs, except for the 3PL form where *ʾəṃṣi would have been expected. It thus shares its own unique set of suffixes with the verbs ‘want’ and ‘know.’ Not all dialects have a šqil-based šaqəl paradigm as is the case in LD however; the historical active participle is used in, for instance, C. Barwar (Khan 2008b, 251). The [ə] in g-əṃṣe is most likely an epenthetic vowel breaking the word-initial CCC cluster. This vowel is preserved in ʾəṃṣe where there is no need for such epenthe- sis, suggesting that šaqəl (and by extension b-šaqəl) have been back-formed from k-šaqəl, as is the case in the verb ʾəbe ‘want’. In J. Betanure, curiously, there is no 40 Thus, the designation šaqəl is suitable here only for reference to the prototypical function, not the morphology, for which šqil would be more suitable. 174 VERBAL MORPHOLOGY dedicated realis form, so that ʾəṃṣe is used for both realis present and non-past irrealis. The fact that the form is used, however, despite the absence of the k-šaqəl form g-əṃṣe in this dialect, could suggest that [ə] is historically not an epenthetic from g-əṃṣe, but rather a device to make the form disyllabic (cf. §5.2 above). šaqəl ʾə́mṣe negated: la ṃṣe b-šaqəl b-ʾəṃṣe k-šaqəl 3MS g-əṃṣe 3FS g-əṃṣa 3PL g-əṃṣe 5.6.6. ‘Be born’ and ‘be’ hawe (*h-w-y) Verbs based on the historical root h-w-y can mean either ‘be’ or ‘be born’. In some inflections, these meanings are formally differentiated, and some use a different lexeme, as compared in Table 32. MORPHOLOGY 175 Table 32: Tense-aspect-modal forms of ‘be’ and ‘be born’ in J. Dohok ‘be’ gloss ‘be born’ gloss non-past hawe ‘he may be’ hawe ‘he may be irrealis born’ realis pre- k-(h)awe ‘he is habitu- k-hawe/k-awe ‘he is born’ sent ally’ =le (basic present cop- ula) ‘he is’ predictive p̮-awe ‘he will be’ p-hawe/p̮-awe ‘he will be future born’ present — — wele huya ‘he has been resultative born’ past wewa (basic past cop- ‘he was’ hwele (perfec- ‘he was born’ ula)/wele tive) ‘he was habitu- k-(h)awe-wa ally’ present — — wele bə-hwaya ‘he is being continuous born’ imperative hwi ‘be!’ hwi ‘be!’ 176 VERBAL MORPHOLOGY The inflections mostly exhibits the regular features of verba mediae /w/ and ter- tiae /y/, but the status of the (historical) C1 h is uncertain. The most prototypical TAM functions of forms from the historical root h-w-y and their associated forms for J. Dohok are summarised below; suppletive paradigms are underlined. The stems in other LD dialects are nearly identical; the grammars of J. Betanure and J. Amediya likewise list two b-šaqəl stems, pawe and bhawe/phawe (Mutzafi 2008a, 84; Greenblatt 2008, 291). In the meaning ‘to be’, the verb h-w-y is part of the copula paradigms, though the k-(h)awe and k-(h)awe-wa paradigms are reserved for the habitual aspect, for ex- ample: 1) 1 (informal conversation) ʾawa k-haw-e kǝs-li kud-rota he REAL-is-he by-me every-Friday ‘He is at mine every Friday.’ The (historical) C1 tends to be elided. This is most advanced in the b-šaqəl paradigm in the meaning of ‘be’, where h is never realised and. The elision of h is common in this verb is common in NENA and in dialects such as C.Urmi has been extended to all the paradigms, so that it is now possible to propose the new underlying root Ø-v-y (Khan 2016, 365–6). In LD, by contrast, h is only elided in paradigms with a preverb, meaning that derivation should be considered as based on different roots or stems. MORPHOLOGY 177 The future prefix in the meaning ‘be’ is voiceless and unaspirated, i.e. [p̮]. The lack of voicing is probably a relic of the time when the future morpheme was followed by the unvoiced h. The lack of aspiration is highly unusual, because NENA stops are typically aspirated before vowels (Khan 2008b, 31–2); aspiration is also expected before h, since the production of this consonant is accompanied by a strong burst of breath. The situation in other LD is unclear, since the presence or absence of aspiration in this verb is not indicated in the existing grammars. The spectrograms in Figures 3 and 4 below demonstrate the difference between two unvoiced stops: [p̮] in p̮awe, ‘he will be’, and ([kʰ] in k-awe, ‘he is’). The area highlighted in blue is represents the consonantal segment; a straight line means no aspiration, contrasting with oscillation indicative of aspiration. The unaspiration of /p/ could indeed be indicative of the reanalysis of the first radical of the b-šaqəl inflection as /ʾ/ from p+ʾawe, in which case aspiration of p would indeed not be expected. However, /ʾ/ is never realised in this paradigm and this verb does not pattern like verba primae /ʾ/ from */ʿ/, or like verba primae /Ø/ from */ʾ/ (see above), both of which employ the b- preverb. = h Figure 3: lack of aspiration in p̮awe [p aːwe] Figure 4: aspiration in kawe [k aːwe] 178 VERBAL MORPHOLOGY Like b-šaqəl, the k-šaqəl paradigm also uses the unvoiced allomorph of the preverb, but in contrast to b-šaqəl preserves the preverb’s aspiration, suggesting a different phonetic process or reanalysis. Towards a typology of verbal derivation in NENA The analysis of irregular and weak verbs has raised several points which challenge the traditional Semitic model of root-and-pattern derivation. At the same time, it demonstrated the continued existence of morphologically distinct conjugational classes whose inflections continue to be based on non-concatenative segments (cf. Khan 2016, 381). The following sections discuss these challenges as well as fea- tures which uphold a pattern-based theory of derivation in an attempt to describe the typology of NENA derivation and the diachronic cycles producing it. It should be noted that the aim here is primarily to identify the most accurate and useful analytical model, not a cognitive processing mechanism; this distinction is not always clear in the Semitic literature. As far as mental processing or storing is concerned, there is solid evidence from experiments with Modern Hebrew that weak verbs are processed as stems, rather than roots (Frost, Deutsch, and Forster 2000). At the same time, linguistic processes such as analogical changes towards particular verbal classes (aka ‘reanalysis’) discussed above are inevitably linked to speakers’ cognitive preferences about the language’s morphology In general, the existence of the distinct verbal Patterns and classes discussed above is remarkable (cf. Hoberman and Aronoff 2003). It is indicative of a conservatism MORPHOLOGY 179 of the NENA verbal morphology (cf. Matras 2009), which exists in the language alongside radical restructurings in areas such as morphosyntactic alignment (Coghill 2016). Such conservatism is furthermore maintained despite the deep his- torical contact with stem-deriving Iranian languages and with rigorously-agglu- tinating Azeri Turkish (Khan 2016, 381). The discussion above revealed no clear difference between TZ and LD in their tendency to preserve the non-concatenative derivation, contrary to other areas of grammar, where TZ is drastically more in- novative (verbal morphosyntax, the lexicon and word order). Within the LD group, J. Amediya and J. Betanure appear to exhibit the greatest tendency to preserve the consonantal segments of stems (‘the radicals’), though this does not necessarily mean the preservation of a greater root uniformity across the various paradigms of a single verb because of reanalysis processes. 5.7.1. Root and/or template opacity As has been shown here for the LD and pointed out by Khan for C. Urmi, the morphology of many NENA weak and irregular classes cannot be synchronically derived from a root and a pattern, because the phonetic processes that formed them are no longer in operation. In this regard, NENA is no different from both Classical and Modern Hebrew and Arabic, or even the derivation of Semitic verbs 180 VERBAL MORPHOLOGY of Maltese.41 In most such cases, however, NENA verbs still group into distinct classes based on the composition of their historical roots. This results in a situation where inflectional stems are still not arbitrary, but remain largely restricted to a set of possible vocalisations. In other words, there is (in most cases) a synchronic pattern, but not a synchronic root.42 It is nevertheless possible to postulate a root- based derivations in the language’s past, as lexemes with a comparable type of root have undergone comparable developments. In more extreme cases still, even the verbal pattern is unique; this occurs in cases where verbs undergo phonetic processes unattested in other, structurally similar verbs. In J. Dohok. this is the case k-šaqəl paradigms of ‘want’ and ‘know’, and arguably in all the inflections of ‘bring’, which pattern differently from other verbs historically with /ʾ/ as C1. The fact that the verbs for ‘want’ and ‘know’ are syn- chronically not derivable from a pattern is also apparent in their inflectional suf- fixes, which are not connected to a single class of weak verbs, and are thus ‘un- motivated’ from a derivational viewpoint. Such unique stems, however, are largely 41 Cf. Hoberman and Aronoff (2003), who seem to imply a difference between Maltese and ‘proto- typical’ Semitic, but it is unclear what makes them different. They state that ‘vowels of the Semitic Maltese theme templates (i.e. patterns), unlike those of Classical Arabic, are not fixed.’ Even in Classical Arabic, however, Pattern I (qatala) has several different vocalisations which are not pre- dictable. 42 This seems to be the definition of ‘templatic morphology’ in Hoberman and Aronoff (2003), but the definition of this term is not entirely clear in the publication. MORPHOLOGY 181 tolerated only in the most frequent paradigms.43 Contrast, for instance, the J. Dohok k-šaqəl paradigm g-əbe which does not conform to any morphological clas- ses, with the šqəlle form ʾəbele, which complies with the morphology of tertiae /y/ and has /ʾ/ as a strong C1. 5.7.2. Stem uniformity across a paradigm and the derivation of stems from stems A few verbal classes exhibit tendencies which are decidedly characteristic of stem- based languages (though not always incompatible with patterns, see below). One example is creation of a uniform paradigm across an inflection in LD are verbs with C1 and C2 in Pattern II, šaqəl, Class II. For instance, the forms mtălǝl (3MS; expected: mtaləl) mtắlǝla (3FS) share the stem mtălǝl. Comparable processes take place elsewhere in NENA.44 is relatively marginal and—to my knowledge—does not occur in strong verbal classes. There are also occasional cases of backformation in NENA, the most notable one of which being the creation of the new stem ʾəbe from g-əbe. Such derivation of one inflectional stem from another one is attested in languages with stem-and- 43 This has a cognitive underpinning, which is nevertheless out of the scope of this study (cf. Bybee 1985; Bybee and Hopper 2001, 10–7). 44 In TZ, verbs with w as C2 have also regularised the stem of the inflected forms to the 3MS form. In J. Arbel, for instance, the 3MS *dayə́q from the root d-y-q ‘beat, crush’ has shifted to deq because of forms such as deq-á (3FS) (Khan 2016, 381) 182 VERBAL MORPHOLOGY affix morphology (Bybee 1985, 50). Particularly productive source for new stems is the uninflected—typically the 3MS form—of the present paradigm. In J. Dohok, it is in fact possible to derive almost all of the inflections of ‘want’ from the single synchronic stem -ʾəb-, sometimes accompanied by TAM preverbs (cf. Table 28 above), though it is equally possible to postulate new synchronic roots. The same stem-based derivation can be argued for most of the paradigms of ‘bring’ and ‘be able to’ in J. Dohok (see above). There is furthermore an entire series of paradigms expressing particular TAM cat- egories which are themselves based on the šaqəl stem (cf. §8). This is a clear ex- ample of deriving new stems from existing ones, and follows the model of prefixing TAM morphemes while suffixing inflectional ones, which is also attested in Kurd- ish (cf. Stilo and Noorlander 2015). By contrast, NENA possesses no new inflec- tional stems that are demonstrably directly root-based.45 The new inflections with TAM preverbs, however, are still compatible with a Pattern-based model, because the stems from which they are derived (i.e. šaqəl) in themselves follow existing verbal classes. 45 Though, naturally, the inherited Pattern-based stems have undergone drastic changes in function and inflection. MORPHOLOGY 183 5.7.3. Analogical change (‘reanalysis’) of the pattern Yet alongside such processes of deriving new paradigms from existing stems we see clear pressures for adherence to existing verbal patterns. This is observed in some high-frequency verbs which exhibit cycles of change. First, certain forms undergo local, surface-phonetic changes, but these changes subsequently feed into a morphological reanalysis, being used to re-fill radical slots and thus create forms which comply with existing patterns. Such change is a cycle, because it uses move- ment away from canonical morphology—idiosyncratic phonetic change—in order to return back to it. In the material presented here, this is clearly observable with the verb ‘want’ in J. Dohok and in TZ. In the latter, the segment [ʾ] has been used 46 to re-fill the C1 slot in almost all the inflections. Arguably, the TZ dialects have undergone an almost full reanalysis yielding the new root ʾ-b-y. Another example of root reanalysis is the verb ‘give’, which can synchronically, this verb may be considered to be derived from a family of (historically related) roots;47 in J. Dohok, h-w-l (hulle), ʾ-h-w (ʾĭhawa) and y-w-l (yawəl) and y-Ø-w (yawa). A similar process is observable with the verb h-w-y ‘be’ and ‘be born’ and is most advanced in C. Urmi. There, the phonetic elision of h in non-word initial positions has been 46 The only remaining inconsistently being that the k-šaqəl paradigm still has a ‘weak’ /ʾ/ while in the other inflections, /ʾ/ behaves like a strong radical. 47 For the concept of ‘family of roots’, see Khan (2016, 376). 184 VERBAL MORPHOLOGY extended also to word initial ones, causing Ø to replace h in the C1 slot. A local phonetic change has given impetus to morphological reanalysis. It is arguably in this area that language contact fails to apply. If NENA were influ- enced by stem-based derivation, we would expect no such pressure to return back to the canonical morphology of patterns, as the language would tolerate stems with unique morphologies. Such process are indicative of a pressure of the morphological system. This, how- ever, is a pressure to maintain patterns and classes, rather than roots. Verbs such as ‘give’ require an entire synchronic root family, not a single root (see Khan 2016, 376, who demonstrates this for C. Urmi). However, it is no less efficient to list different roots for each inflection to listing various stems. This lack of efficiency is applied here to the linguist’s analysis, but almost certainly holds also for cogni- tive organisation. 5.7.4. Loanwords Another area which is highly telling for a language’s typology is loan morphology. Briefly, NENA is this regard decidedly more ‘Semitic’ than Maltese in that its loan verbs—even from the stem-based Kurdish—are also made to conform to existing verbal templates (Hoberman and Aronoff 2003). This reflects a clear endeavours to ‘mould’ loanwords into existing patterns, and inevitably entails a root ‘extrac- tion’. MORPHOLOGY 185 5.7.5. Summary The verbal morphology of NENA—and most likely also of other Semitic lan- guages—seems best characterised as pattern-based: there is a clear pressure to maintain adherence to existing vocalic templates. On the other hand, there is little pressure from any ‘underlying’, i.e. root-based, mechanism, because few verbs ex- hibit the tendency to create a single new synchronic root. This is consistent with Goldenberg’s typology, who considers the primary feature of Semitic the existence of radical slots within the various inflectional stems (Goldenberg 2013, 116–125). At the same time, a partial exception is attested in high-frequency verbs which— precisely because of their frequency—allow for the preservation of stems that do not conform to existing patterns. Whether or not this is more frequent than in other Semitic languages would have to be established through a systematic study. There are further arguments for a stem-based typology of NENA dialects such as C. and J. Urmi (Khan 2016, 381-9), but they are not relevant for the dialects dis- cussed here. CHAPTER 6 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS Introduction This chapter describes the mapping of grammatical roles of objects and related arguments (see below) onto morphosyntactic structures in J. Dohok, supplemented by comparative notes. It is primarily concerned with pronominal arguments, which in NENA are expressed with pronominal affixes on the verb or through independent prepositional pronouns. However, this discussion is expanded to include differential object marking, that is, the cross-referencing of a highly referential nominal (or pronominal) object by adding an object affix on the verb.1 Following Malchukov, Haspelmath and Comrie (2011), the present analysis distinguishes the following object roles:2 Patient (‘P’): the object of a monotransitive verb, like him in ‘I killed him’); Theme (‘T’): the directly affected or transferred object of a ditransitive verb, like it in ‘give it to me!’; 1 For referential prominence, see §9 and the references therein. 2 A full list of grammatical roles and their justification can be found in the chapter on word order (§9). 190 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS Recipient (‘R’): recipient and addressees of verbs classified here as ditransitive, like to me in ‘give it to me!’; beneficiary (‘BEN’): recipients and other beneficiaries of verbs not considered ditransitive (like for him in ‘make it for him’); beneficiaries are therefore not considered a ‘core argument’ like P, T and R. In general, the properties of P in NENA are representative of the broad category ‘object’ which subsumes also Themes and Recipients, and arguably even arguments such as beneficiaries which receive object-like encoding in NENA. In cases with no other pronominal object in the clause, any pronominal can be expressed with a pronominal (L-)suffix on the verb in most finite verb paradigms. However, it is Themes which are especially close to Patients; whether pronominal or nominal, a Theme can always be expressed in the same way like a (pronominal or nominal) Patient in NENA.3 Both arguments are therefore subsumed here under the grammatical relation ‘O(bject)’, as is the convention in the word order chapter for nominal and independent pronominal Themes and Patients (§9). Recipients, on the other hand, are discussed separately in order to show their partial difference from Themes and Patients. For instance, in J. Dohok, only Recipients can be expressed with the independent prepositional pronouns ṭa (‘to, for’)+L- suffix with any finite verbal form, and with an L-suffix on the šqəlle (past 3 For nominal objects and their position on the clause, see the chapter on word order (§9). MORPHOLOGY 191 perfective) base. The terms ‘Objects’ and ‘Recipients roughly correspond to the categories ‘direct objects’ and ‘indirect objects’ in some other NENA studies, respectively. In this chapter, the first section (§6.2) is devoted to ‘Objects’. The examples involve monotransitive clauses with Patients, but an identical encoding could have been selected for the Theme of a ditransitive construction. The subsequent section (§6.3) surveys ditransitive constructions, focusing in particular on the distinctive features of Recipients (as opposed to Objects), and on clauses with both a pronominal R and T. It is also shown that beneficiaries, experiencers and possessors may be expressed like Recipients in NENA (cf. Noorlander 2021). This is followed by a comparative survey of the competing past transitive constructions in LD and C. Barwar (§6.4) and by an overview of differential object marking (§6.5). The indexing of objects (Ts, Ps and Rs) and related arguments (e.g. beneficiaries) in NENA is sensitive to the stem and/or particular TAM category, and even to the person in paradigms such as šqil-a-le (internal objects in perfective past) and in ditransitive constructions based on the šaqəl stem. In general, NENA possesses a large set of strategies to express object and related arguments, especially in the perfective past (Khan 2017; Noorlander 2018), which is typologically striking. This, in turn, results in a diversity of morphosyntactic alignments – sometimes coexisting side-by-side within a single NENA dialect (Coghill 2016; Noorlander 192 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS 2018). Furthermore, the referencing strategies—especially for Recipients, beneficiaries and experiencers—are highly pertinent to the drastic alignment shifts which took place in Eastern Aramaic.4 Unanswered questions about the various object-expressing strategies remain, particularly relating to the pragmatic features influencing the distribution of the competing object-marking strategies. The answers to such questions will unveil some of the diachronic motivations for the rise of these competing strategies. Objects of monotransitive 6.2.1. Šaqəl-based forms except qam-šaqəl-la with L-suffixes In J. Dohok and throughout NENA, pronominal Objects on forms based on šaqəl are expressed with L-suffixes, as illustrated for J. Dohok below: k-šāqə̀l-Ø-le REAL-take-A.he-P.him ‘he takes him’ k-šaqəl-Ø-la REAL-take-A.he-P.her ‘he takes her’ k-šaqəl-Ø-lu REAL-take-A.he-P.them ‘he takes them' 4 For alignment from a diachronic perspective, see Coghill (2016); for a typology of NENA and Ṭuroyo alignment, see Noorlander (2018). MORPHOLOGY 193 An exception is the qam-šaqəl-la construction, discussed separately in the section on the perfective past below. 6.2.2. Past perfective forms Below, various strategies for object marking in past perfective are surveyed. In LD, using L-suffixes to express Objects—in contrast to Recipients—is not possible. 6.2.2.1. Qam-šaqəl-la with L-suffixes In J. Dohok, the most common and nearly exclusive way of expressing transitive perfective past with a pronominal or a differentiatively-marked nominal Object is with the qam-šaqəl-la construction. Its use is person-unrestricted, as illustrated in (1) below: (1) 1a/15 baxt-e qam-qaðr -à -le ʾud-la ʾĭxala basìma ṭa-le.ˈ wife-his PERF-host- A.she- P.him PERF.make-she food good for- him ‘He (the wolf) came, his wife hosted him and made delicious food for him.’ 194 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS Recently, Noorlander has argued that dialects with qam-šaqəl-la as the default past transitive construction (i.e. also J. Dohok) have an alignment in the perfective past which is ergative, or at least one that is not canonically accusative (see Noorlander 2018, 239). Qam-šaqəl-la is a recent arrival on the Aramaic scene – much more recent than the competing šqil-a-le construction (see below). According to Mengozzi (2012, 32), qam-šaqəl-la is rare in the 18th century Christian NENA poetry (Alqosh and Telkepe), occurring regularly only in 19th century texts. It should be remembered, however, that the Christian NENA poetry is archaising in nature. Qam-šaqəl-le is unattested in the Early J. Nerwa corpus, though has a possible predecessor in Jewish Babylonian Aramaic (cf. Fassberg 2015 and references therein). A proposed grammaticalisation pathway of this form can be viewed in Khan (2021). This situation suggests that this construction spread in the past four centuries, reaching a remarkably large number of NENA dialects. 6.2.2.2. Šqəlle with E-suffixes A different strategy for expressing Objects in the past is with E-suffixes infixed between the past perfective šqəl- base and an inflectional L-suffix. This is not possible with Recipients or beneficiaries in J. Dohok. Most NENA dialects including J. Dohok constrain this use of the E-suffixes in the case of 1st and 2nd MORPHOLOGY 195 person objects. The possible E-suffixes for objects on the šqəl- stem in J. Dohok are illustrated below, using the 3MS verbal form: šqəl-Ø-le PERF.take-P.him-A.he ‘he took/he took him’ šqəl -a-le PERF.take-P.her-A.he ‘he took her’ šqəl -i-le PERF.take-P.them-A.he ‘he took them’ These examples also show that the 3MS object is encoded with Ø. This is common in NENA, though exceptions exist (e.g. J. Arbel - Khan 1999, 119). Several (compatible) explanations of this object person restriction have been offered. In general terms, the 1st/2nd persons are prototypically more agent-like and thus pragmatically more topic-worthy, the 3rd persons being more closely associated with the patient (and theme) roles. Correlating these semantic/pragmatic associations with the person restrictions, this constraint can be formulated that ‘if the P(atient) references the highest-ranking person, it cannot be marked through an E-suffix’ (Noorlander 2018, 166). This, in turn, may be indicative of a greater association of the L-suffixes with the agent/topic role than is the case in the E-set. Khan’s explanation (2017, 880) suggests a similar function- form disconnect, but involves the concept of morphological markedness. It relies on the premise that E-suffixes are morphologically less ‘marked’ (more prototypically agent/topic-like?) than the L-suffixes, though the reason for considering them ‘less marked’ is not explicit in the publication. Using the 196 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS morphologically unmarked (i.e. E-)suffixes for the semantically marked function of first and second person patients is therefore a mismatch, and is disallowed.5 There are other (compatible) explanations for this restriction, which invoke diachrony and cross-system harmonisations (see Noorlander 2018, 161-8). Unrestricted object marking using the E-series only occurs in a handful of dialects, including the north-western Iraqi J. Amediya, J. Challa and J. Aradhin (all LD), and C. Ashitha and C. Marga in south-eastern Turkey, and the dialects in Urmia in north-western Iran (Khan 2017; Noorlander 2018, 161). It also occurs in the Early J. Nerwa corpus. In J. Dohok, šqil-a-le constructions with 3FS and 3PL objects could be elicited for most speakers and were accepted by all, but are virtually unattested in the corpus. The handful attested šqil-a-le forms with the 3FS E-suffix -a are used with intransitive verbs – to express a middle voice (see below). This is illustrated in (2) below with the verb ʾ-r-q (ʾarəq) ‘flee’, whose meaning in forms lacking the 3FS suffix is ‘run’. (2) 2/15 ʾărəq-a-li mən Hbet sèferH,ˈ pθəx-li dəkkàna,ˈ dəkkan-ət HnaʿalàyimH.ˈ 5 MORPHOLOGY 197 PST.run-P.her-S.I6 from school, PERF-open-I shop shop-GEN shoes. ‘I ran away from school and opened a shop, a shoe shop.’ Once, it is attested with a 3FS suffix anaphorically referencing a situation – a non- prominent referent: (3) 6/conversation Context: ‘There, Muslims were our neighbours.’ qṭiʾ-a-lan θe-lan ʾàxxa.ˈ PERF.cut-P.her-A.we PERF.come-we here. ‘We ended this (i.e. good relations with Muslims when) we came here.’ A handful of forms with E-suffixes was also produced during informal conversations, for instance: (4) 5/conversation Context: ‘Oh, so you have been to Dohok?’ Hʾaz betaḥH xəzy -a -lax rab-θa sqəl-ta.ˈ so sure PERF.see -P.her -A.youFS big-FS beautiful-FS. ‘So you must have seen it, (how) big and beautiful (it is).’ 6 This transitive-like parsing is adopted to illustrate the ‘transitive’ morphology of such constructions. 198 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS 6.2.2.3. Šqəlle with the independent prepositional pronoun ʾal- The independent prepositional pronouns ʾal- are not attested in the corpus, though a few such forms were produced during interviews, for instance grǝš-le ʾal-a PERF.pull-A.he P-her ‘he pulled her’. They are apparently associated with a very high register, and could perhaps constitute a relic from an earlier language stratum. This pronoun series is not mentioned for J. Betanure and J. Challa. In J. Amediya, this set is said to exist in a highly formal register, such as translations of Hebrew religious texts (Hoberman 1989, 101). 6.2.3. The Imperative with L-suffixes Pronominal objects on the Imperative are also expressed with L-suffixes, as illustrated in (6). The same strategy is available for Recipients. (6) 1 (interview) mắkuš -u -le IMP.bring_down -A.youPL -P.him ‘bring it down!’ MORPHOLOGY 199 6.2.4. Wele šqila (resultative) with ʾəll- Since the J. Dohok resultative construction wele šqila occurs in the corpus mostly with intransitive verbs, there are no attestations thereof with a pronominal object. Still, examples like the one in (6) below could be elicited. The object in such cases (including the Recipient) is expressed with the independent prepositional pronoun ʾəl+L-suffix, as illustrated in (6) below. (5) 1/interview wele šqìl -ə -lle.ˈ COP.DEIX.he RES.take -A.MS -P.him ‘He has borrowed it.’ A parallel strategy is attested in other LD dialects (e.g. J. Challa - Fassberg 2010, 123), with the exception of J. Zakho. The latter dialects makes use of the possessive pronoun series did- to express object on the Participle; the same object marking in J. Zakho is attested with bə-šqala, which expresses the continuous aspect (Cohen 2012, 143-4). (6) J. Zakho (Cohen 2012, 144) …ṭamaha we-lu mukim-e dida ʾay bažər. why PST.COP-they RES.blacken-PL P.her thisFS town ‘…why they had blackened the town.’ 200 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS As for the other non-finite verb construction in J. Dohok, the progressive wele bə- šqala, in the corpus, it is not attested with transitive verbs (and therefore with object suffixes), and this appears to be disallowed (see §8.15). Ditransitive and constructions with object-like arguments 6.3.1. Šaqəl-based forms except qam-šaqəl-la 6.3.2. Šaqəl with L-suffixes Recipients and beneficiaries can also be expressed with an L-suffix on the šaqəl base in J. Dohok. On ditransitive verbs and monotransitive verbs with a beneficiary, it is relatively common to express both arguments with an L-suffix, always in the order [V]-T-R ([V]-P-BEN) as illustrated in (7a) below. Such stacking of two object suffixes is disallowed in the majority of NENA dialects (e.g. C. Barwar - Khan 2008b, 289; Noorlander 2018, 123-4). In addition to the LD dialects such as J. Dohok and J. Zakho, stacking is also recorded for, inter alia, C. Hertevin on the north-western periphery of NENA (Noorlander 2018, 123-4). When the stacking of L-suffixes occurs, a person restriction applies in J. Dohok, as anticipated by Noorlander (2018, 147-8): the expression of the Recipient with an L-suffix is dispreferred with 1st or 2nd person Themes (7b). Such forms could not MORPHOLOGY 201 be elicited and were rejected by informants. In such cases, the independent prepositional pronouns ṭa- or ʾəll- must be used for the Recipient, as shown in (7c): (7) 1/interview a. b-yaw-ən -nu -lax FUT-give-IM -T.them -R.you ‘IM will give them to youFS.’ b. *b-yaw-ən -nax -lu FUT-give-IM -T.youFS -R.them c. b-yaw-ən -nax ṭalu/ʾəllu FUT-give-IM - T.you Rto.them ‘IM will give youFS to them.’ This constraint is likely due to a mismatch between the attested person and its prototypical role, heightened by the identity of the suffixes which could lead to interpretational difficulties. Prototypically, the 1st and 2nd persons are more Recipient-like while the 3rd persons are more Theme-like, but in cases such as (7b- c), the 1st/2nd person referent has the Theme role. An arguably parallel mismatch between a prototypical role (agent/topic) and the person in the construction (1st/2nd) is attested in the šqil-a-le paradigm. 202 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS 6.3.3. Šaqəl with the independent prepositional pronoun ṭa- (and ʾəll-?) As shown in (7c) above, the R can also be expressed with the independent prepositional pronouns; in the corpus, this is attested with the preposition ṭa-, but ʾəll- was also accepted by informants. Past perfective forms 6.4.1. Šqəlle with L-suffixes The LD dialects—in contrast to many other NENA varieties such as C. Barwar (Khan 2008b, 289–90)—allow the use of the L-suffixes for Recipients and beneficiaries also in the šqəlle paradigm.7 In such cases, the L-suffixes occurs in the order [V]-A-R/BEN, which is illustrated in (8) below. Rs contrast in this regard with Ts and Ps, which cannot be referenced with an L-suffix on šqəlle. (8) 3a/34 húl -lu -lan baṭane. PERF.give -A.they -R.us T.blankets ‘They gave us blankets.’ 7 This is also recorded for a few Christian varieties in regions adjacent to the Iraq-Turkey border (e.g. C. Marga - Noorlander 2021, 59). MORPHOLOGY 203 In such constructions, A (and S) aligns morphosyntactically with R, since both are marked with an L-suffix. Such coding has apparently arisen in analogy to other verbal paradigms, in which L-suffixes encode all types of objects as well as beneficiaries. Diachronically, however, it can simply be viewed as a retention of the original benefactive (‘dative’) meaning of l- (‘to, for’) – preserved despite the changes in the marking of subjects across the paradigms. Whether or not the use of L-suffixes for the R on šqəlle is lexically restricted across the dialects which allow it, is yet to be investigated. Mutzafi (2008a, 86) reports that in (the corpus of) J. Betanure, šqəlle+L-suffix is only attested with Ø-m-r (ʾamər) ‘say’ and *y-h-w (yawəl) ‘give’. In J. Dohok, by contrast, the stacking of L- suffixes on šqəlle is attested with a series of ditransitives (e.g. ‘say’, ‘give’, and ‘answer’), as well as with beneficiaries, e.g. ‘make for’, as shown in (9) below. In J. Dohok, therefore, Recipients and beneficiaries are characterised by the same coding throughout the paradigms. (9) 3b/29 ʾùd -le -li ʾədyo kurkamàna.ˈ PERF.make -A.he -BEN.me today golden ‘He has created (i.e. conjured up) for me today a golden coin.’ By contrast, in the TZ dialects and in some Christian dialects in Turkey and Iran, the set of ʾəll/L-suffixes is available for all objects in the past (Khan 1999, 119– 133; Noorlander 2018, 282–3). In dialects such as J. Urmi and J. Saqqiz, the 204 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS historical ʾəll- set formally overlaps with the L-suffixes, so synchronically, it is best to consider them a single set in these dialects (Noorlander 2018, 175–76). Such coding of Patients, Themes and Recipients as well as Subjects and Agents results in a completely neutral alignment in the perfective past (Noorlander 2018, 186- 7). Referencing all arguments with an L-suffix on šqəlle is also possible in the transitional J. Barzani cluster, as shown in (10) below. In this regard, therefore, J. Barzani is closer to TZ than to LD, though in areas of syntax such as word order, J. Barzani resembles LD more closely. This notwithstanding, it is likely that historically, these L-suffixes in J. Barzani come from the ʾəll- prepositions (ʾəl- in J. Barzani); cf. Mutzafi (2002a). (10) J. Barzani (Mutzafi 2002a, 53) gré -wā -le -le PERF.shave -PST -A.he -P.him ‘He had shaved him.’ 6.4.2. Šqəlle with the independent prepositional pronouns ṭa- (or ʾəll- and ʾəbb-) Recipients and beneficiaries in J. Dohok are commonly expressed with the prepositional pronoun ṭa-. ʾəll- and ʾəbb-, respectively ‘on, about, at’ and ‘in’, were not accepted by informants for most verbs, despite the fact that ʾəll- seemed acceptable in the šaqəl-based paradigms. In LD, ʾəll- and ʾəbb- are lexically MORPHOLOGY 205 restricted (Cohen 2012, 158-163); ʾəll-, for instance, occurs with the verb m-x-y (maxe) ‘hit’, as illustrated in (11): (11) 1c/90 Context: ‘Look at my wounded head…’ u-xzì-Ø,ˈ dukt-ət mxe-lox ʾō narà ʾəl-la.ˈ and-IMP.see-youSG place-GEN PERF.hit-youMS T.thisM axe R.on-her ‘…and look at the place which you hit with your axe (lit. the place you hit your axe on it).’ 6.4.3. Qam-šaqəl-la with L-suffixes In J. Dohok, Recipients and beneficiaries, like Themes, can be expressed with an L-suffix on the verbal stem, though with restrictions. A pronominal R can be referenced with an L-suffix as long as the T is also not (cross-referenced by a pronominal affix. In (12), for instance, the Theme is expressed nominally, so the Recipient can be referenced with an L-suffix. By contrast, if the Theme is (cross-) referenced with an L-suffix, the Recipient is as a general rule expressed with the prepositional pronoun ṭa+L-suffix, though an exception is shown in (14) below. These general rules indicate that the stacking of L-suffixes on qam-šaqəl-le is dispreferred. (12) 1c/57 măr-e beθa qam-maxze -Ø -le ʾurxa u-mpə̀q-le.ˈ 206 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS owner-GEN house PERF-show -A.he -R.him T.way and-PERF.leave- he ‘The father of the house showed him (i.e. to the wolf) the way, and (the wolf) left.’ Affixal cross-referencing of referentially prominent nominal Recipients on qam- šaqəl-la8 is attested in the corpus in direct object constructions that involve verbs of filling and naming (cf. §9.6.2), as illustrated in (13). In this group, the Recipient is the primary—i.e. agreement-triggering—object. (13) 1b/76 bət9 ʾana lʾine qam-ṃāḷè -Ø -lu ṭa-le zùze.ˈ through these jugs PERF-fill -A.he -R.them BEN.for-him T.money ‘He filled these jugs for him with money.’ Though the above-mentioned restriction on pronominal Recipients in the presence of a pronominal Theme is strong, it is not absolute. Of the 91 attested qam-šaqəl- la constructions, one (shown in (14) below) has two L-suffixes, the first one cross- referencing a nominal object and the second referencing a beneficiary. Such 8 This phenomenon is referred to in this thesis as differential object marking and object agreement; see §6.5 below. 9 This preposition is not expected here. MORPHOLOGY 207 stacking is considered to be ‘disfavoured or disallowed’ in NENA in general (Noorlander 2018, 231). (14) 3i/36 Context: ‘The wife left the woman alone. Then the wife said to her husband: “I told her to eat, but she didn’t answer me.”’ qam -šoq -an -ne -la ʾĭxàla.ˈ PERF -leave -IF -P.him -BEN.her food. ‘(So) I left the food for her (there).’ Beyond such verbs with beneficiaries and the above-mentioned double object constructions with a verbal agreement affix for a nominal R, there are few cases of L-suffixes on qam-šaqəl-la expressing Recipients. Thus, the high frequency ditransitive verbs Ø-m-r (ʾamər) ‘say’ and *y-h-w (yawəl) ‘give’ do not express a pronominal R with an L-suffix on qam-šaqəl-la; instead, šqəlle with an L-suffix or an independent pronoun is preferred (see above). Constructions like the one in (15) are unattested in the J. Dohok corpus, as is also the case in the J. Betanure and J. Amediya corpora in Mutzafi (2008a) and Greenblatt (2008). By contrast, expressing the R of ‘say’ or ‘give’ with an L-suffix on other inflections in J. Dohok is relatively common (see §6.3.1.1 above). (15) unattested *qam -amər -ra dət… PERF -say- A.he -R.her REL 208 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS ‘He told her that…’ Object (P, T and R) affixes on past perfective forms in LD in general In LD in general, there are two main strategies in the perfective past for the expression of pronominal Ps, Ts and Rs: one is with L-suffixes on qam-šaqəl-la, the other with E-suffixes on šqəl-, i.e. šqil-a-le.10 As shown above, there is another strategy available for Recipients only: the L-suffixes on šqəlle, i.e. šqəl-le-la. The LD dialects differ with regard to the use of qam-šaqəl-le and šqil-a-le, as shown in Table 33 below; C. Barwar is also included for comparison. Within LD, J. Dohok represents arguably the most advanced stage of development, internal E-suffixes having been almost completely lost. J. Betanure and J. Amediya retain a more productive šqil-a-le paradigm, but still—like J. Dohok—overwhelmingly prefer the qam-šaqəl-la construction.11 J. Challa represents a more conservative stage in that 10 The quantitative data in this section (Tables 33 and 34) are the outcome of my collaboration with Paul Noorlander and were presented in a joint paper at a workshop hosted by Freie Universität Berlin on 14th February 2020. 11 J. Zakho appears to be similar, but could not be investigated quantitatively due to the lack of an accessible corpus. MORPHOLOGY 209 it lacks qam-šaqəl-la altogether;12 šqil-a-le is available in this dialect for all persons of the object. Table 33: Frequencies of the two competing transitive constructions in LD and the middle voice construction (šqəlle with construction E-suffixes) qam- griš-a-le (including middle šaqəl-la voice usage) J. Challa 0% 100% (E-suffixes for all persons) J. Amediya 76% 24% (E-suffixes for all persons) J. Dohok 95% 5% (3rd person E-suffixes) J. Betanure 71% 29% (3rd person E-suffixes) 12 However, the extension of the E-suffixes to the 1st and 2nd person could itself be an innovation which did not take place in all dialects (cf. Noorlander forthcoming 2021), meaning that the dialects possessing this restriction (e.g. J. Dohok and J. Betanure) might perhaps be conservative in another sense. 210 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS C.Barwar 66% 34% (3rd person E-suffixes) Under the surface of these general tendencies, a few specific correlates of šqil-a-le have been identified. First, in dialects that use both constructions, a strong tendency to encode reflexive objects through the šqil-a-le construction is observable, as shown in Table 3413 and illustrated in (16) below. Table 34: Frequencies of constructions for encoding reflexive objects in LD šqil-a-le qam-šaqəl-la J. Amediya 70% 30% J. Betanure 100% 0% C. Barwar 100% 0% 13 These results remain preliminary; their statistical significance in the context of past object marking in general is yet to be confirmed. MORPHOLOGY 211 (16) J. Amediya (Greenblatt 2011, 324:311) gyana mtultiy -a -la (…)! self-her PERF.hang -P.her -A.she ‘She hanged herself (…)!’ Second, all the dialects listed in Table 33 above make exclusive use of the šqil-a-le construction with intransitive verbs such as ʾ-r-q (ʾăriq-a-le ‘he fled’) mentioned above.14 In J. Zakho, this is also attested with verbs such as pṣix-a-le PERF.rejoice- P.her-S.he ‘he rejoiced’ and gxik-a-le PERF.laugh-P.her-S.he ‘he laughed’. Such constructions have the (historical) morphosyntactic coding of transitive verbs— the 3FS E-suffix -a—but are semantically intransitive. Their special morphosyntactic coding can thus be considered a type of middle voice (cf. Mengozzi 2007), though this strategy is demonstrably lexically restricted. In the synchronic system of J. Dohok, the 3FS E-suffix on šqil- is in fact almost exclusively a middle marker, due to the all but complete loss of internal object marking in the perfective past. This loss, in turn, has almost entirely obscured the (likely) historical functional overlap of šqil-a-le and qam-šaqəl-la in J. Dohok.15 14 This, in turn, is a calque from Kurdish (Garbell 1965). 15 The association of šqil-a-le with reflexive objects and non-referential constituents in the middle voice construction could suggest a (historical) association of this construction with referentially non-prominent objects, but this has not been investigated further. 212 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS Differential object marking All NENA dialects possess strategies for differential object marking, that is, for encoding objects which are referentially prominent (Coghill 2014; Al-Saka 2019).16 The most common of these strategies—and one characteristic of LD—is known as direct object agreement, and involves the cross-referencing of the object through an agreement morpheme on the verb. This is illustrated for J. Dohok in example (17) below. Another strategy common cross-linguistically is direct object flagging, which involves a special marking on the object itself – by case or adposition (see Coghill 2014 and the references therein). In NENA, flagging is attested in dialects such as C. Barwar and in the TZ group (Coghill 2014; Noorlander 2018, 177-185), alone or in combination with verbal agreement. In LD, prepositional flagging does not occur. The above-discussed adpositional expression of pronominal Recipients and beneficiaries (e.g. with ṭa+L-suffix) is not differential. The distribution of differential object marking is likely dialect-specific in NENA, though it has been studied in detail only for a few dialects.17 Quantitative studies such as Multiple Correspondence Analysis (cf. §9.1.2) promise especially interesting results. While traditional ‘manual’ methodologies reveal individual 16 See Khan (1988) for differential object marking in Semitic languages in general. 17 Notable exceptions are Khan’s grammars and the other works mentioned in this section. MORPHOLOGY 213 agreement triggers, statistical studies can determine the strength of individual triggers and show which features ‘conspire together’ in making agreement more or less likely. The situation in J. Dohok is yet to be investigated quantitatively; the sections below present patterns observed in a brief survey. Existing NENA studies on differential object marking typically list definiteness as a key factor making agreement or flagging likely, but the distribution of definite vs indefinite objects is asymmetrical. Thus, in C. Telkepe, J. Zakho and C. Barwar, both definite and indefinite objects are found among those lacking differential marking (respectively, Coghill 2014, 342–45; Cohen 2012, 142 and Khan 2008b, 782). However, indefinite objects can also trigger agreement, as reported for C. Bariṭla by Al-Saka (2019, 84–101). This, in turn, is likely due to their discourse topical status, as argued by Khan (2008b). In Khan’s analysis, a ‘topical’ object is one which serves as a key discourse topic (i.e. is discourse prominent), rather than necessarily being the topic of the clause in question. Thus, ‘topical’ objects subsumes also discourse-new arguments. Factors rendering differential marking unlikely include narrow focus (listed for C. Telkepe, C. Bariṭla and C. Barwar) and the semantic integration of the object with its predicate (mentioned for C. Telkepe).18 For C. Barwar, inalienable possession is 18 E.g. objects like ‘weed’ in ‘to weed the weed’. 214 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS also said to ‘occasionally’ block differential marking (Khan 2008b, 774). Animacy, on the other hand, is not considered a meaningful factor in C. Telkepe (Coghill 2014, 342–45). In J. Dohok, the most apparent correlates of agreement are definiteness (as in J. Zakho, C. Barwar and C. Telkepe) and the pre-verbal position. Narrow focus and inalienable possessees do trigger agreement (with definite inalienable referents, this is very common), apparently in contrast to C. Barwar and C. Telkepe. Unlike in C. Telkepe, moreover, the lack of referential prominence does seem to make agreement less likely. 6.6.1. Word order As shown in the chapter on word order (§9.5.1.1), pre-verbal objects are significantly more likely to be differentiatively marked than post-verbal ones. The pre-verbal position in J. Dohok correlates with several pragmatic functions, and thus potentially with differentially-marked objects in various information- structural roles. Pre-verbal objects can be discourse-old and topical, topicalised and discourse-new, or carry narrow focus. In general, however, the clause initial or preverbal position in J. Dohok is characteristic of referentially prominent constituents and (discourse and clausal) topics (§9.4.1). MORPHOLOGY 215 6.6.2. Definiteness in animate objects Almost all of the differentiatively marked objects in the J. Dohok corpus are grammatically definite. This includes referents mentioned for the first time, as shown in (17) below. This points to definiteness as a key trigger of differential marking. (17) 3i/1 gòraˈ g-əb-é -wā -la ṛaḅa bàxt-e.ˈ man REAL-love-A.he -PST -P.her much wife-his ‘[A certain] man loved his wife very much.’ 6.6.3. Definiteness in inanimate objects Not all inanimate objects are differentially marked, even when definite (and discourse-old), as illustrated in (18) below. This suggests that low referential prominence of an argument makes its differentiative marking less likely. (18) 1c/21 Context: ‘The wolf said: “every day, I will give you a golden coin.”’ šqǝl-le lira kurkamana dide mǝn gùrga.ˈ PERF.take-he coin golden his from wolf ‘(So the man) took his golden coin from the wolf.’ 216 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS 6.6.4. (Grammatically) indefinite objects Cross-indexing on the verb is also attested with objects which are lack grammatical definiteness and are inanimate, though they are arguably semantically definite, because they represent a general category (here, ‘shoes’): (19) 2/28 Context: ‘In Dohok, I used to repair shoes.’ ṭamằ kundare Ø-mtakn -ən -wā -lu? why shoes IRR/FUT -A.IM -PST -P.them ‘Why did I use to repair shoes?’ 6.6.5. Inalienable (definite and inanimate) objects With inalienable possesses, agreement is common in J. Dohok as illustrated in (20) below: (20) 1a/3 u-là Ø-ʾamr -ǝ́tū -liˈ baθǝr reš-oxun Ø-mafer -ə̀n -ne.ˈ and-NEG IRR-say -A.youPL -R.me then heads-yourPL IRR-make_fly -A.IM -P.them ‘And (if) you don’t tell me, then I will cut off your heads.’ MORPHOLOGY 217 6.6.6. Objects in narrow focus The J. Dohok corpus includes cases of objects in narrow focus which are marked differentially. In the example in (21), the focus is on the numeral of the object: (21) 1d/4 Context: ‘So the Arabs colluded with the Nazis in killing the Jews.’ tre-ʾammà huðaye qam-qaṭl -i -lu go baġdad.ˈ three-hundred Jews PERF-kill -A.they -P.them in Baghdad ‘They killed two hundred Jews in Baghdad.’ Summary The most common strategies for referencing pronominal objects and for cross- referencing nominal ones in J. Dohok are L-suffixes on the verb, and the independent prepositional pronoun ṭa+L-suffix, used for Recipients. Marginal strategies include the ʾəll- series and the internal E-suffixes on šqəlle. A similar situation obtains in other LD varieties, which have their own variants of the independent prepositional pronoun (e.g. ṭaθ+possessive suffix in J. Amediya). A relatively unique feature of the LD cluster is its tolerance for stacked L-suffixes. On šqəlle, this is allowed for the R (expressed with affixally alongside A), and for all objects on šaqəl-based stems except qam-šaqəl-le. Like elsewhere in NENA, 218 THE EXPRESSION OF (PRONOMINAL) OBJECTS AND RELATED ARGUMENTS however, such stacking is dispreferred on qam-šaqəl-le, being only marginally attested. Restrictions on these object-expressing strategies in J. Dohok are both TAM- and inflection-specific. First, the functional restriction of the L-suffix is inflection- specific: the prohibition to express Themes and Patients with an L-suffix only applies in šqəlle, but not in the other past perfective, qam-šaqəl-le. The suffixal expression of Recipients of common ditransitives, such as ‘give’ and ‘say’, is extremely rare in the case of qam-šaqəl-le in J. Dohok, where the L-suffix is preferred for Themes and Patients. On the other hand, the restriction to express two objects with verbal affixes is TAM-specific: this is disallowed or strongly dispreferred in the perfective past (both in šqəlle and qam-šaqəl-le), while being tolerated elsewhere. This, in turn, also means that qam-šaqəl-le is different from all other šqəlle-based paradigms, which do allow for the stacking of L-suffixes. In dialects such as C. Barwar, by contrast, the prohibition of stacking L-suffixes is wider, applying to all šaqəl-based inflections (present and past), as well as to the šqəlle paradigm. CHAPTER SEVEN THE MORPHOLOGY AND SYNTAX OF THE COPULA Introduction he function of the copula in NENA is to connect the subject with its predicate when the predicate is not a finite verb. Typically, copula clauses involve nominal or pronominal predications of state, attribute, category or identification. The copula also occurs with adverbial predications (adverbs proper and prepositional phrases), which mostly predicate about a contingent state such as temporal or locational situations.1 Finally, the copula in NENA is used with non-finite verbal forms. In J. Dohok, in combination with b- ‘in’ and the Infinitive (i.e. wele bə-šqala, cf. §8.15), it expresses the continuous, and seems to be used especially with intransitive stative verbs. In other dialects, parallel constructions commonly expresses the progressive (i.e. dynamic continuous). In combination with the resultative Participle (i.e. wele šqila, cf. §8.14), it conveys the resultative in J. Dohok, as well as the more general perfect or even past elsewhere in NENA. The next section presents the morphology of J. Dohok copulas and basic rules about their placement in the clause. This overview is followed by a detailed discussion of semantic and pragmatic functions of particular types of copula clauses, accompanied by references to other dialects. Finally, the morphology and 1 For a typological overview of copula clauses, see Dryer (2007). 220 THE MORPHOLOGY AND SYNTAX OF THE COPULA syntax of the copula is studied from a comparative and diachronic perspective. It is shown that the NENA copulas and their syntax in general reflect a convergence with Kurdish, but such convergence is far from complete, and differs across the NENA dialects. Yet even though convergence might be the broad explanation, a few features are NENA-specific. The most notable one is the existence of the object-like copula which—in J. Dohok—results in the marking of the copula’s subject identically to the object in a transitive clause. The greatest variety in types of copula clauses in J. Dohok and in NENA in general is found in the present tense domain, where the differences concern the type of the copula, its position, the person (1st/2nd vs 3rd person predicates) and the grammatical type of the predicate (part of speech). Most dialects possess (at least) two present tense copulas – a ‘basic’ and a ‘presentative’ one (also called ‘deictic’ or ‘emphatic’ in the NENA literature; Khan 2018, 224).2 For J. Dohok, these two are referred to as, respectively, the ‘basic present copula’ and the wele copula. In J. Dohok and several other dialects, the presentative copula is restricted to the third persons, which is a conservative feature. The exact semantic, pragmatic and syntactic distribution of the ‘presentative’ copula varies across dialects, though some core features are (likely) shared across NENA. One basic feature of presentative copula constructions is that they are associated with events, which typically involves verbal and adverbial predicates. 2 The ‘presentative’ copula is also termed ‘deictic’ or ‘emphatic’ in NENA studies. SYNTAX AND MORPHOSYNTAX 221 J. Dohok copulas: morphology and basic syntax rules 7.2.1. Basic copula J. Dohok has two present tense two copula paradigms. The paradigm of the ‘basic’ copula has independent forms (which occur before the predicate) and enclitic forms, transcribed with ‘=’. Within this paradigm, the 1st/2nd person forms are arguably more clitic-like in that they are altogether invisible to stress, which falls on the penultimate syllable of the word which the copula is cliticised to. By contrast, the 3rd person forms (henceforth ‘L-copulas’) cause the lengthening of the vowel in the penultimate syllable of the whole unit (predicate=copula) (cf. §2.3.2 for more details). This difference is apparent in the minimal pair: gúrgā=le [ˈgurgaːle] (102 ms) ‘he is a wolf’ vs gúrga wən [ˈgurgawən] (61 ms) ‘I am a wolf’ 222 THE MORPHOLOGY AND SYNTAX OF THE COPULA Table 35: The basic present copula in J. Dohok independent/enclitic enclitic only (pre/postpredicate) 3MS (=)ile =le 3FS (=)ila =la 3PL (=)ilu =lu 2MS wǝt 2MS wat 2PL wetun 1MS wǝn 1FS wan 1PL wax The 1st/2nd forms are reminiscent of the šaqəl inflection of the verb ‘to be’, e.g. wən ‘I am (MS)’ (copula) vs hawən ‘I may be (MS)’ (h-w-y ‘be’). This similarity, however, likely reflects a formal convergence, rather than a shared morphological origin, as is suggested later in this chapter. The L-copulas beginning with i are used in predicates ending with a consonant, e.g. ḅāš=ile ‘he is good’, while forms beginning with l occur after vowels, e.g. ʾamuqa=le ‘it is deep’. 7.2.2. The wele copula (contingent state and verbal) The alternative present copula in J. Dohok exists only for the third persons as an independent prepredicative copula, which formally overlaps with the alternative past copula wele. SYNTAX AND MORPHOSYNTAX 223 Table 36: The wele (contingent state/eventive) copula in J. Dohok 3MS wele 3PL welu 3FS wela 7.2.3. Negative present copula The negative copula is always prepredicative and independent. Unlike the declarative one, it has only one paradigm, which consists of the negated form of the basic copula (though from a likely earlier form, e.g. 1MS la iwən > laywən > lewən). In the third persons, however, the L-copula is not used, and the forms share the stem with the 1st/2nd persons. Two allomorphs are attested, the latter occurring in fast speech. 224 THE MORPHOLOGY AND SYNTAX OF THE COPULA Table 37: The negative present copula in J. Dohok 3MS lewe 3PL lewu 3FS lewa 2MS lewǝt/lēt 2PL léwetun/letun 2FS lewat(/lāt?) 1MS lewǝn/lēn 1PL lewax/lēx 1FS lewan(/lān?) 7.2.4. Basic past copula The past copula comprised of the basic present copula (attested in the present in the 1st/2nd persons), followed by the past convertor wa. The forms are thus reminiscent of the past imperfective k-šaqəl-wa inflection of verba tertiae /y/ (cf. §5.3.6). SYNTAX AND MORPHOSYNTAX 225 Table 38: Basic past copula in J. Dohok 3MS wewa 3PL wewa 3FS wawa 2MS wǝtwa 2PL wétuwa 2MS watwa 1MS wǝnwa 1PL waxwa 1FS wanwa 7.2.5. Alternative past copula (wele) J. Dohok possesses an alternative past copula paradigm, whose distribution appears to overlap with that of the basic past copula. The 1st/2nd person forms of the past wele paradigm could be elicited (e.g. 1SG weli), but are unattested in the corpus. This paradigm is the šqəlle inflection of the verb ‘be’ (root h-w-y) Table 39: Alternative past copula in J. Dohok 3ms wele 3pl welu 226 THE MORPHOLOGY AND SYNTAX OF THE COPULA 3fs wela 7.2.6. Negative past copula The negative past copula is formed with the negative particle la followed by the basic past copula, and the stress is always placed on the negator, for instance, là wawa ‘she was not’. 7.2.7. Future and irrealis copulas In the future and in the irrealis, the copula is formed by the respective form of the verb ‘be’ h/Ø-w-y (hawe), for example: (1) 3 (elicited) balki Ø-hoy-a go beθa maybe IRR-be-she in house ‘She might be at home.’ SYNTAX AND MORPHOSYNTAX 227 The semantics and pragmatics of NENA copula clauses 7.3.1. Semantic features: permanent vs contingent states The most important semantic distinction for the present discussion is that of permanence vs contingency. In general, constructions with the presentative copula are characteristic of contingent states, while the basic copula correlates with permanent property predications (cf. Khan 2008b, 625–6). This semantic distinction tends to correlated with the part of speech of the predicate: nouns are likely to predicate about the referent’s permanent features, such as its category or identity (e.g. ‘a Turk’, ‘the president of Turkey’). Adjectives can describe states and properties of entities which are either permanent (e.g. ‘Turkish’) or contingent (e.g. ‘tired’). Adverbials (whether prepositional phrases or adverbs proper) typically describe a referent’s ‘location’ within space or time. Locations are intuitively unstable, especially when the referent is animate. Though ‘permanence’ is a relative notion which covers an entire temporal continuum, there is a test for the permanence of a property, known as the ‘lifelong effect’. When converted to the past tense, a predication about a truly permanent state ought to imply that the referent is no longer in existence (Carlson 1980), for instance: ‘the inhabitants of this village were Kurdish’. 228 THE MORPHOLOGY AND SYNTAX OF THE COPULA 7.3.2. Pragmatic features of copula clauses: thetic vs categorical Khan's (2018) definition of clauses with the presentative (‘deictic’) copula is that they are ‘thetic’. Though no uniform definition of theticity can be found in the literature, Khan relies on Sasse’s classical notion that thetic statements are deictic in nature: they draw attention to a situation and present it as ‘an undivided whole’, while categorical statements ‘name an entity and make a statement about it’ (Khan 2018, 245, paraphrasing Sasse 1987).3 In Khan’s view, theticity is thus motivated by discourse pragmatics: a speaker chooses to package a discourse unit as thetic in order to, for instance, draw attention to new situations or provide additional information to the discourse, rather than developing it (Khan 2018, 252-5). Khan’s notions of theticity should thus be kept separate from slightly different definitions, such as those closely bound with information structure in the strict sense. According to Rosengren (1997) and Lambrecht (2000), for example, thetic clauses contain no information which is presupposed, and are thus entirely in focus. In Khan’s framework, the thetic function of ‘drawing attention to situations’ fits the profile of contingent states and eventive predications, which licenses considering such constructions thetic. Namely, both the continuous/progressive (copula + in- 3 As such, theticity is a concept different from—though related to—the notion of sentence focus, used in the study of post-verbal Subjects (§9.7). Theticity is arguably a matter of choice (choosing to package a clause as ‘highly assertive’, surprising etc.), while sentence focus suggests that the clause contains no presupposed information. SYNTAX AND MORPHOSYNTAX 229 Infinitive) and the perfect/resultative (copula + Participle) serve to draw attention to spatial and temporal situations, such as ‘look, he has come!’ or ‘look, she is sleeping’. Khan’s notion of ‘drawing attention to situations’ is useful in that it certainly captures the pragmatics of many contingent state copula clauses (which often use the presentative copula). However, theticity as defined by Khan is arguably difficult to measure for individual cases, being primarily defined by the speaker’s intention (see above). Thus, the existence of particular copula type in a sentence and word order should certainly not make us assume that the clause in question is thetic. Indeed, the multiplicity of approaches to theticity in the literature shows that the exact definition depends on the parameters set to measure it (e.g. syntax, prosody, information structure or discourse pragmatics). Moreover, we should not assume a uniform function of particular types of copula clauses across NENA. If one wanted to measure theticity, possible criteria would include syntactic context and the type of speech acts, so these criteria are considered in this study. For instance, wh- questions typically inquire about a situation which is presupposed (e.g. ‘why didn’t you come yesterday?’), so we do not expect that attention should be drawn to that situation by a marked structure (cf. Rosengren 1997, 472). As for syntactic context, it is shown in the chapter on verbal semantics that subordinate clauses such as relative ones are prototypically pragmatically unassertive: they do not present information considered new to the listener (Cristofaro 2003, 30–50). For instance, in the sentence ‘the man that you saw with me yesterday is my doctor,’ the main assertion is the phrase ‘is my doctor’, 230 THE MORPHOLOGY AND SYNTAX OF THE COPULA so why should the speaker draw attention to the clause ‘that you saw with me yesterday’? The following sections study the semantic features of the different types of copula clauses in J. Dohok, accompanied with comparative notes. Reference is also made to their pragmatic features, if these can be described by the above-introduced parameters or by context. Copula clauses are grouped by the predicate’s part of speech category and the subject’s person, because the wele copula in J. Dohok only exists in the third persons. Wele/basic copula + verbal predicate 7.4.1. 3rd person clauses In J. Dohok, verbal predicates4 can only occur with the pre-predicative wele copula, apparently with the exception of clauses with an extraposed subject (i.e. subject in narrow focus). Constructions with the basic (i.e. L-) copula either before or after the predicate are unattested. The fact that the basic copula is not permissible in most environments and that wele occurs even in ‘non-thetic’ 4 The Participle (šqila) is formally ambiguous, since it also occurs with adjectives which are unambiguously stative in meaning, that is, are adjectives proper. In the present study, verbal or potentially verbal uses of the Participle comprise of those lexemes which can also occur in finite verbal forms. Thus, the adjective quya ‘strong’ is excluded from the verbal section, since forms such as *qweli ‘I became strong’ (šqəlle; q-w-y) do not exist. SYNTAX AND MORPHOSYNTAX 231 environments suggests that the primary association of the wele copula construction is not pragmatic, but rather grammatical: associated with verbal predicates in J. Dohok. The use of the wele copula is illustrated in (2): (2) 1b/57 mər-ru ta mằləkˈ flan welu ʾə̀θy-e,ˈ Ø- mdagəl-Ø xa-dùgla.ˈ PERF.say-they to king some_people COP.DEIX.they RES.come-PL IRR-lie- he certain-lie They said to the king ‘some people have come, [one] shall tell you a lie.’ The wele copula is clearly favoured even in contexts which are pragmatically non- assertive, such as relative clauses and questions, for instance:5 (3) 5 (elicited) ṭamằ wele ʾəθy-a ʾaxon-ox?ˈ why COP.DEIX.he RES.come-MS brother-yourMS ‘Why has your brother come [here]? (4) 1 (elicited) Context: ‘I have two sisters. One lives in Tel Aviv, the other in Jerusalem.’ 5 These examples come from elicitation, because no certain examples of such constructions are attested in the corpus (for possible examples, see §7.9.3 below). Clauses with the basic copula were accepted by informants, but never produced, suggesting that they are strongly dispreferred. 232 THE MORPHOLOGY AND SYNTAX OF THE COPULA xaθ-i d-wela skən-ta go tel avìv, g-lep-a Ø-peš-a dòxtor.ˈ sister-my REL-COP.DEIX.she RES.live-FS in Tel Aviv REAL-study-she IRR- become-she doctor. ‘My sister who lives in Tel Aviv is studying to become a doctor.’ This distribution indicates that in J. Dohok, the construction wele + predicate has been grammaticalised in verbal clauses, rather than being a marked structure to convey theticity. This, in turn, means that there is no way to formally encode a thetic verbal clause in J. Dohok, or to emphasise a situation’s contingency as opposed to a persistent verbal predication like the one in (4) above.6 This overwhelming preference for wele in verbal clauses also results in a situation in which the L-copula—in contrast to many other dialects—is almost always postverbal (see below). The L-copula is also much strongly associated with stable and permanent states in J. Dohok than is the case in many other NENA varieties. LD dialects such as J. Betanure and J. Amediya (e.g. Hoberman 1991, 174) allow for a pre-predicate L-copula with verbal predicates, including clearly contingent states, as illustrated for J. Betanure in (5): (5) J. Betanure (Mutzafi 2008a, 50) 6 Still, there is one case of COP-SUBJ-PRED is attested (cf. (6) below), and this clause corresponds to a high degree of pragmatic presentativity. SYNTAX AND MORPHOSYNTAX 233 le bə-rgada COP.PRS.he in-INF.shiver ‘he is shivering’. According to Mutzafi (2008b, 57), the presentative copula in J. Betanure (wele or wəlle) adds a ‘deictic-presentative meaning’, though at other times apparently adds no such presentative nuance. This indicates that the two types of copula clauses (i.e. wele + predicate and le + predicate) are to some extent grammatically and pragmatically in free distribution: they can both occur with verbal predicates and do not necessarily add a thetic nuance. J. Zakho—though in principle allowing both copulas with verbal predicates (‘rhemes’)—exhibits a strong preference for the ‘presentative’ (wele) copula (Cohen 2008, 60). When the ordinary copula is used, it is apparently postverbal (Hoberman 1991, 174), which contrasts with the position of the copula in verbal clauses in J. Dohok, J. Betanure and J. Amediya. In clauses with the Infinitive in C. Barwar, both the presentative, pre-predicative and the basic, post-predicative copulas occur, i.e. hole qṭala and qṭalɛle ( (Khan 2008b, 704–5). C. Barwar is thus capable of expressing a greater variety of 7 ‘Salience’ is defined as the speaker’s wish ‘to draw particular attention to an activity that exists in the current speech situation’ (ibid). 234 THE MORPHOLOGY AND SYNTAX OF THE COPULA pragmatic nuance in verbal copula clauses than J. Dohok. This comparative discussion is summarised in Table 40 below. SYNTAX AND MORPHOSYNTAX 235 Table 40: Features of verbal copula clauses in LD and C. Barwar Ordinary (L- Ordinary Cases A dedicated copula) used (L-) with no deictic/presentative/thetic alongside copula copula structure ‘presentative’ attested in main in post- clauses predicate reported positions J. Dohok no (except no no — clause with extraposed subject) J. Zakho yes; yes no wele + predicate (but not infrequently always with this function) J. Amediya yes no no wəlle + predicate (always with this function?) J. Betanure yes no no wəlle + predicate but not always with this function) C. Barwar yes yes yes hole + predicate 236 THE MORPHOLOGY AND SYNTAX OF THE COPULA In clauses with narrow focus on the subject when the subject is extraposed in J. Dohok, speakers accept and produce both types of copula, for instance: (6) 1 (elicited) Sàra d-wela /t-ila skən-ta ʾaxxa.ˈ Sarah REL-COP.DEIX.she /REL-COP.PRS-she RES.live-FS here ‘[It is] Sarah who is living here. In NENA in general, the use of the independent copula is common in such cases; the copula typically follows the constituent in narrow focus (Khan 2018). In clauses such as (5) above, the focus is a proper noun – a permanent identity label of an entity. The use of the basic copula can thus be explained semantically (see below). Clause-medial subjects (COP-SUBJ-PRED) also occur, but only marginally and likely pragmatically highly marked. 7.4.2. 1st/2nd persons: ordinary copula–predicate In the 1st/2nd persons, the basic copula is used in all types of verbal clauses and— like the wele copula in the 3rd persons—it is always pre-predicative: (7) 2 (conversation) wən lip-a ṭòṛa.ˈ COP.PRS.IM RES.study-MS Torah ‘I have studied the Hebrew Bible.’ SYNTAX AND MORPHOSYNTAX 237 This structure (i.e. basic copula + predicate) is also characteristic of J. Betanure and J. Amediya (though not J. Zakho, which has a post-predicative basic copula), and of NENA more broadly. Wele/basic copula + non-verbal predicate clauses 7.5.1. 4.1. 3rd person clauses: presentative copula + predicate The clause consisting of the wele copula followed by a predicate is attested with adverbial and adjectival predicates.8 With adverbial predicates, like with verbs, the wele copula is attested, even for states that are stable (e.g. (9) below), being avoided only with permanent states, suggesting primarily a grammatical association (the wele copula + adverbial predicates). With adjectives, it is used with contingent states in some cases, and appears to retain its thetic force. Adverbial predicates (8) 3a/25 Wì,ˈ žàri,ˈ papùki,ˈ wi welu d-la ʾĭxàla!ˈ oh poor miserable oh COP.DEIX.they without food ‘Oh, poor ones, miserable ones, oh look, they are without food!’ 8 For adjectives in the šqila (Participle) pattern, see below. 238 THE MORPHOLOGY AND SYNTAX OF THE COPULA The following sentence is also interpreted as having an adverbial predicate (it predicates about where the subject lives), though the expected preposition (go ‘in’) is missing. It demonstrates the preference of wele even with stable properties. (9) 3f/11 g-əmr-i ʾiθən xa-huðàyaˈ wele b-amr-an-nax Hšikunim bèt.H REAL-say-they EXIST certain-Jew COP.DEIX.he FUT-say-IF-youFS estate no.2 ‘They say “there is a Jew, he lives—let’s say—[in the] housing estate number 2.”’ The wele copula is even used in ‘non-thetic’ contexts such as relative clauses:9 (10) 2,3 (elicited) ʾo masḥaf dət wele {t-ile} rəš šùlḥanˈ dìdī =le.ˈ thisM book REL COP.DEIX.he {REL-COP.PRS.HE on table my =COP.PRS.he ‘this book which is on the table is mine.’ Adjectival predicates 9 The basic copula was accepted by informants (indicated in curly brackets in the example), but not actively produced. Possible cases of wele in relative clauses also feature in the corpus, but the copula is morphologically ambiguous in those instances (cf. §7.9.3 below). SYNTAX AND MORPHOSYNTAX 239 With adjectival predicates, the wele copula occurs only with contingent states, and apparently retains its thetic force, in contrast to verbal and adverbial predicates. The predication in (11), for instance, is arguably a good candidate for a thetic construction: it draws attention to the striking situation that the girl Khsisene remains visually impaired, despite having received treatment. The thetic force of wele with adjectives is also implicit in the fact that relative clauses with wele + adjectival predicate could not be elicited.10 (11) 3c/25 Context: ‘You went to Mosul, to Baghdad and they couldn’t cure her? How could I cure her eyes?’ kappàra, xsisə̀nne,ˈ hàr wela šahar-ta.ˈ Hàr la k- xazy-a.ˈ poor Khsisene still COP.DEIX.she blind-FS still NEG REAL-see-she ‘Poor thing, Khsisene, look, she is still blind. She still cannot see.’ 7.5.2. 4.2. 1st/2nd person clauses: basic copula + predicate In the 1st/2nd persons, predications about non-permanent states have a basic copula before the predicate. Only one noun is attested within this type of clause, and it has a highly referentially prominent referent. 10 I tested unambiguous adjectives (i.e. not of the šqila pattern). 240 THE MORPHOLOGY AND SYNTAX OF THE COPULA Adverbial predicates With adverbials, the order ‘copula + predicate’ is attested in all environments, including (asyndetic) subordinate clauses: (12) 3d/81 ʾatta Ø-yaʾ-e ʾana wən ʾaxxa b-qaṭḷ-ì-li!ˈ now IRR-know-they I COP.PRS.IM here FUT-kill-they-me ‘now [if] they find out [that] I am here, they will kill me!’ Adjectival predicates (13) 3d/37 g-əmr-a ʾe, wat kurkamàn-ta!ˈ REAL-say-she oh COP.PRS.YOUFS yellow-FS ‘She says “oh, your face is yellow!”’ Noun predicates The only noun predicate following the basic copula has a pronominal (i.e. referentially prominent) subject: (14) 3d/84 là māx- ə́tu -li,ˈ ʾana wən wazīr̀ ˈ go ḥukùma!ˈ NEG IRR-beat -youCPL -me I COP.PRS.IMS minister in government ‘Don’t beat me, I am a minister, in the government!’ SYNTAX AND MORPHOSYNTAX 241 According to Khan, the construction SUBJ-COP-PRED is a marked structure for highly topical referents in NENA (e.g. C. Barwar - Khan 2008b, 626). As the C. Barwar example in (15) below demonstrates, it even occurs even with predicates that express a permanent property: (15) C. Barwar (2008b, 626) Context: ‘We had a fast (…)’ ṣawma ʾile xamší yomàθa. fast COP.PRS.he fifty days ‘[This] fast lasts fifty days.’ 7.5.3. Clauses with the Participle (wele + predicate in 3rd persons) In the J. Dohok corpus, clauses including a Participle form of a verb in the vast majority of cases follow the ‘verbal’ copula syntax: (wele) copula + predicate.11 Such predications cover the entire semantic spectrum from events to non- permanent states, and thus include both ‘verbal’ and ‘adjectival’ interpretations of the Participle form.12 This means that in J. Dohok, the meaning of such 11 This discussion only applies to the Participle forms which can also occur in finite verbal forms, e.g. kpina ‘hungry’ > kpənni ‘I became hungry’, but not quya ‘strong’, since forms such as *qweli ‘I became strong’ do not exist. The latter word is thus unambiguously an adjective. 12 For a more detailed discussion on the range of semantic properties of the Participle in NENA, see Khan (2008b, 653–660). 242 THE MORPHOLOGY AND SYNTAX OF THE COPULA predicates—‘eventive’ vs ‘stative’—cannot be determined through formal syntactic features such as word order and copula type, but only on the verb’s actionality and the broader context. This ambiguity in meaning is especially pertinent to intransitive verbs which are telic, and thus typically agentive. In the Participle form, such verbs can refer either to a dynamic event or to the state following that event (cf. Khan 2008b, 653–660), as illustrated in example (16) below. The distinction between eventive (dynamic) and stative meaning can remain ambiguous in cases where it cannot be determined from the context. (16) Id/14 wan- wa tu-ta… COP.PRS.IFS -PST RES.sit-FS ’I had sat down/I was sitting…’ By contrast, atelic predicates (with experiencer subjects) always have a stative, adjective-like interpretation in the Participle form, as demonstrated in (17) below. Such predicates, however, still prefer the construction wele (3rd persons)/basic copula (1st/2nd) + predicate in J. Dohok, which is due to their semantic connection with contingent states. (17) 3c/22 Wi! ʾilaha Ø-naṭər-Ø, wat smə̀x-ta!ˈ oh God IRR-keep-he COP.PRS.youFS pregnant-FS ‘Oh! God keep you, you are pregnant!’ SYNTAX AND MORPHOSYNTAX 243 A similar formal-syntactic ambiguity arguably applies to transitive events which launch a state, especially in clauses with the patient as the grammatical subject. This ambiguity is demonstrated in (18) and applies to verbs such as t-w-r (tawər) ‘break’ and ṃ-ḷ-y (ṃaḷe) ‘fill’. (18) 1b/18 ʾe kǝsta wela mliθa zùze. thisF bag COP.DEIX.she RES.fill-FS money ‘This bag has been filled/is full [of] coins.’ A similar ambiguity is reported for J. Betanure in clauses which have a pre- predicative copula, while the order predicate=copula apparently suggests a ‘stative meaning’ (Mutzafi 2008b, 56–7).13 In J. Dohok, however, the order ‘predicate=basic copula’ is extremely rare with the Participle, meaning that most cases involving the Participle in this dialect are ambiguous. Hoberman argues that in J. Amediya, the position of the copula can help disambiguate a stative from a dynamic interpretation of the Participle, though in his examples, this difference seems to be lexical rather than usage-dependent. He thus lists kpina ‘hungry’ among stative adjectives, because kpina precedes the copula, rather than following. However, a single lexeme can have both a stative and an eventive interpretation, as shown above, so in principle, it would be expected to occur both before and after the copula. 13 For formal ambiguity in C. Barwar, see Khan (2008b, 653–60). 244 THE MORPHOLOGY AND SYNTAX OF THE COPULA Non-verbal predicate=basic copula 7.6.1. 3rd person copula clauses: predicate=basic copula The enclitic copula occurs with adverbial, adjectival, pronominal and noun predicates. With adverbial predicates, it is only possible with properties that are truly permanent. With nouns and pronouns, only this construction is attested in the corpus, which is expected in light of the association of nouns and pronouns with the permanent property. With adjectives, the order ‘predicate=basic copula’ occurs not only with permanent states, but also contingent ones, where both the basic and the wele copula are possible. Adverbial predicates (19) 4a/1 yə̀mmiˈ mən mòṣul=ila.ˈ moṣul g-maḥke ʿằrabi.ˈ ‘My mother is from Mosul. They speak Arabic [in] Mosul.’ Adjectival predicates Adjectival predicates describing stable or permanent properties typically have an enclitic copula: (20) 3b/71 qurbana rušalàyim,ˈ basə̀mtā=la.ˈ SYNTAX AND MORPHOSYNTAX 245 ‘Beloved Jerusalem – it is good.’ In other cases, the copula is cliticised to adjectival predicates which clearly predicate about a contingent state (cf. (21) below), showing that both type of copula clauses are available in such cases. As was suggested above, the wele copula still retains its thetic force with such predicates. (21) 1c/82 gurga ži qam-yāsər-Ø-re reš gyàne,ˈ brìndar=ile.ˈ ‘As for the wolf, he bandaged his head – he is wounded. Noun and pronoun predicates With nouns, only the ‘predicate=basic copula’ order is attested (except for two specific pragmatic context, see below). Most nouns predicate about a permanent property, but stable yet non-permanent features are also attested (e.g. in (22) below). The absence of wele even when the relevant semantic or pragmatic parameters are met (non-permanence and spatial deixis as in (22)) suggests a grammaticalisation of the basic enclitic copula with noun predicates. (22) 3d/32 Context: ‘Today, a Jew won’t be seen wandering in the streets. Look at this one!’ ʾe yàltā=la,ˈ mà huðeθa!ˈ ‘This one is a child, what [do you mean a] ‘Jew’! 246 THE MORPHOLOGY AND SYNTAX OF THE COPULA 7.6.2. 1st/2nd person copula clauses: predicate + basic copula In the corpus, the construction ‘predicate + basic copula’ is only attested with a noun, but is expected to occur also with adjectives and adverbs expressing permanence. (23) 1c/41 gùrga wən,ˈ k-əxlən nàše.ˈ ‘I am a wolf. I eat people.’ Basic copula + predicate (non-verbal, permanent property) In J. Dohok, the basic copula + predicate order is attested in non-assertive (conditional) clauses, for example:14 (24) 1b/86 u-ʾēn ila dùgle,ˈ hàl-li kəsta.ˈ ‘And if it is a lie, give me the bag [of money].’ 14 For the notion of assertion, see §8.12-13. SYNTAX AND MORPHOSYNTAX 247 Summary of present tense copula clauses in J. Dohok A notable feature of J. Dohok is the strong complementary distribution of wele and the basic copula in the 3rd persons: while the former is always pre-predicative, the latter is almost always enclitic to the predicate. Moreover, the semantic and pragmatic correlates of particular copula constructions are apparently specific to the type of the predicate. In verbal and adverbial predicates, the construction ‘wele + predicate’ (3rd persons) is not restricted to clauses which draw attention to situations with spatial or temporal deixis, that is, thetic constructions, likely in contrast to other NENA dialects (e.g. LD varieties and C. Barwar). Rather, with verbal and adverbial predicates, the pre- predicative (wele) copula is often the only possible one (even in predications of stable property), and occurs even in ‘non-thetic’ environments such as wh - questions. In the 1st/2nd persons, only the basic copula in the pre-predicative position is available in most environments, which makes J. Dohok similar LD dialects such as J. Betanure and J. Amediya, though distinct from J. Zakho, which has a post-predicative basic copula with verbal predicates. The grammaticalisation of the ‘presentative’ wele copula in J. Dohok with verbal and adverbial predicates, including ‘non-thetic’ and in pragmatically non-assertive environments, likely distinguishes this dialect form the other LD varieties. No LD grammars mention subordinate use of the wele/wəlle copulas; in J. Zakho, this reportedly does not occur (Cohen 2017, 2–3). Such an extension of the originally presentative copula could be a local development within NENA. The use of the 248 THE MORPHOLOGY AND SYNTAX OF THE COPULA ‘presentative’ copula hole in non-assertive clauses such as relative and interrogative ones has also been noted for Christian dialects in the vicinity of Dohok, for instance Mar Yaqo and Aradhin.15 With adjectival predicates, the two copula constructions (‘wele + predicate’ and ‘predicate=basic copula’) seem to be in free distribution with states that are not permanent. However, the construction wele + predicate likely still retains some thetic force, as it appears to be favoured especially in discourse contexts which license a thetic construction. Nouns (and likely personal pronouns) in the corpus do not occur in the construction ‘wele copula + predicate’ (3rd person), and there’s only one attestation of basic copula + predicate (1st/2nd). Even clauses predicating about a non-permanent state have an enclitic copula, suggesting that the key factor determining the choice of clause is grammatical (noun as a predicate). Past tense copula clauses 7.9.1. Clauses with the basic past copula In the overwhelming majority of cases, past copula clauses are formed with the basic past copula (cf. Table 38 above). Other types of past copula exist, and appear to be in free distribution with the basic past copula (see below). Unlike in the 15 Geoffrey, Khan, personal communication. SYNTAX AND MORPHOSYNTAX 249 present, there are hardly clear syntactic patterns for different (semantic and grammatical) types of clauses. The only clearly observable rule is that verbal predicates come after the copula (in clauses both with and without an explicit, independent subject), as is the case in the present domain, for instance: (25) 1d/5 ʾana wən-wa tiw-a qapràna.ˈ I COP.IMS-PST RES.sit-MS rooftop_platform I had sat/been sitting [on] a rooftop platform. This situation in past copula distribution (only fixed rule being pre-predicative copula in verbal clauses) is also reported for J. Amediya (Hoberman 1989, 178). With adverbial predicates, the copula tends to be preverbal (cf. example in (26) below), but since there is only a handful of such adverbial clauses, it would be premature to claim that this is a categorical rule. (26) 3g/1 gòr-iˈ we-wa go xastaxàna,ˈ xastaxan-ət har tsofìm.ˈ husband-my COP.he-PST in hospital, hospital-GEN Mt Scopus. ‘My husband was in the hospital, the Mount Scopus hospital.’ In nominal predicates, there is no categorical rules for copula placement (cf. (27) and (28) below), but the post-predicate position predominates. In the case of nouns, the past situation contrasts with the present, where the order predicate + copula is fixed. 250 THE MORPHOLOGY AND SYNTAX OF THE COPULA (27) 3d/103 Context: relations between Jews and Muslims in Iraq. we-wa jiràne,ˈ xa g-əb-e-wa u-xeta…ˈ COP.they-PST neighbours one REAL-love-he-PST and-another ‘[In Iraq,] they were [all] neighbours, they loved each other…’ (28) 3d/104 kùll-u jirane we-wa,ˈ məšəlmàne,ˈ mux HʾaḥìmH wewa.ˈ all-them neighbours COP.they-PST Muslims like brothers COP.they-PST ‘[In Iraq, they] were all neighbours, [the] Muslims were like brothers [to us].’ The particular word order does not seem to correlate to information-structural parameters either, because the order subject + noun predicate + copula is also attested with nucleus stress on the predicate, which is in narrow focus: (29) 1b/3 bə-daw wàxtˈ kud mà-θaˈ dawə̀l-ta wa-wa.ˈ in-GEN-thatM time every town state COP.she-PST ‘At that time, every city was [a] state.’ SYNTAX AND MORPHOSYNTAX 251 7.9.2. Clauses with k-hawe-wa (*h-w-y, hawe) ‘used to be’ as a copula There is one attestation of the verb h-w-y in the k-šaqəl-wa form, apparently serving as a past copula. This particular form may have been used to emphasise the habituality (cf. (30) below) of the predication, perhaps especially with adverbial predicates such as the one below, since these are prototypically associated with contingent states. Still, the basic past copula is also attested in adverbial predicates referring to long periods as demonstrated in example (31), which indicates at the least that the basic copula also occurs with habitual meanings. (30) 3b/51 Baxtàθaˈ k-aw-e-wa l-bèθa.ˈ women REAL-be-they-PST to-home. ‘[As for] women, they used to be at home.’ (31) 3d/105 go daʿăwətyàθa muxde we-wa.ˈ in weddings together COP.they-PST ‘On weddings, they [Jews and Muslims] used to be together.’ 252 THE MORPHOLOGY AND SYNTAX OF THE COPULA 7.9.3. Alternative past copula and its identity with the present wele copula In the LD dialects, the past copula construction can also be expressed through apocopated šqəlle forms of the verb ‘be’, *h-w-y (cf. Table 39 above), e.g. 3MS wele.’ This usage contrasts with the unapocopated forms such as hwele, meaning ‘be born.’ Formally, this past copula is identical with the present presentative copula in J. Zakho and J. Dohok. In the J. Dohok corpus, it is consequently difficult to distinguish the two, which, in turn, renders it challenging to determine the distributional and functional differences between the basic and the wele past copulas. According to Cohen (2017), in the J. Zakho corpus, the two are distinguished in word order: the presentive copula is always pre-predicative, while the past copula is post-predicate.16 In J. Dohok, however, all the wele, wela and welu copulas are pre-predicative. Since the examples are few, however, this should not be taken as evidence that only this order is permissible. Moreover, no 1st 2nd forms (e.g. welox 2MS) are attested in the corpus; such forms would be unambiguously past. This notwithstanding, the entire paradigm could be elicited. This suggests that in J. Dohok, wele with the past meaning exists, but is highly infrequent (and perhaps obsolete). Unlike in J. Zakho, however, its position is not 16 This analysis seems accurate; the order predicate + wele past copula occurs even in clauses with verbal (or at least contingent state) predicates, in which the order copula + predicate would be expected. For instance: sməxta we-la RES-pregnant-FS PST-COP-he ‘she was pregnant’ (data from Cohen 2017, 5). SYNTAX AND MORPHOSYNTAX 253 fixed, causing sometimes a complete overlap with the wele presentative copula clause. In J. Zakho, the apparent restriction of the wele past copula is likely an attempt to disambiguate between the two. The following could be examples of the usage of the past copula wele, or of the present wele copula, whose past temporal values is inferred from the context. The copula clause in (32) is subordinate, suggesting that this might be a past copula. In J. Dohok, however, the present (verbal/eventive) wele copula can occur in subordinate contexts with adverbial predicates, so the past interpretation is not completely certain. (32) 1d/27 θe-le ʾaxxa, g-əb-e Ø-qāṭəl-lu kùllu huðaye PERF-come-he here REAL-want-he IRR-kill-he-them all Jews ‘(The Mufti of Jerusalem) came here, wanting to kill all the Jews…’ d-welu/we-lu go d-a mə̀nṭaqa.ˈ REL-want-he COP.DEIX.they/PST.COP-they in GEN-thatF region. ‘…who were in this region.’ The sentence in (33) may be another example of the past copula wele; the narrative context is past, so a past copula would be suitable. On the other hand, clauses setting the scene or describing background/circumstances often occur in the 254 THE MORPHOLOGY AND SYNTAX OF THE COPULA present within past narrative,17 which leaves open the possibility that this sentence contains a presentative copula: (33) 5a/32 ʾur-ra šàbθa,ˈ naše welu/we-lu go knə̀šta.ˈ PERF.enter-she Shabbat, people COP.DEIX.they/PST.COP-they in synagogue ‘Shabbat had begun, people were/are in the synagogue.’ Negated copula clauses (all clause types) A negated copula always precedes the predicate (so its position corresponds to the pre-verbal position of the verbal negator la).18 The negated copula furthermore always carries nucleus stress (cf. (34) and (35) below), which highlights its focal role (negated clauses typically carry narrow, e.g. counter-presuppositional, focus). (34) 1b/30 ʾeha lè-wa dugle.ˈ ThisF NEG-COP.PRS.she lie This is not [a] lie. 17 A certain example of a present copula within past narrative is: gurga ži qam-yāsər-Ø-re reš gyàne,ˈ brìndar=ile.ˈ wolf also PERF-bind-he-it head himself, wounded=COP.PRS.he ‘As for the wolf – he bandaged his head – he was [lit. is] wounded.’ 18 This is common across NENA (cf. e.g. J. Challa - Fassberg 2010, 101). SYNTAX AND MORPHOSYNTAX 255 (35) 3d/75 là wewa mux daxxa…ˈ NEG COP-he-PST like here ‘(In Iraq,) it wasn’t like here.’ Irrealis copula clauses The irrealis copula is formed through the verb h-w-y ‘be’. In the non-past domain, šaqəl forms are used (as illustrated in (36) below), while clauses with past temporal reference employ šaqəl-wa forms. (36) 4a/1 Context: ‘The Greeks used to force the Jews to abandon (their) religion.’ gəbe Ø-hawe-Ø mux yawnàne.ˈ MUST IRR-be-they like Greeks ‘They must become like the Greeks.’ Clauses lacking a copula In most NENA dialects, the copula is occasionally omitted. In some dialects, this is grammaticalised in particular constructions: in C. Barwar, the absence of the copula with the Infinitive predicate signals the perfective (rather than the progressive) meaning of the predicate. This pattern does not occur in J. Dohok corpus, likely because of the less developed (in scope and frequency) use of the non-finite verbal forms. In J. Betanure, copula omission has a pragmatic correlate: 256 THE MORPHOLOGY AND SYNTAX OF THE COPULA it occurs occasionally when ‘the subject is clear from the context’ (Mutzafi 2008b, 50; in the examples, there is no explicit subject referent in the clause). The copula may even be missing in the TZ varieties such as J. Sanandaj which—as has been argued here—exhibit the greatest degree of convergence with Iranian in copula clauses in NENA. For J. Sanandaj, this is reported in parallel contexts arguably parallel to other NENA – when the referent is clear from context (in circumstantial clauses and in adjacent parallel non-verbal clauses) (Khan 2009, 340-1), as illustrated in (37) below: (37) J. Sanandaj (Khan 2009, 39, A:77) bšəlmané trè jorén.ˈ xá šíʿá xá sunnì-ye.ˈ Muslims two kinds one Shia, one Sunni-COP.PRS.he ‘Muslims (are) of two kinds, one (is) Shiite and the other Sunni.’ In Iranian, the (clause-final) copula is reportedly obligatory (Haig 2017, 404). Exceptions might exist, but I have not seen a discussion on this in the literature. Thus, a comparison of the degrees of copula obligatoriness in NENA (and other languages in contact) could constitute a fruitful research avenue, revealing degrees of areal convergence. In the J. Dohok corpus, the copula omission correlates with particular styles and pragmatic functions. Overall, however, such omission is rare – only about seven such clauses are attested in the corpus, so the copula can be considered near- SYNTAX AND MORPHOSYNTAX 257 obligatory.19 In (38) below, the copula omission occurs in a particularly salient phrase at the end of a folktale. The sentence in question is the culmination of the story: the speaker answers the king’s question about where the middle of the world is.20 This sentence could reflect an older layer of the language, being transmitted together with the folktale in this fixed form. (38) 1a/18 Context: ‘So the king asks “where is the middle of the world?”’ kursi dìdoxˈ palgut-Ø dùnye.ˈ chair yourMS middle-GEN world ‘Your throne is the middle of the world.’ In a few cases, the copula is omitted in the second of parallel clauses (perhaps for the economy of expression or for stylistic elegance): (39) 3f/22-3 g-emər-Ø hakar ʾana ʾùd-li dla ləbb-ox,ˈ baxt-i Ø-hoy-a mṭolàq-ta.ˈ REAL-say-he if I PERF-do-I against heart-yourMS wife-my IRR-be- she RES-divorce-FS 19 In some cases, the copula omission could also be due to Modern Hebrew interference. 20 Its word order is also noteworthy: the focal constituent (the predicate ‘your throne’) is fronted, the topic (‘middle of the world’) is clause-final. 258 THE MORPHOLOGY AND SYNTAX OF THE COPULA He says ‘if I did something against your wish [lit. against your heart], my wife shall be divorced.’ g-emər-Ø ṭa-le tṑv, beseder,ˈ baxt-ox mṭolàq-ta.ˈ REAL-say-he to-him good well wife-yourMS RES-divorce-FS ‘He says to him “good, well then, your wife (shall be) divorced.”’ Comparative and diachronic discussion 7.13.1. Introduction: Distinctive and areal features of NENA copulas and their clauses As shown in this chapter, the presence of a copula in clauses lacking a finite verb is near-obligatory in J. Dohok, as is the case throughout NENA. Such copula frequency in the present tense realm is a development from earlier (Late) Aramaic, where a (pronominal) copula was in the present restricted to particular information-structural functions, such as narrow focus on the predicate (see further Khan 2006 and Cohen 2008). A widespread usage of a (post-predicate) copula is an areal feature, occurring in the various genetically unrelated languages north Iraq, south-eastern Turkey and western Iran (Haig 2017; cf. also Haig and Khan 2019). It is generally considered to have emerged under the influence of the local Iranian varieties, which have an obligatory copula in non-verbal clauses. The presence of a copula in clauses lacking a finite verb in NENA may also be viewed as a verbalisation of such clauses (cf. Retsö 1987, 243) in that the copula cross-references the subject – often by being cliticised to the predicate itself, much SYNTAX AND MORPHOSYNTAX 259 like a verbal subject index. Such verbalisation is compatible with the claim about Iranian convergence, since the copula in Iranian is verbal in origin (Haig 2017; Khan 2018 and Matras 2009, 270). Still, this convergence with Iranian is not complete in NENA; the synchronic Aramaic varieties and their Iranian linguistic neighbours differ in the copula’s position, its obligatoriness and its verbal vs pronominal origin. The degree of convergence in NENA (as well as in Ṭuroyo and Mlaḥso) is therefore dialect-specific, with the greatest convergence being exhibited by the Trans-Zab Jewish dialects.21 As for the morphology of NENA copulas, its distinctive feature is the existence of the so-called L-copula, listed above. In some environments, this copula is morphologically identical to the verbal L-suffixes. Such a copula does not exist in any other Neo-Aramaic variety, including the closely-related Ṭuroyo and Mlaḥso.22 In the vast majority of dialects, this L-copula is restricted to the 3rd persons, where it first developed. Another characteristic of the NENA copula system is the partial resemblance of the 1st/2nd person basic copula to the šaqəl inflection of the verb h/Ø-w-y (hawe) ‘be’. This similarity likely reflects a partial convergence in form with Iranian, where the copula is verbal in origin, based on the verb ‘be’. 21 For a cross-linguistic areal study of the copula and its clauses featuring Qəltu Arabic, North- Eastern, Central and Western Neo-Aramaic and Kurdish, and for further references, see the conference paper by Noorlander and Molin (2020). 22 For the present copula paradigms in Ṭuroyo and Mlaḥso, see Jastrow (1992 and 1994 respectively). 260 THE MORPHOLOGY AND SYNTAX OF THE COPULA A hallmark of Eastern Neo-Aramaic is the copula’s post-predicate (enclitic) position. This contrasts with Western Neo-Aramaic, where clauses with locative predicates have no copula, while adjectival predicates have a proclitic copula, for instance: (40) Western Neo-Aramaic (based on Arnold 1990, 58:6) (ana) n-uzʕur-∅ (I) I-small-MS ‘I am small.’ 7.13.2. The L-copula (most dialects third persons only) Morphologically, the NENA L-copula (J. Dohok 3MS le or ile) is (nearly) identical to the verbal L-suffixes (3MS le),23 though there are phonotactic differences, at least in J. Dohok. L-copulas not restricted to the third persons are found only in two peripheral dialects: J. Urmi in north-western Iran (Khan 2008a, 61) and in C. Baz in eastern Turkey.24 J. Amediya is also reported to have an L-copula paradigm in the 1st/2nd persons, which apparently coexisting with the more typical (i)wət 23 Retsö (1987, 220) suggests—after Nöldeke—that the L-copula is historically derived from ʾīθ l- e EXIST to-him ‘he has’ > ile COP.PRS.he ‘he is’, though synchronically, he also considers the L- copula similar to the L-suffix (‘object-like’). 24 Geoffrey Khan, personal communication. SYNTAX AND MORPHOSYNTAX 261 copula type.25 The Trans-Zab varieties in western Iran and in the Iraqi region adjacent to it (e.g. J. Suleimaniya) have no L-copula (e.g. J. Sanandaj - Khan 2009, 85; J. Suleimaniya - 2008a). However, the western-most Trans-Zab dialects – J. Arbel and the ‘transitional’ J. Barzani cluster – resemble the rest of NENA, having L-copulas restricted to the 3rd persons (Mutzafi 2002a, 426; 2008c, 54). In addition to the L-copula, the L-suffixes (or morphemes with which they are synchronically identical) also occur within the various presentative copula paradigms, for instance, we-le/wəlle in LD. As for the relationship of the L-copula with the L-suffixes, despite their morphological (near) identity in J. Dohok, the two are not identical phonologically.. The L-copula is a semi-clitic: though it does not take lexical stress, it attracts stress ‘as close as possible’ to itself. By contrast, the L-copula is a clitic proper, meaning that it is altogether invisible to lexical stress. To my knowledge, this has not been pointed out for other dialects. This difference is apparent in the examples in (41): the stress in a) is on the syllable before the first L-suffix, while in b), stress remains on the penultimate syllable of the form to which the copula is cliticised: (41) a) k-pāθə̀x-Ø-le-li [kpaːˈθəxle(ː)li] (not *k-pāθəx-lé-li or *k-páθəx-le-li) 25 Like the (i)wət copula, it is said to be enclitic. However, further details regarding the frequency and distribution of the L-copula are missing. 262 THE MORPHOLOGY AND SYNTAX OF THE COPULA REAL-open-A.he-P.it-BEN.me ‘he opens it for me’ b) ʾiláhā=le [ʔiˈlaːhaːle] God=COP.3MS ‘(he) is God’ This lesser degree of phonological integration attested in the L-copula likely reflects its more recent emergence in comparison to the L-suffix. The L-suffix has various functions, including objects (of monotransitive and for Themes of ditransitive), for beneficiaries and (predicative) possessors (cf. §6), as well as for subjects in the šqəlle paradigm. Though the near-obligatoriness of the copula in NENA is due to contact with Iranian, its object-like morphology is not, because such object-like coding of the referent in copula clauses does not occur in Iranian. Instead, the development of the object-like encoding is specific to NENA, as well as occurring in some Arabic varieties in the area.26 Several possible diachronic scenarios explaining the similarity between the L- copulas and the L-suffixes have been proposed, but unanswered questions about this development remain. According to Khan, the L-suffixes were extended to non- verbal clauses ‘by analogy from deictic constructions that drew attention to a referent’ (Khan 2006, 162; 2018, 242). Their analogy’ presumably lies in the fact that they are both non-verbal clauses that often have a single referent (i.e. a subject). The object-like coding of the referent in a deictic construction, in turn, apparently occurs because such a referent is treated as an object of sensory 26 I am indebted to Paul Noorlander for pointing out these areal parallels to me; see below. SYNTAX AND MORPHOSYNTAX 263 perception (e.g. ‘look, him!’) (cf. Khan 2018). In Khan’s view, therefore, copula clauses are historically linked with sentential demonstratives (e.g. ‘here he is’). This means that a construction such as ‘look, my father (is here)’ is analogous to the clause ‘this is my father’. Sentential demonstratives with referents that are (synchronically) marked like objects are attested in Ṭuroyo, e.g. kalé [kaˈleː] ‘here he is!’, comparable with the 3MS object suffix -le (Jastrow 1992, 106). These Ṭuroyo sentential demonstratives must historically be based on a deictic/presentative particle such as kal, apparently followed by an L-suffix. If they came simply from the Ṭuroyo particle ko27 with an L-suffix, we would have expected the form *ko-le.28 The NENA presentative copulas such as wele (J. Dohok), hole (C. Barwar) or kilə (C. Qaraqosh) thus also likely originate in sentential demonstratives. Khan’s hypothesis notwithstanding, the presentative copulas (which are synchronically inflected with an L-suffix) and the L-copulas have different distributions. The former is pre-predicative, typical of contingent states and even of thetic constructions, which are arguably similar to sentential demonstratives; both draw attention to a situation. By contrast, the basic copula is often post- predicative, and in that position, often associated with categorical clauses that describe a permanent or a stable property. Therefore, extending the object-like 27 This is the realis present particle and largely corresponds to the NENA preverb k-/ʾi. 28 Thus, it is not certain whether ko (realis preverb) and ka-le (sentential demonstrative) in Ṭuroyo share their etymology. Ka(-le) could perhaps be related to the Kurmanji presentative particle ka. 264 THE MORPHOLOGY AND SYNTAX OF THE COPULA marking from sentential demonstratives to categorical copula clauses would have required a significant reinterpretation of the sentential demonstrative’s (‘copula’s’) semantics and pragmatics. In his pioneering study, Retsö (1987) analyses several Neo-Aramaic and Arabic varieties which use the same morpheme for both an enclitic copula and an object (Theme) affix. Both ‘nominative’ and ‘oblique’ morphemes occur. The most notable parallel to the NENA object-like encoding is found in Christian Baghdadi Arabic, where the object marker yā has also been extended to the function of a post-predicative copula and is available for all persons (Retsö 1987; Abu-Haidar 1991). In this Qəltu Arabic variety, this post-predicate copula yā is optional, apparently used for ‘emphasis’ (e.g. ‘indeed, X has the property Y’), for instance: (42) C. Baghdadi Arabic (Abu-Haidar 1991, 122) hiəyyi ḥəlw-i yā-ha she pretty-FS OBJ-her ‘She really is pretty.’ Like Khan (2018), Retsö also points out the original shared deictic function of the copula and the object (Theme). He also notes a positional parallel. In the Arabic and Aramaic varieties which have an identical object marker an ‘oblique’ copula, the copula is enclitic to the predicate and the pronominal object (Theme) is always the last pronominal argument to be expressed (i.e. occurs after a pronominal Recipient in ditransitive constructions). SYNTAX AND MORPHOSYNTAX 265 In NENA, however, the copula coincides in its position not only with a pronominal object of a transitive clause, but also the subject of an intransitive clause or of a transitive clauses without an expressed object. Thus, subject-like encoding could also in theory also have been selected. Moreover, only in NENA is the object-like encoding of a copula clause subject restricted to the third persons. The 1st/2nd person indeed have more subject-like encoding. These parallels are shown in (43). (43) Positional overlap of subject and object (=Theme) affixes and the enclitic copula in J. Dohok SUBJECT PREDICATE COPULA ʾaxoni dòxtor =ile brother-my doctor COP.PRS.he ‘My brother is a doctor.’ TRANS. VERB SUBJ.affix OBJECT affix k-xaz -ən -ne REAL-see -A.IM P.him ‘I see him.’ INTRANS. VERB SUBJ. affix g-zadʾ -ən ‘I fear.’ PREDICATE COPULA gurga wən 266 THE MORPHOLOGY AND SYNTAX OF THE COPULA wolf COP.PRS-IM ‘I am a wolf.’ Even synchronically, the L-copulas can be considered a type of object-like subject encoding in relation to realis present tense clauses. Object-like encoding of the referent in categorical copula clauses is attested in other languages, besides the Arabic dialects discussed in Retsö (1987). In dialects such as J. Dohok which employ the copula form le (contrasting with ile in other dialects), third person copula clauses are even formally analysable as having an ergative alignment. This is illustrated in (44); there, the 3MS subject of the intransitive clause is marked identically to the 3MS object of the transitive clauses in the realis present tense; thus aligning S (the subject of intransitive clauses) with P (the object of monotransitive clauses).29 (44) k-tor-á -le REAL-break-A.she -P.him ‘She breaks him.’ tóra =le bull =S.COP.PRS.he ‘It is a bull.’ 29 For the terminology of grammatical relations, see §9.1.2. SYNTAX AND MORPHOSYNTAX 267 A similar type of overt ergative marking of the third person is attested in a language isolate in Brazil, called Trumai (Guirardello 1999). In this language, the same third person enclitic -n/-e that marks objects of transitive verbal clauses also expresses subjects of non-verbal predicates in copula clauses.30 Such an analysis, however, does not hold for other dialects such as J. Amediya, where the L-copula occurs also in verbal copula clauses, including transitive ones (see §7.4.1 above). This leads to the alignment of S with A (subject of transitive) and P, all expressed with L-copulas (or L-suffixes). 7.13.3. The quasi-verbal basic copula (1st/2nd persons)31 As shown above, the basic copula in the 1st/2nd persons of the present realis realm in most NENA varieties resembles the verbal E-suffixes (cf. Cohen 2008; Khan 2006, 162). The segment [w], present in most dialects in this copula paradigm, is moreover reminiscent of the inflectional endings of the šaqəl paradigm of h/Ø-w- y ‘be’ (e.g. J. Dohok 2MS hawət), though the dialects of the Mosul plain such as C. Qaraqosh and C. Karamlesh, have [y] in place of [w], for instance iyət (2MS), iyən (1MS) and iyax (1PL) (C. Qaraqosh - Khan 2006, 162). 30 I thank Paul Noorlander for pointing me to this reference. 31 The term ‘quasi-verbal’ which I coined refers to the formal parallel between the inflection of ‘to be’ and the copula in NENA. 268 THE MORPHOLOGY AND SYNTAX OF THE COPULA Despite their formal correspondence with the verbal E-suffixes, these copulas (both of the iyət and the (i)wət type) are most likely not verb-derived, because the feminine singular endings do not match. This is illustrated for J. Dohok in Table 41 below: Table 41: 1st/2nd person copulas and inflection of h-w-y 'be' in J. Dohok copula šaqəl of h-w-y ‘be’ copula šaqəl of h-w-y ‘be’ 2MS wǝt haw-ət 2PL wetun haw-etun 2MS wat hoy-at (< *hawy-at) 1MS wǝn haw-ən 1PL wax haw-ax 1FS wan hoy-an (< *hawy-an) The [w] in the copula paradigm is thus most likely secondary (cf. Khan 2006, 162). It likely reflects a ‘verbalisation’ of this copula paradigm, and specifically making it parallel to the h-w-y šaqəl inflection. This, in turn, yields a partial convergence with Iranian, which—as previously noted—has a copula originally based on the inflection of the verb ‘be’. Thus, even though the process yielding the 1st/2nd person NENA copulas is distinct from the Iranian development path ‘to be’ > copula, both language families produce two paradigms which formally SYNTAX AND MORPHOSYNTAX 269 correspond: in both Kurdish and NENA, the inflection of ‘be’ and of the copula are similar. In many NENA varieties, this type of copula also has the segment [i], e.g. iwət. According to Khan (2006, 161), this segment could originate in a 3rd person singular enclitic pronoun, which has been generalised to the entire paradigm.32 This segment [i] is missing in many Jewish varieties, except for its relic in the negated form (e.g. le-wət > la-iwət). J. Dohok, J. Zakho (Cohen 2008, 47), J. Betanure (Mutzafi 2008a, 50) and the western-most TZ dialect of J. Arbel (Khan 2006, 159) have the form wət (2MS). An exception is J. Amediya, which has forms such as iwət (Greenblatt 2008, 391). In the transitional J. Barzani varieties, both forms exist and are distributed as follows: C-iwət, but V-wət (Mutzafi 2002, 54–5). 7.13.4. The presentative (‘deictic’) copula There are several distinct types of the presentative (‘deictic’) copula in NENA, the most common ones being hole, dule and wele and variants thereon (for overviews, see Khan 2006, 2018; Napiorkowska 2015a, 170-6; Cohen 2017, 2). There are also dialects with several presentative copulas; C. Diyana-Zariwaw possess three distinct deictic copula paradigm, enabling speakers to further differentiate between near and far deixis (see Table 42 below below). According to Khan (2018, 32 This idea is supported by the situation in Ṭuroyo, which has yo for the 3CS copula, and Neo- Mandaic, which has ye or ī. 270 THE MORPHOLOGY AND SYNTAX OF THE COPULA 243)(Noorlander 2017, 196–97), the copula hole/hawle comes from *ha-aw-le, DEIX-thisMS-him. This reconstruction would mean the historical presence of two cross-referencing elements (‘thisMS’ and ‘him’). The LD group is characterised by the wele/wəlle copulas, which are presumably related to each other (for details, see Table 42 below).33 Like the other presentative copulas in NENA, the LD copulas are most likely derived from a demonstrative pronoun or a presentative/deictic particle in combination with an L(L)-suffix, as shown in Table 42 below (Khan 2008b, 149; Noorlander 2017, 196-7). The copula wele/wəlle might also be related to the Bahdini presentative particle wa (this suggestion—to my knowledge—is not found in the existing literature), or at least have developed to resemble it formally. In a partial parallel to the L-copula, the object-like encoding of the subject in clauses with the presentative copula (e.g. 3MS we-le) is not always readily extended to the 1st/2nd persons. The person restriction of the wele/wəlle copula or its lack in the synchronic LD dialects is shown in Table 42 below. In J. Betanure, wəlle could apparently be elicited for all persons (e.g. 2MS wellox ‘here you are’; Mutzafi 2008a, 52). However, the published corpus contains no 1st/2nd person forms, suggesting that if they do occur, they are extremely infrequent. The Early J. Nerwa 33 J. Zakho also possesses the hole copula. Only wele is given by Cohen (2008 and 2017), but I have also come across hole in the informal, spoken register of J. Zakho in my fieldwork. In J. Dohok, hole also occurs – apparently in the fossilised 3MS form (hole), used with k-šaqəl as the marker of the progressive. SYNTAX AND MORPHOSYNTAX 271 has attestations of the 1st person presentative copula, but such forms take the E- set, e.g. we-wən COP.DEIX.-IMS ‘here I am’ (Cohen 2017, 2-3, citing data from Sabar 1976), which reflects a similar avoidance of the L-suffixes for the 1st/2nd person referents. Table 42: Presentative copulas and their properties across LD Dialect form persons used proposed reconstruction J. Dohok wele 3rd *wal-le? (independent particle wal attested) *wa-le (Kurdish presentative particle/demonstrative pronoun)? J. Zakho wele 3rd *ʾwa(ha)-le? ‘that (over there) (Cohen 2017) him’ J. Betanure wəlle, 3rd only? (cf. w-əlle and-LL-suffix; apparently (Mutzafi occasionally above) proposed because ʾəlle attested 2008b, 52, wele alongside the copula – 393) apparently with deictic force 272 THE MORPHOLOGY AND SYNTAX OF THE COPULA J. Amediya wəlle all persons (Greenblatt 2008, 394– 96) rd J. Challa wəlle, 3 only (Fassberg marginal apparently rd 2010, 102) wele (only 3 person examples given) Outside LD, full presentative copula paradigms (e.g. 3MS hole, 2MS holux) occur in many dialects. It is still noteworthy, however, that in dialects such as C. Barwar and C. Diyana-Zariwaw, the 1st/2nd person presentative copula forms are also inflectable with the E-suffixes, alongside L-suffixes. This also suggests a partial dispreference for the L-suffixes in those persons. The C. Diyana-Zariwaw copula paradigms are shown in Table 43 below. Table 43: Copula paradigms in C. Diyana-Zariwaw (Napiorkowska 2015a, 170–6) basic copula near-deictic far-deictic I far-deictic II (endings -a, - e) SYNTAX AND MORPHOSYNTAX 273 3MS ele dule hole wele 2MS et/ewət dut/duwət/dulux hot/hoyət wet/weyət CHAPTER 8 THE TENSE-ASPECT-MOOD SYSTEM 8.1. Introduction The aim of this chapter is to study the tense-aspect-modal functions (henceforth ‘TAM’) of the verbal forms in J. Dohok with references to other dialects. The NENA dialects possess a large inventory of verbal forms, the Aramaic verbal system having undergone radical innovations (cf. Stilo and Noorlander 2015; Khan 2021). Particular issues discussed in greater detail and in a comparative perspective are the semantic restriction and infrequent use of the constructions copula + Participle (wele šqila), and copula + in-Infinitive (wele bə-šqala) in J. Dohok. The NENA future form (J. Dohok b-šaqəl) is also shown to exhibit functional and syntactic restrictions. The J. Dohok imperfective realis present and past, respectively k-šaqəl and k-šaqəl-wa, are considered in a comparative and diachronic perspective. Such comparative and diachronic surveys help understand grammaticalisation pathways of the NENA verbal system, which, in turn, suggests which grammatical features in J. Dohok are conservatisms, and which innovations. Though many existing NENA grammars remain descriptive and synchronic, a diachronic study of the NENA verbal system could not only further illuminate, but perhaps even challenge existing grammaticalisation models (cf. §8.15 below). 276 THE TENSE-ASPECT-MOOD SYSTEM 8.1.1. Defining ‘meaning’: theoretical assumptions and methodology A general introduction on tense, aspect and mood is due here in order to avoid terminological confusion, and to ground the present chapter in the theory of verbal semantics. When studying the ‘meaning’ of a verbal form, one should differentiate between: • lexical semantics: a basic meaning of a verb shared across all the different morphological paradigms; • grammatical semantics: the meaning particular to a given morphological category of a verb, interpreted on the basis of the verbal form itself in combination with the syntactic and discourse context within which it occurs; • pragmatic information: information which is not derived from the given clause, but is inferred from the broader discourse context or even the social situation. 8.1.1.1. Grammatical meaning At the core of this chapter is the second type of ‘meaning’. However, in line with Function or Construction Grammar (Dik, e.g. 1997),1 it is assumed here that the 1 Thus, phrases such as ‘the form conveys the meaning of…’ in this chapter are used as shorthand for describing such syntactically cooperative meaning. For a useful summary of Function Grammar, SYNTAX AND MORPHOSYNTAX 277 meaning of a given verbal predicate emerges from the cooperation between the verb and the other arguments in the clause, rather than from the verbal construction only. Thus, attention is always paid to meaning of the whole clause and its type (i.e. main vs subordinate). Moreover, the meaning of a verb is often dependent on discourse parameters (cf. Kuty 2008, 155-172). The first of these parameters is genre, the relevant of which for the present investigation is narrative and descriptions relating to the past, and dialogue (cf. Cohen 2012). This distinction is crucial because in past narrative and descriptions, the default tense is usually the past. In dialogue, on the other hand, such a tense assumption does not exist, leading to a stronger need for explicit temporal reference in the clause. The difference between past narrative and past description (of traditions, lifestyle etc.) is that while the former is associated with perfective events, in the latter, the default aspect is habitual. The second discourse parameter used here is that of foreground vs background. All of these parameters are applied in the context when they are relevant. see Kuty (2008, 155-172). Being a top-to-bottom approach, Function Grammar focuses more on the meaning of the entire construction (usually termed ‘State of Affairs’), and is less concerned with recovering the ‘original’ or ‘underlying’ meaning of the verbal form itself, as is the case in other approaches to verbal semantics. For Construction Grammar, see Booij (2010). 278 THE TENSE-ASPECT-MOOD SYSTEM 8.1.1.2. The cooperation between lexical and grammatical meaning The two types of meaning listed above, lexical and grammatical, are to a large degree separate.2 For example, the inchoativity of a given predication (e.g. ‘stand up’) is sometimes due to the internal meaning of the verb, rather than due to the form in which it occurs. In the existing literature, this distinction between grammatical and lexical aspect is often termed ‘Aktionsart’ (e.g. Bhat 1999, 58). In other cases, however, verbal semantics interact with lexical meaning.3 For example, a particular verbal form is sometimes only attested with a particular semantic category of verbs. In this chapter, this is exemplified by the copula + Participle construction (J. Dohok wele šqila), which is typically only used with intransitive or possessive transitive verbs (cf. §8.14). In J. Dohok, therefore, the copula + Participle construction is resultative, rather than a fully-developed perfect. This, in turn, demonstrates that the meaning of verbal forms must be interpreted in light of the semantics of the lexemes that can occur within it, rather than by applying pre-existing functional labels. In other cases, the study of a form in its broader context reveals that a given verbal form conveys a meaning distinct 2 For a discussion on grammatical vs lexical meaning, see Comrie et al. (1976, 41–51). 3 This is also in line with Construction Grammar, which posits that ‘lexical and rule-based components of grammar are part of the same spectrum and can freely interact’ (Noorlander 2018, 31–3). In our case, these two components of grammar are, respectively, the internal lexical semantics of a verb, and the morphological with its various grammatical meanings. SYNTAX AND MORPHOSYNTAX 279 from that which is most commonly conveyed by the form in question. For instance, the k-šaqəl-wa form (§8.10) generally encodes past habituality, but in a group of experiencer-subject verbs of cognition and emotion, its primary function is to signal a stative interpretation, or reinforce a verb’s lexical stativity. Thus, grammatical semantics are sometimes dependent on lexical semantics. 8.1.1.3. The cooperation between grammatical meaning and pragmatic information So far, it has been proposed that the meaning of a verbal predicate is derived from the verbal form with the other arguments of the clause and from its discourse context, and may be shaped also by the lexical semantics of the verb used. In some cases, the context ‘cooperates’ with the verbal form to create a certain meaning (e.g. to show that a verb should be interpreted statively), while in others, it implies some information which is not encoded in the verbal form at all. In such cases, the verbal form in question is considered ‘unmarked’ for that piece of information (Comrie 1976, 111–122).4 In NENA, for instance, the Imperative is unmarked for aspect, so that only the broader context can demonstrate whether a form should be interpreted perfectively or imperfectively. 4 In this chapter, expressions such as ‘contextual interpretation’ ‘inferred from context’ or ‘implicated’ all refer to such context-derived information. 280 THE TENSE-ASPECT-MOOD SYSTEM With frequent usage, however, contextually implied meaning can become grammaticalised, that is, integrated within the semantics of the form due to an increased association of the two. This process is called entrenchment (Glynn 2010). This possibility of grammaticalisation implies, in turn, that there is not always a sharp dichotomy between pragmatic implication and grammatical meaning (Dahl 1985, 11), and the difference is especially difficult to determine for infrequent forms. A case in point is the šqəl-wa-le form attested a handful of times in the J. Dohok corpus, whose dialectal counterpart is typically described as anterior past. In the J. Dohok corpus, however the contextual interpretation is also perfect, raising a question whether the ‘perfect’ meaning is implied from context or grammaticalised. A similar question can be raised about the NENA predictive future (b-šaqəl in J. Dohok). Since NENA dialect typically lack a dedicated future imperfective, should b-šaqəl be considered aspectually unmarked? 8.1.1.4. The synchronic system as a point within a diachronic process From a diachronic perspective, a given synchronic system represents just a single stage within an ongoing diachronic process (cf. Bybee, Perkins and Pagliuca 1994), which in turn means that diachrony can be used for explaining the synchronic system. In this chapter, this is relevant for k-šaqəl (§8.12), b-šaqəl (§8.13), copula + in-Infinitive (wele bə-šqala; §8.15) and copula + Participle (wele šqila; §8.14). For instance, the future b-šaqəl in J. Dohok is functionally restricted: occurs only with predictive modality, rather than with other types of epistemic SYNTAX AND MORPHOSYNTAX 281 mood, except for specific types of clauses such as why questions. Syntactically, it is apparently restricted to main clauses. 8.1.2. Categories of tense, aspect and mood and their terminology Functional definitions of verbal forms are mostly based on Comrie (1985) for tense, Comrie (1976) and Dahl (1985) for aspect and Palmer (2001) for mood. Short definitions of specific TAM categories relevant for this chapter are given below: 8.1.2.1. Tense • Present: in highly simplified terms, this refers to a situation temporally overlapping with the time of reference (Comrie 1985, 36–41). • Past: a situation temporally prior to the time of reference • Future: a situation temporally future to the time of reference These temporal categories can be either absolute or relative. When the time is absolute, the reference time is the time of speech. When the time is relative, the reference time is the time of another situation.5 5 In Reichenbach’s model of three-way time reference, ‘absolute tense’ corresponds to cases of overlap between Reference and Speech time, while ‘relative tense’ corresponds to the case when Reference and Speech time are distinct and the designation ‘relative’ refers here to the relationship between the Event time and the Reference time (cf. van de Haan 2010, 446-7). 282 THE TENSE-ASPECT-MOOD SYSTEM 8.1.2.2. Aspect • Perfective: a situation whose internal temporal constituency is not expressed in the predicate (though see below for external complexity of perfective events in the section on iterative). Iterative: this term has various meanings in the literature. Here, following Bertinetto and Lenci (2012, 854-5) and in contrast to the earlier work by Comrie (1976) or Bybee, Perkins and Pagliuca (1994, 160), an ‘iterative event’ is defined as a perfective action which is repeated on several different occasions; for instance: ‘whenever I visited my home town, I went for a swim.’ Thus, while an iterative event is temporally complex, this complexity is external rather than internal.6 It is also distinct from—although related to—the imperfective category of habituality, which refers to an ongoing situation which may consist of events repeated over a period of time (see below). In a habitual form, the emphasis is on the ongoing ‘state’, and such a state is often conceived of as a property of an entity participating in it (e.g. ‘I used to swim as a teenager’ = ‘I was a swimmer’). In iterative forms, by contrast, the ‘emphasis’ is on the even itself 6 By contrast, in a semelfactive event, the repetition takes place on a single occasion, e.g. ‘cough’ or ‘knock’. In Comrie, the terms ‘iterative’ and ‘semelfactive’ are apparently synonymous (cf. Comrie 1976). In NENA, a similar meaning for the notion of ‘iterative actions’ is implicit in Napiorkowska (2015a, 340). SYNTAX AND MORPHOSYNTAX 283 (Bertinetto and Lenci (2012, 854-5). Needless to say, however, the dichotomy between habituals and iteratives is not always very sharp. • Imperfective: a situation with verbally-encoded internal temporal complexity Habitual: a situation which holds (or an event which is repeated) during an extended period of time. Crucially, such a situation is presented as a feature characterising a particular period of time and/or an entity participating in it (Bertinetto and Lenci 2012, 854-5). This means that not every form referring to a repeated situation is automatically habitual. Conversely, habituality as defined above does not require iterativity – it also subsumes the ‘generic’ aspect, that is states which persist over time. Habitual states, however, have to extend beyond the present tense (either actual present in dialogue, or a present tense relative to another event in narrative). For example, the verb ‘want’ in a sentence such as ‘do you want coffee or tea?’ is not habitual, since it refers to a state of short duration, roughly corresponding to speech time.7 Admittedly though, because states lack iteration, finding the boundary between ‘habitual’ (persisting) and ‘non-habitual’ states can prove challenging. 7 A parallel categorisation in NENA is found in Khan (2008b, 587) who for stative (cognition) verbs differentiates between ‘a continuous situation at a particular period’ (continuous in the past) and ‘a habitual property of the subject’. 284 THE TENSE-ASPECT-MOOD SYSTEM Gnomic: a situation which always occurs if necessary circumstances arise. Though related to habituality, gnomic situations are not restricted to a particular subject or a specific period of time, but rather are considered universally applicable, for instance ‘Hope dies last.’ Continuous: a situation (state or event) which is ongoing at the time of reference or at speech time; in the present, the continuous overlaps with the category of ‘actual present’. Progressive: a dynamic event which is ongoing at the time of reference; this category therefore excludes stative verbal forms. For example: ‘I heard the noise when I was swimming.’ 8.1.2.3. Mood • Realis: an event or state presented as factual • Irrealis: an event or state of uncertain factuality Epistemic: it expresses the speaker’s judgement about the factual status of a situation such as the likelihood of its occurrence, for example: ‘He might not come back again.’ In reality, epistemic modality is a continuous scale, with realis and completely unlikely events on the opposite ends of the spectrum (cf. Akatsuka 1985). SYNTAX AND MORPHOSYNTAX 285 Event: modality in which the realisation or non-realisation of a situation is dependent on someone’s will (the agent’s or someone else’s); this can be further subdivided into: Deontic: an externally imposed wish, obligation etc (e.g., a command, for instance: ‘you must not come back.’ Dynamic: an internal wish, obligation etc; for example: ‘I don’t want to come back.’ 8.2. Šaqəl Clauses with šaqəl convey various types of irrealis mood (epistemic and event). It is used in both main and subordinate clauses; in the latter, it is subordinated to a modal particle or a verb. For 1st person dynamic modality, both šaqəl and b-šaqəl are possible. The use of šaqəl for realis mood in main clauses—in contrast to other NENA dialects and to the J. Dohok past form šaqəl-wa— is unattested. 8.2.1. Main clause irrealis mood In expressions of epistemic modality, šaqəl typically occurs with an adverb, for instance: (1) 5 (dialogue) balki Ø-ʾaθe-Ø bə̀nhe.ˈ maybe IRR-come-he tomorrow 286 THE TENSE-ASPECT-MOOD SYSTEM ‘He may come tomorrow.’ In the 3rd persons, šaqəl expresses deontic wishes and obligations,8 as illustrated by the wish in (2): (2) 3c/22 wì!ˈ ʾilaha Ø-nàṭər-Ø,ˈ wat smə̀xta!ˈ Oh God IRR-protect-he, COP.PRS.youFS pregnant ‘Oh! God keep you, you are pregnant!’ Affirmative verbs conveying obligation occur subordinated to a deontic modal particle šud or lazəm (see the following section). By contrast, negative obligations as a general rule do not have a modal particle, as illustrated in (3). A bare negated šaqəl can thus express both obligations and wishes. (3) 1b/58-59 suw-un mùθ-unˈ xằˈ băle là Ø-hawe-Ø huðaya.ˈ IMP.go-youPL IMP.bring- youPL one but NEG IRR-be-he Jew ‘Go, bring me someone, but he should not be a Jew.’ In the 1st person, by contrast, šaqəl conveys a dynamic wish,9 though this can also be expressed with a modal verb with a complement. This interchangeability is 8 In the literature, this function is also known as jussive. 9 This function is also known as cohortative in the literature. For the difference between b-šaqəl in 1st person dynamic mood clauses, see section on b-šaqəl below. SYNTAX AND MORPHOSYNTAX 287 illustrated in example (3) below. By contrast, for the expression of deontic mood (typically obligation), a bare šaqəl form is unattested in the corpus, though such usage is reported for other NENA dialects. In J. Dohok, a deontic particle (e.g. gəbe, lazəm) is preferred in such cases. This inversion of šaqəl between 1st and 3rd persons (respectively, for dynamic and deontic wishes) likely reflects a bias towards the speaker as the prototypical agent (i.e. source of will).10 (4) 1c/32-33 ʾana g-ǝb-an Ø-ʾoð-an-ne qàðre,ˈ Ø-qaðr-an-ne u-Ø-ʿazm- ax-le kəs-lan l-bèθa.ˈ I REAL-want-IF IRR-do-IF feast IRR-host-A.IF-P.him and-IRR-host -A.we-P.him by-us to-house. ‘I want to make a feast for him, we shall honour him and host him at our house.’ 8.2.2. Subordinate clause irrealis mood Subordinate usage of šaqəl is more common than main clause usage,11 and šaqəl is the only form used following verbs, verboids and particles expressing deontic 10 First and second persons are prototypically more agent- than patient-like, in contrast to third persons; cf. §6 and Haspelmath (2015, 28). 11 Still, the fact that šaqəl does occur in main clauses suggests that its defining feature is mood, not its syntactic environment. This, in turn, prevents us from labelling the form ‘subjunctive’, though this terminology is used in some NENA grammars. 288 THE TENSE-ASPECT-MOOD SYSTEM modality. It is used with particles expressing obligation, typically expressed with the fossilised form gəbe (from the 3MS k-šaqəl form of ‘want’) and lazəm. It is also used for permission and wishes, following the particle šud (a fossilised form from the imperative from *šwuq d- ‘let that’), and with modal verbs such as ʾəbe ‘want’. For example: (5) 1c/101 lazəm Ø-yàʾe-Ø naša maṭo Ø-maḥk-e. Must IRR-know-he man how IRR-speak-he ‘A man must know how to speak.’ (6) 1c/44 ʾaz g-əmra šud ʾaθe-Ø b-lèle, xə̀ška.ˈ So REAL-say-she let IRR-come-he in-night, darkness. So she says ‘let him come at home at night, (when it’s) dark.’ In expressions of epistemic mood, šaqəl appears subordinated to verbs such as ‘know’ (yaʾe). 8.3. K-šaqəl K-šaqəl is formed by adding the habitual realis preverb g- to the šaqəl base. Most typically, k-šaqəl occurs in dialogue to convey the present tense. In J. Dohok, k- šaqəl is unmarked for a specific imperfective aspect, conveying all the various SYNTAX AND MORPHOSYNTAX 289 types of imperfectivity in the present: continuous (actual), habitual and gnomic. In many other NENA dialects, by contrast, the dialectal counterpart of k-šaqəl is associated with habituality, rather than with actual present. Occasionally, k-šaqəl in J. Dohok is used in past narrative to convey narrative present or past habituality. 8.3.1. Dialogue: continuous, habitual and gnomic The k-šaqəl form is used for the habitual present, paralleling the situation in most other NENA dialects. Some NENA varieties use the bare šaqəl form (identical to the irrealis form) while others—like J. Dohok—attach a habitual-realis preverb (typically ʾi- or k-). In several dialects (e.g. TZ varieties like J. Sanandaj, C. Barwar and C. Marga) use both the form with and without a preverb, depending on morphological or pragmatic factors. In J. Dohok, however, the k-šaqəl form is also the default way of expressing present continuous including the progressive, while many other dialects have dedicated progressive constructions.12 Since dedicated continuous form in NENA such as copula + in-Infinitive are—diachronically speaking—younger than k-šaqəl, the frequent use of k-šaqəl for progressivity in J. Dohok is in itself a conservative feature. In dialects such as C. Urmi, C. Barwar, C. Qaraqosh and J. Arbel, the new 12 See §8.15 below for dedicated continuous forms in NENA. J. Dohok does have a marked progressive and a continuous, but these are used extremely infrequently. 290 THE TENSE-ASPECT-MOOD SYSTEM ‘continuous’ constructions have even been extended into non-continuous domains (e.g. habitual present or even perfective past in the narrative; e.g., C. Urmi – Khan 2016, 185–200), following a common development path of verbal functions from actual present to a more general imperfective (Khan 2007b). K-šaqəl—in contrast to b-šaqəl—does not display any clause-type restriction, being attested in subordinate (e.g., relative) clauses. The use of k-šaqəl for various aspectual types of the present tense are shown in (7)-(9) below. Continuous (7) 3d/63 wi g-əmr-i mằ g-oð-at rəš gare?ˈ ṭamằ yĭsəq-lax rəš gare?ˈ oh REAL-say-they what REAL-do-youFS on roof why PERF.ascend- youFS on roof ‘Oh’, they say ‘what are you doing on the roof? Why did you go up on the roof?’ Habitual (8) 3a/23 Bǝt HʿaṣmaùʾtH g-raqd-i go HkvišimH…ˈ bujàlde.ˈ On independence REAL-dance-they in roads… roads (it is) on Independence Day (that people) dance on the roads… roads. SYNTAX AND MORPHOSYNTAX 291 Gnomic (9) 1c/91 šwir-ət xàbraˈ là k-eθe-Ø nšaya.ˈ šwir-ət də̀rbaˈ naša g-našè-Ø-le.ˈ wound-GEN word NEG REAL-come-he INF-forget wound-GEN blow man REAL-forget-A.he-P.him. ‘A wound (caused) by words will not be/is not forgotten. A wound (caused by) a blow one forgets.’ 8.3.2. Past narrative and description 8.3.2.1. Habitual past Though past habituality is typically expressed through the k-šaqəl-wa form (i.e. with the past-shifting wa morpheme), a few cases of k-šaqəl for past habituality are also attested. This form explicitly marks the aspect (imperfective), while the discourse context (past narrative) specifies that the temporal interpretation of the form is past. With only a few forms attested, it is not possible to identify a particular (discourse) function vis-à-vis the more frequent k-šaqəl-wa forms. Sometimes, k-šaqəl forms occur at the end of a chain of k-/b-/-Ø-šaqəl-wa verbs (e.g. (10) below), probably because the temporal value is apparent from the verbs at the beginning of the 292 THE TENSE-ASPECT-MOOD SYSTEM chain. Still, chains which use k-šaqəl-wa (sometimes with b-šaqəl-wa or šaqəl-wa) throughout are also commonplace. (10) 5b/65 u-gyatwì-wa,ˈ g-zamrì-wa,ˈ k-əxlì-waˈ u-k-šate-Ø (and-REAL-drink-they),ˈ u- ʿằraqˈ u-mằ ʾāt gəbat.ˈ ‘(At weddings,) they used to sit, sing and eat, drink – arak and whatever you’d like.’ 8.3.2.2. Narrative present In the context of past narrative, k-šaqəl can also be used as narrative (i.e. ‘historical’) present, though this is infrequent in the J. Dohok corpus.13 This use of k-šaqəl takes advantage of the general association of this form with present tense and imperfectivity in order to place the listener ‘in the midst of the narrative’. It thus arguably expresses ‘discourse present and imperfectivity.’ This is parallel to 13 There are also past narratives which are almost entirely narrated in the k-šaqəl form. In such cases, it is apparently assumed that given the genre (e.g., a folk tale), the listener knows that the temporal reference is past, and narrates the story entirely as if the events were unfolding in the here-and-know. This—in parallel to cases discussed in this section—also has the effect of enlivening the narrative, but this is applied to the entire story, and not simply to discourse- significant events. SYNTAX AND MORPHOSYNTAX 293 many other languages, where a present form is used to ‘enliven the narrative’ (cf. Coghill 2009, 270). In example (11) below, the narrator switches from šqəlle (prototypically past perfective) to k-šaqəl when retelling the first moment of the encounter with the long-awaited and dreamt-about Land of Israel, apparently in order to make this memorable moment more vivid: (11) 3a/2-4 θè-lan (PERF.come-we) rušalajim,ˈ ʾara mqudàšta,ˈ mkoyǝfè,ˈ k-eθ-ax (REAL- come-we) rušalàjim. ‘We came to Jerusalem, the holy land, joyful, (finally) we are coming to Jerusalem.’ g-maṭ-ax (REAL-arrive-we) l-Hšaʾar ha-ʿaliyàH,ˈ let-lan ʾĭxala la štàya.ˈ ‘We arrive at the Aliyah Gate – we have neither food nor drink.’ g-maṭ-ax (REAL-arrive-we) saʿa ṭlaha b-lèle.ˈ ‘We arrive at three at night.’ In the sentence in (12), k-šaqəl overlaps with the beginning of a new story unit: (12) 1c/71-3 zəl-le (PERF.go-he) l-ṭùra.ˈ ‘He went to the mountain.’ 294 THE TENSE-ASPECT-MOOD SYSTEM durdət yom q-qayəm-Ø (REAL-get_up-he) măre bèθa,ˈ g-ezəl-Ø (REAL-go-he) ta ṣìwe,ˈ ‘The next day the father of the house gets up, goes to (get) wood.’ gurga ži θè-le (PERF.come-he)… ‘(Then) the wolf came…’ The existence of the narrative present is not discussed in the grammars of the LD dialects of J. Challa, J. Betanure and J. Amediya (respectively Fassberg 2010; Mutzafi 2004; Greenblatt 2008). The narrative present apparently does not exist in the narrative genre of the J. Zakho corpus (Cohen 2012, 434). 8.3.2.3. Stative past K-šaqəl in past narrative also occurs with two experiencer-subject verbs of cognition and emotion yaʾe ‘know’ and ʾəbe ‘want’, where it does not primarily convey aspect (habituality) or the narrative present, but rather stativity. This phenomenon is related to the use of k-šaqəl-wa in those same experiencer-subject verbs in the past, discussed in more detail in §8.10.2 below. When k-šaqəl is used, the past temporal value has to be inferred from the context. (13) 1b/53 SYNTAX AND MORPHOSYNTAX 295 htòvhˈ. g-əb-e Ø-maxəlṣ-i b-gyànu,ˈ zun-nu ṭa- le šoʾa lʾìne…ˈ good REAL-want-they IRR/FUT-save-they in-themselves, PERF.buy-they for-him seven jugs ‘Well then. They wanted to save themselves, (so) they bought for him seven jugs…’ 8.4. Šaqəl and k-šaqəl for irrealis mood in past purpose clauses In purpose clauses following a predication in the past, tense marking on the subordinated verb is often omitted as illustrated in (15) below. Thus, the prototypically present irrealis šaqəl form occurs in such contexts alongside the prototypically irrealis past šaqəl-wa form. (14) 5b/34-35 xə̀tna u-HḥaverìmH dide k-eθe-wa l-beθa p-šaql-ì-wa-la,ˈ dət Ø-ʾoð-i Hševa braxòtHˈ groom and-friends his REAL-come-PST to-home FUT-take-A.they-PST- P.her, that IRR-do-they seven blessings. ‘The groom and his friends would go to her house and take her (i.e. the bride) to have the wedding ceremony.’ More unusually, however, the prototypically present realis k-šaqəl form is also encountered; it is attested in purpose clauses as a complement to verbs of 296 THE TENSE-ASPECT-MOOD SYSTEM movement in cases when the subject is the same in both verbal forms. This is despite the fact that such clauses have irrealis modal value, expressing deontic intention. K-šaqəl in such clauses cannot be interpreted to convey habitual past, since the sentences in question refer to a single event. Such sentences are likewise unlikely to convey narrative present, since they’re bound within one intonation unit with the preceding verb of movement, suggesting subordination. This is illustrated in (15) below: (15) 3a/24 kuš-lan xapča g-ràqd-ax,ˈ θe-lu trumbele mən ṣùba…ˈ PERF.descend-we little REAL-dance-we PERF-come-they cars from Tzuba ‘We went down to dance a little, (then some) cars from Tzuba came…’ 8.5. B-šaqəl B-šaqəl is formed by adding the future preverb b- (devoiced to p- before voiceless consonants) to the šaqəl base. This preverb, however, is only clearly audible in Pattern I verbs. In other patterns, all of which start with m, it has assimilated to m (e.g. *b-mbašəl > m-mbašəl ‘he will cook’), but the gemination of m cannot be clearly discernible (so the form is realised mbašəl). Thus, the present study takes into account only Pattern I verbs, though there certainly are many more underlying b-šaqəl forms. Determining whether a given Pattern II, III or IV form SYNTAX AND MORPHOSYNTAX 297 should be underlyingly b-šaqəl or šaqəl would require a great deal of speculation; future and irrealis mood are intrinsically linked, so determining which forms are marked for futurity (b-šaqəl) and which are not (šaqəl) would require a great deal of speculation.14 B-šaqəl is most commonly used in dialogue for predictive future, that is, high- certainty epistemic mood. Generally, therefore, it is not employed for lower- certainty epistemic mood with future time reference such as possibility or uncertainty, except for ‘why’ and ‘either—or’ clauses. Its occasional overlap with event mood (1st person dynamic modality) is considered here contingent on the category of prediction. B-šaqəl does not occur in most types of subordinate clauses. This is apparently due to its yet incomplete grammaticalisation and in particular, due to its persisting association with the domain of pragmatic assertion (see further §8.13 below). According to Khan (2021), b-šaqəl and its past counterpart b-šaqəl-wa are ‘discourse-depended’. This is to say, such forms are not necessarily (or no longer) dependent in the narrow syntactic sense of being subordinate. Still, they retain a ‘looser’ dependence on their discourse context. For instance, a b-šaqəl form can be dependent temporally (which in Khan’s definition appears to be synonymous to the concept of ‘relative future’), modally (i.e. its occurrence is conditioned by the 14 See Khan (2016, 212–3) for the methodology of determining the underlying future form in C. Urmi. 298 THE TENSE-ASPECT-MOOD SYSTEM presence or absence of another event) or even ‘logically’ (e.g., by a relationship of logical consequence). 8.5.1. Dialogue 8.5.1.1. Future tense of predictive (epistemic) mood By far the most common use of b-šaqəl is to indicate the future tense in the main clause. In J. Dohok, b-šaqəl is apparently unmarked for aspect; imperfective use can be elicited, but is not attested in the corpus. In terms of modality, b-šaqəl typically expresses prediction, as illustrated in (16). By definition, all types of future are irrealis (Palmer 2001, 104–5), but some are more certain than others. ‘Prediction’ is a high-certainty type of epistemic mood, so on the continuum of epistemic modality, it is close to the category of realis. (16) 1c/48, 50 dammət payəš-Ø (IRR-stay-he) xə̀ška,ˈ šud Ø-ʾaθe-Ø (LET IRR-come-he).ˈ ‘Let him come when it gets dark.’ b-aθe-Ø (FUT-come-he) kəslan beθa u-b-àzəl-Ø (FUT-go-he).ˈ čuxxa la k- xāzèle.ˈ ‘He will come to our house and go. No one will see him.’ In some NENA grammars, the dialectal variants of b-šaqəl are said to also express dynamic modality in 1st person agentive verbs, that is, a speaker’s will to undertake SYNTAX AND MORPHOSYNTAX 299 an activity.15 However, in many dialects, including J. Dohok, b-šaqəl is not used for all 1st person dynamic modality. As shown in example (4) above, for instance, ‘I want (+future event)’ is generally expressed not with b-šaqəl, but rather through ʾəbe ‘want’ + šaqəl, or simply šaqəl. This is because such predications are modally primarily dynamic, not predictive (epistemic). While the clause in (4) above expresses a desire, its fulfilment is contingent on the cooperation of other individuals and is thus uncertain. This sentence comes from a story in which a wolf is being invited to the speaker’s house, but is very reluctant to come. Later in the story, when the plan is becoming more certain and the future is therefore predictive, the b-šaqəl form is used for that same future eventuality (17) Ic/51 baxti b-oð -a-lox xa-ʿazime ḅāš̀ .ˈ wife-my FUT-do -A.she-BEN.you a-feast good ‘My wife will make a great feast for you.’ Thus, it seems more appropriate to consider b-šaqəl and its dialectal parallels primarily a predictive future form. This will automatically also include 1st person dynamic modality verbs which are of higher certainty and thus ‘predictable’. This definition of b-šaqəl fits the attested 1st person verbs, which refer to situations 15 For example, C. Barwar (Khan 2008b, 598–602), or C. Diyana-Zariwaw (Napiorkowska 2015a, 352); in both of these, this function is described as deontic wish or intention. 300 THE TENSE-ASPECT-MOOD SYSTEM whose fulfilment is only dependent on the speaker, so the predictive and dynamic meanings overlap: (18) 1a/24 Context: ‘There was a certain Jew in the market. He says:’ nàbl -ū -li,ˈ ʾàna b-amr -ǝn -ne ʾékē=la palgut dunye.ˈ IMP.take -A.youPL -P.me I FUT-say -A.IM -R.him where=COP.PRS.she middle-GEN world ‘Take me, I will/shall tell him where the middle of the world is.’ 8.5.1.2. Conventionalised habitual present (of Ø-m-r,ʾamər ‘say’) With the verb ʾamər ‘say’, b-šaqəl is used for conventionalised ways of responding (by speech) to a given situation, which is illustrated in (19) below. The reason for the use of the prototypically future b-šaqəl is apparently because such situations are conceived of as future to or causally contingent on an implied circumstance in which the predication holds (i.e. ‘discourse-dependent’ in Khan’s analysis, see above).16 Arguably, however, the dependence is very loose here: the circumstance 16 See Hoberman (1989, 140), who describes this function of b-šaqəl in J. Amediya as a type of conditional and habitual: ‘(such clauses are) hypothetical situations’ stemming from a ‘timeless (generic) possible condition’. SYNTAX AND MORPHOSYNTAX 301 on which the predication is contingent is in fact not stated explicitly in the discourse. In theory therefore, one might also expect in such context k-šaqəl, emphasising the habitual or gnomic nature of such events. (19) 1e/2 u-tama b-àmr-axˈ mən yom-ət fhə̀m-li,ˈ ʾəsrà šənne qăbəl.ˈ and-there FUT-say-we from day-GEN PERF.understand-I ten yearS before. ‘And there – (as) we say – “from the day I became aware” – ten years earlier.’ 8.5.1.3. Modal future: ‘why‘ and ‘either-or’ clauses The predominant predictive use of b-šaqəl notwithstanding, in certain syntactic environments, this form also conveys irrealis (low-certainty epistemic or event) mood. First, it is used in all but one of the eight attested ‘why’ clauses; for instance: (20) 5a/17-8 Context: ‘The Kurds would say to the Jews “if not today or tomorrow, one month you will leave your houses and go.”’ ʾaz bate-tun ṭamắ b-zon-àx-lu ṭamắ b-yāw-áx-loxun pare?ˈ so houses-yourPL why FUT-buy-A.we-P.them why FUT-give-A.we- R.you money 302 THE TENSE-ASPECT-MOOD SYSTEM ‘So your houses – why would/should we buy them? Why would we give you money?’ B-šaqəl forms are also typically used in ‘either-or clauses’ (which usually express epistemic modality): (21) 3c/16 har ʾaql-e b-qaṭʾi-la…ˈ ʾan b-ṭaṛṣ-a ʾàn b-qāṭ- ax-la.ˈ just leg-his FUT-cut-A.we-P.her either FUT-heal-she or FUT-cut- we-her. ‘We just amputate his leg. (That is,) it either heals or we amputate it.’ 8.5.1.4. Modal relative future: apodosis of real conditional clauses B-šaqəl is typically used in the apodosis of conditional clauses. Such clauses convey futurity which is relative to the predication in the protasis, meaning that the function of b-šaqəl is to convey this relative futurity. A more comprehensive study of conditional clauses can be found in Cohen (2021), who studies the J. Zakho dialect. The predication in the protasis, by contrast, is almost always expressed through the prototypically irrealis šaqəl form. For example: (22) 1c/68 SYNTAX AND MORPHOSYNTAX 303 ʾēn là Ø-māx -ət -te năra go reš-iˈ ʾana b-axl - ə̀n -nox.ˈ if NEG IRR/FUT-hit -youMS -him axe in head-my I FUT-eat - A.IMS -P.youMS ‘If you don’t hit me with the axe on my head, I will eat you.’ 8.5.1.5. Modal future: protasis of real conditional clauses (vividness effect?) Even though predication in the protasis is most often expressed with šaqəl, one b- šaqəl form is also attested in a folktale. In this sentence, the wolf is describing what will happen if he appears in town: (23) 1c/42 b-aθ-ən go maθa kull-u Ø-mbàrbəʿ-i.ˈ FUT-come-IM in town all-them IRR/FUT-be_alarmed-they ‘(If) I come to town, everyone will be alarmed.’ B-šaqəl likely functions here to create the discourse effect of ‘realis’ mood, that is, to paint the situation more vividly. We may thus paraphrase: ‘Imagine me coming to town: everyone will be alarmed.’ 304 THE TENSE-ASPECT-MOOD SYSTEM 8.5.2. Past narrative 8.5.2.1. Relative future Very occasionally, b-šaqəl is used within a narrative sequence to convey futurity which is relative to the previous verb, as is illustrated by (24) below. Note that the two verbs are separated by an intonation boundary and not linked by the subordinating d-. Thus, b-qaṭe (b-šaqəl) should not be considered subordinate to the previous verb, meaning that it does not express dynamic modality (i.e. ‘he went in order to cut wood’). In this case, we would expect a šaqəl(-wa) form. In the present case, the focus is thus on the relationship of time, and not of intention. The J. Dohok corpus contains no clear cases of subordinated b-šaqəl forms. (24) 1c/11 xă yoma zǝl-le l-ṭùṛa,ˈ b-qaṭ-e ṣìwe.ˈ one day PERF.go-he to-mountain FUT-cut-he wood ‘One day he went to the mountain, he was going to cut wood.’ 8.6. La k-šaqəl La k-šaqəl has more functions than its affirmative counterpart k-šaqəl, and it is this additional functions which is discussed here. SYNTAX AND MORPHOSYNTAX 305 8.6.1. Negated predictive future La k-šaqəl—in parallel to the affirmative form b-šaqəl discussed above—expresses predictive future: (25) 1c/50 b-aθe-Ø kəs-lan beθa u-b-àzəl-Ø.ˈ čŭxa la k-xāzè-Ø- le. FUT-come-he by-us home and-FUT-go-he nobody NEG REAL-see- A.he-P.him He will come to our house and go. No one will see him. The use of la k-šaqəl in its dialectal variants peculiar to NENA in the Bahdini area, and parallels the situation in Bahdini Kurdish. This points to contact with Kurdish as the source of this phenomenon in NENA, or perhaps to a shared Aramaic- Kurdish convergence.17 There is also a possible pragmatic factor, relating to the fact that negated clauses make a weaker assertion. For a further discussion on this, see §8.13 below. 17 More specifically, it is likely a case of pattern borrowing. In Matras's terminology (2009: 235– 39), ‘pattern borrowing’ refers to the cross-linguistic transfer of ‘syntactic patterns’, which includes various formal features of syntax (e.g. word order), as well as their various semantic and discourse- pragmatic functions; it also includes processes such as calques and grammaticalisation paths. 306 THE TENSE-ASPECT-MOOD SYSTEM 8.7. Šqəlle (and qam-šaqəl-la) Šqəlle prototypically conveys perfective past in both dialogue and past narrative. The form qam-šaqəl-la is its transitive counterpart, used with high-prominence objects (see §8.2.2 and §8.4). Thus, the subsequent description of šqəlle is also pertinent to qam-šaqəl-la. 8.7.1. Perfective past (narrative foreground, dialogue) Since šqəlle expresses perfective past, it is the most frequently used form in narrative foreground. A characteristic narrative device is to set the scene with imperfective forms such as (k-)šaqəl-wa or k-šaqəl, and to subsequently switch to šqəlle, indicating the opening of the main story line. This is illustrated the beginning of the folktale ‘The Wolf’: (26) 1c/1,11-12 ʾǝθwa xa-beθa d-Ø-ʿāyə̀š-Ø-wa (IRR-live_off-he-PST)…ˈ ‘There was a household who used to live off…’ xă yoma zǝl-le (PERF.go-he) l-ṭùra,ˈ b-qate ṣìwe,ˈ ‘One day he went to the mountain’ xze-le (PERF.see-he) xa-gùrga. ‘And saw a wolf.’ SYNTAX AND MORPHOSYNTAX 307 8.7.2. Relative past in clauses with future time reference (including protasis of real conditionals) Occasionally, a šqəlle clause conveys pastness which is relative to another past event. This is also attested in the protasis of real conditional and is illustrated in (27) below. The use of šqəlle in such contexts, however, is by no means obligatory; for instance, examples (27) and (28) refer to the same future eventuality, but the former uses šqəlle, and the latter šaqəl. The subtle difference between these two sentences regards the information that is encoded in the verbal form, and what is implicit in the context: šqəlle explicitly encodes the relative tense (past), but not absolute tense or irrealis mood, while šaqəl only expresses explicitly irrealis mood and absolute non-pastness. (27) 3f/22 hakar ʾana ʾùd-li d-la ləbb-ox,ˈ baxt-i Ø-hoy-a mṭolàq-ta.ˈ if I PERF.do-I without heart-yourMS wife-my IRR-be-she RES.divorce-FS (The king said) ‘if I did something against your wish (lit. against your heart), my wife shall be divorced.’ 308 THE TENSE-ASPECT-MOOD SYSTEM (28) 3f/20 hakar Ø-māx-ə̀t-tiˈ Ø-mṭaḷq-ət-ta bàxt-ox.ˈ if IRR-hit-youMS-me IRR-divorce-A.youMS-P.her wife-yourMS ‘(The Jew said) if you beat me, you will divorce your wife.’ 8.8. Šqəl-wa-le Šqəl-wa-le is formed by infixing the past-shifting wa morpheme between the past base and its inflectional suffix. The primary function of this form is to convey pastness relative to another situation in the past; such a function is common in other dialects (e.g. C. Barwar - Khan 2008b, 109; J. Zakho - Cohen 2012, 414). Khan points out that the primary function of this form is not past perfect (see Khan 2008b, 109). At the same time, the domains of past perfect and relative pastness are doubtless bound very closely together. In other words, an event in the anterior past is likely to be menioned precisely because its effects were still relevant at the time when another past situation obtained. As a result, the NENA šqəl-wa-le form is likely associated with a past perfect meaning, at least by pragmatic implication. This overlap of anterior past and perfect is apparent from the šqəl-wa-le examples attested in the J. Dohok corpus, for example: (29) 1e/20 soti wawa b-kə̀slan.ˈ sawoyi mə̀θ-wā-le (PERF.die-PST-he)ˈ u-babi wewa brona dida bàs…ˈ ‘My grandmother was (staying) at our house. My grandfather had died and my father was his only son…’ SYNTAX AND MORPHOSYNTAX 309 8.9. Šaqəl-wa Šaqəl-wa is formed by adding the past-shifting wa morpheme to the prototypically irrealis šaqəl base. Šaqəl-wa expresses past irrealis mood, but sometimes also realis in the past. In fact, šaqəl-wa is the only attested form expressing realis mood in past tense subordinate clauses. By contrast, realis use of the non-past counterpart šaqəl does not occur in J. Dohok. A methodological point is due here. It is not certain whether all forms which are pronounced by speakers as [šaqəlwa] also have the šaqəl-wa morphology on the underlying level. As mentioned in the section on b-šaqəl above, Pattern II, III and IV b-šaqəl(-wa) forms are realised [šaqəl(wa)]. This in turn means that Pattern II, III and IV forms which are realised [šaqəlwa] can on the underlying level be either šaqəl-wa or b-šaqəl-wa. This overlap is especially problematic in the past, where b- šaqəl-wa (and to some extent also šaqəl-wa) also function in the (habitual) realis domain, so their realis meaning does not help determine the underlying morpheme. In the present, the functional domains of šaqəl, b-šaqəl and k-šaqəl are more distinct. Thus, the interpretation of a form realised [šaqəlwa] in the past which conveys a realis meaning is challenging. As shown above, only in Pattern I verbs can the underlying form be established with relative certainty. Thus, by and large, only Pattern I forms are considered in this study to ensure an accurate analysis. Unfortunately, this results in exclusion of a large group of verbs. 310 THE TENSE-ASPECT-MOOD SYSTEM 8.9.1. Irrealis Šaqəl-wa is the default way of expressing irrealis mood in the past. It is attested, inter alia, in complement clauses conveying event modality (30) and in purpose clauses (31): (30) 1f/17-18 u-kùll-u ži g-əb-e-wa Ø-ʾāθ-e-wa.ˈ and-all-them indeed REAL-want-they-PST IRR-come-they-PST ‘And all (the Jews) wanted to come (to Israel). (All the time they had thought “we are not (from) here (i.e. from Iraq), we will go to our own county.”)’ (31) 1e/25,27 u-gə-mzābə̀n-Ø-wa-lu,ˈ d-Ø-ʿāyəš-Ø-wa mən… ʾə̀b-bu.ˈ And-REAL-A.sell-P.he-PST-them REL-IRR-live-he-PST from in- them. ‘And he would sell them [the grapes] to make a living from… through them.’ SYNTAX AND MORPHOSYNTAX 311 8.9.2. Realis Šaqəl-wa is occasionally used also with realis meaning, in both main and subordinate clauses. Sometimes, the realis preverb is missing at the end of a verb a series of verbs, in which case habituality and realis mood is inferable from the preceding k-šaqəl-wa forms. In other cases, however, šaqəl-wa is the first verb of a series, as shown in (32) below. (32) 5a/5-6 Context: ‘In Dohok, there was a place by the river.’ Ø-yatw-ì-waˈ gə-mbašl-ì-wa,ˈ g-oð-i-wa pəsra rəš nùra…ˈ IRR-sit-they-PST REAL-cook-they-PST, REAL-do-they-PST meat on fire… ‘(There,) they would sit, cook, make barbeque…’ Šaqəl-wa is also attested in subordinate relative clauses, including those modally realis, as shown in (33). The state of affairs in such clauses overlaps temporarily with the situation in the main clause. In non-past contexts, temporally overlapping relative clauses use k-šaqəl forms. In the past, by contrast, only šaqəl-wa (and not k-šaqəl-wa) is attested in such relative clauses in the corpus. (33) 1b/39 xa wewa huðaya go šùqa,ˈ Ø-ṭāləb-Ø-wa (IRR-ask-he-PST) HnedavòtH,ˈ Ø- ʿāyəš-Ø-wa (IRR-live-he-PST) bət HnedavòtH.ˈ 312 THE TENSE-ASPECT-MOOD SYSTEM There was Jew in the market (who) used to beg, (who) used to live off the alms.’ Unfortunately, due to the exclusion of all Pattern II, III and IV verbs from this study, the exact frequency of this phenomenon cannot be determined. Even impressionistically, however, it is clear that the realis use of šaqəl-wa is significantly more common than that of šaqəl for main clause realis, which is virtually unattested. Overall, therefore, šaqəl-wa is used more widely than its non- past counterpart šaqəl: while the latter is largely restricted to the realm of irrealis, šaqəl-wa also occurs for realis past. This most likely reflects a grammaticalisation in progress: the k- preverb—which originated in the present domain—has not yet become obligatory in all contexts of past realis. This is further discussed in §8.12 below. 8.10. K-šaqəl-wa K-šaqəl-wa is formed by adding the habitual-realis preverb k- and the past-shifting suffix wa to the šaqəl base, and prototypically denotes the habitual aspect in the past. In the vast majority of cases, this form occurs in past narrative. It is used especially often in narrative backgrounds and in in descriptions of lifestyle, traditions and customs within the Dohok community, as such descriptions refer to situations that characterised long periods of time. It is also used with a few cognition and emotion verbs to indicate their stative, i.e. non-inchoative, meaning. It is possible to regard some cases of k-šaqəl-wa as iterative, rather than habitual. SYNTAX AND MORPHOSYNTAX 313 In J. Dohok, however, there is apparently no dedicated iterative form. There is one context where k-šaqəl-wa denotes the continuous aspect in the J. Dohok corpus. As for the other LD dialects, in. J. Amediya, k-šaqəl-wa is said to carry the broader meaning of past imperfective.18 In the narrative of J. Zakho, this form as apparently associated with kud-yom ‘every day’, that is, presumably, with specifically with habituality, rather than imperfectivity in general. In dialogue, it is said to have a broader imperfective meaning (‘non-specific, repetitive or ongoing action’; Cohen 2012, 422–4). In NENA in general, habitual past is expressed by two forms: one has a dedicated habitual realis marker (e.g. in C. Barwar, ʾi-qaṭəl-wa, largely parallel to J. Dohok k-šaqəl-wa), the other one covers both habitual and irrealis past (e.g. C. Barwar qaṭəl-wa, formally parallel to J. Dohok šaqəl-wa). In C. Barwar, however, the former is the ‘marked’ (i.e. used for discourse-prominent situations). This comparative discussion is continued in §8.12 below. In J. Dohok, k-šaqəl-wa is the most common form for the expression of past habituality, though šaqəl-wa and b- šaqəl-wa also occur in this function. 18 In Hoberman’s language, k-šaqəl-wa has ‘the narrow meaning of [-Øtotal]’, while šqəlle has the meaning of [Total] (Hoberman 1989, 128-9). 314 THE TENSE-ASPECT-MOOD SYSTEM In the Early J. Nerwa corpus, (k-)šaqəl-wa is commonly used for past continuous (including the progressive), that is, for situations overlapping with another event. In many contemporary dialects such as C. Barwar, a new progressive/continuous form has developed (see below), replacing k-šaqəl-wa in this narrow imperfective function. 8.10.1. Habitual past (narrative background & descriptions of lifestyle narrative) Being prototypically habitual, k-šaqəl-wa is the most common form used for describing general states of affairs at the start of a narrative, or for providing supplementary information on general situations in the course of a narrative, for example: (34) 1d/2-3 go turkya ʾud-lu farman…ˈ In-Turkey PERF.do-they decree ‘In Turkey, they passed a decree…’ qamàye kull-a ʾaya mənṭaqa dət Hmizráḥ tixónM ži wawa HtùrkimH,ˈ tùrkya Hk-šalṭa-waH.ˈ before all-it thisFS region GEN Middle East indeed COP.PST.she Turks, Turkey REAL-rule-she-PST SYNTAX AND MORPHOSYNTAX 315 ‘Earlier, all of this region of the Middle East was Turkish, Turkey used to rule.’ mupəq-la farman ʾəl-suraye go kull-a mənṭaqa dida Ø-qaṭl- i-luˈ PERF.go_out-she decree on-Christians in all-her region her IRR- kill-A.they-P.them. ‘A decree went out about the Syriacs in all of (Turkey’s) region and in Armenia (that) they would kill them.’ A similar habitual usage of k-šaqəl-wa is attested in descriptions of lifestyle (alongside b-šaqəl-wa and šaqəl-wa), as illustrated by the opening section on life in Dohok in the past: (35) 1e/13-13 tama kull-u k-xāye -Ø-wa mən šul-ət ʾiðàθ-u.ˈ there all-them REAL-live -they-PST from work-GEN hands-their ‘There, everyone used to live by manual labour.’ ʾəθ-wa…ˈ g-oð-i-wa čammàne.ˈ EXIST-PST REAL-do-they-PST orchards ‘There were… they would plant orchards (and there were people who had a craft...)’ 316 THE TENSE-ASPECT-MOOD SYSTEM K-šaqəl-wa is even used to refer to an imaginary individual who represents a community characterised by a given state of affairs. In such cases, the habitual aspect of k-šaqəl-wa cannot be understood as referring to those imaginary individuals, since in some cases, those individuals participated in a given activity only once. Rather, habituality applies to the whole community represented by those individuals. At most, the activity could be iterative if the prototypical individual took part in an event more than once. For instance, sentence (36) below contains the form də-g-napl-ì-wa REL-fall-they-PST, which out of context would be translated ‘those who used to fall’. However, this translation does not sound natural, even if one wanted to emphasise that the situation no longer holds: the sentence in question does not mean that there were people used to fall repeatedly, but rather that many cases of falling took place within the group in general. The translation ‘those who would fall’ is more suitable, indicating that this usage of k- šaqəl-wa is closer to the English would in such cases: (36) 3c/2 bàbiˈ k-eθe-Ø-wa kəs-le ʾana də-g-napl-ì-wa…ˈ father-my REAL-come-they-PST by-him those REL-REAL-fall-they-PST ‘My father – those who had fallen (lit. used to fall) would come to him (with their legs or arms broken)…’ k-eθ-e-wa kəs bab-i u-awa g-ewə̀ð-Ø-wa-lu ṭa-lu.ˈ SYNTAX AND MORPHOSYNTAX 317 REAL-come-he-PST by father-my and-he REAL-do-A.he-PST-P.them for- them ‘…they would come to him and he would cure (lit. do) them for them.’ 8.10.2. Stative past in the verbs yaʾe ‘know’ and ʾəbe ‘want, to love’ In a few experiencer-subject verbs of cognition and emotion, k-šaqəl-wa apparently does not primarily encode aspect (i.e. habituality), but rather stativity. It either reinforces a verb’s lexical stativity, or signals that a verb which is lexically unmarked for dynamicity functions with stative meaning. In the corpus, this is most clearly discernible with the verb yaʾe ‘know, to be able to’ (lexically unmarked for dynamicity) and ʾəbe ‘want’, which is lexically stative in the meaning ‘want’, while in the meaning ‘love’ is lexically unmarked for dynamicity. First, some k-šaqəl-wa forms of yaʾe or ʾəbe can be subsumed under the prototypical habitual (or, better here, generic) meaning of k-šaqəl-wa. In (37), for instance, habitual aspect is apparent from the other (underlined) k-šaqəl-wa form in the sentence: (37) 2a/4 ʾàxniˈ g-maḥk-ax-wa mən bab-an,ˈ bas bət-kùrdi,ˈ bət ʾò lišana.ˈ we REAL-speak-we-PST with father-our, only in-GEN-Jewish_Neo-Aramaic, in- GEN thisM language 318 THE TENSE-ASPECT-MOOD SYSTEM ‘We used to speak with our father only in Jewish Neo-Aramaic, in this language. la k-iʾ-e-Ø-wa Ø-maḥk-e-wa ču lišana xḕt.ˈ NEG REAL-know-he-PST IRR-speak-he-PST any language other. He couldn’t speak another language.’ In other cases, however, the verbal form describes a state which held over a short period of time at one particular occasion. In such cases, there is no need to present the situation in question as persisting over time. Moreover, the verbs do not match the definition of habituality as presenting a characteristic feature of an individual over a long period of time. Indeed, in English, such forms are more naturally rendered with perfective past than with ‘used to’ forms, as the translations below demonstrate. As for the verb yaʾe, when stative in meaning, it displays a strong preference for k-šaqəl-wa, as illustrated in (38) below, though one šqəlle form is also attested. (38) 5a/12 Context: ‘All the houses of the Jews in Kurdistan were left behind without being sold. Why?’ là g-əb-e-wa Ø-zon-i-wa-lu.ˈ NEG REAL-want-they-PST IRR-buy-they-PST-them ‘(The Kurds) didn't want to buy them.’ SYNTAX AND MORPHOSYNTAX 319 k-iʾ-è-wa Ø-ʾəb-e u-là Ø-ʾəb-e b-šoq-i-la dūq.ˈ REAL-know-they-PST IRR-want-they NEG IRR-want-they FUT-leave- they-it anyway. ‘They knew that whether or not they wanted to, (the Jews) would have to leave them behind anyway. (‘So why would the Kurds buy the houses from them?)’ By contrast, the šqəlle form (or in this case, qam-šaqəl-la) is apparently associated with dynamicity. Forms with the dynamic meaning of ‘recognise someone’ are not attested in the corpus, but could be elicited. In the case of ʾəbe, we should distinguish between the meaning ‘want to’ (followed by a verbal complement) and ‘love’ (followed by a nominal complement). In the meaning of ‘want’, the verb is apparently lexically stative. It always occurs in the k-šaqəl-wa form, even when describing a single and relatively short situation in such cases, the verbal form is used to grammatically reinforces the verb’s lexical stativity: (39) 5b/1 xa g-əb-e-wa Ø-gawə̀r-Ø-wa,ˈ mằni g-ezəl-Ø-wa?ˈ One REAL-want-he-PST IRR-marry-he-PST who REAL-go-he-PST? 320 THE TENSE-ASPECT-MOOD SYSTEM ‘(When) someone wanted to get married, who would go (ask for the girl’s hand)?’ When this verb is used with the meaning ‘love (+nominal complement)’, it can be stative or dynamic. When stative, it is always expressed in the k-šaqəl-wa form. There is one attestation of a clearly dynamic (i.e. inchoative) meaning, and the qam-šaqəl-la form is used (identical to šqəlle in TAM). This form occurs in a story in which Muslims encounter a wounded stranger in the street. The form in question describes the moment when they saw him and ‘began to love him’, or, more suitably translated here, ‘took pity on him’: (40) 2d/68 ʾəθ-wa məšəlmane qam-ʾəbè-Ø-le,ˈ qam-məθe-Ø-le kəs xàθ-e.ˈ EXIST-PST Muslims PERF-love-A.they-P.him, PERF-bring-A.they-T.him at sister-his ‘There were Muslims (who) took pity on him and took him to his sister’s.’ A parallel phenomenon is mentioned in Khan’s grammar of C. Barwar. Khan mentions the verbs b-ʾ-y (baʾe) ‘want’ and y-ð-ʾ (yaðe) ‘know’. In their šaqəl-wa SYNTAX AND MORPHOSYNTAX 321 form,19 these verbs ‘may denote a continuous situation at a particular period in the past rather than a habitual property of the subject’ (Khan 2008b, 587–8). The term ‘continuous’ presumably refers to an unchanging state, which requires a stative interpretation, rather than to a progressive dynamic event. As shown above, in the Early J. Nerwa corpus, (k-)šaqəl-wa is commonly used for the past continuous (including the progressive), that is, for situations overlapping with another event. However, in synchronic NENA varieties including J. Dohok, this function of k-šaqəl-wa is infrequent. The stative usage of k-šaqəl-wa for cognition and emotion verbs in the synchronic J. Dohok dialect therefore most likely reflects a relic of that stage. States are by definition atelic and non-punctual (Comrie 1976, 44), while (lexical) telicity and punctuality is more associated with perfectivity than imperfectivity. Therefore, the retention of ‘continuous’ morphology is more likely for states than for events. 8.10.3. Past continuous In one instance, k-šaqəl-wa conveys the past progressive, as shown in (41) below. The form appears in dialogue.20 19 In Khan (2008b), the form is called qaṭəl-wa, and corresponds to k-šaqəl-wa and šaqəl-wa in J. Dohok. 20 Cf. the use of k-šaqəl-wa for ‘ongoing action’ in J. Zakho dialogue, apparently in contrast to narrative (Cohen 2012, 422–4). 322 THE TENSE-ASPECT-MOOD SYSTEM (41) 1d/19-20 Context: ‘The Turks had passed a decree against Christians.’ ʾana xzeli b-ʾèniˈ g-meθe-Ø-wa (REAL-bring-PST) xa-ʾəsri, ṭḷaθi, ʾàrbi,ˈ xà b-xa b-qaṭḷ-ì-wā-lu.ˈ ‘I saw with my own eyes: they were bringing some twenty, thirty, forty (of them) and then one-by-one, they would kill them.’ 8.11. B-šaqəl-wa B-šaqəl-wa alongside k-šaqəl-wa (and, less frequently, šaqəl-wa) in passages which describe habituality in the past. In Cohen’s J. Zakho corpus, in contrast to J. Dohok, b-šaqəl-wa is said to be very rare (Cohen 2012, 343). Habitual use of b- šaqəl-wa is also reported for J. Betanure (Mutzafi 2008a, 61). In his grammar of C. Barwar, Khan (2008b, 607-8) describes the counterpart of b-šaqəl-wa, which he refers to as bəd-qaṭəlwa, as ‘in some way dependent on the preceding discourse’; for instance, it elaborates on the preceding discourse or follows a previous situation temporarily (see §8.5 above). In J. Dohok, however, there often is no apparent difference between k-šaqəl-wa and b-šaqəl-wa, suggesting that the latter can simply express habituality in the past. In many NENA dialects including J. Dohok, b-šaqəl-wa is used in unreal conditional clauses. SYNTAX AND MORPHOSYNTAX 323 As stated above, the šaqəl-wa and b-šaqəl-wa forms of Pattern II, III, and IV verbs are identical on the surface and to some extent overlap functionally. Thus, Patterns II, III and IV have largely been excluded from this study to ensure greater accuracy. 8.11.1. Future in the past (perfective past) In the corpus, b-šaqəl-wa with the meaning of future in the past in a perfective context is attested only a few times. It has the meaning of an event which is about to happen as a result of an intention in the past. As with b-šaqəl, none of the cases are subordinate. One example is given in (42) below, another may be found in (41) above. (42) 3d/67 Context: ‘In the end, they brought the brother of my boss, his belly had been cut.’ gemər b-qaṭḷì-wā-li (FUT-kill-A.they-PST-P.me).ˈ ‘He says “they were going to/about to kill me.”’ 8.11.2. Unreal conditional The attested patterns of conditional clauses are šaqəl-wa (protasis) – b-šaqəl-wa (apodosis), but also b-šaqəl-wa (protasis) – šaqəl-wa (apodosis). In other LD dialects (e.g. J. Zakho - Cohen 2012, 408–9) the structure of unreal conditionals is said to be šaqəl-wa (protasis) – b-šaqəl-wa (apodosis). 324 THE TENSE-ASPECT-MOOD SYSTEM (43) 1 (elicited) Ø-ʾaθ-ə̀t-wa,ˈ b-yāw -ə́n -wa -lox xa-làxma.ˈ IRR-come-youMS-PST FUT-give -A.IM -PST -R.you some-bread ‘If you had come, I would have given you some bread.’ (44) 5b/9 b-qabl-ì-waˈ Ø-āzəl-Ø-wa kəs bàb-a.ˈ FUT-accept-they-PST IRR-go-he-PST at father-her ‘(If) they agreed, (the boy’s father) would go to her father.’ 8.11.3. Habitual past B-šaqəl-wa occurs in the corpus alongside k-šaqəl-wa (and, less frequently, šaqəl- wa forms) in passages which describe habitual actions in the past. Sometimes, it occurs in the final verb of a series, when its function may be specifically to mark futurity relative to the previous events, presenting a given situation as an intention or result of the previous events. It should be noted, however, that in neither of these forms is b-šaqəl-wa syntactically dependent (i.e. subordinated) to the previous verb, despite the fact that it is logically dependent on the previous discourse (i.e. is presented as an effect thereof). This is inferred from the fact that it always occurs within its own intonation unit and is never preceded by the subordinating d-. In the whole corpus, subordinated b-šaqəl-wa is unattested. SYNTAX AND MORPHOSYNTAX 325 (45) 1e/14-16 Context: ‘There were people who had a craft.’ g-zaqr-i-wa (REAL-weave-they-PST) julle...ˈ ‘they used to weave clothes...’ g-oð-i-wa (REAL-make-they-PST) ... pēlàveˈ kundare,ˈ b-ʿeš-i-wa (FUT-live_off- they-PST) mə̀nnu.ˈ gə-mzabn-ì-wā-lu (REAL-sell-A.they-PST-P.them) .ˈ ‘They used to make… shoemakers (would make) shoes, and used to make a living from these. They used to to sell them.’ Still, in other cases, b-šaqəl-wa occurs in the middle or beginning of a verb series, apparently interchanging freely with k-šaqəl-wa and thus simply conveying past habituality. This is illustrated in (46) below, where b-šaqəl-wa also opens the section. In the passage, k-šaqəl-wa forms are underlined. (46) 5b/13-9,21-2 qăbəl kùlluˈ b-oði-wa ḥə̀nna.ˈ ‘Before everything (else), they would do the henna ceremony.’ b-oði-wa qə̀dduš.ˈ 326 THE TENSE-ASPECT-MOOD SYSTEM ‘They would do (the) engagement (lit. blessing).’21 Ø-mqādə̀š-wa-la,ˈ b-amri-wa lemašàl, əštá yarxe xeta b-gòri.ˈ ‘(When) he betrothed her, they would say, for example, in six months, they will get married.’ ʾanna pare dət baba k-šāqəl-wa mòhar,ˈ ‘(With) the money which the father had taken (lit. used to take) for the bride price...’ g-ezəl-wa g-zāwən-wa ṭala ʾawàye.ˈ jùlle,ˈ dèhwa,ˈ ‘...he would (use them to) go and buy things for her. Clothes, gold.’ p-peša-wa yomət daʿăwa.ˈ ‘(When) the wedding day would arrive...’ b-lele… goði-wa-le ḥənna.ˈ ‘in the evening before… they would hold the henna ceremony.’ 21 According to the informant, Kiddush was done ahead of the wedding as a form of engagement ahead of the wedding, in contrast to modern Jewish weddings, where it constitutes a part of the wedding ceremony. For the Kurdish Jews, this engagement was considered binding; the breaking of engagement necessitated a divorce. SYNTAX AND MORPHOSYNTAX 327 kùlle leleˈ g-raqðì-wa,ˈ g-zamrì-wa,ˈ k-ṭarpi-wa,ˈ ‘All night long they would dance, sing, clap...’ ḥil p-peša-wa bənòke.ˈ ‘...until the morning (lit. until it would become morning).’ This free interchangeability of b-šaqəl-wa with k-šaqəl-wa in J. Dohok makes it unlikely that in this dialect, b-šaqəl-wa expresses a future eventuality relative to another past event and is therefore ‘discourse dependent.’ The cases where b-šaqəl- wa does occur at the end of a series might therefore be coincidental. Conversely, it is by no means obligatory to use b-šaqəl-wa at the end of a sequence to denote a situation which is an effect. The development of b-šaqəl-wa from future in the past to past habitual is likely linked to Khan’s notion of ‘discourse dependency’. Originally, b-šaqəl-wa was apparently reserved for denoting the logical consequence of another past event in realis and conditional clauses (see the preceding sections). In this function, it would become associated especially with imperfective situations (habituals and conditionals), which show typical consequences in specific situations or in given periods of time. Over time, the habitual context of b-šaqəl-wa would become entrenched within the grammatical semantics of b-šaqəl-wa, which, in turn, would 328 THE TENSE-ASPECT-MOOD SYSTEM turn this form into an independent past habitual form.22 It is possible that an intermediate step in this development would have been iterative meaning, but this has not been investigated systematically. 8.12. K-šaqəl(-wa) in a diachronic and comparative perspective In this section, the functional restrictions of k-šaqəl(-wa) in J. Dohok are and compared with the functional features of the corresponding verbal form in other NENA dialects. Diachronically, these cross-dialectal differences can be viewed as different degrees of grammaticalisation, which—in turn—can be used to compare the dialects in terms of their degrees of development relatively to each other. The features considered here are obligatoriness of the realis imperfective preverb (typically k- or ʾi-) in the past and present, and in subordination. Across NENA, there exist two main (widespread) types of realis imperfective markers.23 The first is related to the LD k- preverb with variant realisations such as ke- in the Upper Ṭiyari dialects, ki- in Literary C. Urmi and či- (ɟi/ci) in some 22 See Bybee, Perkins and Pagliuca (1994, 161) for the reconstruction of the English would in its role as a marker of past habituality. The authors show the usage of the volitional/future for habituals. Arguably, many examples are analysable as iteratives. 23 See Rubin (2018, 57, 130–9) for a general overview of realis imperfective preverbs in NENA, and for some historical reconstructions alternative to those found in Khan (2007b). SYNTAX AND MORPHOSYNTAX 329 rural varieties of C. Urmi (Khan 2007b, 86).24 The second marker type is related to the preverb ʾi- in C. Barwar. However, some dialects (e.g. the varieties spoken in the Lower Ṭiyari) possess no realis imperfective preverb (Khan 2008b, 174). Moreover, the realis imperfective marker is not obligatory in all dialects; in J. Dohok, the non-obligatoriness is apparent in the past, and especially, it seems, in past subordination. As regards the source of the k- preverb, it is likely related to the Jewish Babylonian Aramaic q(ā) and/or the Syriac kā. The Jewish Babylonian Aramaic q(ā) itself derives from qāʾəm, the historical active participle of q-y-m ‘stand up’ (Fassberg 2015, 178). Originally a locative/positional circumstantial, it is used in Jewish Babylonian as a presentative present continuous in combination with an active participle Aramaic q(ā)-ʔāmar-Ø PROG PTCP.say-he ‘he is saying’.25 The Syriac kā- apparently has a presentative function (‘look!’),26 associated with the present continuous aspect, e.g. kā ʾāzəl PROG PTCP.go-he ‘look, he is going’. Both forms, therefore, first denoted the continuous present, being later generalised to the generic present (Khan 2007b, 96 and the references therein). 24 It is possible, however, that forms such as the C. Urmi či- reflect a merger of both types of preverbs (Khan 2007b, 86). 25 This construction is also used for the present perfect (or – more likely – lexically restricted resultative?), with both of the functions sharing the ‘ongoing-ness’ feature. 26 The presentative function is related to the thetic function; see §7.3. 330 THE TENSE-ASPECT-MOOD SYSTEM It is not certain whether q(ā) and k- are related, and thus, whether the NENA k- preverb is related to both or just one. Remarkably, the J. Barzani dialect cluster has two preverbs: qam- for the present continuous and (y)g- for the generic present (Mutzafi 2002a, 42). If both are descended from qaʾəm via the Jewish Babylonian Aramaic particle q(ā), it would indicate that J. Barzani contains a recurrence of the same grammaticalisation process, which produced both the synchronic qam- and (and (y)g. Since qam- in J. Barzni has not yet been generalised to the general imperfective from the continuous, and is phonetically closer to the presumed historical qāʾəm, it represents the later cycle. As for the NENA preverb ʾi-, it originates, according to Khan, in a present copula, likely suggesting an original presentative function. This is suggested on the basis of the fact that presentative (or ‘deictic’) copulas form the basis of new continuous constructions in dialects such as C. Qaraqosh (cf. Khan 2002, 331). The schema below shows the features relevant to the history of k-/ʾi- (top two), based on Khan (2007b, 96), and those relevant to the current developmental stage, discussed later in this section. Pragmatics: assertive > non-assertive Aspect: progressive > general imperfective (including habituality) ß Tense: present > [bleaching of temporal function] SYNTAX AND MORPHOSYNTAX 331 Mood: realis > irrealis The pragmatic notion of ‘assertion’ describes a unit in discourse whose purpose is to present information considered new to the listener. This results in a formal constraint: since the listener is expected to recognise the piece of information as new, the unit conveying it has to be ‘cognitively independent’. For this reason, assertion generally does not occur in subordinate clauses (Cristofaro 2003, 30– 50). Assertion is thus related to the notion of ‘discourse salience’. Discourse salience, in turn, tends to provide fertile ground for the emergence of new continuous/progressive constructions (Khan 2007b, 96), which draw attention to situations in the here and now. Thus, the NENA imperfective markers ʾi- and likely also k- are arguably historically linked with the function of assertion or presentativity (see above). Synchronically, pragmatic assertion does not play a discernible role in J. Dohok. In the present, the realis preverb is obligatory, and in the past, form with and without it interchange within one narrative without apparent discourse functional differences. In C. Barwar, on the other hand, ʾi-qaṭəl and ʾi-qaṭəl-wa, in contrast to qaṭəl and qaṭəl-wa, are used to indicate ‘discourse prominence’. In other words, perhaps in C. Barwar, the ʾi-qaṭəl and ʾi-qaṭəl-wa forms are still restricted to clauses with a high degree of assertiveness (Khan 2008b, 590-1). This distributional difference could be indicative of a distinct origins of k- vs ʾi-. 332 THE TENSE-ASPECT-MOOD SYSTEM Another domain where development of the forms with k-/ʾi- is apparent in addition to aspect and assertion is tense. First, in certain dialects, the k-/ʾi preverb is not obligatory in the present. This is apparently the case in the LD dialects of J. Challa, which exhibits a tendency to omit the marker in the second verb of a series (Fassberg 2010, 107), and in dialects with the ʾi- preverb such as C. Barwar (Khan 2008b, 590-1).27 In other words, in J. Challa, k- has not yet been fully grammaticalised as a marker of imperfective realis present. This contrasts with J. Dohok, where the use of the bare šaqəl for realis present is virtually unattested, indicating fuller grammaticalisation. K- is apparently obligatory even in long verb chains, for instance: (47) 3c/31 hatxa bass-a k-xàzy-a,ˈ g-ez-a k-əθy-a g-ʾàrq- a,ˈ Hk-sàrg-aH,ˈ bàss-a.ˈ Thus enough-her REAL-see-she, REAL-go-she, REAL-come-she, REAL- run-she, REAL-knit-she enough-her ‘This is enough for her, she can see, she goes and comes, runs, knits, it is enough.’ Third, as shown above, k-/ʾi- r has now in many NENA dialects been extended to the past, where it is prefixed to šaqəl-wa. This, in turn, indicates that the preverb 27 This may also be the case in J. Betanure (Mutzafi 2008a, 54 and 60), but it is not entirely clear whether the omission is common in both the present and past domains in this dialect. SYNTAX AND MORPHOSYNTAX 333 k-/ʾi has for most part lost its temporal meaning, now serving simply as a marker of imperfective realis. A different situation obtains in C. Alqosh dialect (south- western periphery of NENA), where the realis preverb is not used in the past,28 suggesting an earlier stage of development – before temporal bleaching. Similarly, in Ṭuroyo (located on the western periphery of ENA), corresponding realis imperfective ko-(šoqəl) can only be used in the present (Jastrow 1992, 106), while imperfectivity in the past is expressed with the bare form šoqəl-wo. A brief survey of the Early J. Nerwa corpus suggests a similar early stage: the preverb k- does not appear to be used in the past (Sabar 1976). To summarise, once the association with the continuous domain—which is especially productive in the present tense—is weakened, the form becomes a general imperfective marker, and its temporal function can in turn be bleached: Aspect: continuous/progressive > imperfective (including habituality) ß temporal function bleaching Tense: present > no single tense association 28 Geoffrey Khan, personal communication. 334 THE TENSE-ASPECT-MOOD SYSTEM In J. Dohok, this temporal bleaching is still incomplete: as noted above, k- in the past—in contrast to the present—is not obligatory, indicating that the past is ‘lagging behind’ the present.29 The final feature of k-šaqəl mentioned here is its aforementioned use in perfective past purpose clauses with verbs of movement in J. Dohok (cf. §8.4 above) and illustrated again in (48). I am not aware of a discussion on such irrealis use of realis forms in the NENA literature. (48) 3e/29 zəl-lu k-sàxe-Ø… PERF.go-they REAL-swim-they ‘They went to swim…’ Cross-linguistically, purpose clauses can be expressed either by ‘independent’ (i.e. main clause) or ‘subordinate’ (i.e. subjunctive) verbal forms (Cristofaro 2003, 85– 6), indicating that they have transitional character with regard to syntactic dependence and independence. This syntactic fluidity is presumably parallel to the modal one. Purpose clauses are by implication realis: it is assumed that if there was an intention present and this intention is not negated in the subsequent discourse, it was realised. 29 The grammatical description of the other LD dialects is insufficient to allow for comparisons in this regard. SYNTAX AND MORPHOSYNTAX 335 Thus, the above-mentioned cases of k-šaqəl can be explained as being pragmatically ‘realis’. Still, the syntactic environment of these forms is certainly ‘subordinate-like’. One is therefore left wondering whether this reflects the beginning of another cycle in which k-šaqəl would eventually replace šaqəl even in subordinate and irrealis environments. Such a replacement process has already occurred in earlier Aramaic: the prefix conjugation was first pushed out into the realm of irrealis (and largely subordination) by the active participle (the predecessor of šaqəl), before being replaced by it altogether (Stilo and Noorlander 2015). This process of change can be schematically represented as follows: Mood: realis > irrealis The environment of clauses which are pragmatically ‘realis’ could certainly serve as a suitable entry point for a verbal form into the irrealis ream. A cross-dialectal study would be necessary to illuminate this issue further. 8.13. B-šaqəl(-wa) in a diachronic and comparative perspective The following section discusses the history of the NENA future forms and these forms’ synchronic features. I review the modal and syntactic restrictions of the future, discussed for J. Dohok above, suggesting that they are due to an incomplete grammaticalisation of this form. These developments within NENA are set in the context of common (cross-linguistically attested) grammaticalisation pathways of 336 THE TENSE-ASPECT-MOOD SYSTEM the future. Such a diachronic approach is also used here to explain some of the particular features of the NENA future forms, such as their inability to receive negation (in most dialects). The future preverb in NENA takes the form b-, bət or t-.30 This preverb does not exist in the synchronic dialects in the Sorani Kurdish area such as TZ or C. Shaqlawa,31 which is apparently due to language contact, since Sorani also does not possess a distinct future form (Fox 2015, 167). All of the synchronic future preverbs (i.e. b-, bət or t-) are related, and originate in the modal auxiliary ‘want’ , historically bʿy, later fossilized as a modal particle in the form b(ā)ʿē (attested, for instance, in Jewish Babylonian Aramaic), followed by the subordinating d-. Noorlander (2017, 190) reconstructs the NENA future in the following way: bʿy > bʿy d- > b(ā)ʿē d- > bəd (> b-, t-) ‘want’ proximative proximative future future inflected inflected uninflected fossilised fossilised According to Noorlander (2017, 187–8), the construction b(ā)ʿē d- developed from an event-modal function into a proximative present marker. The proximative conveys the sense that something is ‘about to happen’. This transitional stage of the proximative seems indeed likely – such a process is in fact currently being 30 The distinct future marker zil, found in the Mosul plain dialects, has its own grammaticalisation source, and is not discussed here; see Coghill (2012). 31 Fieldwork by author. SYNTAX AND MORPHOSYNTAX 337 recycled in NENA, including the J. Dohok dialect (see §8.16.1). The proximative function is also attested in Jewish Babylonian Aramaic and Syriac.32 Later, this proximative developed into a future form, apparently through the semantic proximity between the proximative present and the future. These functional changes can be represented as follows: Aspect: proximative (present) > (loss of proximative function) Tense present > future Turning to the synchronic stage of the NENA future, the following parameters will prove relevant here: Modality event > epistemic desire, obligation > intention > prediction > all types Pragmatics assertive (> neg. assertive?) > non-assertive Clause type main > subordinate These parameters are adapted from Bybee, Perkins and Pagliuca (1994, 243–79; 1991), expanded by the pragmatic dimension. 32 For references, see Noorlander (2017). 338 THE TENSE-ASPECT-MOOD SYSTEM As regards modality, according to Bybee, Perkins and Pagliuca (1994, 243–79), epistemic-modal futures typically develop later out of event modality.33 This is also the case in NENA, where the predecessor of b-šaqəl conveyed desire and obligation: event > epistemic Modality desire, obligation > intention > prediction > all types More specifically, there are two reasons to consider the synchronic NENA future form is an intermediate stage between event and epistemic modality. First, b-šaqəl is no longer simply a marker of event modality. As shown above (§8.5), dynamic modality with the 1st person is only expressed through b-šaqəl if the predication denotes an intention, and, consequently, overlaps with the epistemic category of prediction. However, the key functional feature is arguably prediction. With the acquisition of predictive function, b-šaqəl has now entered the epistemic realm, but has not yet been extended into other types of epistemic future. For instance, 33 Bybee, Perkins and Pagliuca (1991 and 1994) use the term ‘agent-oriented modality’ for event modality in which the agent is the main locus of the obligation, desire etc. Clearly, this category does not subsume imperatives, which is termed by Bybee, Perkins and Pagliuca ‘speaker-oriented’. According to these authors, speaker-oriented modality is a feature of ‘more fully developed futures’, and shouldn’t therefore be subsumed under the typologically ‘early’ event modality. SYNTAX AND MORPHOSYNTAX 339 as shown in §8.2.1 above, uncertainty is conveyed only with the šaqəl form, for instance: (49) 3 (elicited) balki Ø-ʾaθe-Ø bə̀nhe...ˈ maybe IRR-come-he tomorrow ‘He may come tomorrow (, I don’t know).’ The modal notion of prediction is associated with pragmatic assertion (see §8.12 above) and assertion, in turn, is relevant to the question of subordination. As the notion of prediction suggests, future forms are typically used to make an assertion about what is going to happen (cf. Bybee et al 1991). Thus, as expected, such forms usually make pragmatic assertions. In NENA, the association of the future form with assertion is still very strong. Namely, the future form is dispreferred even in environments where the mood is ‘predictive’ (it is assumed that a given state of affairs will obtain), but the ‘prediction’ in question is syntactically subordinate and thus ‘non-assertion-like’. For example, in J. Dohok, b-šaqəl is apparently ungrammatical in subordinate restrictive relative and in adverbial clauses. The following sentences come from elicitation, but were obtained from several informants. When rephrased with b-šaqəl by the interviewer, they were rejected by the informants. (50) 1 (elicited) 340 THE TENSE-ASPECT-MOOD SYSTEM ḥmol ʾàxxa,ˈ ʾaw naša d-Ø-ʾaθe-Ø (REL-IRR-come-he) ʾaxxa b-xamša daqiqe xeta bàbī=le,ˈ šqulle mənnox u-θằ.ˈ ‘Wait here, the person that will come here in five minutes is my father, take him with you and come.’ (51) 5 (elicited) ʾana Ø-maḥk-ən (IRR/FUT-speak-IM) mənnox dammət Ø-daʾăr-ət (IRR-return- youMS) l-bèθa. ‘I will speak to you when you get home.’ In English, adverbial clauses like the one in (48) above also tend to use non-future forms; in restrictive relative clauses like in (47), future forms are possible. A similar situation obtains in other NENA dialects. The following is an example from C. Dohok which predicates about a future situation. The first two main-clause verbs use the b- preverb to mark a predictive future, while a šaqəl form follows the adverbial subordinator ʾegət ‘when’: (52) C. Dohok b-zalux hàm ṭali b-aθət,ˈ ʾaygət aθət māθətti ʾixàla.ˈ ‘You will go and come back to me. When you come, you will bring me food.’ By contrast, in languages such as Slavic or (Modern Standard) Arabic, future forms are not restricted to pragmatic assertion; future verbs can co-occur with particles SYNTAX AND MORPHOSYNTAX 341 of uncertainty or possibility (e.g. MSA rubbamā; Polish może (+ Future), both meaning ‘perhaps’), that is, with non-predictive types of epistemic mood. This occurs not only in main clauses, but also in subordinate ones, including those which are parallel to the NENA sentences above. Still, in certain subordinate clauses in NENA, b-šaqəl is used. In J. Dohok, this has been found for complement clauses of the verbs ‘say’ and ‘understand’, for example: (53) 5 (elicited) Context: Did he say he would come? he, mərre-li ʾawa d-bàθe-Ø. yes PERF.say-he he REL-FUT-come-he ‘Yes, he said he would come.’ On the other hand, b-šaqəl was deemed ungrammatical with verbs of will or perception, where šaqəl is used instead, for example: (54) 1 (elicited) mqorər-re (dət) Ø-ʾàzəl-Ø. PERF.decide-he (REL) IRR-go-he ‘He decided to go.’ In pragmatic literature, there is an ongoing debate regarding whether complement and adverbial clauses should be classified as assertive and non-assertive, and both empirical and conceptual arguments are invoked (Cristofaro 2003, 33-5). 342 THE TENSE-ASPECT-MOOD SYSTEM Empirically, as the above NENA examples demonstrate, languages differ in what they choose to grammatically encode as ‘assertive’ or ‘non-assertive’, or at least which subordinate clauses display the grammar of main clauses. The best an empirical linguist can do is therefore to map the distributional differences and make suggestions about the directionality of change. In the case of NENA, if b- šaqəl is to become widespread in subordinate clauses, complement clauses will likely be the first to be affected. Finally, the notion of assertion can be helpful for explaining the negation of the future in NENA. As previously mentioned, the vast majority of the NENA dialects which possess a future form cannot negate it.34 An exception to this is appears to be J. Challa (LD, Turkey).35 In J. Urmi, a form with the b-/be- preverb is used for an ‘emphatic’ negation of the future (Khan 2008a, 257) (with ‘emphasis’ likely being equivalent to making a strong assertion). In most other NENA varieties, however, the negated future is expressed by negating either the bare form (parallel to J. Dohok šaqəl) or the realis imperfective one (i.e. the form parallel to J. Dohok k-šaqəl).36 The latter strategy is also found in the Early J. Nerwa texts (for an example, see Sabar 1976, 64, 18:20). 34 Dialects in the Sorani Kurdish area typically do not have a designated future form, in parallel with Sorani. 35 Cf. the examples in Fassberg (2010), e.g. 105, example f. 36 For an overview of future negation across nine different dialects, see Fox (2015, 165). SYNTAX AND MORPHOSYNTAX 343 In light of this and contrary to Fox's suggestion (2015, 165), it seems likely that the future form *la b-šaqəl never (fully) developed in NENA. This lack of development would also align with the pragmatic properties of negation, pointing again to a limited grammaticalisation of b-šaqəl.37 According to Givón, negated sentences (‘sentential negations’) make a ‘weaker assertion’ than affirmative ones. This is because ‘negative assertions are, it seems, made on the tacit assumption that the hearer (…) is at least familiar with the corresponding affirmative’ (Givón 1978, 7). In other words, negated sentences build on some presupposed information, which they subsequently label as false. As shown above, the NENA b-šaqəl is generally restricted to making pragmatic assertions. Perhaps, therefore, the form has not yet been extended to negative assertions, because such assertions are ‘weak’ – in between assertion and non- assertion (cf. Givón 1978, 22). For the development of the NENA future, this can be schematically represented as follows: Pragmatics assertive > negative assertive > non-assertive 37 I owe this insight to a discussion with Paul Noorlander. If this explanation is correct, it is also applicable to Bahdini, under whose influence the negated future form in NENA developed. 344 THE TENSE-ASPECT-MOOD SYSTEM 8.14. Copula + Participle (wele šqila) The wele šqila construction in J. Dohok (i.e. wele copula + resultative Participle)38 expresses resultativity and—in stative verbs—the semantically-overlapping continuous aspect. This means that the focus of the predication is on the persisting state which directly follows an event, rather than on that event itself (Nedjalkov 2001, 928–930).39 Across NENA, the functional scope of the construction copula + Participle varies. In many other—more progressive—dialects, it has been extended to a fully-fledged perfect.40 As a perfect, this construction refers to an action or a state which happened in the past. Its results are still relevant for the present, but may be non-specific and less closely bound with the predication than in the case of the resultative (Nedjalkov 2001, 930). Based on typological studies (e.g. Bybee, Perkins and Pagliuca 1994, 81-5) and the evidence of the early NENA sources, this functionally restricted (resultative) construction in J. Dohok is best understood as a conservative feature within this dialect. 38 For the syntax of this construction, see §7. In the past, the presentative copula is replaced by the past copula, e.g., 3Ms wewa. In the future and in the irrealis domain, forms of the verb ‘be’ are used before the Participle; respectively p-awe šqila and Ø-hawe šqila. 39 I am indebted to Paul Noorlander for drawing my attention to the distinction between the perfect and the resultative, and for helping me test various verbs in the resultative construction during my fieldwork in Jerusalem in September 2019. 40 For a historical study of this construction in Aramaic, see Noorlander (2018, especially 328- 333). SYNTAX AND MORPHOSYNTAX 345 In the corpus, there only are eight certain attestations of wele šqila, but other examples have been elicited. The first category of verbs that occur in wele šqila are statives; such constructions predicate about a state which had commenced in the past and continues until the given reference time, as shown in (52) below. In this function, wele šqila overlaps semantically with the continuous aspect, because both refer to a state ongoing at a time of reference. In fact, stative continuous in J. Dohok is apparently be default expressed with wele šqila. The other continuous construction, wele bə-šqala (see §10.15 below), is less frequent and much more difficult to elicit.41 (55) 1d/14 gəmra ʾana wan-wa (COP.IF-PST) tu-ta (RES.sit-FS) ʾàxxa,ˈ θele lòri…ˈ ‘(My aunt) said “I had sat/I was sitting here (on the rooftop platform, when) a lorry came...”’ In addition to stative verbs, wele šqila is also used with the verbs of movement such as Ø-θ-y (ʾaθe) ‘come’, where the emphasis is on the arrival: (56) 1b/53 flan welu (COP.DEIX.they) ʾə̀θy-e (RES.come-PL)ˈ mdagəl xa-dùgla. ‘Some men have come (i.e. are here) to tell you a lie.’ 41 A similar situation whereby a single form is used for both resultative and (stative) continuous is attested in languages such as Chinese or Japanese (cf. Shirai 1998). 346 THE TENSE-ASPECT-MOOD SYSTEM Other functions of wele šqila are closer to the perfect domain. For instance, with the verb of movement ʾazəl ‘go’, this construction can express the resultative/perfect of experience, in which case, its result is less directly linked to the preceding situation: (57) 3 (conversation) Context: ‘Have you been to Mosul?’ wan-wa (COP.IF-PST) zəl-ta (RES.go-FS) go mòṣəl. ‘I have been to Mosul.’ Transitive verbs attested in wele šqila are only possessive transitives, where the focus is on the result of having an object in one’s possession or on one’s body, for instance: (58) 1b/85 sawoyox we-le (PERF.be-he) šqil-a (RES.take-MS) mən sawoyi -i par-e… ‘Your grandfather had borrowed money from my grandfather (i.e. had a loan).’ Other types of transitive verbs are unattested and could not be elicited. For instance, one informant rejected the following construction as ‘belonging to the dialect of J. Amediya.’ (56) 5 (not accepted) *we-le gríš -ə -llu DEIX.COP.he RES.pull -A.MS -P.them SYNTAX AND MORPHOSYNTAX 347 ‘he has pulled them’ As mentioned above, in other dialects, copula + resultative Participle has become a full perfect. This is apparently the case in Christian Barwar, where this construction can be used with non-possessive transitives such as ‘kill’ (Khan 2008b, 735). In the Khabur region on the western periphery of NENA, the copula + Participle construction is typologically even further developed: it is apparently in the process of developing into the preterit.42 Even the other LD dialects are apparently more progressive than J. Dohok in this regard. J. Zakho has—according to Cohen—a present perfect which also subsumes the resultative (Cohen 2012, 429–433). Sentence (56) is one example of a non- possessive transitive in the prefect construction from Cohen (2012): (59) J. Zakho (Cohen 2021, 430) le-wən qṭil-a dida. NEG-COP.PRS-IM RES.kill-MS P.her ‘I haven’t killed her.’ In J. Betanure, perfect-type usage is also attested; for example: (60) J. Betanure (Mutzafi 2008a, 56) təmmal babi le (COP.PRS.HE) wíd-ə-lle (RES.make-A.MS-P-him) mən kepa 42 Paul Noorlander, personal communication. Bybee et al (1994, 81-85) proposed the following development of the resultative: stative > resultative > perfect > preterit. 348 THE TENSE-ASPECT-MOOD SYSTEM ‘my father only yesterday made it from stone.’43 Similar examples can be found in J. Amediya (Hoberman 1997, 89–90). The J. Betanure corpus (Mutzafi 2008a) furthermore indicates that this dialect has another strategy for expressing the present resultative (/perfect): wəlle (presentative copula) + šqəlle. For instance: (61) J. Betanure (Mutzafi 2008a, 220, XIII/273) wəlle (COP.DEIX.he) θe-le (PERF.come-he) mušəlmana. ‘Look, a Muslim has come! (Do not speak very much.)’ 8.15. Copula + in-Infinitive (wele bə-šqala) and wal/hol/hole k- šaqəl In many NENA dialects, constructions consisting of copula + in-Infinitive (in J. Dohok, wele bə-šqala) are commonly used to express the continuous aspect.44 In J. 43 Mutzafi’s uses the terminology of ‘dynamic present perfect’ and ‘stative present perfect’. Constructions termed here ‘resultative’ fall within both of those categories, since ‘dynamic’ subsumes dynamic intransitive verbs, which are considered here as belonging to the resultative. 44 The term ‘progressive’ is commonly used in the NENA grammars (e.g. J. Betanure - Mutzafi 2008a; C. Barwar - Khan 2008b; C. Diyana-Zariwaw - Napiorkowska 2015a). Still, the broader term ‘continuous’ is preferred here, since in many dialects, the construction in question can also SYNTAX AND MORPHOSYNTAX 349 Zakho, for instance, there are two forms, structurally highly similar to the J. Dohok ones, and these are said to convey ‘actual or specific present’ (Cohen 2012, 435). In dialects such as C. Urmi or J. Arbel, this construction has even been extended into non-progressive and non-continuous domains – the habitual present or even perfective past in the narrative (C. Urmi - Khan 2016, vol.2, 185–200).45 On the other hand, in dialects of the southern NENA periphery, the construction copula + Infinitive does not exist. This applies to the Mosul plain Christian dialects (see below) and the South-Eastern Trans-Zab dialects in Western Iran such as J. Sanandaj (Khan 2009, 7). The TZ dialects in north-western Iran (J. Urmi) and across the border in eastern Iraq (J. Rustaqa and J. Ruwanduz) also have a continuous form also based on the Infinitive, e.g. garoše/garoš-le ‘he is pulling’ in J. Urmi (Khan 2009, 7; 240). There is also an altogether different continuous construction in NENA, based on a presentative element combined with the present form – in J. Dohok, wəl/hole/hol k-šaqəl. A parallel construction – wal k-šaqəl – is also attested in J. Zakho (LD; Cohen 2012, 435-7). This type of continuous construction is widely used in the Mosul plain dialects of C. Qaraqosh (Khan 2002, 333–5), C. Karamlesh and Baretḷa, and in the Ankawa region (Borghero 2008). These dialects are neither be used with stative verbs. Such verbs are incompatible with the progressive aspect, which implies a ‘progress’, and therefore dynamicity (cf. Comrie 1976, 25). 45 On a general discussion on progressive constructions in NENA, see Khan (2007b, 95–7). 350 THE TENSE-ASPECT-MOOD SYSTEM geographically nor typologically close to LD. In C. Qaraqosh, this construction is kilə k-šaqəl, and it can occur in some cases with the broader habitual meaning (Khan 2002, 334). As mentioned above, J. Dohok possesses two constructions which convey the continuous aspect. These two constructions apparently have different functions and are both highly infrequent in the corpus as is shown below, pointing to their limited grammaticalisation. 8.15.1. Wele bə-šqala This construction is attested only twice in the J. Dohok corpus; in both cases, the clause is negated and the predication concerns an (observable) state. The forms are formally present, but occur in past narrative and are therefore past continuous: (62) 3f/10 mà d-g-məθele dərmaneˈ u-mà d-goðiˈ lè-we (NEG-COP.PRS.he) bə-ṭraṣa (in- INF.heal).ˈ ‘Whatever medicine they bring and whatever they do, he is not getting better.’ (63) 3f/36 Context: ‘They told the Jew to go home: “we will see tomorrow, whether the king wakes up or not.”’ SYNTAX AND MORPHOSYNTAX 351 zəlle HmiskenH ʾo huðaya l-bèθa,ˈ lewe (NEG-COP.PRS.he) bə-dmàxa (in- INF.sleep)ˈ mən zdòʾŏθe.ˈ ‘The poor Jew went home, (but) he wasn’t sleeping for fear.’ With other verbs, wele bə-šqala could not be elicited with most speakers. When the speakers were presented with affirmative or negative stative verbs of cognition (e.g., t-x-r, taxər ‘remember’) or sensory perception (e.g., x-z-y, xaze ‘see’) in the wele bə-šqala construction, they accepted them, but described them as unnatural or ‘reminiscent of J. Zakho or J. Amediya’, and would often rephrase with a k- šaqəl form. Similarly, with dynamic verbs such as ʾaθe ‘come’, this construction was judged as intelligible, but speakers would often rephrase with wəl k-eθe-Ø DEIX REAL-come-he ‘he is coming’ (see below). This suggests that in J. Dohok, wele bə-šqala is—in contrast to other dialects where this form is typical of dynamic events— precisely not a progressive construction. Rather, it tends to convey a non-dynamic continuous aspect. Even in this function, however, it is not very productive, and could even be restricted to particular types of states or to negated sentences.46 However, since only a few examples represent this construction in the natural data, this conclusion remains tentative. 46 In C. Barwar, cognition (i.e. stative) verbs are more likely to occur in the continuous/progressive construction than their affirmative counterpart (cf. Khan 2008b, 571). 352 THE TENSE-ASPECT-MOOD SYSTEM 8.15.2. Wəl/hol/hole k-šaqəl In contrast to wele bə-šqala, the construction wəl/hol/hole k-šaqəl is attested only with lexically dynamic verbs, and can thus be labelled progressive. In this construction, a k-šaqəl form is accompanied by the particles wəl or hol, or the apparently fossilised 3MS form hole (e.g. hole g-baxš-a-lu DEIX REAL-stir-A.she- P.them ‘look, she is stirring’). (64) 3h/21 θela kàlo,ˈ g-markəwí-wā-la rəš sùsa,ˈ rəš koðə̀nta,ˈ (…) dohola u-zə̀rnaˈ hol (DEIX) g-ràqð-i (REAL-dance-they)…ˈ ‘The bride came – they would put her on a horse, on a mule. (They had a procession with) a drum-and-flute ensemble, they were dancing…’ (65) 3i/4 Context: ‘(But) she would tell him “leave me.”’ ʾana wəl (DEIX) gə-mbàšlan (REAL-cook-IF),ˈ ʾatta gəbe Ø-ʾoðan ʾixàla,ˈ Look, I am cooking, I need to make food now.’ The above-mentioned association of wəl/hol/hole k-šaqəl constructions with lexically dynamic verbs is also suggested by interviews: speakers showed a tendency to rephrase constructions such as *wele b-iθaya DEIX.COP.he in-INF.come ‘he is coming’ which were produced by the interviewer to wəl k-eθe-Ø DEIX REAL- come-he ‘look, he is coming’. SYNTAX AND MORPHOSYNTAX 353 The progressive aspect in these constructions is expressed by the aforementioned presentative elements wəl, hole or hol (cf. Cohen 2017; Khan 2018).47 Presentative particles typically draw attention to events which are happening in the here-and- now, and which are therefore aspectually continuous. Similarly in narrative, presentative particles have the effect of placing the listener in the midst of the unfolding events, as if he or she were witnessing them. Still, the J. Dohok construction wəl/hole/hol k-šaqəl are not fully grammaticalised progressive forms; the progressive/continuous aspect is often expressed simply with k-šaqəl (cf. §8.3.1 above). Thus, the primary function of these constructions is the presentative function, meaning that they are reserved for uses such as narrative vividness (e.g. in (60) above), or for drawing attention to an ‘important’ event. This is the case in (61) above, which could be paraphrased: ‘(Leave me alone.) Don’t you see that I’m cooking?’ The aspectual (progressive) meaning, in turn, is secondary, being contingent on the presentative function. Within LD, J. Zakho is also noted to possess both of these continuous aspect constructions (Cohen 2012, 435), though Cohen’s study does not note whether these two constructions are functionally different. 47 See §7, and especially §7.3 and §7.13 for a larger discussion on presentative particles, copula clauses presenting situations in the ‘here-and-now’, and sentential demonstratives. 354 THE TENSE-ASPECT-MOOD SYSTEM 8.15.3. The history of the NENA continuous forms It is widely recognised that the general developmental path of imperfective forms is as follows: progressive (dynamic events) > continuous > general imperfective (cf. Bybee et al 2015). As has been demonstrated for many languages, progressive forms typically originate in constructions conveying the aspect of ‘ongoingness’ at a given time of reference (Hengeveld 2011, 590-1) – the speech time, the time of another event or a time specified within the clause. Eventually, these progressive constructions develop into general imperfective forms, subsuming progressive, continuous and habitual forms. Another idea often encountered in the literature is that continuous constructions start as progressives, that is, occur with dynamic verbs (Bybee, Perkins and Pagliuca 1994, 132-5). Many languages, including several NENA dialects such as C. Barwar, still exhibit a preference for the continuous construction to be used with dynamic verbs. This should not lead us to assume, however, that there is only one possible source construction of the continuous form, namely a dynamic progressive.48 In fact, if 48 Even in English, the ‘progressive’ was originally (in Old English) not restricted to dynamic events (cf. Kranich 2010, 88-90 and see below). SYNTAX AND MORPHOSYNTAX 355 ongoingness at a reference time is indeed a key feature of progressive/continuous source constructions as suggested by Hengevald (2011, 590), we can expect these constructions to be used especially often in circumstantial clauses (‘While X is in activity A, event B occurs’). Such circumstantial use, in turn, seems to be especially suitable for atelic rather than telic verbs: the former lack a natural endpoint and are therefore better predisposed to stretch over a longer time period, accompanying another event. Such circumstantial/continuous use, in turn, does not preclude states; indeed, states, are more likely to be of longer duration than telic dynamic events (cf. Kranich 2010, 30). On the other hand, Bybee, Perkins and Pagliuca (1994, 132-5) suggest that the source progressive construction in languages such as English would have been more typical of main clauses and dynamic events (Bybee, Perkins and Pagliuca 1994, 133). Instead of another event, its reference time would likely be speech time. For NENA, questions about the source of the continuous/progressive constructions since remain; data from synchronic dialects such as J. Dohok caution against assuming an original dynamic function, despite the fact that this is the predominant function in many synchronic NENA varieties. Below, I offer a few remarks on the semantics of the ‘continuous’ construction in the early NENA sources, and on their syntax. A comprehensive grammaticalisation model for the NENA continuous/progressive is yet to be established, and its implications for general grammaticalisation theories considered. Unfortunately for the present study, the early NENA material includes little dialogue, and therefore provides 356 THE TENSE-ASPECT-MOOD SYSTEM few contexts for the use of the actual present, i.e. ‘Look, X is in the middle of activity A.’ In general, in the early NENA sources, bə-šqala (with or without a copula) is used both predicatively (in main clauses) and circumstantially. This usage does include states, but is more common with (transitive or intransitive) events, which are often atelic. If we assume an original circumstantial usage, the preference of bə-šqala for atelic verbs makes sense and the assumption of an exclusively dynamic original use is unjustified. In early Christian poetry (the Mosul plain, 17th-19th centuries, but archaising in nature), bə-šqala (for the continuous) is very and occurs mostly without a copula, functioning therefore ‘as a circumstantial modifier or a complement of the predicate’ rather than as an independent predicate (Poizat 1999, 173, cited in Mengozzi 2012, 34). In most of the examples cited by Mengozzi (2012, 34–35), this construction refers to atelic, dynamic (both intransitive and transitive) events occurring at the time of another event, as shown in (61) below. The examples also include a state. (66) Early Christian NENA poetry (Mengozzi 2012, 35) zmerran beštaya ‘we sang while drinking’ In the early J. Nerwa homilies, whose original text likely goes back to the early 16th century, bə-šqala is also infrequently (in the continuous function). In all cases, SYNTAX AND MORPHOSYNTAX 357 the verbs are dynamic (transitive and intransitive) and almost always atelic. Apparently in contrast to the early Christian NENA poetry, it occurs both with a copula (4x) and without it (5x). It is found in circumstantial (4x), relative (2x) and main (3x) clauses. The lack of copula is not an automatic sign of non-predicative use; there is one case of predicative use without the copula. Circumstantial (or subordinate) use is more common, but predicative use (e.g. actual present) is also attested and should thus not be excluded from the original functional scope of this construction. Indeed, predicative use with actual present as reference time is more typical of dialogue, and this genre is absent in the sources. Thus, the predicative use of (copula +) bə-šqala is perhaps underrepresented in the corpora. Comparing early NENA to the synchronic J. Dohok dialect, it is apparent that in the former, the (copula +) bə-šqala form is not restricted to stative forms, though it does not exclude them either. As previously suggested, stative and atelic dynamic verbs share a predisposition for a longer duration (than telic verbs). It is possible that the development of the wele bə-šqala in J. Dohok took place along lines of atelicity. The Early J. Nerwa corpus also includes one attestation of the predecessor of the LD construction wal/wəl/hole k-šaqəl: (67) Yā Moše, ʾāna wē-le gmaqw-ın-ne lıbbıd Parʿo. ‘Oh Moses, look, I am hardening the heart of the Pharaoh.’ 358 THE TENSE-ASPECT-MOOD SYSTEM This is a clearly presentative use of wele (+ k-šaqəl), used to match in the NENA behold I am’ in Ex.14:17. As‘ יִנְנִה translation the Biblical Hebrew presentative argued above, this presentative function is still preserved in J. Dohok in the construction hole/hol/wele/wəl k-šaqəl. Presentativity is assertive (cf. §8.10 above) – it draws attention to an event taking place in the here-and-now. Assertion, in turn, is generally incompatible with subordination, which suggests that a presentative construction such as wele k-šaqəl would as a general rule not occur in subordinate clauses such as circumstantial ones. If this functional description of the construction hole/hol/wele/wəl k-šaqəl is accurate, it suggests a very different development path in comparison to the construction (copula +) bə-šqala construction, which in the early NENA sources is associated with subordinate usage. The construction wel/wal/hole k-šaqəl could in fact be more similar to the original English function of the -ing progressive. In Old English, this construction had not yet become associated with dynamic events and was most strongly associated with ‘vividness of narrative’ (Kranich 2010, 88- 90), that is, presentativity. 8.16. Expression of other TAM categories 8.16.1. Proximative There are a number of strategies to express the imminent future: SYNTAX AND MORPHOSYNTAX 359 xapča xēt (pīš/pəšle) + b-šaqəl (68) 3 (elicited) ʾīθ yalunke rəš gàreˈ g-zadʾan xapča xēt (PROX) b-nàpl-i (FUT-fall-they).ˈ ‘There are children on the roof, I am afraid (that) they are about to fall.’ waxta + šaqəl (69) 3 (elicited) Context: ‘There is a rock on top of the mountain.’ ʾenoxun l-gyànoxunˈ wàxta (PROX) Ø-napəl-Ø (IRR-fall-he).ˈ ‘Watch out, it is about to fall.’ gəbe + šaqəl (with inanimate subjects) (70) 3 (elicited) Context: ‘There is a rock on top of the mountain.’ gəbe (PROX) Ø-nàpəl-Ø (IRR-fall-he).ˈ ‘It is about to fall.’ Gəbe is a fossilised (3MS) k-šaqəl form of the verb ‘want’. In combination with a šaqəl form, it can express the imminent future,49 but this construction is restricted 49 In this fossilised form, it also functions as a deontic modality particle of obligation. 360 THE TENSE-ASPECT-MOOD SYSTEM to inanimate subjects,50 indicating incomplete grammaticalisation. Inanimate subjects are non-agentive. Thus, when gəbe occurs alongside an inanimate subject, it cannot be interpreted as conveying event modality, which would require the subject to be agentive. Such a construction thus unambiguously expresses imminence, not event modality. 8.16.2. Inchoative Inchoativity is expressed through an inflected form of the verb p-y-š (payəš) ‘stay, become’ followed by an Infinitive form. In Pattern I verbs, the Infinitive is preceded by the preposition b-, as is the case in (67) below, while in other Patterns, b- has been lost through assimilation to the following m (e.g. pəšli maḥkòye ‘I began speaking’). In the corpus, these inchoative constructions are attested only with semantically dynamic and agentive verbs. (71) 3d/12 ‘Context: In the end, The Farhud broke out, on the eve of Shavuot.’ mpəqlu pəš-lu (PERF.begin-they) bə-qṭàḷa (in-INF.kill) naše.ˈ ‘At night, they went out and started killing people.’ 50 I am indebted to Paul Noorlander for alerting me to this fact during an interview. SYNTAX AND MORPHOSYNTAX 361 8.17. Conclusion The present chapter has surveyed the J. Dohok verbal system with references to other dialects and suggestions about likely grammaticalisation pathways for particular forms. It has been suggested that the NENA verbal system can help understand large-scale grammaticalisation pathways occurring cross- linguistically. For instance, I noted the use of the prototypically realis k-šaqəl form in purpose clauses with movement verbs in the main clause (‘I went in order to do X.’). Such purpose clauses allow the use of k-šaqəl despite the fact that they express an intention, and are thus technically irrealis. The extension of a realis form into irrealis took has already taken place in earlier Aramaic: the prefix conjugation was first pushed out into the irrealis realm (and largely to subordination) by the active participle (the predecessor of šaqəl), before being replaced by it even in modal environments (Stilo and Noorlander 2015). The data from synchronic varieties such as J. Dohok could suggest that such an extension into the modal domain can take place via clauses which are pragmatically interpretable as realis: the predication in purpose clauses—unless negated—is inferred to have been realised. CHAPTER NINE WORD ORDER 9.1. Introduction 9.1.1. Aims This chapter is a study of constituent word order in verbal clauses, accompanied by a few remarks on word order in existential clauses. It takes a quantitative and comparative approach, drawing on tagged corpora of J. Dohok and J. Sanandaj to represent, respectively, LD and TZ. The most frequent order in LD is subject-verb- object, and in TZ, it is subject-object-verb. The aim is to present a general typology of word order in J. Dohok, and to contrast it with that of J. Sanandaj. It is anticipated that many of the conclusions for J. Dohok will turn out to be applicable to other dialects from the LD cluster, if not other varieties of Iraqi NENA. At the same time, the comparative method helps highlight shared features of LD and TZ, which persist despite fundamental word order differences in these two NENA groups. It is hoped that insight into such shared features will prove useful in the future for establishing a NENA-wide word order typology, and for understanding stages of diachronic change in Aramaic. To-date, no studies of word order and information structure in LD have been carried out. The greatest contribution to the study of word order in NENA at large has been made by Khan, whose grammars (e.g. C. Barwar – 2008b; J. Sanandaj – 364 WORD ORDER 2009; C. Urmi – 2016) feature both observation-based generalisations and meticulous studies of information structure and discourse pragmatics in particular examples. However, quantitative, corpus-based and comparative studies of word order in NENA are still lacking, and it is this gap which this chapter comes to fill. With its quantitative methodology, this investigation is capable of demonstrating the strength of patterns observed in a ‘manual’ study, and of revealing quantitative relationships between individual patterns. This, in turn, enables the writer to set the NENA dialects of J. Dohok and J. Sanandaj in the context of general-linguistic clause typology. 9.1.2. Data, methodology and terminology The present chapter draws from an article on word order in NENA by Paul Noorlander and the author of this thesis (forthcoming 2022), which has been significantly expanded in the writing of this chapter (from 7,500 words to 14,500 words). The J. Sanandaj corpus of ca. 7,600 words is based on Khan’s data (2009, excl. poem E) with four informants. It consists of narrative texts which include spontaneous spoken language about past customs, anecdotes and life stories, as well as folktales. Paul Noorlander is credited here for tagging, analysing and interpreting the J. Sanandaj data. The J. Dohok data are taken from my own corpus, some sections of which were excluded for the present investigation. I was responsible for tagging, analysing and interpreting the data. The comparative work was done jointly. The section on intransitive and existential clauses, on SYNTAX AND MORPHOSYNTAX 365 Agents and most of the section on ditransitives is my own work with the exception of the J. Sanandaj data, which were supplied by Paul Noorlander.1 The counts of annotated verbal clauses for both dialects are shon in Table 44 below. The second row represents the percentage of verbal clauses with overt independent arguments (noun phrases and independent pronouns). Table 44: Total counts of annotated data J. Dohok J. Sanandaj Total verbal clauses 1,577 1,742 of which with full arguments 44% 70% The two corpora were tagged for transitivity and argument structure. The value of the transitivity tag (‘intransitive, ‘transitive’ and ‘ditransitive’) was for the most part established lexically, rather than on the basis of explicit argument marking in the clause. The meaning of the few verbs with variant valency was determined 1 This work complies therefore with the University’s policy on collaborative research, cf. https://www.cambridgestudents.cam.ac.uk/your-course/examinations/graduate-exam- information/submitting-and-examination/phd-msc-mlitt/research#collaboration. 366 WORD ORDER through context; for instance, ‘make’, can function transitively in one context (‘make bread’), and ditransitively in another (‘make grains into flour’). Following Andrews (2007), the following types of argument were distinguished: the S(ubject) of intransitive clauses, the A(gent)-like subject of mono- and ditransitives, and the P(atient)-like object of monotransitive verbs. While many word order studies focus exclusively on transitive clauses, I made a special effort distinguish As from Ss, since they can exhibit distinct syntactic behaviour (cf. Haig and Schnell 2016). In the present chapter, this holds especially for J. Dohok, and merits the treatment of intransitive clauses as a syntactic structure to be characterised on its own terms. In this chapter, the term ‘subject’ refers to the subject of all clauses (intransitive and transitive), while ‘Subject’ is reserved for intransitive constructions. For ditransitives, T(hemes) and R(ecipients) are distinguished (cf.§6.1) after Malchukov, Haspelmath and Comrie (2011). When discussing objects in general, however, both T and P are subsumed under O(bjects).2 This is because Ts, like Ps, tend to agree with the verb in diffential object marking (see §6.4). In some types of ditransitives, it is however R which triggers agreement (see §9.6 below). 2 Distinguished from the general category ‘objects’, which also subsumed Recipients. SYNTAX AND MORPHOSYNTAX 367 Prepositionally expressed arguments (of intransitive or transitive; e.g., ‘look at’) are termed here ‘oblique’. One particular type of oblique discussed here in its own right are Goals (see §9.8.1). Thus, ‘oblique’ is used here for all prepositional arguments other than Goals. The category G(oal) subsumes: all prepositional goal- like arguments expressing the destination of movement verbs and of caused motion (like ‘bring’), the recipient of ditransitive verbs like ‘give’ and ‘show’, the addressee of ‘say, tell’, ‘speak, tell’ and ‘ask’, beneficiaries and the complement of ‘look at’. Occasionally, the prepositional marking on Goals of movement verbs is dropped in J. Sanandaj,3 but such constituents are still subsumed under Goals here. The non-prepositional recipient-like argument in double object constructions, however, is treated distinctly from Goals. Goals of intransitive movement verbs are generally considered an argument here, since they arguably display some argument-like properties (see below). Finally, the TZ varieties use a number of complex predicates ('phrasal verbs') particularly with ʾ-w-l ‘do’ and d-r-y ‘throw’ (Khan 2009, 153).4 Such nominal elements are not considered objects in the study, unless they trigger agreement with the verb. The following grammatical types of arguments are distinguished here: 3 For examples, see §9.8.1 below. 4 This itself is a Kurdish and/or Persian pattern borrowing (Khan 2009, 153). In the Kurdish literature, such arguments (of ‘do’) are generally known as light verb complements (Haig 2019, 8). 368 WORD ORDER a) full: nominal and independent pronominal arguments, possible for all subjects and objects affixal: objects expressed as verbal affixes b) pronominal: personal pronouns, whether affixal or full, but not indefinite pronouns (e.g. xa ‘a certain’) nominal: full arguments other than definite pronouns: noun phrases, indefinite pronouns and numerals Different expressions of the same argument tend to show distinct syntactic behaviours. Affixal arguments appear in a fixed position as a bound verbal morpheme, so they cannot be included in the positional counts for arguments. Within the group of full arguments, pronominal arguments tend to differ from nominal ones in that the latter prefer the clause-initial position. To demonstrate the significance of frequencies attested for various word order permutations, tests of statistical significance (‘p’) are provided.5 The Chi square is employed when the total number of observations is above 20 and none of the expected frequencies are below 5. Otherwise, the Fisher exact test is used (Levshina 2015, 213–4). For all tests, the value of degrees of freedom is 1, unless stated otherwise. Highly significant probability values (including more than four decimal points) were kept at p<0.0001. 5 For statistical significance for corpus linguistics, see Stefanowitsch (2020). SYNTAX AND MORPHOSYNTAX 369 When useful, the phi coefficient (‘Φ’) is also supplied. This coefficient calculates Figure 5: Phi coefficient (Stefanowitsch 2020, 153) the strength of association between two variables,6 and its extreme values are either 1 for a complete association or 0 for no relationship:7 Results of significance tests and counts of attested data are presented in contingency tables, such as the illustrative one provided below. The table presents the attested frequencies and—if useful—expected frequencies in parenthesis. It is accompanied by statistical significance and the phi coefficient: Table 45: Illustrative contingency table VS SV total old 5 (10) 11 (6) 16 new 23 (18) 4 (9) 27 6 http://methods.sagepub.com/reference/the-sage-encyclopedia-of-communication-research- methods/i10603.xml 7 Technically, the value of the coefficient can be positive or negative, but this differentiation is not necessary here. 370 WORD ORDER total 28 15 43 (p<0.001, Φ=-0.55 > strong relationship) For intransitive word order, this study also employs the Multiple Correspondence Analysis (‘MCA’).8 This tool is used to establish which dependent variables are especially closely associated with one another (i.e. ‘change together’), and which are more independent (i.e. change independently of other variables). 9.2. Word order and information structure Word order reflects the syntacticisation (i.e. syntactic grammaticalisation) of semantics and information structure (e.g. Givón 1977). This study aims to identify the most frequent types of word order and deviations thereof and to suggest the pragmatic correlates of both. This, in turn, is often done by the identification of particular semantic or syntactic features of arguments. When appropriate, this analysis is furnished with comments on alignment of various arguments and the semantic-pragmatic factors which this is likely due to (cf. Donohue 2008). As for information structure, this study differentiates between broad focus (i.e. predicate focus), narrow focus and sentence focus (for sentence focus, see §9.7). Broad focus is the pragmatically unmarked type of focus and subsumes also the 8 https://methods.sagepub.com/reference/encyclopedia-of-measurement-and- statistics/n299.xml?fromsearch=true. SYNTAX AND MORPHOSYNTAX 371 object, under the assumption that the object is a part of the verbal construction (cf. Lambrecht 1994, 296-7). In narrow focus clauses, by contrast, only one argument is in focus. Clauses with narrow focus on the predicate (excluding its object arguments) also occur, and should not be confused with predicate focus clauses. Narrow focus may be counter-presuppositional, additive or feature in an answer to a wh- question. Nucleus stress of broad focus clauses typically falls on the clause-final constituent. This holds for both dialects, meaning that the stress falls on the object in J. Dohok in prototypical monotransitive clauses, since the dialects has the default subject- verb-object order. In J. Sanandaj where the order is subject-object-verb, the nucleus stress tends to be placed on the verb. In narrow focus clauses, stress falls on the constituent in narrow focus, though this appears to be is less consistent for J. Sanandaj. The most frequent word order permutations (henceforth ‘default word order’) generally reflect the grammaticalised positions of particular grammatical functions (and grammatical types such as parts of speech). Consequently, this position is associated with the prototypical pragmatic functions of those arguments, and is therefore often pragmatically unmarked. Therefore, deviations from the most frequent pattern often correlates with―and therefore signals―a shift from the unmarked word order. On the other hand, the default word order is not always pragmatically unmarked: sometimes, there is a 372 WORD ORDER formal overlap between a marked and an unmarked pragmatic structure. Moreover, a marked pragmatic structure does not always trigger an alternative word order, even when available: for instance, a speaker can choose to keep an object that is in narrow focus in its ‘unmarked’ position, rather than shifting it to a dedicated ‘marked’ slot. There are apparent differences between J. Dohok and J. Sanandaj with regard to deviations from the default word order. In J. Dohok, only Objects commonly shift when in narrow focus, moving to the preverbal position. Thus, a change in the pragmatic function of Objects in J. Dohok is typically reflected in word order. Such a position shift can also occurs for the Agent and Subject (see below), though less frequently. The handful of obliques which are in narrow focus in the J. Dohok corpus do not always shift from their default postverbal to the preverbal position. This is illustrated for the oblique ‘only in Jewish Neo-Aramaic’ in (1) below: (1) 4a/4 Context: ‘My mother could speak a bit of Jewish Neo-Aramaic, which she had learnt through speaking.’ ʾàxniˈ g-maḥk-ax-wa mən bab-an bas b-ət kùrdi,ˈ b-ət ʾò lišana.ˈ We REAL-speak-A.we-PST with father-our only in-GEN Jewish_Neo_Aramaic, in-GEN thisMS language. ‘We used to speak with our father only in Aramaic, in this language.’ In J. Sanandaj (object-verb order), Objects are usually preverbal (though see §9.6.2 below for postverbal Objects), regardless whether they are in broad or SYNTAX AND MORPHOSYNTAX 373 narrow focus. Thus, in contrast to J. Dohok, Objects in narrow focus rarely shift in position. Instead, broad versus narrow (object) focus in J. Sanandaj can be distinguished through intonation, because constituents in narrow focus tend to receives nucleus stress. In this TZ dialect, other arguments in narrow focus are also generally immediately preverbal, as is the case in AOV languages such as Korean and Armenian (Comrie 1988). This apparently applies to all arguments, including beneficiaries marked by the preposition baq- ‘for, to’, as shown in (2) below. This is despite the fact that arguments expressed by this preposition are overwhelmingly postverbal in J. Sanandaj (cf. §9.8.1 below). (2) J. Sanandaj; Khan 2009; D:9 TOPIC NARROW FOCUS PREDICATE ʾəlha xa-tăra baqa-didàn-əč k-ol-Ø-o. God a-door for-us-too REAL-do-A.he-back. ‘God will open a door also for us.’ In J. Dohok, however, only Objects regularly shift in their position when in narrow focus. With other arguments (e.g. S and A and G), this tendency is less strong and apparently does not occur altogether with obliques. ‘Clause topic’ is defined here as a referent about which an assertion in the clause is made (Lambrecht 1994). By contrast, ‘discourse topic’ is taken as a referent featuring ‘prominently’ (i.e. repeatedly) in discourse. 374 WORD ORDER 9.3. Default word order and its flexibility in clauses with a subject and another argument 9.3.1. Frequency of clauses by number of arguments The designation ‘clauses with two arguments fully expressed’ refers to clauses which contain (minimally) the subject and another argument. In J. Dohok, only 8.5% of the verbal clauses contain two arguments fully expressed, which is comparable to 7.3% for the TZ representative J. Sanandaj. 9.3.2. Word order in two-argument clauses Table 46: Overall word order typology by verb position J. Dohok J. Sanandaj V-initial V-medial V-final V-initial V-medial V-final AO 0 69/70 1/70 0 7/86 79/86 (98.5%) (92%) SG and 5/57 53/57 0 0 89/100 11/100 (93%) (89%) AG Other 4 0 2 0 3 13 The word order of such two argument clauses is relatively rigid. J. Dohok is verb- medial, having ca. 92% (122/133) clauses with preverbal S/A and postverbal O/G Verb-final clauses are virtually unattested. The AVO order is the predominant one SYNTAX AND MORPHOSYNTAX 375 in NENA in general,9 being inherited from earlier Aramaic. The transitional J. Barzani cluster (between LD and TZ; Mutzafi 2008b), resembles LD in being predominantly AVO, as I established impressionistically on the basis of the texts in Mutzafi (2002a). This basic typology stands in sharp contrast with the TZ dialects such as J. Sanandaj: for monotransitives, the most frequent word order in this dialect is AOV. The OV character of the TZ dialects is itself the interference of Kurdish, which is predominantly OV (Haig 2015; Stilo 2018). The differences between the two dialects notwithstanding, they both display a clear preference for clause-final goals, i.e., A/SVG. This order is found in 93% of clauses with S/A and G in J. Dohok, and in 89% in J. Sanandaj. In general, however, Gs are more flexible in J. Sanandaj than in J. Dohok, which is discussed below. Table 46 above shows the permutations that occur for A followed by O, and for A or S followed by G, relative to the position of the verb. The “Other” row includes other permutations like OAV, OVA or GVS etc. 9 In addition to TZ discussed below, another exception known to me are the Chaldean dialects of the South- Eastern NENA periphery such as C. Shaqlawa. 376 WORD ORDER As expected for OV languages (Siewierska 1997, 493), ‘verb first’ order does not occur in J. Sanandaj. Other attested orders—in addition to AOV—are AVO (7/202), OAV (13/202) and OVA (3/202). The latter variant is highly infrequent in languages with predominant AOV order in general (ibid). At face value, these findings seem to suggest word order is slightly more flexible in J. Sanandaj. However, they only pertain to clauses with two arguments fully expressed. In general, Objects and Subjects are more flexible in J. Dohok, while Goals and obliques are more flexible in J. Sanandaj. In general, Table 46 above shows that arguments are more likely to be expressed postverbally in J. Dohok (65% against 35%) but more likely preverbally in J. Sanandaj (69% against 31%). The preferences, however, naturally differ per argument class (S, A, O, G). While OV and VA do occur in Dohok (see further below), they are not attested together (*OVA). Similarly, VO and VA order are possible variants in Sanandaj, but apparently do not occur together (e.g. *VAO/VOA). 9.4. The core arguments S, A and P 9.4.1. Position The preverbal placement of A is strongly preferred in J. Dohok as well as J. Sanandaj, while the predominant position of P is postverbal in J. Dohok and immediately preverbal in J. Sanandaj. However, Patients also occurs in the preverbal position in 11% of clauses in J. Dohok, showing a greater flexibility than SYNTAX AND MORPHOSYNTAX 377 J. Sanandaj. The latter dialects prefers the PV order, only allowing the VP order in 4.4% of cases. The difference between these two dialects in their deviation frequency of P is statistically significant (p<0.0001). The placement of objects in ditransitive constructions is more complex and is discussed in .§9.6 below. The Subject in Dohok has a tendency to occur postverbally, featuring in that position significantly more often than A (p<0.0001), while its (preverbal) position in J. Sanandaj is relatively more rigid. The verb-subject order should therefore be considered a feature of intransitive clauses. The general clause-initial placement of subjects in both dialects means that in pragmatic terms, the clause-initial position is associated with the topic role. This is consistent with the cross-linguistic findings for subject-initial languages (cf. Siewierska 1998, 501). This reality is also observable in the placement of the highly topical independent personal pronouns. Regardless of their grammatical function (S, A or P), independent personal pronouns are typically placed in the clause initial position. Table 47: Position of the core arguments S, A and P relatively to the verb J. Dohok Total J. Sanandaj Total _V V_ _V V_ A 170 16 (8.6%) 186 208 7 (3.3%) 215 s 98 64 (40%) 162 185 11 (5.6%) 196 P 32 (11%) 254 286 393 18 (4.4%) 411 378 WORD ORDER 9.4.2. Frequency of full expression Another apparent difference between S, A and P concerns the frequency of their full expression in the clause and in the argument type used to express them (pronominal vs nominal for all, and full vs affixal for P arguments), as displayed in Table 48 below. Since cross-indexing of S and A co-nominals is obligatory in NENA (cf. Noorlander 2018, 133–4), only P can be left unexpressed. Thus, the category ‘affixal’ is only relevant for P, whose expression is optional. ‘Affixal’ further means ‘expressed by affix only’, and therefore excludes Patients marked differentially. It is apparent that the Subject is more likely to be fully expressed than Agent (for both dialects, this difference is statistically significant (p<0.0001), with Patients being expressed fully most frequently. Moreover, P is much more likely to be expressed by a noun phrase, while A is least likely to be expressed in this way. This is consistent with cross-linguistic corpus data (Haig and Schnell 2016). Table 48: Argument frequency per argument’s grammatical type J. Dohok J. Sanandaj A S P A S P Total verbal clauses 1039 538 614 1223 524 700 of which full nominal 10% 24% 47% 11% 28% 55% of which independent pronouns 8% 6% 0.4% 7% 10% 3% SYNTAX AND MORPHOSYNTAX 379 of which affixal pronouns 25% 15% The higher frequency of affixal expressions of P in J. Dohok vs J. Sanandaj is due to the set of independent object pronouns used in the TZ varieties. Such independent object pronouns do not exist in the LD dialects and independent personal pronouns generally not used for this function. In these dialects, animate topical referents for the P are therefore expressed affixally. The basis of such pronouns is preposition (ʾəl)l- and/or the particle did-. The ʾəl-based pronominal objects always follow the perfective past verb in the Sanandaj corpus.10 When postverbal, however, the ʾəl-series tends to coalesce with the preceding verbal form, e.g. grəš-li ʾəlaf > grəš-li-laf (cf. Khan 2009, 158).11 This unmarked postverbal position for pronominal objects in Sanandaj against preverbal full noun phrases is typologically unusual and may reflect the original AVO structure of this language. This ʾəll-series is used independently typically only in preverbal position in pragmatically marked contexts (e.g. narrow focus). This ʾəll-series is excluded from further discussion. 10 See §6 for further details on the use of ʾəll- in NENA. 11 See Noorlander (2018, 410) for possible motivations for this. 380 WORD ORDER 9.4.3. Position of independent personal pronouns Full non-prepositional personal pronouns are overwhelmingly preverbal in both dialects, regardless of their grammatical function (subject or object). Pronouns (except indefinite ones) are inherently definite, and personal pronouns typically refer to animate entities. Being definite, pronouns are thus typically ‘discourse salient’; for instance, topical or in narrow focus. In J. Dohok, Patients expressed as independent personal pronouns (two cases) and as demonstrative or anaphoric pronouns (six instances) are always clause initial or at least preverbal. An example of a topical Patient in J. Dohok is given in (3) below. In J. Sanandaj, definite pronouns for subjects and objects are also preverbal, but this is less triking given the A/SOV order. An example of a preverbal P in J. Sanandaj in narrow (additive) focus is presented in (4) below. (3) 3d/28 Context: ‘I saw two men in the street, they looked at me and one said to the other:’ ʾeha qṭùḷ -Ø -laˈ huðèθā =la.ˈ thisF IMP.kill -A.youSG -P.her Jewess =COP.PRS.she ‘This one – kill her, she’s a Jew.’ (4) J. Sanandaj (Khan 2009, 412, A:17) TOPIC P=NARROW FOCUS PREDICATE ʾana did-ì-č Ø-qaṭl -i I P-I-too IRR-kill-A.they ‘They will kill me too.’ SYNTAX AND MORPHOSYNTAX 381 9.5. Transitive clauses 9.5.1. Os and general comments on information structure In J. Dohok, change in the position of Objects (Patients and Themes) is more frequent, apparently because there are more pragmatic functions which trigger such a change, relatively to J. Sanandaj j. In J. Sanandaj, the ‘default’ preverbal position covers several other marked pragmatic functions. 9.5.1.1. Position and differential object marking In J. Dohok, there is a noteworthy correlation between differential object marking (cf. §6.4) and word order in that differential object marking is significantly more likely for preverbal objects, as shown in Table 49 below. No such correlation of position and differential object marking exists J. Sanandaj (p>0.05), meaning that J. Dohok shows a greater degree of flexibility in the placement of definite objects than J. Sanandaj. Table 49: Differential object marking through verbal agreement and object position in J. Dohok OV VO Total (row) Differential object marking 27 (8) 46 (65) 73 No differential object marking 16 (35) 303 (284) 319 Total (column): 43 349 392 [p<0,0001, Φ=0,4] 382 WORD ORDER 9.5.1.2. Object defocalisation Both dialects use the strategy of changing the position of the O (to the preverbal position in J. Dohok and postverbal in J. Sanandaj) to indicate that the object does not belong within the focus group, but is discourse-old, and likely topical. In sentences such as (5) below, such object defocalisation results in narrow on the predicate. Many sentences with preverbal objects in J. Dohok—like the one in (5) below—can also be regarded as topicalisation through clause-initial position (see following paragraph). The reason for this is that the preverbal position is often identical with the clause-initial position, and that the argument involved is often a discourse-old topic. However, in contrast to object defocalisation, topicalisation through clause-initial position often involves a prosodic separation of the argument, and can also involve new discourse topics, which is not the case in (5) below. The two types of clauses are thus kept separate, though some sentences can arguably be categorised as either. As implied above, object defocalisation in J. Dohok is also signalled by the lack of nucleus stress on the object (cf. example (5) below). In this dialect, 60% (25/42) of nominal preverbal OS do not carry nucleus stress, suggesting that such object defocalisation is the most frequent reason for its shift to the preverbal position. (5) 1d/16 Context: ‘A lorry came, a lorry car, full of Syriacs.’ P=TOPIC V=FOCUS u-kull-u ʾan suraye qam-qaṭḷ-ì-lu. SYNTAX AND MORPHOSYNTAX 383 And-all-them these SyriacS PERF-kill-A.they-P.them ‘And they killed all of these Syriac Christians.’ In J. Sanandaj, defocalised objects also change their position, appearing postverbally. Clause-final position has been argued to be the preferred slot for foci cross-linguistically, though this does not generally hold for verb-final languages (Herring and Paelillo 1995). In J. Sanandaj, postverbal Objects are focal only in 4/22 (18%) cases,12 indicating that the postverbal slot is also not the prototypical focus position in this dialect. However, there is one particular type of focus which does occur postverbally in J. Sanandaj, as shown in §9.6.2 below. 9.5.1.3. New topics or topic reactivation through clause-initial position and prosodic separation (all arguments) In both dialects, any constituent can be placed clause-initially and thus coded as a new discourse topic or a reactivated discourse-old topic in cases of topic switch. This is almost always signalled prosodically by placing the noun phrase in its own intonation unit and giving it nucleus stress. Such topicalisation is illustrated below in example (6) for a discourse-new Object in J. Dohok, and for a Goal serving as a reactivated topic in J. Sanandaj (7). A similar strategy of fronting and prosodic separation also occurs for contrastive topics, as illustrated in the sentence in (8) from J. Dohok, which includes a subject and an object contrasted with each other. 12 According to Khan, this is done in order to bind it to the immediately preceding discourse (2009, 345). 384 WORD ORDER Topicalised objects or Goals of ditransitive tend to be resumed later in the clause through a verbal agreement affix. Topicalisation through clause-initial position is attested cross-linguistically (cf. Lambrecht 1994, 160–2), likely excluding verb- and object-initial languages, whose clause-initial position can be expected to be associated with foci. (6) 5a/1 Sehràneˈ g-oð-ax -wa… Sehrane REAL-do-A.we -PST… ‘(As for) Sehrane—we used to celebrate (it) (…after we finished Passover).’ (7) J. Sanandaj (Khan 2009, 436, A:106) Context: ‘In the evening, the merchant’s neighbour had to go home…’ tajər-ăkèˈ ḥăsab-ef Ø-haw-Ø-le-o baq-èf merchant-DEF account-his IRR-give-A.he-T.it to-him. ‘(And as for) the merchant—(the neighbrour) would give his account to him.’ (8) 1c/98-9 Šwir-ət13 xàbraˈ là k-eθ-e nšaya.ˈ Wound-GEN word NEG REAL-come-S.he INF.forget. ‘A wound caused by word is not/will not be forgotten.‘ Šwir-ət ḍə̀rbaˈ naša g-našè-Ø-le.ˈ wound-GEN blow man REAL-forget-A.he-P.him. ‘(but) a wound (caused by) a blow one can forget.’ 13 This form apparently means ‘wound’, but its origin is uncertain. SYNTAX AND MORPHOSYNTAX 385 In the case of subjects—which tend to be clause-initial in both dialects—their status as a new topic is only signalled by prosodic separation. This is illustrated in the sentence in (9) below, which opens a story about speaker’s father: (9) 1c/a Bàb-iˈ g-marpe-Ø-wa də̀mma.ˈ Father-my REAL-treat-A.he-PST blood. ‘(As for) my father, he used to treat (people by releasing) blood.’ 9.5.1.4. Arguments in narrow focus In J. Dohok, an Object which is preverbal and bears nucleus stress is narrow focus. This occurs in 5/42 instances (12%) of preverbal objects in J. Dohok and is illustrated in (10) below with an object bearing additive focus. The predicate is out of focus, since the previous clause has already introduced the imperative ‘bring/buy!’ (10) 1b/51-2 Context: ‘Buy me seven jugs!’ u-šoʾa ḥammàre žik muθ-un!ˈ And-seven donkey_drivers also IMP.bring-A.youPL ‘And bring me also seven donkey drivers!’ For other arguments in J. Dohok, a regular change in position to indicate narrow focus is difficult to observe. Sometimes, As in narrow focus shifts to the postverbal position (§9.5.2) and Goals—like Objects—move before the verb (§9.8.1). 386 WORD ORDER In J. Sanandaj, an argument in narrow focus tends to be immediately preverbal. Such an argument bears nucleus stress if it is indefinite (cf. Khan 2009, 344-5). Narrow focus on objects, therefore, does not result in the change of the argument’s position, as illustrated by (11) below. For Agents and oblique, a change may be observable, depending on the number of arguments, while the movement of Goals to preverbal position is apparent. (11) J. Sanandaj (Khan 2009, 454, B:47) pəsrà la k-əxl-i-wa hulae. meat NEG REAL-eat-A.they-PST Jews ‘(It was) meat (that) the Jews did not eat.’ 9.5.1.5. Clause repetition and word order inversion in narrative A stylistic device characteristic of NENA narrative—including dialogue within narrative—is clause repetition, illustrated in (12) and (13) below.14 Such repeated clauses are sometimes characterised by the inversion of the position of object, relatively to the verb. This in turn is apparently motivated by the stylistic preference to avoid a verbatim clause repetition. In many NENA dialects, such inversion is also attested for other constituents, but in the J. Dohok corpus, this strategy is only attested with objects (cf. example (13) below). However, in neither 14 There are several types of this repetition formally and functionally (see Coghill 2009, 277). SYNTAX AND MORPHOSYNTAX 387 of the two dialects studied is such an inversion obligatory (cf. the examples in Khan 2009, 376). (12) J. Sanandaj (Khan 2009, 439, B:4) ba-ʾaql-e ʾay jəl-e ʕùč-Ø-lu! with-foot-PL these garment-PL IMP.trample-A.youSG-P.them ‘Trample the clothes with your feet!’ ʕúč-Ø-lu ʾe jəl-ăkè! IMP.trample-A.youSG-P.them these garments-DEF ‘Trample the clothes! (And I shall wring them out).’ (13) 1c/27-8 băle šqul-Ø-le năr-a dìdox.ˈ But IMP.take-A.youSG-P.him axe-MS yourMS 'But take your axe.’ năra didox šqùl-le…ˈ axe yourMS IMP.take-A.youSG-P.him ‘Take your axe (and hit me on the head with it as strong as you can)!’ 9.5.2. Agents As noted above, the postverbal position is associated in a statistically significant way with the subjects of intransitive, rather than transitive clauses. Still, the VA order does occur in 8.6% cases (16/186). Moreover, in 3/16 cases, this order occurs with an independent personal pronoun, despite such pronouns’ preference for the clause-initial position. For the subject of intransitive, the VS order with pronouns is unattested. It is also noteworthy that in 6/16 cases, a postverbal A occurs with the verb Ø-m-r ‘say’. 388 WORD ORDER The sentences in (14) below contain two cases of postverbal A, the latter is in narrow focus. The speaker responds to a situation where no one could tell the king where the middle of the world is, claiming that he will be the one to say it. One of the functions of the VA order is thus apparently to code the Agent as carrying (narrow) focus. This would be to some extent parallel to the shifting objects to the preverbal position. This, in turn, would suggest that with narrow focus, the key feature in J. Dohok is the change of the argument’s default position, while in topicalisation, the strategy is always clause initial placement. It is possible that with the verb ‘say’, the Agent tends to be postverbal precisely because of its narrow focus function: in clauses accompanying dialogue (e.g., ‘X said…’ ‘Y replied…’), the act of speech is often presupposed, the focus being therefore only on the speaker. Still, not all postverbal As are in narrow focus, and narrow focus does not always trigger a change in place. Occasionally, the VA order even result in the ‘V-argument-argument’ order, which clearly violates the preference for the verb to be medial in J. Dohok. (14) 1a/13 g-emǝr-Ø ṭa-lu huðàya,ˈ g-ēr-Ø, ʾàna Ø-maxalṣ-ən-nu,ˈ REAL-say-A.he to-them Jew, REAL-say-A.he, I IRR/FUT-save- A.IMP.them ‘(The) Jew says to them, he says ‘I will save them…’ b-amr-ǝn-ne ʾàna ʾeke=la palgut-Ø dunye. FUT-say-A.IM I where=COP.PRS.she middle-GEN world. ‘(it is) I (who) will tell him’ where the middle of the world is.’ SYNTAX AND MORPHOSYNTAX 389 9.6. Ditransitive clauses The class of ditransitives includes a set of verbs which take two object-like arguments (Haspelmath 2015, 20). Two distinct types of ditransitives occur in NENA and are referred to as constructions with a direct and an indirect object (‘constructions with an indirect object’ for short) and double-object constructions. Constructions with an indirect object are by far the most common type, while double-object constructions are lexically restricted (Noorlander 2018, 394). In the latter group, the differentiation between ‘Recipients’ and ‘Themes’ is semantic in this investigation: the R-like object represents is a (physical or abstract) recipient or an endpoint. Syntactically, these two types of ditransitives in both J. Dohok and J. Sanandaj differ in the position of the object arguments, their coding (zero or prepositional) and in differential object marking (whether it is triggered by T or R).15 The primary object is the one which triggers verbal object agreement.16 In constructions with an indirect object, the T-argument is primary, while in double object constructions, R is usually primary. 15 For the morphosyntax of ditransitives which have affixal objects, see §6.3. 16 Cf. Dryer (2007, 1254-5). Languages differ regarding which object of ditransitive (if any) aligns with P. The alignment diagnosis also depends on the criteria selected to determine this. Cross-linguistically, however, it is T which is semantically more patient-like (i.e. it is directly affected). 390 WORD ORDER These distinct patterns (indirect object vs double-object constructions) are generally due to the referential prominence of the arguments.17 More specifically, double-object constructions are used with verbs whose Recipient argument is likely to have an inanimate referent (thus, R=T, or R 9.6.1. Constructions with an indirect and a direct object In most ditransitive constructions in J. Dohok—as in NENA in general—the Recipient is expressed with a preposition meaning ‘to, for’, that is, indirectly: either ta + nominal, or ṭa-L-suffix for pronominal arguments. As Table 50 illustrates, the clause-final position of R is preferred in J. Dohok, and both arguments tend to occur postverbally: 17 For prominence hierarchies, see Haspelmath (2015, 28). SYNTAX AND MORPHOSYNTAX 391 Table 50: The position of R (indirect object) and T (direct object) in J. Dohok _V_ V_ _ RT 0 7 TR 7 8 In J. Sanandaj, the most frequent order for all types of ditransitive is TVR, meaning that the Recipient is prototypically postverbal. In indirect object constructions in NENA in general including J. Dohok, the Theme is the near-exclusive trigger of verbal object agreement, for example: (15) 5a/5 ʾaval ʾana ʾatta Ø-mabl -i -li l-Dŏhok̀ ,ˈ But I now IRR/FUT.take -A.they -T.me to-Dohok ‘But as for me, if they took me to Dohok now…’ k-taxr-ən-na kùll-a… REAL-remember-IM-her all-it. ‘I’d remember all of it.’ Thus, the T is primary object, and parallels with the Patient of monotransitive clauses. The Theme also shares with the Patient its zero-marking (lack of a preposition). In J. Dohok, the objects’ linear order is pragmatic: a topical object always precedes the focal one. In fact, topical Ts are preverbal in this NENA dialect, as shown in (15) above. One of the attested preverbal Ts is in narrow focus. In C. Barwar, topical Themes (‘anaphorically bound to the preceding context’) may also occur preverbally (Khan 2008b, 879), for instance: 392 WORD ORDER (16) C. Barwar (Khan 2008b, 879, A:29) Context: ‘I’ll divide my share into two parts and give one to you and one to the vizier.’ ʾana sami t-ya -n -ne ṭla-lɛ̀xu.ˈ I portion-my FUT-give -A.IM -T.it to-youPL ‘I shall give my share to you.’ The linear ordering of objects in indirect verb constructions in J. Dohok is consistent with cross-linguistic findings for VO languages (Haspelmath 2015, 33): topical objects precede focal ones. In a related fashion, more referentially- prominent objects tend to precede less referentially-prominent ones (Heine & Koenig 2010, 94). This is parallel to the clause-initial placement of independent personal pronouns, which are inherently prominent. However, the ordering relative to the verb (TVR) in NENA is apparently typologically unusual. In ditransitive constructions cross-linguistically, both objects are typically placed in the usual position of the P, that is, postverbally in VO languages (Haspelmath 2015, 31). From the internal viewpoint of J. Dohok, however, the preverbal position of topical (and narrow focus) objects of ditransitive is not surprising, since similarly motivated shifts to the preverbal position are attested for the object of monotransitive. Perhaps the word order in J. Dohok and in NENA is more flexible and more pragmatically determined than in some other VO languages.18 18 This is impossible to determine due to the low numbers of ditransitives attested for J. Dohok. Moreover, Haspelmath does not comment on the effect of the association of the clause-initial position with topicality on SYNTAX AND MORPHOSYNTAX 393 Similarly, the Recipient precedes the Theme if the former is topical. Furthermore, recipients expressed as prepositional pronouns (ṭa-L-suffix) are always immediately postverbal. Pronominal objects are inherently ‘prominent’ within the referential prominence hierarchy, having definite and human referents,19 and thus having topical referents. Such arguments are thus apparently likely to be placed earlier in the clause. However, their placement preverbally is apparently dispreferred due to their prepositional marking. This immediately postverbal position of prepositional pronouns for the R contrasts with the position of other Goals (Recipients or otherwise), which are typically clause-final. The following two clauses illustrate this difference: (17) 3c/18 Zun-nu diyàre ta bab-i.ˈ PERF.buy-A.they gifts for father-my ‘They bought presents for my father.’ (18) 3c/20 ʾaxni məθe -lan ṭa-lox diyàre.ˈ We PERF.bring -A.we to-youMS gifts ‘We brought you presents.’ the placement of objects of ditransitive. In principle, preverbal placement of topical objects should be possible due to such an association. 19 Nominal hierarchy: pronominal > definite nominal (proper, kin) > indefinite nominal > common; animacy hierarchy: human > animate > inanimate. 394 WORD ORDER A noteworthy exception to the above-mentioned rule that that only T triggers agreement in constructions with a direct and indirect object is found in J. Betanure (LD). In this dialect, the R-like argument of Ø-m-r (ʾamər) ‘give R the name T’ can be expressed either directly—which results in a double-object construction—or indirectly. Even when indirect, however, the R-like object can trigger agreement, for example: (19) J. Betanure (Mutzafi 2008a, XIII: 278) ta qaš-əd suraye g-əmr -áx -wa -le kòma.ˈ for priest-GEN Syriacs REAL-say -A.we -PST -R.him black. ‘We used to call a Christian priest the black (one).’ The verb Ø-m-r ‘give R the name T’ in J. Betanure are therefore ‘in between’ indirect object constructions and double-object clauses. Agreement is reminiscent of direct-object constructions (it is triggered by the R-like argument), but the R itself can be coded either as a direct or indirect object. As a general rule, the indirect object construction in NENA is characteristic of objects in which Recipient is referentially referentially more prominent than Theme; in other words, the R is more likely to be human. However, as mentioned above, the split between different ditransitive constructions is lexical (i.e. verb- rather than argument-based) in NENA: only specific verbal lexemes can occur within the double-object constructions, rather than any verbs which have a non- prominent R. Consequently, one also finds non-human R arguments in constructions with an indirect object if the verb in question does not take two SYNTAX AND MORPHOSYNTAX 395 direct objects. Still, there is a special strategy to distinguish between prominent and non-prominent Recipients. An animate beneficiary-like Recipient is distinguished from a destination Recipient through the preposition used: ta/ṭa- ‘for’ is associated with animate entities, while kəs ‘at (someone’s place)’ triggers a destination-like interpretation, even with a human referent. This is illustrated in (19) below with the verb m-θ-y (meθe) ‘bring’, which always takes a direct and an indirect object. Its Recipient is non-prominent (a location). (20) 3c/14 Qam-məθ -e -le kəs bàb-i.ˈ PERF-bring -A.they -T.him R.at father-my (Then) they brought him to my father’s (house) (not ‘for my father’). In fact, the group of indirect object constructions even includes some verbs that prototypically have an inanimate R argument: d-r-y ‘place’ and m-x-y ‘hit T on R’, i.e. ‘hit R with T.’ The Recipients of these verbs are coded with prepositions such as go ‘in’ or rəš ‘on’: (21) 3b/23 g-oð-i-wa qàlya,ˈ g-dār -e -wa -lu go HkadimH hatxa zòre…ˈ REAL-do-A.they qalya, REAL-place -A.they -PST -T.them in jugs so small ‘They used to make qalya. They would put them (i.e. the pieces of meat) in jugs this small…’ 396 WORD ORDER 9.6.2. Constructions with two direct objects There are two types of double-object construction in NENA, distinguished by which object can trigger verbal agreement. 9.6.2.1. Verbs of filling and naming The J. Dohok corpus includes three verbs in which the R-like object triggers agreement (and is thus primary): ṃ-ḷ-y (ṃaḷe) ‘fill’, and Ø-m-r (ʾamər) and ṣ-ṛ-x (ṣaṛəx), both meaning ‘give R the name T’. A similar situation with lexically- restricted double-object construction is attested in other NENA dialects. For C. Barwar, for instance, Khan lists the verbs ʾ-w-ð ‘make T into R’, ṃ-l-y ‘fill’, d-w-q ‘put’ (Khan 2008b, 785). In Khan’s grammar of C. Urmi, the T-like argument of the verb ‘fill’ is considered a ‘predicative complement’ rather than an object (Khan 2016, vol.2, 363); apparently because—like in J. Dohok—such a ‘Theme’ is optionally introduced by the preposition b- ‘with’. As regards the arguments’ prominence, verbs of calling tend to have a proper or a categorising noun (e.g. species name) as the T, meaning that the T argument can be more prominent than R.20 In verbs of filling, both objects are prototypically inanimate (i.e. T=R). Thus, direct object constructions as a general rule occur 20 This, however, does not apply to clauses such as 1e/7 below, where the R-like argument is a proper noun, and the T-like one is indefinite. SYNTAX AND MORPHOSYNTAX 397 with verbs whose prototypical R-like argument is not more prominent than the T- like argument (though see below). In the J. Dohok corpus, the R-like object usually precedes the T-like one, because it tends to be topical, as illustrated by the example in (22) below: (22) 1e/7 Dǒhok k-sarx-í-wa-la MHʿìrMH,ˈ bằlad,ˈ màθa.ˈ Dohok (R) REAL-call-A.they-PST-R.it city, town, town (T). ‘Dohok used to be called city (word repeated three times).’ For verbs of filing, the attested permutations are RVT (3 cases) and VRT (2). For naming, only RTV occurs (4). The RT (and even RVT) order is due to R’s topicality. This is especially apparent in verbs of naming: in these verbs, the R always occupies the topic (i.e. clause-initial) slot (cf. example (22) above); such clause predicate about what name a Recipient is given. This preverbal position of R in the verbs of naming is also characteristic of other LD dialects. In J. Betanure, for instance, the RVT order is found in 14/17 of ‘giving R the name T’ clauses (counted for clauses with full nominals R and T). The RVT order occurs not only with definite referents, but also with indefinite category representatives (e.g. ‘a priest’). These clauses include cases of both directly and indirectly expressed Rs (see above). The corresponding double-object constructions in J. Sanandaj are represented in the corpus by the verbs Ø(ʾ)-m-r (hămə́r) ‘give R the name T’ and z-m-ṭ (zămə́ṭ) ‘fill’. In verbs of naming, a pronominal Recipient of the name can be expressed 398 WORD ORDER either with an L-suffix or with the independent preposition baq- ‘to, for’ with a pronominal suffix, while in J. Dohok, such objects are always direct (i.e. not prepositional). In such cases, the R is therefore an indirect object (but for convenience, these clauses are included in the present section). The attested permutations are shown in Table 51 below. With both verbs, the T is often postverbal, which is unusual for this dialect.21 In OV languages, both objects of ditransitive are expected to be preverbal, with the topical or referentially more prominent one preceding the focal one (cf. Haspelmath 2015, 31). In all of the attested clauses of naming, the R is topical and T focal. The more ‘expected’ TVR order with an indirect Recipient is illustrated in (23) below, while an example of VRT is given in (24). (23) J. Sanandaj (Khan 2009, 445, B:68) šaplultà k -əmr -í -wa baq-éf.ˈ šaplultà REAL -say -we -PST to-him ‘They used to call it “šaplulta”.’ (24) J. Sanandaj (Khan 2009, 454, B:75) k -əmr -əx -wa baq-ef tuḷèˈ ʾilan-è xelapa 21 Cf. Khan (2009, 379), who states that in double-object constructions, ‘one of the objects is generally placed after the verb.’ In the verb of naming, ‘the nominal complement’ (i.e., mostly the Theme, as Khan’s examples suggest) is usually postverbal.’ Khan’s findings for clauses on naming apply to all clauses with such T-like argument. Such an overwhelming preference for the postverbal position, however, is only clearly observable in all clauses taken together, which includes cases of the R-like argument expressed affixally. SYNTAX AND MORPHOSYNTAX 399 REAL -say -we -PST to-him shoots trees-GEN willow ‘We called them shoots, of willow trees.’ Information structure does not offer a comprehensive explanation for the postverbal Themes either. In J. Sanandaj, (narrow) focus constituents are immediately preverbal (see above), while topics are clause-initial. According to Khan, objects can be placed postverbally when they are discourse-new arguments which become significant for the ensuing discourse (Khan 2009, 345), that is, when they carry presentational focus (Herring and Paolillo 1995, 165). Presentative foci can indeed occur postverbally in other AOV languages (Herring and Paolillo 1995, 191-2). While some cases of the RVT and VRT order in J. Sanandaj ditransitives can be ascribed to presentational focus on the Theme,22others cannot.23 In fact, this postverbal placement of a focal argument is highly reminiscent of a VO language such as J. Dohok, in which postverbal arguments are associated with focus. The Themes in such constructions could also be regarded as predicative complements (cf. Khan's analysis of C. Urmi - 2016, vol.2, 363). In such case, the example in (24) above would be translated ‘We used to say (that they were) shoots, of willow 22 E.g. the clause in (24) above; the object ‘shoots of willow trees’ becomes the topic of the following discourse. 23 E.g. the object jážn-e făṭir ‘unleaven bread’ in Khan (2009, 450; B:36), which is not a key topic in the subsequent discourse. 400 WORD ORDER trees.’24 Ditransitive and predicative complement constructions in Kurdish dialects (of the Sanandaj area) should also be investigated to determine whether the word order attested in J. Sanandaj is characteristic of other OV languages of the area. Verbs of ‘making T into R’ The verbs of ‘making T into R’ also employ double-object constructions in J. Dohok and J. Sanandaj, but show a pattern distinct from other types of ditransitive. Their primary object is T. Agreement with the T, in turn, is likely due to the subject’s frequent topical function. In J. Sanandaj, the TVR order is used. This order is as expected, given the preference for Goals to be postverbal. In J. Dohok, only TVR (illustrated in (25)) is attested (1/1). (25) 1d/9 ʾan šəble g-oð -í -wa -lu hə̀ṭṭe.ˈ those ears_of_grain REAL-do- A.they -PST -T.them grains ‘Those ears of grain they would thresh into grain.’ 6.3. Summary of ditransitives Table 51 summarises the findings of this section: 24 I am indebted to Geoffrey Khan for this suggestion. SYNTAX AND MORPHOSYNTAX 401 Table 51: Properties of ditransitive constructions in J. Dohok and J. Sanandaj Constructi Attested word Primary R Argument Most often on type order (J. object prominence attested in Dohok above, (triggers topic role J. Sanandaj agreeme below) nt) Indirect VTR (8), VRT T indirect R>T, object (7), TVR (7) also R=T TVR (vast majority) Double- RVT (3), VRT R direct R=T, R object (2) verbs of filling R (dir.) VT VR (ind.) T Double- RVT (4) R Direct, T>R, also R>T R object except verbs of RVT (3), VRT naming pronominal (2), TVR (4) in J. Sanandaj Double- TVR (1), VTR T direct R=T T object (unattested, verbs of likely possible) making T TVR (majority) into R) As demonstrated above, most of the word order permutations (including the typologically unexpected ones such as TVR in J. Sanandaj) are explained with recourse to information structure. The role of information structure is especially 402 WORD ORDER clear for J. Dohok, where (zero-marked) topical objects are commonly preverbal. This explains the predominant RVT order for naming verbs, and means that word order patterns are different for different types of ditransitive, depending on the prototypical roles of the objects involved. This reality prevents us from creating purely argument-based generalisations about alignment of ditransitive clauses in general. The second factor which explains some of the data, in the case of J. Sanandaj is the tendency for Recipients (i.e. Goals) to be clause-final. This accounts for the TVR order, attested in indirect object constructions and in verbs of ‘making T into R.’ The feature which is yet to be satisfactorily explained is the postverbal placement of the Theme attested in double-object constructions with the verbs of filling and naming in J. Sanandaj. Some cases of such postverbal Ts order could be explained as presentative foci. However, this order is also highly reminiscent of the VO/predicate-focus order in dialects such as J. Dohok. A possible solution is also to regard them not as arguments, but as predicative complements. Another conclusion is that in double-object constructions, agreement is triggered by the object which is more prototypically topical. Thus, in verbs of naming, the R is always topical in the corpora of both dialects, and it triggers agreement. This tendency (R as trigger of agreement and often a topic) is also apparent (to a lesser degree) for verbs of filling in J. Dohok. In verbs of making T into R, in both SYNTAX AND MORPHOSYNTAX 403 dialects, only T is attested in the topical role, and only T triggers agreement. This observation is compatible with the insight that definiteness (and therefore often discourse-topicality) is the primary trigger of verbal object agreement in J. Dohok, and in many other NENA dialects (cf. §6.4). Indirect object constructions, however, clearly do not conform to such generalisations about prototypical prominence of Themes and Recipients. Most verbs are likely to have a topical and/or a referentially prominent R (e.g. a human recipient of the verb ‘give’ or an addressee of ‘say’). Such (prepositionally marked) Recipients, however, cannot be cross-referenced with a verbal agreement affix, with the exception of verbs of naming in J. Betanure). In J. Dohok, only the T can trigger agreement when it is definite, and thus often discourse-topical. 9.7. Intransitive clauses: Subjects (Ss) and the argument of existential clauses 9.7.1. Sentence focus’ A distinctive feature of J. Dohok in contrast to J. Sanandaj is the aforementioned tendency for S to occur postverbally. The tendency is especially strong for nominal S arguments: Table 52: Position of nominal Subjects in J. Dohok SV VS total 404 WORD ORDER All full 98 63 161 arguments (39%) Nominal 64 62 126 arguments (49%) In many languages with AVO order (e.g., Italian), the VS order is a syntactic device for marking, broadly speaking, sentence focus. In general, ‘sentence focus’ is also the explanation for the VS order in J. Dohok. In J. Dohok, the association of VS with sentence focus is not exclusive; this order likely has other functions too (e.g. narrow focus on the subject or discourse topic reactivation), but the main question of this section is whether there is a meaningful (i.e. statistically significant) association between the VS order and sentence focus. Sentence focus clauses contain no presupposed information, and thus no topical constituent (Lambrecht 1994, 233). Consequently, all of the constituents are focal (Lambrecht 2000, 635). In languages such as Italian, therefore, the VS order has the pragmatic function of marking the subject as a non-topic by removing it from the clause-initial position, and perhaps even coding it as a focus by placing it postverbally. Such postverbal subjects are therefore also more ‘object-like’ in their position.25 Cross- linguistically, there are various strategies for marking a subject as a non-topic; the inversion of word order is only one (Lambrecht 2000). 25 Cf. Sasse (1987, 516-520) for a discussion on discourse newness in thetic constructions. SYNTAX AND MORPHOSYNTAX 405 The concept of sentence focus can be traced back to Sasse’s notion of thetic clauses (Sasse 1987). As shown in §7.3, Sasse defines the function of thetic clauses as presenting a situation as an undivided whole, fusing ‘an element denoting an individual and an element denoting an event’ (Sasse 1987, 522). This, however, is a definition of a discourse-pragmatic effect, rather than of identifiable syntactic and information-structural properties, and is consequently hard to apply cross- linguistically.26 Therefore, Lambrecht’s model is favoured here, as it enables the researcher to apply measurable criteria such as discourse-newness. As regards their semantics and pragmatics, sentence-focal clauses tend to be ‘simple’. Lambrecht (2000, 617) states that sentence-focal clauses must be intransitive, that is, possessing only one (canonical) argument.27 Lambrecht (2000, 617) also adds that sentence-focal clauses tend to have a non-agentive subject. The likely explanation for this (though this is not explicit in Lambrecht) is that an agentive (and thus prototypically human) subject is more likely to be topical, and thus less likely to feature in a sentence-focal construction. This generalisation notwithstanding, agentive Subjects are in fact attested within structures labelled as ‘sentence-focal’ in several languages, including those studied by Sasse (1987, 529) and in J. Dohok. Indeed, even transitive clauses are attested 26 Sasse’s study begins with English, where theticity is conveyed by placing the nucleus accent on the subject. 27 See §9.8.1 below on Goals in intransitive clauses. 406 WORD ORDER in sentence-focal constructions in some languages, though this is apparently rare. The reason for this rarity is likely the semantic-pragmatic complexity of transitive clauses. Since transitive clauses—in contrast to intransitive ones—have at least two arguments, the requirement of all information being new is ‘more difficult’ to meet. Thus, (at least) one of these arguments is likely to be topical. In J. Dohok, even if both arguments of a transitive clause are discourse-new, the preferable strategy of packaging such information is through the topicalisation of the subject (see §9.5.2 above), rather than through shifting the A to the postverbal position. In other words, the communicative content of such clauses is ‘broken apart’, rather than presented as a single, all-new pragmatic unit. The likely motivation for this is the easing of the clause’s cognitive processing, considering the semantic complexity of transitive clauses relatively to intransitive ones. In his grammars of C. Barwar and C. Urmi (both VO dialects), Khan (2008b and 2016 vol.2) cites a few examples in which a subject is placed postverbally to argue that such order is used to introduce new, ‘prominent’ subjects to the discourse (Khan 2008b, 858). However, it is not always completely clear why the subjects in the sentences quoted are ‘prominent’. The present study, therefore, aims to identify the specific environments in which the verb-subject order occurs, showing that in J. Dohok, this order has other functions beyond the introduction of ‘discourse-prominent’ topics. At the end of this section, it is suggested that the J. Dohok tends towards an ergative alignment in word order. Word order alignment has—to the author’s knowledge—not been studied for NENA. SYNTAX AND MORPHOSYNTAX 407 9.7.2. The presence of Goals The strongest factor influencing the position of the Subject is the presence of a Goal (i.e. goals of movement verbs). In clauses with nominal and full pronominal Subjects, the VS order is highly infrequent if a Goal is present, as shown in Table 53: Table 53: Position of V and S in J. Dohok intransitive clauses SVG 30 VSG 4 VGS 1 % clauses with VS which include a G 13% Total clauses with S and G 39 Likely, the presence of Goals is not only a formal-syntactic constraint, but also a semantic-pragmatic one. Goals of intransitive movement are arguably more argument-like than other obliques, being semantically more central to the predication than, for instance, locatives (cf. §9.8.2 below). The implication of treating them as argument would be that clauses with both S and G have in fact two arguments. Such clauses are thus semantically-pragmatically more ‘complex’ than clauses with the S only, and parallel the semantically ‘complex’ transitive clauses. Being complex, therefore, like transitive clauses, they are not ‘good candidates’ for sentence-focal constructions. 408 WORD ORDER Clauses with Gs are excluded from further discussion to ensure that their presence or absence does not obscure the effect of other variables. 9.7.3. The role of animacy and discourse status Two other factors were found relevant for word order of intransitive clauses: animacy and discourse status. MCA of these two variables along with the word order variable (VS/SV) revealed that the variable more closely associated with word order is discourse status (they appear closely together), rather than animacy: Figure 6: Association between discourse status (‘d-new’ and ‘d-old’), animacy (‘animate’ and ‘inanimate’) and word order (‘VS’ and ‘SV’) in intransitive clauses with a full nominal Subject in J. Dohok (MCA) Variable categories - MCA 1.0 inanimate 0.5 SV contrib d-old 14 16 18 20 Dim2 (27.6%) Dim2 0.0 22 d-new VS -0.5 animate -0.5 0.0 0.5 Dim1 (55.9%) SYNTAX AND MORPHOSYNTAX 409 While an association between the VS word order and non-animacy does exist, it is weak and statistically not highly significant (cf. Table 54 below). Thus, the VS order is also well attested with animate referents, as illustrated by example (26) below, while (27) has an inanimate referent. Table 54: The association between word order and animacy in J. Dohok (contingency table) VS SV Total Animate 26 (32) 34 (28) 60 Inanimate 28 (23) 15 (21) 43 Total 54 49 103 (p<0.05, Φ=-0.22 > weak relationship) (26) 3b/57 k-eθe-Ø-wa HzmanH mux băθər sùkka…ˈ REAL-come-he-PST time like after Sukkoth… ‘(Then) a time would come, somewhere after Sukkoth…’ (27) 3b/33 …k-əθy-a-wa yəmm-i g-māqəlw-à-wa-lu.ˈ REAL-come-she-PST mother-my REAL-clean-A.she-PST-P.them ‘…my mother would come and clean them. The association of the VS order with inanimate referents can be explained with recourse to sentence focus. Inanimate subjects, in contrast to animate ones, are unlikely to function as discourse topics, and are thus more likely to occur only 410 WORD ORDER once and be discourse-new. Animate subjects, by contrast, are more likely to serve as discourse topics and thus to be reoccur several times within the discourse. Thus, it is inanimate subjects which are more likely to feature postverbally, assuming that this position is indeed associated with non-topical subjects. 9.7.3.1. Discourse status An explanation which is stronger (moderate strength) than animacy is found in the discourse status of the Subject: the VS order is more characteristic of discourse- new Subjects : Table 55: The association between word order and discourse status in J. Dohok (contingency table) VS SV Total d-old 16 (29) 39 (26) 55 d-new 38 (25) 10 (23) 48 Total 54 49 103 (p<0.0001, Φ=-0.5 > moderate relationship) These findings corroborate that the idea that VS is a tool for conveying sentence focus, though the correlation is far from complete. In contrast to some other SYNTAX AND MORPHOSYNTAX 411 languages, stress placement does not appear to be key to sentence focus:28 in both VS and SV clauses, nucleus stress tends to be clause-final. This does not apply to clauses with narrow focus on the S, in which the subject is stressed, regardless of its position.29 A likely reason why discourse-old subjects are placed postverbally, at least in some cases, is apparently because they are added as an afterthought for clarification, especially in cases of topic shift.30 This is apparent in example (28) below: the nucleus stress falls on the verb, suggesting tha the Subject was indeed added later, as an afterthought. In the previous sentence, another character is speaking and the topic subsequently shifts back to ‘ruler’: (28) 1b/80 munə̀x-le ḥakoma,ˈ g-emər-Ø… PERF.sigh-S.he ruler REAL-say-A.he ‘The ruler sighed and said (to the man)…’ 28 For instance, in languages with a rigid word order such as English, sentence focus is created through stress: in sentence-focus clauses, stress falls on the subject (cf. Lambrecht 2000, 617). 29 Word order of narrow focus on the S has not been studied systematically, though an impressionistic study suggests that both VS and SV order is possible. 30 This function is also mentioned by Khan for C. Urmi (2016, 323) 412 WORD ORDER 9.7.3.2. Discourse status in inanimate subjects This association between word order and discourse status is even stronger in the case of inanimate subjects. In other words, the position of inanimate Subjects is especially likely to be determined based on their discourse status: they are preverbal if discourse-old and postverbal if discourse-new. Table 56: The association between word order and discourse status of inanimate Subjects in J. Dohok VS SV Total old 5 (10) 11 (6) 16 new 23 (18) 4 (9) 27 Total 28 15 43 (p<0.001, Φ=-0.55 > strong relationship) Sentence (29) below illustrates the use of VS to express sentence focus: (29) 3f/6 xă yoma ʾo naša mpəq-la guləkta go kàs-e,ˈ gò kas-e.ˈ one day thisM man PERF.go_out-S.she ball in belly-his in belly- his ‘One day this man – a swelling31 appeared on his belly, inside his belly.’ 31 The italics indicates nucleus stress and thus also sentence focus. SYNTAX AND MORPHOSYNTAX 413 This notwithstanding, as Tables 55 and 56 above demonstrated, the distribution of sentence focus and non-sentence focus along the lines of word order is not complete. Indeed, as observed previously, the VS order is characteristic of inanimate subjects in general. Thus, the VS order is also attested with discourse- old subjects, as illustrated by (30) below. This sentence comes at the end of the description of Passover celebrations, meaning that the referent is discourse-old: (30) 3b/8 k-xalṣa-wa pèsaḥˈ g-oð-i-wa xa-ga xə̀t.ˈ REAL-finish-S.she-PST Passover, REAL-do-they-PST once-again. ‘When) Passover finished, they would start (lit. do) once again.’ Conversely, there are cases of a new referent placed preverbally: (31) 3f/43 Context: ‘The king says “where is that Jew?” His servants replied “look, we sent him home.”’ Wuḷḷa ləbb-e bqèʾĕ-le.ˈ g-əmr-i b-qāṭə̀ḷ-le.ˈ Look heart-his PERF.broke-S.it. REAL-say-A.they FUT-kill-A.he-P.him. ‘Look, he was dying from fear (lit. his heart broke). They said (that) he was going to kill him.’ In J. Dohok, therefore, the function of the VS order is not restricted to introducing new, discourse-prominent subjects (cf. C. Barwar - Khan 2008b, 858). As demonstrated, this order is very closely associated with discourse-new inanimate subjects which are not necessarily resumed in the subsequent discousre (an example of an unresumed Subject is ‘his heart’ in (31) above). 414 WORD ORDER 9.7.3.3. Discourse status in animate Subjects In the case of animate subjects, the correlation between word order and discourse status is weaker (moderate correlation). The VS order is furthermore attested for both agentive and experiencer Subjects (e.g., there are many examples with the verb Ø-θ-y (ʾaθe) ‘come’. Table 57: The association between word order and discourse status of animate Subjects in J. Dohok (contingency table) VS SV Total old 11 (17) 28 (22) 39 new 15 (9) 6 (12) 21 Total 26 34 60 (p<0.001, Φ=-0.42 > moderate relationship) As shown in Table 57, the relationship between discourse newness and word order is far from exclusive in animate subjects. This means that, for instance, discourse- old subjects (including definite ones) also occur postverbally, as illustrated in (32) below. It is not instantly apparent what other pragmatic function the VS order in animate Subjects could correlate with; for instance, the intransitive sentence in (32) does not have narrow focus (either on the Subject or the predicate). The subject of this sentence is different from that of the preceding clause, so the function of the VS order in this case could be the reactivation of a discourse topic. SYNTAX AND MORPHOSYNTAX 415 The function of discourse topic reactivation has not been investigated systematically, however. (32) 1c/72 durdət yom q-qayəm-Ø măr-e bèθa,ˈ g-ezəl-Ø ta ṣìwe.ˈ next day REAL-get_up-he owner-GEN house, REAL-go-he for wood ‘The next day the father of the house gets up and goes to (get) wood.’ In general, therefore, the VS order is associated with discourse-new subjects, but this association is not exclusive. First, the VS order may perhaps have other pragmatic functions, or could even be pragmatically unmarked in some cases.32 Moreover, the VS order is more likely to occur when the discourse-new subject in question is inanimate. In other words, the VS order is more characteristic of inanimate discourse-new subjects than of animate discourse-new ones. Conversely, discourse-new Subjects can also feature in the preverbal position. These findings can in turn be used to argue that clauses with inanimate discourse- new subjects are most prototypical type of sentence-focal constructions, and as such are most likely to employ the VS order. This is comparable with Lambrecht’s assumption that sentence focus is associated with non-agentive subjects (cf. 32 The extent to which word order correlates with sentence focus in AVO languages differs. Lambrecht, for instance, notes that such correlation is weaker in Latin than it is in Italian (Lambrecht 2000, 673). J. Dohok apparently belongs to those languages where the correlation is relatively loose. 416 WORD ORDER Lambrecht 2000, 617), though Lambrecht assumes that sentence focus should be restricted to such subjects. Arguably, there is also another difference between inanimate and animate subjects. Namely, it is animate subjects which are prototypical discourse topics. Thus, a more nuanced functional definition of the VS order with animate subjects is that it serves to introduce new discourse topics (which are therefore focal in the first instance they occur). This is illustrated well in the sentences in (33), in which the postverbal subjects doxtor and năsax introduce new characters within a story: (33) 3g/2 Context: ‘My husband was in the hospital, the Mount Scopus hospital.’ b-lèleˈ θe-le dòxtor,ˈ θe-le xă nằsaxˈ mən HlevanònH.ˈ in-night PERF.come-he doctor, PERF.come-he one sick from Lebanon. At night, a doctor came, (and with him) came a patient from Lebanon.’ 9.7.4. Clauses with p-y-š (payəš) ‘break out, occur’: sentence focus and non-canonical subjects The previous section concluded that the correlation between VS order and sentence focus in J. Dohok is not perfect. An exception to this are clauses with p- y-š in the meaning ‘break out, occur’ (illustrated in sentence (34) below), in which a complete association is attested. SYNTAX AND MORPHOSYNTAX 417 (34) 4b/17 qam-ʾoð -i -la ʾèða,ˈ hatxa pəš-le mən-nu bùrhun.ˈ PERF-do -A.they -T.it R.feast so PERF.occur-S.he with-them miracle.ˈ ‘They made it into a feast, (because) a miracle happened among them in this way.’ The verb p-y-š has several lexical meanings as well as the grammatical meaning of inchoativity, expressed with an inflected form of p-y-š + in-Infinitive (bə-šqala). The meaning relevant in this context is that of conveying the appearance of a phenomenon. Such constructions are thus a type of existential clauses, mostly with inchoative nuance. Crucially, however, the subject of a p-y-š is always an inanimate, often abstract referent (e.g., an atmospheric phenomena or a (historical) event). In other words, p-y-š is not used in the meaning ‘be born’. Prototypical existential clauses, on the other hand, can have an animate or an inanimate argument. According to Lambrecht, existential clauses are a salient sub- category of sentence-focal clauses (Lambrecht 2000, 617). This also applies to p- y-š clauses in J. Dohok: since they predicate about the beginning of a phenomenon, such a phenomenon is assumed to have had no prior existence, and should thus be discourse-new when it occurs in a clause with p-y-š. In the J. Dohok corpus, the Subject of p-y-š ‘break out, occur’ is always postverbal. Thus, such clauses exhibit a complete match between the pragmatic category of sentence focus and the VS order. This order occurs even when a given referent reoccurs several times in the discourse (with the predicate p-y-š), and is thus 418 WORD ORDER already known to the listener (though, in principle, the clause still predicates about the phenomenon’s beginning): (35) 3d/9-11 pəš-le farhūd,̀ ˈ qṭə̀ḷ-lu naše…ˈ PERF.occur-he Farhud, PERF.kill-A.they P.men ‘The Farhud broke out, they killed people.’ qamaye Hmutḥəl-luH… mxè-lu naš-e,ˈ gnù-lu.ˈ first PERF.begin-they… PERF.beat-they P.men PERF.steal-A.they ‘First (i.e. before the Farhud broke out), they began… they beat people, robbed (them).’ xaràye,ˈ pəš-le farhūd̀ ˈ yom-ət ʾe-ðət zyàra,ˈ b-lèle.ˈ Eventually, PERF.occur-he Farhud, day-GEN feast-GEN Pentecost in- night. ‘Eventually, the Farhud broke out, on the eve of the Feast of Pentecost.’ In clauses with p-y-š, therefore, the S aligns with the P of transitive clauses in word order (i.e. both are postverbal). Indeed, the Subject of p-y-š clauses is highly ‘object-like’ in its pragmatic and semantic properties. Both arguments are prototypically focal; both are also non-agentive and highly affected: the subject of a p-y-š is an inanimate entity brought into existence by an external factor. This means that such subjects are low on Givón’s hierarchy of semantic ‘subject- likeness’: agent > recipient/benefactive > experiencer > patient (=affectee) (Givón 1984, 151; terminology adapted) SYNTAX AND MORPHOSYNTAX 419 The subject of a p-y-š clause, therefore, is only the grammatical subject insofar as it controls agreement. Such ‘non-subject-like subjects’ contrast with arguments which do have semantic subject-like properties and which have undergone an opposite process: a promotion to the subject role. In the history of Aramaic/NENA, such a process took place with predicative possessors and experiencers.33 Possessors are semantically ‘subject-like’ by virtue of being recipient-like (cf. Givón’s hierarchy above). Possessors in NENA trigger agreement on the verb – they are cross-referenced by an L-suffix on the existential particle. Moreover, when independent, predicative possessors are zero-marked. Both of these are clear subject features (Noorlander 2021, 52). Moreover, possessors, like subjects, are generally preverbal in NENA. The latter appears to be in fact a stable—rather than an acquired—feature of many Aramaic varieties, being attested, for instance, in Syriac. In sum, in the history of Aramaic, possessors have acquired the morphosyntactic features of subjects, and appear to have (always) occupied the subject position. By contrast, the subject of p-y-š clauses lack the semantic and syntactic (word order) properties of subjects in dialects such as J. Dohok. 33 For a historical typology of non-canonical subjects and their acquisition of subject-like properties, see Seržant (2013). 420 WORD ORDER 9.7.5. Existential clauses and their parallels with verbal intransitive clauses In NENA in general, existential clauses consist of a form of the existential particle (or, in the case of irrealis or habitual, an inflected form of the verb h/Ø-w-y (hawe) ‘be’) and of the argument whose existence is being predicated.34 The present tense affirmative forms of the existential particle in J. Dohok and J. Sanandaj are, respectively, ʾiθ and hit. The function of existential constructions is to introduce a new referent to the discourse (Dryer 2007, 241). It is for this reason that cross- linguistically, the argument of existential clauses is usually indefinite (Abbot 1997). Since the argument is discourse-new, moreover, the entire clause is in focus. Existential clauses in LD exhibit one apparent difference when compared to the TZ varieties: in LD, the predicate-argument order predominates, while in TZ, the default order is argument-predicate. In J. Dohok, the order argument-predicate is only found when the predicate is in narrow focus (in 2/51 cases). Thus, in the example in (36), the entity ‘reading’ arguably evokes also the concept ‘writing’ in the listener’s mind. Thus, ‘reading’ is 34 Existential clauses are not considered verbal clauses in this thesis. Their argument, moreover, does not display subject-like properties. Still, there are parallels between existential and intransitive clauses, which merits their mention in the present context. SYNTAX AND MORPHOSYNTAX 421 likely presupposed at the point when its existence (in the school education) is negated: (36) 1e/43 maʿalmin-ət huðaye qam-mālə́p-Ø-wa-lu qràya,ˈ kθawa lə̀-θ-wa,ˈ bas qràya…ˈ Teacher-GEN Jews PERF-teach-A.he-PST-R.them reading, writing NEG-EXIST-PST only reading ‘A Jewish teacher would teach them to read, there wasn’t writing, just reading…’ In J. Sanandaj, the predicate-order is much more common. According to Khan, nucleus stress typically falls on the argument (Khan 2009, 342–3), though in negative clauses, the existential particle tends to be stressed. In J. Dohok, therefore, the position of the argument in an existential clause overlaps with that of the P in a transitive clause and of the S in sentence-focal constructions. This overlap is apparently due to the prototypical focal role of the argument of existential clauses. For TZ, however, it is not possible to make such alignment generalisations regarding existential clauses. This is because in J. Sanandaj, both S/A and O are preverbal. What still holds true, however, is that the shift of the argument of the existential clause to the preverbal position in TZ took place in parallel with the shifting of the O to that same position, both apparently under the influence of Kurdish (for existential clauses in Northern Kurdish, cf. Thackston 2006, 31). VS order in J. Sanandaj 422 WORD ORDER As stated above, the VS order is extremely rare in J. Sanandaj (fewer than ten cases). It occurs with animate discourse-old Subjects, as well as with discourse-old inanimate ones. This order is also attested with the verb x-Ø-r (xar), which largely corresponds semantically to the verb p-y-š in J. Dohok and with other verbs of similar meaning (‘appear’). In J. Sanandaj, x-Ø-r is used both inchoatively (‘appear’) or non-inchoatively (‘be’). (37) J. Sanandaj (Khan 2009, 46, B:72) xir-Ø xà yarxa…ˈ appear-he one month ‘One month passed…’ Still, even with x-Ø-r, the SV order is more frequent in the J. Sanandaj corpus: (38) J. Sanandaj (Khan 2009, 450, B:40) yoma xar-ø-wà-uˈ kŭl-e g-ez-i-wa baqa kništa. day appear-he-PST-and all-them REAL-go-S.they-PST to synagogue ‘The dawn came and they would all go to the synagogue.’ Postverbal subject placement is not characteristic of A(/S)OV languages such as J. Sanandaj (cf. Lambrecht 2000), or at least is not associated with the sentence-focal function. The reason for this is apparent: given that objects in these languages are preverbal, placing a subject postverbal does not help make it more ‘object- and focus-like’, and thus mark the whole clause as in focus. Word order ergativity SYNTAX AND MORPHOSYNTAX 423 In sum, the VS order in J. Dohok largely corresponds to the pragmatic category of sentence focus, and can therefore be considered a marked structure for conveying such focus. Inanimate discourse-new Subjects are most frequently attested in the postverbal position, which fits with the idea that such Subjects are ‘best candidates’ for sentence focus, since they are unlikely to be topical. However, the VS order is also well attested with focal animate referents, and is in fact not restricted to sentence focus, since it also occurs with discourse-old Subjects. This has implication for word order alignment in J. Dohok. This dialect exhibits a tendency towards an ergative alignment for all Subjects: S and P tend to be postverbal, while A is overwhelmingly preverbal. Once again, this tendency is especially strong for inanimate discourse-new subjects. Like objects, such subjects are prototypically focal and non-agentive. These syntactic tendencies may be quantified using a mathematical formula which was created by this thesis’ author, drawing from a formula presented by Haig and Schnell (2016, 601). This formula shall be called here ‘word order ergativity index’. Haig and Schnell’s formula calculates lexical discourse ergativity (i.e. the extent to which S shares with P its tendency to be expressed nominally, rather than pronominally). This is done using percentages for the frequency with which a given argument is expressed nominally (as opposed to having no independent argument or a pronoun) and subtracting the difference in these percentages in the following way: 424 WORD ORDER Lexical discourse ergativity index = (P – S) – (S – A) A negative value indicates that the ‘distance’ (expressed in percentages) between P and S is smaller than between S and A, and consequently, that there is a tendency towards an ergative alignment. A positive value means a tendency towards an accusative alignment. It is proposed here that this formula can be modified in order to calculate ergativity in word order. Such a formula uses the frequencies of postverbal occurrences for a given argument, calculated as a percentage of the total count of all full nominal arguments. The frequencies of postverbal occurrence per argument in the J. Dohok corpus are given below: Patients 89% Subjects (Ss), including clauses with Goals 51% Subjects (Ss), clauses without Goals 53% Agents 13% Using these numbers, the word order ergativity index is calculated. Since this formula is concerned with the overall frequencies, all clauses with the Subject— including those with a Goal—are included. Word order ergativity index = (P – S) – (S – A) = (89 – 51) – (51 – 13) = –2 The result is negative, indicating that in the J. Dohok corpus, word order exhibits a tendency towards an ergative alignment. SYNTAX AND MORPHOSYNTAX 425 9.8. Goals and obliques Oblique constituents have previously were previously as prepositionally expressed arguments other than Goals. Oblique arguments can be further subdivided into: Locations. Some obliques are more argument-like, being semantically more closely bound with the predicate. For instance, goals of movement verbs are the ‘aim’ or ‘purpose’ of an activity (e.g., ‘I went to the synagogue’), while locatives are simply adjuncts (e.g. locatives, ‘I’m singing in the synagogue’). Goals subsume here both destinations of movement verbs and Recipients of ditransitive. In J. Dohok, both Goals and obliques tend to be postverbal. In J. Sanandaj, by contrast, Goals are generally postverbal, while obliques preverbal. 9.8.1. Goals In both dialects, the majority of Goals is postverbal, but tendency is much stronger in J. Dohok, as shown in Table 58 below. This difference in the positional flexibility of Goals is statistically significant, suggesting a slightly different Goal typology in J. Dohok and J. Sanandaj, despite ther shared VG tendency. Table 58: Position of Goals in J. Dohok and J. Sanandaj Postverbal Preverbal Total Frequency of preverbal (rows) Gs 426 WORD ORDER J. Sanandaj 450 (463) 43 (28) 493 9% J. Dohok 299 (283) 2 (18) 301 0.6% Total 749 45 794 (columns) [p<0.0001, Φ=0.17] In J. Dohok, a Goal is sometimes fronted when in narrow focus, as illustrated in (39) below. This behaviour is parallel to that of objects in this dialect, which also occur preverbally when in narrow focus, but apparently more regularly. (39) 1d/21 Context: ‘The Turks had passed a decree (against) Christians.’ ta huðàye ži ʾud-lu farman.ˈ for Jews also PERF.do-they decree. Also against the Jews they passed a decree. The VG order is commonplace across NENA—occurring in both VO and OV dialects—and is also used in Bahdini and Sorani, that is, in Iranian varieties within the NENA territory.35 In Kurdish, the frequency of preverbal Goals differs per 35 In the Iranian literature, the categories corresponding to ‘oblique’ here are ‘Place’ (Haig 2019) and ‘Locative’ (Stilo 2018), Instruments (‘Instrumental’ in Stilo 2018), Sources of movement verbs (‘Ablative’ in Stilo 2018) and other non-local obliques (‘Oblique’ in Haig 2019 and ‘comitative’ in Stilo 2018). SYNTAX AND MORPHOSYNTAX 427 dialect, but is generally more frequent than in J. Dohok (and likely other NENA dialects). The VG order in Kurdish and in the NENA dialects such as J. Sanandaj is typologically unusual, given the OV profile of these languages (Haig 2015). According to Haig, the VG order in Kurdish may reflect ‘the ancient imprint of Aramaic’ (Haig 2015). Still, this word order convergence is apparently not complete, since Kurdish shows a weaker tendency for postverbal Goals than NENA dialects such as J. Dohok. Once again, J. Sanandaj shows a closer affinity with the Iranian word order: both exhibit a weaker tendency for Goals to be postverbal. In the J. Dohok corpus, there are three clauses in which both Source and Goal are expressed and in all three, the Source precedes the Goal. This linear order is best understood as linguistic iconicity: the temporal sequence of the movement ‘from…to’ is mapped onto the temporal sequence of the utterance, so that endpoints occur after start-points (cf. Haig 2018, 36; Givón 1985). This is illustrated in sentence (40) below with the order source-V-G: (40) 3d/73 ʾana mən daxxa b-yasq-ən l-gàreˈ… I from here FUT-climb-IM to-roof ‘I will climb up from here on the roof (i.e. from my house)…’ In the J. Dohok corpus, destinations of movement verbs which include a nominal are always prepositionally marked. In J. Sanandaj, on the other hand, such prepositional marking is occasionally omitted. This is illustrated by the sentence in (41), where baqa ‘to’ is expected before kništà: 428 WORD ORDER (41) J. Sanandaj (Khan 2009, 646, B:74) g-ez-ex-wa Ø kništà… REAL-go-we-PST Ø synagogue ‘We used to go (to) the synagogue.’ This omission does not occur for any other type of Goals (i.e. Recipient) or for oblique constituents. Such zero-marked Goals morphosyntactically resemble (indefinite) objects. 9.8.2. Other obliques In J. Dohok, almost all non-Goal obliques are postverbal, except for 24/166 cases (15%), all of which are locatives (cf. (42) below). The tendency for preverbal placement appears to be especially strong for adverbial locatives. Such oblique locatives are adjuncts rather than arguments. (42) 1d/2 go turkya ʾud-lu farman…ˈ In Turkey PERF.do-A.they P.decree ‘In Turkey, they passed a decree…’ A similar tendency for non-Goal obliques to be postverbal has been found for C. Barwar, another VO NENA dialect, where 20% of non-Goal obliques are preverbal according to Stilo (2018, 6). In J. Sanandaj, obliques—in contrast to Goals—prefer the preverbal position, but the position of obliques in J. Sanandaj is less rigid than in J. Dohok. In J. Sanandaj, 66% of obliques are preverbal, compared with 85% of postverbal obliques in J. SYNTAX AND MORPHOSYNTAX 429 Dohok.36 Stilo (2018) attributes the tendency for obliques to be preverbal in the TZ dialects to contact with West Iranian languages, including Kurdish, which exhibit a general preference for obliques to be preverbal. 9.9. Conclusion This chapter has studied word order in J. Dohok, contrasting and comparing it with that of J. Sanandaj. The general word order typology of these dialects is respectively A/SVO and A/SOV. In clauses possessing a full subject and at least one other full argument, word order is rigid in both dialects, being especially fixed in J. Dohok. Thus, transitive clauses have the AVO order in J. Dohok in 98.5% of cases, while the AOV order is found for 92% of transitive clauses in J. Sanandaj. In general, however, word order is relatively flexible in both dialects, being to a great extent pragmatically determined. Thus, in both dialects, the clause-initial (or at least preverbal) position tends to be reserved for (zero-marked) topics, while the immediately preverbal position is associated with arguments in narrow focus (though in J. Dohok, this is clearly observable only with objects). In J. Dohok, moreover, the postverbal position in broad focus clauses is strongly associated with focal arguments (typically objects). 36 For more details on obliques in J. Sanandaj, see Noorlander and Molin (forthcoming 2022). 430 WORD ORDER This pragmatic component of word order is especially apparent in ditransitive clauses: different types of ditransitives have different default (i.e. frequent) word order patterns, depending on the prototypical functions of the arguments involved. In J. Dohok, these general pragmatic preferences also result in the relative flexibility of objects and Subjects. In this dialect, 11% of objects are preverbal, and the majority of these are definite topical constituents (which trigger verbal object agreement). This contrasts with only 4.4% of postverbal objects in J. Sanandaj. Such greater ‘flexibility’ of objects in J. Dohok is due to the fact that in this dialect, marked pragmatic functions of the object always correlate with a changed position. Specifically, narrow focus and topical (including topicalised) objects are preverbal, while new or reactivated topics appear clause-initially. In J. Sanandaj, by contrast, the default preverbal position is used in both broad focus clauses as well as in clauses with narrow focus on the object, meaning that the OV order accounts for both pragmataic patterns. Thus, both dialects are characterised by the preverbal placements of objects in narrow focus. In J. Dohok, moreover, the OV order correlates with definiteness in a statistically significant way. This observation is likely relevat for the shift from VO to OV which has taken place in the TZ NENA varieties, and perhaps for comparable shifts cross-linguistically. Likely, the preverbal position first became the norm for narrow-focal and/or topical objects, gradually extending to other types of objects, perhaps along semantic lines such as definiteness or animacy. SYNTAX AND MORPHOSYNTAX 431 Subjects of intransitive verbs are also more flexible in J. Dohok than in J. Sanandaj. In 49% of cases, nominal Subjects are postverbal, meaning that J. Dohok tends towards an ergative alignment in word order (for nominal arguments). Placing a Subject postverbally makes it object/focus like, and therefore creates sentence focus. However, this association between word order and discourse-new Subjects is not complete – the VS order is also attested with discourse-old Subjects. Indeed, this word order also sometimes correlates with clarificatory subjects (added as an ‘afterthought’ to clarify the topic, e.g. in cases of topic shift). In J. Sanandaj, the VS order is extremely infrequent. This is unsurprising, since in an OV dialect such as J. Sanandaj, placing a Subject postverbally would not have the same effect of making the referent more object-/focus-like. Both J. Sanandaj and J. Dohok share the areal VG preference. This tendency, however, is stronger for J. Dohok in a statistically significant way, indicating a greater resistance to the influence of Kurdish. Non-Goal obliques tend to be preverbal in J. Sanandaj and postverbal in J. Dohok. However, this word order is once again more rigid in J. Dohok than in J. Sanandaj. CHAPTER TEN CONCLUSION This thesis has described the fundamental linguistic features of the North-Eastern Neo- Aramaic dialect of the Jews of Dohok, north-western Iraq, and located this variety on the linguistic ‘map’ of NENA by bringing to bear comparative dialectological data. It has also addressed several major issues in NENA linguistics pertinent to cross-linguistic debates on diachronic development and typology. The following is a summary of the main topics in the previous discussion. In the phonemic study (Chapter 1), I pointed out the disparity between /a/ and /ă/, versus other vowel phonemes. I argued for a lag in the loss of phonemic distinction between long and short a relatively to other vowels, likely due to the distinct phonetic properties of a. Overall, however, I argued for a minimalist system of vowels, in contrast to the grammars of other Lishana Deni dialects. I showed that the distribution of vowel quantity can be satisfactorily explained, but requires more than a simple dichotomy between ‘phonetically predictable’ vs ‘phonemic’, still encountered in the phonetics and phonology literature. Derived contrast, diachronic change in progress and phonotactic rules must be taken into account. Chapter 2 concentrated on the phonetics and phonology of units larger than segments. For example, I surveyed the phonetic properties of emphasis and its spread, showing that the manifestations of emphasis in sonorants are stronger than those in obstruents. I also suggested that even though emphasis spread continues 434 CONCLUSION to be a linear phenomenon in NENA dialects of Iraq such as J. Dohok, synchronically, it is often anchored in several phonemes susceptible to it within a single word. This leads effectively to the anchoring of emphasis in a unit larger than a single segment, such as a syllable. I suggested that environments with emphasis throughout would have facilitated the reassignment of emphasis from phonemes to entire phonological words, which occurred in C. and J. Urmi. Chapters 3 and 4 provided overviews of pronominal and nominal morphology respectively. Chapter 5 focuses on verbal derivation and inflection, and especially on ‘weak’ and ‘irregular’ verbal patterns. I demonstrated that several stems are synchronically not directly derivable from a root. In other cases, the various inflections of a given verbal lexeme cannot be derived from a single root. On this basis, I proposed that instead of the traditional Semitic ‘root-and-pattern’ derivation, we should primarily speak of a ‘pattern’ derivation, while also allowing for a stem-based derivation in some cases. Yet although roots are secondary to NENA derivation, patterns remain central. Throughout the long cycles of phonetic and analogical change (‘reanalysis’) that the verbal morphology of NENA has undergone, there has been a clear pressure for new forms to adhere to existing verbal patterns. Such a pressure reflects a resistance to the interference of contact languages such as Kurdish, whose derivation is concatenative, that is, not based on abstract patterns. CONCLUSION 435 The study of the expressions of pronominal objects in Chapter 6 showed that the Lishana Deni group including J. Dohok is distinguished from its co-territorial Christian dialects such as Barwar in allowing the so-called stacking of L-suffixes. Thus, in LD, a Recipient of a ditransitive can be expressed with L-suffixes alongside a Theme on the šaqəl-based stems (except the past perfective qam-šaqəl-le): V –E-suffix –L-suffix –L-suffix V –A –T –R g-yāwə́l –Ø –le –li give he it me ‘He gives it to me.’ However, the expression of an R with an L-suffix is blocked if the T has a 1st/2nd person referent (and is thus also expressed with an L-suffix). When the stacking occurs on šqəlle, the first L-suffix expresses the Agent and the second one can only express a Recipient, but not a Theme or a Patient: V –L-suffix –L-suffix V –A –R húl –le –li gave he me ‘He gave it to me.’ In the perfective past in LD including J. Dohok, there are two strategies available for all pronominal objects (Ps, Ts and Rs). One is with L-suffixes on qam-šaqəl-la, the other with E-suffixes on šqəl-, i.e. šqil-a-le. J. Challa only possesses the latter strategy, while in J. Dohok, the latter is extremely infrequent. In dialects such as 436 CONCLUSION J. Betanure and J. Amediya, both strategies are available. I conducted a survey identifying basic functional differences between these two competing past perfective transitive constructions in LD, and in C. Barwar. In dialects where both strategies are available, šqil-a-le is strongly preferred for reflexive objects. In all the dialects surveyed, šqil-a-le, and specifically the fossilised 3FS E-suffix -a, has also developed into a lexicalised middle voice. This is also the case in J. Dohok, where šqil-a-le is used almost exclusively for middle voice, which in turn obscures its historical connection with qam-šaqəl-la (as an alternative transitive construction). Chapter 7 offered a detailed description of the morphology and syntax of the various copulas in J. Dohok. This description was supplemented by a general discussion on the history, syntax, semantics and pragmatics of the copulas. A particularly interesting issue in NENA is the (enclitic) 3rd person basic copula, e.g. 3MS =le, termed here ‘L-copula’. The use of this copula leads to an object-like encoding of the subject of copula clauses in the present tense realis domain; in J. Dohok, for instance, a 3FS copula subject and object are both expressed with the morpheme la on the predicate. Moreover, a feature of J. Dohok which distinguishes it from other LD dialects is that this L-copula occurs virtually only in intransitive clauses. This, in turn, results in a morphological alignment of the S(ubject) of intransitive copula clauses with the P(atient of monotransitive) in the 3rd persons, which is characteristic of ergative systems. The implications of NENA CONCLUSION 437 copula syntax for alignment typology have—to my knowledge—not been discussed yet in the literature. Chapter 8 studied verbal semantics and revealed some differences between J. Dohok and other NENA varieties, including the closely related LD group. First, in J. Dohok, the continuous and resultative/perfect constructions have undergone only partial grammaticalisation relatively to other dialects. The continuous J. Dohok construction, wele bə-šqala (eventive copula + in-Infinitive), is only attested a handful of times, and exclusively with stative (atelic) verbs. This contrasts sharply with dialects such as J. Betanure, J. Zakho or J. Amediya, where the corresponding construction is primarily ‘progressive’, that is, a dynamic continuous. In the early NENA sources, the parallel construction—which sometimes lacks the copula—is associated with atelic verbs. This distribution is compared with the alternative NENA continuous/progressive form, which comprises of a demonstrative/deictic element and a present form – in J. Dohok, wəl k-šaqəl. This situation in J. Dohok and in the early NENA texts suggests the existence of different grammaticalisation pathways for progressive/continuous constructions cross-linguistically, and cautions before assuming an original dynamic function of such constructions. The other J. Dohok construction of limited functional scope alluded to in the previous paragraph is wele šqila (eventive copula + resultative Participle), which remains restricted to the experiential perfect and the resultative. In other dialects 438 CONCLUSION (e.g. J. Betanure, J. Zakho, C. Barwar, C. Urmi), by contrast, it has developed into a general perfect. The study of the functional and syntactic features of the NENA future form (J. Dohok b-šaqəl) demonstrated that, inter alia, the future is generally avoided in subordinate positions. This likely points to limited grammaticalisation of this form – its persisting assertive function. Chapter 9 investigated the typology of word order in J. Dohok. I confirmed statistically that the basic word order profile of J. Dohok is S/A–V–O, and offered a detailed comparison of J. Dohok with the Trans-Zab dialect of J. Sanandaj, which is characterized by the S/A–O–V order. Despite this difference in basic typology, both dialects share the preference for definite, referentially prominent objects to be preverbal, which has implications for the diachronic shift from V–O to O–V. Such a shift likely took place along the lines of referential prominence – from prominent to non-prominent. Furthermore, postverbal subjects of intransitive verbal clauses in J. Dohok are associated in a statistically significant way with discourse new referents. As a result, discourse-new Ss tend to align with objects (Ps) as opposed to As in their position. This insight was quantified with a new mathematical formula, which demonstrated that J. Dohok has a tendency to an ergative alignment in word order. A similar conclusion will likely be reached for other S/A–V–O NENA dialects, but these dialects still await their investigation in this regard. CONCLUSION 439 This examination of dialectal microvariation has also revealed several unique features of J. Dohok, which distinguish this variety even from the other LD varieties. For instance, in J. Dohok, the basic present 3rd person copula, e.g. 3MS le, almost never occurs with eventive/contingent state (verbal and adverbial) predicates. At the same time, the presentative copula has been bleached of its deictic (or ‘thetic’) function. For most verbal and adverbial predicates, it is the only copula available, occurring in environments that do not fit into the prototypical deictic/thetic category, such as why questions. By contrast, in J. Amediya, J. Zakho and J. Betanure, the 3rd person basic copula occurs also with eventive predicates, while the presentative copula tends to retain its deictic/‘thetic’ force. In general, however, the affiliation of J. Dohok to the other LD dialects has been confirmed, and parallels with other dialects have also emerged. One example of such a parallel is the morphology the historical primae /ʾ/ verbs in the present tense realis paradigm (J. Dohok k-šaqəl) in the Jewish dialects (Lishana Deni and Trans-Zab) and in some Christian varieties, several of which are on the periphery of NENA. In these dialects, the historical primae /ʾ/ verbs have a raised front vowel after the realis preverb k-. This sound contrasts with the vowel a in the inflection of strong verbs, and in the primae /ʾ/ verbs in several Christian dialects, e.g. C. Urmi c-+axəl ‘he eats’. Compare the J. Dohok form k-exəl ‘he eats’ with the J. Sanandaj k-əxlá ‘she eats’ and C. Alqosh k-ixəl (southern periphery of NENA). Such features could be conservatisms, remaining unaffected by the analogical change 440 CONCLUSION from e/a > a taking place in the core Christian NENA dialects. Presumably, the aforementioned dialects remained unaffected due to their communal and/or geographical ‘distance’ from the epicentre of change – either the ethno-religious boundary between Jews and Christians in the case of LD and TZ, or the location on the geographical periphery of NENA in the case of C. Alqosh and TZ. The conservative form of the independent genitive particle did (as opposed to the innovative diy) encountered in Jewish and some Christian varieties is another case in which geographical and communal ‘distance’ has blocked the spread of a linguistic change. Overall, the NENA dialects bear witness to a staggering and fascinating linguistic diversity. Such a diversity cautions before creating firm generalisations about restrictions on language variation and change, at least before all the linguistic diversity is thoroughly investigated. This diversity also shows the limitations of functionalist models. The functionalist explanation for language change can be summarised in the statement that ‘today’s grammar is yesterday’s semantics’.1 Even though function is often a trigger of language change, other factors are also at play: change due to formal analogy and reanalysis, local (i.e. irregular) phonetic change, dialectological influence and language contact. These processes sometimes obscure the original motivation for a given change. For instance, in J. Dohok the L-copula is morphologically identical to the L-suffix, but this identity is 1 This is my own quote inspired by Givón’s idea that ‘today’s morphology is yesterday’s syntax.’ CONCLUSION 441 most likely not (directly) caused by a functional-semantic correspondence of the subject of states with the object of events. Instead, this identity is the product of one possible line of change in which positional analogy of pronominal objects and the copula may have played a role, or which may be an incidental formal convergence of the two morphemes in question. In sum, this description and analysis of J. Dohok demonstrates that the NENA dialects constitute a goldmine for linguists. Still, some of their potential remains untapped. For instance, the newly emerging data published in grammars and sketches hold information concerning dialectal networks of the NENA speech communities, which in turn has implications for these communities’ historical origins and migrations. In the future, these connections can be disentangled using statistical and computer-generated models of classification. Such methods will certainly go beyond tendencies observed in ‘manual’ research, able to verify degrees of connectedness among the different NENA varieties. The NENA-Iranian contact represents another promising research avenue. BIBLIOGRAPHY Abbott, Barbara. 1997. Definiteness and existentials. Language 73 (1): 103–08. Abu-Haidar, Farida. 1991. Christian Arabic of Baghdad. Wiesbaden: Harrasowitz. Akatsuka, Noriko. 1985. “Conditionals and the Epistemic Scale.” Language 61 (3): 625-50. Al-Saka, Vean M. 2019. “Verbal Syntax and Differential Object Marking in Baritle Neo-Aramaic.” Ph.D. Dissertation, Newcastle University. Al-Zebari, Aziz E. 2018. “The Documentation of the Neo-Aramaic Dialect of Aqra: A Holistic Approach.” Ph.D. dissertation, Salahaddin University - Erbil. Barry, Daniel. 2019. “Pharyngeals in Kurmanji Kurdish: A reanalysis of their source and status.” In Current issues in Kurdish linguistics, edited by Songül Gündoǧdu et al, 39-72. Bamberg: University of Bamberg Press. Barry, William. J. 2006. “Phoneme.” In Encyclopedia of Language & Linguistics, edited by Keith Brown, 345–50. Amsterdam: Elsevier. Ben Yaacov, Abraham. 1981. The Jewish Communities of Kurdistan. 2nd ed. Jerusalem: Kiryat Sefer (in Hebrew). Bertinetto, Pier Marco, and Alessandro Lenci. 2012. “Habituality, Pluractionality, and Imperfectivity.” In The Oxford Handbook of Tense and Aspect, edited by Robert E. Binnick, 852–90. Oxford: Oxford University Press. Bhat, D. N. S. 1999. The Prominence of Tense, Aspect and Mood. Amsterdam: John Benjamins. Booij, Geert. 2010. “Construction Morphology.” Linguistics and Language Compass 4 (7): 543–55. 444 BIBLIOGRAPHY Borghero, Roberta. 2008. “Some Features of the Neo-Aramaic Dialect of Karamlesh.” In Neo-Aramaic Dialect Studies, edited by Geoffrey Khan, 75–90. New Jersey: Gorgias Press. Brauer, Erich. 1993. The Jews of Kurdistan. Detroit: Wayne State University Press. Bybee, Joan. 1985. Morphology: A Study of the Relation Between Meaning and Form. Amsterdam/Philadelphia: John Benjamins Publishing Company. ———. 2015. Language Change. Cambridge: Cambridge University Press. Bybee, Joan, and Paul Hopper. 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam/Philadelphia: John Benjamins Publishing Company. Bybee, Joan, William Pagliuca, and Revere D. Perkins. 1991. “Back to the Future.” In Approaches to Grammaticalization: Volume II. Types of Grammatical Markers, edited by Elizabeth C. Traugott and Bernd Heine, 17–58. Amsterdam/Philadelphia: John Benjamincs. ———. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago/London: University of Chicago Press. Carlson, Gregory N. 1980. A Reference to Kinds in English. Outstanding Dissertations in Linguistics. New York/London: Garland. Coghill, Eleanor. 2003. “The Neo-Aramaic Dialect of Alqosh.” Ph.D. Dissertation, University of Cambridge. ———. 2009. “Four Versions of a Neo-Aramaic Children’s Story.” ARAM Periodical 21: 251-80. ———. 2012. “Parallels in the Grammaticalization of Neo-Aramaic zil- and Arabic raḥ- and a Possible Contac Scenario.” In Grammaticalization in Semitic. Journal BIBLIOGRAPHY 445 of Semitic Studies Supplement 29, edited by Domenyk Eades, 127-43. Oxford: Oxford University Press. ———. 2014. “Differential Object Marking in Neo-Aramaic.” Linguistics 52 (2): 335–64. ———. 2016. The Rise and Fall of Ergativity in Aramaic. Oxford: Oxford University Press. Cohen, E. 2008. “The Copular Clause in Jewish Zakho Neo-Aramaic.” Journal of Semitic Studies 53 (1): 43–68. ———. 2012. The Syntax of Neo-Aramaic: The Jewish Dialect of Zakho. Gorgias Neo- Aramaic Studies. Piscataway: Gorgias Press. ———. 2017. “The Functions of Presentative Constructions in Jewish Zakho Neo- Aramaic.” Journal of Jewish Languages 5 (1): 1–21. ———. 2021. Conditional Patterns in the Jewish Neo-Aramaic Dialect of Zakho. 195-220. In Studies in the Gramar and Lexicon of Neo-Aramaic. Cambridge Semitic Languages and Cultures, edited by Geoffrey Khan and Paul Noorlander, 143–93. Cambridge: Open Book Publishers. Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. Comrie, Bernards. 1976. Aspect. Cambridge: Cambridge University Press. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Currie Hall, Kathleen. 2013. “A Typology of Intermediate Phonological Relationships.” The Linguistic Review 30 (2): 215–75. Dahl, Osten. 1985. Tense and Aspect Systems. Oxford: Blackwell. Deo, Ashwini. 2015. “The Semantic and Pragmatic Underpinnings of 446 BIBLIOGRAPHY Grammaticalization Paths: The Progressive to Imperfective Shift.” Semantics and Pragmatics 8 (0): 1–52. Dik, Simon C. 1981. Functional Grammar. 3rd ed. Dordrecht: Foris. ———. 1997. The Theory of Functional Grammar. 2nd ed. Berlin: De Gruyter. Donohue, Mark. 2008. “Semantic Alignment Systems: What’s What, and What’s Not.” In The Typology of Semantic Alignment, edited by Mark Donohue and Søren Winchman, 24–76. Oxford University Press. Dryer, Matthew S. 2007. “Clause Types.” In Language Typology and Syntactic Description Second Edition. Volume 1: Clause Structure, 224–75. Cambridge University Press. Fassberg, Steven. 2015. “The Origin of the Periphrastic Preterite Kəm/Qam- Qāṭəlle in North-Eastern Neo-Aramaic.” In Neo-Aramaic and Its Linguistic Context. Gorgias Neo-Aramaic Studies 14, edited by Geoffrey Khan and Lidia Napiorkowska, 172–86. Piscataway: Gorgias Press. Fassberg, Steven E. 2010. The Jewish Neo-Aramaic Dialect of Challa. Leiden: Brill. Fox, Samuel E. 2015. “The History of the Future.” In Neo-Aramaic and Its Linguistic Context. Gorgias Neo-Aramaic Studies 14, edited by Geoffrey Khan and Lidia Napiorkowska, 162–71. Piscataway: Gorgias Press. ———. 2002. “A Neo-Aramaic Dialect of Bohtan.” In Sprich Doch Mit Deinen Knechten Aramaisch, Wir Verstehen Es!: 60 Beitrage Zur Semitistik. Festschrift Fur Otto Jastrow Zum 60. Geburtstag, edited by Werner Arnold and Hartmut Bobzin, 165–80. Wiesbaden: Harrasowitz. Fox, Samuel E. and Fassberg, Steven E. 2020. “ ‘The verb yhb in Aramaic.” Aramaic BIBLIOGRAPHY 447 Studies 18( 2): 268-78. Garbell, Irene. 1965. “The Impact of Kurdish and Turkish on the Jewish Neo- Aramaic Dialect of Persian Azerbaijan and the Adjoining Regions.” Journal of the American Oriental Society 85 (2): 177. Givón, Thomas. 1977. “From Discourse to Syntax: Grammar as Processing Strategy.” Syntax and Semantics 12: 1-23. ———. 1978. “Negation in Language: Pragmatics, Function, Ontology.” In On Understanding Grammar, edited by Thomas Givón. Amsterdam/Philadelphia: John Benjamins. ———. 1984. “Direct object and dative shifting: Semantic and pragmatic case.” In Objects: Towards a Theory of Grammatical Relations, edited by Frans Plank, 151-82. New York: Academic Press. ———. 1985. “Iconicity, isomorphism, and non-arbitrary coding in syntax.” In Iconicity in Syntax, edited by John Haiman, 187-220. Amsterdam: John Benjamins. Glynn, Dylan. 2010. “Corpus-Driven Cognitive Semantics: Introduction to the Field.” In Quantitative Methods in Cognitive Semantics: Corpus-Driven Approaches, edited by Dylan Glynn and Kerstin Fischer, 1–42. Berlin: De Gruyter. Glynn, Dylan, and Kerstin Fischer, eds. 2010. Quantitative Methods in Cognitive Semantics: Corpus-Driven Approaches. Berlin: De Gruyter. Goldenberg, Gideon. 2013. Semitic Languages: Features, Structures, Relations, Processes. Oxford: Oxford University Press. 448 BIBLIOGRAPHY Greenblatt, Jared. 2008. “The Neo-Aramaic Dialect of the Jews of Amǝdya.” Ph.D. dissertation, University of Cambridge. Grice, Mark. 2006. “Intonation.” In Encyclopedia of Language and Linguistics, edited by E. K. Brown and Anne Anderson, 2nd ed., 778–87. Amsterdam/London: Guirardello, Raquel. 1999. “A Reference Grammar of Trumai.” Ph.D. dissertation, University of Houston, Texas. Gutman, Ariel. 2018. Attributive Constructions in North-Eastern Neo-Aramaic. Berlin: Language Science Press. van de Haan, Ferdinand. 2010. “Typology of Tense, Aspect, and Modality Systems.” In The Oxford Handbook of Linguistic Typology, edited by Jung Song, Jae. Oxford: Oxford University Press. Haig, Geoffrey. 2015. “Verb-Goal (VG) Word Order in Kurdish and Neo-Aramaic: Typological and Areal Considerations.” In Neo-Aramaic and Its Linguistic Context. Gorgias Neo-Aramaic Studies 14, edited by Geoffrey Khan and Lidia Napiorkowska, 407–25. Piscataway: Gorgias. ———. 2017. “Western Asia: East Anatolia as a Transition Zone.” In The Cambridge Handbook of Areal Linguistics, edited by Raymond Hickey, 396–423. Cambridge: Cambridge University Press. ———. 2019. “Post-Predicate Constituents in Kurdish.” In Post-Predicate Constituents in Kurdish, edited by Yaron Matras, Ergin Öpengin, and Geoffrey Haig. London: Palgrave Macmillan. Haig, Geoffrey, and Khan, Geoffrey, eds. 2019. The Languages and Linguistics of Western Asia. Berlin: De Gruyter Mouton. BIBLIOGRAPHY 449 Haig, Geoffrey, and Schnell, Stefan. 2016. “The Discourse Basis of Ergativity Revisited.” Language 92 (3): 1–14. Haspelmath, Martin. 2012. “How to Compare Major Word-Classes across the World’s Languages.” Theories of Everything: In Honor of Edward Keenan 17 (16): 109–30. ———. 2015. “Ditransitive Constructions.” Annual Review of Linguistics 1: 19–41. Hassan, Zeki M., and Esling, John. “Investigating the emphatic feature in Iraqi Arabic: Acoustic and articulatory evidence of coarticulation.” In Instrumental Studies in Arabic Phonetics, edited by Zeki M. Hassan and Barry Hesselwood, 217-34. Amsterdam: John Benjamins. Hengeveld, Kees. 2011. “The grammaticalization of tense and aspect.” In The Oxford Handbook of Grammaticalization, edited by Bernd Heine and Heiko Narrog. Oxford: Oxford University Press. Herring, Susan C., and Paolillo, John C. 1995. “Focus Position in SOV Lanugaes.” In Word Order in Discourse, edited by Pamela E. Downing and Michael Noonan, 163–98. Amsterdam: John Benjamins. Hoberman, Robert D. 1988. “Emphasis Harmony in a Modern Aramaic Dialect.” Language 64 (1): 1–26. ———. 1989. The Syntax and Semantics of Verb Morphology in Modern Aramaic : A Jewish Dialect of Iraqi Kurdistan. New Haven: American Oriental Society. ———. 1997. “Modern Aramaic Phonology.” In Phonologies of Asia and Africa, edited by Alan S. Kaye, 313–35. Winona Lake: Eisenbrauns. ———. 1985. “The Phonology of Pharyngeals and Pharyngealization in Pre- 450 BIBLIOGRAPHY Modern Aramaic.” Journal of the American Oriental Society 105 (2): 221–31. ———. 2006. “Semitic Triradicality or Prosodic Minimality? Evidence from Sound Change.” In Studies in Semitic and Afroasiatic Linguistics Presented to Gene B. Cragg, edited by Cynthia Miller, 139–54. Chicago: The Oriental Institute. Jastrow, Otto. 1992. Lehrbuch Der Turoyo-Sprache. Semitica Viva - Seria Didactica 2. Wiesbaden: Harrasowitz. ———. 1994. Der Neuaramäische Dialekt von Mlaḥsô. Semitica Viva 14. Wiesbaden: Harrassowitz. Kapeliuk, Olga. 1999. “Bgdkpt in Neo-Aramaic.” Handout to paper presented at the 25th NACAL, Philadelphia, 1996. Khan, G. 1999. A Grammar of Neo-Aramaic: The Dialect of the Jews of Arbel. Leiden: Brill. Khan, Geoffrey. 1988. Studies in Semitic Syntax. Oxford: Oxford University Press. ———. 1999. A Grammar of Neo-Aramaic: The Dialect of the Jews of Arbel. ———. 2002. The Neo-Aramaic Dialect of Qaraqosh. Leiden: Brill. ———. 2004a. The Jewish Neo-Aramaic Dialect of Sulemaniyya and Ḥalabja. Brill. ———. 2006. “Some Aspects of the Copula in North West Semitic.” In Biblical Hebrew in Its Northwest Semitic Setting: Typological and Historical Perspectives, edited by Avi Hurvitz and Steven Fassberg, 156–75. Jerusalem/Winona Lake: Magness/Eisenbraun. 2007a. “The North-Eastern Neo-Aramaic Dialects.” Journal of Semitic Studies 52 (1): 1–20. BIBLIOGRAPHY 451 ———. 2007b. “Indicative Markers in North Eastern Neo-Aramaic.” In XII Incontro Italiano Di Linguistica Camito-Semitica, edited by Marco Moriggi, 85–98. Rubbettino: Soveria Mannelli. ———. 2008a. The Jewish Neo-Aramaic Dialect of Urmi. Piscataway: Gorgias Press. ———. 2008b. The Neo-Aramaic Dialect of Barwar. Volume 1: Grammar. Leiden/Boston: Brill. ———. 2009. The Jewish Neo-Aramaic Dialect of Sanandaj. Piscataway: Gorgias Press. ———. 2013a. “Phonological Emphasis in North-Eastern Neo-Aramaic.” In Base Articulatoire Arrière: Backing and Backness, edited by Jean L. Léonard and Samia Naïm. Munich: Lincoln Europa. ———. 2013b. “Syllable Structure: Biblical Hebrew.” In Encyclopedia of Hebrew Language and Linguistics, edited by Geoffrey Khan, online. Leiden: Brill. ———. 2013c. “Guttural Consonants: Masoretic Hebrew.” In Encyclopedia of Hebrew Language and Linguistics, edited by Geoffrey Khan, online. Leiden: Brill. ———. 2016. The Neo-Aramaic Dialect of the Assyrian Christians of Urmi. Volume I: Grammar: Phonology and Morphology. Leiden: Brill. Volume II: Syntax. Leiden: Brill. In the text, no volume number is given if the reference is to volume I. ———. 2017. “Ergativity in Neo-Aramaic.” In The Oxford Handbook of Ergativity, edited by Jessica Coon, Diane Massam, and Lisa D. Travis, 873-99. Oxford: Oxford University Press. ———. 2018. “Remarks on the Historical Development and Syntax of the Copula 452 BIBLIOGRAPHY in North-Eastern Neo-Aramaic Dialects.” Aramaic Studies 16 (2): 234-69. ———. 2021. “Verbal Forms Expressing Discourse Dependency in North-Eastern Neo-Aramaic.” In Studies in the Gramar and Lexicon of Neo-Aramaic. Cambridge Semitic Languages and Cultures, edited by Geoffrey Khan and Paul Noorlander, 143–93. Cambridge: Open Book Publishers. Khaykin, Kira. 2011. “The Verbal System of the Jewish Neo-Aramaic Dialect of Dohok.” Master's dissertation, Tel Aviv University. Kiparsky, Paul. 2003. “Syllables and Moras in Arabic.” In The Syllable in Optimality Theory, edited by Caroline Féry and Ruben van de Vijver, 147–82. Cambridge: Cambridge University Press. Knudsen, Ebbe. 2015. Classical Syriac Phonology. Piscataway: Gorgias Press. Kranich, Svenja. 2010. The progressive in Modern English: A Corpus-Based Study of Grammaticalization and Related Changes. Leiden: Brill. Krauss, Michael. 1992. “The World’s Languages in Crisis.” Language 68 (1): 1–42. Kuty, Renaud Jean. 2008. “Studies in the Syntax of Targum Jonathan to Samuel.” Ph.D. Dissertation, Leiden University. Ladd, D. R. 2011. “Phonetics in Phonology.” In The Handbook of Phonological Theory, 348–73. Oxford: Blackwell. Ladefoged, Peter, and Johnson, Keith. 2011. A Course in Phonetics. 6th ed. Boston: Wadsworth/Cengage Learning. Ladefoged, Peter and Maddieson, Ian. 1997. The Sounds of the World’s Languages. Oxford: Blackwell Publishers. Lambrecht, Knud. 1994. Information Structure and Sentence Form. Cambridge: BIBLIOGRAPHY 453 Cambridge University Press. ———. 2000. “When Subjects Behave like Objects.” Studies in Language 24 (3): 611–82. Leezenberg, Michael. 1993. “Gorani Influence on Central Kurdish: Substratum or Prestige Borrowing?” ILLC - Department of Philosophy, University of Amsterdam. Lehiste, Ilse. 1970. Suprasegmentals. Cambridge, USA: MIT Press. Malchukov, Andrej, Haspelmath, Martin, and Comrie, Bernard. 2011. Studies in Ditransitive Constructions. Berlin: De Gruyter. Matras, Yaron. 2009. Language Contact. Cambridge: Cambridge University Press. McCarthy, John J. 1982. “Nonlinear phonology: An overview.” GLOW Newsletter 8. Mengozzi, Alessandro. 2007. “Middle Markers: Neo-Aramaic and Italian Verbal Forms with a 3rd Singular Feminine Pronominal Object.” In XII Incontro Italiano Di Linguistica Camito-Semitica, edited by M. Moriggi, 105–15. Soveria Mannelli: Rubbettino. Mengozzi, Alessandro. 2012. “The Contribution of Early Christian Vernacular Poetry from Northern Iraq to Neo-Aramaic Dialectology: Preliminary Remarks on the Verbal System.” ARAM 24: 25–40. Molin, Dorota. 2017. “The Language of the Biblical Hebrew Quotations in the Aramaic Incantation Bowls.” MPhil dissertation, University of Cambridge. ———. 2020. “Biblical Quotations in the Aramaic Incantation Bowls and Their Contribution to the Study of the Babylonian Reading Tradition.” In Studies in 454 BIBLIOGRAPHY Semitic Vocalisation and Reading Traditions. Cambridge Semitic Languages and Cultures, edited by Aaron D. Hornkohl and Geoffrey Khan, 148–70. Cambridge: Open Book Publishers. Morgenstern, Matthew. 2011. Studies in Jewish Babylonian Aramaic Based upon Early Eastern Manuscripts. Winona Lake: Eisenbrauns. Muraoka, Takamitsu. 2005. Classical Syriac: A Basic Grammar with a Chrestomathy. Wiesbaden: Harrassowitz. Mutzafi, Hezy. 2002a. “Barzani Jewish Neo-Aramaic and Its Dialects.” Mediterranean Language Review 14: 41–70 ———. 2002b. “On the Jewish Neo-Aramaic Dialect of Aradhin and Its Dialectal Affinities.” In Sprich Doch Mit Deinen Knechten Aramäisch, Wir Verstehen Es!, edited by Werner Arnold and Hartmut Bobzin. Wiesbaden: Harrasowitz. ———. 2004a. The Jewish Neo-Aramaic Dialect of Koy Sanjaq. Wiesbaden: Harrasowitz Verlag. ———. 2004b. The Jewish Neo-Aramaic Dialect of Koy Sanjaq (Iraqi Kurdistan). Wiesbaden: Harrassowitz. ———. 2004c. “Two Texts in Barzani Jewish Neo-Aramaic’.” Bulletin of the School of Oriental and African Studies. Vol. 67. ———. 2008a. The Jewish Neo-Aramaic Dialect of Betanure (Province of Dihok). Wiesbaden: Harrassowitz. ———. 2008b Trans-Zab Jewish Neo-Aramaic. Bulletin of the School of Oriental and African Studies 71 (3): 409-31. ———. 2016. “Jewish Neo-Aramaic Innovations.” Journal of Jewish Languages 4: BIBLIOGRAPHY 455 85–108. ———. 2018. “Further Jewish Neo-Aramaic Innovations.” Journal of Jewish Languages 6: 145–181. Napiorkowska, Lidia. 2015. A Grammar of the Christian Neo-Aramaic Dialect of Diyana-Zariwaw. Leiden: Brill. ———. 2015. “A Diachronic Perspective on Emphasis in Christian Diyana- Zariwaw.” In Neo-Aramaic and Its Linguistic Context, edited by Geoffrey Khan and Lidia Napiorowska, 130–45. Piscataway: Gorgias Press. Nedjalkov, Vladimir P. 2001. “Resultative Constructions in Syntactic Typology.” In Language Typology and Language Universals, edited by Martin Haspelmath, 928–40. Berlin: De Gruyter. Noorlander, Paul M. 2017. “The Proximative and Its Correlatives in North Eastern Neo-Aramaic.” In Prospective and Proximative in Turkic, Iranian and Beyond, edited by Agnes Korn and Irina Nevskaya, 187–210. Wiesbaden: Reichet. ———. 2018. “Alignment in Eastern Neo-Aramaic Languages from a Typological Perspective.” Ph.D. dissertation, Leiden University. ———. 2021. “Towards a Typology of Experiencers and Possessors in Neo- Aramaic: Non-Canonical Subjects as Relics of a Former Dative Case.” In Studies in the Grammar and Lexicon of Neo-Aramaic. Cambridge Semitic Languages and Cultures, edited by Geoffrey Khan and Paul M. Noorlander, 29–94. Cambridge: Open Book Publishers. ———. forthcoming 2021. “Source Constructions as a Key to Alignment Change: The Case of Aramaic.” Journal of Historical Linguistics 11. 456 BIBLIOGRAPHY Noorlander, Paul M. and Molin, Dorota. 2020. “Contact, conservatism and convergence in the development of Arabic and Aramaic dialects.” Paper presented at Tracing Contact in Closely Related Languages, University of Zurych, 19-20 November 2020. ———. forthcoming 2022. “Word Order Typology in North Eastern Neo-Aramaic: Towards a Corpus-Based Approach.” Language Typology and Universals (STUF). Ozerov, Pavel. 2018. “Tracing the Sources of Information Structure: Towards the Study of Interactional Management of Information.” Journal of Pragmatics 138: 77-97. Palmer, Frank R. 2001. Mood and Modality. Cambridge: Cambridge University Press. Polotsky. 1961. “Studies in Modern Syriac.” Journal of Semitic Studies 6: 1–32. Polotzky, Hayim. 1967. “Eastern Neo-Aramaic: Urmi and Zakho.” In An Aramaic Handbook. Porta Linguarum Orientalium 2, part 2, edited by Franz Rosenthal, 69–77. Wiesbaden: Harrasowitz. Retsö, Jan. 1987. “Copula and Double Pronominal Objects in Some Semitic Languages.” Zeitschrift Der Deutschen Morgenländischen Gesellschaft 137 (1987) 219-245. Rubin, Aaron D. 2018. Studies in Semitic Grammaticalization. Harvard Semitic Studies 57. Leiden: Brill. Sabar, Yona. 1976. Pešat Wayehî Bešallaḥ : A Neo-Aramaic Midrash on Beshallaḥ (Exodus). Wiesbaden: Harrassowitz. ———. “The Christian Neo-Aramaic Dialects of Zakho and Dihok: Two Text BIBLIOGRAPHY 457 Samples.” Journal of the American Oriental Society 11 (1), pp. 33-51. ———. 2002. A Jewish Neo-Aramaic Dictionary. Wiesbaden: Harrasowitz Verlag. Sasse, Hans-J. 1987. “The Thetic Categorical Distinction Revisited.” Linguistics 25: 511–80. Seržant, Ilja A. (2013). The Diachronic Typology of Non-Canonical Subjects and Subject-like Obliques. In The Diachronic Typology of Non-Canonical Subject, edited by Ilja A. Seržant and Leonid Kulikov 314–60. Amsterdam: John Benjamins. Shirai, Yasuhiro. 1998. “Where the Progressive and the Resultative Meet: Imperfective Aspect in Japanese, Chinese, Korean and English.” Studies in Language 22 (3): 661–92. Siewierska, Anna. 1998. “Variation in Major Constituent Word Order: A Global and a European Perspective.” In Constituent Order in the Languages of Europe. Typology of Languages in Europe, edited by Anna Siewierska. Berlin: De Gruyter. Stefanowitsch, Anatol. 2020. Corpus Linguistics: A Guide to the Methodology. Berlin: Language Science Press. Stilo, Donald. 2018. “Preverbal and Postverbal Peripheral Arguments in the Araxes-Iran Linguistic Area. Invited Lecture at the Conference Anatolia- Caucasus-Iran: Ethnic and Linguistic Contacts, Yerevan University, 10-12 May 2018.” Stilo, Donald, and Noorlander, Paul M. 2015. “On the Convergence of Verbal Systems of Aramaic and Its Neighbors. Part II: Past Paradigms Derived from 458 Present Equivalents.” In Neo-Aramaic and Its Linguistic Context. Gorgias Neo- Aramaic Studies 14, edited by Geoffrey Khan and Lidia Napiorkowska, 453−484. Piscataway: Gorgias. Thackston, Wheeler M. 2006. Kurmanji Kurdish: A Reference Grammar with Selected Readings. Diyarbakir: Renas Media. Vennemann, Theo. 1987. Preference Laws for Syllable Structure. Berlin, Boston: De Gruyter. Watson, Janet C. E. 1999. “The Directionality of Emphasis Spread in Arabic.” Linguistic Inquiry 30 (2): 289–300. ———. 2007. The Phonology and Morphology of Arabic. Oxford: Oxford University Press.