A grammar of Japhug

A grammar of Japhug Second edition Guillaume Jacques lan guage sc ie n ce pr ess Comprehensive Grammar Library 1 Comprehensive Grammar Library Editor: Martin Haspelmath In this series: 1. Jacques, Guillaume. A grammar of Japhug. 2. Grimm, Nadine. A grammar of Gyeli. 3. MaurerCecchini, Philippe. A grammar of T uatschin: A Sursilvan Romansh dialect. 4. Visser , Eline. A grammar of Kalamang. 5. Caballero, Gabriela. A grammar of Choguita Rarámuri: In collaboration with Luz Elena León Ramírez, Sebastián Fuentes Holguín, Bertha Fuentes Loya and other Choguita Rarámuri language experts. 6. Barlow , Russell. A grammar of Ulwa (Papua New Guinea). 7. T erhart, Lena. A grammar of Paunaka. 8. Pebley , Carol J. & Thomas E. Payne. A grammar of Kagayanen. 9. Rohleder , Jean. A grammar of V amale. 10. Kasak, Ryan. A grammar of Mandan. ISSN (print): 2748971X ISSN (electronic): 27497798 A grammar of Japhug Second edition Guillaume Jacques lan guage sc ie n ce pr ess Jacques, Guillaume. 2025. A grammar of Japhug : Second e dition ( Comprehensive Grammar Library 1). Berlin: Language Science Press. This title can be downloaded at: http://langsci- press.org/catalog/book/549 © 2025, Guillaume Jacques Published under the Creative Commons Attribution 4. 0 Licence (CC BY 4.0): http://creativ e commons.org/licenses/by/4. 0/ ISBN: 978-3-96110-538-0 (Digital) 978-3-98554-158-4 (Hardcov er) ISSN (print): 2748-971X ISSN ( ele ctr onic): 2749-7798 DOI: 10.5281/zenodo.16811879 Source code available from ww w .github .com/langsci/295 Errata: paperhive.org/documents/remote?type=langsci&id=295 Cover and concept of design: Ulrike Harbort T ypesetting: Guillaume Jacques Proofreading: Anton Antonov , Curtis Bartosik, Aimée Lahaussois, Laura Arnold, T om Bossuyt, Michael Daniel, V alérie Guérin, Ariel Gutman, Andreas Hölzl, Katarzyna Janic, Seppo Kittilä, Li Shang 李上 , Liu Huiyong 刘慧勇 , Joseph Lovestrand, Lachlan Mackenzie , Bruno Olsson, Bastian Persohn, Y vonne T reis, Jeroen van de W eijer and Y e Jingting 叶婧婷 Fonts: Libertinus, Arimo, DejaV u Sans Mono T ypesetting software: X Ǝ L A T E X Language Science Press Scharnweberstraße 10 10247 Berlin, Germany http://langsci- press.org [email protected] Storage and cataloguing done by F U Berlin Pour mon amour , Mirabelle 题献给爱妻米哈 Contents Ackno wledgments x xiii Abbreviations xxv 1 Introduction 1 1 . 1 T h e J a p h u g l a n g u a g e ........................ 1 1 . 2 T h e J a p h u g c o r p u s .......................... 2 1.3 Structure of the grammar . . . . . . . . . . . . . . . . . . . . . . 4 1.4 Language and culture . . . . . . . . . . . . . . . . . . . . . . . . 4 2 A grammatical sketch 7 2.1 Phonology and wor d structure . . . . . . . . . . . . . . . . . . . 7 2 . 2 P a r t s o f s p e e c h ............................ 8 2 . 3 N o m i n a l m o r p h o l o g y ........................ 9 2.3.1 Non-attested nominal morphological categories . . . . . 9 2 . 3 . 2 N u m b e r ........................... 9 2 . 3 . 3 C a s e m a r k i n g ........................ 1 1 2 . 3 . 4 P o s s e s s i o n .......................... 1 2 2 . 3 . 5 C o m p o u n d i n g ........................ 1 5 2 . 3 . 6 D e r i v a t i o n s ......................... 1 5 2 . 4 V e r b a l m o r p h o l o g y ......................... 1 6 2 . 4 . 1 O v e r v i e w .......................... 1 6 2 . 4 . 2 I n d e x a t i o n .......................... 1 7 2.4.3 Orientation preverbs and T AME . . . . . . . . . . . . . . 21 2.4.4 Non-finite verb forms . . . . . . . . . . . . . . . . . . . 24 2.4.5 Associated motion . . . . . . . . . . . . . . . . . . . . . 28 2 . 4 . 6 V o i c e ............................. 2 8 2.4.7 Denominal deriva tions . . . . . . . . . . . . . . . . . . . 33 2.5 Core and oblique arguments . . . . . . . . . . . . . . . . . . . . 33 2.5.1 Neutral alignment . . . . . . . . . . . . . . . . . . . . . 34 2.5.2 Nominative-accusative alignment . . . . . . . . . . . . . 34 2 . 5 . 3 S u b j e c t h o o d ......................... 3 4 Contents 2 . 5 . 4 O b j e c t h o o d ......................... 3 5 2.5.5 Absolutive-ergative alignment . . . . . . . . . . . . . . . 36 2.5.6 Ditransitive verbs . . . . . . . . . . . . . . . . . . . . . . 37 2 . 6 W o r d o r d e r .............................. 3 7 2 . 7 S u b o r d i n a t i o n ............................ 3 8 2.7.1 Relative clauses . . . . . . . . . . . . . . . . . . . . . . . 38 2.7.2 Complement clauses . . . . . . . . . . . . . . . . . . . . 38 2.7.3 Other subordinate clauses . . . . . . . . . . . . . . . . . 39 2.8 Remarkable features . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.8.1 Consonant clusters . . . . . . . . . . . . . . . . . . . . . 41 2 . 8 . 2 D i r e c t - i n v e r s e ........................ 4 1 2.8.3 Inflectionalization . . . . . . . . . . . . . . . . . . . . . . 42 2 . 8 . 4 P r e fi x a l c h a i n ........................ 4 2 2.8.5 Hybrid indirect spe ech . . . . . . . . . . . . . . . . . . . 43 2.8.6 The expr ession of degree and comparison . . . . . . . . 44 2.8.7 Japhug morphology and T rans-Himalayan comparative l i n g u i s t i c s .......................... 4 5 3 P honology 47 3 . 1 I n t r o d u c t i o n ............................. 4 7 3 . 2 C o n s o n a n t s .............................. 4 7 3 . 2 . 1 O n s e t s ............................ 4 7 3 . 2 . 2 C o d a s ............................ 5 0 3 . 3 V o w e l s a n d r h y m e s ......................... 5 2 3 . 3 . 1 V o w e l s ............................ 5 2 3 . 3 . 2 R h y m e s ........................... 5 4 3.3.3 Historical phonology . . . . . . . . . . . . . . . . . . . . 55 3.4 Syllabic constraints . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.4.1 Rhotic dissimilation . . . . . . . . . . . . . . . . . . . . 57 3.4.2 Uvular harmony . . . . . . . . . . . . . . . . . . . . . . 58 3.5 Neutralization, quasi-neutralization and free variation . . . . . . 59 3.5.1 The contrast between /-oŋ/ and /-aŋ/ .......... 5 9 3.5.2 The contrast between /ɯ/ and /i/ after palatal and alveolo-palatal consonants . . . . . . . . . . . . . . . . . 60 3.5.3 The contrast between /ɤ/ and /e/ after palatal and alveolo-palatal consonants . . . . . . . . . . . . . . . . . 65 3.5.4 The contrast between /ɤ/ and /a/ ............ 6 5 3.6 Spee ch errors and self-corrections . . . . . . . . . . . . . . . . . 66 3 . 7 S u p r a s e g m e n t a l s ........................... 6 8 ii Contents 3 . 8 W o r d s t r u c t u r e ............................ 6 8 3 . 8 . 1 W o r d h o o d .......................... 6 8 3.8.2 Non-final syllables . . . . . . . . . . . . . . . . . . . . . 70 4 Consonant clusters and partial reduplication 71 4.1 Partial Re duplication . . . . . . . . . . . . . . . . . . . . . . . . 71 4.2 Inventory of consona nt clusters . . . . . . . . . . . . . . . . . . 72 4 . 2 . 1 P r e i n i t i a l s .......................... 7 3 4 . 2 . 2 M e d i a l s ........................... 8 7 4 . 2 . 3 S u m m a r y .......................... 9 9 4 . 3 S a n d h i ................................ 1 0 4 4.4 Tibetan script-based orthography . . . . . . . . . . . . . . . . . 106 5 Nominal morphology 109 5.1 Possessive prefixes . . . . . . . . . . . . . . . . . . . . . . . . . 109 5.1.1 Possessive paradigm . . . . . . . . . . . . . . . . . . . . 109 5.1.2 Inalienably p ossessed nouns . . . . . . . . . . . . . . . . 116 5.1.3 Indefinite vs. generic possessor . . . . . . . . . . . . . . 134 5.1.4 Prenominal modifiers .................... 137 5.2 Unpossessible nouns . . . . . . . . . . . . . . . . . . . . . . . . 139 5 . 2 . 1 P l a c e n a m e s ......................... 1 3 9 5 . 2 . 2 C o l o u r n o u n s ........................ 1 4 2 5.2.3 Other unp ossessible nouns . . . . . . . . . . . . . . . . . 143 5 . 3 P e r s o n a l n a m e s ........................... 1 4 4 5 . 3 . 1 V o c a t i v e ........................... 1 4 4 5 . 3 . 2 T i b e t a n n a m e s ....................... 1 4 5 5.3.3 Alienably p ossessed or unp ossessible nouns? . . . . . . 146 5.3.4 Personal names and modifiers . . . . . . . . . . . . . . . 147 5 . 4 B o u n d s t a t e ............................. 1 4 9 5.4.1 V owel alternations in non-final members of compounds 149 5.4.2 Other alternations . . . . . . . . . . . . . . . . . . . . . 150 5.4.3 Final memb er of compounds . . . . . . . . . . . . . . . . 153 5 . 5 C o m p o u n d n o u n s .......................... 1 5 5 5.5.1 Noun-Noun comp ounds . . . . . . . . . . . . . . . . . . 155 5.5.2 V erb- V erb compounds . . . . . . . . . . . . . . . . . . . 159 5.5.3 Noun-Ideophone compounds . . . . . . . . . . . . . . . 160 5.5.4 Adv erb- V erb compounds . . . . . . . . . . . . . . . . . . 161 5.5.5 Noun- V erb compounds . . . . . . . . . . . . . . . . . . . 161 5.5.6 V erb-Noun compounds . . . . . . . . . . . . . . . . . . . 168 iii Contents 15 Orientation and associated motion 627 15.1 Orientation pre verbs . . . . . . . . . . . . . . . . . . . . . . . . 627 1 5 . 1 . 1 M o r p h o l o g y ......................... 6 2 7 15.1.2 Orientable verbs . . . . . . . . . . . . . . . . . . . . . . 638 15.1.3 The tridimensional system . . . . . . . . . . . . . . . . . 647 15.1.4 Extended uses of orientations . . . . . . . . . . . . . . . 658 15.1.5 Lexicalized orientations . . . . . . . . . . . . . . . . . . 668 1 5 . 2 A s s o c i a t e d m o t i o n .......................... 6 8 8 15.2.1 AM prefixes: morphology . . . . . . . . . . . . . . . . . 688 15.2.2 Argument of motion . . . . . . . . . . . . . . . . . . . . 694 15.2.3 Motion verbs and AM prefixes . . . . . . . . . . . . . . . 696 15.2.4 Orientation and AM . . . . . . . . . . . . . . . . . . . . 697 1 5 . 2 . 5 R o u n d t r i p s ......................... 6 9 9 15.2.6 The nature of the motion . . . . . . . . . . . . . . . . . 700 1 5 . 2 . 7 G o a l ............................. 7 0 2 15.2.8 Echo phenomena . . . . . . . . . . . . . . . . . . . . . . 704 15.2.9 The verb ru ‘fetch, bring’ . . . . . . . . . . . . . . . . . . 708 15.2.10 Associated motion vs. motion verb construction . . . . . 711 16 Non-finite verbal morphology 721 1 6 . 1 P a r t i c i p l e s .............................. 7 2 1 16.1.1 Subject participles . . . . . . . . . . . . . . . . . . . . . 722 16.1.2 Object participles . . . . . . . . . . . . . . . . . . . . . . 740 16.1.3 Oblique participles . . . . . . . . . . . . . . . . . . . . . 753 1 6 . 2 I n fi n i t i v e s .............................. 7 7 0 16.2.1 V elar infinitives . . . . . . . . . . . . . . . . . . . . . . . 770 16.2.2 Bare infinitives . . . . . . . . . . . . . . . . . . . . . . . 782 16.2.3 Dental infinitives . . . . . . . . . . . . . . . . . . . . . . 784 1 6 . 3 D e g r e e n o m i n a l s ........................... 7 8 6 16.3.1 Polarity prefixes . . . . . . . . . . . . . . . . . . . . . . 788 16.3.2 Argument structure . . . . . . . . . . . . . . . . . . . . 788 16.3.3 Nominal predicates . . . . . . . . . . . . . . . . . . . . . 789 16.3.4 Degree construction . . . . . . . . . . . . . . . . . . . . 790 16.3.5 Complementation . . . . . . . . . . . . . . . . . . . . . . 790 16.4 Action nominals and abstract noun . . . . . . . . . . . . . . . . 791 16.4.1 tɯ- a c t i o n n o m i n a l s .................... 7 9 1 16.4.2 tɤ- a b s t r a c t n o u n s ..................... 7 9 3 1 6 . 4 . 3 S i m u l t a n e o u s ........................ 7 9 5 16.4.4 Denominalization of action nominals . . . . . . . . . . . 796 x Contents 16.4.5 Lexicalized action nominals . . . . . . . . . . . . . . . . 797 16.4.6 Inalienably possessed bare action nominals . . . . . . . 799 16.4.7 Action nominal compounds . . . . . . . . . . . . . . . . 802 16.5 Other deverbal nouns . . . . . . . . . . . . . . . . . . . . . . . . 805 16.5.1 Nominalization -z s u ffi x .................. 8 0 5 16.5.2 Nominalization ɣ-/x- p r e fi x ................ 8 0 7 1 6 . 6 C o n v e r b s............................... 8 0 8 1 6 . 6 . 1 G e r u n d ........................... 8 0 8 1 6 . 6 . 2 P u r p o s i v e .......................... 8 1 4 1 6 . 6 . 3 I m m e d i a t e .......................... 8 1 6 16.6.4 Adverb from finite verb . . . . . . . . . . . . . . . . . . 818 1 6 . 7 D e f e c t i v e v e r b s ........................... 8 1 9 16.8 Historical p erspectives . . . . . . . . . . . . . . . . . . . . . . . 819 16.8.1 V elar non-finite prefixes . . . . . . . . . . . . . . . . . . 819 16.8.2 Sigmatic non-finite prefixes . . . . . . . . . . . . . . . . 822 16.8.3 Dental non-finite prefixes . . . . . . . . . . . . . . . . . 823 17 V alency-increasing derivations 827 1 7 . 1 I n t r o d u c t i o n ............................. 8 2 7 17.2 Sigmatic causative . . . . . . . . . . . . . . . . . . . . . . . . . . 827 17.2.1 Regular allomorphy . . . . . . . . . . . . . . . . . . . . 827 17.2.2 Irregular allomorphs . . . . . . . . . . . . . . . . . . . . 830 17.2.3 Lexicalized sigmatic causatives . . . . . . . . . . . . . . 836 17.2.4 Morphosyntax . . . . . . . . . . . . . . . . . . . . . . . 838 17.2.5 Semantics of the causative . . . . . . . . . . . . . . . . . 846 1 7 . 2 . 6 S t a t i v e v e r b s ......................... 8 5 4 1 7 . 2 . 7 R e c u r s i o n .......................... 8 5 5 17.2.8 Compatibility with other derivations . . . . . . . . . . . 856 1 7 . 3 V e l a r c a u s a t i v e ............................ 8 5 7 17.3.1 Irregular allomorphs . . . . . . . . . . . . . . . . . . . . 859 17.3.2 Morphosyntax . . . . . . . . . . . . . . . . . . . . . . . 861 1 7 . 3 . 3 S e m a n t i c s .......................... 8 6 4 17.3.4 Compatibilities with other derivations . . . . . . . . . . 867 1 7 . 4 A p p l i c a t i v e .............................. 8 6 7 17.4.1 The syntactic and semantic functions of the promoted a r g u m e n t .......................... 8 6 8 1 7 . 4 . 2 A l l o m o r p h y ......................... 8 7 4 17.4.3 Lexicalized applicatives . . . . . . . . . . . . . . . . . . 874 17.4.4 Applicatives and other derivations . . . . . . . . . . . . 876 xi Contents 1 7 . 5 E s t i m a t i v e .............................. 8 7 6 1 7 . 5 . 1 A l l o m o r p h y ......................... 8 7 8 17.5.2 Past imperfective . . . . . . . . . . . . . . . . . . . . . . 878 17.5.3 Lexicalized estimatives . . . . . . . . . . . . . . . . . . . 879 17.5.4 Compatibility with other derivations . . . . . . . . . . . 880 17.5.5 Other estimative constructions . . . . . . . . . . . . . . 882 18 V alency-de creasing derivations 883 1 8 . 1 P a s s i v e ................................ 8 8 3 18.1.1 The interaction of the Passive derivation with Dynamicity and T AME . . . . . . . . . . . . . . . . . . . 885 18.1.2 Lexicalized passives . . . . . . . . . . . . . . . . . . . . 886 18.1.3 Agent demotion . . . . . . . . . . . . . . . . . . . . . . . 888 18.1.4 Passive from ditransitive verbs . . . . . . . . . . . . . . 892 18.1.5 Reduplicate d passive . . . . . . . . . . . . . . . . . . . . 892 18.1.6 Compatibility with other derivations . . . . . . . . . . . 893 1 8 . 2 R o g a t i v e ............................... 8 9 5 1 8 . 3 R e fl e x i v e ............................... 8 9 6 18.3.1 The reflexiv e prefix: form and basic function . . . . . . . 896 18.3.2 Reflexive vs. anticausative . . . . . . . . . . . . . . . . . 901 18.3.3 Lexicalized reflexiv es . . . . . . . . . . . . . . . . . . . . 903 18.3.4 Reflexive causative . . . . . . . . . . . . . . . . . . . . . 904 18.3.5 Reflexive and autive . . . . . . . . . . . . . . . . . . . . 908 18.3.6 Reflexive , estimative and applicative . . . . . . . . . . . 909 18.3.7 Historical origin . . . . . . . . . . . . . . . . . . . . . . 910 1 8 . 4 R e c i p r o c a l .............................. 9 1 1 18.4.1 Reduplicate d reciprocal . . . . . . . . . . . . . . . . . . 911 18.4.2 Reciprocal amɯ- p r e fi x ................... 9 1 8 18.4.3 Reciprocal andʑɯ- p r e fi x .................. 9 2 3 18.4.4 V erbs of co-participation . . . . . . . . . . . . . . . . . . 923 1 8 . 5 A n t i c a u s a t i v e ............................ 9 2 4 1 8 . 5 . 1 M o r p h o l o g y ......................... 9 2 4 1 8 . 5 . 2 F u n c t i o n ........................... 9 2 8 18.5.3 Anticausative and dummy subject constructions . . . . 930 1 8 . 5 . 4 C o l l o c a t i o n ......................... 9 3 1 1 8 . 5 . 5 V o l i t i o n a l i t y ......................... 9 3 2 18.5.6 Compatibility with other derivations . . . . . . . . . . . 934 18.5.7 Other cases of voicing alternation . . . . . . . . . . . . . 937 xii Contents 1 8 . 6 A n t i p a s s i v e .............................. 9 3 7 18.6.1 rɤ- a n t i p a s s i v e ....................... 9 3 7 18.6.2 sɤ- a n t i p a s s i v e ....................... 9 4 1 18.6.3 Lexicalized antipassive . . . . . . . . . . . . . . . . . . . 942 18.6.4 Antipassive forms of ditransitive verbs . . . . . . . . . . 944 18.6.5 Reduplicate d antipassive . . . . . . . . . . . . . . . . . . 947 18.6.6 Antipassive and past imperfective . . . . . . . . . . . . . 948 18.6.7 The uses of the antipassive derivations . . . . . . . . . . 949 18.6.8 Other strategies used to express indefinite patients . . . 954 18.6.9 Compatibility with other derivations . . . . . . . . . . . 957 18.7 Distributed property . . . . . . . . . . . . . . . . . . . . . . . . . 958 1 8 . 8 P r o p r i e t i v e .............................. 9 6 1 1 8 . 8 . 1 A l l o m o r p h y ......................... 9 6 3 18.8.2 Proprietive derivation and generic marking . . . . . . . 963 18.8.3 Proprietive derivations from semi-transitiv e verbs . . . 965 18.8.4 Proprietive derivations from transitive verbs . . . . . . . 966 18.8.5 Lexicalized proprietiv e . . . . . . . . . . . . . . . . . . . 967 18.8.6 Compatibility with other derivations . . . . . . . . . . . 968 18.8.7 Relationship with other derivations . . . . . . . . . . . . 969 1 8 . 9 F a c i l i t a t i v e .............................. 9 6 9 18.9.1 Subject-oriente d facilitative . . . . . . . . . . . . . . . . 970 18.9.2 Object-oriente d facilitative . . . . . . . . . . . . . . . . . 972 19 Other verbal derivations 975 1 9 . 1 A u t i v e ................................ 9 7 5 19.1.1 The autive prefix and verb transitivity . . . . . . . . . . 975 19.1.2 Position in the verbal template and allomorphy . . . . . 976 19.1.3 A utobenefactive/self-affectedness function . . . . . . . . 978 19.1.4 Spontane ous function . . . . . . . . . . . . . . . . . . . 981 19.1.5 Permansive function . . . . . . . . . . . . . . . . . . . . 985 19.1.6 Lexicalized autives . . . . . . . . . . . . . . . . . . . . . 987 19.1.7 Historical relationship with other derivations . . . . . . 990 1 9 . 2 V e r t i t i v e ............................... 9 9 0 1 9 . 3 A b i l i t a t i v e .............................. 9 9 2 19.3.1 Lexicalized abilitatives . . . . . . . . . . . . . . . . . . . 994 19.3.2 Historical origin . . . . . . . . . . . . . . . . . . . . . . 994 19.4 Distributed action . . . . . . . . . . . . . . . . . . . . . . . . . . 996 19.4.1 Lexicalized distributed action verbs . . . . . . . . . . . . 1000 19.4.2 Irregular partial reduplication . . . . . . . . . . . . . . . 1002 xiii Contents 19.4.3 Compatibilities with other derivations . . . . . . . . . . 1003 19.5 A uto-evaluativ e . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003 19.6 Attenuative reduplication . . . . . . . . . . . . . . . . . . . . . . 1004 19.7 Fossil affixes and marginal derivations . . . . . . . . . . . . . . 1005 19.7.1 V olitional mɯ- p r e fi x .................... 1 0 0 5 19.7.2 Applicative -t s u ffi x ..................... 1 0 0 6 19.7.3 Antipassive -t s u ffi x .................... 1 0 0 8 19.7.4 Other detransitive prefixes . . . . . . . . . . . . . . . . . 1010 19.7.5 rɤ- p r e fi x .......................... 1 0 1 0 19.7.6 ɣɯ-/ɣɤ- p r e fi x ....................... 1 0 1 2 19.7.7 a- p r e fi x ........................... 1 0 1 3 19.7.8 Abilitative j- p r e fi x ..................... 1 0 1 4 19.7.9 Prenasalization . . . . . . . . . . . . . . . . . . . . . . . 1014 19.7.10 Comparative derivation? . . . . . . . . . . . . . . . . . . 1015 19.7.11 Reduplication . . . . . . . . . . . . . . . . . . . . . . . . 1017 19.7.12 V owel alternation . . . . . . . . . . . . . . . . . . . . . . 1017 20 Denominal derivations 1019 2 0 . 1 I n t r o d u c t i o n ............................. 1 0 1 9 20.1.1 Morphological properties of denominal prefixes . . . . . 1020 20.1.2 Denominal derivations and light verb constructions . . . 1020 20.2 Contracting prefixes . . . . . . . . . . . . . . . . . . . . . . . . . 1022 20.2.1 Stative a- .......................... 1 0 2 2 20.2.2 Similative arɯ- ....................... 1 0 2 4 20.2.3 Proprietive arɤ- ....................... 1 0 2 6 20.2.4 Proprietive aɣɯ- ...................... 1 0 2 7 20.2.5 Collective andʑi- ...................... 1 0 3 2 20.3 Sigmatic denominal prefixes . . . . . . . . . . . . . . . . . . . . 1033 20.3.1 Proprietive sɤ- ....................... 1 0 3 3 20.3.2 Causative/instrumental sV- ................ 1 0 3 5 20.4 Rhotic denominal prefixes . . . . . . . . . . . . . . . . . . . . . 1039 20.4.1 Intransitive denominal verbs . . . . . . . . . . . . . . . 1039 20.4.2 Transitiv e denominal verbs . . . . . . . . . . . . . . . . 1042 20.4.3 Pairing with other denominal prefixes . . . . . . . . . . 1044 20.5 V elar denominal prefixes . . . . . . . . . . . . . . . . . . . . . . 1046 20.5.1 Intransitive denominal verbs . . . . . . . . . . . . . . . 1046 20.5.2 Transitiv e denominal verbs . . . . . . . . . . . . . . . . 1048 20.5.3 Pairing with other denominal prefixes . . . . . . . . . . 1050 20.6 Labial nasal denominal prefixes . . . . . . . . . . . . . . . . . . 1052 xiv Contents 20.7 Dental n asal denominal prefixes . . . . . . . . . . . . . . . . . . 1054 2 0 . 7 . 1 I n t r a n s i t i v e ......................... 1 0 5 4 2 0 . 7 . 2 T r a n s i t i v e .......................... 1 0 5 8 20.7.3 Pairing with other denominal prefixes . . . . . . . . . . 1060 20.8 Other denominal verbs . . . . . . . . . . . . . . . . . . . . . . . 1061 20.8.1 Zero-derivation or backformation? . . . . . . . . . . . . 1061 20.8.2 V owel alternation . . . . . . . . . . . . . . . . . . . . . . 1063 20.9 Deideophonic verbs . . . . . . . . . . . . . . . . . . . . . . . . . 1064 20.9.1 ɣɤ- and sɤ- deideophonic verbs . . . . . . . . . . . . . . 1064 20.9.2 nɯ- deideophonic verbs . . . . . . . . . . . . . . . . . . 1066 20.9.3 a- and nɤ- deideophonc verbs . . . . . . . . . . . . . . . 1067 20.10 The denominal origin of voice prefixes . . . . . . . . . . . . . . 1068 20.10.1 The origin of the rɤ- antipassive prefix . . . . . . . . . . 1068 20.10.2 The proprietive and antipassiv e sɤ- prefixes and their estimative counterpart . . . . . . . . . . . . . . . . . . . 1072 20.10.3 Applicative , sigmatic causative and passive . . . . . . . 1073 2 0 . 1 1 L o a n v e r b s .............................. 1 0 7 4 2 0 . 1 2 C o m p o u n d v e r b s ........................... 1 0 7 5 2 0 . 1 3 I n c o r p o r a t i o n ............................ 1 0 7 8 20.13.1 Incorporation and denominal derivation . . . . . . . . . 1078 20.13.2 Incorporation and other derivations . . . . . . . . . . . 1083 20.13.3 Syntactic funct ion of the incorp orated wor d . . . . . . . 1084 20.13.4 Incorporation and other constructions . . . . . . . . . . 1086 21 T ense , asp ect, modality and evidentiality 1091 2 1 . 1 I n t r o d u c t i o n ............................. 1 0 9 1 2 1 . 1 . 1 P r e v e r b s ........................... 1 0 9 2 21.1.2 Stem alternation . . . . . . . . . . . . . . . . . . . . . . 1095 2 1 . 1 . 3 O t h e r a ffi x e s ......................... 1 0 9 6 21.1.4 Evidentiality and p erson . . . . . . . . . . . . . . . . . . 1097 2 1 . 2 I m p e r f e c t i v e ............................. 1 0 9 7 2 1 . 2 . 1 M o r p h o l o g y ......................... 1 0 9 8 21.2.2 Use in p eriphrastic T AME categories . . . . . . . . . . . 1099 21.2.3 Use in temp oral clauses . . . . . . . . . . . . . . . . . . 1103 21.2.4 Use in complement clauses . . . . . . . . . . . . . . . . 1104 2 1 . 2 . 5 H o r t a t i v e .......................... 1 1 0 5 2 1 . 2 . 6 I n c h o a t i v e.......................... 1 1 0 6 21.2.7 Perception verbs . . . . . . . . . . . . . . . . . . . . . . 1107 xv Contents 21.3 Non-past categories . . . . . . . . . . . . . . . . . . . . . . . . . 1110 21.3.1 Factual Non-Past . . . . . . . . . . . . . . . . . . . . . . 1110 2 1 . 3 . 2 S e n s o r y ........................... 1 1 1 4 21.3.3 Egophoric Present . . . . . . . . . . . . . . . . . . . . . 1122 21.3.4 Tripartite contrast . . . . . . . . . . . . . . . . . . . . . 1128 21.4 Modal categories . . . . . . . . . . . . . . . . . . . . . . . . . . . 1129 2 1 . 4 . 1 I r r e a l i s ............................ 1 1 2 9 2 1 . 4 . 2 I m p e r a t i v e .......................... 1 1 3 6 2 1 . 4 . 3 P r o h i b i t i v e .......................... 1 1 4 0 2 1 . 4 . 4 D u b i t a t i v e.......................... 1 1 4 3 2 1 . 5 P a s t c a t e g o r i e s ............................ 1 1 4 4 2 1 . 5 . 1 A o r i s t ............................ 1 1 4 4 2 1 . 5 . 2 I n f e r e n t i a l .......................... 1 1 5 3 21.5.3 Past Imperfective and Inferential Imperfective . . . . . . 1161 2 1 . 5 . 4 A r c h a i c f o r m ........................ 1 1 6 7 21.6 Secondar y Aspectual categories . . . . . . . . . . . . . . . . . . 1168 2 1 . 6 . 1 P r o g r e s s i v e ......................... 1 1 6 8 2 1 . 6 . 2 P r o x i m a t i v e ......................... 1 1 7 5 21.7 Secondar y Mo dal categories . . . . . . . . . . . . . . . . . . . . 1178 21.7.1 Apprehensive . . . . . . . . . . . . . . . . . . . . . . . . 1178 21.7.2 Probabilitative . . . . . . . . . . . . . . . . . . . . . . . . 1182 21.7.3 Rhetorical interrogativ e . . . . . . . . . . . . . . . . . . 1185 21.7.4 Interrogative . . . . . . . . . . . . . . . . . . . . . . . . 1189 21.7.5 Periphrastic modal constructions . . . . . . . . . . . . . 1190 21.7.6 Pseudo-concessive . . . . . . . . . . . . . . . . . . . . . 1191 21.8 Non-verbal T AME markers . . . . . . . . . . . . . . . . . . . . . 1192 21.8.1 T AME adverbs . . . . . . . . . . . . . . . . . . . . . . . 1192 21.8.2 Sentence-final particles . . . . . . . . . . . . . . . . . . . 1192 21.8.3 Nouns and the expression of modality . . . . . . . . . . 1192 22 Simple clauses 1197 2 2 . 1 W o r d o r d e r .............................. 1 1 9 7 22.1.1 Basic wor d order . . . . . . . . . . . . . . . . . . . . . . 1197 22.1.2 O v ert and non-ov ert arguments . . . . . . . . . . . . . . 1203 22.1.3 Right dislocation . . . . . . . . . . . . . . . . . . . . . . 1207 22.2 Sentential adverbs . . . . . . . . . . . . . . . . . . . . . . . . . . 1209 22.2.1 T ense and aspe ct . . . . . . . . . . . . . . . . . . . . . . 1210 22.2.2 Quantification . . . . . . . . . . . . . . . . . . . . . . . . 1213 2 2 . 2 . 3 I d e n t i t y ........................... 1 2 1 8 xvi Contents 22.2.4 Adverbial Intensifiers . . . . . . . . . . . . . . . . . . . . 1218 22.2.5 Epistemic modality . . . . . . . . . . . . . . . . . . . . . 1220 22.2.6 Orientation adverbs . . . . . . . . . . . . . . . . . . . . 1220 22.2.7 Postverbal elements . . . . . . . . . . . . . . . . . . . . 1223 22.3 Non-verbal predicates . . . . . . . . . . . . . . . . . . . . . . . . 1224 22.4 Noun-verb collocations and light verb constructions . . . . . . . 1226 22.4.1 Intransitive verbs . . . . . . . . . . . . . . . . . . . . . . 1226 22.4.2 Transitiv e verbs . . . . . . . . . . . . . . . . . . . . . . . 1233 22.4.3 Frozen collo cations . . . . . . . . . . . . . . . . . . . . . 1247 22.5 Copulas and existential verbs . . . . . . . . . . . . . . . . . . . . 1252 22.5.1 Basic functions . . . . . . . . . . . . . . . . . . . . . . . 1252 22.5.2 Possessive constructions . . . . . . . . . . . . . . . . . . 1256 22.5.3 Postverbal copulas . . . . . . . . . . . . . . . . . . . . . 1259 22.5.4 Postverbal negative existential verb . . . . . . . . . . . . 1263 22.5.5 V erb doubling . . . . . . . . . . . . . . . . . . . . . . . . 1266 22.5.6 Other constructions . . . . . . . . . . . . . . . . . . . . 1266 23 Relative clauses 1267 2 3 . 1 I n t r o d u c t i o n ............................. 1 2 6 7 23.2 Subtypes of relative cla uses . . . . . . . . . . . . . . . . . . . . . 1267 23.2.1 Participial relativ e clauses . . . . . . . . . . . . . . . . . 1267 23.2.2 Finite relativ e clauses . . . . . . . . . . . . . . . . . . . . 1268 23.2.3 Genitival relativ e . . . . . . . . . . . . . . . . . . . . . . 1269 23.2.4 Relator nouns . . . . . . . . . . . . . . . . . . . . . . . . 1270 23.2.5 Interrogative pr onouns and correlative constructions . . 1271 23.3 Morphosyntactic sp ecificities of relative clauses . . . . . . . . . 1273 23.3.1 Resumptive pronouns . . . . . . . . . . . . . . . . . . . 1273 23.3.2 T otalitative reduplication . . . . . . . . . . . . . . . . . 1274 23.3.3 Genitive p ossessor . . . . . . . . . . . . . . . . . . . . . 1275 23.3.4 Possessive prefix neutralization . . . . . . . . . . . . . . 1276 2 3 . 3 . 5 D e t e r m i n e r s ......................... 1 2 7 7 23.3.6 Postverbal elements . . . . . . . . . . . . . . . . . . . . 1279 23.4 Position of the relativized element . . . . . . . . . . . . . . . . . 1280 2 3 . 4 . 1 H e a d l e s s ........................... 1 2 8 1 2 3 . 4 . 2 P r e n o m i n a l ......................... 1 2 8 1 23.4.3 Head-internal . . . . . . . . . . . . . . . . . . . . . . . . 1281 2 3 . 4 . 4 P o s t n o m i n a l ......................... 1 2 8 4 23.5 Function of the relativized element . . . . . . . . . . . . . . . . 1285 23.5.1 Intransitive subje ct . . . . . . . . . . . . . . . . . . . . . 1285 xvii Contents 23.5.2 Transitiv e subject . . . . . . . . . . . . . . . . . . . . . . 1287 2 3 . 5 . 3 O b j e c t ............................ 1 2 8 8 23.5.4 Quasi-objects . . . . . . . . . . . . . . . . . . . . . . . . 1291 23.5.5 Goal and lo cative . . . . . . . . . . . . . . . . . . . . . . 1294 2 3 . 5 . 6 I n s t r u m e n t ......................... 1 2 9 7 2 3 . 5 . 7 C o m i t a t i v e .......................... 1 2 9 8 2 3 . 5 . 8 D a t i v e ............................ 1 2 9 9 23.5.9 Time adjuncts . . . . . . . . . . . . . . . . . . . . . . . . 1299 2 3 . 5 . 1 0 P o s s e s s o r .......................... 1 3 0 0 23.5.11 Relativization out of complement clause . . . . . . . . . 1304 23.5.12 Constraints on relativizability . . . . . . . . . . . . . . . 1308 23.6 Relative clauses and focalization . . . . . . . . . . . . . . . . . . 1308 23.6.1 Pseudo-cleft constructions . . . . . . . . . . . . . . . . . 1308 23.6.2 Non-equative pseudo-cleft . . . . . . . . . . . . . . . . . 1311 2 3 . 7 Q u a n t i fi c a t i o n ............................ 1 3 1 2 23.8 Relative vs. complement clauses . . . . . . . . . . . . . . . . . . 1314 23.8.1 Ambiguity (finite clauses) . . . . . . . . . . . . . . . . . 1315 23.8.2 Ambiguity (non-finite clauses) . . . . . . . . . . . . . . 1316 23.8.3 Participial clauses in core argument function . . . . . . 1317 23.8.4 Relativized complement clauses . . . . . . . . . . . . . . 1318 24 Complement clauses 1321 2 4 . 1 I n t r o d u c t i o n ............................. 1 3 2 1 2 4 . 2 C o m p l e m e n t t y p e s .......................... 1 3 2 1 24.2.1 V elar infinitive clauses . . . . . . . . . . . . . . . . . . . 1322 24.2.2 Bare infinitives and dental infinitives . . . . . . . . . . . 1324 24.2.3 Finite complements . . . . . . . . . . . . . . . . . . . . . 1326 24.2.4 Multiclausal complements . . . . . . . . . . . . . . . . . 1330 24.2.5 Reporte d sp eech . . . . . . . . . . . . . . . . . . . . . . 1332 24.3 Morphosyntactic properties of complement clauses . . . . . . . 1337 24.3.1 W ord order and constituency . . . . . . . . . . . . . . . 1337 2 4 . 3 . 2 C a s e m a r k i n g ........................ 1 3 3 7 2 4 . 3 . 3 D e t e r m i n e r s ......................... 1 3 3 9 24.3.4 Restrictive and additive fo cus . . . . . . . . . . . . . . . 1340 24.3.5 Raising of preverb orientation . . . . . . . . . . . . . . . 1340 24.4 Complementation strategies . . . . . . . . . . . . . . . . . . . . 1343 24.4.1 Relative clauses in core argument function . . . . . . . . 1343 24.4.2 Participial clauses . . . . . . . . . . . . . . . . . . . . . . 1344 24.4.3 Action nominals . . . . . . . . . . . . . . . . . . . . . . 1350 xviii Contents 24.4.4 Coordination . . . . . . . . . . . . . . . . . . . . . . . . 1354 24.5 Complement-taking verbs . . . . . . . . . . . . . . . . . . . . . 1355 24.5.1 Causative verbs . . . . . . . . . . . . . . . . . . . . . . . 1356 24.5.2 pa ‘ d o ’ ............................ 1 3 6 2 2 4 . 5 . 3 M o d a l v e r b s ......................... 1 3 6 2 24.5.4 V erbs of cognition . . . . . . . . . . . . . . . . . . . . . 1370 24.5.5 V erbs of perception . . . . . . . . . . . . . . . . . . . . . 1372 24.5.6 P hasal verbs and other aspectual auxiliaries . . . . . . . 1374 24.5.7 Similative verbs . . . . . . . . . . . . . . . . . . . . . . . 1380 24.5.8 Complement-taking adjectival verbs . . . . . . . . . . . 1381 24.6 Complement-taking nouns . . . . . . . . . . . . . . . . . . . . . 1383 24.6.1 Complement-taking nouns and denominal verbs . . . . 1383 24.6.2 Adnominal complement clause and p ossessor . . . . . . 1384 24.6.3 Over view of complement-taking nouns . . . . . . . . . 1385 24.6.4 Relative clauses as a complementation strategy . . . . . 1389 24.7 Syntactic errors . . . . . . . . . . . . . . . . . . . . . . . . . . . 1391 25 Other types of multiclausal constructions 1393 25.1 Classification of multiclausal constructions . . . . . . . . . . . . 1393 25.1.1 Marked subordinate clauses . . . . . . . . . . . . . . . . 1393 25.1.2 Correlative clauses . . . . . . . . . . . . . . . . . . . . . 1394 25.1.3 Periphrastic tenses and sub or dination . . . . . . . . . . 1395 25.1.4 Unmarked embe dded clauses . . . . . . . . . . . . . . . 1396 25.1.5 Serial verb constructions . . . . . . . . . . . . . . . . . . 1397 25.1.6 Coordination and parataxis . . . . . . . . . . . . . . . . 1397 25.1.7 T ail-head linkeage . . . . . . . . . . . . . . . . . . . . . 1398 25.2 Conditional construct ions . . . . . . . . . . . . . . . . . . . . . 1399 25.2.1 Real conditionals . . . . . . . . . . . . . . . . . . . . . . 1400 25.2.2 Necessar y condition . . . . . . . . . . . . . . . . . . . . 1402 25.2.3 Concessive conditional . . . . . . . . . . . . . . . . . . . 1402 25.2.4 Counterfactuals . . . . . . . . . . . . . . . . . . . . . . . 1407 25.3 T emporal clauses . . . . . . . . . . . . . . . . . . . . . . . . . . 1408 25.3.1 Iterative coincidence . . . . . . . . . . . . . . . . . . . . 1408 2 5 . 3 . 2 P r e c e d e n c e ......................... 1 4 0 9 2 5 . 3 . 3 S u b s e q u e n c e ........................ 1 4 1 1 2 5 . 3 . 4 C o n c u r r e n c e ........................ 1 4 1 4 2 5 . 4 M a n n e r c l a u s e s ........................... 1 4 1 7 25.4.1 Serial verb construction . . . . . . . . . . . . . . . . . . 1417 25.4.2 Manner converbs . . . . . . . . . . . . . . . . . . . . . . 1421 xix Abbreviations Abbre viations 1 first person 2 second person 3 third person [II], [III] stem II, stem III abil abilitative aca us anticausativ e add additive aff affirmative anthr anthroponym aor aorist ap ass antipassive appl applicative appr apprehensive a ut o autive a ux auxiliar y bare.inf bare infinitive ca t aph cataphoric ca us causative cisl cislocative cl counted noun coll collective comit comitative comp standard of comparison cond conditional conv converb da t dative deiph deideophonic dem demonstrative denom denominal dim diminutive distr distributed action du dual dub dubitative epenth epenthetic linking -j- emph emphatic erg ergative estim estimative f a cil facilitative f a ct factual non-past gen genitive genr generic ger gerund hort hortative hum human idph(X) pattern X ide ophone ifr inferential imm immediate imp imperative indef indefinite inf infinitive inf:II dental infinitive interj interjection inv inverse intr intransitive ipfv imperfective irr irrealis iter iterative coincidence lnk linker neg negative nmlz nominalization obj object obl oblique onom onomatopoeia p ass passive pcp participle peg peg circumfix pfv perfective pl plural poss possessive prob probabilitativ e prog progr essive prohib prohibitiv e xxvi prop proprietiv e pro x proximal pro xm proximativ e prs egophoric present pst past purp purposive qu question recip reciprocal refl reflexiv e rh.q rhetorical interrogative rog rogativ e sens sensory sfp sentence final particle sg singular simul t simultaneous (action nominal) st a t stativ e subj subject t op topic t opo toponym t ot al totalitative reduplication tr transitive tral translocative unexp.deg unexpe cted degree vert vertitiv e † XXX incorrect form *XXX reconstructed form X* Kleene star xxvii 1 Introduction 1.1 The Japhug language The Japhug language (lo cal name kɯrɯskɤt ) is spoken by several thousand speak- ers in Mbarkhams county (in Chinese Maerkang 马尔康 ), Rngaba prefecture, Si- chuan province , China, in the 乡 <xiāng> ‘townships’ of ʁdɯrɟɤt ( གདོང་བད་ gdoŋ.- brgʲad , 龙尔甲 lóng'ěrjiǎ ), sarndzu ( གསར་ ོང་ gsar.rdzoŋ , 沙尔宗 shāěrzōng ) and tatsʰi ( ད་ ཚང་ da.tsʰaŋ , 大 藏 dàzàng ), colle ctiv ely calle d tɕɤpʰɯ or tɕʰɤpʰɯ (fr om which the name Japhug is taken). 1 All speakers of Japhug are classified as ethnic Tibetans ( 藏族 zàngzú ). 2 This work focuses almost exclusively on the dialect of kɤmɲɯ village ( 干木鸟 gānmùniǎo , 32° 12′ 43″ N, 101° 57′ 42″ E), henceforth written as Kamnyu. Japhug is one of the four Core Gyalrong languages (Sun 2000a; Gong 2018; Zhang 2020; Sangsrgyas T shering 2025), T shobdun, Zbu and Situ, represented in Figure 1.1. It is particularly closely related to the W est Gyalrongic languages: Khroskyabs, Stau, Phosi, Geshiza, Nyagrong Minyag, T angut (Sun 2000b; Jacques et al. 2017; Honkasalo 2019; Lai et al. 2020; Liu 2023; 2025) and more remotely to the Smar languages spoken in Tib et (T ashi Nyima & Suzuki 2019; Zhao 2024). Core Gyalrong and W est Gyalrongic form the Gyalrongic gr oup (Sun 2000b), itself a subbranch of Burmo-Gyalrongic (Jacques & Michaud 2011) and possibly the larger Tibeto-Gyalrongic group (Sagart et al. 2019). Japhug is in contact with both Standard Mandarin and a local variety of Si- chuanese. Since no study has focused on the Sichuan Mandarin as spoken by Gyalrong speakers in Mbarkham, the present w ork represents all Chinese wor ds in Standard Mandarin and pinyin (b etw een chevrons), except for highly nativized Chinese words, presented in IP A. 1 The township of Gdongbrgyad is how ever not traditionally included under the name tɕɤpʰɯ , and is referred instead by the term sɤŋu . 2 The language can be referred to as 茶堡话 chápùhuà in Chinese, reading the character 堡 as pù rather than the more usual bǎo . In W estern languages, the ‘ph’ should be pronounced as a stop rather than labiodental fricative, and the ‘g’ read as a v oiceless stop. In French for instance, I call the language [dʒapuk] . 1 Introduction 30.5 31.0 31.5 32.0 32.5 33.0 101 102 103 104 Language Amdo Hor pa J aphug Khams Khrosky abs Mandarin Qiang Situ Tshobdun Y elong Zb u Figure 1.1: A map of Gyalrong languages Situ Gyalrong and Amdo Tibetan used to b e the tw o dominant languages in the area, and many speakers of Japhug also have a passive understanding of Situ, and sometimes also of T shob dun and Zbu. Ther e is no immediate intercompre- hensibility between Gyalrong languages, and speakers of Japhug reporte dly need several months to acquire basic conversational skills in T shobdun. 3 1.2 The Japhug corpus With the exception of a chapter in Lin (1993: 468–486) on phonology and a few articles by Lin Y oujing (Lin & Luo erwu 2003; Lin 2011) (all on the T atshi dialect), the available data on Japhug are essentially fr om my own fieldwork, based on nine trips ( July- A ugust 2002, April-July 2003, July- A ugust 2005, July 2010, July- A ugust 2012, April-May 2014, July- A ugust 2015, July- A ugust 2016, May 2018) and constant contact by phone with my main language consultants. Some texts have be en r e cor ded with the help of my former PhD students Gong Xun and Lai Y un- fan. 3 See for example (145) (§8.2.9), (35) (§23.3.6) and (126) (§24.5.3.1), where T shendzin recounts her experience as a teacher in T shobdun village, wher e she had to learn T shobdun to make herself understood by the pupils’ parents. 2 1.2 The Japhug corpus A short grammar (Jacques 2008), some texts (Jacques & Chenzhen 2010) and a dictionary (Jacques 2015b, Bonnet et al. 2017) have been previously published. A corpus of transcribe d stories on the Pangloss ar chive (Michailovsky et al. 2014; Michaud et al. 2016) is available on https://pangloss.cnrs.fr/corpus/Japhug. An update of the dictionar y is in preparation (Galliot 2025). My main consultant is T shendzin (Chenzhen 陈珍 , female, born 1950), a retir e d schoolteacher (a native speaker of Japhug, bilingual in Sichuan Mandarin since childhood), whose sp eech and grammaticality judgements are taken as the norm of this grammar . Stories have also been colle cted from her husband χpɤltɕin Dpalcan 柏 尔 青 , her maternal uncle Andzin, her nephew ( sister’s child) A yang/Kunbzang Mtsho, and other Kamnyu people, in particular T shering Skyid. Most traditional stories are typical of Tibetan folklore and not specific to Japhug-speaking areas, but some procedural texts, descriptions of local plants and animals, as well as ac- counts of local topography , include unique insight into lo cal cultur e. In addition, a few conversations hav e b een recorded. The majority of examples in this grammar ( except in the grammar sketch in chapter 2) are taken from te xts or conversations rather than elicited. Most of the texts cited are already available on Pangloss. T emp oral synchronization between transcription and audio was automatically accomplished using forced alignment (Macaire 2020), and whenev er available, a hyperlink to the doi of the segment (V asile et al. 2020) is pro vided for each example sentence in the grammar . 4 In addition, I indicate for all examples a title in ASCII characters: Hill (2017: 305) reports that this citation format is sufficient to find the relevant data. Conversations ar e not include d in the ar chive at the moment of writing, but individual sound files of the sentences quoted in this grammar are preserved in a private database. In addition, a certain number of sentences have been note d down during participant observation, and lack recordings. Since speakers I have worked with only recount a limited number of tradi- tional stories (some of which in any case are fr om written sources; in particular , the story ‘The fox’ {0004087} told by Kunbzang Mtsho is clearly adapted from the Grimm fairy tale ‘Der goldene V ogel’), I have resorted to translation from Chi- nese to collect a larger corpus. T shendzin pr ovided surprisingly idiomatic render- ings of various stor ybooks for children (including Grimm tales, Andersen tales, Liaozhai zhiyi 聊斋志异 , Arabian nights and Xiyouji 西游记 ). These documents 4 The doi were semi-automatically added using the program DOI-Mer cy written by Alexander Delaporte. Only the informative part of the URL is indicated. T o obtain the complete URL, the string https://doi.org/10.24397/pangloss- has to be prefixed to it. 3 1 Introduction are not given the same value as mor e spontane ous texts, and ar e systematically indicated by adding the extension “-zh” to the document name. Systematic com- parison with the original text is offered whenev er any suspicion of calque from Chinese exists. How ever , the immense morphosyntactic difference b etw een Chi- nese and Japhug makes it necessar y to completely re work the structure of the sentences in most cases (see for instance 265, §21.7 .1.2), so that even if there is un- doubtedly influence from Chinese, e ven the way Chinese is adapted into Japhug is itself an interesting topic of resear ch. A system of automatic spee ch recognition has been devised for Japhug (Wis- niewski et al. 2020; Guillaume et al. 2022), which makes it possible to enrich the Japhug corpus more effectively with less manual labor . 1.3 Structure of the grammar Rather than being intende d as the final w ord on Japhug grammar , this book is conceived as a tool for exploring the Japhug corpus and learning the language. The reader is invited to start with the grammar sketch ( chapter 2). The rest of the grammar contains two chapters on phonology (3, §4), five on nouns and noun phrases (5 to 9), eleven on v erbal morphology (11 to 21), five on syntax (§22 to 26), as well as a chapter 10 on expr essives and another one 27 on kinship. In writing this grammar I have always preferr ed to abstain from definite judge- ment rather than provide incorr ect data, and the description remains thus incom- plete in many aspects, in particular in accounting for minute semantic differences between similar constructions, or on the grammaticality of b orderline sentences (about which speakers sometimes change their mind). There are many points of uncertainty in many aspects of the morphosyntax (and even in some topics of phonology ), and further research on finer points of phonology , morphosyntax and historical linguistics are much needed, with additional data from different speakers. 1.4 Language and culture This work focuses on the grammar of the language, and only contains a single chapter (27, on the kinship system) dedicate d to ethnology . Ho wever , the corpus that has be en collected contains many traditional stories, as well as procedural texts describing the traditional life before massive sinicization. Examples from these texts are found on nearly all the pages of this grammar , and their interest 4 1.4 Language and culture is not purely linguistic. Thus, the section on place names presents a mythologi- cal story explaining the origin of a name (79, §5.2.1), the chapter on orientation prev erbs contains an account of the traditional living-room/ kitchen (§15.1.4.4) and an introduction to weaving (§15.1.4.5, Jacques et al. 2025), and the section on antipassive derivations includes a passage on rain-calling and mountain tutelary spirits (see example 161, §18. 6.7 .3). In this sense , traditional culture permeates this grammar from beginning till end. Further ethnographical work is necessary , in particular on kinship (Zhang & Fan 2020), traditional houses (Dong 2018), weaving (Jacques et al. 2025) and eth- nobotany , and it is hope d that this grammar (together with the dictionary) will make it possible to properly describe lo cal culture thr ough the me dium of Japhug rather than Chinese. 5 2 A grammatical sketch This chapter offers a short introduction to the phonology and morphosyntax of the Japhug language. It is written for typologists and comparative linguists wanting to get a quick general picture of this language , b efor e delving into the core of the grammar . Morphosyntactic phenomena are illustrated mainly using simplified elicited examples rather than examples gleaned from texts and conv ersations as in the rest of the grammar . 2.1 P honolog y and word structure Japhug has 8 vow els (§3.3.1) and 50 consonant phonemes (§3.2.1), which can b e combined into more than 400 biconsonantal or triconsonantal clusters in the on- set (§4.2). Additional clusters ar e attested across syllable b oundaries (§4.2.3.1). In coda position, only 12 consonants are found, and no clusters ar e possible: several phonological contrasts are neutralized (§3.2.2). The IP A -base d transcription in this grammar is a spelling system that is not strictly phonological: it uses different symbols to represent the allophones of some phonemes (in particular /w/ , §3.2.1, §4.2.1.1). An alternative Tibetan-based orthography for use by native speakers (§4.4) is also provided. Japhug is very far removed fr om the isolating, tonal and “monosyllabic” typ e once considered to be typical of Sino- Tib etan languages. Japhug lacks tonal con- trasts (§3.7), and monosyllabic and monomorphemic wor ds are a minority (§11.6). W ords of six syllables or more are not rare in the corpus, as sho wn by the verb form ( 1) below . (1) a-kɤ-tɯ-nɯ-rɤʑí-nɯ irr-pfv-2-a ut o-stay-pl ‘May you stay (here). ’ Non-final syllables have str ong phonotactic constraints at least in the native vocabulary (§3.8.2). The last syllable of verbal and nominal stems generally re- ceives stress ( as the syllable -ʑi- in 1, §3.7) and allows the maximal number of vow el and consonant contrasts. 2 A grammatical sketch 2.3.4.3 Generic possessors The indefinite possessor prefix has three allomorphs tɯ- (as in tɯ-ku ‘head’), tɤ- (as in tɤ-se ‘blood’) or ta- (as in ta-ma ‘work’). It has to be distinguished from the generic possessor prefix tɯ- (§5.1.3). Generic possessors are identical with indefinite possessors in the case of inalienably possesse d nouns selecting the tɯ- prefix, for instance tɯ-ku can either mean ‘head’ or ‘ one ’s head’ . With inalienably possessed nouns sele cting the tɤ- or ta- allomorphs, a contrast is found between tɤ-se ‘blood’ and tɯ-se ‘one ’s blood’ (11) for instance. (11) qajɯsmɤnba leech kɯ erg tɯ-se genr.poss-blood ku-tsʰi ipfv-drink ŋu be:f act ‘The leech drinks p eople ’s (i.e. one ’s) blood. ’ The generic possessor prefix can also occur on alienably possesse d nouns, as in tɯ-kʰa ‘ one ’s house ’ . No more than one generic refer ent is possible p er clause , so that if a noun with generic possessor prefix is found in the same clause as a verb with generic indexation, there is obligatory co-reference (§14.3.2.5), as in (12) between the pos- sessor of tɯ-rpɯ ‘ one ’s mother’s brother’ (indefinite form tɤ-rpɯ , §5.1.2.4) and the transitive subject of the verb tu-kɯ-ti ‘ one says’ (§14.3.4). (12) tɯ -rpɯ genr .poss-mother’s.brother ɣɯ gen ɯ-rʑaβ 3sg.poss-wife ɯ-ɕki 3sg.poss-da t tɕe loc “a-ɬaʁ” 1sg.poss-aunt tu- kɯ -ti ipfv- genr -say ŋu be:f act ‘ One 𝑖 calls one 𝑖 ’s mother’s brother’s wife ‘my aunt’ (i.e . one says ‘my aunt’ to one ’s mother’s brother’s wife). ’ 2.3.4.4 Possessive existential construction Predicative possession can be expressed by the verb aro ‘ own’ (§14.2.3, §22.5.2), encoding the possessor as subje ct, but the most frequent construction involv es an existential verb (§22.5.1.2) with the possessum as subject and the possessor marked by a possessive prefix on the possessum, optionally with a genitive phrase (§22.5.2.1). The construction is the same for alienably (13a) and inalienably (13b) possessed nouns. (13) a. kɯβʁa noble ra pl ɣɯ gen nɯ-nɯŋa 3pl.poss-cow kɯ-dɤn sbj:pcp-be.many pjɤ-tu ifr.ipfv-exist ‘The nobles had many cows. ’ 14 2.3 Nominal morphology b . kɯβʁa noble ra pl ɣɯ gen nɯ-tɕɯ 3pl.poss-son χsɯm three pjɤ-tu ifr.ipfv-exist ‘The nobles had three sons. ’ In this construction, the existential verbs rar ely agree in number with the p os- sessum: in both (13a) and (13b), the singular verb form pjɤ-tu is by far more com- monly used than its plural counterpart (§14.6.1.1). 2.3.5 Compounding Compounding can b e r ealize d by the simple concatenation of noun ( or verb) stems. In compounds comprising two noun stems (§5.5.1), whose order is modifier- modified, no p ossessive pr efixes occur b etween the nominal stems, ev en if the second noun is inalienably p ossessed. For instance, the compound built from kɯrɯ ‘Tibetan’ and the inalienably possesse d tɯ-ŋga ‘ clothes’ is kɯrɯ-ŋga ‘Ti- betan clothes’ , rather than † kɯrɯ-tɯŋga . In many cases, the non-final elements of the compound appear in a b ound form, the bound state (§5.4), which is characterize d by a vocalic change to either -ɤ or -ɯ (§5.4.1) For instance, compounding tɯ-ku ‘head’ with tɤ-rme ‘hair’ yields tɤ-kɤ-rme ‘head hair’ with the bound state kɤ- from -ku . The compound inherits the allomorph tɤ- of the indefinite possessor prefix of the head of the compound tɤ-rme ‘hair’ . 2.3.6 Derivations Some suffixal nominal derivations come from compounds with a grammatical- ized noun as second element, in particular the diminutive suffixes (§5.7 .4). The recent origin of these suffixes can be shown by the fact that a free form still co- exists with the corr esp onding compound in some cases. For instance, the com- pound χpɯn-pɯ ‘little monk’ from χpɯn ‘monk’ with the diminutive -pɯ occurs in free variation with the phrase (14), in which the main noun is followed by the alienably possessed property noun ɯ-pɯ ‘little one’ (§ 5.1.2.7) deriv ed from tɤ-pɯ ‘ offspring, young’ . (14) χpɯn monk ɯ-pɯ 3sg.poss-little.one ‘(The/a) little monk’ 15 2 A grammatical sketch Not all suffixal derivations have transparent origins. For instance , the privative -lu suffix (§5.7 .1) is not related to any indep endently attested nominal or verbal root. It turns an inalienably possessed noun into a non-p ossessible one mainly used as postnominal modifier (§9.1.8.1), removing possessive pr efixes and sub- jecting the nominal stem to b ound state alternation ( tɯ-ku ‘head’ → kɤ-lu ‘head- less’). Prefixal nominal derivations, how ever , do not originate from elements of com- pounds, but rather from participial forms of denominal verbs (§20.2.4, §20.2.5). The social relation collective kɤndʑi- prefix (§5.7 .9.1) occurs on the bare stem of kinship terms and a few other terms of social relationship to indicate a group of people. When the group members are related to each other by a symmetrical relationship ( tɤ-xtɤɣ ‘br other’ (of a male) → kɤndʑi-xtɤɣ ‘group of br others’), the social relation collective is based on only one nominal stem, but when the relationship is asymmetrical, kɤndʑi- can be prefixed to two compounded noun stems, as in kɤndʑi-wɤmɯ-snom ‘group of siblings’ from tɤ-wɤmɯ ‘brother’ ( of a female) and tɤ-snom ‘sister’ (of a male). The comitative derivation (§5.8.1) is built by adding the prefixes kɤ  - or kɤɣɯ- to reduplicated noun stems. It derives an adverb meaning ‘together with 𝑋 ’ which can have scope over the whole clause , or b e r estricted to a noun phrase. It can apply to both inalienably p ossessed nouns ( tɯ-ŋga ‘ clothes’ → kɤ  -ŋgɯ~ŋga ‘to- gether with his/ her clothes’) or alienably possesse d ones ( jla ‘male hybrid yak’ → kɤ  -jlɯ~jla ‘together with the/his/her hybrid yak’). Comitative derivation can preserve the indefinite possessor prefix of inalienably possessed nouns, causing alienabilization (§2.3.4.2, §5.1.2.9). For instance, tɤ-rte ‘hat’ has two comitativ e forms: kɤ  -rtɯ~rte ‘together with his/ her hat’ ( wearing it) and kɤ  -tɤ-rtɯ~rte ‘together with a/the hat’ (not wearing it). 2.4 V erbal morphology 2.4.1 Over vie w V erbs have a considerably more elaborate morphology than all other parts of spee ch (§11.1). V erbal morphology is strongly prefixal (§11.2), with some vo wel contractions (§12.3). Non-concatenative morphology includes stem alternations (§12.2) as well as infixation (§11.2.2, §14.2.2, §19.1.2). Japhug verbal morphology is considerably mor e regular than that of other Gyalrong languages, in particular Zbu (Sun 2004; Gong 2018) and Situ (Zhang 2018). Most alternations are productiv e and predictable, and irregular v erbs are limited in numb er (§14.2.2, §14.3.4). 16 2.4 V erbal morphology As illustrated by (15), a verb form can comprise six inflectional prefixes (in blue) arranged in a rigid template (§11.2.1) and several derivational prefixes (in red). (15) a −6 -mɤ −5 -ɣɯ −4 -nɯ −3 -tɯ  −2 -wɣ −1 - z-nɤ -re irr −6 -neg −5 -cisl −4 -pfv −3 -2 −2 -inv −1 - ca us-denom -laughter ‘Don’t let him come and make you laugh. ’ The suffixal chain (§11.3) only includes inflectional suffixes, with a maximal number of four slots (16). (16) ɯmɤ −6 -pɯ −3 -kɯ −2 -mto- t +1 -a +2 -ndʑi +3 -ci +4 prob −6 -aor −3 -peg −2 -see- pst:tr +1 -1sg +2 -du +3 -peg +4 ‘It looks like I have seen the two of them. ’ Numerous non-adjacent dependencies (§11.5) are observed across the prefixal and the suffixal chains. Inflectional verbal morphology encodes person and numb er of one or two core arguments ( chapter 14, §2.4.2.2), orientation (§15.1), asso ciated motion (§15.2), negation (§13.1) and T ense- Aspect-Mo dality-Evidentiality ( chapter 21). 2.4.2 Indexation All finite verb forms in Japhug have obligatory person indexation. Since the 3sg has zero marking as in many languages of the world (Benv eniste 1966b: 227 – 236), indexation is not conspicuous on intransitive v erbs requiring a 3sg subje ct (§14.2.7), but indirectly obser vable e ven on transitive dummy verbs (§14.3.5) due to the presence of stem alternation (§12.2.2). T ransitive and intransitive v erbs are clearly distinguished by a series of seven morphological parameters (§14.3.1). Intransitive verbs only index one argument (the intransitive subject, S), and transitive verbs inde x two arguments (the tran- sitive subject A and the object O). Semi-transitive verbs hav e intransitive indexa- tion (§14.2.3), but select a second core argument (§8.1.5). Only a handful of verbs are labile , and can b e conjugated either transitively or intransitiv ely (§14.5). 2.4.2.1 The intransitive paradigm Person indexation in Japhug is best introduced with the intransitive paradigm (§14.2.1), since it is considerably smaller than the transitive one. The r egular pa- radigm is illustrated in T able 2.3 with the verb mbɣom ‘be in a hurr y’ in the Factual Non-Past, the only T AME without orientation prev erb (§21.3.1). 17 2 A grammatical sketch T able 2.3: The intransitive indexation paradigm Person Form Example 1sg Σ - a mbɣom-a 1du Σ - tɕi mbɣom-tɕi 1pl Σ - ji mbɣom-i 2sg tɯ - Σ tɯ-mbɣom 2du tɯ - Σ - ndʑi tɯ-mbɣom-ndʑi 2pl tɯ - Σ - nɯ tɯ-mbɣom-nɯ 3sg Σ mbɣom 3du Σ - ndʑi mbɣom-ndʑi 3pl Σ - nɯ mbɣom-nɯ The intransitive paradigm has different forms for singular , dual and plural. First persons have dedicated suffixes enco ding both person and numb er; there is no inclusive/exclusiv e contrast. Second and third person forms are distinguished by the second person tɯ- prefix (on the historical significance of this pr efix, se e Jacques 2012a and DeLancey 2014). There is no o vert third person marker on intransitive verbs. Number markers are shar ed by second and third person forms: absence of suffix for the singular , -ndʑi for the dual and -nɯ for the plural. Slightly different forms ar e found in diale cts of Japhug other than K amnyu (§14.8.1). In addition to the paradigm in T able 2.3, a generic person kɯ- prefix also occurs on intransitive verbs (§14.3.2.5). A handful of intransitive verbs infix rather than prefix the second person (§14.2.2). It is the only irregularity related to person indexation in intransitive verbs in Japhug. 2.4.2.2 The transitive paradigm The transitive paradigm is too large to be describ ed in this introductory chapter in its entirety (§14.3.2). T able 2.4 presents the singular forms of the paradigm of the transitive verb sat ‘kill’ ( a verb lacking stem alternations) in the Factual Non-Past, which are sufficient to illustrate its basic structur e. In this table, the columns represent the objects (O ), and the rows the subjects ( A). The shaded cells indicate configurations with corefer ent subje ct and object, which are e xpressed by the reflexiv e derivation (§18.3) and do not belong to the transitive paradigm. 18 2.4 V erbal morphology T able 2.4: The transitive paradigm (singular forms) of the non- alternating verb sat ‘kill’ in the Factual Non-Past 1O 2O 3O 3 ′ O 1A ta- sat sat -a 2A kɯ- sat -a tɯ- sat 3A ɣɯ  - sat -a tɯ  -wɣ- sat sat 3 ′ A ɣɯ  - sat The comparison of T ables 2.3 and 2.4 shows that the 1sg -a suffix and the second person tɯ- prefix have neutral alignment: they can index intransitiv e subject, transitive subject or object. In the following, the person configurations are referr ed to by 𝑋 → 𝑌 , wher e 𝑋 represents the subject and 𝑌 the object (for instance 1 → 3 means ‘first person subje ct, thir d p erson object’). The eight non- shaded cells of the paradigm in T able 2.4 can b e divided into three groups. First, 1 → 3, 2 → 3 and 3 → 3 ′ are the direct configurations with a third person object, whose forms resemble the 1, 2 and 3 forms of the intransitive paradigm (at least in the Factual Non-Past). Second, 3 → 1, 3 → 2 and 3 ′ → 3 are the inverse configurations, which have the same prefixes or suffixes as the corresponding intransitive and dir ect forms (at least in this paradigm), but take in addition the inverse prefix ɣɯ- / wɣ- (the allo- morphy of this prefix is explained in §14.3.2.7). Third, 1 → 2 and 2 → 1 ( without third person) are the local configurations (§14.3.2.3). They are characterized by the presence of the portmanteau ta- and kɯ- prefixes (§14.8.3) not found in the intransitive paradigm. The reasons for using the terms “ direct” and “inv erse” to describe the transitive paradigm are discussed in §14.3.2.8. When both arguments are third person (§14.3.2.2), there is a contrast between direct 3 → 3 ′ and inverse 3 ′ → 3 configurations, whose meaning is not entir ely straightforward (§14.3.3). The subject of the direct configuration, and obje ct of the inverse one is called pro ximate (3), and the other argument obviative (3 ′ ). The direct 3 → 3 ′ configuration is by far the most common one in narratives and con- versation. The inverse 3 ′ → 3 is mor e restricted; it occurs in particular to index a generic subject with a third person obje ct (§14.3.2.5), and also when the subject is inanimate and the object animate (§14.3.3.1). The majority of transitive verbs ( all verbs ending in closed syllables or with front vo wels) are non-alternating like sat ‘kill’ . Howe ver , about a third of all transitive verbs (those ending in -a , -u , -o and -ɯ ) hav e stem alternation in the 19 2 A grammatical sketch direct configurations with a singular subject in the Factual Non-Past and a few other tenses (§12.2.2). T able 2.5: The transitive paradigm (singular forms) of the alternating verb ʁndɯ ‘hit’ in the Factual Non-Past 1O 2O 3O 3 ′ O 1A ta- ʁndɯ ʁndi -a 2A kɯ- ʁndɯ -a tɯ- ʁndi 3A ɣɯ  - ʁndɯ -a tɯ  -wɣ- ʁndɯ ʁndi 3 ′ A ɣɯ  - ʁndɯ T able 2.5 illustrates the singular forms of the alternating verb ʁndɯ ‘hit’ , which has additional stem -ʁndi in the 1sg → 3, 2sg → 3 and 3sg → 3 ′ configurations (with blue colouring), which are thus different fr om the corresponding intransitive forms. In the Aorist (§21.5.1.1), a slightly differ ent paradigm is found, illustrated in T able 2.6 with the ʁndɯ ‘hit’ . Unlike the Factual Non-Past shown in the pre vious tables, the Aorist r equires an orientation prev erb (§15.1.1.1), here the upw ards tɤ- . There is no stem alternation marking the direct forms, but all verbs hav e a different series of pre verbs (here ta- ) in the direct 3 → 3 ′ forms (§21.5.1.1, §15.1.1.1). In addition, the 1sg → 3 and 2sg → 3 forms require a -t suffix which redundantly encodes b oth person and tense-aspect (§11.3). T able 2.6: The transitive paradigm (singular forms) of the v erb ʁndɯ ‘hit’ in the Aorist 1O 2O 3O 3 ′ O 1A tɤ-ta- ʁndɯ tɤ- ʁndɯ - t -a 2A tɤ-kɯ- ʁndɯ -a tɤ-tɯ- ʁndɯ - t 3A tɤ  -wɣ- ʁndɯ -a tɤ-tɯ  -wɣ- ʁndɯ ta - ʁndɯ 3 ′ A tɤ  -wɣ- ʁndɯ There is no ambiguity in person indexation in Japhug (unlike for instance in Khaling where 2 → 1 and 3 → 1 configurations are identical, Jacques et al. 2012), but there are str ong restrictions on number indexation: unless the 1sg suffix is present, only one of the two arguments can be indexed for both p erson and 20 2.4 V erbal morphology number , the subject in direct configurations, and the object in inverse and local configurations. Double number indexation only occurs in forms with the 1sg -a suffix, to which additional number suffixes can b e added (§ 14.3.2.6), for instance ɣɯ  -ʁndɯ-a-nɯ ‘they will hit me ’ (3pl → 1sg) where the plural morpheme -nɯ , indexing the number of the subject, follows the 1sg. 2.4.2.3 Person indexation and finiteness Person indexation markers (including person-indexing stem alternation and ori- entation prev erbs) are not found on participles, infinitives and other non-finite verb forms ( chapter 16). A handful of phatic and exclamative w ords of nominal origin have how ever de veloped the ability to take numb er suffixes (§2.8.3, §14.7). Apart from these , words belonging to parts of spe ech other than verbs are incom- patible with the indexation affixes described in this se ction. 2.4.3 Orientation preverbs and T AME This section focuses on morphology . Since the use of the T AME categories in- volv e sometimes subtle semantic nuances, and have to be explained on the basis of examples with a clear context, the discussion of the semantic function of each category is deferred to chapter 21. 2.4.3.1 The morphology of orientation preverbs The main morphological exponents of tense, aspect, modality and evidentiality (henceforth T AME) in Japhug ar e the orientation preverbs (§15.1). All r egular finite verb forms require one and only one preverb (§11.2.1), e xcept the Factual Non-Past (§21.3.1) which does not take any preverb . The stacking of tw o or more prev erbs is ungrammatical. Only a handful of irregular defective v erbs are in- compatible with orientation prev erbs (§14.2.2, §14.3.4). Prev erbs encode one out of seven orientations (T able 2.7), divided into three dimensions: vertical (§15.1.3.1), riverine (§15.1.3.2) and solar (§15.1.3.3), to which an unspecifie d orientation is added. This tridimensional system is not restricted to verbal morphology: locative relator nouns (§8.3.4.1), egressiv e postp ositions (§8.2.10) and locative adverbs (§15.1.1.4) have similar systems with six orienta- tions, built from morphemes that are historically related to the prev erbs. There are four series of pre verbs (T able 2.7 includes two of them, the series A and B) in the Kamnyu dialect, used in different T AME categories (§15.1.1.1). Some dialects of Japhug have a slightly different system (§15.1.1.3). 21 2 A grammatical sketch T able 2.7: Orientation preverbs in K amnyu Japhug Dimension Orientation A B V ertical Up tɤ- tu- Down pɯ- pjɯ- Riverine Upstream lɤ- lu- Downstream tʰɯ- cʰɯ- Solar Eastwards kɤ- ku- W estwards nɯ- ɲɯ- Unspecifie d jɤ- ju- Orientable verbs (§15.1.2) ar e compatible with all orientations; this includes in particular motion verbs like ɣi ‘ come ’ and ɬoʁ ‘ come out’ (§15.1.2.1). With this type of verbs, the pre verbs indicate either the absolute direction of the motion (§15.1.3), for instance upw ards in (17) or have extended meanings (§15.1.4), such as the illative function (§15.1.4.2) of the upstream prev erb in (18). (17) tɤŋe sun tɤ-ɬoʁ aor:up-come .up ‘The sun rose . ’ (18) ɯ-ŋgɯ 3sg.poss-in lɤ-ɣi imp:upstream-come ‘Come in! (for instance, inside a house)’ {0003884#S90} Non-orientable verbs only select a restricted number of lexically determined orientations, sometimes only one (§15.1.5). For instance, the v erb ndza ‘ eat’ and mto ‘see ’ require the upw ards (§15.1.5.4) and do wnw ards (§15.1.5.9) pr everbs, respectively . 2.4.3.2 The mor phology of T AME categories There are ele ven primary T AME categories, which can be divided into four main groups: Non-Past (Factual Non-Past, Egophoric Present, Sensor y , §21.3), Imper- fective (§21.2), Past ( Aorist, Inferential, Past Imperfective , and Inferential Imper- fective, §21.5) and Modal (Irrealis, Imperative , Prohibitive , Dubitative, §21.4) cat- egories. 22 2.4 V erbal morphology In the finite T AME categories, the B-type preverbs ar e found in the Imper- fective (§21.2) and the A -type pre verbs in the Imperative (§21.4.2.1), the Irrealis (§21.4.1.1) and the Aorist (§21.5.1), though with slightly differ ent vowel contrac- tion rules (§12.3). In the Aorist paradigm of transitiv e verbs, another series of prev erbs (C) is found in the direct 3 → 3 ′ forms (§2.4.2.2), based on the A -type prev erbs but with -a vocalism instead of -ɯ and -ɤ (originating from fusion with another prefix, §15.1.1.3). For instance , the 2pl → 3 Aorist of ndza ‘ eat’ is tɤ-tɯ- ndza-nɯ (aor:up-2-eat-pl ‘y ou 𝑝 𝑙 ate it’) with the upw ards A -type tɤ- preverb , but the corresponding 3pl → 3 ′ form is ta -ndza-nɯ ( aor:3 → 3 ′ :up -eat-pl ‘they ate it’) with the C-type prev erb ta- . It is the only case when an orientation pre- verb encodes person in addition to T AME. The Inferential (§21.5.2) has a series of pre verbs (series D ) based on series B, but with -o and -ɤ vocalism instead of -u and -ɯ , respectively (§15.1.1.3). For instance, the upw ards and downw ards D-type pre verbs are to- and pjɤ- , corresponding to the B-type tu- and pjɯ- , respectively (§15.1.1.1). Five T AME categories neutralize the orientation contrast, and requir e the same marker for all verbs: the Sensory evidential ɲɯ- (§21.3.2), from the B-type west - w ards preverb (T able 2.7), the Egophoric Present ku- (§21.3.3), the Dubitativ e ku- (§21.4.4), from the B-type eastw ards preverb , the Past Imperfective pɯ- from the A -type downw ards prev erb (§21.5.3.1, Lin 2011), and the Inferential Imperfective pjɤ- from the D-type downw ards prev erb (§21.5.3.1). In addition to orientation prev erbs, T AME categories are marked by several morphological exponents, including stem alternations (§12.2), allomorphy of ne- gative prefixes (§13.1.1) and additional affixes: the Irrealis a- prefix (§11.2.1) and the Past transitive -t suffix (§11.3). In the Non-Past, Imperfe ctiv e and Mo dal categories ( all except Past), transitive alternating verbs have a specific stem in direct configurations with a singular subject (see T able 2.5 ab ov e and §12.2.2). Another stem is found in the Aorist of a handful of verbs (§12.2.1). Some of the primary categories can b e combined with the copula ŋu ‘be’ to form periphrastic T AME categories (§21.2.2), for instance the Periphrastic Past Imperfective (§21.5.3.5) illustrated in (19), built from the Imperfective ( tu-ndze-a , with the B-type upw ards pr everb tu- , the alternating stem ndze from ndza ‘ eat’ and the 1sg suffix) and the Past Imperfective of the copula pɯ-ŋu . (19) tɤ-mtʰɯm indef.poss-meat tu-ndze-a ipfv-eat[III]-1sg pɯ-ŋu pst.ipfv-be ‘I was eating meat/I used to eat meat. ’ 23 2 A grammatical sketch The sigmatic causative is also used to mark instruments (§17 .2.5.8), for example in (29), where the instrument taqaβ ‘needle’ r eceives ergative marking (§8.2.2.4) like a causee (§8.2.2.6): the construction literally means ‘s/ he made the nee dle sew the clothes’ . (29) ki dem.pro x taqaβ needle ki dem.pro x kɯ erg tɯ-ŋga indef.poss-clothes cʰɤ- sɯ -tʂɯβ ifr-ca us-sew ‘S/he sewed the clothes with this needle. ’ In addition to the productive causatives, ther e are irregular causative forms (§17 .2.2, §17 .3.1), some of which co-exist with their r egular counterparts, but with a more lexicalized meaning. For instance , the verb tsʰi ‘ drink’ has the irregular causative jtsʰi ‘give to drink’ (§17 .2.2.5) as opposed to the regular one sɯ-tsʰi ‘make drink, drink with’ . The sigmatic causative pr efix can b e combined with nearly all other deriva- tional prefixes (§17 .2.8). It can precede the velar causative, as in z-ɣɤ-mpja ‘heat up 𝑋 with 𝑌 , make/ let 𝑌 heat up 𝑋 ’ from ɣɤ-mpja ‘heat up ’ , causative of mpja ‘be warm’ . It is also the only prefix that can occur more than once in a single verb form, 5 as shown by examples such as sɯ-sɯ-spoʁ ‘make a hole with’ fr om sɯ-spoʁ ‘make a hole ’ , causative of the intransitive verb spoʁ ‘hav e a hole’ (§17 .2.7). 2.4.6.3 Estimative The estimative nɤ- prefix (§17 .5), like a causative derivation, turns an intransi- tive verb into a transitive , but its meaning differs: the adde d argument is not a causer , but an experiencer feeling/ perceiving the state expressed by the base verb . For instance, the estimativ e of mpɕɤr ‘b e beautiful’ is nɤ-mpɕɤr ‘find beau- tiful’ , with the experiencer encode d as subject and the stimulus (corr esponding to the intransitive subject of the base verb) as object, as shown by (30). (30) ɲɯ-ta-nɤ-mpɕɤr sens-1 → 2-estim-be.beautiful ‘I find you very beautiful. ’ The estimative can be used to define adje ctiv es as a sub-class of stative verbs: only adjectival stative verbs can undergo this derivation, unlike for example e x- istential verbs and copulas (§22.5). 5 A double reciprocal form is attested (§18.4.2.5), but not with the same reciprocal prefixes. 30 2.4 V erbal morphology 2.4.6.4 Antipassive When the object of a morphologically transitive verb (thus excluding labile v erbs in intransitive conjugation, §14.5) is non-ov ert, it is necessarily interpreted as definite (§22.1.2.1). For instance, e xample (31a) can only b e used if the refer ent that has be en se wn has b een previously mentioned or is retrievable fr om the context, and cannot be understo od as ‘sew e d something’ with an indefinite object. T o express this meaning, several strategies ar e possible, including the antipassive rɤ- derivation (§18.6.1). (31) a. tɕʰeme girl nɯ dem kɯ erg cʰɤ-tʂɯβ ifr-sew ‘The girl sewed it. ’ b . tɕʰeme girl nɯ dem cʰɤ-rɤ-tʂɯβ ifr-ap ass-sew ‘The girl sewed / did sewing. ’ The rɤ- turns a transitive verb into an intransitiv e one, whose only argument is semantically the agent, but does not take ergative marking (31b). 2.4.6.5 Reflexive and reciprocal Japhug has a dedicated reflexive pr efix ʑɣɤ- (§18.3), different from other valency- decreasing derivations. The reflexiv e verb is conjugated intransitively , as shown by (32) 6 (32) tʰɯ-ʑɣɤ-rku-a aor:do wnstream-refl-put.in-1sg ‘I put myself (in the bag). ’ The reflexiv e is frequently combined with the sigmatic causative (§18.3.4) to express an unintentional indir ect causation affe cting oneself (33). (33) tɤ-ʑɣɤ-sɯ-mpɕa-a aor-refl-ca us-scold-1sg ‘I got myself scolded. ’ 6 If the verb in (32) wer e transitive, a past transitive -t suffix would be inserted (§11.3, §14.3.2.1) and the expected form would be † tʰɯ-ʑɣɤ-rku-t-a . 31 2 A grammatical sketch The reciprocal derivation (§18.4.1), entirely differ ent from the reflexiv e, is built by prefixing a- and reduplicating the verb stem (§4.1). For instance the transitiv e verb rqoʁ ‘hug’ yields a-rqɯ~rqoʁ ‘hug each other’ . Reciprocal verbs generally require a non-singular intransitiv e subje ct, and can also select a comitative post- positional phrase (§8.2.5). 2.4.6.6 Autive The autive nɯ- (§19.1) is a highly productive derivation, which does not affect verbal transitivity unlike the pre vious ones (§19.1.1). Its most basic function is self-affectedness or autob enefactive (§19.1.3). In par- ticular , the autive on transitiv e verbs taking a inalienably possessed obje ct can be use d to specify that the subject and the p ossessor of the object are cor eferent (34a), whereas the absence of the autiv e is generally interpreted as indicating the absence of coreference (34b). This is not an absolute syntactic rule , howe ver , since the A utive has additional unr elated uses (see b elow and §19.1.4, §19.1.5) which can interfere with this particular function. (34) a. ɯʑo 3sg kɯ erg ɯ-sroʁ 3sg.poss-life ko-nɯ-ri ifr-a ut o-save ‘S/he 𝑖 saved his/her 𝑖 own life . ’ b . ɯʑo 3sg kɯ erg ɯ-sroʁ 3sg.poss-life ko-ri ifr-save ‘S/he 𝑖 saved his/her 𝑗 life. ’ Another function of the A utive is to indicate spontaneous or non-volitional actions (§19.1.4), occurring for example by mistake, as illustrated by the minimal pair between (35a) and (35b). (35) a. tɤ-rɣe indef.poss-necklace pɯ-nɯ-prat-a aor-a ut o-br eak-1sg ‘I broke the pearl necklace ( by mistake). ’ b . tɤ-rɣe indef.poss-necklace pɯ-prat-a aor-br eak-1sg ‘I broke the pearl necklace (on purpose). ’ The third main function of the A utive is to express permansive aspect (§19.1.5). 32 2.5 Core and oblique arguments 2.4.7 D enominal derivations V erbalizing denominal (chapter 20) and deideophonic (§20.9) derivations are rich and productive in Japhug. A considerable number of denominal prefixes can be identifie d. Some of them have a w ell-identifiable meaning, for instance the proprietive aɣɯ- (§20.2.4) de- riving verbs meaning ‘having a lot of 𝑋 ’ or ‘pr o ducing a lot of 𝑋 ’ ( such as aɣɯlu ‘producing a lot of milk’ ( of a cow ) from tɤ-lu ‘milk’) or the similative arɯ- (§20.2.2). For the prefixes rɯ/ ɤ- (§20.4), nɯ/ ɤ- (§20.7) and ɣɯ/ ɤ- (§20.5), sev- eral different functions have to be postulated, since these prefixes can derive both intransitive (§20.4.1, §20.7 .1, §20.5.1) and transitiv e (§20.4.2, §20.7 .2, §20.5.2) verbs. Some denominal derivations occur in pairs (§20.7 .3, §20.4.3). For instance, when a noun has both rɯ/ ɤ- and nɯ/ ɤ- denominal verbs, the former is usually dy- namic intransitive , and the latter transitive, as illustrated by the pair comprising the intransitive verb rɤ-ma ‘ do ( some) work’ and its transitive counterpart nɤ-ma ‘ do (a work)’ , both from the noun ta-ma ‘work’ . Denominal derivations compete with light verb constructions to turn nouns into predicates (§20.1.2, §22.4). For instance, the meaning ‘tell lies, cheat’ from the noun kʰramba ‘lie ’ can be expressed either by a collocation with the light verb βzu ‘make ’ (§22.4.2.1) or by the denominal verbs rɯkʰramba ‘tell lies’ (§20.4.1, intransitive) and nɯkʰramba ‘ cheat’ (§20.7 .2, transitive). An important proportion of voice prefixes originate from the r eanalysis of denominal derivations from bare nominalized forms (§20.10), most clearly in the case of the rɤ- antipassive (§20.10.1, Jacques 2014b). When applied to noun-verb compounds (§5.5.5, §16.4.7), denominal derivations can serve to build incorporating verbs (§20.13.1). For instance, the intransitive verb ɣɯsɯpʰɯt ‘ cut fire wood’ is derived by the prefix ɣɯ- (§20.5.1) from the compound sɯpʰɯt ‘firew o od cutting’ , itself made from the bound state sɯ- of the noun si ‘wood’ compounde d with the transitive v erb pʰɯt ‘take out, cut’ , ‘cut’ . A tripartite contrast exists between the basic transitive construction ( si+pʰɯt ), a light verb construction combining βzu ‘make ’ with the corr esp onding action nominal compound ( sɯpʰɯt+βzu ) and the denominal incorporating verb ɣɯ- sɯpʰɯt , all three meaning ‘ cut firew oo d’ (§20.13.4). 2.5 Core and oblique arguments The morphosyntactic properties of arguments can be studie d from the point of view of flagging (§8, §8.2, §8.3) and indexation (§14.2, §14.3.2), but also r elativiza- 33 2 A grammatical sketch tion (§23.5) and coreference r estrictions between complement and matrix clauses (§24.2). Core arguments are defined as those that are indexed by the verb morphology . In the case of intransitive verbs (§14.2), the only argument inde xe d on the v erb is the intransitive subject , but the argument structure of intransitive v erbs can contain up to two additional oblique arguments (§14.2.5). Morphologically tran- sitive v erbs index two arguments (including in some cases a dummy one, §14.3.5). Although there are no differ ent morphological slots for transitive subjects and objects (§2.4.2.2, §14.3.2.8), the transitive paradigm contains no ambiguity in per- son configurations (§14.3.2), and the person of the agentive and patientive core arguments can always be clearly identifie d in finite verb forms. The cor e argu- ment indexed like the agentive argument of v erbs of action such as sat ‘kill’ or ʁndɯ ‘hit’ (as in §2.4.2.2) is called transitive subject , and the other one is the object . 7 Outside of verb indexation, the thr ee basic core arguments (intransitive sub- ject, transitive subject and object) are encode d in various ways, and pr esent dif- ferent types of alignments. 2.5.1 Neutral alignment Person indexation affixes in general have neutral alignment: for instance the 1sg suffix -a indexes the subject of intransitive verbs, and is also found in 1sg → 3, 3 → 1sg and 2 → 1sg configurations (§2.4.2.2, §14.3.2.8). The absence of strict coreference r estrictions b etw een matrix and subordinate clauses in some categories of complement clauses (§24.2.1.2, §24.5.6.1, §24.2.3.2) and in manner (§ 25.4.2) and temporal clauses (§25.3) could b e interpr ete d as a type of neutral alignment, but it appears that the neutralization in those cases is not limited to the three core arguments: there can also be complete absence of corefer ence, or coreference with an oblique or a possessor of an argument. 2.5.2 Nominative-accusative alignment 2.5.3 Subjectho od Nominative-accusative alignment appears in several unrelated constructions in Japhug. The first piece of evidence for this type of alignment in Gyalrong languages to have been proposed (Sun 2003) is the fact that the kɯ- (subject) participle 7 The term object in this grammar is restricted to this particular core argument, to the exclusion of all object-like patientive arguments. 34 2.5 Core and oblique arguments (§16.1.1.4) is the only form that can be used to relativize b oth intransitiv e (§23.5.1) and transitive subjects (§23.5.2). Howe ver , the kɯ- participles are not exclusiv ely used to relativize intransitive and transitive subjects: they can also r elativize p os- sessors of intransitive subjects (§23.5.10), and are the only option to do so, r esult- ing in ambiguities (see examples 177 and 178 in §22.5.2.1). The fact that nearly all labile verbs are subject-preserving (§14.5.1.3) could also be adduce d as e vidence of nominative-accusative alignment, but the e xistence of a handful of object-preserving labile verbs (§14.5.1.4) makes it less compelling. Clearer cases of constructions wher e strict nominative-accusative alignment is observed include asso ciated motion and complementation. First, the argument performing the motion encode d by the associated prefixes (§15.2.1) is always the intransitive subject in the case of intransitive verbs, and the transitive subject in the case of transitive ones (§15.2.2), and can nev er b e the object, a possessor of the subje ct or any oblique argument (but it can be the causee of a causative verb). Second, several subtypes of complement clauses require corefer ence between the (intransitive or transitive) subjects of the complement clause and that of the main clause, in particular bar e and dental infinitives (§24.2.2.2). 2.5.4 Objectho od While evidence for subjects independent from person indexation can be identi- fied, evidence for objects (as opposed to other non-subje ct arguments in absolu- tive form, §8.1.5, §8.1. 6, §8.1.8) is more elusive . In subject participles of transitive verbs, the object is marked by a possessive prefix (§16.1.1.1), obligatorily if no other pr efix (of orientation, negation or asso- ciated motion) is present, as in (36b). (36) a. spjaŋkɯ wolf nɯ dem kɯ erg ɕa meat ɲɯ-ndze sens-eat[III] ‘The wolf eats meat. ’ b . ɯ-kɯ-ndza 3sg.poss-sbj:pcp-eat ‘The one/someone who eats it’ Howe ver , possessive prefixes on subje ct participles cannot be used as evidence for objects outside of finite indexation, as they can mark other grammatical func- tions. For example the verb rga ‘like ’ is conjugated intransitively as sho wn by (37b): its subject spjaŋkɯ has no ergative marking, and no stem alternation is 35 2 A grammatical sketch observed on the verb (a form like † ɲɯ-rge would b e expected if this verb wer e transitive as in 36a, §12.2.2.1, §14.3.1). How ever , in addition to its subje ct, this verb takes an absolutive argument ( ɕa ‘meat’ in 37 a) that cannot be indexed: the semi-object . This argument is also marked as a p ossessive prefix on the subject participle like a real object, as in (37b). (37) a. spjaŋkɯ wolf nɯ dem ɕa meat ɲɯ-rga sens-like ‘The wolf likes meat. ’ b . ɯ-kɯ-rga 3sg.poss-sbj:pcp-like ‘The one/someone who likes him/her/it. ’ Moreov er , objects do not have any relativization construction that is specific to them. They can be relativized by object participial relatives (§16.1.2.4) and finite relativ es (§23.2.2), but these categories of clauses can also b e used to relativize semi-objects (§23.5.4.1, §16.1.2.5) and other participants such as goals (§23.5.5.2, §23.5.5.1) and possessors of objects (§23.5.10). For instance, headless participial clauses relativizing the object of ndza ‘ eat’ (38a) or the semi-object of rga ‘like’ (38b) have the same structure . (38) a. a-kɤ-ndza 1sg.poss-obj:pcp-eat ‘[The things] that I eat. ’ b . a-kɤ-rga 1sg.poss-obj:pcp-like ‘[The things] that I like. ’ 2.5.5 Absolutive-ergative alignment Although the terms “absolutive ” (§8.1) and “ ergative ” (§8.2.2) are used in this grammar to refer to flagging, Japhug does not display perfect absolutive-ergative alignment in case marking. Absolutive form ( absence of postp osition or relator noun) does indeed ser ve to mark intransitive subjects (§8.1.1) and objects (§8.1.3) by default, and transitive subjects normally take the ergative kɯ postposition (§8.2.2.1). How ever , both absolutive (§8.1.5, §8.1.7, §8.1.8) and ergative forms (§8.2.2.4, §8.2.2.7) have many functions other than marking core arguments. In addition, ther e is some fluidity in the use of the ergative postposition: some intransitive subjects can also be 36 2.6 W ord order marked by it (§8.2.2.3), in particular due to anticipation of a transitive verb in the following clauses (§8.2.2.2). The clearest evidence of absolutiv e-ergative alignment in Japhug is found in generic person indexation (§14.3.2.5): generic intransitive subjects and objects are both indexed by kɯ- , while generic transitive subjects are marked by the inverse wɣ- (§14.3.2.8). In addition, in local configurations (§14.3.2.3), the fact that the suffix closest to the verb stem indexes the object (like an intransitive subject) can also be analyze d as ergative alignment in this sub-part of the indexation system. There ar e no cases of syntactic pivot with exclusive neutralization of intransi- tive subject and object in complementation or relativization. 2.5.6 Ditransitive verbs Most ditransitive verbs have indirective alignment (§14.4.1): the theme is inde xed (treated as the object of a monotransitive verb), and the recipient marked with the genitive (§8.2.3) or the dative (§8.3.1). There ar e howe ver a few highly common verbs such as mbi ‘give ’ which hav e secundative alignment (§14.4.2) and index the recipient. Causativized transitive verbs have a different indexation pattern: either the causee or the patientive argument is indexed as object, dep ending on factors such as person hierarchy (§14.4.3). 2.6 W ord order Japhug has a strict verb-final order , with only a limited numb er of exceptions: right dislocated constituents (§22.1.3), only a few adverbs, particles (§10.4) and ideophones (§10.1.7) can occur p ostverbally (§22.2.7). When both the subje ct and the object of a monotransitive verb ar e overt as in (39), the former is placed by default b efor e the latter (§22.1.1.2). For ditransitive verbs, the or der b etween theme and r e cipient is not rigid (§22.1.1.3, §22.1.1.5). (39) lɯlu cat nɯ dem kɯ erg pɣa bird nɯ dem to-ndza ifr-eat ‘The cat ate the bird. ’ Noun phrases have by default (Demonstrativ e)-Noun- Adjective-Numeral-De- monstrative order (§9.3) as illustrated by (40). 37 2 A grammatical sketch (40) (kɯki) dem.pro x tɕʰeme girl kɯ-mpɕɤr sbj:pcp-be.beautiful χsɯm three kɯra dem:pl ‘These three beautiful girls’ 2.7 Subordination 2.7.1 Relative clauses Japhug lacks relative pr onouns, and most relative clauses are participial (§23.2.1). Finite relative clauses ar e also attested (§23.2.2). Most arguments and adjuncts are relativizable , though there are constraints on relativizability (§23.5.12). In the text corpus, an important proportion of relativ e clauses are headless (§23.4.1). When the head noun is overt, it is either internal to the r elative clause (§23.4.3) or follows it (§23.4.2), depending in part on the syntactic function of the relativized element and on the type of relative clause . Example (41) illustrates a head-internal finite relative: the relativized object pɣa ‘bird’ is located between the transitive subject lɯlu nɯ kɯ and the finite verb ( an object participle tɤ-kɤ-ndza (aor-obj:pcp-eat) would also be possible to express the same meaning), at the position it would normally occupy in the corresponding independent sentence (see 39 above). (41) [lɯlu cat nɯ dem kɯ erg pɣa bird ta-ndza] aor:3 → 3 ′ -eat nɯ dem ‘The bird that the cat ate ’ In (42) on the other hand, the relative clause pr ece des the relativized transitive subject lɯlu ‘ cat’ . (42) [pɣa bird ɯ-tɤ-kɯ-ndza] 3sg.poss-aor-sbj:pcp-eat lɯlu cat nɯ dem ‘The cat that ate the bird’ 2.7.2 Complement clauses Complement clauses in Japhug (chapter 24) occur in intransitive subject (43a, §24.5.8), object (43b, §24.5.3.4) and semi-obje ct (43c, §24.5.3.5) functions. There are also complement-taking nouns (43d, §24. 6). 38 2.7 Subordination (43) a. [kʰramba lie kɤ-βzu] inf-make mɤ-pe neg-be.good:f act ‘T elling lies is bad. ’ b . ɯʑo 3sg kɯ erg [kʰramba lie kɤ-βzu] inf-make mɤ-spe neg-be.able[III] ‘S/he is not able to tell lies. ’ c. [ɯʑo 3sg kɯ erg kʰramba lie kɤ-βzu] inf-make mɤ-nɤz neg-dare ‘S/he does not dare to tell lies. ’ d. [kʰramba lie kɤ-βzu] inf-make a-ʁjiz 1sg.poss-wish mɤ-ɣi neg-come:f a ct ‘I do not want to tell lies. ’ As sho wn by (43c), when the complement-taking verb is semi-transitive (and thus morphologically intransitive , §14.2.3), the verb in the complement clause is transitive and they share their subjects, the subje ct can take ergativ e case follow- ing the verb in the complement clause rather than absolutive (§24.2.1.1). Not all complements are infinitive clauses as in (43) (§2.4.4.2, §24.2.1). Some verbs are also compatible with finite complements (§24.2.3) as in (44), with sub- ject coreference . (44) [pjɯ-ta-sɯxɕɤt] ipfv-1 → 2-teach ɲɯ-spe-a sens-be.able[III]-1sg ‘I am able to teach you. ’ V erbs of spee ch also take reported spee ch clauses (§24.2.5), which present mor- phosyntactic properties different fr om regular finite complements (§24.2.5.2). Aside fr om complement clauses proper , various complementation strategies are also found, including r elative clauses in core argument function ( which re- semble, and can ev en be ambiguous with, complement clauses, §24.4.1), partici- pial clauses (§24.4.2), action nominals (§24.4.3) and simple coordination (§24.4.4). 2.7.3 Other sub ordinate clauses Subordinate clauses other than relative and complement clauses (tr eated in chap- ter 25) tend to be expressed by finite clauses headed by a relator noun or a post- position. For instance, clauses of temporal precedence (§25.3.2.1) require a finite verb in the Imperfective regar dless of the T AME category of the verb in the main clause (§21.2.3) followed the postposition ɕɯŋgɯ ‘b efor e ’ (§8.2.11) as in (45). 39 2 A grammatical sketch voiced one intransitive . It is not obvious which one is the derived form, and the direction of derivation is still being debate d (Sagart & Baxter 2012; Mei 2012). Japhug and other Northern Gyalrong languages (Jacques 2004: 411–412, Gong 2018: 271) howe ver pro vide a crucial piece of evidence showing that the direction- ality was from the transitive verb to the intransitive one (§18.5.1.2), confirming evidence from unr elated sources (Sagart 2003): the Tib etan borrowing χtɤr ‘scat- ter’ (from གཏོར་ gtor ‘scatter’) has an intransitive counterpart ʁndɤr ‘be scattered’ (§18.5.1) with prenasalized onset without equivalent in Tib etan. This intransi- tivizing derivation can be describ ed as anticausative (§18.5.2). There is further evidence that the prenasalization alternation comes fr om a nasal prefix, very probably a lexicalization of the autiv e nɯ- prefix (§19.1.7), whose spontaneous function (§19.1.4) is very close to that of anticausative verbs. Evidence for autive derivation only exists in Gyalrongic (Lai 2017: 357 –368, Gong 2018). Howe ver , the presence of traces of the anticausative derivation as voicing alternations in various branches of T rans-Himalayan, for instance in Kiranti (Jacques 2015a), Old Chinese (Sagart & Baxter 2012) or Tibetan (Jacques 2012b), implies that the autive derivation by extension must also be of proto- T rans-Himalayan age, and thus represents a unique ar chaism of Gyalrongic. The richness and high productivity of morphology in Japhug and other Gyal- rong languages, comparable with that of Sanskrit in Indo-European or Meskwaki in Algonquian, offer a framework to explore the fossil morphology of other T rans-Himalayan languages, in particular those belonging to the Burmo-Gyalron- gic and Tibeto-Gyalrongic branches (Jacques & Michaud 2011; Sagart et al. 2019), but also potentially for the family as a whole. Commenting on the irregularity of correspondences between Tibetan, Old Chi- nese and Burmese, Hill (2019: 212) concludes that “the phonetic influence of de- funct morphology will one day explain these complicated correspondences, but this possibility will manifest only when more languages, particularly archaic lan- guages such as those of the Rgyalrong and Kiranti branches, are br ought within purview . ” One of the aims of this grammar is precisely to provide comparativists with sufficient data on this language to make it systematically usable in T rans- Himalayan etymological resear ch, in the hope that this field can one day reach the degree of sophistication of Indo-European (Fellner & Hill 2019). 46 3 P honolog y 3.1 Introduction Japhug syllables follow the template ( C) (C)(C)V(C) with initial clusters contain- ing at most three consonants, and at most one consonant in the coda. Given the complexity of possible onsets, it is not practical to provide an e xhaustive list of possible syllables in the language (unlike Naish languages for instance, see Michailovsky & Michaud 2006). How ever , onsets and rhymes can b e listed ex- haustively . This chapter , partly based on previous publications (in particular Jacques 2004 and Jacques 2019c), presents the inventory of consonants and vow els, offers a fo- cused discussion on syllabic structure and quasi-neutralization, and describes suprasegmental phenomena and spee ch errors. The complete inv entor y of con- sonant clusters is listed and analyzed in §4.2 in the following chapter . This chapter does not treat the phonology of loanwords fr om Chinese (e xcept highly nativized ones). Non-nativized Chinese loanwords are r epresented in this grammar in pinyin ( even though this system is an imperfect way of rendering Sichuanese Mandarin) between angle d brackets. 3.2 Consonants 3.2.1 Onsets There are fifty consonantal phonemes in Japhug (T able 3.1). All can occur as simple onsets. Stops and affricates have a four-way contrast between voiceless unaspirated, voiceless aspirated, voiced and prenasalized series. The voiced fricatives /ɣ/ and /ʁ/ should be classified as non-nasal sonorants (§4.2.2), alongside the glides /j/ and /w/ , the lateral /l/ and the rhotic /r/ . T able 3.2 provides e xamples of each of these phonemes, followed whenev er possible by the vowel /ɯ/ . Among these consonants, four are only attested in borrowings from Tibetan and/or ideophones: /ʂ/ , /dʐ/ , /dʑ/ and /g/ . The phoneme /w/ is realized as a fricative [f] or [ɸ] before voiceless obstru- ents and as [v] or [β] before voiced ones, and can also be fricativize d when it 3 Phonology T able 3.1: Consonantal phonemes Bilabial Dental/ Retroflex Alv eolo- Palatal V elar Uvular Glottal Alveolar palatal Plosive unv . /p/ /t/ /c/ /k/ /q/ asp. /pʰ/ /tʰ/ /cʰ/ /kʰ/ /qʰ/ voi. /b/ /d/ /ɟ/ /g/ pren. /mb/ /nd/ /ɲɟ/ /ŋg/ /ɴɢ/ Affricate unv . /ts/ /tʂ/ /tɕ/ asp. /tsʰ/ /tʂʰ/ /tɕʰ/ voi. /dz/ /dʐ/ /dʑ/ pren. /ndz/ /ndʐ/ /ndʑ/ Nasal /m/ /n/ /ɲ/ /ŋ/ Fricative unv . /s/ /ʂ/ /ɕ/ /x/ /χ/ /h/ voi. /z/ /ʑ/ /ɣ/ /ʁ/ Approximant /w/ /j/ Rhotic /r/ Lateral voi. /l/ unv . /ɬ/ occurs as coda. In the orthography use d in this work, it is transcribed as < f > when followed by an voiceless stop, affricate or fricative , and as < β > when fol- lowed by a v oice d one (§4.2.1.1), or in coda position (§3.2.2). As in many languages of the Tibetan area, the /r/ is a trilled retr oflex voiced fricative [ɽʐ] in onset position, sometimes realized as a simple voiced fricative [ʐ] . It is devoiced to [ʂ] ( with neutralization of the contrast with /ʂ/ ) when fol- lowed by a v oiceless consonant in clusters (§4.2.1.4). The prenasalized voiced stops and affricates /mb/ , /nd/ , /ndz/ , /ndʑ/ , /ndʐ/ , /ɲɟ/ , /ŋg/ and /ɴɢ/ all have voiceless and voiceless aspirated counterparts such as /mp(ʰ)/ , /nt(ʰ)/ , /nts(ʰ)/ , /ntɕ(ʰ)/ , /ntʂ(ʰ)/ , /ɲc(ʰ)/ , /ŋk(ʰ)/ and /ɴq(ʰ)/ (§4.2.1.9). Y et, there are sev eral pieces of evidence showing that the prenasalized voiced stops and affricates are of a differ ent nature from the pr enasalize d v oice- less ones. First, the former can appear in clusters precede d by fricatives or non-nasal sonorants, as in /ʑmbr/ , /jndʐ/ or /rɴɢl/ , while the latter cannot. Clusters such as * /ʑmp(ʰ)r/ , * /jntʂ(ʰ)/ or * /rɴq(ʰ)l/ are not permitted in Japhug. Clusters of this type may have existed, but have been removed by voicing the stop/affricate (§18.5.7). Second, the uvular voiced prenasalized /ɴɢ/ has no simple voiced counterpart * /ɢ/ , a fact which therefore precludes analyzing /ɴɢ/ as a cluster /n+ɢ/ . There is a three-way contrast between /tɕ/ , /c/ and /k/ before the front vow el /i/ , as shown by the triplet comprising the correlativ e additive focus marker tɕi 48 3.2 Consonants T able 3.2: Examples of the consonant phonemes /p/ /ɯ- pɯ / ‘its young’ /tɕ/ /ɯ- tɕɯ / ‘his b oy’ /pʰ/ /ɯ- pʰɯ / ‘its price ’ /tɕʰ/ /to tɕʰɯ / ‘it gore him/her’ /b/ /ba bɯ / ‘blackcurrant’ /dʑ/ / dʑɯ / ‘it is oily’ /mb/ / mbɯt / ‘ collapse’ /ndʑ/ /ko ndʑɯ / ‘s/he accused him/her’ /m/ /tɯ mɯ / ‘sky’ /ɕ/ / ɕɯ / ‘who ’ /w/ / wɯ wɯ/ ‘ Boletus sp. ’ /ʑ/ /mɤ ʑɯ / ‘not only’ /t/ / tɯ boʁ/ ‘ one group’ /c/ / cɯ / ‘stone ’ /tʰ/ / tʰɯ / ‘be serious ’ /cʰ/ /tɤ cʰɯ / ‘wedge ’ (of a disease) /d/ / dɯ dɯt/ ‘turtledove ’ /ɟ/ /wa ɟɯ / ‘earthquake ’ /nd/ / ndɯ / ‘appear (rainbow)’ /ɲɟ/ / ɲɟɯ / ‘ open (it)’ /ts/ /konɤ tsɯ / ‘s/he hid it’ /ɲ/ / ɲɯɣ ɲɯɣ/ ‘soft and powdery’ /tsʰ/ / tsʰɯ tʰo/ ‘kid’ /j/ /ɯ- jɯ / ‘its handle ’ /dz/ / dzɯr dzɯr/ ‘straight’ /k/ / kɯ ki/ ‘this’ /ndz/ / ndzɯ pe/ ‘way of sitting’ /kʰ/ / kʰɯ na/ ‘ dog’ /n/ / nɯ ŋa/ ‘ cow’ /g/ / gɯ gɯɣ/ ‘very dark (sky)’ /s/ / sɯ mat/ ‘fruit’ /ŋg/ /ɯ- ŋgɯ / ‘inside ’ /z/ / zɯ mi/ ‘almost’ /ŋ/ /ɕa ŋɯ / ‘heat (deer )’ /l/ /rɯ lɯ / ‘medicine’ /x/ / xɯr xɯr/ ‘r ound’ /ɬ/ / ɬɯɣ nɤɬɯɣ/ ‘breathing mov ement’ /ɣ/ / ɣɯ / ‘genitiv e’ /tʂ/ / tʂɯm pa/ ‘apron’ /q/ / qɯ qli/ ‘staring’ /tʂʰ/ / tʂʰɯɣ / ‘maybe’ /qʰ/ /kɯ-sɤ qʰɯ qʰa/ ‘naughty’ /dʐ/ / dʐɯɣ dʐɯɣ/ ‘strong ( of tea)’ /ɴɢ/ /mɯ ɴɢɯ / ‘ Ligularia fischeri ’ /ndʐ/ / ndʐɯ nbu/ ‘guest’ /χ/ / χɯχɯ / ‘having big nostrils’ /ʂ/ / ʂɯŋ ʂɯŋ/ ‘ clear’ /ʁ/ /naŋ ʁɯ / ‘shirt’ /r/ / rɯ / ‘temporar y place ’ /h/ / ha nɯni/ ‘a little ’ (nomads) ‘also ’ (§9.1.6.2), the highly polyfunctional ci ‘ one ’ (§6.6.1, §6.8, §6.7 .2, §9.1.7, §7 .1.1, §9.1.4.1 and §22.2.1) and the demonstrative ki ‘this’ (§6.9). The palatal stops /c/ , /cʰ/ , /ɟ/ and /ɲɟ/ cannot be analyze d as velar+ /j/ clus- ters, as a clear contrast exists between the palatal series and velar stops follow e d by /j/ (§4.2.2.2), in minimal pairs such as ɲɟo ‘have damages’ and ŋgio ‘slip ’ , ‘glide ’ . 1 That the onsets /ɲɟ-/ and /ŋgj-/ have a different syllabic structure is confirmed by their reduplication patterns (§4.1): while in the former the palatal- ization is present on the reduplicant pɯ-nɤ-ɲɟɯ~ɲɟo ‘have damages e very where ’ (in the distributed action derivation, §19.4), in the latter the /j/ is not reduplicate d as pɯ-nɤ-ŋgɯ~ŋgio ‘he slipped every where ’ . The alveolo-palatal affricates /tɕ/ , /tɕʰ/ , /dʑ/ and /ndʑ/ are also contrastive with dental affricates+ j clusters, as shown by the minimal pair ndziaʁ ‘be tight’ 1 The grapheme <i> represents an allophone of /j/ in medial position with dental and dorsal initials (§4.2.2.2). 49 3 Phonology ( of knot) vs. ndʑaʁ ‘swim’ (§ 4.2.2.2). Ther e is also a contrast with dental stops+ j , though no good minimal pairs can b e found due to the rarity of these clusters. The attested contrasts b etw e en cor onal affricates and dorsal stops with and without the j medial are illustrated in T able 3.3 in combination with the v owel /o/ . T able 3.3: Palatalization contrasts in coronal and dorsal onsets Onset Example /ts/ tɤtsoʁ ‘ Potentilla anserina ’ /tsj/ tɤ-mtsioʁ ‘beak’ /tʂ/ tʂoʁ ‘add water’ /tɕ/ mtɕoʁ ‘be sharp’ /c/ co ‘valley’ /k/ ko ‘prevail o ver’ /kj/ kio ‘ cause to glide ’ /q/ rqoʁ ‘hug’ /qj/ qioʁ ‘vomit’ The voiceless lateral /ɬ/ (realized by some speakers as a postaspirated lateral [lʰ] ), is a marginal phoneme in Japhug, which does not appear in clusters (except heterosyllabic ones, as in /cɯɣɬaj/ ‘symptom in which the oral cavity becomes white ’) and is very rare in the native vocabulary . Y et, its phonemic status is justi- fied by the fact that it contrasts with /lx/ . There are no minimal pairs contrasting the two , but the contrast can b e indirectly illustrated by examples such as alxaj ‘not properly put’ ( of clothes) and lxɯlxi ‘thick and cumb ersome ’ on the one hand, and ɬɤt ‘be come old’ and ɬɤndʐi ‘ghost’ on the other hand. 3.2.2 Codas The inventory of consonants in coda p osition in Japhug is more r estricted than in initial position. In particular , the voicing and aspiration contrasts are neutralized in codas. Only twelv e consonants out of fifty app ear as codas: /-p/ , /-w/ , /-m/ , /-t/ , /-z/ , /-n/ , /-l/ , /-r/ , /-j/ , /-ɣ/ , /-ŋ/ , /-ʁ/ . The stop /-p/ is restricted to a few ideophones (§10.1.5.2), and is found neither in the inherite d non-ideophonic vo- cabulary nor in Tib etan loanwor ds, except as the first element of the heterosyl- labic cluster /pt/ in the word /sqap.tɯɣ/ ‘ ele ven’ (§4.2.3.1 ). The codas /-n/ , 50 3.2 Consonants /-l/ and /-ŋ/ are extr emely rare (but not entirely absent) in the non-ideophonic native vocabulary . A list of possible combinations b etw een co das and vo wels in Japhug is de- scribe d in §3.3.2. In word-final p osition, codas are v oiced when followed by a word beginning with a voiced consonant or a vow el, but are dev oiced in phrase-final p osition, before a pause or before a voiceless segment ( even acr oss word boundaries, §4.3). In isolation, word-final /-z/ , /-r/ , /-j/ , /-ɣ/ and /-ʁ/ in particular ar e realized as [s] , [r  ] , [j  ] , [x] and [χ] , respectively . The coda /-ʁ/ can also b e realized alternatively as pharyngealization of the preceding vowel. Since the voicing contrast between the voiceless fricatives /s/ , /x/ , /χ/ and the voiced ones /z/ , /ɣ/ , /ʁ/ is neutralized in co da position, it could seem b et- ter to argue that the fricative codas, whatever their phonetic r ealization, are archiphonemes { s,z} , { x,ɣ} and { χ,ʁ} , and that any discussion of their under- lying voicing is futile (Hill 2016). How ever , in the case of Japhug at least, some morphophonological rules are easier to describe if one assumes that fricative co- das are underlyingly voiced. 2 First, when the 1sg -a suffix is added to a verb stem ending in a fricative coda, that coda is resyllabified, be coming the onset of the syllable with a as rhyme. In these cases, the voiced allophone always surfaces (§14.2.1.1): for instance the 1sg → 3sg Imperfective of ntɕʰoz ‘use ’ is tu-ntɕʰóz-a , syllabifie d as /tu.ntɕʰo.za/ . If one were to assume that the fricativ e coda -z wer e voiceless or underspe cified for voicing, a context-specific voicing rule would have to be assumed to have taken place, since V  sa is a p ermissible sequence in Japhug, as in pjɤ  -wɣ-sat ifr- inv-kill (see for instance example 36 in §20.4.2). It is mor e economical to assume that -z and the other fricative codas are underlying voiced, and become devoiced in the same contexts as the sonorants. Second, the lo cativ e p ostposition zɯ (§8.2.4.1) is the result of the degrammati- calization of the locative *-s suffix still attested in Situ (§8.2.1). The fact that it has a voiced, rather than an voiceless onset, suggests that the fricative was v oiced when it was a suffix. 2 I owe the idea that final -z is v oiced to Sun (2005), where a similar analysis is implicitly pro- posed ab out T shobdun. 51 3 Phonology 3.3 V ow els and rhymes 3.3.1 V owels 3.3.1.1 V ow el phonemes Japhug has eight vow el phonemes, listed in T able 3.4. The mid-open unrounded vow els /ɤ/ and /e/ are only marginally contrastive: /ɤ/ does not occur in word- final open syllables except in unaccented clitics ( like the additiv e nɤ , §8.2.6), and /e/ only occurs in the last (accented) syllable of a word. They ar e clearly con- trastive only with the coda /-t/ (§3.3.2). T able 3.4: List of vowels in Japhug V owel Example Meaning /a/ /qa la / ‘rabbit’ /e/ /qa le / ‘wind’ /i/ /ɟu li / ‘flute ’ /ɤ/ / lɤ pɯɣ/ ‘radish’ /ɯ/ /rɯ lɯ / ‘medicine’ /y/ /qa ɟy / ‘fish’ /o/ / lo / upstream /u/ /tɤ lu / ‘milk’ Not all speakers of Kamnyu Japhug hav e a phoneme /y/ in the native vocabu- lary . Even for the speakers who ditinguish it, it is only atteste d in the w ord ‘fish’ and the verbs derived from it. It ne vertheless contrasts with /ɯ/ and /u/ , as shown by the quasi-minimal pairs /qaɟy/ ‘fish’ , /waɟɯ/ ‘ earthquake ’ and /ɟuli/ ‘flute ’ . Other speakers pronounce ‘fish’ with a medial /w/ as /qaɟwi/ . However , [y] is found in the spee ch of all Japhug speakers in Chinese loanwords such as 洋芋 <yángyù> ‘potato ’ . 3.3.1.2 V ow el assimilation When followed by a syllable containing a r ounded vowel ( /u/ or /o/ ), the back unrounded vow els /ɯ/ and /ɤ/ optionally undergo rounding harmony to [u] and [o] , respectively . For instance , /ɣɤʑu/ ‘e xist (sensory)’ (§22.5.1.2) is gene- rally pronounced as [ɣoʑu] , and the reduplicated form tɯ~tu-dɤn ‘more and more ’ (§12.4.1.4) is realized as [tutudɛn] . 52 3.3 V owels and rhymes The phoneme /ɤ/ in prefixes tends to be pronounced more open as [ɐ] when followed by a syllable whose main v owel is /a/ , making it sometimes difficult to perceive the contrast, for instance between ta-ma ‘work’ and tɤ-ma ‘mother’ (honorific). In the verbal system, the 1sg -a suffix triggers obligator y regr essive assimilation ɤ → a on the preceding syllable (see T able 14.2, §14.2.1.1). 3.3.1.3 Synizesis While no true diphthongs exist in Japhug, when the 1sg -a suffix is added to a verb stem ending in an open syllable, the two syllables undergo synizesis (§14.2.1.1). When the verb stem contains the mid vo wels /-e/ and /-o/ , they become the cor- responding high vow els /-i/ and /-u/ due to merging with a , and the contrasts between /e/ and /i/ on the one hand, and /o/ and /u/ on the other hand, are neutralized. For instance, tso-a [tsua] ‘I understand’ and βze-a [βzia] ‘I (will) do it’ are homophonous with tsu-a ‘I have time ’ and βzi-a ‘I ( will) be drunk’ , r espe c- tively . Thr ee pseudo-diphthongs are thus atteste d: [ia] (fr om /e-a/ and /i-a/ ), [ua] (from /o-a/ and /u-a/ ) and [ɯa] (from /ɯ-a/ ). Synizesis results in syllables homophonous to the rhymes -wa and -ja in clus- ters ending in -w- (§4.2.2.1) or -j- (§4.2.2.2) followed by the vo wel -a . For instance, aro-a ‘I own’ (§14.2.3) is homophonous with a-rwa ‘my tent’ (§4.2.2.1). Apart from synizesis, tw o other typ es of v owel contraction are found in Japhug. First, verb stems with initial a- have sp ecific conjugation patterns, where initial a- merges with the immediately preceding prefix following rules that are not completely trivial (§12.3). Second, the 1sg -a suffix merges with -a stems as a , without vow el lengthening in Kamnyu Japhug (§14.2.1.1). 3 3.3.1.4 Zero onset There are str ong phonotactic restrictions on vow el-initial stems and words in Japhug. In word-initial position, only a- and ɯ- are found. These v owels can merge with previous open syllables (§4.3), and no glottal stop appears, unlike in many T rans-Himalayan languages such as Khaling (Jacques et al. 2012). P honetic [u-] , [o-] , [i-] do appear in word-initial position, but are preferably analyzed as /wu-/ , /wo-/ and /ji-/ with a glide. The only possible stem-initial vowel in v erbal stems is a- (§12.3). It is relatively common due to the fact that it appears in many derivational prefixes (§20.2, §18.1, 3 Howe ver , vowel lengthening is found in some dialects, such as that of Sarndzu. 53 3 Phonology §18.4, §18.7). Stem-initial a- undergoes vowel contraction with all pr ece ding pr e- fixes. By contrast, stem-initial wu- or ji- (for instance, wum ‘gather’ and ji ‘plant’) never merge with preceding prefixes. In noun stems, initial a- also exists and interacts with prefixes (§5.1.1.1), but is considerably rarer . Comparison with other Gyalrongic languages indicates that w ord-initial a- and ɯ- (in the native vocabulary , excluding loanwor ds and interje ctions) are se c- ondary . W ord-initial ɯ- is only attested in a few interjections (such as ɯtɕʰɯtɕʰɯ , §10.2.1) and in the third possessive prefix (§5.1.1). The form † ɣɯ- (homophonous with the inverse prefix, §14.3.2.7) would be expe cted, and the unexpected form ɯ- in Japhug might be due to false segmentation in sandhi (§5.1.1.5). W ord-initial a- originates mainly from *ŋa- (Jacques & Chenzhen 2007), due to loss of the simple initial *ŋ- in non-stressed syllables (including wuma ‘real, really’ fr om ངོ་ མ་ ŋo.ma ‘real, true ’ , but excluding monosyllabic verb stems such as ŋu ‘be ’ and ŋa ‘buy on credit, owe ’). 3.3.2 Rhymes There ar e strong phonotactic constraints on possible rhymes in Japhug. T able 3.5 lists all attested rhymes; the coda /-w/ is transcrib ed as -β in the orthography used in this grammar . T able 3.5: List of possible rhymes in Japhug /w/ /p/ /m/ /t/ /n/ /z/ /l/ /r/ /j / /ɣ/ /ŋ/ /ʁ/ /a/ /aw/ /ap/ /am/ /at/ /an/ /az/ /al/ /ar/ /aj/ /aŋ/ /aʁ/ /e/ /et/ /i/ /it/ /il/ /ɤ/ /ɤw/ /ɤm/ /ɤt/ /ɤn/ /ɤz/ /ɤl/ /ɤr/ /ɤj/ /ɤɣ/ /ɯ/ /ɯw/ /ɯp/ /ɯm/ /ɯt/ /ɯn/ /ɯz/ /ɯl/ /ɯr/ /ɯj/ /ɯɣ/ /ɯŋ/ /y/ /yt/ /o/ /om/ /ot/ /on/ /oz/ /ol/ /or/ /oj/ /oŋ/ /oʁ/ /u/ /ut/ /uz/ /uj/ The only coda attested with all vowels is /-t/ (T able 3.6). Among these rhymes, /-et/ and /-yt/ are only attested in verb forms with the past transitiv e -t suffix (§11.3, §21.1.3, §14.3.2.1), which occurs in word-final position only in 2sg → 3 forms. In closed syllables with an alveolo-palatal or a palatal consonant preceding the vow el, /ɯ/ is fronted and its contrast with /i/ is neutralized (§3.5.2.1). It is only maintained b efor e /-t/ in forms with the past transitive -t suffix. For instance, 54 3.3 V owels and rhymes T able 3.6: Examples of closed syllable rhymes in /-t/ V owel Rhyme Example Meaning/Gloss /a/ /at/ /tɤtɯsɤlat/ ‘you boiled it’ tɤ-tɯ-sɯ-ɤla-t aor-2-ca us-boil-pst:tr /e/ /et/ /tɤtɯnɤmɤlet/ ‘you did it’ tɤ-tɯ-nɤmɤle-t aor-2-do-pst:tr /i/ /it/ /tɤtɯrɤlit/ ‘you reimbursed it’ tɤ-tɯ-rɤli-t aor-2-r eimburse-pst:tr /ɤ/ /ɤt/ /jɤtɯlɤt/ ‘you thr ew it’ jɤ-tɯ-lɤt aor-2-r elease /ɯ/ /ɯt/ /tʰɯtɯplɯt/ ‘you destr oyed it’ tʰɯ-tɯ-plɯt aor-2-destr oy /y/ /yt/ /lotɯznɯqaɟyt/ ‘y ou let him fish’ lo-tɯ-z-nɯ-qaɟy-t ifr-2-denom-fish-pst:tr /o/ /ot/ /nɯtɯsɤwlot/ ‘you took care of him’ nɯ-tɯ-sɤβlo-t aor-2-take .care-pst:tr /u/ /ut/ /pɯtɯnɤlut/ ‘you milked it’ pɯ-tɯ-nɤ-lu-t aor-2-denom-milk-pst:tr we find the minimal pair /tɤ-tɯ-cɯ-t/ ‘you opened it’ (aor-2-open-pst) and /lɤ-tɯ-cit/ ‘you mov ed’ (aor-2-mov e). With the coda /-j/ , the contrasts b etw e en /ɯ/ and /i/ on the one hand, and /ɤ/ and /e/ on the other hand, are neutralized. The rhyme /-aj/ is realized as [ɛj] or [æj] . 3.3.3 Historical phonolog y Some notions of Japhug historical phonology are useful to account for the gaps in the distribution of rhymes (§3.3.2) as well as some v owel alternations (§12.2.2.1). Comparison of inherited vocabular y between Japhug and extra-Gyalrongic languages shows that the codas *-l , *-n and *-ŋ have been lost in the native vo- cabulary (T able 3.7). W ords with these codas are either b orr owed from Tibetan or have an ideophonic origin. 55 3 Phonology 3.5.2.2.1 tɕʰi° vs. tɕʰɯ° Many nouns in Japhug have tɕʰi° or tɕʰɯ° as first ele- ment, in particular wor ds of Tib etan origin containing ་ tɕʰu ‘water’ or མ་ mtɕʰu ‘lip ’ . The variant tɕʰi° is found in the great majority of these words. In some cases like tɕʰira ‘water jar’ (from ་ར་ tɕʰu.ra ‘water container’), I have used an etymol- ogizing transcription with ɯ ( tɕʰɯra ) in previous w orks (in particular Jacques 2015b), but changed my mind after careful rechecking (using for instance the stems I and III of the verb tɕʰɯ ‘gor e, stab ’ for comparison). The variant tɕʰɯ° is essentially restricted to wor ds with a cluster with a uvular or a labial fricative as first element ( such as rtɕʰɯʁjɯ ‘ caterpillar’ , tɕʰɯχpri ‘sala- mander’ , tɕʰɯβroʁ ‘type of tsampa’), but also found in tɕʰɯmɲɯɣ ‘water hole’ (from ་མིག་ tɕʰu.mig ‘water hole ’) and tɕʰɯzɯ ‘ cloth beam’ . The noun tɯ-mtɕʰi ‘mouth’ (from མ་ mtɕʰu ‘lip ’) has an unexpected i (the ex- pecte d form would be † mtɕʰɯ ). It is possible that it was extracted from a com- pound ( like tɯ-mtɕʰirme ‘moustaches’) wher e /ɯ/ and /i/ are neutralized as [i] , and that this phonetic variant was then reinterpreted as an independent root. 3.5.2.2.2 ndʑi° vs. ndʑɯ° Compounds with ndʑi° or ndʑɯ° as non-final ele- ment are not common, but social relation collective nouns (§ §5.7 .9.1) containing the prefix kɤndʑi- , are particularly numer ous. W ords with the variant ndʑɯ° are very rare: the two verbs ndʑɯrpɯt ‘be numb ’ and andʑɯβri ‘protect each other’ ( with the rare reciprocal prefix andʑɯ- / andʑi- of denominal origin, see §20.2.5), and the nouns ndʑɯrwɯz ‘ Sonchus sp. ’ and ndʑɯnɯ ‘ A ngelica sp . ’ (minimal pair with ndʑi-nɯ ‘their breast’ , see § §3.5.2.4). 3.5.2.2.3 ci° vs. cɯ° Compounds comprising cɯ ‘stone’ or the root of tɯ-ci ‘water’ as non-final element are common, and the majority of these wor ds have the variant ci° , even when the noun comes fr om cɯ ‘stone ’ (for instance ciχɕiz ‘stony earth’). Exceptions include cɯrmbɯ ‘stone heap ’ , scɯʁzɯɣ ‘app earance ’ (from ེ་གགས་ skʲe.gzugs ‘physical appearance ’) and ɣɤcɯqʰlɯβ ‘making noise (of water when agitated)’ . 5 3.5.2.2.4 cʰi° vs. cʰɯ° Nouns with cʰi° or cʰɯ° include in particular Tibetan loanwor ds in ི་ kʰʲi ‘dog’ . The form cʰi° occurs in most cases (for instance cʰisɲu ‘rabies’ from ི་ ོ་ kʰʲi.smʲo ‘rabies’), but exceptions include cʰɯmu ‘female dog’ , cʰɯrdom ‘roaming dog’ and nɯcʰɯra ‘keep guard’ (rechecked using the minimal pair cʰi ‘be sweet’ vs. tɤ-cʰɯ ‘wedge’). 5 This verb is a denominal verbalization fr om a noun-ideophone comp ound (§5.5.3). 62 3.5 Neutralization, quasi-neutralization and free variation 3.5.2.2.5 ɟi° vs. ɟɯ° W ords containing the status contructus of ɟu ‘bamboo’ can have either the vo wel /i/ (as ɟispjɤt ‘person making bamboo baskets’) or /ɯ/ ( ɟɯmɢom ‘bambo o tw e ezers’). Other w ords in ɟi° include ɟiga ‘tortuous path’ and its derived forms. The root of tɯ-rɟɯ ‘fortune ’ commonly occur as first element of comp ounds, and have the vo wel /ɯ/ , as in rɟɯrŋom ‘ coveting other people ’s fortune ’ . The syllable -rɟɯ- also occurs as the status constructus of tɤ-rɟit ‘child’ in the w ord tɤrɟɯsti ‘ only child’ . 3.5.2.2.6 ji° vs. jɯ° Compounds with non-final jɯ° are extremely rar e, limited to the two Tibetan loanwords jɯɣi ‘writing’ (fr om ཡི་ གེ་ ji.ge ‘letter’) and pjɯrɯ ‘ coral’ (from ་་ bʲu.ru ‘ coral’). 3.5.2.3 V erbal prefixes and reduplicated forms The vow els /ɯ/ and /ɤ/ are by far the most common ones in prefixes in Japhug. In the case of verbal prefixes with palatal and alv e olo-palatal onsets, the same phonological problem as in compounds (§ §3.5.2.2) is observed, namely the ques- tion whether the contrast between /ɯ/ and /i/ has b een neutralized or not, and whether the vow el should be transcrib ed as i rather than as ɯ . The prefixes in which this problem arises ar e few , and are e xhaustively listed in T able 3.12. For prefixes in ɕV- and cʰV- , testing the vow el is possible using minimal pairs ( ɕɯ ‘who ’ vs. ɕi ‘polar interrogative ’ and cʰi ‘be sweet’ vs. tɤ-cʰɯ ‘wedge ’ as in § §3.5.2.2.4). For pjV- and jV prefixes, comparison is possible with the first syllable of the nouns jɯɣi ‘writing’ and pjɯrɯ ‘ coral’ (§ §3.5.2.2.6). The main consultants on the basis of whose expertise this work has been writ- ten have presented conflicting judgments regar ding the nature of the vo wel in these prefixes. After some hesitation, T shendzin considers it to be i for all of these prefixes (for instance , she considers the vowel of pjɯrɯ ‘ coral’ to be different from that of pjɯ-ɣi ‘he comes down’). I ne vertheless keep here the orthographic ‘ etymological’ transcription with the vow el ɯ for three reasons. First, the contrast is marginal, if existent at all, for most speakers, and given the limited numb er of pr efixes affecte d, it is possible to automatically change the transcription from ɯ to i only in these prefixes without loss of information (this can be ne cessary for instance as a pre-treatment for the purposes of automatic transcription training). Second, the prefixes in T able 3.12 are affected by vowel fusion (§12.3) and co- alescence with the inverse prefix -wɣ (§14.3.2.7) like other pr efixes in Cɯ- , and 63 3 Phonology these morphophonological phenomena are easier to state by assuming the un- derlying vow el ɯ for all prefixes. Third, the assimilation from ɯ to i follo wing palatal and alveolo-palatal con- sonants appears to b e general in the morphology of the language , including par- tially reduplicated forms, whose rhyme is replaced by -ɯ (§4.1). For instance, T shendzin judges that the reduplicant here transcribed as -ɕɯ- of the redupli- cated form nɤɕɯɕe ‘go around’ (from ɕe ‘go ’) to be phonetically closer to ɕi ‘polar interrogative ’ than to ɕɯ ‘who ’ . Adopting a transcription with i in the prefixes in T able 3.12 would ther efore only make sense if reduplicated forms of syllables with palatal and alveolo-palatal consonants are also transcribed with i (thus nɤɕiɕe instead of nɤɕɯɕe ). Howe ver , such an orthography would cause unnecessar y problems when automatically r esearching for reduplicated forms in the corpus. Since also in this case the pronunciation of the grapheme ɯ as /i/ is completely predictable, and could be corrected automatically without difficulty , I choose to keep an etymological transcription. T able 3.12: V erbal prefixes with palatal or alveolo-palatal onsets and high unrounded vow els Prefix Function Reference pjɯ/i- Orientation prefix, B-type, downwar ds §15.1.1.1 cʰɯ/i- Orientation prefix, B-type, downstr eam §15.1.1.1 ɲɯ/i- Orientation prefix, B-type, westwar ds; sensor y §15.1.1.1 jɯ/i- Proximativ e §21.6.2 ɕɯ/i- translocative associated motion § §15.2.1.2 ɕɯ/i- apprehensive § §21.7.1 ɕɯ/i- causative §17.2.2.1 ɕɯ/i- denominal §20.3.2 In other wor ds, in the transcription adopted in this work I transcrib e the con- trast between ɯ and i after palatal and alveolo-palatals when non-predictable (in noun and verb stems), but neglect it and use ɯ throughout in verbal pr efixes and reduplicated syllables. 3.5.2.4 V erbal suffixes and possessive prefix es A few unstressed verb suffixes hav e alveolo-palatal onsets and high unrounded vow els: the dual indexation suffixes 1du -tɕi and 2/3du -ndʑi (§14.2.1.1, §14.3.2.1) 64 3.5 Neutralization, quasi-neutralization and free variation and the suffix -ci (§11.4). It is clear that these suffixes originally had a schwa-like vow el rather than a front vo wel as in most Gyalrong languages ( see §14.8.1), but here too T shendzin considers the vow el of these suffixes to b e /i/ rather than /ɯ/ (the suffix -ci , for instance, according to her r esembles ci ‘ one’ mor e than cɯ ‘stone ’), and since no vow el fusion or other phenomena takes place with these suffixes (unlike verbal prefixes), using an etymological notation w ould have be en an unnecessar y complication. In the prefixal possessive paradigm (§ §5.1.1), we also find a series of prefixes with palatal and alveolo-palatal onsets and high unrounded vo wels: 1du tɕi- , 1pl ji- and 2/3du ndʑi- . Unlike verbal prefixes, but like the v erbal indexation suffixes, possessive prefixes in the K amnyu dialect of Japhug never undergo v owel fusion and remain invariable . The vowel is phonologically /i/ , as sho wn by the minimal pair between ndʑɯnɯ ‘ A ngelica sp. ’ and the dual possessive ndʑi-nɯ ‘their 𝑑 𝑢 breast’ (§ §3.5.2.2.2). For these reasons, I pr efer a phonological (with -i ) rather than etymological ( with -ɯ ) transcription for these prefixes. 3.5.3 The contrast b etween /ɤ/ and /e/ after palatal and alveolo-palatal consonants The vow els /ɤ/ and /e/ in Kamnyu Japhug contrast in v er y few conte xts. In wor d-final stressed op en syllables, only /e/ is found, /ɤ/ being attested only in clitics and unstressed final syllables in words such as kɯnɤ ‘also , even’ . In closed syllables, /e/ is only found with the coda /-t/ in the past 2sg → 3 of transitive -e stem verbs (§3.3.2); the wor d-final -et vs. -ɤt are the only cases where /ɤ/ and /e/ are clearly contrastive . In non-final open syllables, there are no e xamples of minimal pairs involving a contrast between /ɤ/ and /e/ . P honetic [e] is clearly heard after palatal and alveolo-palatal consonants when the vow el of the following syllable is /e/ , for instance in wor ds such as tɕʰeme ‘girl’ or sɤrɲɟele ‘e xtend ( limbs)’ , but in other contexts I generally transcribe ɤ throughout, except when the w ord is an obvious compound whose elements are recognized by the speakers (for instance tɕetʰa ‘later’ , ‘soon’ from tɕe ‘then’ and tʰa ‘later’). 3.5.4 The contrast b etween /ɤ/ and /a/ The contrast between /ɤ/ and /a/ , while clear in most contexts, is difficult to perceive in unstr esse d syllables when follo wed by a uvular (in particular when followed by a cluster with a uvular pr einitial) or when followed by a syllable containing a ( due to regressive assimilation of height, §3.3.1.2). 65 3 Phonology In slow syllable-by-syllable pronunciation, T shendzin makes it clear that some unstressed syllables preceding a uvular have /a/ rather than /ɤ/ . This concerns the indefinite possessor prefix in wor ds such as ta-ʁi ‘younger sibling’ , ta-ʁrɯ ‘horn’ , ta-ʁri ‘ dirt’ , ta-ʁjɯβ ‘shadow’ , ta-ʁrɯm ‘light, shadow’ , ta-χpi ‘shap e , mo del’ , ta-ʁrɤt ‘ charcoal’ , ta-ʁɟaz ‘soot’ and ta-ʁa ‘free time ’ (§ §5.1.2), and the fossilize d prefix in nouns such as taqaβ ‘needle ’ . The noun tɤ-ʁar ‘wing’ , howev er , has tɤ- rather than ta- . Another context where the variant ta- is found is with the m- initial nouns ta-ma ‘work’ and ta-mar ‘butter’ . Note that a contrast exists with the honorific noun tɤ-ma ‘mother’ (§5.1.2), showing that the two allomorphs tɤ- and ta- ar e synchronically contrastive in the Kamnyu variety at least for some speakers. With some verbal prefixes, in particular the antipassive prefixes rɤ-/ra- and sɤ- /sa- (§18.6) , the pr oprietive sɤ-/sa- (§18.8) and the estimative nɤ(ɣ)-/na- (§17 .5.1) and several denominal pr efixes), the a allomorph is found where followed by a stem with a uvular preinitial. For instance, the antipassiv e of χtɯ ‘buy’ is raχtɯ ‘ do shopping’ with the variant ra- rather than rɤ- . Not all prefixes in ɤ present this alternation: for instance, the causative , facilitative and denominal prefixes ɣɤ- never hav e a variant † ɣa- . Howe ver , in the case of orientation pre verbs (§15.1.1.1) and other inflectional prefixes, no such phonologically determined alternation is found. For instance, the A -type pre verb tɤ- upw ards r emains unchanged when prece ding stems with a uvular preinitial as in tɤ-χtɯ-t-a (a or-buy-pst:tr-1sg) ‘I b ought it’ , where the prefix tɤ- is different fr om the corresponding C-type preverb ta- in ta-χtɯ (a or:3 → 3 ′ -buy ) ‘he b ought it’ . 3.6 Sp eech errors and self-corrections The Japhug corpus, being exclusively an oral one, contains many sp eech errors, and immediate self corrections by the speakers, which have be en systematically transcribe d. T able 3.13 presents e xamples of phonological errors that are follow e d by self-corrections in the corpus. Phonological errors can result in the replacement of a w ord by a similar-soun- ding one (as in the case of sqi ‘ten’ for sqʰi ‘tripod’), but more often by a non-sense form. Errors generally involv e one phonological feature at a time, including aspi- ration, nasality and place of articulation, especially the contrast b etween dental and alveolo-palatal affricates, and that between uvular and velars. A more puz- zling replacement is that of /χ/ by /r/ ( saχaʁ → sara ) despite the absence of any common phonological feature between these two phonemes. 66 3.6 Spe ech errors and self-corr e ctions T able 3.13: Examples of self-corrections in the Japhug corpus Erroneous Correct form Error Reference form type [sqi] ‘ten’ sqʰi ‘tripod’ aspiration (160703 poucet3) {0006107#S47} [tɯsŋɤr] tɯ-skɤrma ‘ one nasal/oral (150825 baishe zhuan-zh) cent’ {0006342#S69} [tsʰo-] tɕʰorzi ‘jar’ dental/alveolo- (160703 araR) palatal {0006101#S48} [tɕʰɯr-] tsʰɯrɟɯn ‘ often’ (160630 abao-zh) {0006197#S107} [sara-] saχaʁ ‘it is rhotic/uvular (160712 smA G) extremely ... ’ {0006073#S20} [qapri] qapi ‘white stone ’ (160630 poucet2) ‘snake ’ {0006155#S2} [ɯmɲo] ɯ-jmŋo ‘his dream’ yod metathesis (160630 abao-zh) {0006197#S109} [pjɤka] pɤjka ‘squash’ (150827 mengjiangnv-zh) {0006290#S15} [ɯʁri] ɯ-ʁrɯ ɯ-ntsi v owels (160715 nWNa) ‘ one of her horns’ {0006067#S10} [pɣɤjmŋɤt] pɣɤjmɤt (unidentified insertion (140505 sWjno) species of plant) {0003919#S21} 67 3 Phonology The phoneme /j/ is particularly prone to metathesis, either within initial clus- ters ( ɯjmŋo → ɯmɲo , with fusion of /j/ and /ŋ/ as /ɲ/ , see §4.2.1.9), or across vow els ( pɤjka → pjɤka ). 3.7 Suprasegmentals Unlike all other Gyalrong languages, including T shobdun (Sun 2005), Situ (Lin 2012) and Zbu (Gong 2018), Japhug has no tonal contrasts. How ever , ther e is mor- phologically determined stress. Phonological words only have one str ess, which is located by default on the final syllable of the word (§3.8.1, §11.6.1). The personal agreement suffixes (§14.2.1, §11.3) and the peg suffix -ci (§11.4) ne- ver receiv e stress, and their vow els are optionally devoiced. For instance, tɤ-ndza- t-a ‘I ate it’ (aor-eat-pst-1sg) is r ealized as [tɤndzátḁ] or [tɤndzáta] . Stress can be antep enultimate in the case of v erb forms with two suffixes as in to-k-ɤmɯ- rpú-ndʑi-ci ‘they bumped into each other’ (ifr-peg-recip-bump-du-peg). There ar e only three verbal prefixes which can attract stress: the inverse -wɣ- (§14.3.2.7), the negative Sensory mɯ  j- (§13.1.1, §21.3.2.1), and the interrogative ɯ- (§21.7 .4.1). Outside of verbal morphology , the only other process that affects stress is the comitative adv erb derivation (§5.8.1). Some ideophones have em- phatic stress on the first syllable (§10.1.5.4), but it is not phonologically distinc- tive . A few function wor ds have penultimate stress: kɯ  nɤ ‘also ’ (§9.1.6.1), cínɤ ‘not even one ’ (§9.1. 6.4), nóʁmɯz and kóʁmɯz ‘ only after’ (§8.2.11). 3.8 W ord structure 3.8.1 W ordhood Although no native expr ession exists to designate ‘wor ds’ in Japhug (as in the immense majority of the world’s languages, Dixon & Aikhenvald 2002), speakers have an intuitiv e notion of a minimal unit which can b e the object of metalin- guistic discourse, as opposed to prefixes and suffixes. The intuitive notion of ‘wor d’ can b e corr elated with some phonological and morphological criteria. The clearest phonological criterion for wor dhoo d is stress. Str ess is by de- fault wor d-final, and words have at most one str ess. Most function words, in- cluding linkers (§25.1.6), determiners (§9.1), postpositions (§8.2) and even r elator nouns (§8.3) lack stress (unless they r e ceiv e spe cial emphasis). In example (1), the 68 3.8 W ord structure stressed syllables are indicated by an acute accent, and the unstressed function words by a grav e accent. (1) tɕè lnk ʁzɤ mí couple cì indef pjɤ- tú -ndʑi ifr.ipfv-exist-du tɕè lnk tɤ- rʑáβ indef.poss-wife nɯ  dem ɬa mú anthr pjɤ- rmí . ifr.ipfv-be.called tɤ- tɕɯ  indef.poss-son nɯ  dem tsʰɯ ráŋ anthr pjɤ- rmí . ifr.ipfv-be.called ‘There was a couple , the wife was calle d Lhamo , and the man T shering. ’ (28-qajdoskAt) {0003718#S2} There are only three groups of exceptions to wor d-final stress placement (for instance the unstressed suffix -ndʑi in 1), as seen in the previous section (§3.7). Given its pr e dictability , stress is not noted in the transcription employed in this grammar , except on str ess-attracting prefixes. Among the morphological criteria for wordhood (§11.6), the scope of partial reduplication can be employed, at least for the word classes that allow it. Initial reduplication, almost exclusively r estricted to verb forms (§12.4.1), applies to the first syllable of the wor d almost without exception. 6 In the case of verbs, sev eral additional morphological tests converge to indicate the same left boundar y . For the right boundar y of the wor ds, partial reduplication is less useful, as it only applies to stems (§12.4.2): inflectional suffixes are never reduplicated. Inflectional suffixes (§11.3), in particular person indexation suffixes (§14.2.1.1, §14.2.1.2), do present clitic-like properties, since they ar e not stressed, and are outside the scope of reduplication. Howev er , these morphemes cannot be iso- lated, follow a rigid order , hav e non-adjacent dep endencies with v erbal prefixes (§11.4), and no external element can be inserted b etween them. Moreov er , the 1sg marker -a , despite being unstressed when following consonant-final verb stems, undergoes fusion with vowel-final stems (§3.3.1.3), causing v owel mergers. The verbal w ord in Japhug is thus slightly larger than the prosodic domains of stress and partial reduplication (Schiering et al. 2010). There are , howe ver , three problems challenging the notion of ‘w ord’ in Japhug: cliticized sentence final particles (§11.6.2), bipartite verbs (§11. 6.3) and prenominal attributes (§9.1.8.2) when they receive no stress. These issues are discussed in more detail in the rele vant sections. 6 The only exception is the reduplicated subject participle kɯ~kɯ-tu of the existential verb tu ‘ exist’ , which can take (non-r eduplicate d) possessive prefixes, but this may b e a lexicalized form (§12.4.1.5). 69 3 Phonology 3.8.2 Non-final syllables Putting aside function wor ds, monosyllables are considerably less common than polysyllables in Japhug; in particular , verbs can be monosyllabic only in one form of their paradigm (3sg Factual Non-past, §21.3.1.1), and inalienably possesse d nouns are always at least disyllabic (§5.1.2). The stem-final syllable (e xcluding inflectional suffixes), in addition to having stress by default, is also generally the part of the w ord with the maximum of phonological contrasts. Since polysyllabic roots are rare , non-final syllables are in the immense majority of cases either prefixes or compound roots. Prefixes have str ong phonotactic constraints: with the sole exception of some orientation prev erbs (§15.1.1.1). Their main vow els are always /ɯ/ (including [i] in contexts where the contrast is neutralize d or quasi-neutralized §3.5.2), /ɤ/ and /a/ , and they cannot have aspirate d or plain voiced consonants. Specific classes of prefixes have e ven stricter constraints: productive nominalization pr efixes can only be obstruents (§16) and verbal derivation prefixes (Chapters 17, 18, 19 and 20) can only contain nine consonants: /m/ , /n/ , /s/ , /z/ , /ɕ/ , /ʑ/ , /r/ , /j/ and /ɣ/ (a subset of the consonants that occur as preinitials, §4.2). Roots occurring as non-final elements of comp ounds do not have such strin- gent constraints on consonants, but due to a process of rhyme reduction (b ound state, §5.4.1), most v owels are conv erted to either /ɯ/ or /ɤ/ . Hence, in non-final syllables, the vo wels are mostly r estricte d to /a/ , /ɯ/ , or /ɤ/ , the latter two being subject to rounding by vow el harmony (§3.3.1.2) or assimilation with /w/ (§4.2.1.1, §4.2.2.4, §14.3.2.7). An example of a typical Japhug word is the v erb form /atɤtɯʑɣɤɣɤ ʑó / (2), which has a vowel other than /a/ , /ɯ/ , or /ɤ/ only in the last syllable. (2) a-tɤ-tɯ-ʑɣɤ-ɣɤ-ʑo irr-pfv-2-refl-ca us-be.light ‘Make yourself light. ’ (fr om 59, §18.3.4.1) V owels other than /a/ , /ɯ/ , or /ɤ/ in non-final syllables of verb and noun stems are only found in four cases. First, there is a handful of synchr onically unanalyzable polysyllabic nominal and verbal stems with /o/ , /u/ or /i/ in non- final syllables, for instance ɴɢoɕna ‘spider’ , ɟuli ‘flute’ or nɯpodɯdi ‘tickle ’ . Sec- ond, compounds whose first element does not undergo bound state can preser ve any vow el (for instance qrormbɯ ‘anthill’ from qro ‘ant’ and rmbɯ ‘heap up ’). Third, Tibetan loanwords, especially in the most recent layers (§3.3.3), ar e not subject to the phonotactic constraints of native nouns and allow any combination of vow els (for instance loŋbutɕʰi ‘ elephant’ from ང་པོ་ཆེ་ glaŋ.po.tɕʰe ‘ elephant’). 7 0 4 Consonant clusters and partial reduplication 4.1 Partial Reduplication A useful test to analyse and classify clusters is partial reduplication (Jacques 2007), a very productive process which can be applied to b oth verb and noun stems and has a variety of morphosyntactic functions depending on the position of the syllable affected by reduplication. Partial reduplication of the first syllable of the verbal comple x can mark the protasis of a conditional construction (§12.4.1.2), iterative coincidence (§12.4.1.3), degree incrementation (§12.4.1.4) and totalitativ e relativization (§12.4.1.5). When the target of partial reduplication is the last syllable of the verb , it can indicate em- phasis (§12.4.1.6, §12.4.3), and also serve as a secondar y e xponent (§ 12.4.2) of se v- eral derivational processes, such as reciprocal (§18.4.1), distributed action (§19.4), auto-evaluative (§19.5), attenuativ e (§19.6), as well as some non-finite v erb forms (gerund §16. 6.1 and purposive converb §16.6.2). Ther e are in addition a few spo- radic cases of reduplication which do not follow the rules described in this chap- ter (§18.6.5, §19.7 .11). Outside of verbal morphology , w e find partial reduplication in comitative ad- verbs (§5.8.1), perlative (§5.8.2) and collectives (§5.7 .9.2). When partial reduplication is applied to a syllable, the rhyme of the redupli- cated syllable is changed to /ɯ/ in the reduplicant (other types of partial redu- plication are tr eate d in §19.4.2). This v owel can undergo rounded assimilation to [u] when the base syllable has the main vow el /u/ or /o/ (§3.3.1.2). Some clusters are only partially copied: when the last consonant of a clus- ter is one of the non-nasal sonorants ( /r/ , /l/ , /j/ , /w/ , /ɣ/ or /ʁ/ ), and the preceding consonant is neither a non-nasal sonorant nor a sibilant or alveolo- palatal fricative , the sonorant is delete d, as in the gerund of mbɣom ‘be in a hurry’ , which yields sɤ- mbɯ ~mbɣom ( example 245, §16.6.1.3) instead of † sɤ- mbɣɯ~mbɣom , or the emphatic form kʰɯ ~kʰro from kʰro ‘much’ (§26.1.1.2), instead of † kʰrɯ~kʰro . 4 Consonant clusters and partial reduplication T able 4.4: List of consonant clusters with /r/ and /ʂ/ as a first element (36+0) /p/ /ʂp/ tɯ-rpa ‘axe ’ /pʰ/ /ʂpʰ/ rpʰɤβrpʰɤβ ‘flapping wings’ /mb/ /rmb/ armbat ‘be near’ /m/ /rm/ rmɤβja ‘peaco ck’ /t/ /ʂt/ rtalu ‘horse year’ /tʰ/ /ʂtʰ/ ɯ-pɤrtʰɤβ ‘between’ /d/ /rd/ rdɤstaʁ ‘stone ’ /nd/ /rnd/ rnde ‘find’ /n/ /rn/ rnaʁ ‘be de ep ’ /ts/ /ʂts/ rtsot ‘vengeance ’ /tsʰ/ /ʂtsʰ/ rtsʰom ‘have a crack’ (bucket) /dz/ /rdz/ rdzardza ‘insolent’ /ndz/ /rndz/ rndzɤkɤŋe ‘shade of the mountain’ /s/ /ʂs/ rsɯβrsɯβ ‘hair y’ /z/ /rz/ tɯ-rzɯɣ ‘ one section’ /tɕ/ /ʂtɕ/ nɯrtɕa ‘tease ’ /tɕʰ/ /ʂtɕʰ/ rtɕʰɯʁjɯ ‘ caterpillar’ /ndʑ/ /rndʑ/ cirndʑi ‘sand’ /ɕ/ /ʂɕ/ rɕɯβrɕɯβ ‘rough’ /ʑ/ /rʑ/ tɤ-rʑaβ ‘wife ’ /c/ /ʂc/ tɤ-rcoʁ ‘mud’ /cʰ/ /ʂcʰ/ ɯ-rcʰarcʰɤβ ‘interstice ’ /ɟ/ /rɟ/ rɟaʁ ‘ dance ’ /ɲɟ/ /rɲɟ/ rɲɟaʁlo ‘bolt’ /ɲ/ /ʂɲ/ ʂɲoʁʂɲoʁ ‘long and thin’ /rɲ/ rɲaŋ ‘be ancient’ /k/ /ʂk/ rko ‘be hard’ /kʰ/ /ʂkʰ/ tɤ-rkʰom ‘feather rachis’ /g/ /rg/ rga ‘like ’ /ŋg/ /rŋg/ rŋgɤm ‘hard piece ’ /ŋ/ /rŋ/ tɯ-rŋa ‘face ’ /ʂŋ/ aʂŋaʁ ‘twist (ankle)’ /q/ /ʂq/ rqoʁ ‘hug’ /qʰ/ /ʂqʰ/ tɤ-rqʰu ‘hull, skin’ /ɴɢ/ /rɴɢ/ ɕirɴɢo ‘ A nisodus tanguticus ’ /χ/ /ʂχ/ ʂχɯʂχi ‘with big nostrils’ 78 4.2 Inventory of consonant clusters T able 4.5: List of consonant clusters with /ɕ/ and /ʑ/ as a first element (18+0) /p/ /ɕp/ ɕpaʁ ‘be thirsty’ /pʰ/ /ɕpʰ/ ɕpʰɤt ‘patch’ (vt) /mb/ /ʑmb/ ʑmbɤr ‘ulcer’ /m/ /ɕm/ ɕmi ‘mix’ /t/ /ɕt/ ɕte ‘ contaminate ’ /tʰ/ /ɕtʰ/ ɕtʰɯz ‘turn towards’ /d/ /ʑd/ ʑdɯɣʑdɯɣ ‘strong, tough’ /n/ /ɕn/ ɕnat ‘heddle’ /ʑn/ ʑ-nɯ-ɕar ‘go and look for it’ /ts/ /ɕts/ ŋɤjɕtsa ‘foreman’ /tɕ/ /ɕtɕ/ sɤɕtɕɯɣ ‘strap ’ (to carr y children on the back) /tʂ/ /ɕtʂ/ ɕtʂaŋlaŋ ‘hanging and swinging’ /k/ /ɕk/ ɕkom ‘muntjac’ /kʰ/ /ɕkʰ/ ɕkʰo ‘spread’ /g/ /ʑg/ ʑgaʁ ‘ exactly’ /ŋg/ /ʑŋg/ ʑŋgu ‘ cross river’ /ŋ/ /ɕŋ/ ɕŋaʁɕŋaʁ ‘bright yellow’ /q/ /ɕq/ ɕqɤjɤr ‘ cross-ey ed’ /qʰ/ /ɕqʰ/ ɕqʰaloʁ ‘latch’ /ɴɢ/ /ʑɴɢ/ ʑɴɢɯloʁ ‘walnut’ ɕtɕ- found in the noun sɤɕtɕɯɣ ‘strap ’ (to carr y children). Ther e is evidence, ho w- ever , that *ɕtsʰ- dissimilate d to jtsʰ- in the irregular causativ e jtsʰi ‘give to drink’ (§ 4.2.1.6, §17 .2.2.5). There are tw o examples of alveolo-palatal preinitials follow ed by a palatal seg- ment: ʑɲ- and ɕcʰ- , but they are not stable and hence not included in T able 4.5. They only occur when the ʑ- and ɕ- allomorphs of the translocative prefix are found before the B-type ɲɯ- and cʰɯ- or the D-typ e ɲɤ- and cʰɤ- ‘westwar d’ and downstream orientation pr everbs. These clusters are always in fr ee varia- tion with zɲ- and scʰ- , respectively , which are considerably mor e common (T a- ble 15.10, §15.2.1.2), since the translocative prefix generally undergoes a palatal dissimilation rule to a dental fricative s- or z- when it precedes a palatal conso- nant. The non-dissimilated clusters ɕcʰ- and ʑɲ- are extr emely rare: the former is only attested in one single example in the corpus (2), and the second in a handful of verb forms ( see for instance 178, §15.2.1.2). These non-dissimilate d forms ar e 79 4 Consonant clusters and partial reduplication undoubtedly due to analogical levelling, which generalized the alveolo-palatal allomorphs of the translocative. (2) a-tʰɯ-ɤ<nɯ>kʰu irr-pfv:downstream-<a ut o>call ɕti be.aff:f a ct ma lnk aʑo 1sg ɕ-cʰɯ-ʑɣɤ-βde-a tral-ipfv:downstream-refl-thr ow-1sg ɕti be.aff:f a ct ma lnk ‘Let her call [me], I am going to throw myself (in the riv er). ’ (22-qajdo) {0003596#S53} Secondar y clusters with alveolo-palatal preinitials ar e created by prefixes with the shape ɕ- or ʑ- , including allomorphs of the translocative as mentioned ab ov e (§15.2.1.2), but also irregular causatives (§17 .2.2.3, §17 .2.2.4) and p erhaps lexical- ized oblique participles (§16.1.3.10). 4.2.1.6 /j/ +C clusters T able 4.6 lists all consonant clusters with the semi-vo wel /j/ as first element not ending in a non-nasal sonorant. It cannot precede alveolo-palatal obstruents and palatal stops. T able 4.6: List of consonant clusters with /j/ as a first element (12+1) /p/ /jp/ jpum ‘be thick’ /m/ /jm/ jmɯt ‘forget’ /t/ /jt/ ajtɯ ‘accumulate ’ /nd/ /jnd/ sɤjndɤt ‘be cute’ /n/ /jn/ jnom ‘be flexible ’ /ts/ /jts/ tɤ-jtsi ‘pillar’ /tsʰ/ /jtsʰ/ jtsʰi ‘give to drink’ /tʂʰ/ /jtʂʰ/ qajtʂʰa ‘ Aegypius monachus ’ /ndʐ/ /jndʐ/ jndʐɤz ‘be thick’ (p ow der) /k/ /jk/ tɤ-jkɯz ‘secret’ /ŋ/ /jŋ/ tɤ-jŋoʁ ‘hook’ /q/ /jq/ jqu ‘be able to lift’ /χ/ /jχ/ ajχoʁ ‘be flat’ ( belly) /jmŋ/ tɯ-jmŋo ‘ dream’ 80 4.2 Inventory of consonant clusters Most of these clusters violate the sonority sequencing principle (§4.2.3.2), but those that are found as the onset of verb r oots (such as jpum ‘be thick’) can o ccur in word-initial position in the Factual Non-Past (§21.3.1). Comparison with Zbu rev eals that one of the sources of the Japhug j- preinitial is proto-Gyalrong *l- (Jacques 2004: 271–272): for instance jmɯt ‘forget’ , tɤ-jtsi ‘pillar’ , tɤ-jme ‘tail’ and tɯ-jmŋo ‘ dream’ correspond to Zbu lmɑt , tɐltsíʔ , -lméʔ and -lmɑʔ , respectively (Gong 2018: 43; 53; 288). In other cases, j- comes from a dissimilated coronal fricative , as in jtsʰi ‘give to drink’ (§17 .2.2.5) and p erhaps jqu ‘be able to lift’ (§19.3.1). The group jmŋ- in tɯ-jmŋo ‘ dream’ has tw o preinitials /j/ and /m/ . The labial nasal is the ancient initial (as sho wn by the Zbu cognate -lmɑ  ʔ ), and the nasal velar arose as a trace of pr oto-Gyalrong vow el velarization *-lmaˠŋ → *-jmɣo → -jmŋo (Jacques 2004: 44; on velarized vo wels in proto-Gyalrong, see §4.2.2.5). When preceding sonorants, /j/ is preinitial in all cases: jw- , jl- , jr- , jɣ- , jʁ- and rʁ- (§4.2.2.1, §4.2.2.3, §4.2.2.4, §4.2.2.5, §4.2.2.6). 4.2.1.7 V elar+C clusters T able 4.7 lists all consonant clusters with the velar fricatives /x/ and /ɣ/ as preinitial and not ending in a non-nasal sonorant. V elar preinitials are not compatible with dorsal ( velar and v elar) initial con- sonants, but are attested with all other places of articulations. The complemen- tary distribution b etween x (b efore unvoiced segments) and ɣ ( before voiced seg- ments, including obstruents and nasals) is complete, and the v oicing contrast can be considered to b e completely neutralized in preinitial position. Comparison with Situ shows that some of these velar fricativ e preinitials origi- nate from velar pr esyllables *kə- (Jacques 2014b: 6). T wo non-productive prefixes are sources of v elar+C clusters: the animal velar prefix (§5. 6.2) and the lexicalized participle x/ ɣ- prefix (§16.5.2). Another source of secondar y velar pr einitials is the intrusive ɣ found as element of some derivation prefixes, including the sɯɣ- (§17 .2.1.4) and ɕɯɣ- (§17 .2.2.2) allomorphs of the causative , the nɯɣ- allomorph of the applicative (§17 .4.2), the nɤɣ- allomorph of the estimative (§17 .5.1) and the sɤɣ- allomorph of the proprietive (§18.8.1). In addition, the velar preinitial in some counte d nouns like tɯ-xpa ‘ one year’ is due to a fausse coupe (§7 .3.1.7). 4.2.1.8 Uvular+C clusters T able 4.8 lists all consonant clusters with the uvular fricatives /χ/ and /ʁ/ as first element and not ending in a non-nasal sonorant. They cannot precede velars 81 4 Consonant clusters and partial reduplication T able 4.7: List of consonant clusters with /x/ and /ɣ/ as a first element (23+0) /p/ /xp/ tɯ-xpa ‘one y ear’ /mb/ /ɣmb/ tɯ-ɣmba ‘ cheek’ /m/ /ɣm/ tɯ-ɣmaz ‘wound’ /t/ /xt/ xtɯt ‘wild cat’ /tʰ/ /xtʰ/ xtʰom ‘put horizontally’ /d/ /ɣd/ ɣdɤso ‘spe cies of grub ’ /nd/ /ɣnd/ ɣnda ‘ram’ /n/ /ɣn/ ɣnɤsqi ‘twenty’ /ts/ /xts/ xtsɤɕna ‘tip of the sho e ’ /tsʰ/ /xtsʰ/ xtsʰɯm ‘b e thin’ /s/ /xs/ xsar ‘goral’ /z/ /ɣz/ ɣzɯ ‘monkey’ /tɕ/ /xtɕ/ xtɕi ‘be small’ /tɕʰ/ /xtɕʰ/ xtɕʰɯt ‘ can contain’ /ndʑ/ /ɣndʑ/ ɣndʑɤβ ‘ disastrous fire ’ /ɕ/ /xɕ/ xɕaj ‘grass’ /ʑ/ /ɣʑ/ ɣʑo ‘be e ’ /tʂ/ /xtʂ/ nɤxtʂɯ ‘bring in passing’ /ʂ/ /xʂ/ xʂɤxʂɤt ‘long and thin’ /c/ /xc/ xcat ‘be many’ /cʰ/ /xcʰ/ tɤlɤxcʰi ‘fresh milk’ /ɟ/ /ɣɟ/ ɣɟaβ ‘ churn’ /ɲ/ /ɣɲ/ ɯ-ɣɲaʁ ‘ disaster’ except in heterosyllabic clusters (§4.2.3.1). The voicing contrast between /χ/ and /ʁ/ is only attested b efore sonorants: a handful of ideophones have the cluster /χɲ/ contrasting with /ʁɲ/ . Uvular preinitials are particularly common in Tibetan loanwor ds, where they correspond to the g- and d- sŋon.ⁿdʑug . The voiced fricative /ʁ/ is more often realized as an epiglottal [ʕ] in this context. Some uvular preinitials come fr om the reduced allomorphs χ- and ʁ- of the uvular animal prefix qa- (§5.6.1). 82 4.2 Inventory of consonant clusters T able 4.8: List of consonant clusters with /χ/ and /ʁ/ as a first element (25+0) /p/ /χp/ χpi ‘stor y’ /pʰ/ /χpʰ/ taχpʰe ‘slap ’ /b/ /ʁb/ ʁbɤʁbɤβ ‘thick and big’ /mb/ /ʁmb/ aʁmbɯm ‘be concave ’ /m/ /ʁm/ ʁmaʁ ‘army’ /t/ /χt/ χtɤrma ‘ offerings’ /tʰ/ /χtʰ/ naχtʰɤβ ‘seize the opportunity’ /d/ /ʁd/ ʁdɯɣ ‘umbrella’ /nd/ /ʁnd/ ʁndɤr ‘be scattered’ (anticausative) /n/ /ʁn/ ʁnaʁna ‘both’ /ts/ /χts/ χtso ‘it is clean’ /tsʰ/ /χtsʰ/ χtsʰɤχtsʰɤt ‘small and active ’ /ndz/ /ʁndz/ ʁndzɤr ‘ cut’ (with scissors) /s/ /χs/ χsɤr ‘gold’ /z/ /ʁz/ ʁzɤβ ‘be careful’ /tɕ/ /χtɕ/ χtɕoŋ ‘rheumatism’ /ɕ/ /χɕ/ χɕu ‘be strong’ /ʑ/ /ʁʑ/ ʁʑɯnɯ ‘young man’ /tʂ/ /χtʂ/ χtʂɯɣdʑa ‘butter tea’ /ʂ/ /χʂ/ χʂɤχʂɤt ‘light ( clothes)’ /c/ /χc/ χcoŋkroŋ ‘ cross-legged ( sitting)’ /cʰ/ /χcʰ/ χcʰa ‘right’ /ɟ/ /ʁɟ/ ʁɟa ‘ completely’ /ɲɟ/ /ʁɲɟ/ ʁɲɟiʁɲɟi ‘ enormous’ /ɲ/ /χɲ/ χɲɤχɲɤr ‘without energy’ /ʁɲ/ ʁɲɤrpa ‘steward (monastery)’ 83 4 Consonant clusters and partial reduplication 4.2.1.9 Nasal+C clusters The T ables 4.9 and 4.10 list all consonant clusters with a nasal consonant as preini- tial and without medial consonant. Unlike the previous pr einitials, nasals can never be directly followed by any non-nasal sonorant: groups such as *mj- , *mw- , *mr- , *ml- , *mɣ- and *mʁ- are not attested in onset position (they are, how ever , found as heterosyllabic clusters, §4.2.3.1). This gap is due to a combination of two sound changes. First, the medials *j and *ɣ were nasalized when preceded by a nasal, so that pre-Japhug *mj- and *mɣ- became mɲ- (in tɯ-mɲa ‘arrow’ , compare མདའ་ mda ‘arrow’ fr om *mla , Hill 2019: 18) and mŋ- (in tɯ-jmŋo ‘ dream’ , see §4.2.1.6), re- spectively . Second, pre-Japhug *mr- b ecame mbr- by epenthesis, as shown by mbro ‘horse ’ , mbro ‘be high’ and mbri ‘cry , sing’ (in Burmese mraṅḥ , mraṅʔ and mraññ , Hill 2019: 60;259). Additionally , nasal preinitials cannot be followed by any fricative. A s in Ti- betan (Li 1933), this absence is due to the conversion of post-nasal fricatives to the corresponding affricates, as in tɯ-mtsʰi ‘liver’ (from *m-si , Burmese asaññḥ , Hill 2019: 56). T able 4.9: List of consonant clusters with a homorganic nasal as first element (14+1) /p/ /mp/ mpɯ ‘be soft’ /pʰ/ /mpʰ/ mpʰɯl ‘reproduce ’ /t/ /nt/ ntaβ ‘b e stable ’ /tʰ/ /ntʰ/ ntʰɤβ ‘be caught b etween’ /ts/ /nts/ ntsɯ ‘always’ /tsʰ/ /ntsʰ/ ntsʰɤr ‘neigh’ (of horse) /tɕʰ/ /ntɕʰ/ ntɕʰoz ‘use ’ /tʂ/ /ntʂ/ ntʂu ‘weed with a hoe’ ( vt) /c/ /ɲc/ ɲcɤr ‘press’ /cʰ/ /ɲcʰ/ ɲcʰoʁ ‘shrink’ /k/ /ŋk/ ŋke ‘walk’ /kʰ/ /ŋkʰ/ ŋkʰor ‘arrive ’ /q/ /ɴq/ ɴqa ‘be difficult’ /qʰ/ /ɴqʰ/ ɴqʰi ‘be dirty’ /mpɕ/ mpɕɤr ‘be b eautiful’ 84 4.2 Inventory of consonant clusters While voiced stops and affricates with homorganic prenasalization are mono- phonemic (§3.2.1), unaspirated and aspirate d ones (T able 4.9) are better analyzed as clusters. All places of articulation are possible, fr om labial ( mp(ʰ)- ) to uvular ( ɴq(ʰ)- ). The special case of mpɕ- is discusse d in §4.2.1.10. T wo non-homorganic nasal preinitials are also attested, /m-/ before all places of articulation except labials ( where m- corresponds to the homorganic nasal preinitial), and n- before ( voiced) labials and velars (T able 4.10). When precede d by /m/ or /n/ , the contrast between plain voiced and prenasalized voiced stops and affricates is neutralized; for instance the b in nb- stands for the archiphoneme { b,mb} . In the case of the cluster mɢ- , since the prenasalized phoneme /ɴɢ/ lacks a plain voiced counterpart (§3.2.1), the only analysis possible is /mɴɢ-/ . In the orthography used in this grammar , the simplified ( but non-ambiguous) transcription mɢ- is used. The other nasal phonemes /ɲ/ and /ŋ/ are not attested as non-homorganic preinitials: for instance , clusters such as † ŋp- or † ɲt- are prohibited. Some m- preinitials come from the nasalization of *w- before nasal or prenasal- ized stops (as in the irregular causativ e mɲo ‘prepare ’ from ɲo ‘be prepared’ , see §17 .3.1), a sound law accounting for the absence of preinitial /w/ in this context (§4.2.1.1). This sound law does not operate across syllables (§4.2.3.1), and hetero- syllabic clusters with a non-nasalized coda -β followed by a nasal do exist. Some n- preinitials originate from former *t- b efor e nasal or prenasalized stops, as in tɯ-nŋa ‘ debt’ (§4.2.1.9; see also Jacques 2014b). Another source is a reduced n- allomorph of denominal nɯ- (as in ngo ‘be ill’ from the noun tɯ-ŋgo ‘ disease ’ , §20.7 .1). 4.2.1.10 Stop+ ɕ clusters A handful of onsets have an unvoiced stop follow ed by ɕ , including kɕ- (as in kɕilu ‘year of the dog’ fr om ི་ལོ་ kʰʲi.lo ‘year of the dog’ , pɕ- , ( tɯpɕi ‘flax’) as well as lpɕ- (§4.2.1.3) and mpɕ- (§4.2.1.9). These clusters, found in both native vocabulary and borrowings from Tibetan, mainly originate from the combination of aspirated stops pʰ and kʰ with the medial -j- . Evidence that these groups should not be synchronically analyzed as /kʰj-/ and /pʰj/ underlyingly comes from two facts. First, the gr oups pʰj- and kʰj- are otherwise atteste d (§4.2.2.2). Second, the clusters (C)(p|k)ɕ- are not affecte d by partial reduplication; for instance mpɕɤr ‘be b eautiful’ is r eduplicate d as mpɕɯ~mpɕɤr (see for instance 304, §21.8.3.2), not † mpʰɯ~mpɕɤr as would hav e b een expecte d if mpɕɤr were r eally † /mpʰjɤr/ . 2 2 In addition, mpɕɤr ‘be b eautiful’ is pr obably b orro wed from མཆོར་ པོ་ mtɕʰor.po ‘handsome ’ , so that the analysis as /mpʰj/ is wrong ev en historically . 85 4 Consonant clusters and partial reduplication T able 4.10: List of consonant clusters with a non-homorganic nasal as first element (24+0) /t/ /mt/ tɤ-mtɯ ‘knot’ /tʰ/ /mtʰ/ mtʰɯ ‘spell’ /nd/ /md/ mda ‘be the time’ /n/ /mn/ mna ‘be b etter’ , ‘heal’ /ts/ /mts/ tɤ-mtsɯ ‘button’ /tsʰ/ /mtsʰ/ mtsʰɤm ‘hear’ /ndz/ /mdz/ mdzadi ‘flea’ /tɕ/ /mtɕ/ mtɕoʁ ‘be sharp’ /tɕʰ/ /mtɕʰ/ tɤ-mtɕʰo ‘wedge’ /ndʑ/ /mdʑ/ tɯ-mdʑu ‘tongue ’ /tʂ/ /mtʂ/ mtʂɤkʰoz ‘bib ’ /ndʐ/ /mdʐ/ mdʐɯɕɯɣ ‘be dbug’ /c/ /mc/ tɤmcar ‘tongs’ /cʰ/ /mcʰ/ tɯ-mcʰi ‘gall’ /ɲɟ/ /mɟ/ tɯ-mɟa ‘jaw’ /ɲ/ /mɲ/ mɲɤm ‘spe cies of tree ’ /k/ /mk/ tɯ-mke ‘neck’ /kʰ/ /mkʰ/ mkʰɤz ‘b e expert’ /ŋg/ /mg/ tɯ-mga ‘advantage ’ /ŋ/ /mŋ/ mŋɤm ‘feel pain’ /ɴɢ/ /mɢ/ tamɢom ‘ clamp ’ /mb/ /nb/ anbaʁ ‘hide ’ /m/ /nm/ tɤ-nmaʁ ‘husband’ /ŋg/ /ng/ ngɯt ‘b e strong’ , ‘be resistant’ /ŋ/ /nŋ/ nŋo ‘lose ’ 86 4.2 Inventory of consonant clusters 4.2.2 Medials This section deals with consonant clusters whose last consonant is one of the six non-nasal sonorants ( /r/ , /l/ , /j/ , /w/ , /ɣ/ or /ʁ/ ). These sonorants can either be me dials (§4.1) or initials, depending on whether they are deleted when the cluster is partially reduplicated. They are always medials when the preced- ing consonant is not one of the possible preinitials (§4.2.1). Clusters ending in a non-nasal sonorant whose penultimate element is a possible preinitial (sibilant/ alveolo-palatal fricative or non-nasal sonorant) 3 are r eferred to as ambiguous clusters , as they need to b e systematically tested with partial reduplication to determine the phonological status of the final sonorant (initial or medial). The secondar y clusters with medial glides created by the effect of synizesis (§3.3.1.3) with the 1sg suffix -a (§14.2.1.1) are not included in the listing below , even though what is transcribed as Ce| i-a and Co|u-a in the orthography is ho- mophonous with Cja and Cwa , respectively . Including them would hav e the ef- fect of introducing an enormous number of se condary and predictable clusters. In addition, synizesis of Cɯ-a generates the [ɰ] allomorph of /ɯ/ , constituting a seventh glide not found in the v ocabular y . 4.2.2.1 C+ /w/ clusters T able 4.11 lists all clusters ending in /w/ . Clusters of this type are very rare, as a combination of two sound changes hav e removed all instances of proto-Gyalr ong *Cw . First, the labiovelars, still pr eser ved in Zbu, have merged with plain velars (for instance nɯŋa ‘ cow’ , Zbu ŋwéʔ , Gong 2018: 40), and the remaining instances of *w have become ɣ (as in tɯ-ɕɣa ‘tooth’ from *-ɕwa , §4.2.2.5). There ar e no clusters with labial consonants followed by w , but there is a triple contrast between hw- , χw- and xw- : it is the only context where minimal pairs between the glottal /h/ , the uvular /χ/ and the velar /x/ unvoiced fricatives can be found. The combination of cluster-final /w/ with the rhyme -a is homophonous with the result of synizesis of the 1sg -a with v erb stems in -o and -u ; for instance, the 1sg possessive a-rwa ‘my tent’ is not distinguishable from aro-a ‘I own’ (§3.3.1.3). Both are phonetically realized as [arwa] . Clusters in w are mainly found in ideophones and Chinese loanwor ds (includ- ing nativized ones such as kwitsɯt ‘ cupb oard’ fr om 柜子 <guìzǐ> ‘ cupb oar d’ with 3 Nasal sonorants cannot directly precede non-nasal sonorants in syllable onset position, due to a series of sound changes (§4.2.1.9), so they are not possible preinitials when the initial is a non-nasal sonorant. 87 4 Consonant clusters and partial reduplication In the ambiguous clusters with a sonorant as first element ( /wr/ , /jr/ , /ɣr/ and /ʁr/ ), /r/ is initial, as shown by the perlative ɯ-jrɯ~jroʁ ‘following X’s trace ’ (§ 5.8.2), the emphatic participle kɯ-wɣrɯ~wɣrum ‘very white (one)’ (98, §5.5.1.2) and the reciprocal a-zɣɤʁrɯ~ʁre ‘respect each other’ from zɣɤʁre ‘re- spect’ . By contrast, in ambiguous clusters with a coronal fricative as first element ( /sr/ , /zr/ , /ɕr/ , /ʑr/ ), /r/ is medial, as illustrate d by the emphatic participles kɯ-zɯ~zri ‘very long (one)’ (in 100, §23.5.10.1), kɯ-ʑɯ~ʑru ‘v er y strong ( one)’ and kɯ-ɣɤχsɯ~χsrɯ ‘very handsome (one)’ (4). 5 T shendzin rejects reduplica- tion of these clusters without deletion: † kɯ-zrɯ~zri and † kɯ-ʑrɯ~ʑru are not grammatical. (4) ʁʑɯnɯ young.man kɯ-ʑɯ~ʑru sbj:pcp-emph~strong ʑo emph kɯ-ɣɤχsɯ~χsrɯ sbj:pcp-emph~handsome ʑo emph ɲɤ-k-ɤβzu-ci ifr-peg-be come-peg ‘(The fox) became a strong and handsome young man. ’ (2012 qachGa) {0004087#S183} The /w/ preinitial in the group wɣr- is not r ealize d as a separate segment (§4.2.1.1). It is an orthographic device used to indicate that the preceding back unrounded vow els /ɯ/ and /ɤ/ turn into their rounded counterparts when ad- jacent. For instance, r e duplicated participle kɯ-wɣrɯ~wɣrum ‘very white (one)’ is phonetically realized as [kuɣruɣrum] . An extra-short svarabhakti vo wel can be heard in some clusters with -r- medial, in particular after coronal affricates and dorsal stops. For instance , cʰɤ-kro ‘s/he distributed it’ can b e r ealize d as [cʰɤkəro] . The retrofle x affricates originate at least in part from clusters of dental stops followed by r ; direct evidence for the sound change *tr- → tʂ- comes from al- ternations between /tʂ/ and /r/ , as that b etween the numeral kɯtʂɤɣ ‘six’ and sqaprɤɣ ‘sixteen’ (§7 .1.3) and between tʂu ‘road’ and ftɕɤru ‘path in the middle of the fields’ is a compound of ftɕar ‘summer’ and tʂu ‘road’ (§5.4.3.2). The gap in the system caused by this sound change has be en filled by some ideophones in dr- (§10.1.5.1); the only non-ideophone with a cluster of this typ e is qɯmdroŋ ‘ crane ’ , whose onset can either b e analyzed as /dr/ or as /ndr/ , as the contrast between /d/ and /nd/ is neutralized due to assimilation with the /m/ coda of the preceding syllable. 5 Reduplication of the simple /sr/ cluster in the comitative adverb kɤ  -sɯ~srɯn ‘together with (its) cotton wool’ (§5.8.1) confirms that /r/ is medial in /sr/ , not only in /χsr/ . 94 4.2 Inventory of consonant clusters T able 4.17: List of consonant clusters with three elements ending in /r/ with a non-nasal preinitial (32) /wkr/ fkrɯz ‘be gree dy’ /wɣr/ wɣrum ‘be white’ /wsr/ fsraŋ ‘protect’ /spr/ sprɯskɯ ‘reincarnated’ /zbr/ zbrilu ‘year of the snake ’ /zmbr/ sɤzmbrɯ ‘anger’ ( vt) /stʰr/ stʰrɯβ ‘ dangling (of snot)’ /scr/ scraʁscraʁ ‘very small’ /zɟr/ zɟraŋzɟraŋ ‘soft and bloated’ /skr/ skraskra ‘impolite ’ /skʰr/ tɯ-skʰrɯ ‘bo dy’ /zgr/ zgrawa ‘leather sack’ /sqr/ sɤsqra ‘limit’ /ɕpr/ aɕprɯm ‘be badly sewed’ /ʑmbr/ ʑmbri ‘willow’ /ɕtr/ ɕtraŋɕtraŋ ‘long and soft’ /ʑdr/ ʑdraŋʑdraŋ ‘long and soft’ /ɕkr/ ɕkrɤz ‘ oak’ /ʑgr/ ʑgrɯɣ ‘ certainly’ /ʑŋgr/ ʑŋgri ‘star’ /ɕqr/ ɕqraʁ ‘be intelligent’ /ʑɴɢr/ ʑɴɢro ‘Jew’s harp ’ /jkr/ jkrɯt ‘ congeal’ /jtsr/ jtsraβ ‘ delay departure ’ /xpr/ ɣɤxpra ‘ dispatch’ /χpr/ tɕʰɯχpri ‘salamander’ /ʁmbr/ taʁmbra ‘jump ’ (of horse) /χsr/ ɣɤχsrɯ ‘be handsome’ /ʁzr/ ʁzraŋʁzraŋ ‘ dishevelled’ /χcr/ χcɯχcri ‘thin, diluted’ /ʁɟr/ ʁɟɯʁɟri ‘fat and soft’ /ʁgr/ ʁgra ‘ enemy’ 95 4 Consonant clusters and partial reduplication T able 4.18: List of consonant clusters with three elements ending in /r/ with a nasal preinitial (9) /ɲcr/ ɲcɯɲcri ‘thin, diluted’ /ŋkʰr/ ŋkʰrɯli ‘screw’ /ngr/ ngrɯβ ‘accomplish’ /ɴqr/ ɯ-ɴqra ‘broken one ’ /mtsr/ mɯmtsrɯɣ ‘ drink with a straw’ /mpʰr/ mpʰrɯmɯ ‘ divination’ /mkʰr/ mkʰroŋ ‘be reincarnated’ /mgr/ mgrɯn ‘treat’ , ‘invite’ /nbr/ nbraʁ ‘loosen the earth’ (vt) In triconsonantal clusters (T ables 4.17 and 4.18), the medial /r/ is compatible with all preinitials except itself. 4.2.2.5 C+ /ɣ/ clusters The T ables 4.19 and 4.20 list all clusters ending in /ɣ/ . The velar fricative can follow all places of articulations except r etroflex affricates, v elars and uvulars. In ambiguous clusters, the /ɣ/ is initial in /zɣ/ and when precede d by a non- nasal sonorant ( /wɣ/ , /jɣ/ , /lɣ/ and /rɣ/ , as illustrated by the distribute d action derivation (§19.4) nɤ-lɣɯ~lɣa ‘ dig every wher e’ fr om lɣa ‘ dig’ (not † nɤ-lɯ~lɣa ). The velar sonorant is medial in /ɕɣ/ , as shown by the comitative adv erb (§5.8.1) kɤ  -ɕɯ~ɕɣa ‘together with its teeth’ (not † kɤ  -ɕɣɯ~ɕɣa ) from tɯ-ɕɣa ‘tooth’ . 6 The group /ʑɣ/ cannot be tested with partial reduplication. Cluster-final ɣ is secondar y , and has at least thr ee proto-Gyalrong origins. First, it comes from proto-Gyalr ong medial *-w- , as in lɣa ‘dig’ (Situ rwâ , Huang & Sun 2002) or tɯ-tɣa ‘ one span’ (Situ tə-təwá ; the Tibetan cognate མཐོ་ mtʰo ‘span’ underwent Laufer’s law , Jacques 2009; Hill 2011b). Most *-w- medi- als in the inherited vocabular y have undergone this sound change ( except after dorsals, where they hav e disappeared), and the present w medials are secondar y (§4.2.2.1). Second, it originates from the lenition of velar stops in clusters with tw o stops in proto-Gyalrong, as in pɣa ‘bir d’ (from *pk- , see Cogtse Situ pká , Huang & Sun 2002). Third, it is a secondary trace of velarized vowels (Jacques 2004: 231), as in tɤjpɣom ‘ice ’ (from *tɐ-lpaˠŋ , see Zbu tɑlvɑmʔ , Gong 2018: 13). 6 In Jacques (2004: 59), I claimed that /ɕɣ/ could b e r e duplicated either with or without deletion of /ɣ/ , but in my more recent data only the variant with deletion is attested. 96 4.2 Inventory of consonant clusters T able 4.19: List of consonant clusters with two elements ending in /ɣ/ (25) /p/ /pɣ/ pɣa ‘bird’ /pʰ/ /pʰɣ/ pʰɣo ‘flee’ /b/ /bɣ/ sɯbɣi ‘spe cies of bush’ /mb/ /mbɣ/ mbɣaʁ ‘turn over’ ( vi) /w/ /wɣ/ βɣa ‘mill’ /t/ /tɣ/ tɯ-tɣa ‘ one span’ /tʰ/ /tʰɣ/ tʰɣe ‘acorn’ /d/ /dɣ/ dɣɤrdɣɤr ‘ dumb ’ /nd/ /ndɣ/ ndɣɤndɣɤt ‘shaking’ /ts/ /tsɣ/ tsɣi ‘rot’ /ndz/ /ndzɣ/ tɯ-ndzɣi ‘fang’ /s/ /sɣ/ sɣa ‘rust’ /z/ /zɣ/ zɣɯt ‘arrive ’ /l/ /lɣ/ lɣa ‘ dig’ /tɕ/ /tɕɣ/ tɕɣaʁ ‘squeeze out’ /tɕʰ/ /tɕʰɣ/ tɕʰɣaʁtɕʰɣaʁ ‘ completely’ /ndʑ/ /ndʑɣ/ ndʑɣaʁ ‘be sque ezed out’ /ɕ/ /ɕɣ/ tɯ-ɕɣa ‘tooth’ /ʑ/ /ʑɣ/ ʑɣɤ- reflexiv e prefix /r/ /rɣ/ tɯ-rɣi ‘see d’ /ʂ/ /ʂɣ/ ʂɣɤlʂɣɤl ‘transparent and round’ /cʰ/ /cʰɣ/ qacʰɣa ‘fox’ /ɲɟ/ /ɲɟɣ/ ɲɟɣɤrɲɟɣɤr ‘plump and huge ’ /j/ /jɣ/ jɣɤt ‘turn around’ 97 4 Consonant clusters and partial reduplication T able 4.20: List of consonant clusters with three elements ending in /ɣ/ (21) /spɣ/ spɣi ‘storehouse ’ /zbɣ/ tɤkɤzbɣaʁ ‘headache ’ /stɣ/ stɣɤrnɤstɣɤr ‘jumping’ /lcʰɣ/ lcʰɣaʁlcʰɣaʁ ‘nice to wear’ /ldzɣ/ stoʁldzɣɤm ‘straw from broad beans’ /ɕpɣ/ ɕpɣo ‘ten bushels’ /ɕpʰɣ/ ɕpʰɣo ‘flee with’ /jmbɣ/ nɤjmbɣom ‘have vertigo ’ /jpɣ/ jpɣom ‘freeze ’ /jndɣ/ nɯjndɣo ‘ echo’ ( vi) /rmbɣ/ tɤ-rmbɣo ‘ drum’ /rpɣ/ rpɣo ‘up on the mountain’ /rpʰɣ/ ɯ-rpʰɣɤt ‘upper door frame’ /χpɣ/ tɯ-χpɣi ‘thigh’ /ʁmbɣ/ ʁmbɣi ‘sun’ /mpʰɣ/ mpʰɣaʁmpʰɣaʁ ‘very tight’ /ntɕʰɣ/ ntɕʰɣaʁ ‘splash’ /ntʰɣ/ antʰɣar ‘bounce’ /ntsɣ/ ntsɣe ‘sell’ /ntsʰɣ/ nɤntsʰɣɤz ‘bump into ’ /ɲcɣ/ ɲcɣɤɲcɣɤt ‘many people, very noisy’ /ɲcʰɣ/ ɲcʰɣaʁ ‘birchbark’ In triconsonantal clusters (T able 4.20), medial ɣ can o ccur with all preinitials except velar fricativ es and the labial w- and m- , a further clue of its diachronic origin from *w . Combinations /-w.Cɣ-/ are possible, ho wev er , in heterosyllabic clusters (§4.2.3.1), as in the noun laβzɣi ‘steamed turnip ’ . 4.2.2.6 C+ /ʁ/ clusters Clusters with ʁ as last element are listed in T able 4.21. The uvular fricative is me- dial only in a handful of clusters with dental or alveolo-palatal affricates, where it contrasts with ɣ , as shown by the minimal pair tɯ-ndzʁi ‘ collar bone’ vs. tɯ- ndzɣi ‘fang’ . In ambiguous clusters, with the dental fricative /zʁ/ and non-nasal sonorants ( /wʁ/ , /lʁ/ , /rʁ/ and /jʁ/ ), /ʁ/ is always initial, as illustrated by rɤβʁɯ~βʁa from rɤβʁa ‘roar’ , kɯ-rʁɯ~rʁom ‘v er y rough ( one)’ from rʁom ‘be rough’ and kɯ-ɤjʁɯ~jʁu ‘very b ent ( one)’ from ajʁu ‘be bent’ . 98 4.2 Inventory of consonant clusters T able 4.21: List of consonant clusters ending in /ʁ/ (8) /w/ /wʁ/ βʁa ‘prevail, win’ /ndz/ /ndzʁ/ tɯ-ndzʁi ‘ collar b one ’ /z/ /zʁ/ zʁɤɲcɯ ‘sling’ /l/ /lʁ/ lʁa ‘gunny bag’ /tɕ/ /tɕʁ/ tɕʁɯznɤtɕʁɯz ‘ crunchy’ /tɕʰ/ /tɕʰʁ/ tɕʰʁɯznɤtɕʰʁɯz ‘ crunchy’ /r/ /rʁ/ rʁom ‘be rough’ /j/ /jʁ/ ajʁu ‘be b ent’ ( of legs, trees, etc) One of the origins of cluster-final ʁ is the result of the lenition of the uvular stop *q in double stop clusters, as in βʁa ‘prevail, win’ (fr om *pq- , Jacques 2004: 330, as shown by the Situ cognate pkâ , Huang & Sun 2002: 603). 4.2.3 Summary T able 4.22 summarizes the numb ers of clusters identified in §4.2.1 and §4.2.2. Since in ambiguous clusters (§4.2.2) partial reduplication is not always available to determine whether the final non-nasal sonorant is medial or initial, no attempt is made at distinguishing between [initial+me dial] and [preinitial+initial] clus- ters in this table. Clusters with four elements can appear at least at the phonetic level if the last syllable of a verb stem with a triconsonantal onset undergoes synizesis with the 1sg -a suffix (§3.3.1.3). The denominal verb ɣɤjmŋo ‘ dream of’ (§20.5.2) from the noun tɯ-jmŋo ‘ dream’ (§4.2.1.6) pr ovides examples like pjɤ  -wɣ-ɣɤ-jmŋo-a (ifr- inv-denom-dream-1sg) ‘s/ he dreamed of me ’ realized as [pjó.ɣe.jmŋwa] with a complex onset [jmŋw] comprising four segments. Attested ambiguous clusters are listed in T able 4.23. Grey shading indicates phonotactically impossible combinations, due in particular to the constraints against the combination of velars and uvulars (§4.2.2.5, §4.2.2. 6), and against the combination of alveolo-palatals with /j/ (§4.2.1.6, §4.2.2.2). Orange colour indi- cates [initial+medial] clusters, whose final sonorant is removed in partial redu- plication. Blue colour marks [preinitial+initial] clusters, whose final sonorant is unaffected by reduplication. Clusters left in white are those for which partial reduplication cannot be tested. The cluster rw has two possible reduplication patterns (§4.2.2.1) and is thus left unmarked. 99 4 Consonant clusters and partial reduplication T able 4.22: Count of consonant clusters type CC CCC total wC 15 8 23 s/zC 23 0 23 lC 17 1 18 ʂ/rC 36 0 36 jC 13 1 14 ɕ/ʑC 20 0 20 x/ɣC 23 0 23 χ/ʁC 25 0 25 NC 14 1 15 m/nC 25 0 25 Cɕ 2 2 Cw 10 0 10 Cj 20 19 39 Cl 19 12 31 Cr 25 41 66 Cɣ 25 21 46 Cʁ 8 0 8 total 320 104 424 T able 4.23: Ambiguous clusters /w/ /j/ /r/ /l/ /ɣ/ /ʁ/ /ɕ/ /ɕr/ /ɕl/ /ɕɣ/ /ɕʁ/ /ʑ/ /ʑr/ /ʑl/ /ʑɣ/ /s/ /sj/ /sr/ /sl/ /sɣ/ /z/ /zw/ /zj/ /zr/ /zl/ /zɣ/ /zʁ/ /l/ /lw/ /lj/ /lɣ/ /lʁ/ /r/ /rw/ /rj/ /rl/ /rɣ/ /rʁ/ /w/ /wj/ /wr/ /wl/ /wɣ/ /wʁ/ /j/ /jw/ /jr/ /jl/ /jɣ/ /jʁ/ /ɣ/ /ɣj/ /ɣr/ /ɣl/ /ʁ/ /ʁj/ /ʁr/ /ʁl/ 100 4.2 Inventory of consonant clusters Even though some of the clusters cannot be subjecte d to testing with partial reduplication, the data in T able 4.23 suggest the following rules: • If the first element of the ambiguous cluster is an alveolo-palatal fricative ( /ɕ/ , /ʑ/ ), the following sonorant is medial . • If the first element of the ambiguous cluster is a glide ( /w/ , /j/ ) or a dorsal sonorant ( /ɣ/ , /ʁ/ ), the following sonorant is initial . • If the first element of the ambiguous cluster is a dental fricative ( /s/ , /z/ ), the rhotic /r/ or the lateral /l/ , the following sonorant is medial if it is a glide or /r/ , and it is initial if it is /l/ or a dorsal sonorant. 4.2.3.1 Heterosyllabic clusters In addition to the clusters attested in onset position describ ed in this chapter , another type of consonant clusters is found whenever a closed syllable is in non- final position in the word. These heter osyllabic clusters, whose first element is the coda of the first syllable, present considerably few er phonotactic constraints than onset clusters. The most obvious difference is the fact that, while a strict pr ohibition against having the same segment as preinitial and a medial is observed in onset clus- ters, this constraint does not apply in heterosyllabic ones. For instance, the noun tɕʰɤrprɯ ‘rain shelter’ , a compound from Tibetan ཆར་ tɕʰar ‘rain’ and the native wor d tɤprɯ ‘rain shelter’ , has the sequence -r.pr- , which would be completely impossible as syllable onset (§3.4.1). The phoneme /w/ , realized as -β in coda p osition (§3.2.2), does not become an unvoiced fricative when follow ed by an unvoiced obstruent in the next syllable, as in the noun slɤβkʰaŋ ‘school’ , which is not realized as † slɤfkʰaŋ as would have been expe cted if the /w/ wer e in preinitial position (§4.2.1.1). The coda -β is generally deleted when it precedes labial stops, esp ecially the voiced /b/ in ideophones. For instance, the pattern II ideophone (§10.1.2.2) from the root |bɤβ| is bɤbɤβ ‘stubborn, bulky’; the alternative realization bɤβbɤβ is also possible. Unlike /w/ in preinitial position (§4.2.1.9), the coda -β does not assimilate with nasals or prenasalized onsets that follow it. For instance, the coda -β in the last syllable of verb stems is not nasalized by the dual -ndʑi or plural -nɯ indexa- tion suffixes (§14.2.1.2): pjɯ-fkaβ-nɯ (ipfv-cover-pl) ‘the y cover it’ ( example 163, §15.1.5.11) cannot be realized as † [pjifkámnɯ] . The same is true in noun com- pounds; for example jlɤβndʑu ‘weft stick’ from tɯ-jlɤβ ‘w eft’ and ndʑu ‘stick’ . 101 4 Consonant clusters and partial reduplication A puzzling cluster -p.t- is obser v e d in the numeral sqaptɯɣ ‘ elev en’ (§7 .1.3). The -p- is a linking element between the b ound form sqa- ‘ten’ and the follow- ing numeral, also found in other Gyalrongic languages (Jacques 2017 c), but its expected form would be the preinitial f- (from /w/ ) in this position. It is b etter analyzed as the coda of the first syllable. The dental /t/ , which exists as coda (§3.2.2), but not as preinitial (§4.2.1), is at- tested in heterosyllabic clusters follow e d by unv oiced labial or velar stops, mainly in Tibetan loanwords such as rgɤtpu ‘ old man’ (from ད་པོ་ rgad.po ‘old man’) or mtɕʰɤtkʰo ‘house shrine ’ (on which se e §3.3.3). In compounds, the coda /t/ is either lost in non-final syllables (§5.4.2.2), or nasalizes to [n] if the following segment is a nasal sonorant or a prenasalized obstruent (§5.4.2.1), in particular with the dual -ndʑi and plural -nɯ indexation suffixes (§14.2.1.2), so that ev en heterosyllabic clusters with /t/ a first element are infr e quent. Exceptions to the loss of -t before coronal stops and affricates do exist, ho w- ever , for instance tɤɕphɤtta , the name of a typ e of sewing method, comp ound of tɤ-ɕpʰɤt ‘patch (n)’ (§16.4.6) with the v erb ta ‘put’ (§22.4.2.6). In this noun, a geminated -t.t- occurs across syllable boundaries. While dental fricative preinitials cannot pr ece de affricates (§4.2.1.2), the coda /z/ does occur b efore the dental affricate tsʰ- in the cluster -s.tsʰ- in mbrɤst- sʰi ‘rice gruel’ (from mbrɤz ‘rice ’ and tɯtsʰi ‘rice gruel’), assimilating in voice (§5.4.2.1). The preinitial /j/ is not compatible with palatal and alveolo-palatal initials (§4.2.1.6), but the coda /j/ can precede those segments across syllable boundar y , as in qajʑmbraʁ ‘ ear of wheat’ from qaj ‘wheat’ and ʑmbraʁ ‘ ear’ ( of corn). The velar coda /ɣ/ is generally deleted b efore v elar stops, even in redupli- cated ideophones. For instance, the pattern II ideophone (§10.1.2.2) from the root |gɤɣ| ‘ cur v e d; mo ving with difficulty’ is gɤgɤɣ , though gɤɣgɤɣ is also p ossible . There is only one example of ɣ pr ece ding a uvular segment ev en across sylla- ble boundaries: the sigmatic causative sɯɣʁaʁ ‘ cause to hatch’ from ʁaʁ ‘hatch’ (§17 .2.1.4). The Tib etan form འག་ོག་ ⁿbrug.glog ‘thunder and lightning’ would hav e be en e xp ected to yield † mbrɯɣ.ʁloʁ in Japhug, but we find instead mbrɯɣloʁ (in dialects other than Kamnyu) or mbɣɯrloʁ ‘thunderstorm’ in K amnyu with irregular metathesis of ɣ and r . The uvular /ʁ/ can surface before velar and uvular stops in heterosyllabic clusters (unlike when it occurs as preinitial, §4.2.1.8), as in taʁki ‘up and down’ (§5.5.2.2; realized as [taχki] ) and qʰoʁqʰoʁ ‘ingot’ . The uvular coda is delete d, howe ver , in the compound paskɤɣ ‘pig to be fattene d’ fr om paʁ ‘pig’ and skɤɣ ‘fatten’ (pig), avoiding the non-attested cluster -ʁ.sk- . 102 4.2 Inventory of consonant clusters When preceding unvoiced obstruents, the fricative codas /z/ , /ɣ/ and /ʁ/ assimilate in voice and are converted to [s] , [x] and [ʁ] , respectively (§5.4.2.1). Onset clusters with nasal preinitials follow ed by non-nasal sonorants have be en completely eliminated by a series of sound changes detailed in §4.2.1.9. How- ever , the nasal co das /m/ , /n/ and /ŋ/ can occur before syllables with l- , r- , j- , w- , ɣ- or ʁ- as onset. For instance, w e find -n.l- in srɯnloʁ ‘ring’ , -m.ɣ- (instead of -mbɣ- ) in nɤtsɯmɣɯt ‘take away and bring back’ (from tsɯm ‘take away’ and ɣɯt ‘bring’ , §20.12), -n.r- in smɤnrɯɣ ‘medicinal plants’ (from ན་ རིགས་ sman.rigs ‘type of materia medica’) or -m.j- (instead of -mɲ- ) in zɯmjɯ ‘barrel handle ’ (com- pound of zɯm ‘bucket’ and ɯ-jɯ ‘its handle’ , from ཟོམ་ zom ‘bucket’ and ་བ་ ju.ba ‘handle ’ , respectively). These clusters, although heterosyllabic, ar e reduplicated as a whole when partial reduplication is applied, as in the comitative adverb (§5.8.1) kɤ  -srɯnlɯ~nloʁ ‘together with his/ her ring’ from srɯnloʁ ‘ring’ (in- stead of † kɤ  -srɯnlɯ~loʁ , an incorrect form). More complex clusters with a nasal as first element ar e found when the se cond member of the comp ound has a complex onset, for instance -n.rʑ- in χɕɯnrʑi ‘Y ama’ (from གཤིན་ ེ་ gɕin.rdʑe ‘Y ama’ , the Buddhist god of the under world) and -m.xtsʰ- in jpumxtsʰɯm ‘thickness’ (a compound noun derived from jpum ‘be thick’ and xtsʰɯm ‘be thin’ , §5.5.2.2). Some heterosyllabic clusters are avoided by insertion of an anaptyctic vow el /ɯ/ , in particular in Tibetan loanwords. For instance, སེམས་ག་ sems.sdug ‘sadness, worry’ is borrowed as sɯmɯzdɯɣ ‘worry’ instead of expecte d † sɯmzdɯɣ , with the ɯ breaking the cluster -m.-zd- . The anaptyctic vo wel can be /i/ if the cluster in the se cond syllable has a palatal or alveolo-palatal preinitial. For instance , ɕom ‘iron’ and tɤ-jŋoʁ ‘hook’ are compounded as ɕɤmiŋoʁ ‘iron hook’ , in which the preinitial /j/ is converted to /i/ . 4.2.3.2 The sonority sequencing principle in Japhug A considerable amount of work in phonology supports the idea that the segments of the world’s languages follow a univ ersal scale of sonority (for instance V enne- mann 1988; Blevins 1995; see Ohala 1990 for an opposing view). Se veral versions of this sonority hierarchy have been proposed, for instance (§5) (Parker 2002: 235). (5) lo w vowels > mid v owels > high vo wels > /ə/ > glides > laterals > flaps > trills > nasals > /h/ > voiced fricatives > v oiced stops > voiceless fricatives > voiceless stops and affricates 103 5 N ominal morphology T able 5.1: Possessive prefixes in Japhug Person Kamnyu dialect T atshi dialect 1sg a- a- 2sg nɤ- na- 3sg ɯ- ə- 1du tɕi- tsə- 2/3du ndʑi- ndzə- 1pl ji- ji- 2/3pl nɯ- nə- indefinite tɯ-/tɤ-/ta- tə-/ta- generic tɯ- tə- not mark that the noun is indefinite, 1 but rather indicate that it lacks a definite possessor , when the possessor is unknown or irrelevant. 2 Possessive prefixes other than the indefinite p ossessor pr efixes are collectively referr e d to as “ definite possessor prefixes” . The semantic difference between in- definite and generic possessors is discussed in §5.1.3. The vow el contrast /ɯ/ vs. /ɤ/ vs. /a/ on the indefinite p ossessor pr efixes is lexically determined (§5.1.2.1). This contrast is neutralized on definite possessor prefixes, which all have the same form r egardless of the noun. This neutralization is an innovation: in Situ, the vo wel contrast is present on all possessive prefixes (Lin 1993: 168–169), 3 Stacking of possessive prefixes is not allow ed in Japhug, with the exception of the combination of a definite possessor prefix with an indefinite possessor prefix tɯ- or tɤ- to turn an inalienably possessed noun into an alienably p ossessed one (see §5.1.2.9). The only irregularities in possessive morphology are found with a- initial nouns (§5.1.1.1). 1 On the various strategies to express definiteness in Japhug, see §9.1.4 and §9.1.4.3. 2 A revie wer suggests the term “neutral” instead, but I find it not spe cific enough, and “absolu- tive ” is to be avoided for obvious reasons, in a language with numerous constructions following ergative-absolutive alignment (§2.5.5). 3 Prins (2016: 118–119) analyzes the vow el as part of the nominal root. 110 5.1 Possessive pr efixes 5.1.1.1 a- initial nouns Unlike verbs ( see §12.3), nouns whose stem b egins with a a- are e xtremely rare in Japhug. Nevertheless, as is the case with verbs, the v owel a- merges with any prefixed element, so that nouns of this type do not have regular possessive forms. The only noun in a- to commonly receive possessive prefixes is araʁ ‘liquor’ ( a loanword fr om ཨ་ རག་ ʔa.rag ‘liquor’). Its possessive forms are highly anomalous: 1sg aʑɤ-raʁ , 2sg nɤʑɤ-nɤ-raʁ and 2pl nɯʑɤ-nɯ-raʁ (as in 1), combining the pr o- noun in bound state (§5.4) follo wed by the p ossessiv e prefix, which takes over the initial a- . (1) nɯʑɤ-nɯ-raʁ 2pl-2pl.poss-liquor ɯ  -ra qu-be.neede d:f act ‘Do you need liquor?’ (elicited) T o account for these forms, it is necessary to assume that the initial a- was re- analyzed as a 1sg p ossessiv e prefix. Howe ver , this analysis did not occur directly . The expected 1sg p ossessive form *a-araʁ (1sg.poss-liquor ) would automatically yield *a-raʁ by vow el fusion (following the rule described in §12.3). The resulting 1sg *a-raʁ then be came the piv ot for the analogical reshaping of the rest of the paradigm, from which for instance *nɤ-raʁ (2sg.poss-liquor ) was generated (re- placing putative earlier forms such as *nɤ-araʁ * /naraʁ/ ). Due to homophony with the base noun araʁ , the 1sg pronoun aʑo was systematically added and *aʑo a-raʁ underwent morphological fusion to the attested aʑɤ-raʁ , followed by the rest of the possessive paradigm. 5.1.1.2 The expression of possession Possession cannot be expressed without a possessive prefix on the possessee, except for a handful of constructions where a pr onoun occurs instead (§6.1.2). Possessive prefixes can b e used on nearly any noun (e xcept the unp ossessible nouns, see §5.2), including recent borrowings from Chinese ( or quasi-code switch- ing), as 老 家 lǎojiā ‘native place ’ in (2). They also occur on sev eral non-finite verbal forms, including participles ( see §16.1.1.1 and §16.1.2.1), bare infinitives (§16.2.2) and degree nominals (§16.3). (2) aʑo 1sg ɣɯ gen a-<laojia> 1sg.poss-old.house ɣɯ gen ɯ-lɤcu 3sg.poss-upstream nɯre there ri loc ku-rɤʑi-nɯ ipfv-stay-pl ŋu be:f act ‘They live in a place upstr eam from my old house. ’ (14-siblings) {0003508#S215} 111 5 N ominal morphology In the case of first or second person p ossessors, it is possible to have simply a possessive prefix on the noun, ( a-ɣɲi ‘my friend’ , a-mbro ‘my horse ’ and a-ʁgra ‘my enemy’ in 3), a personal pronoun and a possessive prefix ( same p erson and number , as in 4) or even a pr onoun, the genitive clitic ɣɯ and a possessive prefix as in (2) (§8.2.3.1). (3) a-ɣɲi 1sg.poss-friend ci indef tɯ~tɯ-ŋu cond~2-be:f act nɤ, lnk a-mbro 1sg.poss-horse ɯ-lwa 3sg.poss-mane ɯ-taʁ 3sg.poss-on kɤ-zo, imp-land a-ʁgra 1sg.poss-enemy ci indef tɯ~tɯ-ŋu cond~2-be:f act nɤ, lnk a-mbro 1sg.poss-horse ɯ-jme 3sg.poss-tail ɯ-taʁ 3sg.poss-on kɤ-zo imp-land ‘If you are my friend, land on my horse ’s mane, if y ou are my enemy , land on my horse ’s tail. ’ (2002 qaCpa) (4) aʑo 1sg a-mbro 1sg.poss-horse nɤrwɯrɯnb otɕʰi anthr ŋu, be:f act tɯ-sŋi one-day χpaχtsʰɤt yojana ci indef ɲɯ  -wɣ-tsɯm-a ipfv:west-inv-take.away-1sg cʰa can:f a ct ‘My horse is Norbu Rinpoche, he can make me cross one yojana per day . ’ (2003 smanmi2) It is possible to have a first singular possessive preceded by a first plural pro- noun, as in (5) (see §14.6.1 for other e xamples of person mismatch involving 1pl pronouns). (5) iʑo 1pl a-mu 1sg.poss-mother nɯ dem tʰamtʰam now kɯrcɤsqaptɯɣ 81 tʰɯ-azɣɯt aor-r each ŋu. be:f act ‘My mother is now 81. ’ (2010-histoir e09-2) In the case of a possessee shared by the sp eaker and the addressee , the 1sg p os- sessive is the pr eferred form. For instance, a couple of parents or grandpar ents talking to each other about their son or their grandchild more often use a-tɕɯ ‘my son’ or a-ɣe ‘my grandchild’ than a 1du or a 2sg possessor such as tɕi-ɣe ‘ our grandchild’ or nɤ-ɣe ‘your grandchild’ , as I have noticed by participant observa- tion. Nevertheless, the use of other possessive pr efixes than 1sg in such contexts is not agrammatical, and systematically occurs in texts translated from Chinese (by calquing), as in (6). (6) tɕi-tɕɯ 1du.poss-son ɲɯ-sɤzdɯxpa, sens-be.pitiful ‘Our poor son!’ (150831 renshen wawa-zh) {0006418#S15} 112 5.1 Possessive pr efixes 5.1.1.3 Definiteness and obviation Nouns with a definite possessor in Japhug can b e indefinite , unlike in most lan- guages of Europe. The y can o ccur with an indefinite determiner ( example 3 above). With a quantifier such as tɯ-rdoʁ ‘ one piece’ as in (7), a noun with a definite possessor is interpreted as referring to a certain number of p ersons out of a group (‘ one of his X’). (7) tɤ-tɕɯ indef.poss-son nɯ dem kɯ erg ɯ-zda 3sg.poss-companion tɯ-rdoʁ one-cl ɯ-pʰe 3sg.poss-da t to-ti, ifr-say tɯ-rdoʁ one-cl nɯ dem kɯ erg li again ci indef ɯ-pʰe 3sg.poss-da t tɕe lnk ɲɤ-k-ɤ-sɯ-ɤmɯ-mtsʰɯ~mtsʰɤm-nɯ ifr-peg-recip-ca us-recip-hear-pl ‘The boy told one of his companions, and that one another one, and [in this way ] they informed each other . ’ (2012 Norbzang) {00037 68#S7 0} Unlike in Algonquian languages, but like in Mapudungun (Haude & Zúñiga 2016), inverse marking on the verb (§14.3.2.2) is not r equired if the subje ct is a possessed noun whose p ossessor is also object of the same sentence, as sho wn by example (8) where the dir ect form na-βde app ears. In other wor ds, nouns with a third person possessor are not automatically obviative (see Jacques 2010a and §14.3.3.2 for additional discussion). (8) ɯ-rʑaβ 3sg.poss-wife nɯ dem kɯ erg na-βde aor:3 → 3 ′ -thr ow .away ‘His 𝑖 wife left him 𝑖 . ’ (14-siblings) The inverse form nɯ  -wɣ-βde is also possible in exactly the same context – example (9) comes from the same text and refers to the same e vent. (9) ɯ-rʑaβ 3sg.poss-wife cʰo comit ɯ-tɕɯ 3sg.poss-son nɯ dem ʁnaʁna both kɯ erg nɯ  -wɣ-βde aor-inv-thr ow .away ‘His 𝑖 wife and his 𝑖 son left him 𝑖 . ’ (14-siblings) {0003508#S266} Although not obligatory , the inverse on the verb (like in example 9) is more common than a direct form ( example 8) in this type of configuration (§14.3.2.2). 113 5 N ominal morphology 5.1.1.4 Other uses of possessive prefixes Possessive prefixes are also used to express beneficiaries, recipients and other oblique arguments, such as the ‘person nee ding’ in the construction with the verb ra ‘be nee ded’ , as in (10). 4 (10) a-mbro 1sg.poss-horse taʁndo spee ch kɯ-tso sbj:pcp-understand ci one tɕi also ra be.neede d:f act ‘I also need a horse who understands sp eech. ’ (2003kAndzwsqhaj2) In the case of beneficiaries and recipients, if a genitive pronoun or genitiv e phrase is present, the presence of a possessive pr efix is possible (11) but not oblig- atory (12), in particular in the case of p ossessed nouns that already have a definite possessor (13). (11) aʑɯɣ 1sg:gen a-kɯ-ra 1sg.poss-sbj:pcp-be.neede d ci indef tu exist:f act tɕe lnk nɯ dem ‘ɣa’ yes tɤ-ti imp-say ra be.neede d:f act ‘There is one thing I need, and you have to say ‘y es’ to it. ’ (140429 qingwa wangzi-zh) {0003890#S47} (12) aʑɯɣ 1sg:gen kɯ-ra sbj:pcp-be.neede d me not.exist:f act ‘I don’t need anything. ’ (2005 Norbzang) (13) aʑɯɣ 1sg:gen ɯ-lu 3sg.poss-milk ra be.neede d:f act ‘I want its milk. ’ (02-deluge2012) {000337 6#S12} W e also find 3sg possessive prefixes ɯ- indexing not a possessor or a benefi- ciary/ recipient, but anaphorically referring to a whole clause, as in (14), wher e ɯ-cʰa does not mean ‘its/ his alcohol’ , but ‘the alcohol made in the fashion de- scribe d in the pr evious clause ’ . 5 4 See §8.2.3.2 for a more detailed account of the expression of beneficiaries in Japhug. 5 As pointed out by a revie wer , this type of construction reminds of the diachronic evolution from third person possessive suffix to definite article in Amharic (Rubin 2010). 114 5.1 Possessive pr efixes (14) kɯɕɯŋgɯ in.former .times tɕe lnk iɕqʰa the.afor ementione d ʑmbrɯβɟaj boat.oar nɯ dem kɯ erg nɯnɯ dem cʰa alcohol nɯ dem tú-wɣ-sɯ-ɕmi ipfv-inv-ca us-mix tɕe lnk ɯ-cʰa 3sg.poss-alcohol mɯm be.tasty:f a ct tu-ti-nɯ ipfv-say-pl pɯ-ŋgrɤl pst.ipfv-be.usually .the.case ‘In former times, people use d to mix the alcohol with boat oars, the alcohol [made this way] is tasty , they used to say . ’ (cha-31) {00037 64#S38} 5.1.1.5 The form of the 3sg possessive prefix Japhug differs from other Gyalrong languages (T able 5.2, data from Sun & Shi- danluo 2002, Gong 2014) in that the third person possessive prefix is not ho- mophonous with the inverse prefix (§14.3.2.7, §14.3.2.8). T able 5.2: The form of the 3sg possessive prefix in Gyalrong languages 3sg.poss inverse Japhug ɯ- ɣɯ- / -wɣ- T shob dun o- o- Zbu wə- wə- Situ və- və- Independently of the question of whether these two prefixes could be historic- ally related (Sansò 2014), it is probable that Japhug is innovativ e here. In the same way as the inverse prefix ɣɯ- has an allomorph transcribed as -wɣ- when preceded by another prefix, realized as vowel r ounding in most cases (§14.3.2.7), there is a possible trace of a vow el rounding allomorph of the posses- sive prefix in the linker núndʐa ‘for this r eason’ (§25.5.1). This linker originates from a phrase combining the demonstrativ e nɯ ‘that’ ( on which se e §6.9.1, §9.1.2 and §9.1.5.4) with the 3sg possessed form of the noun ɯ-ndʐa ‘ cause ’ (§25.5.2). The form núndʐa possibly reflects earlier *nɯ-w-ndʐa , *-w- being a frozen allomorph of the 3sg possessive prefix in non-initial position. There is a possible trace of the expected allomorph † ɣɯ- (from proto-Gyalrong *wə- ) in the noun ɣɯfsu ‘friend’ , etymologically ‘his equal’; the inalienably p os- sessed noun ɯ-fsu ‘ of the same size’ , which shares the same root, has a regular possessive prefix that is corefer ent with the standard of comparison (as in 15 with the 3pl; see §26.3.1.3 on this construction). The alienably p ossessed noun ɣɯfsu 115 5 N ominal morphology ‘friend’ is thus possibly a lexicalized equivalent of ɯ-fsu ‘of the same size ’ (used in one of the equative constructions, §26.3.1.3), whose 3sg prefix was frozen be- fore the change fr om *ɣɯ- to ɯ- o ccurr e d. (15) tɯrme person kɯ-mbro sbj:pcp-be.high ra pl nɯ-fsu 3pl.poss-equal jamar about tu-zɣɯt ipfv-reach ma apart.from mɤ-cʰa neg-can:f a ct ‘It can only grow about as high as a tall human. ’ (15-bab W) {0003512#S4} Furthermore , additional evidence for the idea that the third person prefix con- tained *w- comes from the etymology of the reflexiv e prefix ʑɣɤ- , which is argued to originate from the third person pronoun (§18.3.7). It remains unclear why the regular allomorph of the thir d person p ossessive is ɯ- rather than expected † ɣɯ- . A p ossible explanation could be false segmen- tation, due to reanalysis with the genitive marker ɣɯ , since the genitiv e can op- tionally occur b etw e en the possessor and the possessee, as in (16). A pre-Japhug form such as *qaɕpa ɣɯ-pɯ could have been misanalyze d as qaɕpa ɣɯ ɯ-pɯ due to vow el fusion sandhi (§4.3), and a new allomorph ɯ- extracted from such constructions. 6 (16) nɯnɯ dem qaɕpa frog ɣɯ gen ɯ-pɯ 3sg.poss-young ŋu be:f act tɕe, lnk ‘It (the tadpole) is the young of the frog. ’ (hist-28-k W p Az) {0003714#S203} 5.1.2 Inalienably p ossessed nouns 5.1.2.1 Morphology Inalienably possessed nouns differ from alienably possessed ones in that they require the presence of a possessive prefix. Unless when used with the indefinite possessor prefixes, inalienably possessed nouns are not formally distinguishable from alienably possessed ones; for instance, a-pi ‘my elder sibling’ and a-mbro ‘my horse ’ both take the 1sg a- prefix and no direct clue indicates that the first noun is inalienably possessed and that the se cond one is alienably possessed. The citation form howe ver differs between inalienably and alienably p ossessed nouns: the former must take an indefinite possessor prefix ( or in some cases a 3sg ɯ- ), while the latter can occur without possessive prefix, as for instance tɤ-pi 6 The weakness of this hypothesis is that some Japhug dialects have kɯ rather than ɣɯ as their genitive marker . 116 5.1 Possessive pr efixes ‘ elder sibling’ (with the indefinite tɤ- ; the bare r oot † pi is not a correct form) vs. mbro ‘horse ’ (without pr efix). Inalienably possessed nouns are divided into four classes dep ending on their citation form. The indefinite possessor prefix has three allomorphs ( tɯ- , tɤ- , ta- ) whose distribution is not completely predictable on the basis of phonology or semantics (though some generalizations are pro vided b elow ). In addition, some inalienably possessed nouns only take definite p ossessor prefixes. The contrast between these four classes is neutralize d when the noun takes a definite posses- sor prefix (unlike in Situ, see Lin 1993: 168–169 and Prins 2016: 118–119). The most common allomorph of the indefinite possessor prefix is tɯ- . Inalien- ably possessed nouns sele cting this allomorph, such as tɯ-jaʁ ‘hand, arm’ , have identical indefinite and generic possessor forms (see §5.1.3). The allomorph tɤ- is also very common, in particular with kinship terms and some bo dy parts ( see §5.1.2.3 and § 5.1.2.4). The form ta- is a phonological variant of tɤ- , occurring mainly with nouns whose stem b egins with a uvular such as ta-ʁrɯ ‘horn’ or ta-ʁi ‘younger sibling’ . The contrast between /ɤ/ and /a/ in this prefix is very difficult to perceive before uvulars with some speakers (see §3.5.4), and the transcription adopted in this grammar (and the online corpus and dictionary) is based on the slow syllable-by-syllable pronunciation of these wor ds by T shendzin. T wo inalienably possesse d nouns, how ever , ta-ma ‘work’ and ta-mar ‘butter’ , have the ta- allomorph with an initial m- , probably originally due to vow el assimilation (§3.3.1.2, §3.5.4). The minimal pair between tɤ-ma ‘mother’ and ta-ma ‘work’ shows that this vow el contrast, howe ver marginal, is distinctive, and that e ven if the two allo- morphs tɤ- and ta- were originally phonologically conditioned, it is no longer the case in Kamnyu Japhug. Some inalienably possessed nouns never occur with indefinite p ossessor pr e- fixes, for instance ɯ-tʰoʁ ‘ground’ is only attested with the 3sg ɯ- prefix ( se e §5.1.2.11). In some cases, the indefinite possessor form is difficult to elicit and in case of doubt the third singular form is given in the dictionary Jacques (2015– 2016) (for instance ɯ-mdoʁ ‘ colour’). Future reasear ch may reveal an indefinite possessor form for some of these nouns. When denominal verbs are deriv ed from inalienably possesse d nouns, the vo- calism of the denominal prefix tends to be the same as that of the indefinite possessor prefix (for instance tɤ-βɟu ‘mattress’ → nɤβɟu ‘use as a mattress’ , not † nɯβɟu ), though there are e xceptions ( tɯ-rpaʁ ‘shoulder’ → mɤrpaʁ ‘ carr y on the shoulder’), as discussed in §20.6. By analogy with several non-finite v erb forms, in particular the subject par- ticiple of transitive verbs and the bare infinitive , which index one argument (the 117 5 N ominal morphology object) by a possessive prefix (§16.2.2), the possessors of inalienably possesse d nouns are considered to be core arguments , while those of alienably possessed nouns are treated as adjuncts . In other wor ds, inalienably p ossessed nouns have a valency of 1 like intransitive verbs, while alienably possessed nouns have a va- lency of 0. 7 The indefinite possessor prefix can be viewed as a valency-decreasing device , the nominal e quivalent of passiv e and antipassive derivations, especially given its use in the alienabilization of inalienably possessed nouns (see §5.1.2.9). Wider implications of the assumption that possessors of inalienably possesse d nouns are core arguments ar e explored in §24.6. 5.1.2.2 Inalienabilization Derivation from alienably possessed to inalienably possesse d nouns is not com- mon in Japhug. An interesting case is that of ɯ-ʁle ‘r eputation’ , which originates from the alienably possessed qale ‘wind’ with a reduce d form ʁ- of the class pr e- fix qa- , as some second memb ers of compounds (see §5.4.3.2 and §5. 6). Conversion of counted nouns (§7 .3) to inalienably p ossessed nouns is a regular process (§7 .3.4.1). 5.1.2.3 Bo dy parts The great majority of body parts are inalienably possesse d nouns with the indefi- nite possessor tɯ- . These include native words, but also borrowings fr om Tibetan such as tɯ-qʰoχpa ‘ organs, state of mind’ from Tibetan ཁོག་པ་ kʰog.pa ‘innards’ (see §3.4.2 on the phonology of this word). Among bo dy parts, inalienably possessed nouns selecting the prefix tɤ- are mainly liquids from the body such as tɤ-se ‘blo od’ , tɤ-spɯ ‘pus’ and tɤ-lu ‘milk’ (though some liquids also take the prefix tɯ- , for instance tɯ-ɕtʂi ‘sweat’), hair ( tɤ-rme ‘hair’ , tɤ-kɤrme ‘head hair’ and some animal b ody parts ( tɤ-jme ‘tail’ , tɤ-ŋkɯ ‘pig skin’ , tɤ-rkʰom ‘feather rachis’). Parts of plants on the other hand mainly have the pr efix tɤ- , as tɤ-jwaʁ ‘leaf’ , tɤ-tsrɯ ‘sprout’ , tɤ-zrɤm ‘root’ , etc. Alienably possessed nouns are rare among bo dy parts. Some nouns with the qa- class prefix ( see §5.6) such as qame ‘mole ’ and qambɣo ‘ ear wax’ referring to physical defects or excretions from the body are alienably possessed nouns. A similar situation is obser v ed in Koyukon Athabaskan, wher e nouns ‘ denoting certain temporar y or abnormal parts of the bo dy’ ar e also alienably p ossessed 7 On nominal valency and inalienability , see also Gutman (2018: 30-31). 118 5.1 Possessive pr efixes noun (Thompson 1996: 660), though in Ko yukon this subclass is considerably larger than in Japhug. The compound tɯciste ‘amniotic sac’ from tɯ-ci ‘water’ and tɤ-ste ‘bladder’ has a tɯ- which is originally an indefinite possessor prefix (see §5.1.2.11), but which has be come fr ozen after b eing integrated into a compound (§5.1.2.10), as can be shown by (17). (17) ɯ-tɯciste 3sg.poss-amniotic.sac cʰɤ-ndʑɣaʁ ifr-aca us:squeeze.out ‘Her waters have broken. ’ (elicited) 5.1.2.4 Kinship terms The great majority of kinship terms select the indefinite possessor prefix tɤ- or ta- (see chapter 27 for a description of the kinship system). The only kinship terms in tɯ- are tɯ-me ‘ daughter’ (but this form is not attested in the text corpus) and tɯlɤt ‘second sibling’; howev er , the tɯ- prefix in the latter wor d has be come non-analyzable and this word has be come an unpossessible noun (see §5.2). There ar e other unp ossessible nouns among kinship terms, including woɬaʁ ‘(bad) stepmother’ , which deriv es from tɤ-ɬaʁ ‘mother’s sister’ by replacing the possessive prefix with an unidentified element wo- , and the social relation col- lectives (§5.7 .9.1). Being a unp ossessible noun, woɬaʁ ‘(bad) stepmother’ cannot take possessive prefixes, and the forms of tɤ-ɬaʁ ‘mother’s sister’ are used instead ( a-ɬaʁ can mean ‘my (bad) stepmother’). Kinship terms do not commonly occur with the indefinite p ossessor pr efix. For those denoting spouses, forms with the indefinite prefix are found in the expression ‘look for a wife/husband’ , as in (18). (18) ‘ŋoj where tɯ-ɕe?’ 2-go:f a ct to-ti, ifr-say ‘aʑo 1sg tɤ-rʑaβ indef.poss-wife ɯ-kɯ-ɕar 3sg.poss-sbj:pcp-search ɕe-a’ go:f a ct-1sg to-ti. ifr-say tɕe lnk ‘ndʑiʑo 2du ŋoj where tɯ-ɕe-ndʑi?’ 2-go:f a ct-du to-ti ifr-say ri, lnk ‘tɕiʑo 1du tɤ-nmaʁ indef.poss-husband ɯ-kɯ-ɕar 3sg.poss-sbj:pcp-search ɕe-tɕi’ go:f a ct-1du to-ti. ifr-say ‘She said: ‘Where are you going?’; He said: ‘I am looking for a wife. Where are you going?’; She said ‘W e are looking for a husband. ” (2003-k WBRa) 119 6 Pronouns (105) kɤsɯfse all kɯ erg ʑo emph ta-nɯ put:f a ct-pl maʁ not.be:f act ‘Not everybo dy puts it. ’ (1607 06 thotsi) {0006133#S21} The less common form mɲɯrɯri ‘ everybo dy , each person’ (from Tibetan མི་ རེ་ རེ་ mi.re.re ‘ each man’ also ser v es as a universal quantifier , as in (106). (106) mɲɯrɯri everybo dy kɯ erg ‘nɯnɯ dem ɲɯ-pe ’ sens-be.good ntsɯ always to-ti-nɯ ifr-say-pl ‘Everyb ody said ‘It is nice!” (140521 huangdi de xinzhuang) {0004047#S192} There are se veral wor ds meaning ‘e very where ’ , such as aʁɤndɯndɤt ‘ every- where ’ , but they are tr eated as adverbs rather than pronouns (§22.2.2.2). 6.7.2 Partitive pronouns The mid-scalar quantifier tsuku ‘some ’ is used b oth as a noun determiner (§9.1.3.2) and as a partitive pr onoun, taking case markers and determiners. This construc- tion expresses a meaning close to that obtained by combining a relative clause with an existential verb (‘there is someone who... ’ , §22.5.1.2), as can be seen in (107) where both constructions are used one after the other . (107) tsuku some kɯ erg zgri plant.name tu-ti-nɯ ipfv-say-pl ŋu, be:f act tsuku some kɯ erg mɯrkuj plant.name tu-ti-nɯ ipfv-say-pl ŋu. be:f act mɯrkuj plant.name tu-kɯ-ti ipfv-sbj:pcp-say tɕi also tu, exist:f act zgri plant.name tu-kɯ-ti ipfv-sbj:pcp-say tɕi also tu exist:f act ma, lnk ‘Some call it zgri , some call it mɯrkuj ; there ar e p eople who call it mɯrkuj , and also people who call it zgri . ’ (19-qachGa mW ntoR) {0003546#S154} The quantifier tsuku ‘some ’ used as a pronoun generally refers to humans in the corpus, but (108) shows that it can also denote plants for instance . (108) tsuku some ɲaʁ, be.black:f a ct tsuku some aqarŋɯrŋe, be.light.yellow:f act ‘Some are black, some are light yellow . ’ (140505 stonka mW ntoR) {0003917#S5} 222 6.7 Quantifiers Numerals (in particular ci ‘ one ’) and also counted nouns (§7 .3.2) can be used without head noun with a partitive meaning ‘ one of (a gr oup)’ as in (109) and (110). (109) ci one ɣɯ gen tɤ-tɕɯ, indef.poss-son ci one ɣɯ gen tɕʰeme indef.poss-son tɯ~tɤ-tu cond~aor-e xist nɤ, lnk ʁzɤmi husband.and.wife ku-kɤ-sɯ-βzu ipfv-inf-ca us-make ‘If one of them has a boy , and the other one has a girl, let us make them husband and wife. ’ (2003 zr AntCW) (110) ci one tʰɯ-kɯ-rgɯ~rgɤz aor-sbj:pcp-emph~be.old ɲɯ-ɕti sens-be.aff tɕe, lnk ci one kɯ-xtɕɯ~xtɕi sbj:pcp-emph~be.small ɲɯ-ɕti sens-be.aff tɕe, lnk ‘One of them has grown v er y old, and one of them is v er y small. ’ (2011-05-nyima) This partitive function is also found in combination with personal pronouns, as in (111). (111) nɯʑora 2pl ci one kɯ erg a-tɯ-ci 1sg.poss-indef.poss-water ci a.little ju-tɯ-ɣɯt-nɯ ipfv-2-bring-pl ɯ  -jɤɣ qu-be.possible:f a ct ‘Could one of you bring me some water?’ (150904 zhongli-zh) {0006348#S47} 6.7.3 Distributive pronouns The pronoun ʑaka ‘ each his own’ and its variant ʑakastaka ‘ each his own’ occur as pronouns, especially as possessors in an p ossessive e xistential construction. It can be correlated with a third singular ɯ- (112) or a third plural nɯ- (113) prefix on the possessum. (112) tɕe lnk nɯnɯ dem li again qaʑo sheep nɯ dem kɯ-ɲaʁ sbj:pcp-be.black tu, exist:f act kɯ-wɣrum sbj:pcp-be.white tu, exist:f act kɯ-ɤɣɯnɯɕɯr sbj:pcp-be.reddish kɯ-fse sbj:pcp-be.like tu, exist:f act 223 6 Pronouns kɯ-ɤrŋɯlɯz sbj:pcp-be.blueish tu, exist:f act tɕe lnk nɯnɯ dem ʑaka each.his.own ɯ-mdoʁ 3sg.poss-colour tu exist:f act ma lnk ‘There are black sheep, white ones, reddish ones, blueish ones, each has his own colour (they come in all types of colors). ’ ( 05-qaZo) {0003404#S58} (113) li again ʑaka each.his.own nɯ-rmi 3pl.poss-name tu, exist:f act ‘Each have their own names. ’ (150903 tW mNu) {0006280#S10} The pronoun ʑaka is built by combining the bound state of the pronominal root -ʑo (§6.1) with the root -ka found in the distributive modifier tɯka ‘ each’ ( which follows possessums, se e §9.1.3.3). 6.8 Identity pronoun The words kɯmaʁ ‘ other’ and kɯɕte ‘ other’ occur as prenominal determiners (see §9.1.7, also for a discussion on the etymology of the former ), but it can also be use d as a pr onoun and take determiners as in (114). (114) ma lnk kɯmaʁ other nɯra dem:pl aj 1sg mɯ  j-sɯχsal-a neg:sens-recognize-1sg ri, but tɤkʰepɣɤtɕɯ bird.sp nɯ dem sɯχsal-a recognize:f a ct-1sg ‘The other ones I don’t recognize them, but the tɤkʰepɣɤtɕɯ bird, I do recognize it. ’ (23-scuz) {0003612#S43} The interpretation of both kɯmaʁ ‘other’ and kɯɕte ‘ other’ can be locative ‘somewhere else ’ as in (115), when the main v erb ( ta ‘put’ in this example) selects a goal or a locative adjunct (since locative noun phrases ar e often unmarked, §8.1.8, §8.1.9). (115) kɯmaʁ/kɯɕte other nɯ-tɯ-ta-t aor-2-put-tr:pst ŋu be:f act ɯ  -maʁ? qu-not.be:f act ‘Did you put it somewher e else?’ (elicitated1) Adding the indefinite determiner ci ‘ one ’ is necessary in this context to convey the meaning ‘something else ’: 224 6.8 Identity pronoun (116) kɯmaʁ/kɯɕte other ci indef nɯ-tɯ-ta-t aor-2-put-tr:pst ŋu be:f act ɯ  -maʁ? qu-not.be:f act ‘Did you put something else?’ ( elicitated1) Example (117) illustrates that both kɯmaʁ and kɯmaʁ ci can occur in the meaning ‘another one ’ in some contexts (here with the verb ɕar ‘sear ch’). (117) χsɯ-sŋi three-day mɤ-kɯ-tsu neg-inf:st a t-pass qʰe lnk li again kɯmaʁ other ci indef ju-ɣɯt ipfv-bring qʰe, lnk li again ɯ-zda 3sg.poss-companion ɲɯ-nɯ-ɕar ipfv-a ut o-search ɲɯ-ɕti. sens-be.aff tɕe lnk nɯnɯ dem maka completely kɯjka chough nɯ dem ŋɤn be.evil:f act ma, lnk ɯ-zda 3sg.poss-companion nɯ dem nɯ-me aor-not.e xist ɯ-qʰu 3sg.poss-after mɤ-kɯ-nɤrʑaʁ neg-inf:st a t-spend.time tɕe lnk kɯmaʁ other ɲɯ-ɕar ipfv-search ɲɯ-ɕti sens-be.aff tɕe lnk mɯ  j-pe neg:sens-be.good tu-ti-nɯ ipfv-say-pl ŋgrɤl. be.usually .the.case:f act ‘Not even thr ee days [after hunters kill its mate, the r ed-bille d chough brings another one, it looks for another mate. People say that the red-billed chough is not nice, because not long after its mate has die d, it looks for another one, it is not good. ’ (22-CA GpGa) {0003586#S79} The indefinite ci ‘ one ’ combined with the demonstrative determiner nɯ (or nɯnɯ ) has the meaning ‘the other one ’ (the definite counterpart of kɯmaʁ ‘ other’), as in (118) and (119). (118) tɕe lnk ɯ-jaʁ 3sg.poss-hand kɯ erg ki dem:pro x tu-ste ipfv-do.like[III] lu-z-naʁje ipfv-ca us-reach.into[III] ɲɯ-ŋu sens-be ri, but tɕe lnk ci indef nɯ dem kɯ erg ɯ-jaʁ 3sg.poss-hand ku-mtsɯɣ ipfv-bite ɲɯ-ɕti sens-be.aff qʰe, lnk ‘[The cat] 𝑖 reaches with its 𝑖 paw [into the hole] like this, but the other one (the weasel) bites its 𝑖 paw . ’ (27 -spjaNk W) {00037 04#S47} 225 6 Pronouns (119) ɯ-me 3sg.poss-daughter ʁnɯz two pjɤ-tu ifr.ipfv-exist tɕe, lnk tɯ-rdoʁ one-piece nɯ dem χsɤrlɤsmɤn gser .la.sman pjɤ-rmi, ifr.ipfv-be.called ci indef nɯ dem rŋɯlɤsmɤn dngul.la.sman pjɤ-rmi ifr.ipfv-be.called tɕe, lnk ‘He had two daughters, one of them was called Gser .la.sman, and the other Dngul.la.sman. ’ (2003-k WBRa) Alternatively to the construction in (119) with tɯ-rdoʁ ‘ one piece ’ and ci nɯ to express the meaning ‘ one of them .... and the other ... ’ , it is possible to use ci nɯ two times in the same sentence to refer to more than one persons or animals, as in (120). (120) tɕe lnk ci indef nɯnɯ dem ju-ɕe ipfv-go ɯ-kʰɯkʰa 3sg-while ci indef nɯ dem kɯ erg ɯ-pu 3sg.poss-intestine tu-ndze, ipfv-eat[III] cʰɯ-rɤɕi. ipfv-pull ‘While one of the two (i.e . the prey ) is (still) going, the other one (i.e . the predator ) eats and pulls its intestine. ’ (20-RmbroN) {0003560#S69} The dual ci nɯni ‘the other two ’ and plural ci nɯra ‘the other ones’ are also attested, as in (121), showing that ci is here completely bleached of numeral mean- ing. (121) ci indef nɯni dem:du ɣɯ gen nɯ, dem ndʑi-ta-mar 3du.poss-indef.poss-butter rɟɤɣi tsampa pjɤ-ŋu ifr.ipfv-be tɕe lnk tɕe lnk nɯnɯ dem ɯʑo 3sg kɯ erg to-ndza ifr-eat ‘The tsampa of the other two [ sisters] was butter tsampa, and she ate it. ’ (2003-k WBRa) {0006346#S67} It is also possible in this function to use other mo difiers such as numerals, as in (122) with χsɯm ‘three ’ . (122) iɕqʰa the.afor ementioned ci indef χsɯm three nɯ dem mɯ-jo-ɣi-nɯ neg-ifr-come-pl kɯ erg ‘The three other ones, without coming, (said...)’ (140515 congming de wusui xiaohai-zh) {0003998#S44} As a pr enominal determiner , ci also has the meaning ‘the other 𝑋 ’ (see §9.1.7). Finally , the noun tɯrme ‘person’ (which also occurs to expr ess generic person, §6.2.2) can be use d in the meaning ‘someone else ’ or ‘ other pe ople ’ , in particular in genitival constructions as in (123). 226 6.9 Demonstrative pr onouns (123) tɯrme people ɯ-kʰɤpa 3sg.poss-yard zɯ, loc ki dem.pro x kɯ-fse sbj:pcp-be.like tɯ-rʑaʁ one-night lu-znɯfsoʁspat-a ipfv-do.the .whole.night-1sg ku-omdzɯ-a. ipfv-sit-1sg ‘I would spend an entire night fr om dusk till dawn sitting in some one else ’s animal yard. ’ (2010-histoire09) 6.9 Demonstrative pronouns There are two basic demonstrativ es in Japhug, the proximal ki ‘this’ and the dis- tal one nɯ ‘that’ , which also o ccur as demonstrativ e determiners (see §9.1.2). T able 6.5 illustrates the various demonstrative pronouns that ar e derived from these basic forms, with reduplicated and emphatic forms. There is in addition a cataphoric pronoun nɤki , discussed in §6.9.2.2. Plural and dual forms, as in the case of determiners, are formed by adding -ra and -ni suffixes (§9.1.1) to the demonstrative root, which undergoes bound state change /i/ → /ɯ/ in the case of proximal demonstrativ es. P lural forms ar e given in the table; dual forms are attested but rare and can be predicted ( kɯni ‘these two ’ , etc). T able 6.5: Demonstrative pronouns Base form Re duplicated Emphatic Proximal, singular ki kɯki ɯkɯki Distal, singular nɯ nɯnɯ ɯnɯnɯ Proximal, plural kɯra kɯkɯra ɯkɯkɯra Distal, plural nɯra nɯnɯra ɯnɯnɯra The distal demonstratives, being the default forms, are simply glossed as dem in the examples: only pro ximal demonstratives are e xplicitely marke d as such in the glosses. 6.9.1 Anaphoric demonstrative pronouns The basic demonstratives ki and nɯ are less often used as pronouns that the other ones (they mainly occur as determiners). They nev ertheless do occur in all syntactic functions, including object (in particular with the verb ti ‘say’ , as in 124, 227 6 Pronouns where it refers to w ords that have been previously told to another animal), and semi-object (in particular with the verb stu ‘ do like ’ as in 125). (124) li again nɯ dem to-ti ifr-say ri, lnk ‘[Gesar ] said the same thing to the [sno w leopard]. ’ (gesar ) (125) ɯ-mu 3sg.poss-mother nɯ dem ku-rqoʁ ipfv-hug tɕe lnk ki dem:pro x tu-ste ipfv-do.like[III] tɕe lnk ‘It hugs its mother like that. ’ (19-GzW) {0003536#S29} The distal demonstratives nɯ and nɯnɯ serve as anaphoric pronouns with any type of referent, including humans, but also abstract concepts, inanimate objects or plants as in (126), though as mentioned in §6.1, third person pronouns such as ɯʑo ‘he ’ can also have inanimate antecedents. (126) tʂʰa tea kɤ-nɯ-ta inf-a ut o-put tɤ-ra, aor-be.needed smi fire kɤ-βlɯ aor-burn tɤ-ra aor-be.needed pɯ-nɯ-ŋu, pst.ipfv-a ut o-be tʰamaka tobacco sko-nɯ smoke:f a ct-pl pɯ-nɯ-ŋu, pst.ipfv-a ut o-b e tɕe lnk nɯnɯ dem kɯ erg smi fire tu-sɯ-tɕɤt-nɯ. ipfv-ca us-take.out-pl ‘When they need to b oil tea, to make a fir e or smoke tobacco, people light up the fire with it. ’ (15-bab W) {0003512#S209} When a third person mentioned in a discussion is present, the pronoun ɯʑo ‘he ’ is not the optimal way of referring to him/her , and a pr oximal demonstrative , in particular the reduplicated kɯki ‘this’ , is use d instead. It can occur to present someone to someone else (127) (note that a similar usage exists in W estern lan- guages such as English in the same context) and ev en to talk about the actions of this person, as in (128) and (129). (127) kɯki dem.pro x a-slama 1sg.poss-student ŋu be:f act ‘This a [former ] student of mine. ’ ( conversation 140510) (128) kɯki dem.pro x kɯ erg ta-βzu? aor:3 → 3 ′ -make ‘Did she make it?’ ( conversation 140510) 228 6.9 Demonstrative pr onouns As other pr onouns (see §6.1), demonstrative pronouns can take the demonstra- tive determiner nɯ , as in (129). (129) mɯ~mɤ-pɯ-jɤɣ cond~neg-pst.ipfv-be.acceptable tɕe lnk mɤ-ɣi-tɕi neg-come:f a ct-1du ma lnk kɯki dem:pro x nɯ dem fstɯn-tɕi serve:f a ct-1du ra be.neede d:f act ma lnk tɕi-βɣe 1du.poss-orphan ɯ-ku 3sg.poss-head tʰɯ-kɯ-ɣɤrndi aor-sbj:pcp-support ‘If it is not possible [to take the old man with us] we will not come, as we hav e to ser ve him, he is the one who adopted us orphans when we wer e in dire straits. ’ (The old man is presumably pr esent when this sentence is uttered) (2003nyima2) The emphatic demonstrative pronouns ( which are also used as determiners, §9.1.2) are built by combining the r e duplicated forms of demonstratives with the third person possessive prefix ɯ- . They ar e ab out fifty times less common than corresponding reduplicated forms, but their function is essentially the same. In (130), ɯnɯnɯ is an anaphoric pronoun whose antecedent is present in the im- mediately preceding clause. (130) tɕe lnk ɯ-rqʰu 3sg.poss-hull kɯ-fse sbj:pcp-be.like ci indef ɣɤʑu exist:sens tɕe, lnk ɯnɯnɯ dem:emph kɯ erg ɯ-rdu 3sg.poss-eyeball nɯ dem tu-ɕɯ-fkaβ ipfv-ca us-cover kɯ-fse sbj:pcp-be.like ɲɯ-ŋu. sens-be ‘It has something like a membrane, and it co vers its eyeball with it. ’ ( description of the nictitating membrane of birds, 140513 sWNg W rmABja) {0003983#S9} The demonstrative nɯ may not refer anaphorically to a particular entity , but also to an entire situation, as in (131) where it occurs as an adjunct in absolutive form, meaning ‘this way , like that’ (in another version of the same story , we find nɯ kɯ-fse ‘like that’ instead of nɯ in the same conte xt). (131) nɯnɯ dem ɯ-mŋu 3sg.poss-bank nɯtɕu dem:loc zɯ loc li, again qapri, snake nɤki, filler kɯ-ɲaʁ sbj:pcp-be.black nɯ dem kɯ erg kɯ-wɣrum sbj:pcp-be.white nɯ dem ɯ-qiɯ 3sg.poss-half ʑo emph cʰɤ-mqlaʁ ifr-swallow tɕe lnk 229 6 Pronouns nɯ dem pjɤ-rɤʑi-ndʑi. ifr.ipfv-stay-du ‘On the bank [ of the lake], there was again a black snake that had swallowed half of a white snake , and they were staying [ stuck] like that. ’ (28-smAnmi) {0004063#S99} 6.9.2 Me dial and cataphoric pronoun 6.9.2.1 Me dial demonstrative In addition to the proximal and distal demonstratives, there is a considerably rarer medial demonstrative nɤki , in examples such as (132) and (133), which means ‘your place , near you’ , as opp osed to ‘here ’ . (132) kutɕu dem.pro x:loc ko-qanɯ ifr-be.dark pʰoʁpʰoʁ idph(II):completely ʑo, emph nɤki dem:medial nɯtɕu dem:loc ɯ-kó-qanɯ? qu-ifr-be.dark ‘Here (in Mbarkhams) it is already dark, is it [ also] dark at your place (in Paris)?’ (phone conversation, 14.12.24, referring to the time lag between Paris and Mbarkham) (133) ki dem:pro x kɯra dem:pro x:pl ɲɯ-kʰam-a ipfv-give-1sg tɕe lnk nɤki dem:medial nɯ dem aʑɯɣ 1sg:gen nɯ-kʰɤm imp-give je sfp ‘I give [ you] these [toys], giv e me that one. ’ (2012 Norbzang) {00037 68#S116} The medial demonstrative nɤki can be historically analyze d as a combination of the proximal demonstrativ e ki with the second p erson possessive nɤ- . How- ever , equivalent dual or plural forms such as † ndʑiki or † nɯki are impossible (the equivalent meaning can only b e e xpressed with the dative, using a form such as ndʑiʑo ndʑi-pʰe ‘at y our 𝑑 𝑢 place’ , §8.3.1 ). 6.9.2.2 Cataphoric pronoun In addition to its function as a medial demonstrative (§6.9.2.1), the demonstrative nɤki also occurs to express cataphoric refer ence. It occurs especially when the speaker hesitates and uses it as a filler , followed by a clause with the same v erb ( examples 134 and 135) or just with the same auxiliar y (136). 230 6.9 Demonstrative pr onouns (134) <luohan> Arhat nɯnɯ dem nɤki dem:ca t aph ɲɯ-ŋu sens-be loβ, sfp skɯ statue ɲɯ-ŋu sens-be loβ sfp ‘ An Arhat is... it is Buddhist statue. ’ (150829 jidian) {0006338#S124} (135) tɕe lnk nɯ dem tɯ-ci indef.poss-water ɣɯ gen ɯ-taʁ 3sg.poss-on nɯnɯtɕu, dem:loc nɤki dem:ca t aph ɲɤ-χtɤr , ifr-spread iɕqʰa the.afor ementionne d <yujinxiang> tulip kɤ-ti obj:pcp-say mɯntoʁ flower nɯ dem ɣɯ gen ɯ-jwaʁ 3sg.poss-leaf nɯ dem ɲɤ-χtɤr . ifr-spread ‘She spilled on the water ... she spilled the p etals of the flo wer called “tulip ” . ’ (150818 muzhi guniang-zh) {0006334#S68} (136) tɕe lnk nɯ-nɯŋa 2pl.poss-cow ra pl nɤki dem:ca t aph ŋu be:f act ɕi, qu tɕe lnk ɲɯ-tɯ-nɤm ipfv:west-2-chase[III] qʰe, lnk tɕe lnk ʑara 3pl ku-n-nɯ-ɣi-nɯ ipfv:east-a ut o-vert-come-pl ŋu be:f act ɕi? qu ‘ And your cows, ar e they (still) like that, y ou let them out of the p en (in the morning), and they come back home on their own (in the e vening)?’ (taRrdo conversation) It is also used when the speaker alerts the addressee that a long description follows as in (137), as in English ‘(he said) the following’ . Given the fact the Japhug is strictly verb-final and has pr e-verbal complements (§22.1.1), this is a strategy employed to avoid relegating the main verb to the end of the description. (137) kʰopi anthr kɯ erg ɴqiazwɤr bitter .wormw oo d ci indef ɲɯ-mɯm sens-be.tasty rca unexp:deg ɲɯ-saχaʁ sens-be.extremely ʑo emph tɕe lnk nɤki dem:ca t aph tu-stu-nɯ ipfv-do.like-pl ɲɯ-ŋu sens-be ɲɯ-ti, sens-say ɲɯ-pʰɯt-nɯ ipfv-take.out-pl qʰe lnk kɯ-zri... sbj:pcp-be.long ki dem:pro x jamar about ʑo emph kɯ-zri sbj : pcp -be.long ɲɯ-pʰɯt-nɯ ipfv -take.out- pl qʰe lnk nɤki, dem : ca t aph ɯ-ku 3sg . poss -head ɯ-mtɯ 3sg.poss-crest kɯ-fse sbj:pcp-be.like nɯtɕu dem:loc kú-wɣ-ndo ipfv-inv-take qʰe lnk tɕe lnk 231 7 N umerals and counted nouns is also probably related, though its morphological relationship with ʁnɯz ‘tw o ’ does not fit any known pattern. The superficial resemblance between ʁnɯz ‘two ’ and kɯɕnɯz ‘seven’ could suggest the existence of a former quinary system. Howev er , the fact that these two numerals have differ ent vow els in Situ (Cogtse kəniɛ  s ‘two ’ and kəɕnəs ‘seven’ , from Huang & Sun 2002) makes this assumption less likely; it is in any case ir- relevant to the synchronic grammar of Japhug. Unlike the other numerals in T able 7 .1, the fr ee numeral χsɯm ‘three’ is iden- tical to the form found in compound loans such as kɯmtɕʰoχsɯm ‘triratna’ from དཀོན་ མཆོག་ གམ་ dkon.mtɕʰog.gsum ‘triratna’ . Therefore , one cannot exclude the pos- sibility that it is a borrowing from Tibetan གམ་ gsum ‘three ’ that replaced the native numeral ‘thr e e ’ ( which was cognate to the Tibetan form, and phonetically similar to it). In this hypothesis, the alternative forms -fsum and fsɯ- for ‘three ’ found in the numerals 11-19 (§7 .1.3) and tens (§7 .1.2) could be remnants of the na- tive numeral. Alternatively , it is p ossible that the native w ord and the borrowing are true cognates, and happen to have the same form by coincidence . The numeral kɯngɯt ‘nine ’ has a coda -t which is not found in the cognates of this numeral in Situ and languages outside of Gyalrongic, suggesting analog- ical spreading of the coda from kɯrcat ‘ eight’ . This innovation is shared by all Northern Gyalrong languages: T shobdun has kənⁿgət (Sun & Blogros 2019) and Zbu kənⁿɡət (Gong 2018: 130). The same analogy independently o ccurr ed in the Siyuewu dialect of Khroskyabs, where ‘nine ’ is ŋgəd (Lai 2017: 174). T able 7 .1: Basic numerals in Japhug and Tibetan Native Japhug Tibetan Tibetan loanwords in Japhug 1 ci or tɤɣ གཅིག་ gtɕig ‘ one ’ χtɕɯɣ 2 ʁnɯz གཉིས་ gɲis ‘two ’ ʁɲiz 3 χsɯm གམ་ gsum ‘three ’ χsɯm 4 kɯβde བཞི་ bʑi ‘four’ βʑi 5 kɯmŋu ་ lŋa ‘five ’ rŋa 6 kɯtʂɤɣ ག་ drug ‘six’ tʂɯɣ 7 kɯɕnɯz ( བན་ bdun ‘seven’) βdɯn 8 kɯrcat བད་ brgʲad ‘ eight’ βɟɤt 9 kɯngɯt ད་ dgu ‘nine ’ rgɯ 10 sqi ( བ་ btɕu ‘ten’) ftɕɯ Numerals from 1 to 99 ar e a sub class of unpossessible nouns (§5.2), and can- 238 7 .1 P lain numerals not take possessive prefixes; they differ in this r egard from the higher numerals (§7 .1.4, §7 .2). 7.1.2 T ens The numerals for tens (T able 7 .2) ar e relatively straightforward. With the excep- tion of ɣnɤsqi ‘twenty’ and fsɯsqi ‘thirty’ , they are pr edictable by combining sqi ‘ten’ with the corresponding numeral prefix (§7 .3.1). The numeral kɯngɯsqi ‘ninety’ is ambiguous, as the same form can also mean ‘nine or ten’ (see T able 7 .5). The element ɣnɤ- in ɣnɤsqi ‘twenty’ is related to the numeral ʁnɯz ‘two ’ , but has a velar ɣ- prefix instead of the uvular ʁ- , and has a different v owel. The adverb ʁnaʁna ‘both’ is also relatable, but the alternations are not e xplainable from a synchronic point of view . T able 7 .2: T ens 10 sqi 20 ɣnɤ-sqi 30 fsɯ-sqi 40 kɯβdɤ-sqi 50 kɯmŋɤ-sqi 60 kɯtʂɤ-sqi 70 kɯɕnɤ-sqi 80 kɯrcɤ-sqi 90 kɯngɯ-sqi Other numerals under one hundred are built by combining the tens in T able 7 .2 (removing the -sqi element) with the units in T able 7 .3. For instance, 37 can be obtained by putting together fsɯsqi ‘thirty’ and sqaɕnɯz ‘seventeen’ as fsɯ-sqa- ɕnɯz . 7.1.3 Numerals 11-19 and units The numerals 11-19, listed in T able 7 .3, serve as the basis for indicating units in all following numerals between 21 and 99, by replacing the -sqi element of the tens (T able 7 .2) by the appr opriate form. As an example of ho w to build numerals above 19 with a unit 1-9, T able 7 .3 illustrates the formation of the numerals 21 to 29 from ɣnɤsqi ‘twenty’ . 239 7 N umerals and counted nouns T able 7 .3: Numerals 11-19 and 21-29 10 sqi 20 ɣnɤsqi 11 sqa-p-tɯɣ 21 ɣnɤ-sqa-p-tɯɣ 12 sqa-m-nɯz 22 ɣnɤ-sqa-m-nɯz 13 sqa-f-sum 23 ɣnɤ-sqa-f-sum 14 sqa-βde 24 ɣnɤ-sqa-βde 15 sqa-mŋu 25 ɣnɤ-sqa-mŋu 16 sqa-p-rɤɣ 26 ɣnɤ-sqa-p-rɤɣ 17 sqa-ɕnɯz 27 ɣnɤ-sqa-ɕnɯz 18 sqa-rcat 28 ɣnɤ-sqa-rcat 19 sqa-ngɯt 29 ɣnɤ-sqa-ngɯt The numerals 11-19 present three morphological changes in comparison with the basic numerals 1-9. First, the form sqi ‘ten’ alternates with sqa- . The origin of this Ablaut is un- known, though it could be a type of b ound state (§5.4); some Gyalrongic lan- guages, such as Khroskyabs have a similar alternation (Lai 2017: 175–17 6). Second, the velar kɯ- and uvular χ-/ ʁ- prefixes found in the base numerals are lost in all numerals 11-19. Third, a labial element /p/ ( sqaptɯɣ ‘ elev en’ , sqaprɤɣ ‘sixte en’), /m/ ( sqam- nɯz ‘twelv e’), or /w/ ( sqafsum ‘thirteen’) is inserted between the sqa- and the following numeral root. It does not occur in seventeen, eighteen and nineteen ( which already have a cluster ), fourteen and fifte en ( which have a cluster with a labial as first element). The form sqaptɯɣ ‘ eleven’ contains an ablauted form of tɤɣ ‘ one ’ as second element. The cluster -pt- in this word is the only case in the language of a /p/ followed by an obstruent (§4.2.3.1). In sqamnɯz ‘twelve ’ , the labial linker is nasalized by the following n . This is not a synchronic rule: for instance, a noun ɕnaβndʑɣi ‘snotty-nosed kid’ has β allomorph of /w/ before a prenasalized obstruent (§5.5.5.1). Howe ver , there ar e other cases of nasalization of labial consonants to /m/ before nasal or prenasal- ized consonants in Japhug (see §17 .3.1). In sqaprɤɣ ‘sixteen’ , not only the prefix kɯ- is lost, the tʂ affricate of the base form kɯtʂɤɣ ‘six’ is replaced by /r/ , preceded by the linking element -p- . This /tʂ/ ~ /r/ alternation is evidence for a sound change *tr- → /tʂ/ (§4.2.2.4). The numeral kɯtʂɤɣ ‘six’ contains two etymological prefixes, kɯ- and a pr efix *t- 240 7 .1 P lain numerals that has fused with the root as -tʂɤɣ . This *t- prefix is possibly related to the d- of its Tibetan cognate ག་ drug ‘six’ . The numeral prefixes corresponding to the numerals between 11 and 99 are discussed in §7 .3.1.2. 7.1.4 Hundred and ab ov e There are two ways of expr essing numb ers above 99 in Japhug. First, the noun- like numeral ɣurʑa ‘hundred’ can occur on its own or be followed by another numeral to express a number between 101 and 199, as in (1). (1) aʑo 1sg kɯ-fse sbj:pcp-be.like kɯ-ɤcʰɯcʰa sbj:pcp-be.capable ʑo emph ʁʑɯnɯ young.man ɣurʑa hundred kɯrcat eight ra need:f act ‘I need one hundred and eight able young men like me. ’ (Norbzang) {00037 68#S16} The numeral ɣurʑa ‘hundred’ cannot be combine d with single digit numerals to express numb ers between 200 and 900. The counted noun 1 tɯ-ri ‘ one hundred’ is used for this purp ose , as in 2. The two suppletive roots for hundreds are shar ed with Pumi (the numeral ɕí ‘hundred’ vs. the counted noun -ɻɛj , se e Daude y 2014: 101; evidence for cognacy with ɣurʑa and tɯ-ri is presented in Jacques 2017 c). (2) χsɯ-ri three-hundred jamar about ndɤre lnk tu-nɯ exist:f act-pl ko, sfp tɯ-tɯpʰu one-hive nɯ dem ‘There ar e ab out three hundred of them in one hiv e. ’ (26-GZo) {0003668#S48} Numerals above the hundreds are all borro wed from Tibetan: stoŋtsu ‘thou- sand’ , kʰrɯtsu ‘ten thousands’ , mbɯmχtɤr ‘hundred thousands’ originate from ོང་ཚ་ stoŋ.tsʰo ‘thousand’ , ི་ཚ་ kʰri.tsʰo ‘ten thousands’ and འམ་ ཐེར་ ⁿbum.tʰer ‘hun- dred thousands’ , respectively . Like other numerals, they are postnominal, as shown by (3). (3) ɴɢoɕna spider ɣɯ gen nɯ, dem ɯ-pɯ 3sg.poss-young stoŋtsu thousand tu exist:f act tu-ti-nɯ ipfv-say-pl ɲɯ-ŋu. sens-be qaɟɯ fish ɣɯ gen nɯ, dem kʰrɯtsu ten.thousands tu exist:f act tu-ti-nɯ ipfv-say-pl ɲɯ-ŋu. sens-be ‘People say that spiders have a thousand offspring, and fishes ten thousands. ’ (26-mY aRmtsaR) {0003674#S111} 1 Counted nouns are nouns that take an obligatory numeral prefix (§7 .3). 241 7 N umerals and counted nouns The numerals thousand and above can take other numerals as multiplicative modifiers, as in (4), where kɯtʂɤɣ ‘six’ follows stoŋtsu ‘thousand’ to expr ess ‘six thousand’ . This use illustrates the difference between the numerals stoŋtsu ‘thou- sand’ and above with ɣurʑa ‘hundred’ , as when the latter is followed by a nu- meral, as ɣurʑa kɯrcat ‘ one hundred and eight’ , it can only b e in additiv e, not multiplicative , relation to it (see example 1 abov e). The multiplicative numeral modifier to stoŋtsu ‘thousand’ cannot b e pr epose d (T shendzin says of a combina- tion like † kɯtʂɤɣ stoŋtsu that nɯ mɤ-sɤtso ‘it do es not make sense ’). (4) rŋɯl money tɯ-xpa one-year tɕe lnk stoŋtsu thousand kɯtʂɤɣ six jarma about ɲɯ-fsoʁ ipfv-earn ɲɯ-cʰa sens-can ‘He can earn about six thousand [renminbi] per year . ’ (14-siblings) {0003508#S159} T o hav e an additive interpretation, the comitative cʰo ‘and, with’ ( or cʰondɤre ‘with’) is used as in (5) and (6). (5) stoŋtsu thousand ci one cʰo comit kɯtʂɤɣ-ri six-hundred ‘One thousand six hundred (1600). ’ ( elicited) (6) stoŋtsu thousand ci one cʰondɤre comit kɯtʂɤɣ six-hundred ‘One thousand and six (1006). ’ ( elicited) In (7), a multiplicative modifier sqi ‘ten’ following kʰrɯtsu ‘ten thousands’ is used instead of mbɯmχtɤr ‘hundred thousands’ . The latter numeral, although known to native speakers, is hardly ever used in speech. (7) ɯ-kɤ-nɯkon 3sg.poss-obj:pcp-rule nɯnɯ, dem rɟɤlkʰɤβ country kʰrɯtsu ten.thousands sqi ten pjɤ-tu. ifr.ipfv-exist ‘There wer e a hundred thousand countries under his rule. ’ (140518 jinyin chengbao-zh) {0004028#S12} Unlike numerals under 100, ɣurʑa ‘hundred’ and above are alienably possessed nouns and can take a third person possessive prefix ɯ- to expr ess an approximate number (§7 .2). 242 7 .1 P lain numerals 7.1.5 Ordinals There are no native ordinal numbers in Japhug, even for ‘first’ , which is ex- pressed by combining the superlative stu ‘most’ (§26.4.1) with the subject partici- ple (§16.1.1) of the verb mɤku ‘be first’ (§20.6). This minimal participial relativ e stu kɯ-mɤku (§23.5.1) can b e used as prenominal modifier as in (8), but most commonly occurs adverbially as in (9). (8) stu most kɯ-mɤku sbj:pcp-be.first tɯ-xpa one-year nɯ dem ɲɯ-rɯmɯntoʁ ipfv-have .flower ma, lnk nɯ dem ma apart.from ɯ-mat 3sg.poss-fruit me not.exist:f act ‘The first year , it has flowers, but no fruits. ’ (08-qaCti) {0003456#S33} (9) stu most kɯ-mɤku sbj:pcp-be.first nɯ dem a-pi 1sg.poss-elder .sibling kɯ erg pɯ  -wɣ-sat-a, aor-inv-kill-1sg nɯ dem ɯ-qʰu 3sg.poss-after tɕe, lnk pɣɤtɕɯ bird tɤ-sci-a, aor-be.born-1sg ‘First, my elder sister killed me, and then I was reborn as a bird. ’ (2005-stod-kunbzang) With counted nouns (§7 .3), it is strictly prenominal, and the counted noun can be converted to an inalienably p ossessed noun (§7 .3.4.1); for instance, the noun phrase (10) can be use d instead of stu kɯ-mɤku tɯ-xpa in (8). The phrase stu kɯ-mɤku does not occur p ostnominally . (10) stu most kɯ-mɤku sbj:pcp-be.first ɯ-xpa 3sg.poss-year ‘The first year’ ( elicitation based on 8) There is no specific word meaning ‘last’ in Japhug either , and the participial relative stu kɯ-maqʰu with the subject participle (§16.1.1) of maqʰu ‘b e after’ is used instead; as in the case of stu kɯ-mɤku , counted noun to inalienably pos- sessed noun conversion is possible, and for instance both (11) and (12) exist. (11) stu most kɯ-maqʰu sbj:pcp-be.after tɯ-sŋi one-day (12) stu most kɯ-maqʰu sbj:pcp-be.after ɯ-sŋi 3sg.poss-day ‘The last day’ ( elicite d) 243 7 N umerals and counted nouns Given the absence of native ordinals, Chinese ordinals ar e use d (§7 .1.6.2). Ti- betan ordinals are only attested for names of months (§7 .1.6.1) and even this use is disappearing, as they are being replaced by their Chinese equivalents. Unlike many languages of the Sino- Tibetan family , the system of birth-order ordinals is very rudimentar y . The eldest son is simply calle d stu kɯ-wxti ‘the one who is the biggest’ as in (13). (13) ɯ-rɟit 3sg.poss-child stu most kɯ-wxti sbj:pcp-be.big nɯ dem rɯntɕʰɯn anthr tʂoma anthr rmi be.called:f a ct ‘Her eldest child is called Rinchen Sgrolma. ’ (12-BzaNsa) {0003484#S59} The only dedicated birth-order ordinal is tɯlɤt ‘second sibling’ ( se e §5.1.2.4 and §5.2.3). The inalienably possessed noun ɯ-pa ‘(the one) b elow’ can be used to designate any following sibling, and the youngest sibling is designated by the superlative form of the participle of xtɕi ‘be small’ (see example 14, which illustrates all these terms). (14) stu most kɯ-wxti sbj:pcp-be.big cʰondɤre comit nɯ dem ɯ-pa 3sg.poss-under nɯ dem tɯlɤt second.sibling ni du wuma really ʑo emph pjɤ-ɕqraʁ-ndʑi. ifr.ipfv-be.intelligent-du tɕe lnk nɯ dem ndʑi-tɕɯ 3du.poss-son stu most kɯ-xtɕi sbj:pcp-be.small nɯnɯ, dem dɯxpa poor ma lnk wuma really ʑo emph pjɤ-kʰe. ifr.ipfv-be.stupid ‘The eldest son and the one under him, the second one, wer e very intelligent. Their smallest son, poor him, was ver y stupid’ (140430 jin e-zh) {0003893#S5} 7.1.6 Tib etan and Chinese numerals In addition to the native numerals pr esente d abov e, Tibetan and Chinese nume- rals commonly occur in Japhug stories and conversations. 7.1.6.1 Tib etan numerals In Japhug, Tibetan numerals (see T able 7 .1 above) ar e mainly used in fixed ex- pressions. 2 Some of these expr essions exclusively comprise Tibetan words, for 2 Only a minority of Japhug speakers are fluent in Amdo or other Tibetic languages, so that Tibetan loanwords in Japhug, and in particular numerals, cannot be considered to be code- switching. 244 7 .1 P lain numerals instance luskɤr χtɕɯʁɲiz ‘twelve y ear cycle ’ from ལོ་ ར་ lo.skar ‘year of the cy- cle ’ and བ་ གཉིས་ btɕu.gɲis ‘twelve ’ . Other compounds combine Tib etan numerals with Japhug roots, for instance the possessive compound rnaftɕɯχa ‘having ears with ten holes’ from tɯ-rna ‘ ear’ ( or Tib etan ་ rna ‘ ear’ , the form is ambiguous between cognate and b orro wing), the numeral ftɕɯ- (from བ་ btɕu ‘ten’) and the verb aχa ‘lacking a piece ’ , which appears as p ostnominal modifier in the name of a trickster character qala rnaftɕɯχa ‘the rabbit with ten holes in his ears’ . The Japhug form -χtɕɯɣ- of the Tibetan numeral གཅིག་ gtɕig ‘one ’ occurs in com- pounds, but is not attested as an indep endent wor d; howe ver , it is possible that it used to exist in Japhug at an earlier stage of the language (though perhaps not as a numeral), as the stative v erb naχtɕɯɣ ‘b e the same ’ is a denominal derivation based on this numeral (§9.1.7, §20.7 .1). The Tibetan ordinal numbers up to ten (and the cardinal numbers for ‘ eleven’ and ‘twelv e’), indicated in T able 7 .4, can b e used for months names (T ournadr e & Suzuki 2023: #406). This usage is not attested in conversations, but does o ccur in texts when speakers consciously tr y to av oid using Chinese words. In example (15), T shendzin first used the Chinese 三 月 份 <sānyuèfèn> ‘third month’ and then corrected it to its Tib etan equivalent. (15) <sanyuefen> third.month jamar about zlawa month χsɯmba third jamar about tɕe lnk ci one cʰɯ-rɤpɯ, ipfv-have .young ‘It bears young once in the third month. ’ (21-lWLU) {000357 6#S61} T able 7 .4: Tibetan ordinals used in Japhug Japhug Tibetan 1 taŋbu དང་པོ་ daŋ.po ‘first’ 2 ʁɲispa གམ་པ་ gɲis.pa ‘second’ 3 χsɯmba གམ་པ་ gsum.pa ‘third’ 4 βʑɯpa བཞི་པ་ bʑi.pa ‘fourth’ 5 rŋapa ་པ་ lŋa.pa ‘fifth’ 6 tʂɯxpa ག་པ་ drug.pa ‘sixth’ 7 βdɯnpa བན་པ་ bdun.pa ‘seventh’ 8 βɟɤtpa བད་པ་ brgʲad.pa ‘ eighth’ 9 rgɯpa ད་པ་ dgu.pa ‘ninth’ 10 ftɕɯpa བ་པ་ btɕu.pa ‘tenth’ 11 ftɕɯχtɕɯɣ བ་གཅིག་ btɕu.gtɕig ‘ eleventh’ 12 ftɕɯʁɲiz བ་གཉིས་ btɕu.gɲis ‘twelfth’ 245 7 N umerals and counted nouns The ordinals can be used with the Tib etan wor d zlawa ‘month’ (from ་བ་ zla.ba ‘moon, month’) as in (15), or without it as in (16) from the same text. (16) rgɯpa ninth jamar about tɕe lnk ci one cʰɯ-rɤpɯ ipfv-have .young ɲɯ-ŋu. sens-be ‘It bears young once in the ninth month. ’ (21-lWLU) {000357 6#S63} In traditional stories of Tibetan origin, complete noun phrases in Tib etan in- cluding numerals can be found. For instance, in (17), the numeral χsɯmrɟatʂɯɣftɕɯ from གམ་བ་ག་བ་ gsum.brgʲa.drug.btɕu ‘three hundred sixty’ occurs with the noun pja (from Tibetan ་ bʲa ‘bird’; not normally used in Japhug except in compounds) and the adverb mɯndʐamɯχtɕɯɣ ‘all kinds’ (from མི་འ་མི་གཅིག་ mi.ⁿdra.mi.gtɕig ‘ di- verse , different’ , §9.1.3.5). (17) pja bird mɯndʐamɯχtɕɯɣ, all.kinds χsɯmrɟatʂɯɣftɕɯ 360 kɯ erg pɣɤmbri bird.song ɲcɣɤnɤɲcɣɤt idph(III):very .noisy ʑo emph to-lɤt-nɯ, ifr-throw-pl ‘Three hundred sixty birds of all kinds sung bir d songs. ’ (2003 smanmi) 7.1.6.2 Chinese numerals Chinese numerals are ubiquitous in Japhug, and many younger people (b orn after 1990) have trouble counting ab ov e twenty , ev en if they other wise speak Japhug fairly fluently . Even the most fluent speakers use Chinese numerals ( with their appropriate Chinese classifiers) when speaking normally . In examples (18) and (19), from a te xt describing the hamlets in the township of Gdong-brgyad, the Chinese expressions 全乡 <quánxiāng> ‘the whole to wnship’ , 三千多人 <sānqiānduōrén> ‘three thousand people and a little more ’ and 两 个 小队 <liǎnggèxiǎoduì> ‘two groups’ ( with the generic classifier 个 <gè> ‘cl’) are used instead of Japhug native numerals and quantifiers to count the number of people in the area. Note the use of the Cultural Revolution term 小队 <xiǎoduì> ‘team, group ’ as a counting unit. The use of Chinese here is due to the fact that administration-related counting, even in villages, is done in Chinese. (18) nɯnɯ dem ʁdɯrɟɤt anthr <quanxiang> the.whole .township nɯnɯ dem <sanqianduoren> three.thousand.people ma apart.from maŋe. not:exist:sens ‘There are only a little more than three thousand people in the township of Gdong-brgyad. ’ (140522 RdW rJAt) {0004061#S109} 246 7 .1 P lain numerals (19) kɤmɲɯ Kamnyu nɯ dem <lianggexiaodui> two .groups ma apart.from maŋe not.exist:sens ‘In Kamnyu ther e are only two xiaodui . ’ (140522 RdW rJAt) {0004061#S111} Given the absence of ordinal numerals in Japhug (§7 .1.5), Chinese expr essions with the ordinal marker 第 /dì/ are used instead. In example (20), the Chinese phrase 第一名 <dìyīmíng> ‘the first (in a competition)’ occurs, directly followed by the participial clause stu kɯ-mɤku ‘the first one ’ . (20) tɕe lnk βʑɯ mouse nɯ dem kɯ erg <diyiming> first tɕe lnk stu most kɯ-mɤku sbj:pcp-be.first pjɤ-mɟa. ifr-take ‘The mouse obtained the first place. ’ (150826 shier shengxiao-zh) {0006284#S96} 7.1.7 Use of the numerals Japhug numerals can occur on their own when counting ( ci , ʁnɯz , χsɯm , kɯβde ...) or be use d as postnominal attributive modifiers (§9.3). The noun can be elided when the context is clear , especially when the same referent occurs in the previ- ous proposition as in (21) with ɯ-rɟit ‘her children’ and (22) with 菜 <cài> ‘ dish’ . In this case the numeral constitutes the head of the noun phrase. (21) ɯ-rɟit 3sg.poss-child kɯngɯt nine tɤ-tu aor-e xist ri, lnk kɯtʂɤɣ six nɯ-si aor-die ‘She had nine children, but six [ of them] died. ’ (14-siblings) {0003508#S16} (22) <cai> dish χsɯm three tu-sɯ-lɤt-i ipfv-ca us-throw-1pl tɕe lnk tɕe lnk kɯβde four nɯ dem kɯ erg χsɯm three tu-ndza-j ipfv-eat-1pl kɯ-fse. sbj:pcp-be.like ‘W e used to order three dishes, and the four of us would eat [the] three [ of them]. ’ (140501 tshering skyid) {0003902#S90} Numerals can also be mo difiers of dual and plural pr onouns as in (23). Since pronouns are nev er obligator y in Japhug (§22.1.2), it is also possible to use a bare numeral in core argument function with first or second person indexation on the verb , as in (22), where the v erb tu-ndza-j ‘we eat’ of the second proposition has the 1pl -j suffix indexing the transitive subject, corefer ent with the ergatively- marked phrase kɯβde nɯ kɯ ‘the four’ (standing for iʑo kɯβde nɯ kɯ ‘the four of us’). 247 7 N umerals and counted nouns alternation is optional and there is free variation between the two forms. In the corpus, we find 15 examples of -sqi- and 14 of -sqɯ- , suggesting that both are about e qually common. (33) tɯ-pʰɯ one-tree nɯ dem tɕe lnk rcanɯ, unexp:deg li, again fsɯsqi-ldʑa thirty-long.object jamar , about kɯβdɤsqɯ-ldʑa forty-long.object jamar about tu. exist:f act ‘On one tree, ther e are about thirty or forty [ branches]. ’ (14-sWNg WJu) {0003506#S186} Example (34) illustrates three alternative forms with the counted noun tɤ-rʑaʁ ‘ one night’: ɣnɤsqaptɯ-rʑaʁ with regular loss of coda, ɣnɤsqaptɯɣ-rʑaʁ with preservation of the coda, and ɣnɤsqamnɯz tɤ-rʑaʁ as two wor ds, the numeral being a kind of prenominal modifier . The first form is regular , while the other ones each are hapax legomena . (34) tɕe lnk nɯ dem ɣnɤsqaptɯɣ-rʑaʁ twenty .one-night tu-tsu ipfv-pass ɲɯ-ra. sens-be.neede d “tɕe lnk ɣnɤsqaptɯ-rʑaʁ twenty .one-night tu-tsu ipfv-pass tɕe lnk ɲɯ-ʁaʁ ipfv-hatch ŋu” be:f act ɲɯ-ti-nɯ sens-say-pl ri, lnk aʑɯɣ 1sg:gen nɯ dem ɣnɤsqamnɯz twenty .tw o tɤ-rʑaʁ one-night mɤɕtʂa until mɯ-nɯ-ʁaʁ. neg-aor-hatch ‘[Eggs] need twenty-two days to hatch; people say ‘They hatch in twenty-two days’ but mine only hatch after twenty two days. ’ (150819 kumpGa) {0006388#S32} 7.3.1.3 Approximate numeral prefixes Approximate numerals also hav e corresponding prefixal forms. T able 7 .9 presents the forms attested in the corpus. By far the most commonly used approximate numeral pr efix is laʁnɯ-X , whose meaning is not ‘ one or two ’ as could have been expe cted from the use of the lɤ-/ la- prefix in the other examples, but ‘a fe w’ (see example 35 – life expectancy of goats and sheep is much ab o ve two years). 254 7 .3 Counted nouns T able 7 .9: Appro ximate numeral prefixes in Japhug Approximate Numeral Approximate Numeral Prefix laʁnɯz ‘a few’ laʁnɯ- lɤβdelɤŋu ‘four or five ’ lɤβdelɤŋu- lɤŋu ‘about five ’ lɤŋu- lɤŋɤtʂɤɣ ‘five or six’ lɤŋɤtʂɤ- , lɤŋɤtʂɤɣ- ɕnɤcat ‘seven or eight’ ɕnɤcɤ- (35) tsʰɤt goat qaʑo sheep nɯnɯ dem tɕe lnk laʁnɯ-xpa a.few-year ma apart.from cʰɯ-mdɯ ipfv-live .up.to mɯ  j-ŋgrɤl neg:sens-be.usually .the.case ma lnk tɕe lnk cʰɯ-rgɤz ipfv-be.old ɕti be.aff:f a ct ‘Goats and sheep only live for a few years, and then be come old. ’ (05-qaZo) {0003404#S128} Just like approximate numbers can also b e expr essed by juxtap osition of nume- rals (§7 .2), it is p ossible to juxtapose counted nouns (the same counted noun with contiguous numeral prefixes, as in 36), or to combine a numeral with a counted noun whose numeral prefix is contiguous to it, as in (37). (36) tɕe lnk nɯnɯ dem χsɯ-zɯm three-bucket kɯβde-zɯm four-bucket jamar about kɯ-xtɕʰɯt sbj:pcp-contain tu. exist:f act ‘There ar e [ jars] that can contain ab out thr ee or four buckets [of water ]. ’ (26-tChW ra) {0003690#S4} (37) <tuolaji> tractor tɯ-ɣjɤn one-time ju-ɣɯt ipfv-bring nɯ dem rcanɯ, unexp:deg ɣnɤsqi twenty fsɯsqi-fkur thirty-burden jamar about ju-sɯ-ɤzɣɯt ipfv-ca us-reach cʰa. can:f a ct ‘Each time the tractor brings [firew oo d], it can mo ve about twenty or thirty loads (the size a person can carr y on his back). ’ (140430 tWfkur ) {0003898#S22} Additionally , the paucal meaning can b e expr essed by converting the counted noun into an inalienably possessed noun with a third person p ossessiv e prefix and reduplication of the noun stem (see §7 .3.4.1). 255 7 N umerals and counted nouns 7.3.1.4 Other numeral prefixes In addition to the numerals mentioned ab o ve, all higher and compound numerals, an interrogative pronoun and a participle form can appear as prefixes of counted nouns. The numerals above 99 (§7 .1.4) occur as numeral prefixes without vow el alter- nation, as ɣurʑa-xpa ‘a hundred years’ and stoŋtsu-xpa ‘a thousand years’ (fr om ɣurʑa ‘hundred’ and stoŋtsu ‘thousand’). 5 The prefixal form of hundreds based on the counted noun tɯ-ri ‘one hundr ed’ have double numeral pr efixes, such as χsɯ-ri-xpa ‘three hundred years’ in (38). (38) tɕendɤre lnk χsɯ-ri-xpa three-hundred-year tɤ-tsu aor-pass tɕe, lnk li again kɯjŋu oath pɯ-ta-t-a aor-put-tr:pst-1sg tɕe, lnk ‘Three hundred years passed, and I made another oath. ’ (140512 yufu yu mogui-zh) {0003973#S87} In the case of complex numerals, such as the v er y common ɣurʑa kɯrcat ‘ one hundred and eight’ , only the last one undergoes b ound state, as in (39) and (40), where we find ɣurʑa as a separate phonological w ord follow e d by the counted noun with the numeral prefix kɯrcɤ- (from kɯrcat ‘ eight’). (39) ɕommbri iron.chain ɣurʑa hundred kɯrcɤ-ɟom, eight-length.of.two .outstretched.arms [...] ... ɕomtsʰoʁ iron.nails ɣurʑa hundred kɯrcɤ-ldʑa eight-long.object ra be.neede d:f act ‘[I] need a chain of one hundred and eight fathoms, and one hundred and eight nails. ’ (2003 tW xtsa) (40) ɣurʑa hundred kɯrcɤ-ɣdɤt eight-section qapri snake nɯ dem a-tɤ-ɕe irr-pfv:up-go ra be.neede d:f act ‘May the snake be cut into hundred and eight se ctions!’ (2012 Norbzang) {00037 68#S243} 5 Even in the absence of vo wel alternation, the prefixal status of these numerals is shown by three pieces of evidence: (i) the fact that they combine with non-free elements like the stem -xpa ‘year’ , which require a numeral prefix, (ii) the absence of stress and (iii) the inability to pause or insert any element in between, except in the case of spee ch errors. 256 7 .3 Counted nouns The interrogative pr onoun tʰɤstɯɣ ‘how many’ has the prefixal form tʰɤstɯ- with counted nouns (§ §6.5.3). The subject participle kɯ-ɤntɕʰɯ of the stative verb antɕʰɯ ‘be many’ , has the prefixal form kɤntɕʰɯ- , as in (41). This prefix is very common in the corpus. (41) tɕe lnk nɯnɯ dem kɤntɕʰɯ-tɯpɯ many-household ɣɯ gen nɯ-tɯrsa 3pl.poss-grave ɯ-sta 3sg.poss-place ɲɯ-ŋu sens-be ma lnk ‘It is the grave-place of many families. ’ (140522 kAmYW tWji) {0004055#S91} It is howe ver not possible to convert any noun, pronoun or quantifier into a numeral prefix; with other wor ds, it is necessar y to convert the counted noun to an inalienably possessed noun, with a third singular ɯ- prefix instead of the numeral prefixes, see §7 .3.4.1. The nouns tɯ-sla ‘ one month’ and tɯ-xpa ‘ one year’ can be use d with a nu- meral prefix kɤrɤ- not attested with other counted nouns (see §7 .5.1). 7.3.1.5 Irregular forms A handful of counted nouns, in particular tɤ-rʑaʁ ‘one night’ , have an alternativ e paradigm with /ɤ/ instead of ɯ in the prefixes, as shown by the forms ʁnɤ-rʑaʁ ‘two nights’ and χsɤ-rʑaʁ ‘three nights’ in (42). (42) ʁnɤ-rʑaʁ two-night jamar , about χsɤ-rʑaʁ three-night jamar about to-tsu ifr-pass tɕe lnk tɕendɤre lnk cʰa alcohol ɯ-di 3sg.poss-smell tu-mnɤm ipfv-have .a.smell ɲɯ-ŋu. sens-be ‘ After two or three nights, one can smell the smell of alcohol. ’ (1607 03 araR) {0006101#S39} The complete paradigm of this noun is presented in T able 7 .10. There is how ever some degr ee of variation, and using the regular paradigm is not considered erroneous. In the corpus, the numeral ‘ one ’ form tɯ-rʑaʁ ‘ one night’ is more common than tɤ-rʑaʁ , possibly be cause of the homophony with the inalienably possessed noun tɤ-rʑaʁ ‘time’ found in e xamples such as (44). For other numerals, the forms in T able (7 .10) ar e considerably more common than the regular ones. In addition, the ɤ vocalism is also found with other numeral prefixes such as the interrogativ e ( tʰɤstɤ-rʑaʁ ‘how many nights’). 257 7 N umerals and counted nouns T able 7 .10: Irregular numeral prefixes in Japhug Numeral Free form -rʑaʁ ‘night’ 1 tɤɣ tɤ-rʑaʁ 2 ʁnɯz ʁnɤ-rʑaʁ 3 χsɯm χsɤ-rʑaʁ 4 kɯβde kɯβdɤ-rʑaʁ 5 kɯmŋu kɯmŋɤ-rʑaʁ 6 kɯtʂɤɣ kɯtʂɤ-rʑaʁ 7 kɯɕnɯz kɯɕnɤ-rʑaʁ 8 kɯrcat kɯrcɤ-rʑaʁ 9 kɯngɯt kɯngɤ-rʑaʁ 10 sqi sqɤ-rʑaʁ (43) tɕe lnk qarma eared.pheasant nɯ, dem tɯ-rʑaʁ one-night tɕe lnk kɯβde four kɯmŋu five jamar about pjɯ-sat-nɯ, ipfv-kill-pl, tɯ-rdoʁ, one-piece tɯrme person tɯ-rdoʁ one-piece kɯ erg ‘Eared pheasants, in one night, each of [the hunters] can kill four or five of them. ’ (23-qapGAmtW mtW) {0003608#S146} (44) tɤ-rʑaʁ indef.poss-time tɤ-rɲɟi aor-be.long tɕe, lnk nɯ-ji 3pl.poss-field ra pl kɯ-dɯ~dɤn sbj:pcp-emph~be.many kɯ-jɯ~jom sbj:pcp-emph~be.broad lo-pɣaʁ-nɯ, ifr-turn.over-pl ɕoŋtɕa timber kɯ-dɯ~dɤn sbj:pcp-emph~be.many pjɤ-pʰɯt-nɯ ifr-remov e-pl ‘ After some time/as time went on, [those people] had ploughe d many broad fields for them, and chopped a lot of timb er . ’ (2002 qajdoskAt) {0003366#S115} Apart from tɤ-rʑaʁ ‘ one night’ , counted nouns following the paradigm in T a- ble (7 .10) are very rare. The counted nouns tɯ-tɣa ‘ one span’ and tɯ-rtsɤɣ ‘one storey’ hav e the irregular forms χsɤ-tɣa ‘three spans’ and χsɤ-rtsɤɣ ‘three storeys’ ( comp eting with r egular χsɯ-tɣa and χsɯ-rtsɤɣ ) as in (45), but not for other nu- meral prefixes. 258 7 .3 Counted nouns (45) nɯ dem χsɤ-tɣa three-span kɯβde-tɣa four-span jamar about tu-rɲɟi ipfv-be.long cʰa. can:f a ct ‘It can grow three or four spans long. ’ (14-sWNg WJu) {0003506#S181} Another unrelated irregularity concerns the counted noun tɯ-ɣjɤn ‘ one time ’: free variation between -jɤn and -ɣjɤn is obser ved for the numerals ‘two ’ and ‘three ’ (b oth χsɯ-ɣjɤn ‘three times’ and χsɯ-jɤn are attested). A similar case is obser v ed with counted noun tɯ-xpa ‘one y ear’; the form -xpa is obligatory for one to three ( ʁnɯ-xpa ‘tw o years’ , χsɯ-xpa ‘three years’), but for ‘four’ on, both stems -xpa and -pa are attested ( both sqɯ-xpa ‘ten years’ and sqɯ-pa ‘ten years’ are possible , though the former is more common), including for approximate numerals ( ɕnɤcɤ-pa ‘se ven or eight years’). The counted noun tɯ-zloʁ ‘one time ’ (see §7 .6 on its use) is attested in (46) with the special form ʁɲɯ-zloʁ ‘two times’ whose first element is a numeral prefix form ʁɲɯ- influenced by the Tibetan numeral གཉིས་ gɲis ‘two ’ (in Japhug pronunciation ʁɲiz- ). The regular form ʁnɯ-zloʁ ‘two times’ also exists. (46) a-zda 1sg.poss-companion ɣɯ gen kɯ-fse sbj:pcp-be.like ʁɲɯ-zloʁ two-times nɯ dem tu-ndze-a ipfv-eat[III]-1sg ɲɯ-tʂaŋ sens-be.fair ‘It would be fair if I had two times as much to eat as the other one. ’ (140426 lv he luozi-zh) {0003816#S9} The stem -zloʁ looks like a blend from the two Tibetan words ོག་ zlog ‘turn around’ and ོ zlo ‘repeat’ . Tibetan numerals are generally post-nominal, but pre- nominal numerals also exist as in ཉིས་བ་ ɲis.ldab ‘two times’ . 7.3.1.6 Distributed numeral prefixes The distributed form of counted nouns is built by reduplicating the numeral ‘one ’ prefix tɯ- , as tɯ-tɯ-rdoʁ from the generic counted noun tɯ-rdoʁ ‘ one piece ’ (see §7 .3.3) in example (47). (47) ʑɯrɯʑɤri progressiv ely qʰe, lnk tɕe lnk tɯ-tɯ-rdoʁ one-one-piece nɯ dem ɲɯ-ʑɣɤqɤr-nɯ ipfv-get.separated-pl qʰe lnk ‘Progr essively , some of them get separated [from the herd]. ’ (20-Rmbr oN) {0003560#S55} 259 7 N umerals and counted nouns Example (48) with tɯ-tɯ-xpa ‘some years’ from tɯ-xpa ‘ one year’ clearly il- lustrates the functional difference with appro ximate numerals (§7 .2 and §7 .3.1.3), as the meaning of the distributed counted noun cannot b e translated here as ‘a few years’ – the years when the income is good are not necessarily contiguous in time. (48) tɯ-tɯ-xpa one-one-year tɕe lnk a-pɯ-pe irr-ipfv-be.good tɕe, lnk kʰrɯtsu ten.thousand ɯ-ro 3sg.poss-excess jamar about ɲɯ-fsoʁ ipfv-earn ɲɯ-cʰa sens-can ‘Some years if [his income] is goo d, he can earn mor e than ten thousands [renminbi]. ’ (14-siblings) {0003508#S161} An alternative distributed form involv es partial reduplication of the stem of the counted noun and replacing the numeral prefix by a third singular posses- sive ɯ- , as in the conversion from counted noun to inalienably possesse d noun (§7 .3.4.1). In (§49), the counte d noun tɯ-pʰɯ ‘ one tr e e ’ is changed to ɯ-pʰɯ~pʰɯ ‘some trees’ . The other distributed form tɯ-tɯ-pʰɯ could also b e used in the same context without meaning difference . (49) zgoku mountain kɯ-mbro sbj:pcp-be.tall tɕe, lnk ɕkrɤz oak kɯ-wxti sbj:pcp-be.big ra pl nɯ dem ɯ-rcʰɤβ 3sg.poss-between ri loc ɯ-pʰɯ~pʰɯ 3sg.poss-p art~tree ʑo emph tu exist:f act tɕe lnk ‘On high mountains, among big oaks, there are some [ of these little trees]. ’ (16-CW rNgo) {0003518#S168} This distributed form puts emphasis on the non-contiguousness and spread over distribution of the entities designated by the reduplicate d counted noun: in (§49) for instance, its pr esence implies that the little trees are not clustered, but rather scattered among the oaks. Distributed counted nouns can also b e repeated to put even more emphasis on the scattered distribution (see §7 .3.2.3). 7.3.1.7 Historical p erspectives on the numeral prefixal paradigm Most non- Tib etan languages of W estern Sichuan/Northern Y unnan have counted nouns (generally called ‘ classifiers’ , see §7 .3.3) with numeral-counted noun order (see for instance Zhang 2014, Michaud 2017: 163–194). It is striking that among the quasi-isolating languages (Lolo-Burmese, Naish), numeral prefix paradigms 260 7 .3 Counted nouns are commonly a pocket of irregular morphology (Bradley 2005, Michaud 2011); this is also true in some Hmong-Mien languages (see Gerner & Bisang 2010). By contrast, while Japhug and the other Gyalrong languages hav e a richer mor- phology in general, the numeral prefixal paradigms are , with only few exceptions (§7 .3.1.5), suspiciously regular . The historical interpretation of this observation is not completely straightforward (Jacques 2017 c), but in any case counted nouns are morphologically like determinative compounds (§5.5.1.1) with a numeral as first element, undergoing bound state in the case of more integrated numerals (some of the numerals under 100, see §7 .3.1.1, §7 .3.1.2 and §7 .3.1.3) and immune from it in the case of higher numerals (§7 .3.1.4). One morphological alternation common to all numeral prefixes under 10 is the loss of the codas, other wise a rare phenomenon in noun compounds (see §5.4.2.2). For numerals above 10, there is some degr ee of free variation in the preser vation of the codas (see for instance example 34 p.254); the forms pr eser ving codas are rarer , and may be ongoing analogical lev elling. Even in the case of low er numerals, there are indir ect traces of the former existence of codas, in particular in irregularities (§7 .3.1.5). A few counted nouns like tɯ-xpa ‘ one year’ and tɯ-ɣjɤn ‘ one time ’ have velar fricative pr einitials x-/ ɣ- which are lacking in some forms of the paradigm. In the case of tɯ-xpa ‘ one year’ , note the alternative stem -pa ( sqɯ-xpa ‘ten years’ vs. sqɯ-pa ‘ten years’), the time or dinals (§7 .5.2), the adverb pakuku ‘e very year’ and the verb pa ‘pass 𝑋 y ears’ (see example 119, §22.4.1.4) from which the counted noun tɯ-xpa ‘ one year’ is historically derived. Cognates of this root -pa appear in other Burmo-Gyalrongic languages such as Naish (Jacques & Michaud 2011) but are restricted to time ordinals, and do not occur as counted nouns. Even in the closely related Stau language, time ordinals such as javə ‘last y ear’ and pəvə ‘this year’ have a r oot -və cognate to Japhug -pa , but the corresponding counted noun e-fku ‘ one year’ has a different root which left no trace in the Gyalrong languages. The generalization of the -pa root to the counted paradigm is probably a common Gyalrong innovation ( see §7 .3.4.3). A possible explanation for this x-/ ɣ- element is that it originates from the coda of the numeral ‘ one’ ( although replaced by ci ‘ one’ , this former numeral is still found in the element -tɯɣ in sqaptɯɣ ‘ eleven’) thr ough false segmentation ( *tɯk-pa → *tɯ-kpa → *tɯ-xpa ‘ one year’) and subsequent generalization to the whole paradigm. 6 It is not an example of x-/ɣ- nominalization (§16.5.2). 6 Note that since proto-Gyalrong *kp- r egularly yields βɣ- with metathesis (Jacques 2004: 272), this false segmentation must have occurred after the *kp- → βɣ- sound change, which is not shared with other Gyalrong languages. 261 7 N umerals and counted nouns 7.3.2 Counte d nouns in quantifying function This section discusses the use of counted nouns in quantifying function, as noun modifiers, head of a noun phrase or a sentential quantifier . The main functions of counted nouns include partitive, distributiv e, restrictive and iterative meanings. 7.3.2.1 Partitive When occurring as postnominal mo difiers, most counted nouns are essentially partitive in meaning, referring to a certain number of individuals from a gr oup, and are not used to express indefiniteness (§7 .3.2.7): the polyfunctional deter- miner ci ‘ one ’ (§9.1.4.1) occurs instead in this meaning. Thus, a phrase such as tɯrme tɯ-rdoʁ combining the noun tɯrme ‘person’ with the generic counted noun tɯ-rdoʁ ‘ one pie ce ’ ( see § §7 .3.3 on counted noun selection) is generally either to be translate d as a partitive ‘ one of them’ as in (50). (50) tɯrme people ʁnɯz two pjɤ-tu ifr.ipfv-exist tɕe, lnk tɯrme people tɯ-rdoʁ one-piece nɯ dem rcanɯ, unexp:deg ɯ-stɤrju 3sg.poss-truth ʁɟa completely ʑo emph tu-βze ipfv-do[III] tɕe, lnk ‘There wer e two persons, one of them always told the truth (and the other one was a liar ). ’ (140427 yuanhou-zh) {000387 0#S3} The partitive meaning is found ev en when the individual counted noun o ccurs on its own without ov ert head noun as in (51). (51) ɯ-tɕɯ 3sg.poss-son kɯβde four pɯ-tu pst.ipfv-exist ri, but ɯʑo 3sg kɯ-fse sbj:pcp-be.like kɯ-ɕqraʁ sbj:pcp-be.intelligent tɯ-rdoʁ one-piece cinɤ even pɯ-me pst.ipfv-not.exist ɲɯ-ŋu sens-be ‘He had four sons, but not even one of them was smart like him. ’ (2005 tA wa k W cqraR) The partitive meaning is found in particular when counted nouns modify mass nouns, such as paʁɕa ‘pork’ in (52). 262 7 .3 Counted nouns (52) pakuku every .year ʑo emph paʁɕa pork tɯ-rdoʁ, one-piece qʰe lnk tɕe lnk tɕe lnk stoʁ broad.bean cʰɯ-ɣɯt, ipfv:downstream-bring ‘Every year , he would bring a piece of pork and broad beans. ’ (140501 tshering skyid) {0003902#S34} 7.3.2.2 Distributive Counted nouns can also have a distributive meaning ‘ each of them’ when another numeral or counted noun occurs in the same clause, indicating that each of the members of the group performs the same action, the verb can either receive plu- ral (as in 53 and 43) or singular indexation ( example 54). This is a particular case of fluid number indexation (§14.6.1). (53) tɕe lnk tɕe lnk tɯrme person tɯ-rdoʁ one-piece kɯ erg kʰɯna dog ʁnɯz two χsɯm three jamar about tu-ndo-nɯ ipfv-take-pl ‘Each [ of the hunters] takes two or three dogs. ’ (150829 KA GW cAno) {0006420#S24} (54) tɯrme people tɯ-rdoʁ one-piece kɯ erg cʰɤmdɤru drinking.straw tɯ-ldʑa one-cl tu-nɯ-ndɤm ipfv-a ut o-take[III] ‘Each person takes one straw . ’ (30-tChorzi) {00037 60#S33} Counted noun with a colle ctiv e meaning (§7 .3.3.2) such as tɯ-tɯpɯ ‘one house- hold’ also frequently have a partitive meaning, especially when combined with a numeral. (55) tɕe lnk tɯ-tɯpɯ one-household tɕe, lnk ɯ-qaʑo 3sg.poss-sheep kɯβdɤsqi, forty kɯmŋɤsqi fifty jamar about pɯ-tu. pst.ipfv-exist ‘Each household used to have about forty or fifty she ep . ’ (160712 smA G) {0006073#S23} T emporal counted nouns such as tɯ-sŋi ‘ one day’ (§7 .5.1) ar e generally re- peated when use d in distributive function in the meaning ‘ e ver y ( day )’ (§7 .3.2.3). Howe ver , in combination with other counted nouns, they can mean ‘per ... ’ as in (56), even without r epetition. 263 7 N umerals and counted nouns In these examples, adding an ov ert noun would be agrammatical, and the quan- tifiers tɯ-tɯ-kɯr cinɤ ʑo ‘not even one mouthful’ and laʁnɯ-ŋka ‘a few w ords’ cannot be analyze d as objects. Rather , they ar e sentential adverbs, whose gram- matical status is comparable to that of temporal counted nouns (§7 .5.1). 7.3.2.6 Other W e also find examples of counted nouns without partitive , distributive, r estricte d or iterative meaning. This the case in particular for counted nouns with a collective meaning, denot- ing a group of entities whose quantity can be precise ( tɯ-tɕʰa ‘ one pair’) or un- specifie d ( tɯ-boʁ ‘ one group ’ , tɯ-tɯpɯ ‘ one household’ or tɯ-tɯpʰu ‘ one hive ’). In (78) for instance, ʁnɯ-tɯpɯ means ‘tw o households’ , not ‘tw o of the house- holds’ or ‘ each of the households’ . Note also the plural (rather than dual) indexa- tion on the verb tu-nɯ , showing the inherent collective meaning of the counted noun tɯ-tɯpɯ ‘ one household’ . P lural indexation her e is optional (§14.6.1.1). In this example , ʁnɯ-tɯpɯ is not a mo difier of the placename taʁrdo : this place- name is an absolutive locative adjunct (§8.1.9). (78) tʰam now taʁrdo placename ʁnɯ-tɯpɯ two-household tu-nɯ exist-pl ‘Now there ar e two households in T arrdo. ’ (140522 tshupa) {0004053#S19} This number quantification meaning is also atteste d with non-collective coun- ted nouns. For instance, in (79), χsɯ-ldʑa ‘three (long objects)’ o ccurs in the same context as the numerals ci ‘ one ’ and χsɯm ‘three ’ . In (80) likewise, it is clear from the context that neither a partitive , distributive, nor restrictiv e interpretation is possible. (79) tɕendɤre lnk tɤ-tɕɯ indef.poss-son nɯ dem kɯ erg tɤɲi staff ci one na-ɕar aor:3 → 3 ′ -sear ch xɕɤndʑu twig χsɯ-ldʑa, three-long.object qapi white.stone χsɯm three kʰɤzɟi feedbag ɯ-ŋgɯ 3sg.poss-inside pa-rku aor:3 → 3 ′ -put.in ɲɯ-ŋu. sens-be ‘The boy looke d for a staff, and put thr e e twigs and thr ee stones in the (horse’s) feedbag. ’ (2005-stod-kunbzang) 27 0 7 .3 Counted nouns (80) tɕe lnk nɯtɕu dem:loc tɕe lnk tʂu path ɯ-rkɯ 3sg.poss-side zɯ loc si tree tɯ-pʰɯ one-tree pjɤ-tu, ifr.ipfv-exist ‘There , on the side of the road, there was a tree . ’ (The divination 2002) {0003364#S10} This function is not very widespread in native texts. In te xts translated from Chinese or elicited material, howev er , this usage is very common, as speakers will easily calque the Chinese noun+classifier construction. In example (81) for instance qaɟy tɯ-ldʑa ‘ one fish’ is very probably calqued from Chinese 一条鱼 yī-tiáo yú ( one-cl fish). 9 (81) qaɟy fish tɯ-ldʑa one-long.object cʰondɤre comit qacʰɣa fox ci one pjɤ-mto ifr-see ɲɯ-ŋu sens-be ‘He saw a fish and a fox. ’ (140505 xiaohaitu-zh) {0003921#S40} Future studies on the use of counted noun as quantifiers should be therefore exclusively based on texts not translated from Chinese and conversation, not on translations and elicitation. 7.3.2.7 Definiteness In Japhug, counted nouns are not specifically use d to mark indefiniteness; the y can even occur with demonstratives such as nɯ ‘that’ in noun phrases with a definite refer ent. In (82) and (83) for instance, the bowl of oil and the tree in question were mentioned earlier in the the stor y and are clearly definite . (82) kʰa house cʰɤ-zɣɯt ifr:downstream-r each tɕe lnk tɯ-kri indef.poss-oil tɯ-kʰɯtsa one-bowl nɯ dem ko-ɕkɯt. ifr-drink.completely ‘He arrived at the house and drank the bowl of oil. ’ (140501 mdzadi) {0003905#S23} (83) si tree tɯ-pʰɯ one-tree nɯ dem ɯ-pʰaʁ 3sg.poss-half ɯ-ntsi 3sg.poss-one.of.a.pair nɯ dem pjɤ-rom ifr.ipfv-be.dry ʑo, emph ɯ-pʰaʁ 3sg.poss-half ɯ-ntsi 3sg.poss-one.of.a.pair nɯ dem 9 The original text from which (81) was translated is 不久 ，老三又看见了一条鱼和一只狐狸 . The counted noun tɯ-ldʑa ‘one long object’ how ever is really applied to fishes, snakes and worms in Japhug ev en in non-translated texts (§7 .3.3). 271 7 N umerals and counted nouns pjɤ-k-ɤrŋi-ci ifr.ipfv-peg-be.green-peg ʑo, emph ‘One half of that tree was dead and the other half was green. ’ (The divination 2002) {0003364#S11} 7.3.3 Counte d nouns and semantic classes This section presents the semantic restrictions on the use of particular counted nouns. It also discusses the cases of nouns that are compatible with sev eral coun- ted nouns. 7.3.3.1 Non-colle ctive counted noun The counted noun tɯ-rdoʁ ‘ one pie ce ’ can occur as postnominal modifier with a considerable variety of nouns, designating people, animals, inanimate objects, including mass nouns; it is the counted noun by default. A minority of nouns select other individual counted nouns referring to specific shapes: tɯ-ldʑa ‘ one long obje ct’ , tɯ-mpɕar ‘ one leaf’ and tɯ-pʰɯ ‘ one tree ’ . The counted noun tɯ-ldʑa ‘one long object’ , like Chinese 一 条 <yitiáo> ‘ one long object’ or 一 根 <yigēn> ‘ one long object’ , occurs in the corpus with nouns belonging to the following semantic categories: • Stick-like objects: ɯ-ru ‘its stalk’ , tɤ-zrɤm ‘r oot’ , tɤɲi ‘walking stick’ , taqaβ ‘needle’ , tɯmɲa ‘arro w’ , dɤrʁɯ ‘fern’ or ndʑu ‘stick’ , ‘ chopsticks’ ( ex. 84) • Limbs, hair and other protruding body-parts: tɯ-mi ‘fo ot, leg’ , tɤ-kɤrme ‘head hair’ , tɤ-muj ‘feather’ , ta-ʁrɯ ‘horn’ . • Limbless animals: qapri ‘snake ’ , qaɟy ‘fish’ • T o wers: ɣɟɯ ‘watchtower’ (84) tɕi-xɕɤndʑu 1du.poss-twig χsɯ-ldʑa three-long.object pɯ-tu pst.ipfv-exist tɕe lnk ‘W e had three twigs. ’ (2003 Kunbzang) The counted noun tɯ-mpɕar ‘ one leaf’ o ccurs with flat objects, including tree leaves ( with tɤ-jwaʁ ‘leaf’ as in 85), sheets of cloth (with raz ‘ cloth’) and snowflakes ( with tɤjpa ‘snow’). 272 7 .3 Counted nouns (85) ɯ-jwaʁ 3sg.poss-leaf nɯ dem ʁnɯ-mpɕar two-leaf ma apart.from me not.exist:f act tɕe lnk ‘It only has two leaves. ’ (16-CW rNgo) {0003518#S17 6} The counted noun tɯ-mpɕar ‘one leaf’ can also r efer to money (in present-day China, bank notes are by far more common than coins e ven for small amounts of money ), meaning ‘ one renminbi’ . The counted noun tɯ-phɯ ‘ one tree’ occurs with the generic noun si ‘tree ’ or names of particular species. 7.3.3.2 Collective counted noun Most nouns referring to humans and animals can occur with the counted noun tɯ-boʁ ‘ one group ’ . Howe ver , the collective counted noun tɯ-ɟɯɣ ‘one pack’ ap- pears to b e exclusiv ely used with horses. Its only attestations in the corpus (for instance 86) are found in a translated text. (86) li again nɯ dem jamar about ki dem:pro x a-jɤ-tɯ-ɕe irr-pfv-2-go tɕe, lnk tɕendɤre lnk nɯtɕu dem:loc tɕe, lnk mɤʑɯ more mbro horse tɯ-ɟɯɣ one-herd tu exist:f act tɕe, lnk ‘Continue going again for about that distance, and there will be a pack of horse. ’ (150824 kelaosi-zh) {000627 6#S193} 7.3.3.3 Multiple counted nouns Some nouns are compatible with more than one counted noun postnominal mo- difier , with different connotations. For instance zgo ‘mountain’ can be used with the generic counted noun tɯ-rdoʁ ‘ one pie ce ’ as in (87), but also with tɯ-ldʑa ‘ one long object’ as in (88) in the meaning ‘mountain range ’ and also with tɯ- tɤɣmbaj ‘ one side ’ to mean ‘mountain face ’ . (87) nɯ dem kʰa house ɯ-rkɯ 3sg.poss-side zgo mountain tɯ-rdoʁ one-piece pɯ-ri aor-be.left ‘When there was only one mountain left [ on his way back] home, ... ’ (Lobzang03) (88) rŋgɯkɤta t opo nɯ dem li again zgo mountain ʁnɯ-ldʑa two-long.object tu exist:f act tɕe, lnk ‘[ As for the placename] Rngukata, ther e are also two mountain ranges. ’ (140522 Kamnyu zgo) {0004059#S62} 273 7 N umerals and counted nouns In other cases, the change of counted noun does not entail a radical semantic contrast. For instance, the noun si ‘tree ’ , while generally used with tɯ-pʰɯ ‘ one tree ’ , is also attested with the counte d noun tɯ-ldʑa ‘ one long object’ , as in ( 89), without clear semantic difference . (89) maka completely rdɤstaʁ stone cʰo comit si tree rcanɯ unexp:deg tɯ-ldʑa one-long.object cinɤ even.one ʑo emph kɯ-me sbj:pcp-not.exist scʰiz indef.loc kɤ-azɣɯt-ndʑi aor:east-r each-du ɲɯ-ŋu, sens-be ‘They arrived at a place wher e there was no stones and not even one tr ee. ’ (2003 Kunbzang) 7.3.4 Counte d nouns and other parts of spee ch This section presents the derivations from other parts of speech (including other nominal classes, such as inalienably and alienably possesse d nouns) into counted noun, and from counted noun into other classes. 7.3.4.1 Counte d nouns and inalienably possessed nouns Counted nouns and inalienably possesse d nouns stand out among other nouns in having an obligatory prefix. Since the citation form of b oth classes of nouns – the numeral ‘ one ’ prefix tɯ- and the indefinite possessor prefixes tɯ-/ tɤ- – are homophonous, it is not une xp ected that conversion occurs between the two classes. Given the fact that numeral prefixes ar e a closed class (see §7 .3.1 and in partic- ular §7 .3.1.4), when one ne eds to use a quantifier without numeral prefix equiv- alent, it is necessar y to convert the counted noun into an inalienably possessed noun in third person singular form, with the quantifier before it. In example (90), the quantifier nɯ tʰamtɕɤt ‘that many’ cannot b e converted to a prefix, and therefor e the counte d noun tɯ-tɯpʰu ‘ one type ’ is converted to an inalienably possessed noun in 3sg p ossessive form ɯ-tɯpʰu . (90) tɕe lnk paʁ pig tɯ-ɣjɤn one-time pjɯ  -wɣ-ntɕʰa ipfv-inv-butcher nɯ dem nɯ dem tʰamtɕɤt all ɯ-tɯpʰu 3sg.poss-type ɲɯ-ɬoʁ ipfv-come.out ra be.neede d:f act ‘Each time one kills a pig, one will get that many types [of foodstuff from it]. ’ (05-paR) {0003400#S89} 274 7 .3 Counted nouns Similarly , in (91), the more complex phrase ɣurʑa ɯ-ro ‘mor e than one hun- dred’ with the emphatic ʑo occurs with the 3sg prefix. (91) ɣurʑa hundred ɯ-ro 3sg.poss-excess ʑo emph ɯ-tɯpɯ 3sg.poss-household tu-j exist:f act-1pl ‘There ar e more than one hundred households of us. ’ (22-kumpGatCW) {0003590#S29} Example (92) illustrates a third case of conv ersion from counted noun to in- alienably possessed noun: a third singular possessive on the converted counted noun indicates here indefinite number , which makes it possible to spe cify the quantity as the predicate ( pjɤ-k-ɤntɕʰɯ-ci ‘they used to be many’) instead of the numeral prefix ( kɤntɕʰɯ-tɯpɯ ‘many households’ , see §7 .3.1.4). (92) kɯɕɯŋgɯ former .time nɯ, dem ɯ-tɯpɯ 3sg.poss-household pjɤ-k-ɤntɕʰɯ-ci ifr.ipfv-peg-be.many-peg nɤ, sfp ‘In former times, the households [there] wer e many . ’ (140522 tshupa) {0004053#S67} Conversion from counted noun to inalienably possessed noun is also obser- ved when a prenominal demonstrativ e or adnominal clause is present, as in (93) and (94). Conversion is how ever not obligatory , as shown by (95) (see additional examples in §7 .1.5). (93) nɯ dem ɯ-xpa 3sg.poss-year nɯ dem taχpa harvest wuma really pjɤ-pe ifr.ipfv-be.good ‘On that year , the harvest was really good. ’ (02-montagnes-kamnyu) {0003378#S60} (94) [arɕo be.finished.up:f a ct pɯ-ŋu] pst.ipfv-be ɯ-sŋi 3sg.poss-day nɯtɕu dem:loc tɕe loc ‘The day when [the appointed time] was ab out to be finished, ’ (2003 sras) (95) tɕendɤre lnk nɯ dem tɯ-sŋi one-day nɯnɯ dem mɯ-pjɤ-ko neg-ifr-defeat ‘On that day , he failed in his attempt [to force her to take him with her ]. ’ (02-deluge2012) {000337 6#S68} Finally , in the case of counted nouns expressing body-based units of length, conversion to an inalienably possessed noun has a very sp ecific meaning. These 275 7 N umerals and counted nouns units have a value that depends on the person of reference (few pairs of people have exactly the same handspan). The possessive pr efix ser ves to indicate the person whose b ody part serves as the reference as in a-tɣa ‘my handspan’ in ( 96). It is the only case that a counted noun converted to inalienably p ossessed noun can take a possessive prefix other than 3sg. (96) kɯ-zri sbj:pcp-be.long nɯra, dem:pl tɯ-tɣa one-span ma apart.from kɯki dem.pro x ɕaŋtaʁ up.fr om mɤ-zri, neg-be.long:f a ct aʑo 1sg a-tɣa 1sg.poss-span jamar about ci one ma apart.from me. not.exist:f act ‘ As for its length, it is at most one handspan, only the length of my handspan. ’ (28-tshA w Are) {0003722#S50} Conversion from inalienably possessed noun to counte d noun also exists, but is very marginal in Japhug. When inalienably p ossessed nouns are conv erted to counted nouns referring to a quantity , they ar e alienabilized (§5.1.2.9) and the nu- meral prefixes are added to the noun stem with its indefinite p ossessor prefix. For instance, the counted noun tɯ-tɯ-kɯr ‘ one mouthful’ (e xample 7 6 p.269) or tɯ- tɤ-ste ‘ one bladder of’ (97) are derived from the bo dy part inalienably possessed nouns tɯ-kɯr ‘mouth’ and tɤ-ste ‘bladder’ . (97) tɕe lnk ʁdɯxpanaχpu anthr ɣɯ gen ɯ-me 3sg.poss-daughter ɣɯ gen nɯ, dem tɯ-ci indef.poss-water tɯ-tɤ-ste, one-indef.poss-bladder tɤrcoʁ clay sɯrna figurine ci indef to-rku-nɯ, ifr-put.in-pl ‘They gave to Gdugpa Nagpo ’s daughter a bladder full of water and a clay figurine (to take with her on the road, as her dowry ). ’ (2003 smanmi2) Direct conversion fr om inalienably p ossessed noun to counted noun without alienabilization is rarer , and it is not always obvious whether the inalienably possessed noun, or the counte d noun is primary . For instance, the counted noun tɯ-qiɯ ‘ one half’ (see §7 .6.1) is likely to have been derived from the inalienably possessed noun ɯ-qiɯ ‘half’ , but the other dir e ctionality cannot be excluded. Clearer cases is pr ovided by the inalienably p ossessed ɯ-mdoʁ ‘ colour’ (from Tibetan མདོག་ mdog ‘colour’), which derives a counted noun with the numeral pre- fix kɤntɕʰɯ- ‘many’ ( on which se e §7 .3.1.4) in (98), and the native inalienably possessed noun tɤ-ʁar ‘wing’ from which the counted noun tɯ-ʁar ‘the length of one arm’ originates (§7 .4). 27 6 7 .3 Counted nouns (98) tɕeri but kɤntɕʰɯ-tɯpʰu, many-types kɤntɕʰɯ-mdoʁ many-colour ɣɤʑu. exist:sens ‘There ar e many typ es [ of the mushrooms called tɯqejmɤɣ ], and with many colours. ’ (24-zw ArqhAjmA G) {0003630#S47} 7.3.4.2 Counted nouns and alienably possesse d nouns Alienably possessed nouns designating containers can b e converted to a partitive counted nouns by adding the numeral prefixes to the noun stem. For instance, the noun kʰɯtsa ‘bowl’ has a corresponding counted noun tɯ-kʰɯtsa ‘one bowl’ as in (99). (99) tɯ-kri indef.poss-oil tɯ-kʰɯtsa one-bowl pjɤ-tu ifr.ipfv-exist ‘They had one bowl [full] of oil. ’ (140501 mdzadi) {0003905#S6} The same is true of nouns borrowed from Tibetan and Chinese such as pʰoŋ ‘bottle’ (fr om 瓶 <píng> ‘b ottle ’). A s an alienably possesse d noun, pʰoŋ designates the bottle itself (not its content), and takes p ostnominal numerals ( as in 100). (100) pʰoŋ bottle kɯβde four ɣɤʑu exist:sens ‘There are four bottles. ’ (elicited; can refer for instance to empty bottles) Converted to a counted noun tɯ-pʰoŋ ‘ one b ottle ’ it refers to the quantity of liquid contained in a b ottle , and typically follows a mass noun as in (101). (101) cʰa alcohol kɯβde-pʰoŋ four-bottle pjɤ-k-ɤ-ta-ci. ifr.ipfv-peg-p ass-put-peg ‘There wer e four bottles [full] of alcohol. ’ (140510 sanpian sheye-zh) {0003945#S49} The derivation process is quite productive , and potentially new counted nouns meaning ‘a ... full of’ can be derived from any alienably possessed noun if a mean- ing can be made out of it, as in tɯ-co from co ‘valley’ in (102), which means in this particular context ‘an entire valle y full of ... ’ . 277 7 N umerals and counted nouns (102) mbro horse tɯ-co one-valley kɯ-fse, sbj:pcp-be.like qaʑo sheep tɯ-co one-valley kɯ-fse sbj:pcp-be.like nɯ, dem fsapaʁ animal rmɯrmi all.kinds ʑo emph nɯ dem tɯ-co one-valley nɤ lnk tɯ-co one-valley ʑo emph pɯ-tu pst.ipfv-exist ɲɯ-ŋu. sens-be ‘There was one valley entir ely for each spe cies of animals, like one valley full of horses, one valley full of sheep . ’ (2005 Kunbzang) 7.3.4.3 Counte d nouns and verbs The derivation of verbs into counted nouns and that of counted nouns into verbs are both productive processes in Japhug. Counted nouns of verbal origin are either built by adding a numeral pr efix to the verb stem ( as tɯ-fkur ‘one load’ fr om fkur ‘carry on the back’) or to the verb stem with an additional pr efix tɤ- ( tɯ-tɤrmbɯ ‘one heap ’ fr om the transitive verb rmbɯ ‘heap up ’ , §16.4). Note also the simultaneous action nominal with two prefixes tɯ-tɯ- , which can be analyzed as involving a counted noun derived from a verb (§16.4.3). Most deverbal counted nouns are fr om transitive verbs, but examples fr om intransitive v erbs are also found. For instance, in (103) the counted noun tɯ- tsʰoz ‘ one complete set’ (from the stative intransitiv e verb tsʰoz ‘be complete’) translates Chinese 一整套 <yīzhěngtào> ‘ one complete set’ . (103) nɯnɯ dem kɯ erg iɕqʰa the.afor ementione d rŋɯl silver kɯ erg tɯ-ŋga indef.poss-clothes tʰɯ-kɤ-βzu aor-obj:pcp-make ci indef tɯ-tsʰoz one-complete.set pjɤ-βde. ifr-throw .down ‘[The bird] thre w her a complete set of clothes that had be en made in silver . ’ (140504 huiguniang-zh) {0003909#S106} Deverbal counted nouns are either partitiv e counted nouns (§7 .3.2.1), desig- nating a quantity of objects resulting from the action of the v erb (for instance tɯ-fkur ‘ one load’ or tɯ-tɤrtsɯɣ ‘ one pile’ fr om rtsɯɣ ‘pile up ’ as in 104) or semelfactive counted nouns (§7 .3.2.5), referring to the number of time an itera- tive/semelfactiv e action takes place (such as tɯ-tɤtɕʰɯ ‘hitting with the hoe one time ’ from tɕʰɯ ‘gore , stab’). 278 7 .3 Counted nouns (104) ɕɤrɯ bone nɯ dem ʁnɯ-tɤrtsɯɣ two-pile to-βzu-ndʑi ifr-make-du tɕe lnk ‘They had made two piles fr om the b ones. ’ (2002nyimav o dzer ) (105) qaʁ hoe tɯ-tɤtɕʰɯ one-hitting.with.hoe ta-lɤt aor:3 → 3 ′ -thr ow ‘He used the hoe one time. ’ (2003 qachGa) {0003372#S86} There also are mor e lexicalized and synchronically less obvious examples of counted nouns derived from verbs. For instance the temporal counted noun tɯ- xpa ‘ one year’ originates from the intransitive v erb pa ‘pass 𝑋 years’ illustrated in (106) (see also example 119 in §22.4.1.4 and the account of the -x- element in tɯ- x pa proposed in §7 .3.1.7). (106) kʰa house na-βde aor:3 → 3 ′ -leav e nɯ dem sqamnɯz twelv e to-pa ifr-pass.X.years tɕe lnk ‘T welve y ears had passed since she had left home. ’ (150828 huamulan-zh) {0006396#S127} Since Situ, Zbu and T shobdun all have cognates of tɯ-xpa ‘ one year’ , this non-trivial derivation must have occurred at the time of their common ances- tor , but not earlier . Other Burmo-Gyalr ongic languages, even Horpa and Khros- kyabs (Jacques et al. 2017) have a root related to Japhug pa ‘pass 𝑋 years’ for time ordinals (§7 .5.2), but not for the corresponding counted noun. Derivation of transitive verbs fr om counted nouns with the denominal prefix rɤ- is well-attested, for instance rɤtɣa ‘measure by handspan’ from tɯ-tɣa ‘ one span’ . A complete list of these verbs and the various meanings of this derivation is presented in §20.4.2. 7.3.4.4 Counted nouns and ideophones Deideophonic counted nouns are quite common, and b elong to all major func- tional categories describe d above . In the following discussion, ideophones (§10.1) are by default cited in pattern II (§10.1.2.2), though other patterns are mentioned in some cases where appropriate . The collective counted nouns tɯ-boʁ ‘one gr oup ’ and tɯ-ɟɯɣ ‘ one pack’ origi- nate from the ideophonic roots found in boʁboʁ ‘in group , in order’ and ɟɯɣɟɯɣ ‘in great number (of long objects)’ , respectively . The presence of plain voiced initial stops in these roots is a clue to their ideophonic origin (§10.1.5.1). 279 7 N umerals and counted nouns 7.5.2 Time ordinals Japhug has two series of time or dinals, one for days (T able 7 .13) and another one for years (T able 7 .14). In the following discussion, I adopt Michailovsky’s (2003) notations: 𝐷 −𝛼 ‘ 𝛼 day(s) ago ’ , 𝐷 +𝛽 ‘in 𝛽 day(s)’ , 𝑌 −𝛾 ‘ 𝛾 y ear(s) ago ’ , 𝑌 +𝛿 ‘in 𝛿 year( s)’ . T able 7 .13: Japhug day ordinals Day Time ordinal expression -2 jɯfɕɯndʐi ‘two days ago ’ , ‘the other day’ -1 jɯfɕɯr ‘yester day’ 0 jisŋi ‘today’ +1 fso ‘tomorrow’ +2 fsɤndi ‘the day after tomorrow’ +3 qʰɤndi ‘in three days’ +4 ɲɤndi ‘in four days’ +5 βʑindi ‘in five days’ +6 pɤtsɤndi ‘in six days’ The Japhug system has relatively few ordinals for past days and years (some Kiranti languages have terms for up to 𝐷 −6 and 𝑌 −4 , as shown by Michailov- sky 2003), but stands out by having six ordinals for futur e years (no Kiranti lan- guages has more than 𝐷 +6 and 𝑌 +4 in Michailovsky’s 2003 data). How ever , the 𝐷 +5 βʑindi ‘in five days’ , 𝐷 +6 pɤtsɤndi ‘in six days’ and the corresponding year ordinals 𝑌 +5 and 𝑌 +6 ar e not use d any mor e in day-to-day sp eech even by the eldest speakers. The 𝐷 0 to 𝐷 −2 and 𝑌 0 to 𝑌 −2 ordinals have a pr efix ji-/jɯ-/ja- , which is proba- bly of demonstrative origin. This prefix also occurs in the derived ordinals jɯɣmɯr ‘this evening’ and jɯxɕo ‘this morning’ ( see T ables 7 .15 and 7 .16 b elow ), as well as in some frozen adverbs, such as jɤxtsʰi ‘this time ’ (from tɯ-xtsʰi ‘ one time ’) and jinde ‘these days’ . Cognates of this prefix are found in other time or dinals elsewhere in Gyalr ongic. In the case of ɣɯjpa ‘this year’ , this prefix surfaces as a preinitial -j- to the root -pa ‘year’ ( see §7 .3.1.7 on the etymology of this noun). The ɣɯ- (from *wə- ) prefix may be cognate to the prefix pə- in pəvə ‘this year’ and pəsɲə ‘today’ in Stau. The expressions jɯfɕɯndʐi ‘tw o days ago’ and japandʐi ‘tw o years ago ’ , which are based on the 𝐷 −1 and 𝑌 −1 forms with a suffix -ndʐi , can be analyzed as time 286 7 .5 Counting time T able 7 .14: Japhug year ordinals Y ear Time ordinal expression -2 japandʐi ‘two years ago ’ , ‘a few y ears ago’ -1 japa ‘last year’ 0 ɣɯjpa ‘this year’ +1 fsaqʰe ‘next year’ +2 fsɤndɤpa ‘in two years’ +3 qʰɤndɤpa ‘in three years’ +4 ɲɤndɤpa ‘in four years’ +5 βʑindɤpa ‘in five years’ +6 pɤtsɤndɤpa ‘in six years’ ordinals ( 𝐷 −2 and 𝑌 −2 , respe ctiv ely), as in (123). 14 (123) nɯ dem ma apart.from kɯ-dɤn sbj:pcp-be.many japa last.year mɯ-tɤ-χtɯ-t-a. neg-aor-buy-tr:pst-1sg japandʐi two .years.ago alo, upstream ɬaltɕɤm anthr kɯ erg z-ɲɤ-ɕar tral-ifr-search tɕendɤre lnk ‘Last year , it did not buy a lot [ of edible fern]. The year before, up ther e (in Kamnyu), Lhalcam collected some. ’ ( conversation, 14. 05.10) Howe ver , in many cases, they have less spe cific meanings such as ‘a fe w days ago ’ and ‘a few years ago ’ , as shown by (124), referring to an e vent that had occurred more than one week before. 15 (124) jɯfɕɯndʐi last.days pɯ-tɯ-χɕu-ndʑi aor-2-be.str ong-du ma, lnk a-ŋga 1sg.poss-clothes lɤ-tɯ-sɯ-ɣɯt-ndʑi aor-2-ca us-bring-du nɯ, dem a-xtsa 1sg.poss-shoe nɯ dem wuma really ɲɯ-pe sens-be.good tɕe lnk nɯ dem ku-nɯ-ŋge-a, prs-a ut o-wear[III]-1sg a-tsa 1sg.poss-adapted wuma really ɲɯ-βze sens-make[III] ‘Thank you both for the other day , the clothes that you hav e sent me, 14 This suffix -ndʐi is not attested elsewhere, but is pr obably related to the syllable -ri of the relator noun ɯ-ʁɤri ‘before, in fr ont of’ , from an earlier *-n-ri with epenthesis ( *-ndri ) and fricativization of the resulting *dr cluster . 15 This use reminds of the English expr ession ‘a couple of days ago’ , which can mean up to a week ago. 287 7 N umerals and counted nouns the shoes are very nice, I wear them, they ar e my size. ’ ( conversation, 15.04.18) Even the 𝑌 −1 ordinal japa ‘last y ear’ can have a less specific meaning if com- bined with dal ‘ early’ and the lo cativ e ri as japa dal ri ‘a few years ago ’ as in (125). No other time ordinal can be used with dal ‘early’ or other modifiers. (125) japa last.year dal early ri, loc ftɕar summer ri loc ʁmbɣɯzɯn solar .eclipse ɣɤʑu exist:sens ɲɯ-ti-nɯ sens-say-pl ma sfp ‘ A few years ago , they said there was a solar eclipse during summer . ’ (29-RmGW zW n2) {0003730#S24} The day ordinals from 𝐷 +2 to 𝐷 +5 contain a suffix -ndi , which attaches to the bound state forms of fso ‘tomorrow’ in 𝐷 +2 fsɤndi and ɯ-qʰu ‘after , behind’ in 𝐷 +3 qʰɤndi . The 𝐷 +5 form βʑindi contains the bound state βʑɯ- of the Japhug pronunciation βʑi of the Tibetan numeral བཞི་ bʑi ‘four’ (see §7 .1.1), presumably meaning ‘the fourth day after tomorrow’ . Y ear ordinals from 𝑌 +2 to 𝑌 +5 (T able 7 .14) are built by simply adding the root -pa ‘year’ ( see §7 .3.1.7 and § §7 .3.4.3) to the bound state of the corresponding day ordinals 𝐷 +2 to 𝐷 +5 . The 𝑌 +1 time ordinal fsaqʰe ‘next year’ comes fr om the bound state of fso ‘tomorrow’ , and the second element may be from the counted noun ɯ-qʰu ‘after , behind’ followed by the locative suffix *-j with regular vo wel fusion (§8.2.4.4). T able 7 .15: Morning ordinals Day Morning ordinal -2 jɯfɕɯndʐisoz ‘the morning of two days ago ’ -1 jɯfɕɯsoz ‘yesterday morning’ 0 jɯxɕo ‘this morning’ +1 fsosoz ‘tomorrow morning’ +2 fsɤndisoz ‘in two days in the morning’ +3 qʰɤndisoz ‘in three days in the morning’ In addition, there are morning or dinals (T able 7 .15) which derive from the cor- responding day ordinals by adding the noun soz ‘morning’ ( with the exception of jɯxɕo ‘this morning’), and the evening and night or dinals (T able 7 .16) which are built by adding the bound form -mɯr (found in tɯ-ɣmɯr ‘ one evening’ and mɯrkɯrku ‘ every evening’). These or dinals are not attested after 𝐷 +3 , though 288 7 .5 Counting time the forms could be built easily; note the absence of b ound state , except for 𝐷 −1 , where jɯfɕɯr ‘yester day’ loses its coda, as in jɯfɕɯsoz ‘yesterday morning’ . Note the isolated qʰuj ‘this afternoon’ T able 7 .16), which can refer to the time period b etween noon and the night (including the early evening). As sho wn by (126), it is also p ossible to combine day ordinals with other time nouns such as tɯrmɯ ‘ dusk’ to indicate particular moments of previous or futur e days. (126) jɯfɕɯndʐi-mɯr two .days.ago-evening nɤ-pi 2sg.poss-elder .sibling kɯ-wxti sbj:pcp-be.big ɯ-taʁ 3sg.poss-on ko-ɴqoʁ-a ifr-hang-1sg ri lnk mɯ-tɤ-wɣ-tsɯm-a, neg-aor:up-inv-take .away-1sg jɯfɕɯr yesterday tɯrmɯ dusk tɕe lnk nɤ-pi 2sg.poss-elder .sibling tɯlɤt second.sibling nɯ dem ɯ-taʁ 3sg.poss-on ko-ɴqoʁ-a ifr-hang-1sg ri lnk mɯ-tɤ-wɣ-tsɯm-a neg-aor:up-inv-take .away-1sg tɕe, lnk jɯɣmɯr this.evening ndɤre t op.advers nɤʑo 2sg tu-kɯ-tsɯm-a ipfv:up-2 → 1-take.away-1sg ra be.neede d:f act ‘T wo days ago, during the e vening, I clung onto your eldest sister but she did not take me away (to heaven), yester day at dusk I clung onto your second eldest sister but she did not take me away , this ev ening take me away . ’ ( 07 -deluge) {0003426#S55} T able 7 .16: Evening ordinals Day Evening Night -2 jɯfɕɯndʐimɯr ‘the evening of tw o days ago ’ -1 jɯfɕɯmɯr ‘yesterday e vening’ jɯfɕɯɕɤr ‘yesterday night’ 0 qʰuj ‘this afternoon’ , jɯɣmɯr ‘this evening’ +1 fsomɯr ‘tomorrow e vening’ +2 fsɤndimɯr ‘in two days in the ev ening’ +3 qʰɤndimɯr ‘in three days in the ev ening’ Time ordinals can be used either on their own as (123), (124) or (126) above, or be followed by the linker tɕe as in (127). They are nev er used with the lo cative postpositions ri , zɯ , or tɕu (which are used to mark some temporal adjuncts, 289 7 N umerals and counted nouns §8.2.4.1), except for the 𝐷 −1 , 𝑆 −2 , 𝑌 −1 and 𝑌 −2 ordinals which can occur with ri , as in (128). (127) fsɤndi in.two .days tɕe lnk li again kɯmaʁ other ji-kɤ-nɤma 1pl.poss-obj:pcp-work ɣɤʑu exist:sens ‘ After tomorrow we hav e something else to do. ’ ( conversation, 2012.12) (128) japa last.year ri loc tɕe, lnk <hongyuan> anthr <caizhengju> finance.office ri loc pɯ-cʰa, aor-can <kaoshi> exam pɯ-cʰa aor-can ‘Last year , he passed the exam for the finance office in Hongyuan. ’ (12-BzaNsa) {0003484#S69} Time ordinals can be converted to inalienably possessed nouns with the third singular prefix ɯ- to change the refer ence time from the present to a point of time in the past or a to hypothetical time reference . For instance, from fso ‘to- morrow’ one can build ɯ-fso ‘the next day’ or ɯ-fsosŋi ‘the next day’ (by adding the counted noun tɯ-sŋi ‘one day’), as in (129), (130) or (131). (129) ɯ-fso-sŋi 3sg.poss-tomorrow-day tɕe lnk tɕe lnk li again pjɤ-nɤtsoʁ-nɯ ifr-dig.up.silv er weed-pl tɕe lnk ‘The next day , the y dug up again silverwee d roots. ’ ( 07 -deluge) {0003426#S46} (130) nɯ dem tɯ-rʑaʁ one-night nɯ dem a-pɯ-nɯ-fse irr-ipfv-a ut o-be.like tɕe, lnk ɯ-fso-soz 3sg.poss-tomorrow-morning tɕe lnk ɯ-ci 3sg.poss-water ʑo emph ɲɯ-ɬoʁ ipfv-come.out ɲɯ-ŋu sens-be ‘If you leave it one night, the next day in the morning the juice comes out. ’ ( conversation 14.05.10) (131) tɕe lnk tɯ-ji indef.poss-field ɯ-ŋgɯ 3sg.poss-inside pjɯ-nɯ-ɕe ipfv:down-a ut o-go tɕe, lnk ɯ-fsaqʰe 3sg.poss-next.year tɕe lnk li again tu-ɬoʁ ipfv:up-come.out ɲɯ-ŋu sens-be tɕe. lnk ‘[Its seed] go es into [the soil of] the field, and it gro ws again the next year . ’ (13-NanWkW mtsW G) {0003492#S105} 290 7 .5 Counting time 7.5.3 Other derived time adverbs Japhug has a series of distributive adverbs meaning ‘ e very/ each 𝑋 ’ (T able 7 .17). These adverbs are derived from time nominals (mainly counted nouns, but also alienably possessed nouns like soz ‘morning’) by adding a suffix -ku and then ap- plying partial reduplication (§4.1). If the nominal root has a coda such as that of -mɯr ‘ evening’ ( as in tɯ-ɣmɯr ‘ one evening’), this coda is resyllabified and un- dergoes reduplication together with the suffix, as in the form mɯrkɯrku ‘e very evening’ (not † mɯrkɯku ). The /ɯ/ of the preceding syllable tends to b ecome [ u ] due to vow el assimi- lation (§3.3.1.2); since vow el assimilation always occurs in the forms without a reduplicated cluster , the transcriptions sɲikuku ‘ every day’ and pakuku ‘ every year’ ar e use d in this grammar (T able 7 .17) instead of the more phonological rep- resentations /sɲikɯku/ and /pakɯku/ . These adverbs are commonly used with the emphatic marker ʑo , as in (132). (132) mɯrkɯrku every .ev ening ʑo emph kɤntɕʰaʁ street a-pi 1sg.poss-elder .sibling ɯ-pʰe 3sg.poss-da t kɯ-nɤpɤri sbj:pcp-have .supp er ɲɯ-ɣi-a, ipfv:west-come-1sg ... ... pɯ-ŋu. pst.ipfv-be ‘Every evening, I would go in the town at my elder brother’s home to have supper and... ’ (140501 tshering skyid) {0003902#S110} This morphological derivation is not productive , and cannot b e used with any other temporal counted noun, even tɯ-sla ‘ one month’ – the distributive modi- fiers raŋri ‘ each’ or rɯri ‘ each’ are used instead (§9.1.3.3). T able 7 .17: Distributive time adverbs Time nominal Adverb soz ‘morning’ soskɯsku ‘ every morning’ tɯ-ɣmɯr ‘ one evening’ mɯrkɯrku ‘ every evening’ tɯ-sŋi ‘ one day’ sɲikuku ‘ ev er y day’ tɯ-xpa ‘ one year’ pakuku ‘ every year’ 7.5.4 Clo ck time Japhug has a series of words borro wed from Tibetan that can b e used to refer to hours and minutes. The alienably possesse d tɯtsʰot ‘time , hour , clock’ from 291 7 N umerals and counted nouns Tibetan ས་ཚད་ dus.tsʰod ‘time, hour’ (note that tɯ- here is not a prefix, but r epre- sents the Tibetan syllable ས་ dus ‘time’), if dir ectly followed by a numeral (or the generic counted noun tɯ-rdoʁ ‘one piece’) r efers to hours, as in examples ( 133) and (134). (133) ɕɤr night tɯtsʰot hour sqamnɯz twelv e ʑo emph tɕe lnk a-jɤ-tɯ-z-nɤtɯɣ irr-pfv-2-ca us-happen.to.be.at tɕe, lnk ‘Y ou will have to make sure to be there at midnight. ’ (2003 qachGa) {0003372#S26} (134) tɕe lnk nɯnɯ dem ci one ta-ɕɯ-mŋɤm aor:3 → 3 ′ -caus-hurt nɯnɯ dem tɯtsʰot hour tɯ-rdoʁ one-piece jamar about mɤɕtʂa until mɤ-ʑi. neg-subside:f a ct ‘When [nettles] start causing pain, it won’t stop for about one hour . ’ (11-mtshalu) {0003472#S7} T o e xpress a length of time in minutes, the noun tɯtsʰot ‘time, hour , clo ck’ is followed by the counted noun tɯ-skɤrma ‘ one minute ’ , as in (135). (135) tɯtsʰot time sqamŋu-skɤrma, fifteen-minute ɣnɤsqi-skɤrma twenty-minute jamar about tɤ-tsu aor-pass tɕe lnk ku-smi ipfv-be.cooked ɕti. be.aff:f a ct ‘ After fifteen or twenty minutes, it is cooke d. ’ (1607 06 thotsi) {0006133#S50} T o ask about clock time, the interrogativ e pronoun tʰɤstɯɣ ‘how many’ , ‘how much’ occurs with the verb zɣɯt ‘arrive ’ as in (136) ( see also §6.5.3). (136) nɤki dem : dist al nɯ dem tɯtsʰot hour tʰɤstɯɣ how .many ko-zɣɯt? ifr -reach ‘ At your place, what time is it?’ ( conversation, 14.12.24 – the question refers to the time lag between Paris and Mbarkham) These expressions, although used in the ev er yday language in Japhug, ar e calqued from Chinese since they follow modern time counting units and are all recent. Another calque from Chinese includes tɯ-tɕɯlɤβ ‘ one unit of tobacco ’ as in (137), from the expr ession 一斗烟的功夫 <yī dǒu yān de gōngfū> ‘time of one unit of tobacco ’ . 292 7 .6 Basic arithmetic op erations (137) tʰamaka tobacco tɯ-tɕɯlɤβ one-tobacco.unit kɤ-sko inf-smoke ɯ-raŋ 3sg.poss-time jamar about ɲɯ-ra sens-be.neede d ‘One needs ab out the time of a tobacco smoke. ’ (elicited) 7.6 Basic arithmetic op erations Although Japhug lacks an elaborate mathematical vocabular y , is it possible to express at least the basic arithmetic operations without recourse to Chinese . The counted noun tɯ-rdoʁ ‘ one pie ce ’ is used instead of the numeral ci ‘ one ’ is calculations (see 138 below ). Additions ar e expressed by the construction in (138) and (139), 16 with the verb ta ‘put’ ( or alternatively , ɣɤjɯ ‘add’) and the locative noun ɯ-taʁ ‘ on, above ’ , literally ‘If one puts Y on the top of X, it makes Z’ corresponding to 𝑋 + 𝑌 = 𝑍 . (138) kɯmŋu five ɯ-taʁ 3sg.poss-on tɯ-rdoʁ one-piece pjɯ  -wɣ-ta ipfv:down-inv-put tɕe lnk kɯtʂɤɣ six tu-βze ipfv-do[III] ŋu. be:f act ‘Five plus one equals six. ’ ( gram140505 math, elicitation) (139) nɤʑo 2sg kɯɕnɯ-kɯrcat seven-eight nɯ dem pɯpɯŋu as.for nɤ, lnk kɯɕnɯz seven ɯ-taʁ 3sg.poss-on kɯrcat eight pjɯ  -wɣ-ta ipfv:down-inv-put tɕe lnk sqamŋu fifteen ɲɯ-ŋu sens-be sqamŋu fifteen tɕe lnk ju-tɯ-nɯ-ɕe ipfv-2-vert-go ɲɯ-ŋu sens-be tɕe, lnk ‘ As for the expr ession ‘seven eight’ [that y our father in law has told] you, sev en plus eight equals fifte en, [it means that] you go back [to y our husband’s home] in fifteen days. ’ (2005 tA wa k W cqraR) In the case of subtractions, the main verb is tɕɤt ‘take out’ combined with the downstream orientation pr everb (the pr everb cʰɯ- in 140, §15.1.4.2). The minuend ( tɯ-rdoʁ in 140) is the object of tɕɤt ‘take out’ , and the subtrahend ( kɯmŋu in 16 The form kɯɕnɯ-kɯrcat (with the numeral pr efix kɯɕnɯ- on the numeral kɯrcat ‘eight’) in example (139) is part of a riddle in the story , and is not a normal way to express additions in Japhug. 293 7 N umerals and counted nouns 140) is encoded with the relator noun ɯ-ŋgɯ ‘inside’ (§8.3.4.4). Alternativ ely , the verb sɯxtɕʰaʁ ‘make diminish’ , causative of tɕʰaʁ ‘ diminish’ (§17 .2.1.4) can also occur (selecting the downw ards orientation). The construction is exactly the same as in (140), replacing cʰɯ  -wɣ-tɕɤt by pjɯ  -wɣ-sɯxtɕʰaʁ . (140) kɯmŋu five ɯ-ŋgɯ 3sg.poss-inside tɯ-rdoʁ one-piece cʰɯ  -wɣ-tɕɤt ipfv-inv-take.out tɕe lnk kɯβde four ma apart.from ɲɯ-me ipfv-not.exist ŋu be:f act ‘If one takes out one from fiv e, only four remain = fiv e minus one e quals four’ (gram140505 math, elicitation) Japhug is rarely used for multiplication. The counted noun tɯ-zloʁ ‘ one time’ can convey a multiplicativ e meaning as in example (141). Note that this counted noun of Tibetan origin has two alternative forms for ‘two times’ , the regular ʁnɯ- zloʁ and the hybrid form ʁɲɯ-zloʁ with an irregular numeral prefix that seems influenced by the Tib etan numeral གཉིས་ gɲis ‘tw o’ ( see §7 .1. 6.1, §7 .3.1.5). Divisions are tr eate d in the section on fractions (§7 .6.1). (141) kɯngɯt nine nɯ dem χsɯm three ɣɯ gen χsɯ-zloʁ three-time ŋu be:f act ‘Nine is three times three. ’ (gram140505 math, elicitation) 7.6.1 Fractions Japhug does not have specific names for fractions other than ɯ-qiɯ ‘half’ . This inalienably possessed noun has no prefixal form, and to express the meaning ‘half a 𝑋 ’ , the only available construction is the one illustrated in (142) and (143), combining a counted noun with the numeral prefix ‘ one ’ , the genitive ɣɯ and ɯ-qiɯ ‘half’ . (142) tɯ-ɟom one-length.of.two .outstretched.arms ɣɯ gen ɯ-qiɯ 3sg.poss-half jamar about ɲɯ-rɲɟi sens-be.long ‘[Its tail] is about half a fathom (i.e. one arm) long. ’ (24-Zmbr W pGa) {0003628#S63} (143) nɯnɯ dem tɯ-xpa one-year ɣɯ gen ɯ-qiɯ 3sg.poss-half nɯ dem ɲɤ-ɕqʰlɤt. ifr-disappear ‘Half a year passed. ’ (150907 laoshandaoshi-zh) {0006398#S97} 294 7 .6 Basic arithmetic op erations The denominal verb nɤqiɯ ‘pass half of’ (§20.7 .1) can be use d with temporal counted nouns to express the same meaning (144). (144) tɯ-sla one-month ɲɤ-nɤqiɯ ifr-pass.half ‘Half a month passed. ’ (elicited) The additive interpr etation ‘and a half’ can b e expr essed by co or dinating the counted noun and the inalienably p ossessed ɯ-qiɯ ‘half’ using the comitative cʰo ‘and, with’ , as in (145). (145) ki dem.pro x kɯ-fse sbj:pcp-be.like tɯ-ɟom one-length.of.two .outstretched.arms cʰo comit ɯ-qiɯ 3sg.poss-half jamar about kɯ-zri sbj:pcp-be.long ra be.neede d:f act ‘They need [a piece of leather] long like one fathom and a half. ’ (24-mbGo) {0003621} Alternatively , the additiv e meaning ‘and a half’ can be expressed with the coun- ted noun tɯ-qiɯ ‘ one half’ , which deriv es from ɯ-qiɯ ‘half’ (§7 .3.4.1). (146) tɕe lnk ɯ-pɤrtʰɤβ 3sg.poss-between tɕe, lnk tɯ-pɤrme one-year tɯ-qiɯ one-half jamar , about, ʁnɯ-pɤrme two-y ear jamar about ma apart.from kɯ-me sbj:pcp-not.exist ʁɟa. completely ‘(Some women had thirteen or fifteen children) Between each of them, there was only one year and a half or two years. ’ (140426 tAp Atso kAnWBdaR, 89; see the preceding sentence in example 55 in §9.1.3.2) For more complex fractions, the counted nouns tɯ-tɯcɯr ‘ one part’ , tɯ-tɤsɯm ‘ one part’ or tɯ-tɯkro ‘ one part’ are used. T wo related constructions are possi- ble; in the following discussion, 𝑋 repr esents the numerator , 𝑌 the denominator (the fraction 𝑋 𝑌 ). The first one is a noun phrase with the structure 𝑌 - tɯcɯr ɯ-ŋgɯ 𝑋 - tɯcɯr , literally ‘ 𝑋 parts among 𝑌 parts’ , illustrated by example (147). This may b e a calque of the Chinese construction 𝑌 分之 𝑋 ( 𝑌 -fēn zhī- 𝑋 ) literally ‘ 𝑋 of 𝑌 parts’ , a way of expressing fractions attested from Han dynasty documents (Anicotte 2015) to standard Mandarin. 295 8 P ostpositions and relator nouns (12) a-me 1sg.poss-daughter ta-mbi 1 → 2-give:f act ra be.neede d:f act ‘I will give you my daughter . ’ (28-smAnmi) {0004063#S281} The theme of a secundative verb cannot be indexed by the verb morphology . In (13), the theme is the numeral ʁnɯz ‘two ’ , but dual indexation on nɯ-ta-mbi ‘I gave 𝑋 to you’ is not possible, as the form nɯ-ta-mbi-ndʑi (a or-1 → 2-give-du) can only mean ‘I gave 𝑋 to both of you’ , the number referring to the recipient and not the theme (§14.4.2). (13) mɤ-ta-mbi neg-1 → 2-give:f act ma lnk ʁnɯz two nɯ-ta-mbi aor-1 → 2-giv e ri, lnk nɤʑo 2sg cʰɤ-tɯ-ɕɣɤz ifr-2-give .back ɕti be.aff:f a ct tɕe lnk ‘I won’t give her [my y oungest daughter] to you, as I gav e you two [ of my daughters], and you sent them back. ’ (2002 qaCpa) Howe ver , the theme of a secundative verb is r elativized like the obje ct of a monotransitive verb (§16.1.2.4, §23.5.4.2), and it can be considered as a sub-type of semi-object (§8.1.5). 8.1.7 Essive Essive noun phrases ar e not arguments of the sentence, but are used to indicate ‘the property of fulfilling the role of an N’ (Creissels 2014: 606; Jacques 2016a: 225). In Japhug, bare noun phrases without any case marker can be interpreted as essive adjuncts, such as nɤ-rʑaβ ‘your wife ’ in (14) and nɤ-kɯmtɕʰɯ ‘y our toy’ (15). These adjuncts are neither recipients (in both examples the recipient is 2sg, encoded as obje ct of the ditransitive v erb mbi ‘give ’ , §14.4.2) nor themes (in 14 the theme is ji-me ‘ our daughter’ , and it ser ves as object of the indirective kʰo ‘give ’ , §14.4.1; in 15 the theme is non-overt). (14) ji-me 1pl.poss-daughter nɯ dem nɤ-rʑaβ 2sg.poss-wife ɲɯ-kʰo-tɕi ipfv-give-1du je, sfp ɲɯ-ta-mbi ipfv-1 → 2-give je sfp to-ti ifr-say ‘He said: ‘W e will give you our daughter in marriage . ” (150831 laoshu jianv-zh) {0006374#S53} 302 8.1 Absolutive (15) nɤ-kɯmtɕʰɯ 2sg.poss-toy ɲɯ-ta-mbi ipfv-1 → 2-give ‘I give it to you as a toy . ’ (28-k W p Az) {0003714#S166} Since essive adjuncts are formally indistinguishable fr om absolutive arguments such as objects and intransitive subjects, some sentences may app ear to be am- biguous. In (16), the noun tɯrme ‘person’ could b e interpreted as the subject, and the phrase kɯki stu kɯ-xtɕi ki as a topic. (16) kɯki dem.pro x stu most kɯ-xtɕi sbj:pcp-be.small ki dem.pro x tɯrme person ɲɯ-pe sens-be.good tɕe lnk ‘The smallest one is very nice as a person. ’ (31-deluge) {0004077#S122} That tɯrme here is not the subject can be se en if one chooses a first or second person form: in (17), the noun tɯrme is still present despite 2sg indexation on the verb . This piece of evidence demonstrates that the analysis as essive adjunct is the only possible one. (17) nɤʑo 2sg tɯrme person wuma really ɲɯ-tɯ-pe sens-2-be.good ‘Y ou are very nice as a p erson. ’ ( elicited) It is common to have a related wor d as essive and as the subject or object which it refers to . In (18), for instance, the inalienably possesse d noun tɯ-ɣli ‘ dung’ occurs as b oth intransitiv e subje ct of the v erb pe ‘b e good’ (with 3sg possessive prefix) and as essive adjunct, in the latter in the specialized meaning ‘fertilizer’ . Note that the essive adjunct is closer to the v erb than the oblique phrase in ɯ-taʁ (§22.1.1.1). (18) tsʰɤt goat nɯ dem ɣɯ gen ɯ-ɣli 3sg.poss-manure nɯnɯ dem tɤ-rɤku indef.poss-crops ɯ-taʁ 3sg.poss-on tɯ-ɣli indef.poss-manure wuma really ʑo emph pe. be.good:f act ‘Goat manure is very good as a fertilizer for the crops. ’ (05-qaZo) {0003404#S27} In (19), the noun mbro ‘horse ’ occurs b oth in the object of rku ‘put in’ (meaning here ‘include the dowry’) and in the essive adjunct a-mbro ‘as a horse (for me)’ . 303 8 P ostpositions and relator nouns (19) mbro horse tɯ-skɤt indef.poss-spee ch kɯ-tso sbj:pcp-understand ci, indef a-mbro 1sg.poss-horse tɤ-rku-nɯ imp-put.in-pl ra be.neede d:f act ‘Give me a horse who understands human speech as my horse. ’ (2003 kandzW sqhaj) The essive adjunct can be either closer to the verb than the core argument it refers to , as in (112), (18) and (19) ab ov e, or further away as in (20), where the intransitive subject of mɤ-ra ‘is not neede d’ , rŋɯl ‘silver’ , occurs between the verb and the essive ɯ-pʰɯ ‘as its price ’ . (20) aʑɯɣ 1sg:gen ɯ-pʰɯ 3sg.poss-price rŋɯl silver mɤ-ra, neg-be.neede d:f act ‘I don’t want money as the price [for these clothes]. ’ (140506 shizi he huichang de bailingniao-zh) {0003927#S235} With the monotransitive verb pʰɯt ‘take out, cut’ , an essive adjunct can b e used to express the purpose of the action as in (21). (21) tsuku some kɯ erg paʁndza hogwash ɲɯ-nɯ-pʰɯt-nɯ ipfv-a ut o-pluck-pl ɲɯ-ŋu sens-be ri, lnk ‘Some people cut [ Sambucus ] as hogwash. ’ (12-ndZiNgri) {0003488#S28} This essive phrase sometimes occurs with the noun ɯ-spa ‘material’ with this verb as in (22), a construction that is in the process of grammaticalizing into a purposive construction when used with a participial clause (§24.4.2.2). (22) zɣɤmbu broom ɯ-spa 3sg.poss-material ɲɯ-nɯ-pʰɯt-nɯ ipfv-a ut o-cut-pl ŋgrɤl be.usually .the.case:f act ‘They cut it to make brooms. (‘as a material for brooms’)’ (140505 sWjno) {0003919#S21} Some verbs, like sɤrtsi ‘ count as’ (§17 .2.8), have an essive argument (rather than adjunct), which is not indexed on the verb and r e ceiv e no case marking, as shown by (23), where the essiv e argument is a-tɕɯ ‘my son’ . (23) aʑo 1sg a-tɕɯ 1sg.poss-son tu-ta-nɯ-sɤrtsi ipfv-1 → 2-a ut o-consider .as ŋu be:f act ‘I consider you as my son. ’ ( elicited) 304 8.1 Absolutive The transitive verb wum ‘gather’ can also select an essive argument differ ent from the object when used in the meaning ‘take as a student’ , as in (24). 3 In this example the object is 1sg, and the essive is the noun nɤ-slama , with a possessive prefix cor eferent with the subject. The verb wum in this meaning requires the orientation eastw ards in its centrip etal function ( see example 92, §15.1.4.3). (24) wortɕʰi please ʑo emph aʑo 1sg nɤ-slama 2sg.poss-student ku-kɯ-wum-a ipfv-2 → 1-gather-1sg ‘Please take me as your student. ’ (150907 laoshandaoshi-zh) {0006398#S35} V erb ellipsis may explain the presence of essiv e noun phrases with verbs that are not usually used with adjuncts of this type. For instance , in (25), the nouns pɣɤtɕɯ ‘bird’ and tɤ-mdzu ‘thorn’ are not objects (there 1sg object indexation on both pɯ  -wɣ-sɯ-sat-a and tʰɯ  -wɣ-sɯ-sat-a ) and are analyzable as essive adjuncts ‘me, as a bir d/thorn’ , ‘while I was a bird/thorn’ . 4 (25) a-pi 1sg.poss-elder .sibling kɯ erg nɯra dem:pl pɯ-fse pst.ipfv-be.like tɕe, lnk pɣɤtɕɯ bird ri also pɯ  -wɣ-sɯ-sat-a, aor-inv-ca us-kill-1sg tɤ-mdzu indef.poss-thorn ri also tʰɯ  -wɣ-sɯ-sat-a aor-inv-ca us-kill-1sg ‘My elder sister was like that, she had me killed [while I was a] bir d, and had also me killed [while I was a] thorn. ’ (2005 Kunbzang) The presence of these adjuncts in this particular context is probably due to the ellipsis of tɤ-sci-a ‘I was born’ , an intransitive v erb which takes essive noun phrases with the meaning ‘be reborn as, b e reincarnated as’ as in (26). (26) nɯ dem ɯ-qʰu 3sg.poss-after tɕe lnk pɣɤtɕɯ bird tɤ-sci-a, aor-be.born-1sg nɯ dem ɯ-qʰu 3sg.poss-after tɕe lnk tɤ-mdzu indef.poss-thorn tɤ-sci-a aor-be.born-1sg qʰe lnk ‘ After that I was reborn as a bird, and after that I was reborn as a thorn. ’ (2003 Kunbzang) 3 This expression is similar to Chinese 收……为 徒 <shōu ... wéitú> ‘take as a disciple’ , where 收 shōu has the same meaning as wum ‘gather’ , and where the essive is overtly marked by 为 wéi . 4 This example comes from a story where the main character was repeatedly killed by her sister , and reborn several times, first as a bird, and then as a thorn, as describ e in (25), from another version of the same story . 305 8 P ostpositions and relator nouns 8.1.8 Goal Motion verbs ( ɕe ‘go ’ , ɣi ‘ come ’ , ɬoʁ ‘come out’ etc), manipulation v erbs ( ɣɯt ‘bring’ , tsɯm ‘take away’ etc) and some p er ception verbs (like ru ‘lo ok at’) have arguments referring to the location towar ds which the action is directed. These arguments are not indexed in the verb morphology (in particular , motion verbs are intransitive , see §14.2.4). They can be marked with locative postp ositions (§8.2.4.1) or dative (§8.3.1), but also occur in absolutive form, as tɯrme-kʰa ‘ other people’s house ’ in (27). (27) tɕe lnk nɯ-mɤrʑaβ aor-marry qʰe lnk tɕe lnk tɯrme-kʰa people-house jɤ-ari aor-go[II] ɕti be.aff:f a ct qʰe lnk ‘She married and went to [live in] other pe ople ’s (her in laws’) house. ’ (14-siblings) {0003508#S9} Goals can be relativized (§23.5.5) using oblique participial relatives (§16.1.3.5) or finite relatives (§23.5.5.1), but not object participial relativ es except in a very restricted context (§23.5.5.2), unlike most absolutive arguments. 8.1.9 Location In addition to goals, absolutive noun phrases expr essing a static location are found in Japhug, in particular with the verb rɤʑi ‘stay’ as in (28). Locative post- positions can also occur , but are optional (§8.2.4.1). (28) tɕe lnk pjɯ-sɤ-sɯxɕat-a ipfv-ap ass-teach-1sg pɯ-ŋu pst.ipfv-be tɕe, lnk rqaco t opo pɯ-rɤʑi-a, pst.ipfv-stay-1sg tsʰuβdɯn t opo pɯ-rɤʑi-a. pst.ipfv-stay-1sg ‘When I was teaching, I lived in Rqakyo and in T shobdun. ’ (150819 kumpGa) {0006388#S2} Absolutive locative phrases are also found with some transitiv e verbs, as nɯ- mtʰɤɣ ‘their waist’ with rtɤβ ‘attach’ in (29). (29) tɯrme people ra pl kɯ erg nɯ-mtʰɤɣ 3pl.poss-waist ɲɯ-rtɤβ-nɯ. ipfv-attach-pl ‘People attach [ badger skin] around their waist (as a r emedy for rheumatism). ’ (27 -spjaNk W) {00037 04#S126} Oblique participial relatives (§16.1.3.5, §23.5.5.1) or finite r elatives (§23.5.5.1) are used to relativize these absolutive locative phrases. 306 8.2 Postpositions 8.2 Postpositions Postposition are invariable wor ds which ne cessarily follo w a noun phrase, and specify the syntactic function of that noun phrase, be it core argument or adjunct. They cannot be used on their own, and at the very least require a demonstrative pronoun such as nɯ (§6.9.1). Although their uses have commonalities with relator nouns (§8.3), the y dif- fer from those in lacking a possessive prefix. Postposition stacking is rare; only two cases exist: the sequence of locative postpositions tɕu zɯ and tɕe (§8.2.4.1) and also the case of genitive postpositional phrases used in the meaning ‘the one from/ of ... ’ ( with an elided head noun), as in (30), where the genitive ɣɯ is followed by the comitative (§8.2.5). (30) tɕe lnk pɤjka pumpkin ɣɯ gen cʰo comit wuma really ʑo emph naχtɕɯɣ. be.similar:f a ct ‘Its [little thorn-like things] are like those of the pumpkin. ’ {0003518#S55} 8.2.1 Indep endent words vs. clitics Since other Gyalrongologists, in particular Sun (1998; 2014a), treat the postposi- tions in related languages as clitics rather than as independent words as is done in the present work, a justification of the present analysis is necessary . In Japhug, the postpositions kɯ ‘ergativ e ’ (§8.2.2) and ɣɯ ‘genitive ’ (§8.2.3) do have some clitic-like characteristics: they cannot be used without a preceding noun phrase ( or a sub ordinate clause , §25.1.1), are unstressed, and in the case of the genitive have special irregular forms with pronouns (§6.3). Howe ver , a pause can occur between these postp ositions (31) and the noun phrase they follow . For instance, in example (31), a two second pause ( with an inspiration) is found between the phrase nɯŋa ra and the following ergative kɯ . (31) tɕe lnk tɯrtsi cow .food nɯ dem pjɯ  -wɣ-βzu ipfv-inv-make tɕe, lnk nɯŋa cow ra, pl kɯ erg nɤ-mɯm-nɯ estim-be.tasty:f a ct-pl cʰo comit wuma really ʑo emph ɣɯ-ɕɯ-fka-nɯ inv-ca us-be.satiated:f act-pl ‘They make cow food with flour , the cows find it tasty , and it satisfies their hunger . ’ (140513 tW rtsi) {0003985#S15} A filler (§10.3) can even be inserted between the noun phrase and the following ergative , as in (32). 307 [Document text truncated for crawler view.]